19

THE DEVELOPMENT OF GEOMETRIC

AND SPATIAL THINKING

Michael T. Battista
MICHIGAN STATE UNIVERSITY

The usefulness of axiomatics . . . goes beyond that of demonstration. . . . It permits us to construct simplified
models of reality and thus provides the study of the latter with irreplaceable dissecting instruments. . . .
[However] axiomatics cannot claim to be the basis of, and still less to replace, its corresponding experiential
science. . . . Axiomatic geometry is incapable of teaching us what the space of the real world is like.
—Piaget, 1950, pp. 30–31

Every spatial representation can depend in part on the geometry of the object and in part on the geometry
of the subject—and the dosages of each are hard to determine.
—Piaget, 2001, p. 223

In this chapter I examine geometric and spatial rea- tion and use of formal conceptual systems to investi-
soning in a broad context. I describe and analyze gate shape and space (Battista, 2001a, 2001b). For
recent research on the nature and development of instance, mathematicians employ a property-based
students’ thinking about geometry, geometric mea- conceptual system to analyze and define various
surement, and aspects of spatial reasoning that are types of quadrilaterals and triangles. This system uses
related to geometry. I concentrate on empirical and concepts such as angle measure, length measure,
theoretical research, not curriculum and technologi- congruence, and parallelism to conceptualize spatial
cal tool design. I focus on key ideas rather than com- relationships within and among the shapes. So, defin-
prehensively surveying the field. Where possible, I ing a square to be a four-sided figure that has four
discuss the cognitive roots of geometric reasoning. right angles and all sides the same length creates an
idealized property-based concept that, with proper
spatial grounding, can help people reason more pre-
cisely about this special class of shapes.
DEFINING THE SUBJECT:
Underlying most geometric thought is spatial
GEOMETRY AND SPATIAL REASONING
reasoning, which is the ability to “see,” inspect, and
reflect on spatial objects, images, relationships, and
Geometry is a complex interconnected network of transformations. Spatial reasoning includes generat-
concepts, ways of reasoning, and representation sys- ing images, inspecting images to answer questions
tems that is used to conceptualize and analyze physi- about them, transforming and operating on images,
cal and imagined spatial environments. Geometric and maintaining images in the service of other men-
reasoning consists, first and foremost, of the inven- tal operations (Clements & Battista, 1992; Presmeg,

843
844 ■ STUDENTS AND LEARNING

1997; Wheatley, 1997). Thus, spatial reasoning pro- 485). Other researchers such as Edelman (1992) have
vides not only the “input” for formal geometric rea- distinguished between perceptual and conceptual ob-
soning, but critical cognitive tools for formal geomet- jects. He argued that perception is nonconscious and
ric analyses. operates on signals from the outside world received
by the body’s receptors, whereas conception is con-
scious and operates on the activity of those portions of
THEORIES OF GEOMETRY LEARNING the brain that record signals from the outside world.
PRELIMINARIES: DRAWINGS, DIAGRAMS,1 Smith (1995) distinguished between a category, which
FIGURES, SHAPES, AND CONCEPTS— is a group of objects in the world that belong together,
TIPS OF AN ENIGMATIC ICEBERG and a concept, which is “a mental representation of
such a group” (1995, p. 5). Finally, consider the math-
ematical formalization of an idea: “The concept defi-
Before examining theories on geometry learning, it is
nition is a form of words used to specify that concept”
important to reflect on and clarify the nature of the
(Tall & Vinner as quoted in Tall, 1992, p. 496).
various “primitive objects” that students operate on in
geometric and spatial reasoning. Integrating and synthesizing these ideas leads
to the identification of five types of basic objects in-
volved in geometric and spatial thought. A physical ob-
The Objects of Geometric Analysis ject is an actual physical entity such as a door, box, ball,
geoboard figure, picture, drawing, or dynamic drag-
Researchers often distinguish two types of geo-
gable computer figure (Coren et al.’s distal stimulus).
metric objects: “Drawing refers to the material entity
A sensory object is the set of sensory activations evoked
while figure refers to a theoretical object” (Laborde,
when an individual views a physical object (Coren et
1993, p. 49). Laborde argued that teaching often “con-
al.’s proximal stimulus). A perceptual object is the mental
fuses drawings and the theoretical geometrical objects
entity perceived by an individual when viewing a physi-
that the former represent” (1998, p. 115). She further
cal object (Edelman’s perception). A conceptual object or
argued that student difficulties in geometry often
conceptualization (Edelman’s conception) is the con-
arise because students reason about material drawings
scious meaning or way of thinking activated by an indi-
when they are expected to reason about theoretical
vidual (a) in response to a perceptual object, memory
geometrical objects (Laborde, 1993). (See Laborde,
of a perceptual object, or concept definition; or (b)
1993 for further elaboration of these distinctions.)
constructed anew from other conceptual objects. A con-
Similarly, Presmeg observed that, “A picture or a dia-
cept definition is an explicit formal mathematical verbal/
gram is by its nature one concrete case, yet for any but
symbolic specification of a conceptual object.2
the most trivial mathematical thinking it is necessary
to abstract and to generalize” (1997, p. 305). Presmeg Another way to distinguish cognitive objects in
claimed that the “one-case concreteness” of drawings geometric reasoning is by their use (Battista, in press
and images is the source of many difficulties in visual- a). An object is a mental entity that is operated on
ization-based mathematical reasoning. (consciously or unconsciously) during reasoning. A
However, the complexity of the issue is only par- representation is something that “stands for” something
tially captured by the drawing-figure dichotomy. Jones, else (Goldin & Kaput, 1996). In general, in geometric
following Holzl, argued that “learners can get ‘stuck’ thought, one reasons about objects; one reasons with
somewhere between a drawing and a figure” (2000, representations.
p. 58). Furthermore, researchers in psychology and To understand geometric and spatial reasoning,
neuroscience have made additional distinctions that one must understand individuals’ cognitions about
illustrate the complexity of identifying the objects of the various objects described above, including how
geometric and spatial analysis. For instance, some re- these cognitions are interrelated. However, in ana-
searchers have characterized the difference between lyzing these cognitions, one must be aware of several
physical and perceptual objects in terms of distal and complications.
proximal stimuli: “A distal stimulus is an actual object
or event ‘out there’ in the world; a proximal stimu- Complication 1: Conception Affects Perception
lus is the information our sensory receptors receive What one “sees” is affected by what one knows
about that object” (Coren, Ward, & Enns, 1994, p. and conceives. For example, in the size-constancy

1
Although one might distinguish the terms drawing (something created freehand by a student or teacher) and diagram (a carefully made
graphic given in a textbook or other printed material), the two terms seem to be used synonymously in the literature.
2
Two additional types of objects discussed in the research literature—mental models and concept images—will be discussed later.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 845

Figure 19.1 Perception of size determined by context (Coren et al., 1994, pp. 488, 497).

phenomenon, the size of an object is conceptualized to relationships between parts of the shape but to its
as constant even though the retinal image it projects position, including implicitly, say, how angles line up
varies as the object’s distance from the eyes varies; the vertically and horizontally.
mind takes into account the apparent distance of the
object from the viewer to adjust perceived size (Coren Complication 2: Diagrams as Data or Representations?
et al., 1994). In Figure 19.1a, the three men appear to Diagrams, or more generally instances of physical
be the same size even though the most distant figure is objects, play two major roles in geometry. On the one
only about a third of the size as the nearest figure. In hand, instances of physical objects can be thought
contrast, in Figure 19.1b, the most distant figure ap- of as the data for geometric conceptualization—it is
pears much larger than the nearest, even though the through analysis of such objects that geometric shape
actual images are identical. Thus, perception of size is concepts are derived. On the other hand, instances of
adjusted based on a contextual conceptualization of physical objects, including diagrams, are used to rep-
distance. Figure 19.1c further illustrates how percep- resent formal geometric concepts. However, both cur-
tion is affected by conception and context. Amazingly, ricula and researchers generally neglect the process
the two dark vertical segments are congruent. But our of forming concepts from physical objects and instead
knowledge of the physical world, and the resulting focus on the “representational” perspective. Consider
mental models created to represent it, cause us to per- the following statements by researchers: “Diagrams
ceive the two segments as very different in length. are intended as models . . . meant to be understood as
Similarly, verbal labeling can affect conception— representing a class of objects. Every diagram has char-
an ambiguous figure that is called a broom when acteristics that are individual and not representative of
presented is drawn to look like a broom, whereas if it the class” (Yerushalmy & Chazan, 1993, p. 25). “Stu-
is called a rifle when presented, it is drawn as a rifle dents often attribute characteristics of a drawing to
(Coren et al., 1994). Might a similar phenomenon oc- the geometric object it represents” (Clements & Battista,
cur in geometry? Does using a different name for a 1992, p. 448). Chazan and Yerushalmy (1998) argued
drawn shape—square, rhombus, diamond—affect a that this “representational perspective” on diagrams
student’s perception/conception of the shape? Also, is rooted in the traditional axiomatic approach to
research shows that drawing a figure, in combination geometry instruction. Geometric “objects” are either
with experience and verbal labeling, can make it easier undefined or are defined formally in terms of unde-
to find when it is embedded in a more complex figure fined terms; “diagrams are aids for intuition and are
(Coren et al.). Thus again, experience (i.e., current not the objects of study themselves” (p. 70). And even
knowledge) affects perception. Finally, research sug- though Chazan and Yerushalmy stated that geometry
gests that current knowledge can affect perception of “should more explicitly build on students’ experience
similarities among instances, and thus classification, of the world around them” (p. 70), they seemed to
possibly because such knowledge can influence what conceptualize diagrams and figures in representa-
stimulus information is attended to (Ross, 1996). For tional terms, citing Aleksandrov: “Geometry has as its
instance, a square with vertical and horizontal sides object the spatial forms and relations of actual bodies,
turned 45° might be perceived as a diamond because removed from their other properties and considered from the
previous experience causes a student to attend not purely abstract point of view [italics added]” (p. 70).
846 ■ STUDENTS AND LEARNING

The representational perspective focuses on ab- do not understand the nature of the objects being
stract, general mathematical concepts, which are taken considered. In particular, students often attribute ir-
as the objects of study. A diagram is a mere indicator relevant characteristics of a diagram to the geometric
of an abstract mathematical concept; so many charac- concept it is intended to represent (Clements & Bat-
teristics of a particular diagram are deemed irrelevant tista, 1992; Yerushalmy & Chazan, 1993). For instance,
distractors. Consequently, most mathematics education students might not identify right triangles drawn in
research has treated diagrams as imperfect represen- nonstandard orientations (having unintentionally
tations of idealized abstract concepts, neglecting the abstracted orientation as an attribute). Another dif-
processes by which formal geometric concepts are ab- ficulty occurs when students use diagrams in proofs
stracted from particular instances. Taking a broader (Clements & Battista, 1992; Presmeg, 1997; Yerush-
perspective can provide additional insights into the almy & Chazan, 1993). For example, students may as-
processes by which students construct knowledge sume that sides that look parallel in an accompanying
about shapes and shape classes. Instead of focusing diagram are parallel (adding unintended conditions
exclusively on diagrams as representations of formal to hypotheses). Or, students might link a theorem too
mathematical concepts, researchers must investigate tightly to the example diagram given with the theo-
more carefully and explicitly two complementary ar- rem statement. For instance, if a theorem is originally
eas: (a) the process by which students progress from illustrated by a diagram of an acute triangle, students
analysis of particular figures to general abstractions might believe that the theorem does not apply, or
about classes of figures (it is from the analysis of par- might not think to apply it, to obtuse triangles. Still
ticulars that abstract conceptualizations arise) and (b) another type of difficulty can occur when drawings do
the mechanisms that enable students to utilize formal not capture appropriate geometric relationships—for
abstract geometric concepts to analyze particular example, a tangent might be drawn freehand so that it
figures. Investigating these areas is critical to under- is not perpendicular to the radius it intersects.3 Final-
standing how students construct understanding of ab- ly, students might mistake geometric diagrams for pic-
stract geometric concepts that are applicable not only tures of objects, profoundly changing their interpre-
within mathematics but to real-world problems. tation. For instance, a bright fourth grader identified
This new perspective reveals several extremely im- nonrectangular parallelograms as rectangles because
portant avenues for investigation. For instance, how he interpreted them as pictures of rectangles viewed
does the ability to abstract what is common among a set “sideways” (Clements & Battista, 1992).
of shapes (or what is invariant in a dynamic diagram)
develop among children, and how can instruction and
curricula promote this ability? How are abstract, gen-
THE VAN HIELE LEVELS:
eral conceptualizations produced? Researchers need
RESEARCH AND REVISIONS
to describe how students move from the particular to
the general. Investigating such abstractive generaliza-
tion is an important but neglected area in research on A considerable amount of research has established
geometric thinking. Although mathematics is about the van Hiele theory as a generally accurate descrip-
abstract concepts, mathematics education research tion of the development of students’ geometric think-
must focus on exactly how individuals derive meaning ing (Clements & Battista, 1992). After briefly review-
for abstract concepts from particular instances. ing the theory, I address recent developments in this
important area of research.

Difficulties With Drawings and Diagrams

The van Hiele Levels of Geometric Thought
Drawings and diagrams are essential for under-
standing geometric ideas. In traditional use, drawings According to the van Hiele theory, students prog-
and diagrams have been used for two major purpos- ress through discrete, qualitatively different levels of
es—(a) to represent classes of shapes (e.g., the set of geometric thinking. The levels are sequential and hi-
rectangles) and (b) to represent geometric relation- erarchical, so for students to function adequately at
ships (e.g., the bisectors of the angles of a triangle one of the advanced levels, they must have “passed
meet at the incenter of the triangle). Students, how- through” the lower levels. Briefly, the levels are as fol-
ever, have difficulty with both uses, often because they lows (see Clements and Battista, 1992 for a complete

3
Of course, a sloppy drawing might not interfere with reasoning if the individual creating it is aware of the relationships that should exist
in it.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 847

description). Level 0 Pre-recognition: Because immature students identify components of solids and informally
perceptual activity causes students to attend to only describe solids using properties, although properties
a subset of a shape’s characteristics, students are un- are not logically related. At Level 3, students are able
able to identify many common shapes. Level 1 Visual: to logically classify solids and understand the logic
Students identify and operate on geometric shapes ac- of definitions. At Level 4, students are able to prove
cording to their appearance. They recognize figures theorems about solids.
as visual gestalts, relying heavily on common proto- In contrast, Gutiérrez and colleagues’ extension
types. Level 2 Descriptive/analytic: Students recognize to 3d visualization is both less straightforward and
and can characterize shapes by their properties. Level more ambitious because it attempts to elaborate the
3 Abstract/relational: Because one property can signal relationship between visualization and geometric
other properties, students form form definitions, dis- conceptualization (Gutiérrez, 1992). At Level 1, stu-
tinguish between necessary and sufficient sets of con- dents compare solids using global perception with
ditions, and understand and sometimes even provide no attention given to properties such as angle size,
logical arguments. Level 4 Formal deduction: Students side length, or parallelism. Students are unable to
establish theorems within an axiomatic system. Lev- visualize solids, their positions, or motions that they
el 5 Rigor/metamathematical: Students reason formally cannot actually see; they manipulate solids by guess
about mathematical systems. and check. At Level 2, students move from global
comparisons to visual analysis of components and
Recent Developments their properties. Students are able to visualize sim-
ple movements from one visible position to another.
New developments will be grouped into three At Level 3, students compare solids by mathemati-
categories: extending the level descriptors beyond cally analyzing their components. They can visualize
2d shapes; reexamining the nature of the levels; and movements involving positions that are not visible,
elaborating the levels phenomenologically and psy- and in reasoning about movements, students can
chologically. match corresponding parts of images and preimages.
At Level 4, students can mathematically analyze and
Extending the Level Descriptors Beyond 2d Shapes formally deduce properties of solids. Visualization is
Early interpretations of the van Hiele levels fo- high and is linked to knowledge of properties.
cused on reasoning about limited aspects of two-di- Two observations are relevant to these extensions
mensional shapes; that is, the theory was investigated of the van Hiele levels to 3d figures. First, the descrip-
primarily for a well-defined but constrained domain. tors for geometric conceptualization (not visualiza-
But a number of researchers have extended the levels tion) at Levels 3 and 4 in the latter article are a bit less
to other geometric domains. Although difficulties in advanced than in the earlier article. One suspects that
interpretation and consistency often arise when re- this may be the case because traditionally there is far
searchers attempt to generalize the levels to other do- less attention given to formal proof in 3d geometry. Sec-
mains, such studies have increased our understanding ond, extending the levels to visualization is complex.
of the levels because generalization typically requires On the one hand, visualization is a capability that is
reconceptualization and refinement. As examples of not necessarily connected to knowledge of properties.
these extensions, Gutiérrez and colleagues extended For instance, some students who are not high visual-
the van Hiele level descriptions to reasoning about izers may develop analytic (property-based) strategies
3d shapes (Gutiérrez, 1992; Gutiérrez, Jaime, & For- to help them compensate for a lack of pure visualiza-
tuny, 1991). Johnson-Gentile, Clements, and Battista tion skills (Battista, 1990). Other students may possess
(1994), Lewellen (1992), and Jaime and Gutiérrez very high visualization skill well before they develop
(1989) extended the descriptions to motions/trans- property-based reasoning about solids. Indeed, some
formations. Clements and Battista (1992) summarized high visualizers can mentally imagine movements of
researchers’ views on the application of the van Hiele solids so well that, for many problems, they simply
levels to deductive/justification reasoning. Aspects of have no need to examine components of the solids.
several of the extensions are discussed below. On the other hand, Gutiérrez and colleagues have
Gutiérrez et al.’s (1991) extension of the levels to broken new ground by illustrating how visualization
3d shapes is straightforward and follows the original and geometric conceptualization can develop hand-
conception of the levels as described by Clements and in-hand, an idea also proffered by Clements and Bat-
Battista (1992). At Level 1, solids are judged and iden- tista (2001). However, much more research is needed
tified visually and holistically, with no explicit con- that clearly describes exactly how geometric concep-
sideration of components or properties. At Level 2, tualization interacts with visualization during develop-
848 ■ STUDENTS AND LEARNING

ment. For instance, at 3d visualization Level 2, how this vector approach, and focusing on 3d geometry,
does progression from global comparison to analysis Gutiérrez et al. described six profiles of level-configu-
of components and properties enable improvements rations in students’ geometric reasoning. For exam-
in visualizing movement, or vice versa? ple, Profile 2 was characterized by complete acquisi-
tion of Levels 1 and 2, high acquisition of Level 3, and
Reexamining the Nature of the Levels low acquisition of Level 4.
Originally, the van Hiele levels were conceptual- However, even though level acquisition was de-
ized as periods of development in geometric reason- scribed in terms of the vector model, the profiles could
ing characterized not only by qualitatively different easily be interpreted in terms of the original van Hiele
thinking, but by different internal knowledge orga- levels (as described by Clements and Battista, 1992,
nization and processing. In this view, students are at with the addition of Level 0). For instance, Profile 2
Level 1 when their overall cognitive organization and could be thought of as Level 2 or transition to Level
processing disposes them to think about geometric 3. Thus, rather than invalidating the levels, the vector
shapes in terms of visual wholes; they are at Level 2 approach merely suggests, as does the research by Leh-
when their overall cognitive organization disposes rer, Jenkins, and Osana (1998), that the original van
and enables them to think about shapes in terms of Hiele levels may need to be elaborated and expanded.
their properties. However, difficulties in classifying Indeed, the profiles suggest a possible decomposition
students’ van Hiele level, oscillation of students be- of van Hiele Levels 0–3 into six steps rather than four.4
tween levels, and students being at different levels for Unfortunately, one difficulty with the profiles approach
different concepts have caused many researchers to (and perhaps all current elaborations of the van Hiele
question whether the levels are discrete. Indeed, one theory) is that it lacks the conceptual coherence and
of the major issues that has arisen is, Should students’ clarity of the original van Hiele levels. That is, the origi-
thinking be characterized as “at” a single level? nal levels have a conceptual-logical coherence and elo-
quence that not only makes them appealing, but has
Degrees of Acquisition: A Vector Approach proved difficult to surpass.5
Arguing that students develop several van Hiele Several other components of the Gutiérrez et al.
levels simultaneously, Gutiérrez et al. (1991) used a (1991) research should be mentioned. First, by sug-
vector with four components to represent the degrees gesting that the van Hiele levels consist of configu-
of acquisition of each of van Hiele levels 1 through 4. rations of different types of reasoning, this research
They identified 5 periods in the acquisition of a level: seemed to encourage other researchers to investigate
no, low, intermediate, high, and complete acquisition. students’ geometric reasoning from a “configuration”
For example, in the intermediate acquisition period approach. Second, Gutiérrez et al. and Clements and
for a given level, students use reasoning characteristic Battista (2001) both attempted to deal explicitly with
of the level frequently and often accurately. However, the tricky issue of how to classify reasoning that fit the
students’ lack of mastery of the level causes them to description of a level but was incorrectly used. That is,
fall back on lower level reasoning when they encoun- if a student clearly is attempting to use properties in
ter special difficulties. In scoring student responses, identifying shapes, but does so incorrectly, is that stu-
Gutiérrez et al. classified each problem solution ac- dent thinking at Level 2? Third, and finally, Gutiérrez
cording to (a) the nature of the reasoning used via et al. have provided a careful and detailed attempt to
the van Hiele level descriptions, and (b) type. Type describe the transition between van Hiele levels.
was quantified by judging degree of attainment of van
Hiele level, correctness of answers, and correctness Elaborating the van Hiele Theory by Merging It With the
of reasoning (including justification). The numerical SOLO Taxonomy
score for degree of acquisition of a van Hiele level was Several researchers have attempted to elaborate
the arithmetic average of type weightings for all items and reformulate the van Hiele theory using the SOLO
that could be answered at that level. Consequently, taxonomy (Olive, 1991; Pegg, 1992; Pegg & Davey,
for example, a student’s degree of acquisition vec- 1998; Pegg, Gutiérrez, & Huerta, 1998). Although
tor might be: 96.67% for Level 1, 82.50% for Level van Hiele stated, “The most distinctive property of the
2, 50.00% for Level 3, and 3.75% for Level 4. Using levels of thinking is their discontinuity, the lack of co-

4
These researchers present the profiles as qualitative descriptions of empirical results, not steps in development (A. Gutiérrez, personal
communication, 12/14/04).
5
Keep in mind that the original levels themselves have had several iterations and interpretations; there have been discrepancies even in
the number of these levels.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 849

herence between their networks of relations” (1986, is seen as a transition between van Hiele Levels 1 to
p. 49), Pegg and Davey argued that several groups of 2.) van Hiele Level 3 occurs at the R2 Level in the
researchers have failed to detect these discontinuities concrete-symbolic mode. Students seem to achieve
and instead have found the levels to be more continu- van Hiele Level 4 as they progress through two UMR
ous in nature. Also, similar to Gutiérrez et al. (1991) level cycles in the formal mode.
and Lehrer et al. (1998), Pegg and Davey argued that The power of integrating the van Hiele theory with
research suggests that students develop more than one the SOLO theory is that it provides a possible framework
level at a time. They further argued that level discrete- for discussing finer delineations of the van Hiele levels.
ness, while appearing valid when the magnification of However, many issues and questions remain. For exam-
the analysis is low, seems blurred when the magnifica- ple, (a) The fact that the correspondence between van
tion is high. That is, as global descriptions, the levels Hiele level and SOLO taxonomy varies, depending on
seem valid, but when one attempts to examine stu- the article read, indicates that researchers are struggling
dents’ thinking more carefully, the level descriptions to make this complex correspondence work. The only
may be inadequate. agreement seems to be that the original van Hiele levels
To deal with the van Hiele theory’s lack of detail are too coarse to provide adequate distinctions in stu-
in describing students’ progression through the lev- dents’ geometric reasoning. (b) Why should researchers
els, Pegg and Davey (1998) integrated the van Hiele believe that the ability to focus on multiple aspects of a
theory with the SOLO theory of Biggs and Collis. In situation is a primary determinant of geometric reason-
their synthesis of the two theories, Pegg and Davey dis- ing? Certainly this ability is important in intellectual de-
cussed three major modes of thinking (see also Pegg, velopment—but is it specific enough to be a primary de-
1992). In the ikonic mode, students form and operate terminant of geometric reasoning, especially at each van
on mental images of objects with which they have had Hiele level? Also, what cognitive mechanisms enable an
contact. In the concrete-symbolic mode, students link individual to focus on more than one aspect at a time?
concepts and operations to written symbols, as long
as the context falls within their personal experience. Waves of Acquisition
In the formal mode, students are no longer restricted Several researchers have posited that different
to concrete referents and can systematically consider types of reasoning characteristic of the van Hiele lev-
principles, theories, and ranges of possibilities and con- els develop simultaneously at different rates, and that
straints; they become capable of formal proof. Each of at different periods of development, different types or
these modes is further differentiated by progression “waves” of reasoning are dominant, depending on the
through levels of complexity, with progression through relative competence students exhibit with each type
each mode requiring at least two cycles (Pegg et al., of reasoning (Clements & Battista, 2001; Lehrer et al.,
1998). At the unistructural (U) level, students focus on 1998; see Figure 19.2). Development depends both
one isolated aspect of a situation. At the multistructural on maturation and instruction.
(M) level, students focus on two or more unrelated as-
pects of a situation. At the relational (R) level, students
interrelate multiple aspects of a situation. Visual reasoning Descriptive-analytic Abstract-relational
dominant reasoning dominant reasoning dominant
According to Pegg et al. (1998),6 van Hiele Level 0
(in the Clements/Battista numbering system) occurs
at the ikonic mode, in the first UMR cycle (U1, M1,
R1). van Hiele level 1 occurs at the ikonic mode in the
Competence

second UMR cycle (U2, M2, R2). At U2, the imaging

process focuses on one isolated aspect of a situation,
such as the sharpness of a vertex. At M2, several unre-
lated aspects of a situation are considered. At R2, the Abstract-relational reasoning
individual has full control over the imaging process Descriptive-analytic reasoning
and is able to identify familiar shapes correctly. van Visual reasoning
Hiele Level 2 occurs at the concrete-symbolic mode
in the second UMR cycle, with Levels U2 (focusing on
Time
a single property) and M2 (focusing on many proper-
ties). (The earlier U1, M1, R1 concrete-symbolic cycle Figure 19.2 Waves of acquisition of van Hiele levels.

6
The correspondence between van Hiele levels and SOLO modes and levels varies by research report. See the cited references in this
section.
850 ■ STUDENTS AND LEARNING

Based on the administration of a single type of analysis and synthesis) is dominant, descriptive-ana-
task,7 Lehrer et al. (1998) argued that “level mixture” lytic knowledge begins to emerge and interacts with
was typical for primary-age students, that students’ jus- visual knowledge. The early dominance of visual-ho-
tifications often jumped across nonadjacent van Hiele listic reasoning gradually gives way to descriptive-ana-
levels, and that reasoning varied greatly with task. They lytic reasoning. As abstract reasoning begins its ascen-
argued against portraying geometric development dance, connections among all types of reasoning are
in terms of discontinuities among levels, and instead strengthened and reformulated. Consistent with this
for it being characterized “by which ‘waves’ or forms “continuous growth” view, these researchers found
of reasoning are most dominant at any single period little evidence in their own research or that of others
of time, without necessarily implying the extinction of of discontinuities between levels as postulated by the
other forms” (p. 163). They found that children’s geo- van Hiele theory; growth in small increments seemed
metric thinking was accomplished by a range of mental to be the norm.
operations including comparing shapes to visual pro-
totypes, mentally “morphing” one shape into another, Stages or Levels?
and finding a variety of attributes of shapes. They ar- Clements and Battista (2001) viewed the van
gued that these varied types of reasoning should not Hiele levels as levels, rather than stages. They described
be described by the single descriptor “visual.” However, a stage as a substantive period of time characterized
one can argue that the types of reasoning that Lehrer by cognition across a variety of domains that is quali-
et al. questioned as being visual are, in fact, examples tatively different from that of both the preceding and
of visual, or Level 1, reasoning, and that Lehrer et al. succeeding stages. In contrast, a level is a period of
have provided us with a richer description of such rea- time of qualitatively distinct cognition, but within a
soning. That is, the Lehrer et al. research suggests that specific domain. In van Hiele research, one of the
Level 1 visual reasoning is not as limited and homo- open questions concerns the size of the “domains”
geneous as the original van Hiele levels suggest. Fur- to which the van Hiele levels apply. For instance, Cle-
thermore, the characterization of students’ thinking as ments and Battista (2001) suggested that many em-
property-based when they refer to number sides should pirical studies indicate that people exhibit behaviors
be questioned (I will return to this issue later in the indicative of different levels on different subtopics of
chapter). And finally, responses the researchers coded geometry (Denis, 1987; Mason, 1989; Mayberry, 1983).
as “classification” were not hierarchical classification This finding, taken together with the observation that
in the van Hiele sense—students were simply naming students sometimes exhibit different levels on differ-
figures (it would be a truly exceptional third-grader ent tasks, suggests that the domains to which levels of
who used a genuine hierarchical classification). So, as thinking apply may be, at most, subtopics within the
in so many van Hiele studies, the question that arises is broad range of geometry.
whether the “level jumping” was due to students’ rea- However, the domain issue may be more compli-
soning or to quirks in the classification system used by cated than it seems, especially when viewed from a lon-
the researchers. gitudinal rather than a static point of view. For instance,
Clements and Battista (2001) proposed the view Lewellen (1992) carefully studied 8 students participat-
that the van Hiele levels (seen as types of reasoning) ing in the Logo Geometry Curriculum (Battista & Cle-
develop simultaneously (albeit at different rates), sim- ments, 1991; Clements & Battista, 2001), which first
ilar to the “overlapping waves” metaphor suggested by deals with geometric shapes then with geometric mo-
Siegler (1996), and by Lehrer et al. (1998). Visual-ho- tions. She found that students entering the motions
listic knowledge, descriptive verbal knowledge, and, unit all started at Level 1 for motions, but after a short
to a lesser extent initially, abstract symbolic knowledge initial learning period, students at Level 2 or in transi-
grow simultaneously, as do interconnections between tion to Level 2 for shapes moved up to Level 2 in mo-
levels. However, although these different types of rea- tions. Lewellen concluded, “These results lend support
soning grow in tandem to a degree, one level tends to to the notion that there is a global van Hiele level for
become ascendant or privileged in a child’s orienta- each student after they have some initial learning in
tion toward geometric problems. Which level is privi- any new domain” (Lewellen, 1992, p. 279).
leged is influenced by age, experience, intentions, In fact, students who entered the motions unit
tasks, and skill in use of the various types of reason- at Level 2 on shapes were able to quickly learn the
ing. In the early years, while visual reasoning (called components and characteristics of objects in this new
syncretic to signify combination without coherent domain. Rather than limiting themselves to specific

7
Detailed discussion of this task will be given in a later section.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 851

values and visual feedback, these students looked for like a square. Orientation of figures may strongly af-
underlying properties, components, characteristics, fect Level 1 students’ shape identifications. In fact,
and rules in problem-solving situations. For example, Level 1 students tend to identify a square only if its
Kelly developed measure-based strategies for verifying sides are horizontal and vertical.
whether or not motions were correct, such as measur- 1.1 Pre-recognition. Students are unable to iden-
ing the distance of corresponding points from the flip tify many common shapes.
line for flips and the distance of corresponding points 1.2 Recognition. Students correctly identify many
from the turn center for turns. She also knew that a common shapes.
slide moves all points of a figure the same distance and
direction. Thus, students like Kelly who were at Level Level 2: Analytic-Componential 8 Reasoning
2 for shapes, having seen the value of analytic con- Students explicitly attend to, conceptualize, and
ceptions in one domain—for instance, using length specify shapes by describing their parts and spatial
and angle measure to describe relationships between relationships between the parts. However, students’
components of shapes—seemed to seek out similar descriptions and conceptualizations vary greatly in so-
conceptions in the new domain of motions. Initially phistication, starting with completely informal and im-
as students worked in the new domain, they did not precise specifications in Sublevel 2.1 and ending with
have such conceptions. The key is that they had de- completely formal geometric specifications in Sublevel
veloped the ability to use such conceptions, and they 2.3 (which corresponds to original van Hiele Level 2).
understood their power, so they looked for properties Two major interrelated factors contribute to the
in the new domain. Of course, an open question is to development of Level 2 reasoning. The first is an in-
describe the degree of acquisition of a level of reason- creasing ability and inclination to account for the spa-
ing—along with enabling mental mechanisms—that tial structure of shapes by analyzing their parts and how
is required to attempt to and be able to implement the parts are related. The second is an increasing ability
that reasoning in another domain. to understand and apply formal geometric concepts in
analyzing relationships between parts of shapes.
An Alternative Route: Elaborating the Original 2.1 Visual-informal componential reasoning. Stu-
van Hiele Levels dents describe parts and properties of shapes infor-
To understand the development of reasoning in mally and imprecisely; they do not possess the formal
geometry, Battista has elaborated the van Hiele levels conceptualizations that enable precise property speci-
to carefully trace students’ progress in moving from fications. Descriptions and conceptualizations are vi-
informal intuitive conceptualizations of 2d geometric sually based, focusing initially on parts of shapes then
shapes to the formal property-based conceptual system on spatial relationships between parts. Examples of
used by mathematicians. The descriptions for Level 2 references to parts are counting the number of sides
given below were developed by Battista in several proj- or angles (usually referred to as “points”) in a shape
ects; the description for Level 3 was developed by Bor- or saying that a figure has straight sides. Examples of
row (2000) and Battista. This new elaboration consid- descriptions of relationships between parts are specifying
erably expands the van Hiele levels in two places—the that “corners” must be “square” or that “all sides must
development of property-based thinking, and the de- be even.” In all cases, students describe parts and their
velopment of inference about properties. relationships using strictly informal language, that is,
language typically learned in everyday experience.
Level 1: Visual-Holistic Reasoning Students’ informal language ranges greatly in pre-
Students identify, describe, and reason about cision and coherence, from using vague and incom-
shapes and other geometric configurations according pletely formulated conceptualizations to informally
to their appearance as visual wholes. They may refer describing a conceptualization that corresponds to a
to visual prototypes, saying, for example, that a figure formal geometric concept. As an example of the latter
is a rectangle because “it looks like a door,” or they case, as students examine a tilted rectangle or square,
might judge that two figures are the “same shape” be- they might informally refer to the property of hav-
cause they “look the same.” Students might also say ing all right angles by saying the shape has “square
that a square is not a rectangle because rectangles corners” or “straight sides.” Students’ descriptions of-
are “long.” Students may justify their responses using ten seem to occur extemporaneously as they inspect
imagined visual transformations, saying for instance shapes (as opposed to describing explicit conceptual-
that a shape is a square because if it is turned it looks izations formulated from previous experiences).

8
I use the term analysis to refer to the process of understanding objects by resolving them into their components.
852 ■ STUDENTS AND LEARNING

2.2 Informal and insufficient-formal componen- as opposite sides equal, must have reached the interi-
tial reasoning. As students begin to acquire formal orized level of abstraction, detaching them from their
conceptualizations that can be used to “see” and de- original contexts so they are applicable to new situa-
scribe spatial relationships between parts of shapes, tions and available for analyzing shapes.
they use a combination of informal and formal de-
scriptions of shapes. The formal descriptions utilize Level 3: Relational-Inferential Property-Based Reasoning
standard geometric concepts and terms explicitly Students explicitly interrelate and make infer-
taught in mathematics curricula. However, the formal ences about geometric properties of shapes. For ex-
portions of students’ shape descriptions are insuffi- ample, a student might say “If a shape has property
cient to completely specify shapes. For example, a stu- X, it also has property Y.” However, the sophistication
dent might explain that a rectangle has “sides across of students’ property interrelationships varies greatly,
from each other that are equal [formal] and square starting with empirical associations (when property X
corners [informal].” Or, a student might say that “a occurs, so does property Y), progressing to componen-
rectangle has sides across from each other that are tial analyses that explain why one property “causes”
equal lengths,” making no mention of right angles. Al- another property, then to logically inferring one prop-
though students often recall properties that they have erty from another, and finally to using inference to
abstracted for classes of shapes (say, “two long sides organize shape conceptualizations into a hierarchical
and two short sides” for rectangles), their reasoning is classification system. Because one property can “sig-
still visually based, and most of their descriptions and nal” other properties, students can logically organize
conceptualizations still seem to occur extemporane- sets of properties, correctly specify shapes without
ously as they are inspecting shapes. naming all their properties, form minimal definitions,
2.3 Sufficient formal property-based reasoning. and distinguish between necessary and sufficient sets
Students explicitly and exclusively use formal geomet- of conditions. At Level 2, students only can use formal
ric concepts and language to describe and conceptu- properties; at Level 3, students can operate on formal
alize shapes in a way that attends to a sufficient set of properties.
properties to specify the shapes. Students have made a At Level 3, the spatial relationships described by
decided shift away from visually dominated reasoning formal property statements reach the second level of
because the major criterion for identifying a shape is interiorization so that they can be symbolized by the
whether it satisfies a precise set of verbally stated for- statements, and so that students can reason meaning-
mal properties. So, for example, at Sublevel 2.3, the fully about the statements, in many cases, without hav-
term rectangle refers to a class of shapes that possesses ing to visually re-present the actual spatial structurings
all the properties the student has come to associate that the statements describe. The verbally-stated prop-
with the set of rectangles. erties themselves are interiorized so that they can be
Students can use and formulate formal defini- meaningfully decomposed, analyzed, and applied to
tions for classes of shapes. However, their definitions various shapes.
are not minimal because forming minimal definitions 3.1 Empirical relations. Students use empirical
requires relating one property to another using some evidence to conclude that if a shape has one property,
type of inferential reasoning (which occurs at Level it has another.
3). Students do not interrelate properties or see that 3.2 Componential analysis. By analyzing how types
some subset of properties implies other properties. of shapes can be built one-component-at-a-time, stu-
They simply think in terms of unconnected lists of for- dents conclude that when one property occurs, anoth-
mally described characteristics. The set of properties er property must occur. Students conduct this analysis
students give for a class of shapes is a list of all the by making drawings or imagining constructing shapes
visual characteristics the student has come to associ- piece-by-piece. For instance, a student might conclude
ate with that type of shape, described in terms of for- that if a quadrilateral has four right angles, its oppo-
mal geometric concepts. (But students seem to recall site sides are equal because if you draw a rectangle by
these property-based specifications rather than discov- making a sequence of perpendiculars, the sides must
er them on the fly as they inspect shapes.) be equal. (The student explicitly sees visually that mak-
Progressing to Sublevel 2.3 requires that formal ing a pair of opposite sides unequal causes some of the
concepts like side length and angle measure have angles to not be right angles.)
been raised to a sufficient level of abstraction so that 3.3 Logical inference. Students make logical in-
they can be used to form relational conceptualizations ferences about properties; they mentally operate on
such as “all sides equal” that describe spatial relation- property statements, not images. For example, a stu-
ships between shape parts. These relationships, such dent might reason that because a square has all sides
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 853

equal, it has opposite sides equal. Such reasoning en- & Battista, 1992; Lehrer, 1998). Gutiérrez et al. (1991)
ables students to make the inferences needed for hi- dealt with this issue by using vectors to indicate the
erarchical classification. For instance, a student whose degree of acquisition of various levels. Clements and
definition for a rectangle is “4 right angles and oppo- Battista (2001) and Lehrer et al. (1998) dealt with the
site sides equal” might infer that a square is a rectangle issue by positing that several levels may be developing
because “a square has 4 right angles, which a rectangle at the same time, but that different types of reasoning
has to have; and because a square has 4 equal sides, become dominant during learning and development.
it has opposite sides equal, which a rectangle has to Although each of these proposed revisions has merit
have.” But students do not use this inferencing abil- and has moved the field forward, they both face two
ity to logically reorganize their conceptual networks important issues.
about shapes, so they do not adopt a logical hierarchi- Issue 1: Type versus level of reasoning. Through-
cal shape classification system (i.e., they still resist the out the research on the van Hiele theory, there is a
notion that a square is a rectangle even though they lack of distinction between type of reasoning and quali-
can follow the logic justifying such a statement). tatively different levels in the development of reasoning. That
Students’ reasoning is “locally logical” in that they is, sometimes the visual-holistic level is used to refer
string together logical deductions based on “assumed- to a type of reasoning that is strictly visual in nature,
true” propositions, that is, propositions that they ac- and sometimes it is used to refer to a period of devel-
cept as true based on their experience, intuition, or opment of geometric thinking when an individual’s
authority. Thus, students at this level use logic, but thinking is dominated and characterized by visual-ho-
they do not question the starting points for their logi- listic thinking. For instance, Gutiérrez et al. (1991)
cal analyses. used vectors to indicate students’ “capacity to use
3.4 Hierarchical shape classification based on logi- each one of the van Hiele levels” (p. 238). This state-
cal inference. Students use logical inference to reorganize ment makes sense only if van Hiele levels are taken
their classification of shapes into a logical hierarchy. as types of reasoning, not qualitatively different levels
They fundamentally restructure their shape classifica- of thought. Similarly, Clements and Battista (2001),
tion networks (as opposed to merely making addition- along with Lehrer et al. (1998), talked about “waves
al connections here and there). It becomes not only of acquisition” of types of reasoning defined by the
clear why a square is a rectangle, but a necessary part van Hiele levels. To complicate matters further, Pegg
of reasoning. Students give logical arguments to justify and Davey used the SOLO taxonomy to describe cog-
their hierarchical classifications. Finally, students’ use nitive processes that enable van Hiele types of reason-
of logic to draw conclusions provides them with a new ing. Thus, broadly speaking, researchers have inter-
way to accumulate knowledge. That is, new knowledge mingled and not yet completely sorted out (a) basic
can now be generated not merely through empirical or enabling cognitive processes, (b) van Hiele levels as
intuitive means, but through logical deduction. types of reasoning, and (c) van Hiele levels as stages
of development of geometric reasoning.
Level 4: Formal Deductive Proof Related to this intermingling, several subissues are
Students can understand and construct formal apparent. First, exactly what are the basic cognitive pro-
geometric proofs. That is, within an axiomatic system, cesses that enable different types of geometric thought?
they can produce a sequence of statements that logi- Are they simply the types of thought described by the
cally justifies a conclusion as a consequence of the van Hiele levels? Or are other, more basic, processes
“givens.” They recognize differences among unde- implicated? For instance, research suggests several pos-
fined terms, definitions, axioms, and theorems. sibilities for basic processes: (a) visualization/imagery;
(b) analysis (decomposing shapes into parts and inter-
Discussion relating those parts); (c) conceptualization in terms of
One of the major criticisms of the van Hiele levels formal concepts; (d) ability to attend to, interrelate,
is that students’ progress does not seem to be charac- and coordinate various aspects of a situation; and (e) in-
terized by jumps from one discrete level to the next, ference/logical deduction. Are there other processes?
but instead, seems to occur in small incremental steps.9 How do each of these interact and produce the types of
To support this criticism, researchers cite the notion reasoning characterized by the van Hiele levels?
of transition between levels and that students often Second, is the vector or configuration of activa-
exhibit characteristics of several levels (e.g., Clements tions perspective viable? Certainly there is precedent

9
In fact, longitudinal studies often show students making progress within a van Hiele level, but not moving to the next level (Gutiérrez &
Jaime, 1998; Clements & Battista, 2001).
854 ■ STUDENTS AND LEARNING

for such a view. For instance, a vector model has been each van Hiele level. For instance, their definition of
used in describing how the brain represents objects in the recognition process focuses on physical attributes
the perceptual field, not by a set of neurons but by a at Level 1 but mathematical properties at Level 2, two
configuration of neuronal firing patterns (Churchland, very different kinds of “recognition” processes that are
2002). Is it then reasonable to think that type of rea- related but certainly not the same at a basic cognitive
soning is enabled by activating a configuration of en- level. Thus, it is one thing to devise broad categories
abling processes? Development of geometric reason- of behavioral descriptors; it is another to determine
ing would then occur (a) as the enabling processes be- the cognitive processes underlying these categories of
come more sophisticated, (b) as new and more useful behaviors. This has been, and will continue to be, a
configurations of processes form and become stable, major challenge facing researchers.
and (c) as metacognitive processes regulate the use of However, despite the difficulty facing researchers
configurations. But this approach leads back to defin- attempting to assess students’ van Hiele levels, each
ing the enabling processes, distinguishing them from serious attempt to do so helps us better understand
the van Hiele levels as types of reasoning, and explain- the nature of the levels and students’ geometric think-
ing how the van Hiele levels as levels occur in this acti- ing. For instance, in discussing van Hiele assessments,
vation approach. For instance, how exactly does one Gutiérrez and Jaime (1998) claimed that even though
type of thinking become ascendant or dominant? researchers most often identify students as belonging
Third, the vector-configuration approach runs the to Level 3 if they can understand and use hierarchical
risk of losing the compelling conceptual eloquence at- classifications of shape families, there are other criti-
tained by the original van Hiele levels. In fact, adopt- cal attributes of this level. They argued that students at
ing the vector-configuration approach might regress Level 3 can also establish logical connections between
the field into older “structure of intellect” models, geometric properties (a key component in thinking
with researchers focused so much on defining and hierarchically and in using logical definitions), and
measuring enabling processes that describing the na- they can accept and identify nonequivalent definitions
ture of different types of reasoning in conceptually of geometric shape classes. Do assessments of Level 3
coherent ways is lost. equally assess all of the characteristics of Level 3, and,
Issue 2: Assessment. Assessment has long been if not, how does that affect the validity of the assess-
an issue in the van Hiele theory, both because it op- ments? In fact, how do assessments affect researchers’
erationalizes researchers’ conceptions of the levels conceptualizations of the levels?
and because it is necessary for applying the theory to A totally different approach to assessing van Hiele
instructional practice. Researchers are still grappling levels was devised by another group of researchers. In
with this issue. For instance, in designing a paper-and- a collaborative effort to find ways to assess elementary
pencil instrument to be used with middle and high students’ acquisition of the van Hiele levels in inter-
school students, Gutiérrez and Jaime (1998) defined view situations, Battista, Clements, and Lehrer devel-
four processes (recognition, definition, classification, oped a triad sorting task, that, with variations, both
and proof) and how those processes were implement- Clements and Battista (2001) and Lehrer et al. (1998)
ed by students at different van Hiele levels. Recognition used in separate research efforts. In this task, students
refers to recognizing types of geometric figures and are presented with three polygons, such as those
identifying components and properties of figures; shown in Figure 19.3, and are asked, “Which two are
it includes use of vocabulary to describe properties. most alike? Why?” Choosing B and C and saying that
Definition refers to formulating definitions of concepts they “look the same, except that B is bent in” was tak-
during the learning process and using textbook defini- en as a Level 1 response. Choosing A and B and saying
tions. Classification refers to classifying geometric fig- either that they both have two pairs of congruent sides
ures into different categories. Proof refers to justifying or that they both have four sides was taken as a Level
geometric statements. Gutiérrez and Jaime described 2 response. The purpose of this task was to determine
how each process occurs at several van Hiele levels so the type of reasoning used on a task that students had
that how a student uses a process can be taken as an in- not seen before (so it was unlikely to elicit instruction-
dicator of the student’s level of reasoning. Assessment ally programmed responses).
tasks required students to describe their reasoning.
However, as always with van Hiele level descrip-
tions and assessments, the devil is in the details. Major
questions for researchers are (a) the appropriateness
A B C
of Gutiérrez and Jaime’s four characterizing processes
and (b) how these processes are defined to occur at Figure 19.3 Triad polygon sorting task.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 855

One difficulty with this analysis is that giving the ers used the triads. Lehrer et al. (1998) construed
number of sides of a polygon is a “low-level” use of each triad task as an indicator of type of reasoning.
properties. That is, there are different types of prop- So students’ use of different types/levels of reasoning
erty use. The simplest property involves describing on different triads was taken as evidence of jumps in
the number of components in a shape. For instance, levels. In contrast, Clements and Battista (2001) used
a quadrilateral has four sides; a triangle has three an- a set of 9 triad items as an indicator of level of students.
gles. A second, somewhat more sophisticated type of To be classified at a given level, a student had to give
property describes global characteristics. For instance, at least 5 responses at that level. If a student gave 5 re-
a quadrilateral is a simple closed curve. Describing the sponses at one level and at least 3 at a higher level, the
straightness of sides also falls in this category. Finally, student was considered to be in transition to the next
and most sophisticated, are properties that describe higher level. The latter approach, by virtue of its use
relationships between parts of shapes. For example, of multiple items, produces a more reliable indication
the following statements describe relational geomet- of level of reasoning. Of course, because it aggregates
ric properties: (a) opposite sides of a parallelogram responses, this approach obscures intertask differ-
are parallel and congruent; (b) a rectangle has four ences and variability in reasoning. It focuses on deter-
right angles; (c) adjacent angles of parallelograms are mining the ascendant level of students’ reasoning on
supplementary; (d) this quadrilateral has one line of a particular type of task. (So this issue is related to the
symmetry. Examples a–c describe relationships be- type-of-reasoning versus level-of-reasoning dichotomy
tween the sides of shapes, (a) explicitly, (b) and (c) previously discussed.)
indirectly by specifying angle measures. Example (d) Another difference between the researchers’ ap-
describes a relationship between two “halves” of a proaches is also important. In analyzing students’ rea-
quadrilateral, which, with further analysis, could be soning on the triad tasks, Lehrer et al. (1998) classi-
described in terms of line segments and angles. fied student responses solely on the basis of the type
The distinction in properties described above sug- of reasoning that students employed. In contrast, in
gests that students’ use of number of sides of a polygon classifying students’ van Hiele levels, Clements and
is not a very good indicator of Level 2 thinking, which Battista (2001) attempted to account for the “quality”
focuses on relational properties. Thus some jumps in of students’ reasoning—each reason for choosing a
levels observed by Lehrer et al. (1998) may have been pair in a triad was assessed to see if it correctly dis-
caused by coding students’ use of number of sides as criminated the pair that was chosen from the third
Level 2. Because a critical factor used in distinguish- item in the triad. In this scheme, the van Hiele levels
ing van Hiele levels is how students deal with geomet- for students were determined based on a complicated
ric properties, clarifying the meaning of properties, as it algorithm that accounted for both type of reasoning
relates to the van Hiele levels, is important. used and discrimination score.
Another factor that should be considered with However, although discrimination is a reasonable
this triad task is that saying Shape B is more like Shape indication of quality of response in the triad context,
C is not necessarily a less sophisticated response than it does not address another component that might be
focusing on number of sides. That is, Shape B is ac- important in assessing van Hiele levels, the validity of
tually more like Shape C if we consider how much reasoning, which involves the accuracy and precision
movement it takes to transform B into C, compared of students’ identifications, descriptions, conceptual-
to B into A. In fact, one could imagine a metric that izations, explanations, and justifications. That is, a stu-
quantifies the amount of movement required. Thus, dent might validly describe a property of two items in
the “morphing” response described by Lehrer et al. a triad that does not discriminate the items from the
(1998), and also observed by Clements and Battista third item. Researchers must determine which, if any,
(2001), may be an intuitive version of a notion whose of discrimination, validity, or mere use of properties
mathematization is far beyond the reach of elemen- are critical characteristics of van Hiele Level 2.
tary students. The last issue in assessment is whether to use in-
Nevertheless, this being said, the original argu- dividual interviews or paper-and-pencil tests. Because
ment that attending to number of sides is more sophis- individual interviews provide a more detailed pic-
ticated than a visual morphing response may still have ture of students’ geometric reasoning, they provide a
some validity. It suggests a shift toward using a more more valid assessment of van Hiele level, but are more
mathematical conceptualization (albeit the primitive time-consuming to administer. In contrast, paper-and-
one of counting number of sides/vertices). pencil instruments are less time-consuming and are
Another shortcoming of the triad-task approach thus more practical for many research projects and
is pointed out by differences in the ways the research- classroom teachers who must deal with large numbers
856 ■ STUDENTS AND LEARNING

of students. But paper-and-pencil instruments, even tion window; it outputs both an image and properties
those that ask students to write about their reasoning, that it extracts from the image. The pattern activation
rarely give as complete a picture of student reasoning subsystem then matches output from the preprocess-
as recorded individual interviews. However, no mat- ing subsystem to stored representations of visual pat-
ter which method is used, creating and interpreting terns that include extracted properties. The pattern
tasks that reliably and validly assess van Hiele levels is activation subsystem is divided into two additional
extremely challenging. subsystems. The category pattern activation subsystem
Final thought on the van Hiele theory. Numerous classifies a stimulus as an element of a category; the
researchers have concluded that the van Hiele levels exemplar pattern activation subsystem registers specif-
accurately, if generally, describe the development of ic examples (e.g., Fido, not dogs in general). The two
students’ geometric thinking, especially about shapes. tasks, “recognizing a stimulus as a specific individual
However, these same researchers, in general, seem and recognizing it as a member of a category, have in-
to find the theory lacking in depth. They would like compatible requirements; information that is needed
a more detailed description of this development. I to recognize the individual must be discarded to clas-
have already described several proposed alternatives sify it as a member of the category” (Kosslyn, 1994, p.
and elaborations. But why does the van Hiele theory, 182). Thus, even at this elemental level, the two es-
though deemed inadequate in detail, also have such a sential geometric tasks of apprehending instances and
strong ring of validity? apprehending categories are differentiated.
I suggest that the reason for this feeling of validity The process of abstraction seems to be at the root
is that the van Hiele levels describe a progression of of construction of entities in the category pattern acti-
thinking that is part and parcel to scientific/mathe- vation subsystem: “To isolate certain sensory properties
matical thinking. That is, we can think of both the lev- of an experience and to maintain them as repeatable
els and the “scientific” method as progressing through combinations, i.e., isolating what is needed to recog-
four phases: perceptualization (resulting in informal, nize further instantiations of, say, apples, undoubtedly
intuitive, surface-level categories and reasoning), constitutes an empirical abstraction” (von Glasersfeld,
conceptualization (resulting in explicit concepts and 1991, p. 57).11 An individual comes to recognize trian-
analysis), organization (resulting in conceptual orga- gles as figural patterns of triangular shapes are abstract-
nization), axiomatization (resulting in formal logical ed and become linked to the word triangle.
treatments). For highly educated individuals in sci-
entifically advanced civilizations—like mathematics Abstraction from the actual sensory material out of
education researchers—moving through these four which particular configurations or sequences are built
phases seems like the natural way for individuals to up, yields figural patterns that have a certain general
progress from intuitive, everyday reasoning to formal applicability and can be semantically associated with
scientific reasoning. specific names. Thus we have, for instance, the notion
of ‘triangularity’ that enables us to see spatial configu-
rations as triangles. (von Glasersfeld, 1982, p. 197)
Psychological Mechanisms Underlying
the Levels Furthermore, von Glasersfeld argues that percep-
tual abstraction is active, not passive.
To understand the cognitive processes underlying
geometric thinking, and movement from one level of
The experiencing subject attends, not to the specific
thinking to the next, one must understand how indi-
sensory content of experience, but to the operations
viduals construct and mentally represent spatial and that combine perceptual and proprioceptive elements
geometric knowledge. Because such knowledge origi- into more or less stable patterns. These patterns are
nates with perception and imagery, I start with these constituted by motion, either physical, or attentional,
processes.10 forming ‘scan paths’ that link particles of sensory ex-
perience. (von Glasersfeld, 1982, p. 196)
Perception of Shape
Kosslyn (1994) has provided a detailed model Individuals form scan paths that enable them to
of the brain processes involved in perception. This recognize similar perceptual objects (e.g., triangles
model consists of several subsystems. The preprocess- that look approximately the same). When several dif-
ing subsystem accepts sensory input from the atten- ferent types of triangular perceptual objects are ab-

10
There seems to be far less relevant literature about cognitive processes underlying thinking at van Hiele levels above Level 2.
11
More will be said about the process of abstraction in a later section.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 857

stracted and associated with the word triangle, an ear- tients suffering brain damage in either the right or left
ly conceptualization of triangle is created. However, hemispheres possessed very different types of cogni-
at this early stage of recognition, an individual’s “con- tive deficits. When asked to draw target stimuli, right-
ceptualization” of triangle is usually limited to a small hemisphere-damaged patients saw the components of
subset of typical prototypes and is holistic in nature. A the stimuli but were unable to correctly put the com-
major problem in placing the van Hiele theory on a ponents together; left-hemisphere-damaged patients
sound cognitive foundation is answering the question, saw the global picture, but not the components of the
How do individuals move to finer, parts-based concep- target stimuli. The patient drawings in Figure 19.4,
tions of shapes? especially the top row, clearly illustrate the global-ver-
sus-parts nature of patients’ cognitive deficits. Patients
From Wholes to Parts with right hemisphere damage saw the components
Kosslyn argued that to form an image of a shape of the original triangle—the segments made from
requires an implicit mental construction in a sequen- rectangles—but were unable to correctly put these
tial, part-by-part manner not only of the parts of the components together spatially. They did not perceive
shape but how they are related; “to represent such or record the overall spatial structure of the triangle.
shapes is to extract specific properties that will not In contrast, left-hemisphere damage seemed to cause
change when the object assumes a new configuration little deficit in global spatial processing; subjects cor-
or a shape-variant appears. One such invariant is the rectly captured the shape of the triangle.12 Left-hemi-
type of spatial relation between parts” (Kosslyn, 1988, sphere damage did, however, cause important details
p. 1623). Kosslyn (1994) further argued that in object to be omitted from figures. Such omissions can be
recognition, global shape is processed first, then parts important geometrically. For instance, Kosslyn (1994)
and characteristics. That is, in perceptual processing, found that the left hemisphere “is critically involved
the mind first attempts to match the overall pattern in detecting target figures when they are embedded
that corresponds to an object; if that does not work, in other patterns” (p. 202).
it attempts to match input to representations of in-
dividual parts. Thus, even when a shape is processed
globally, the perceptual system identifies the shape’s
parts and computes relatively abstract spatial relations
among them. Initially, however, the parts are not ex-
plicitly accessible.
Kosslyn’s research is consistent with that of Smith
(1989). She suggested “that represented objects are
built in a bottom-up fashion from a feature level of
processing” but “that the represented object is a co-
hesive unit and is experienced as a unit. Represented
objects as wholes are given to conscious experience as
unitary entities, so that one cannot get to the repre-
sented parts without some work” (p. 127). But in addi-
tion to Kosslyn’s claim that task demands may cause a
holistic-to-parts shift in processing, Smith argued that
as children get older, they acquire an increased abil- Figure 19.4 Global-versus-parts deficits from hemispheric
ity to selectively attend to single dimensions, and thus damage (Posner & Raichle, 1994).
the characteristics of wholes. Clements and Battista
(1992, 2001) described the processes by which parts Posner and Raichle (1994) explained these hemi-
of shapes gradually become accessible to reflection, spheric differences in terms of attentional scale—
but their theory focused on experience, reflection, zooming in or out: “The right hemisphere is more
and learning. likely to control large or global scales [zoomed out]
Research on brain processes goes even further, sep- while the left does the reverse [zoomed in]” (p. 162).
arating the identification of shape parts, the creation Kosslyn (1994) too argued that a visual object “is rep-
of spatial relations between parts, and the creation of resented at multiple scales; at one scale the parts are
the whole. Posner and Raichle (1994) found that pa- represented, whereas at another they are embedded

12
Posner and Raichle (1994) reported, however, some research showing that left-hemisphere damage can cause deficits in assembling parts
into overall representations.
858 ■ STUDENTS AND LEARNING

in an overall pattern” (p. 161). Furthermore, research sion relation between parts and the wholes to which
has found that observers can selectively focus on ei- they belong is characteristic of Level 1 thinking and
ther global or local features in a visual form. “When is similar to a type of conceptualization described by
we are set to attend to one level of detail, our process- Piaget (Gruber & Voneche, 1977), who claimed that
ing of features at the other level is poorer” (Coren “an arm drawn alongside of a manikin is conceived by
et al., 1994, p. 518). Thus the global-local processing the child as ‘going with’ the manikin not as ‘forming
separation is built into the architecture of the brain. part of’ his body” (p. 102).
However, as individuals learn and become more so- Second, students must mentally construct explicit
phisticated in their reasoning, the separate processes spatial relationships between the parts of a shape that
get coordinated and integrated into higher level rec- give it its spatial structure. This requires the mind to
ognition schemes that are sensitive to the whole, its “break into” the components of perceptual process-
parts, and relationships between parts. Furthermore, ing for identifying a shape, abstract structural spatial
this integration is so automatic in educated adults that relationships that characterize the shape, and con-
it is difficult for them to appreciate the world of the sciously attempt to conceptualize these relationships.
child in which the processes remain separate and un- According to the theory of abstraction to be discussed
integrated. later, breaking into the perceptual processes requires
In summary, the psychological/neurological re- intentional reflection on the nature of the shape, and
search described above sheds light on several find- that the perceptual processes be abstracted at the in-
ings in research on geometric thinking. It indicates teriorized level. Once spatial relationships become ac-
that perceptual wholes are built up from parts-based cessible to reflection, they can be verbally described,
representations, but that accessing these parts-based refined through empirical testing, and later, linked to
representations is initially impossible and only occurs interiorized formal conceptualizations.
“with some work.” Furthermore, there is a separation Several factors can engender the move from im-
in perceiving parts of shapes and relationships be- ages to explicit relationships. First, as has been seen,
tween parts, and there is a separation in perceiving the ability to break into perceptual processes can be
global or local features of shapes. These findings help affected by development and experience. Second,
explain three important issues in research on the de- Kosslyn discussed an imagery-to-verbalization devel-
velopment of geometry reasoning: (a) why students opmental shift. He claimed that even though children
first attend to whole shapes; (b) why students often rely predominantly upon imagery in accessing and us-
notice parts of shapes, but not relationships between ing information, because processing capacity for ver-
parts (Clements & Battista, 1992); and (c) how stu- bal knowledge increases with development, children
dents shift from focusing on wholes to focusing on increasingly translate observations derived from im-
relationships between parts, as described in the van ages into verbal form so facts can be accessed either
Hiele theory. imagistically or verbally (Kosslyn, 1980, 1983; Kosslyn,
Margolis, & Barrett, l990). In fact, Kosslyn (1980) sug-
From Images to Explicit Relationships gests that the use of imagistic representation may be
How does a child progress from viewing shapes reduced as children gain experience and repeatedly
merely as visual images to thinking about them pri- access the same ideas. That is, a child “may recode
marily in terms of verbal statements about their parts information into propositional form as he or she
and relationships between parts? First, students must accesses that information repeatedly” (p. 412). Fur-
gain mental access to the parts of shapes, explicitly thermore, because “certain facts are communicated
considered as parts of shapes, so that they can mental- linguistically and presumably memorized in a propo-
ly construct spatial relationships between the parts. It sitional format. . . . Retrieval . . . should not require
is not that before this time students cannot notice the consultation of an image” (p. 412).
parts of shapes; it is that they do not see these parts as
parts of wholes—“parts” are merely seen as self-stand- Comment
ing entities. In fact, initial processing likely consid- Although a number of theories and studies have
ers parts and wholes on equal footing—the process- been reviewed in an attempt to describe the cogni-
ing system may not detect whether perceived entities tive processes by which students progress through the
are wholes or parts of wholes (Kosslyn, 1994, p. 168). early van Hiele levels, this area of research is still in
Thus, a student’s description that a shape is “pointy” its infancy. This is due in great part because research-
does not imply that the student is attending to a part ers are investigating cognitive processes that cannot
of the whole; the student may simply be focusing on a be observed. To achieve progress in this domain, it
salient entity in the visual field. This lack of an inclu- is important for mathematics education researchers
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 859

to heed the work of researchers in other fields such erated on unless they are physically present. Further-
as cognitive science and neuroscience. Such research more, “the abstracted operational pattern necessary
can provide valuable insights into these difficult-to-ob- to recognize things of a kind does not automatically
serve processes. But we must also continue to conduct turn into an image that can be re-presented” (von Gla-
studies that expose students to instructional treat- sersfeld, 1995, p. 93).
ments and tasks and carefully investigate phenom- Internalization. When material has been suffi-
enologically how students construct meaning for geo- ciently abstracted so that it can be re-presented in the
metric concepts, always attempting to build cognitive absence of perceptual input, it has been internalized.
models that explain what we observe. It is through the Internalization is
integration of cognitive and phenomenological inves-
tigations that we can make true progress in this vital the process that results either in the ability to re-pres-
area of research. ent a sensory item without relevant sensory signals be-
ing available in actual perception or in the ability to
reenact a motor activity without the presence of the
kinesthetic signals from actual physical movement.
OTHER THEORIES RELEVANT Internalization leads to ‘visualization’ in all sensory
TO GEOMETRIC LEARNING modalities. (Steffe & Cobb, 1988, p. 337)

Several other theories seem particularly relevant for According to von Glasersfeld, it is at the internal-
understanding geometric reasoning. ized level that a concept14 has been formed, a concept
referring “to any structure that has been abstracted
from the process of experiential construction as re-
The Theory of Abstraction13
currently usable. . . . To be called ‘concept’ these con-
Learning occurs as individuals recursively cycle structs must be stable enough to be re-presented in
through phases of action (physical and mental), re- the absence of perceptual ‘input’” (von Glasersfeld,
flection, and abstraction in a way that enables them 1982, p. 194). However, for material that has only been
to develop ever more sophisticated mental models. internalized, one cannot reflect upon its re-presenta-
These mental models consist of abstracted objects and tion to consider how it is composed (Steffe & Cobb,
abstracted actions that can be performed on those ob- 1988). That is, one can only re-present internalized
jects. Once objects and actions have been sufficiently material, not analyze its structure.
abstracted, they themselves become mental objects Interiorization. “Reflecting upon the results of a
that can be mentally operated on (e.g., compared, re-presentation requires detachment and placing the
decomposed, connected). The levels of abstraction re-presented activity at a distance in order to analyze
described below indicate major landmarks in the pro- its structure and composition” (Steffe & Cobb, 1988,
cess of abstracting mathematical objects, affording in- p. 17). To do so requires interiorization, which is “the
creasingly powerful reasoning. most general form of abstraction; it leads to the isola-
tion of structure (form), pattern (coordination), and
Levels of Abstraction operations (actions) from experiential things and ac-
Perceptual abstraction. At the perceptual/recog- tivities” (p. 337). Material has been interiorized when
nition level, abstraction isolates an item in the experi- abstraction has disembedded it from its original per-
ential flow and grasps it as an object: “Focused atten- ceptual context and it can be freely operated on in
tion picks a chunk of experience, isolates it from what imagination, including “projecting” it into other per-
came before and from what follows, and treats it as ceptual material and utilizing it in novel situations. To
a closed entity” (von Glasersfeld, 1991, p. 47). When use an abstracted item for thinking about a new situa-
items are perceptually abstracted, we become con- tion requires that the item be reprocessed, not simply
scious of them, entering them into working memory. recalled as it has been encountered in the past—re-
Perceptual abstraction isolates those sensory proper- processing is the operationalization of interiorization
ties of an experience that are needed to recognize fur- (Steffe, 1998).
ther instantiations of the experience. However, items Note that researchers should be careful with the
that have been abstracted only at the perceptual level notion that abstraction “disembeds” material from its
cannot be re-presented (visualized) and cannot be op- original context. As Hoyles and Healy (1997) have

13
Other views of abstraction in geometry can be found in (Hollebrands, 2003; Hoyles & Healy, 1997) and (Mitchelmore & White, 2000).
14
In this case, von Glasersfeld’s “concept” is the same as my “conceptual object” or “conceptualization.”
860 ■ STUDENTS AND LEARNING

noted, abstraction can be situated in that, at least at individuals able to reason meaningfully about these
lower levels, it is a process that occurs within, and re- symbolic statements without having to re-present the
mains linked to, the context in which it first occurs. actual operations for which they stand.
That is, at the perceptual and internalized levels of
abstraction, context is an inseparable part of an ab- A Transformational Perspective on Relational Operations
straction. At the interiorized level, abstracted material What is the nature of the operations that establish
becomes separable from its original context, although interrelationships between components of shapes?
links to that context may remain. In fact, these links One hypothesis is that the relationships are estab-
may be an important part of applying interiorized ma- lished with unconscious visual transformations. For
terial in novel situations. Generalized forms of links example, but differing in detail from Leyton (1992),
may form part of the structure of an abstraction that Battista (in press) posits that in a parallelogram, seeing
helps individuals recognize new situations in which the relationship that opposite sides are parallel might
the abstraction might be applied. be based on mentally (but unconsciously) translating
Second level of interiorization. When material one side onto the opposite side. Similarly, seeing that
reaches the second level of interiorization, operations opposite angles in a parallelogram are congruent may
can be performed on the material without re-present- be based on an unconscious 180° rotation.
ing it and symbols, acting as “pointers” to the original- This analysis is consistent with the observation
ly abstracted material, can be used to substitute for it. that parallelism is more difficult for students to “see”
in typical trapezoids than in rectangles and parallelo-
The Theory of Abstraction Applied to the van Hiele Levels grams. Seeing parallelism in scalene trapezoids can-
To apply the theory of abstraction to the van not be accomplished using the same mental transfor-
Hiele theory, recall that action-based mental process- mation that establishes this property for rectangles
es enable students to recognize and construct images and parallelograms because in a scalene trapezoid,
of shapes—Kosslyn’s sequential part-by-part mental one side is not the translation image of its opposite
constructions, or von Glasersfeld’s scan paths. During side. Also consistent with this hypothesis is the com-
the first van Hiele level, these processes are abstracted mon student belief that the property “opposite sides
at the perceptual and internalized level to recognize parallel” goes with “opposite sides equal” because the
and visualize common shapes, respectively. To move first property is established with a translation that also
up to the second van Hiele level, shape recognition establishes the second. To disassociate these two prop-
processes must be raised to the interiorized level, al- erties, parallelism might need to be established, for
lowing these whole-shape recognition processes to example, with a translation and a stretch.
be decomposed, and thus the components of shapes
to be recognized, operated on, and interrelated. Be- Special Types of Abstraction
cause interiorization of whole-shape recognition pro- Three special forms of abstraction are funda-
cesses permits relational operations to be performed mental to geometry learning and reasoning (Battista,
on shape components, relationships between compo- 1999a; Battista & Clements, 1996). Spatial structuring
nents of shapes can be noticed, starting another cycle is the mental act of constructing and abstracting an
of abstraction, focused not on recognition but on re- organization or form for an object or set of objects.
lationships. As these relational operations between It determines an object’s nature or shape by identi-
components of shapes are abstracted at the percep- fying its spatial components, combining components
tual and internalized levels, students can describe re- into spatial composites, and establishing interrela-
lationships between components of individual shapes tionships between and among components and com-
informally. Subsequently, when relational operations posites. Mental models are sets of abstractions that are
become interiorized, they can be applied to sets integrated to form nonverbal mental representations
of shapes so that the student can see that a class of that are activated to interpret and reason about situ-
shapes has a set of characterizing properties. At this ations (more will be said about mental models in a
point a student is capable Level 2, descriptive/ana- later section). A scheme is an organized sequence of
lytic, reasoning. General properties, applied to classes actions or operations that has been abstracted from
of shapes, can be described. To reach the third van experience and can be applied in response to similar
Hiele level, relational operations must achieve “sym- circumstances. It consists of a mechanism for recog-
bol” status (von Glasersfeld, 1995)—that is, they must nizing a situation, a mental model that is activated to
reach the second level of interiorization. At this level, interpret actions within the situation, and a set of ex-
verbal/symbolic statements of relationships can act as pectations (usually embedded in the behavior of the
substitutes for, or pointers to, these operations, with model) about the possible results of those actions.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 861

Three Types of Structuring is, “parts of the model correspond to the relevant
Geometry learning can be viewed as involving parts of what it represents, and the structural rela-
three types of structuring (Battista, in press b). (a) tions between the parts of the model are analogous to
Spatial structuring constructs a spatial organization or the structural relations in the world” (Johnson-Laird,
form for an object or set of objects. It determines a 1998, p. 447). Individuals reason about a situation by
perception/conception of an object’s nature or shape activating mental models that enable them to simu-
by identifying the object’s spatial components, com- late interactions within the situation so that they can
bining components into spatial composites, and estab- explore possible scenarios and solutions to problems.
lishing interrelationships between and among com- When using a mental model to reason about a situa-
ponents and composites. (b) Geometric structuring tion, a person can mentally move around, move on
describes spatial structurings in terms of formal geo- or into, combine, and transform objects, as well as
metric concepts. That is, in geometrically structuring perform other operations like those that can be per-
a spatial situation, a person uses geometric concepts formed on objects in the physical world. Individuals
such as angles, slope, parallelism, length, rectangle, reason by manipulating objects in mental models and
coordinate systems, and geometric transformations to observing the results. “The behavior of objects in the
conceptualize and operate on the situation. In order model is similar to the behavior of objects that they
for a geometric structuring to make sense to a person, represent, and inferences are based on observing the
it must evoke an appropriate spatial structuring. (c) effects of the operations” (Greeno, 1991, p. 178). Im-
Logical structuring formally organizes geometric con- portantly, individuals’ use of mental models is con-
cepts (i.e., geometric structurings) into a system and strained by their knowledge and beliefs. That is, much
specifies that interrelationships must be described of what happens when we form and manipulate a
and established through logical deduction. To accom- mental model reflects our underlying knowledge and
plish a logical structuring, individuals must logically beliefs about what would happen if we were dealing
organize sets of properties. The construct of structur- with the objects they represent. So the properties and
ing, construed more generally, provides another per- behavior of objects in a mental model simulate the
spective on the van Hiele theory: Spatial structuring is properties and behavior we believe the objects they
the major focus in van Hiele Level 1 and the transition represent possess.
to Level 2, geometric structuring is the major focus I will now discuss several examples of mental-mod-
at Level 2 (when it has been broadly accomplished, el-based reasoning in geometry. The first two examples
Level 2 is fully attained), and logical structuring is the are reformulated from Clements and Battista (2001).
focus at Levels 3 and 4 (“axiomatic” structuring is the A second-grade student was investigating geom-
focus at Level 5). etry in a Logo environment. After she had discovered
that she needed 90° turns to make a tilted square, she
was asked to draw a tilted rectangle. She immediately
Construction and Use of Mental Models15,16
used 90° turns at the vertices. She was asked, Why?
An abundant amount of research indicates that
many forms of reasoning are accomplished with men- CA: Because a rectangle is just like a square but
tal models (Battista, 1994; Calvin, 1996; English & just longer, and all the sides are straight. Well,
Halford, 1995; Greeno, 1991; Johnson-Laird, 1983; not straight, but not tilted like that (makes an
Markovits, 1993). According to the mental model view acute angle with her hands). They’re all like
of the mind, individuals understand or make sense of that (shows a right angle with her hands) and
a situation, including a set of connected verbal propo- so are the square’s.
sitions describing a situation, when they activate or
construct a mental model to represent the situation In later discussions, she also stated that a square
(Johnson-Laird, 1983). Mental models are nonver- is a rectangle.
bal recall-of-experience-like mental versions of situa-
tions that have structures isomorphic to the perceived Obs: Does that make sense to you?
structures of the situations they represent (Battista, CA: It wouldn’t to my [4 year old] sister but it sort
1994; Greeno, 1991; Johnson-Laird, 1983, 1998). That of does to me.

15
Researchers employing the theory of mental models have actually created computer models of mental models for specific situations.
16
Mathematics education researchers using the notion of concept image have reached conclusions similar to researchers embracing the
idea of mental models, specifically that students most often do not use definitions of concepts, but rather concept images—a combination
of all the mental pictures and properties that have been associated with the concept—to make decisions. (See Clements & Battista, 1992
for more details.) Concept images are similar to mental models.
862 ■ STUDENTS AND LEARNING

Obs: How would you explain it to her? rotated from the vertical-horizontal-side position of-
CA: We have these stretchy square bathroom ten do not look like squares. M’s mental-model-based
things. And I’d tell her to stretch it out and transformation was unconstrained by her knowledge
it would be a rectangle. (Battista & Clements, that all sides of a square have the same length. Thus,
1991, p. 18) these episodes raise the question of how students’
conceptual knowledge gets incorporated into their
It “sort of made sense” that a square could be mental models and visualization.
thought of as a rectangle because CA’s mental model A different kind of example is provided by a sec-
of a square enabled her to envision a square being ondary mathematics teacher attempting to decide on
stretched into a rectangle. Because CA had just demon- the validity of the following statement: If all the angles
strated her knowledge that squares and rectangles are in a polygon are equal, then all the sides in the poly-
similar in having angles made by 90° turns, this response gon are equal. The teacher said that in thinking about
suggests that she may have understood intuitively that the problem he first imagined a regular pentagon,
all rectangles can be generated from one another by which he attempted to transform so that it had un-
certain transformations, ones that preserve 90° angles. equal angles. But he could not imagine how to change
CA’s mental models of shapes such as squares and rect- some but not all of the sides of the pentagon without
angles incorporated allowances and constraints on changing the angle measures. The mental model this
how they could be acted upon and transformed. In this teacher activated to think about this problem made
instance, CA reasoned by performing a simulation of the problem impossible for him to solve. When I asked
changing a square into a rectangle using a special type him to think about a rectangle, he immediately solved
of visual transformation, one that incorporated formal the problem. Interestingly, had the teacher originally
geometric constraints—preserving 90° angles—on her activated a mental model of a square or regular hexa-
mental objects and actions. gon, it would have been much easier to implement
Sometimes, however, students’ mental models his transformational strategy. This example illustrates
are inadequate to draw valid conclusions. For in- that an important question for researchers is, What
stance, a second grader was examining her attempt determines which mental models are activated during
to draw a tilted square in Logo. (She too had suc- problem solving, and what kinds of instruction guide
cessfully taught the turtle to draw squares of various students to activate the most useful models?
sizes, all of which had sides vertical and horizon-
tal.) Because she did not use 90° turns for the tilted Mental Models and Imagery17
square, her figure was only a crude approximation
Images are mental representations “that ‘stand in’
to a square. In fact, though M had earlier stated that
for (re-present) the corresponding objects” (Kosslyn,
all sides of a square are the same length, in drawing
1994, p. 1). Images can be evoked in the absence of
her square, she abandoned this constraint to close
perceptual material (re-presentation or visualization)
the figure (because her angles were not 90°). But M
or by perception (e.g., looking at a diagram). Their
intended to draw a square and concluded that it was
formation relies on regions of the brain that support
a square, reasoning as follows:
depictive representations.
Int: How do you know it’s a square for sure?
In a depictive representation, each part of an object
M: It’s in a tilt. But it’s a square because if you
is represented by a pattern of points, and the spa-
turned it this way it would be a square. tial relations among these patterns . . . correspond
to the spatial relations among the parts themselves.
The difference in the validity of CA’s and M’s rea- Depictive representations convey meaning via their
soning seems to have been due to the degree to which resemblance to an object, with part of the represen-
they incorporated relevant knowledge into their men- tation corresponding to parts of the object. (Kosslyn,
tal models. On the basis of a variety of tasks, the re- 1994, p. 5)
searchers found that CA’s thinking about shapes was
based more on properties than M’s, so CA’s mental A number of researchers have suggested that
models were more property-based and her imagery there are two major types of imagery. Following La-
more influenced by property-based considerations. M koff, Wheatley (1998) distinguished between rich im-
based her thinking on simple visualizations—she, like ages and image schemata. Rich images are static, fixed,
all students in her class, had seen that squares that are and contain much visual detail. Image schemata repre-

17
Although I focus on visual imagery, similar discussion can be given for imagery in other modalities such as auditory or kinesthetic.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 863

sent spatial relationships and can be transformed in through 3. I will summarize some of the relevant find-
various ways.18 ings of that research.
The distinction between rich images and image Not only do human minds naturally categorize
schemata can be explained by the more general theo- objects (Pinker, 1997), categorization is essential for
ry of mental models. Johnson-Laird and colleagues ar- intelligent thought and action (Ross, 1996). What the
gued that it is important to distinguish between visual mind gains from forming categories “is inference. Ob-
images (which correspond to rich images) and spatial viously we can’t know everything about every object.
mental models (which correspond to image schema- But we can observe some of its properties, assign it to
ta). Visual images are derived from perception or gen- a category, and from the category predict properties
erated from mental models, with images being views that we have not observed” (Pinker, 1997, p. 307). For
of mental models from particular viewpoints. Images example, if, on seeing round, fist-sized reddish-green
are mentally manipulable when appropriate mental objects attached to the branches of a tree, you catego-
models have been constructed from them (Johnson- rize the objects as apples, you can then infer that they
Laird, 1998). Thus, in visual reasoning, sometimes are edible and have seeds.
people merely construct and scan visual images, and Research suggests that people form two kinds of
other times they operate on spatial mental models categories (Pinker, 1997). Fuzzy or natural categories,
(Knauff, Fangmeier, Ruff, & Johnson-Laird, 2003). such as games, are formed in everyday activity. Such
Furthermore, sometimes, when aspects of visual im- categories generally have no clear definitions, have
ages are irrelevant to a task, the evocation of visual fuzzy boundaries, and are conceptualized mainly in
images can actually interfere with construction and terms of stereotypes and family-like resemblances—
use of an appropriate mental model and thus impede some examples are deemed “better” than others.
reasoning (Knauff et al., 2003). Identification of instances seems to be the major goal.
The distinction between images and mental mod- The second type of category consists of formal, explic-
els can also be interpreted in terms of the theory of itly defined categories, such as odd numbers, in which
abstraction. A re-presentation of a visual image results all instances are logically equivalent as representative
from an internalization of a perception of a spatial ob- of the class. Formulating and studying precise defini-
ject. A mental model of a spatial entity results from in- tions is the goal.
teriorization of the mental processes used to conceive Learning geometry involves the formation of
of the entity: “The operations that are carried out in both fuzzy and formal categories. Before geometry
reasoning with models . . . are conceptual and seman- instruction, and even in primary-grade teaching,
tic” (Johnson-Laird, 1998, p. 457). Thus, to reason as children encounter shapes in the physical world
geometrically about an image, one must construct a and those shapes are named by adults, children treat
proper mental model of the image, one that captures shapes as fuzzy categories—shapes are identified but
the image’s relevant spatial structure (and the depth not defined. Subsequent school instruction demands
of that structure depends on the sophistication of the that students conceptualize shapes in terms of verbally
concepts used in this structuring). That is, geometric stated, property-based definitions. Difficulties often
reasoning about figures requires a reconstitution of occur as students attempt to reconcile fuzzy categories
images in terms of appropriately structured mental with formal concepts. The language used in fuzzy and
models. Indeed, if figures are thought about as pic- formal concepts may be the same, but the underlying
tures (using images), not as sets of spatial/geometric cognitive entities are vastly different. For instance, log-
relationships (using mental models), students are not ically, a square is as much a rectangle as a nonsquare
thinking about formal geometry (cf. Laborde, 1993). rectangle. But students who possess fuzzy concepts of-
ten resist the idea that a square is a rectangle—they
think it is only a square. Progressing from the visual
Concept Learning
to the descriptive/analytic van Hiele level seems very
Because concept learning, use, and analysis play much like moving from fuzzy to formal categories for
a significant role in the development of geometric a set of concepts in a domain.
thinking, research on the formation and use of con- There seem to be several competing theories for
cepts provides insights into that development, espe- how people develop fuzzy categories. The most prom-
cially for understanding thinking at van Hiele Levels 1 inent seems to be that people develop prototypes—

18
The distinction between rich images and image schemata seems related to Piaget’s notion of figurative knowledge, which results from
empirical abstraction, and operative knowledge, which results from reflective abstraction: “Perception yields a series of static, centration-
afflicted snapshots. The knowledge we get from perception is figurative, not operative” (Piaget, 2001, p. 21).
864 ■ STUDENTS AND LEARNING

mental images or examples that include, usually im- commonalities” (p. 276). Indeed, structure mapping
plicitly, features that are relatively common among is a central learning mechanism that enables individu-
members of a category. People decide whether an ob- als to notice and store abstract relational properties,
ject belongs to a category by determining if it is suf- derive abstract knowledge from instances, and extend
ficiently similar to prototypes (Smith, 1995). that knowledge to new cases. This happens as the rela-
Researchers describe two ways that prototypes tions examined in structure mappings are abstracted
might be formed (Krascum & Andrews, 1993). Ac- to become objects in and of themselves.
cording to the abstractionist model, after experiencing Because Gentner and Medina’s (1998) theory
a number of category examples, a perceiver abstracts focuses on structure mappings, it has the potential to
the category features that occur most frequently into shed light on the mechanisms by which students for-
a prototype representation. According to the exemplar mulate properties of geometric shapes when formal
model, category examples are not decomposed into definitions are not given. The developmental/expe-
attributes but instead are abstracted to form a repre- riential shift from perceptual to relational similarity
sentation that consists of a set of discrete, more-or-less seems particularly relevant in progressing from van
intact instances. Krascum and Andrews (1993) stated Hiele Level 1 to 2 because Level 2 reasoning focuses
that many researchers favor the examplar view be- on relationships between parts of shapes. Gentner and
cause of findings that show that young children are Medina’s theory suggests that giving common names
insensitive to component features of objects, show de- to like geometric shapes should cause students to start
velopmental improvement in the ability to selectively performing structure mappings that can enable them
attend to single dimensions of stimuli, and perceive to see, at least informally, spatial relations common
objects in global, unanalyzed fashion. However, Kras- to the shapes. But we do not know how spontaneous
cum and Andrews also cited evidence that is incon- these are. Perhaps for most students, the mappings
sistent with the exemplar model. For instance, they occur only after explicit prompts such as, “How are
reported that young children do not necessarily fail to these shapes alike?” Additional research is needed to
decompose objects into their component parts and in examine the potential of the structure-mapping the-
fact often fixate on single features. The ability to dis- ory in explaining the transition to the early phases of
tribute attention over many stimulus dimensions de- van Hiele Level 2.
velops over time. Interestingly, as researchers observe Vygotsky’s (1986) work sheds further light on the
students’ learning of geometric shapes, we see many relationship between fuzzy and formal concepts. He
of the behaviors described above—students form pro- argued that children become conscious of their spon-
totypes, often limited; they fail to notice important taneous (i.e., fuzzy) concepts relatively later, with the
spatial attributes; they sometimes fixate on particu- ability to define them in words appearing long after
lar features. Also, Kosslyn’s description of perception they have “acquired” the concepts. In contrast, the
seems more consistent with the exemplar model and development of scientific (formal) concepts usually
addresses many of the issues raised about it. begins with their verbal definition. However,
In considering how fuzzy categories of (nongeo-
metric) shapes are formed, Gentner and Medina though scientific and spontaneous concepts develop
(1998) argued that there is a developmental/experi- in reverse directions, the two processes are closely con-
ential shift from perceptual similarity (e.g., recogniz- nected. The development of a spontaneous concept
ing a mobile that is a very close perceptual match to must have reached a certain level for the child to be
one seen before) to relational similarity (e.g., recog- able to absorb a related scientific concept. . . . In work-
ing its slow way upward, an everyday concept clears a
nizing the similarity of a mitten covering a hand and
path for the scientific concept and its downward devel-
a shoe covering a foot). These researchers argued opment. Scientific concepts, in turn, supply structures
that the perception of similarity is accomplished by for the upward development of the child’s spontaneous
a process of “alignment or structure-mapping,” and concepts toward consciousness and deliberate use. Sci-
that performing such mappings increases the likeli- entific concepts grow downward through spontaneous
hood of establishing relational rather than purely concepts; spontaneous concepts grow upward through
visual similarities.19 They hypothesized that giving scientific concepts. (p. 194)
a common name to shapes in a class can encourage
individuals to make comparisons that activate struc- Thus, applying this theory to geometric concepts,
ture mappings that “promote the discovery of deeper we see that fuzzy, recognition-based conceptualiza-

19
A structure mapping is isomorphic-like; it is one-to-one and if two elements are linked in one representation, their images are linked in
the corresponding representation.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 865

tions of geometric shapes develop first as the shapes I will discuss three major types of computer en-
are perceived and named in school and out. For com- vironments for investigating plane shapes (see Clem-
mon shapes, and most students, such early experi- ents and Battista, 1992 for a discussion of other types):
ences “naturally” lead to van Hiele Level 1 reasoning. Logo-based, Geometric Supposers, and Dynamic Ge-
But getting to Level 2, in which shapes are reasoned ometry Environments (DGEs).21 Two essential features
about in terms of formal geometric concepts, requires have been noted for such computer environments.
a great deal of work—work that, to be meaningful to First, they require students to provide explicit specifi-
students, must connect to the fuzzy concepts devel- cations for geometric shapes—using menu selections
oped in Level 1. in Supposers and DGEs, and command lists in Logo.
Finally, at the formal end of the spectrum on cat- Thus, unlike using paper and pencil, in these com-
egorization, is the issue of hierarchical classification. puter environments, students cannot make drawings
According to Jones (2000), following De Villiers, clas- without some level of conceptual and representational
sifications can be hierarchical or partitional. Hierar- explicitness. Researchers have argued that this explicit-
chical classifications use inclusive definitions such as ness promotes and supports reflection on and abstrac-
specifying that trapezoid is a quadrilateral with at least tion of geometric concepts, and movement toward
one pair of sides parallel—which means that a par- van Hiele Level 2. For instance, Clements and Battista
allelogram is a special type of trapezoid. Partitional (1992), following Papert, claimed that,
classifications use exclusive definitions such as specify-
ing that a trapezoid is a quadrilateral with only one Writing a sequence of Logo commands, or a proce-
pair of sides parallel—which excludes parallelograms dure, to draw a rectangle . . . obliges the student to
as trapezoids. In general, in mathematics, inclusive externalize intuitive expectations. When the intuition
is translated into a program it becomes more obtru-
definitions (and thus hierarchical classifications) are
sive and more accessible to reflection. . . . Students
preferred, although exclusive definitions and parti- must analyze the spatial aspects of the rectangle and
tional classifications are certainly not mathematically reflect on how they can build it from parts. (p. 450)
incorrect. However, a number of studies show that
many students have great difficulty with the hierarchi- Similarly, Laborde (2001) claimed that when con-
cal classification of quadrilaterals (de Villiers, 1994; structing a square with a given side in a DGE,
Jones, 2000). Although research on the van Hiele
levels has shed light on this issue, more research is With paper and pencil the . . . task is controlled by
needed that investigates why this difficulty occurs and perception. . . . The same task in Cabri . . . cannot
how to overcome it. be obtained by eye . . . but uses a circle as a tool
for transferring a given distance. The task in Cabri
requires more mathematical knowledge about the
properties of a square and the characteristic prop-
STUDIES OF LEARNING erty of a circle. (p. 294)
AND TEACHING GEOMETRY
In the case of DGE research, the amount of ex-
plicitness used by students has been, indirectly, a
Computer Environments for Learning
frequent focus of discussion. For instance, in The
Geometry20
Geometer’s Sketchpad (GSP) or Cabri, one can “draw” a
Early research on the use of computer environ- rectangle by creating four line segments in a rectan-
ments for learning geometry focused primarily on in- gular configuration. Little conceptual explicitness is
vestigating predictions about student learning made required, and the drawing does not remain rectangu-
by environment designers. More recent research, lar when its vertices are dragged. Alternately, one can
informed by initial studies and extensive experience use DGE commands to “construct” a rectangle that
with the environments, has investigated more closely preserves its rectangularity with dragging. In this case,
the learning processes that occur within the environ- it can be argued that each step in the construction
ments and how the environments affect and shape requires some conceptual explicitness. The question
student learning. is, how much? For instance, after creating a segment,

20
Even though research shows that computer environments have great promise for improving geometry instruction, computer use among
mathematics teachers is low. Becker (2000) reported that only 1 of 9 secondary mathematics teachers said that a typical student in their
classes used computers on more than 20 occasions during a 30-week period.
21
Most of the relevant research has been conducted on these three environments. Specialized restricted applets and general drawing
programs are not discussed.
866 ■ STUDENTS AND LEARNING

students might create a segment perpendicular to it shape created may not be a rectangle at all, merely
using the perpendicular construction. So they must a visual approximation of a rectangle; so even if the
know, at a minimum, that a specific spatial charac- set of commands is considered, the properties of the
teristic of a rectangle is created by the perpendicular rectangle are not accurately embedded in the com-
construction. Then they must use either perpendicu- mands. Thus, the conceptual explicitness that occurs
larity or parallelism to construct the remainder of the in computer environments for geometry learning
figure. Because only three right angles are explicitly depends on a complex interaction between the com-
constructed, how do students know that the last angle mands given in constructing figures, the availability
must be a right angle? Do they deduce it logically, con- of those commands to student inspection, the rea-
duct an informal componential analysis (see the previ- soning of the student, and instruction.
ous description of the Borrow/Battista elaboration of The second feature attributed to computer envi-
van Hiele Level 3), or merely make a visual assump- ronments for facilitating geometry learning is the “re-
tion? Other properties, such as parallelism or congru- peatability” of drawings (Laborde, 1992). “In this kind
ence of opposite sides, are not explicitly constructed, of software a necessary condition for a construction
so may not be recognized in the construction (but to be correct is that it produce several (or an infinity
might be visually induced). of) drawings which preserve the intended properties
Understanding the nature of the conceptual ex- when variable elements of the figure are modified”
plicitness that occurs in students’ constructions of (Laborde, 1992, p. 129). Repeatability is achieved
draggable DGE drawings requires much additional very differently in the different programs. In Logo, re-
research. For instance, are students’ conceptualiza- peatability requires the use of procedures, and often
tions of their constructions manifestations of their variables, adding more syntax requirements. In Sup-
understanding of abstract mathematical concepts, posers, repeatability is discrete and constrained. In a
or are they proceduralized methods for creating DGE, repeatability is continuous and dynamic. Again,
visual configurations? That is, when students con- however, a major issue for research is not only what
struct a perpendicular in a DGE, is the perpendicu- kinds of repeatability engender what kinds of learn-
lar simply a particular kind of visual configuration, ing, but how students use and, more importantly, con-
or is it an instantiation of a geometric concept (i.e., ceptualize repeatability. Do students really see drag-
two lines intersecting at 90° angles)? Another diffi- gable constructions in DGEs as representing multiple
culty with DGEs is that gaining access to commands elements in a given class of shapes? Or, are such con-
used in a completed construction is not always structions simply another kind of geometric object
straightforward. For instance, to determine how a that can be studied?
draggable rectangle was constructed in GSP, one This latter point is critical and returns the fo-
can check all properties of various objects on the cus to diagrams. In paper-and-pencil approaches to
rectangle (but the sequence of constructions is not studying geometry, there are two major types of enti-
readily available), or one can create a tool for the ties. First, there are diagrams, mere perceptual ob-
construction and examine its script. However, if stu- jects. Second, there are shape conceptualizations—
dents do not sufficiently understand the syntax of conscious meanings activated for a class of shapes.
GSP—and learning syntax can be a major barrier for DGEs introduce a whole new object—draggable
students in DGEs (Laborde, 1992)—students may drawings, which researchers often view as represen-
not be able to use these tools to keep track of what tations of shape conceptualizations: “A [draggable]
they do, which considerably lessens conceptual ex- figure captures the relationships between the objects
plicitness. Similarly, with Logo, although each move- in such a way that the figure [i.e., considered as a for-
ment of the turtle requires a command, what the mal concept] is invariant when . . . the construction
commands mean to students should not be taken for is dragged (in other words, that it passes the drag
granted. As an example, a student might correctly test)” (Jones, 2000, p. 58).
construct a rectangle, for instance, using trial and er- Furthermore, because draggable drawings are
ror and several turn commands to make each right constructed to possess specific geometric properties
angle—so there are no explicit commands to make (and consequently possess additional properties ne-
right angles. Without appropriate instructional in- cessitated by the given properties), researchers often
tervention, the student might not construe the set of see all the properties as “included” in the drawings.
RIGHT commands at a vertex as a set for which the However, students may not see any of the properties
sum of the inputs determines the size of vertex an- embedded in draggable drawings and may not even
gle; instead, the commands are taken as separate acts think that the drawings have specific properties. Nov-
used to visually adjust the turtle’s heading. Or, the ice students usually reason about draggable drawings
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 867

as geometric entities in their own right, not explicitly and analysis of ideas in order to form more general
as representations. In analyzing draggable drawings, ideas” (p. 82). She reported that use of the Supposer
students first notice movement constraints, then later, focused many students’ attention on the nature of the
they might conceptualize these constraints in terms set of examples—what information was needed, the
of regularities or invariants, and finally, and often variability of the information, and the need to orga-
only with great effort, these constraints are concep- nize information gathering. Students learned to ap-
tualized as formal geometric properties. The ques- preciate the value of generating many examples and
tion of how students interpret DGE drawings, both of formulating a data-collection strategy that would
draggable and not, is not only critical, but in need of prevent them from getting lost. Students learned “to
much research attention. use extreme cases, negative examples, and non-stereo-
typic evidence to back up their conjectures” (p. 82).
However, Yerushalmy found that some Supposer stu-
Geometric Supposers dents did very little genuinely independent work, gen-
Geometric Supposer construction programs allow erating many examples, but with little sense of direc-
students to choose a primitive shape, such as a type of tion—often asking the teacher what to look for in the
triangle or quadrilateral, and perform measurement examples. Students also struggled with knowing the
operations and geometric constructions on it. The types of examples they should examine (Yerushalmy,
programs record the sequence of constructions and 1993), which suggests that they had naïve ideas about
can automatically perform it again on other shapes. general mathematical statements and their validity.
The focus of Supposer programs, therefore, is to facili- One can pose many questions about the student
tate making, testing, and justifying conjectures, based learning reported by Yerushalmy. For instance, how
on examples. Most of the research on Supposer pro- deliberate is students’ use of examples? What is the na-
grams has taken place in the context of guided-inqui- ture of the reasoning that guides students’ choices of
ry teaching (Wilson, 1993). examples? How much of the generation of examples
Research generally, but not always, shows that use is student initiated, and how much teacher generated?
of Supposer programs can enhance geometry learning If a lot of it is teacher generated, does the generation
(Clements & Battista, 1992; Yerushalmy, 1993; Yerush- become interiorized by students, and, if so, how? What
almy & Chazan, 1993). Supposer interacting students do students think as they generate and manipulate ex-
were better able to identify non-stereotypical examples amples? What are the elements of students’ analyses
of shapes, and their learning often extended beyond of examples—how do conceptualizations of relevant
standard geometry content, for example, to formulat- ideas or beliefs affect these analyses?
ing definitions, making conjectures and arguments, Chazan (1993b) investigated teachers’ observa-
posing and solving significant problems, and devising tions that some students using the Supposer did not ap-
original proofs. After Supposer experience, students preciate that measurement evidence could not prove
more frequently used diagrams in their thinking, a proposition true for all cases within an infinite do-
treated diagrams flexibly, were less often confused main. Despite the fact that the instructional unit used
by diagrams in nonstandard orientation, used many in the study argued that measurement of examples is
diagrams to describe classes of figures, often rea- not a valid method for verifying the truth of geometri-
soned about a single diagram as a model for a class cal statements, instructional “activities in favor of de-
of shapes, and understood that Supposer constructions ductive proof were not successful with a sizable por-
can include characteristics that are not shared by all tion of the students in the study” (p. 109). Also, by the
members of the class of figures under consideration. end of the Supposer instructional treatment, although
Supposer activities can also engender movement away “fewer students considered [empirical] evidence to be
from considering measurement evidence as proof proof . . . more students were skeptical about the . . . ap-
(Clements & Battista, 1992). Indeed, Supposer work plicability of deductive proofs” (p. 109). However, Cha-
can (but does not have to—see below) help students zan argued that the latter finding may not be as nega-
appreciate the need for proof; unlike for textbook tive as it seems. For instance, such skepticism may be
theorems, many students consider Supposer-generated warranted (a) because students realized that they were
conjectures as statements that need to be proved be- novices at constructing valid deductive proofs, or (b)
fore they can be accepted as true. because, following Lakatos, being skeptical about proof
Yerushalmy (1993) claimed that three major is actually a mathematically healthy behavior. But there
processes are involved in geometric generalization: is an even deeper issue here. Almost no geometry cur-
“formation of samples of examples to serve as a data ricula actually support students’ invention of the de-
base for conjectures, manipulations of the samples, ductive method (see Borrow, 2000 for an exception).
868 ■ STUDENTS AND LEARNING

That is, even students who are given the opportunity Not all research on Logo has been positive, how-
to see the shortcomings of the empirical approach are ever. Some studies show no significant differences
generally shown a “better way” rather than allowed to in- between Logo and control groups or limited transfer
vent for themselves essential elements of the deductive of learning (Clements & Battista, 1992, 2001; McCoy,
method. Because students are merely handed the de- 1996). For example, ninth-grade Logo students did
ductive method, not given the opportunity to construct not differ significantly from control students in high
a personal need and meaning for it, they are naturally school geometry (Olive, 1991).
skeptical of it. Nevertheless, the issue of whether Sup-
poser use helps or hinders students’ movement away Logo as Representational Medium
from justifying conjectures with empirical evidence to- If one of the major reasons for the instruction-
ward the use of deductive proof remains unresolved. al efficacy of Logo lies in its representational power
(Clements & Battista, 1992), it is important to analyze
how this representation facility works for students. Is
Logo
the representational efficacy of Logo due to its capa-
Instructional Effectiveness bility to support reflection, to its encouragement of
In general, research supports the use of Logo measurement-based analysis and conceptualization,
in geometry instruction (Clements & Battista, 1992, or to the fact that multiple representations enrich
2001; McCoy, 1996). Students can understand and be- conceptualizations by enlarging the web of mean-
come competent with representing and reflecting on ings into which an idea fits? Critical in answering this
geometry problems using appropriate Logo environ- question is how students command the turtle to make
ments and instructional activities. For instance, in the shapes, and, more importantly, how they conceptu-
Logo Geometry (LG) Project, 28 teachers (980 K–6 stu- alize these commands. The following episodes illus-
dents) used LG, which provided instructional activi- trate students using Logo commands to support (a)
ties and Logo tools to help students progress through analytic, measure-based reasoning (Episode 1), (b)
the early van Hiele levels (Battista & Clements, 1991; visual/spatial and trial-and-error reasoning (Episode
Clements & Battista, 2001). Students in LG scored 2), and (c) the transition from visual-spatial to ana-
significantly higher than control students on a paper- lytic reasoning (Episode 3). Some researchers claim
and-pencil test that was not directly connected to LG, that one reason that students sometimes do not learn
making about double the gains of control students. optimally in Logo environments is that students often
Overall, LG helped students progress from van Hiele reason strictly visually despite the fact that Logo allows
Level 1, visual thinking, toward Level 2, descriptive/ more analytic and measure-based reasoning (e.g., Mc-
analytic thinking. Observations of LG students also Coy, 1996).
showed that they moved from (a) non-analytical and/ Episode 1. Two fifth graders were drawing a tilted
or authoritarian-based notions of knowledge, to (b) square using equal forward commands and 90° turns.
more autonomous, empirical bases for establishing When asked if their figure was a square, the students
truth, and finally to (c) knowledge as reasoning about replied, “Yes, a sideways square. . . . It has equal edges
mathematics as a logical system. and equal turns.”
Several general comments should be made about Episode 2. Second grader M was examining her
this study. First, students spent 4–6 weeks doing LG. attempt to draw a tilted square in a version of Logo
This extensive period of time counterbalanced the that allowed students to “undo” commands (Clements
“investment-time difficulty” required for learning & Battista, 2001). She used a trial-and-error approach,
geometry with Logo. That is, to learn geometry with undoing (signified by E for erase) or inserting addi-
Logo, students must learn the Logo language, which tional commands when part of the figure looked in-
can subtract from the time used to learn geometry correct. Her commands were: L 35 F 30 R 45 F 30 E E
per se. However, when such a large amount of time
is spent in learning geometry, as with LG, the amount
of time learning the Logo language is proportion-
ally smaller compared to the amount of time spent
learning geometry. Second, LG not only provided a
carefully developed sequence of activities, it incorpo-
rated into the Logo language numerous supports that
lessened the “learning Logo” load on students. Third,
participating teachers were given extensive training Figure 19.5 M’s attempts at making (a) a tilted square
on the use of the LG curriculum. and (b) a tilted rectangle in Logo.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 869

R 45 R 35 R 35 E F 30 L 35 R 90 R 35 F 30 L 35 R 90 understand the nature of the components of a square

R 35 F 30 F 5 F 5. Because M did not use all 90° turns, or how those components were spatially related. There
her figure was only a crude approximation of a square was no evidence that M consolidated the commands
(see Figure 19.5a). But she concluded that it was a into groups that represented the components of the
square, reasoning, “It’s in a tilt. But it’s a square be- shape—she merely moved the turtle bit-by-bit, seem-
cause if you turned it this way it would be a square.” ingly guided only by visual processes. In fact, even
Episode 3. M was then asked to make a tilted though M had previously stated that all the sides in
rectangle (previously unreported raw data, Battista a square are equal, she explicitly violated this notion
& Clements, 1987). She entered the following com- with her last set of forward commands. M used the
mands: same type of visual reasoning in the first part of Epi-
sode 3. However, during the second part of Episode
M: R 45 F 50 R 45 R 10 R 10 R 10 F 20 L 90 Whoa! 3, a change occurred. The interviewer’s questioning
R 180 F 50 Whoa! It’s supposed to go like this seemed to promote reflection that changed M’s con-
and it goes like this. Because I turned the tur- ceptualization of how to make a rectangle—she ex-
tle too far. E E L 25 L 160 F 50 E L 20 R 40 plicitly used 90° turns. What caused this change? Did
F 50 Phew! R 90 F 20 END [getting the path M explicitly reflect on the movement commands she
shown in Figure 19.5b]. used in making her first rectangle and draw some spe-
Int: Is there anything special about the turns in a cific conclusions from them—that is, was her realiza-
rectangle? tion based on unverbalized, even unconceptualized,
M: They’re all 90. abstractions from those experiences? This is possible,
Int: But you didn’t always use 90s. but, given the number of commands, it seems unlikely
M: Because I went like this and like this (motions) that she conducted a command-by-command analysis.
and then I didn’t turn and I wanted to go like Alternately, did M change her conceptualization by
this. And I hadn’t turned 90 and I had to make interrelating her current experiences in giving Logo
a couple of guesses and I finally got it right. commands for a tilted rectangle to the knowledge she
Int: Could you teach it to make another tilted exhibited when she stated that all the turns in a rect-
rectangle? angle are 90? And, what was the basis for her statement
M: R 45 F 60 R 90 F 10 L 90 R 180 F 60 R 90 F 10 that rectangles have 90° turns? Was it her experience
END in making rectangles that had vertical and horizontal
sides—most students in the class had used 90 for these
In Episode 2 and the first part of Episode 3, M turns? Or, was it the recollection of statements in class,
seemed to take a strictly visual, trial-and-error ap- by students and the teacher, that rectangles have 90°
proach. Her thinking was far different than the fifth turns? What caused and enabled M to connect the
graders in Episode 1, who used specific Logo com- concept of 90° turn to a specific type of spatial con-
mands to ensure their figure had specific properties. figuration—a right angle? Whatever it was, it resulted
But at the end of Episode 3, M seemed to come to a in an interiorized abstraction of this relationship, an
realization that the turtle needed to make 90° turns abstraction that allowed the relationship to be applied
to make a rectangle. M’s work brings up a critical is- in this new situation of giving commands to make the
sue in the use of Logo, How exactly does Logo move final rectangle.
students to higher levels of reasoning? When investigating processes that enable discov-
One possibility is that Logo, by virtue of requiring eries like those of M, it is important to understand
a more explicit, analytic, and symbolic representation how students use and understand sequences of Logo
of geometric ideas, supports and encourages reflec- commands. For instance, M twice used the sequence
tion and reformulation of those ideas. Clements and L 90 R 180 to accomplish a R 90 turn (because her
Battista (2001) have argued that programming the initial left turn was the wrong way). Did M explicitly
Logo turtle to make shapes encourages and facilitates understand the equivalence between the L 90 R 180
students’ reflection on the composition of the shapes. sequence and the R 90 command? That is, did she see
Students analyze the spatial components of shapes this sequence as a substitute for R 90, or was the sec-
and explicitly represent how they can be built from ond command in the sequence merely a visual adjust-
these components, including explicitly conceptual- ment, with the sequence not being conceptualized for
izing how the components are spatially related. This its net result? On the one hand, because this substi-
leads to recognition of the shape’s properties. tution was a common occurrence among students, it
In Episode 2, however, there was no evidence that is tempting to think that they saw the equivalence in
giving Logo commands enabled or encouraged M to some way. On the other hand, some students give a
870 ■ STUDENTS AND LEARNING

sequence of commands such as F 50 F 10 F 10 merely make accessible, and appropriately support measure-
as a way to move the turtle, seemingly unable to inte- based reasoning for this student.
grate the sequence into a length composite equivalent A final episode further suggests how students’
to F 70 (Clements, Battista, Sarama, Swaminathan, & work with Logo can move them toward property-based
McMillen, 1997). Other students, although able to reasoning about shapes. Prior to the episode, JN had
combine commands when directed to do so, often successfully used a variable-input rectangle procedure
have difficulty combining commands in more com- to draw a tilted rectangle. He was now reflecting on his
plex situations. For instance, late in an instructional unsuccessful attempt to use this procedure to make a
sequence that explicitly attended to encouraging stu- nonrectangular parallelogram.
dents to combine commands (Clements et al., 1997),
a teacher asked if the following commands produced JN: Maybe if you used different inputs. [He types
a rectangle:22 F 40 R 90 F 65 R 90 F 20 F 20 R 90 F 50 in a new initial turn, then stares at the paral-
F 15 R 90. Monica and Nina agreed that it would be lelogram on the activity sheet.] No, you can’t.
a rectangle until they attempted to draw the resulting Because the lines are slanted. . . .
path on paper—Nina drew a rectangle; Monica did Tchr: Yes, but this one’s slanted [indicating the tilt-
not. Even when Nina explained to Monica that 20 and ed rectangle that JN had successfully drawn
20 sum to 40, which matches the other 40, Monica with the Logo procedure].
needed to try the path on the computer before she JN: This one [the parallelogram]—the thing’s
accepted it. She was not convinced by strictly abstract slanted. This thing [the rectangle] ain’t
reasoning. slanted. It looks slanted, but if you put it back
The next episode shows how the Logo medium [shows how to turn it so that its sides are ver-
can help students develop analytic, measurement- tical and horizontal], it wouldn’t be slanted.
based conceptualizations of shapes. Anyway you move this [the parallelogram], it
wouldn’t be a rectangle. So, there’s no way.
RE: The rectangle is like a square, except that
squares aren’t long. But on rectangles, they JN was beginning to develop a conceptualization
are long. of what is formally captured by the property-based
Int: What does a shape need to be a rectangle? concept that the adjacent sides of a rectangle are per-
RE: All of the sides aren’t equal. These two [op- pendicular. In so doing, his reasoning was progressing
posite sides] and these two [other opposite] from the visual reasoning of van Hiele Level 1 to the
sides have to be equal. descriptive-analytic reasoning of Level 2. However, at
this time, JN did not possess an interiorized formal
Int: How about 10 on two sides and 9 on the other
concept of perpendicularity (or right angles) that he
two? Would that make a square?
could apply to the situation.
RE: Kind of like a rectangle.
Int: Would it be a square too? Zooming in on Conceptual Change
RE: [Shaking head negatively] It’s not a square. In my final example of Logo research, I summa-
’Cause if you make a square, you wouldn’t go rize a study of Hoyles and Healy (1997), who inves-
10 up, then you turn and it would be 9 this way, tigated how one 12-year-old student, Emily, changed
and turn and 10 this way. That’s not a square. her conceptualizations of geometric reflections while
working in a Logo-based microworld. On a pretest
RE’s first comment indicated visual reasoning— asking students to draw reflections of figures, two of
rectangles are long. On further questioning, RE talk- Emily’s answers were correct, three were incorrect.
ed about opposite, but not all, sides being equal—a However, in the three incorrect drawings, the reflect-
property-based response. But we cannot attribute this ed images were congruent to their preimages, which
increase in level of reasoning directly to the Logo con- is consistent with Emily’s description of reflections,
text—even students not using Logo make such chang- “the line . . . has got something on one side and it
es in reasoning in response to appropriate question- would be exactly the same on the other.” For instance,
ing. Finally, however, RE reasoned analytically about when drawing the reflection of a vertical line segment
side lengths using the action-based command context through an oblique mirror line, after Emily correctly
of Logo. The Logo environment seemed to promote, drew one endpoint of the image, she drew the rest of

22
For consistency, forward commands are always denoted by F and right commands by R. The actual commands given by students varied
by microworld.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 871

the image parallel to the preimage. Hoyles and Healy “in the middle” of the two congruent figures and that
reported that when Emily did reflection problems, she somehow the turtles drawing the figures made equal
used a ruler placed orthogonally to the mirror line to angles with the line when they met it.
measure equal distances between some, but not all, The next task required students to construct the
corresponding points. Thus, Emily’s initial concep- reflection of a flag when both the flag and mirror line
tualization of reflection seemed to have two compo- were oblique and no Logo commands were provided.
nents: First, a reflection consists of a mirror line along (Colored turtles were used to mark flag vertices so stu-
with congruent images and preimages; second, some dents could find distances between vertices by asking
points on the preimages correspond to points on the Logo to give distances between specified turtles.) In
image via perpendicularity and equidistance (but this this task, the girls changed their strategy, trying to set
idea was vague and inconsistent). up a correspondence between selected preimage and
In the first computer task, a blue turtle was sup- image points. As shown in Figure 19.6a, Emily project-
posed to draw the preimage and a red turtle the re- ed the endpoints of the flag vertically downward, as if
flection image about a vertical mirror line. The two the mirror line were horizontal, making each project-
turtles were shown with the same position and head- ed point and its preimage equidistant from the mir-
ing along the mirror. The commands (but no figure) ror line, in the vertical direction. However, we do not
for the blue turtle were shown; the problem for the know how the girls arrived at this conceptualization. Is
students was to give the appropriate commands for it another manifestation of the vague correspondence
the red turtle. The initial idea of Emily and her part- idea that Emily used on the pretest?
ner Cheryl was to reverse both movement (F, B) and
turn commands (R, L), but they quickly decided, with- Brown Brown
out trying their idea, that they only needed to inter-
Black Black
change right and left turn commands, which they did, Pink Pink
completing the task correctly. Green Green
Orange Orange
Because the girls’ notions of reflections seemed to 64
29 88 60
29°
be tightly connected to Logo commands, Hoyles and 29
45° 29°
Healy next investigated what happened when Logo Red 45° 60
commands for figures were absent. Students were 64 Blue
given two reflection-congruent turtle paths (with the 88

turtles showing), but no Logo commands, and were Blue

Red
asked to find the mirror line. Cheryl simultaneously
traced the path of both turtles, stopping where she (a) (b)
thought they would meet on the mirror line (so she
Figure 19.6 Emily and Cheryl’s reflection attempts in
was using a mental model of two linked and simulta-
Logo (Hoyles & Healy, 1997).
neously moving turtles). After using an experimenter-
provided tool for getting the turtles to meet on the
When the girls discovered that the distance be-
mirror line, the girls turned the blue turtle through
tween the red and blue turtles was unequal to that
a series of L 10s until it looked as if it were heading
between the green and pink turtles, they ignored the
along the mirror line. They then turned the red turtle
blue turtle and gave commands to the red turtle. The
the same angle to the right, and when they saw that it
resulting image flag was congruent to the preimage
was facing exactly the same direction as the blue, they
flag, but when the girls recognized that the image
were convinced that they had found the direction of
flag’s orientation was incorrect, they devised a new
the mirror line.
strategy for finding the endpoints of the flagpole.
Several questions about these first two tasks arise. They used Logo commands to determine the angle
Precisely how was the girls’ left/right reversal strat- and distance of a turtle on the mirror line to the pink
egy connected to or imbedded in their conceptions turtle (the preimage turtle); they then turned the tur-
of reflections? For instance, had the girls previously tle the opposite direction from the mirror and went
abstracted that reflection reverses orientation? Or, forward the given distance to the image red turtle (see
was turn reversal an element of Logo procedural Figure 19.6b). The authors observed,
knowledge that enabled the girls to produce congru-
ent shapes on both sides of a mirror line? How did Across all the [student] pairs and the many different
they discover this knowledge? How exactly did the methods observed, our analysis showed how the mi-
girls conceptualize the mirror line? From the second croworld offered students a way of talking about and
task, we might conjecture that they thought that it was operationalising the angle properties of reflection, by
872 ■ STUDENTS AND LEARNING

thinking about turtles turning from a heading along things happened. First, placing the turtle enabled her
the mirror through the same amount in opposite di- to activate her previous turtle-based scheme for deal-
rections. (Hoyles & Healy, 1997, p. 44) ing with the situation. She was, through the process of
interiorization, able to disembed her abstraction from
Finally, to investigate how situated the girls’ its original Logo context and apply it in this new, pa-
knowledge was in the Logo context, Emily was asked per-and-pencil situation. This interiorized notion, to-
to reflect a triangle in an oblique mirror using only gether with her reflection on the arbitrary placement
a ruler and a protractor. She was completely puzzled of the turtle, seemed to enable her to develop an ex-
until she drew a turtle on the reflection line to use as plicit notion of the placement arbitrariness—creat-
a reference point (see Figure 19.7a). She measured ing a more general and abstract scheme (and her aha
the angle and distance of vertex A to the mirror line,
moment). However, her new, more general conceptu-
then copied this angle and distance to the other side
alization still seemed tied to the Logo context—she
of the line (locating point A’). As she considered the
drew a turtle for C. And the effect of the discovery
placement of the image of triangle vertex C, as shown
on Emily’s orthogonal-correspondence conception
in Figure 19.7a, her interiorization of what she did for
of symmetry is unknown. Was it abandoned, or had it
A enabled her to progress even further, to a general-
been transformed?
izing abstraction.
Although the data presented by the authors pro-
vide an extremely important view of how Emily’s con-
C′ ceptualizations evolved, several questions remain un-
C C answered. For instance, in the initial episodes, both
Emily and Cheryl seemed to possess a complex inter-
connected web of knowledge consisting of holistic vi-
sual knowledge of reflections as congruent halves on
either side of the mirror line, a vague notion of or-
thogonal corresponding parts in preimage and image,
B A′ B A′ and developing knowledge of producing reflection
A A paths via Logo commands. By the final episode, and
as a result of carefully chosen instructional activities,
(a) (b) Emily’s conceptualization of reflections seemed to
change from a turtle-centric path perspective focusing
Figure 19.7 Emily’s reflection using ruler and protractor on Logo movement commands to an exocentric view
and the turtle (Hoyles & Healy, 1997, p. 52).
in which the mirror line served as a reference point
for measuring angles and distances. This change may
have been mediated by her initial vague notion of find-
Emily: That’s one [A], now I suppose I have to do the
same (indicates C). Can I (she places protractor on
ing corresponding parts in reflections. Or, it may have
turtle) . . . does the turtle have to be there? It’s hard been encouraged by the constant demand to translate
to measure to there so, could I . . . hold on . . . I think between the turtle perspective and an exocentric view
I’ve got it. It doesn’t matter does it, you can be any- of reflections (looking “down” on and comparing fig-
where as long as the angle and the distance is the ures). Somehow Emily’s initial vague notion of corre-
same, it doesn’t matter where you actually measure sponding parts became connected to and clarified by
from . . . so say there (draws another turtle), the angle the concept of “same angle/same distance.”
is . . . (measures angle and distance to C) so that like
(constructs angle and distance to position C′) [see Emily significantly increased the level of sophis-
Figure 19.7b]. (Hoyles & Healy, 1997, p. 52) tication in her thinking about reflections. Her rea-
soning progressed from visual to descriptive-analytic;
Hoyles and Healy noted that “This last vignette il- from thinking vaguely and visually, to a synthesis of
lustrates the extent to which, for Emily, the tools used visual, analytic, and measurement-based reasoning,
in her previous explorations had become part-and- much of it couched in the language of Logo. The
parcel of her sense making activities” (1997, p. 53). At Logo environment, including the instructional activi-
first, this non-Logo task presented a major obstacle for ties, seemed to provide Emily with tools for making
Emily because it removed the Logo context in which her conceptions of reflections more analytic, precise,
much of her relevant conceptual knowledge resided. and general. Consistent with what Clements and Bat-
Emily could not think of the problem without turtles tista (2001) found in their Logo studies, this micro-
for reference points. But when she drew a turtle, two world helped Emily and her partner to become more
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 873

cognizant of and to reshape their thinking in terms of the one hand, it has been claimed that DGEs “provide
precise, measurement-based properties. a revolutionary means for developing geometrical un-
However, as is often asked of geometry learning derstanding. . . . This software seems to make the ex-
in Logo, we might ask how Emily’s final conceptual- ploration of geometrical configurations and the iden-
ization of reflections compares to traditional math- tification of meaningful conjectures more accessible
ematical descriptions? That is, Emily constructed a to pupils” (Mariotti, 2001, p. 257). On the other hand,
nontraditional conceptualization of reflections—that skeptics worry that DGEs weaken the role of proof in
preimage and image points have equal angles and dis- high school geometry (Mariotti, 2001). The following
tances to the mirror line. Although Emily’s conceptu- discussion deals with all aspects of students’ learning
alization is mathematically correct, it is not as concep- in DGEs. As in the section on Logo, the discussion will
tually illuminating as the traditional concept because begin with research on students’ work at the elemen-
the traditional concept can more easily be related tary school level, then move to the middle school and
to the intuitively appealing conception of reflection secondary level.
produced by motion. Indeed, thinking of image and
preimage points as endpoints of segments bisected by, DGE at the Elementary Level
and perpendicular to, the mirror line, can be directly Because DGE software such as Cabri and GSP
related to a (nonrigid) motion, with these segments were originally designed for the secondary level, use
being the paths of preimage to image points. It would of this software in elementary classrooms requires ac-
have been interesting to see if and how Emily’s final commodations. For instance, the Shape Makers com-
conceptualization could have been evolved into the puter microworld is a special add-on to GSP that pro-
traditional concept. (For a discussion of students’ de- vides students with geometric shape-making objects
velopment of comparable reflection concepts in dy- (Battista, 1998a). An example is the Parallelogram
namic geometry, see Hollebrands, 2003.) Maker that can be used to make any desired paral-
Finally, let us return to the question, What char- lelogram that fits on the computer screen, no matter
acteristics of the Logo environment are essential for what its shape, size, or orientation—but only paral-
encouraging and supporting students’ learning? As lelograms. The appearance of a Parallelogram Maker
illustrated in the other studies discussed in this chap- is changed by dragging its vertices with the mouse. In
ter, students in the Hoyles and Healy study seemed the Shape Makers environment, when measurement is
motivated to work on problems in the Logo context. introduced, it is done so by displaying relevant mea-
But beyond that, was it the Logo-command language sures on the screen automatically; so students do not
per se that supported students’ learning, or was it one have to use GSP measurement tools directly. The ma-
of several other characteristics of Logo that provided jor focus of Shape Makers units is for students to move
such support? For instance, Logo provides quick, ac- from van Hiele Levels 0 and 1 to Levels 2 and 3. To il-
curate, and interesting visual depictions of students’ vi- lustrate the type of geometric reasoning and learning
sual-geometric conceptualizations—a type of uninter- exhibited by elementary students in DGEs, I describe
preted feedback. It provides and encourages, through several episodes with fifth graders’ working with the
its use of measurement specifications in movement Shape Makers microworld (Battista, 1998a, 2001b, in
and turn commands, an accessible way to analyze press b) in their regular classrooms.
shapes in terms of their components’ measures—an Episode 1. Three students were investigating the
important step in moving from pure visual to analytic Square Maker at the beginning of their Shape Maker
thought. Thus, it might be conjectured that it is not work: MT, “I think maybe you could have made a rect-
Logo code per se that supports the development of angle.” JD, “No; because when you change one side,
students’ geometric understanding but its provision they all change.” ER, “All the sides are equal.”
of (a) appropriate motivation; (b) accurate, quick, MT, JD, and ER abstracted different things from
and uninterpreted visual feedback; and (c) easy-to- their Square Maker manipulations. MT noticed the
use and precise measurement specifications. These visual similarity between squares and rectangles, caus-
same components also appear in dynamic geometry ing him to conjecture that the Square Maker could
environments. make a rectangle. JD abstracted a holistic movement
regularity—when one side changes length, all sides
Dynamic Geometry Environments change (thus, he could not get the sides to be dif-
ferent lengths, which he thought was necessary for a
Currently, DGEs seem to be one of the most pop- rectangle). Only ER conceptualized the movement
ular types of software used by mathematics teachers regularity with complete precision by expressing it in
(Becker, 2000) and investigated by researchers. On terms of a formal geometric property.
874 ■ STUDENTS AND LEARNING

From the van Hiele perspective, one might say moves. It will never get exactly the right size.
that MT’s reasoning on this task was at Level 0, JD’s [Manipulating the Rhombus Maker] Let’s see
at Level 1, and ER’s at Level 2. From the structuring if I can make the square with this. Here’s a
perspective, MT and JD had constructed spatial struc- square. I guess it could maybe be a square.
turings for the Square Maker, with JD’s more sophis- But I’m not sure if this is exactly a square. It’s
ticated than MT’s. ER, in contrast, had constructed a sort of leaning. The lines are a little diagonal.
property-based geometric structuring for the Square [Continuing to manipulate the Rhombus
Maker by applying a previously interiorized concept Maker] Yeah, I think this is a square may-
of equal sides to a spatial structuring similar to that be. . . .
described by JD. Finally, the abstractions made by MT Res: You said the Rhombus Maker could make the
and JD seemed completely situated in the DGE micro- same shape as Shape C, what do you mean by
world, whereas that made by ER was not. that?
Although the above episode illustrates that dif- NL: It could make the same shape. It could make
ferent levels of reasoning that students can achieve this shape, the one with 2 diagonal sides and
on a task in the Shape Maker environment, it does 2 straight sides that are parallel. It could have
not show how a student progresses from one level to been almost that shape and it got so close I
the next. The following episodes provide insight into thought it was that shape.
this process. NL: [Continuing to manipulate the Rhombus
Episode 2. NL was using the seven quadrilateral Maker] Oh, I see why it didn’t work, because
Shape Makers to make the design shown in Figure 19.8 the 4 sides are even and this [Shape C] is
while a researcher was observing and asking questions. more of a rectangle.
Res: How did you just come to that?
kite
maker
parallelogram
maker
NL: All you can do is just move it from side to side
rectangle
and up. But you can’t get it to make a rect-
square
maker
maker trapezoid rhombus
maker
quadrilateral angle. When you move it this way it is a square
maker maker
and you can’t move it up to make a rectangle.
A And when you move this, it just gets a bigger
B C
square.
Res: So what made you just notice that?
D E
NL: Well I was just thinking about it. If it [the
Rhombus Maker] was the same shape, then
F
there is no reason it couldn’t fit into C. But
I saw when I was playing with it to see how
you could move it and things like that, that
G whenever I made it bigger or smaller, it was
always like a square, but sometimes it would
Figure 19.8 Make the design consisting of Shapes A–G be leaning up, but the sides are always equal.
with the 7 Shape Makers (Battista, 1998a).
This episode clearly shows how a student’s ma-
nipulation of a Shape Maker and reflection on that
NL: The Rhombus Maker on [Shape] B. It doesn’t manipulation can enable the student to move from
work. I think I might have to change the thinking holistically to thinking about interrelation-
Rhombus Maker to [Shape] C. ships between a shape’s parts, that is, about its geomet-
Res: Why C? ric properties. Indeed, NL began the episode think-
NL: The Rhombus Maker is like leaning to the ing about the Rhombus Maker and shapes holistically
right. On B, the shape is leaning to the left. and vaguely, saying that she was trying to make the
I couldn’t get the Rhombus Maker to lean to Rhombus Maker “lean to the right” and get “bigger or
the left, and C leans to the right so I’m going smaller,” and that “the whole thing moves.” The fact
to try it. [After her initial attempts to get the that NL could not make the nonequilateral parallelo-
Rhombus Maker to fit exactly on Shape C] I gram with the Rhombus Maker evoked a perturbation
don’t think that is going to work. that caused her to reevaluate the spatial structuring
Res: Why are you thinking that? contained in her mental model of the Rhombus Mak-
NL: When I try to fit it on the shape and I try to er. Originally, because her model did not include the
make it bigger or smaller, the whole thing constraint “all sides equal,” her mental simulations of
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 875

changing the shape of the Rhombus Maker included to achieve—understanding why the Rhombus Maker
transforming it into nonequilateral parallelograms. would not make Shapes B or C. (This was a personal
Her subsequent attempts to make a nonequilateral goal for NL not only because the problem interested
parallelogram with the actual Rhombus Maker tested her but because she was a participant in an inquiry-
her model, showing her that it was not viable. As she based classroom culture and curriculum.)
continued to analyze why the Rhombus Maker would Finally, this episode illustrates how geometry learn-
not make the parallelogram—why it would not elon- ing with the Shape Makers is much richer and more
gate—her attention shifted to its side lengths. This powerful than that which occurs in many traditional
new focus of attention enabled her to abstract the curricula. First, because NL’s learning of properties
regularity that all the Rhombus Makers’ sides were was purposeful—connected at the outset with attain-
the same length. As she incorporated this abstraction ing a personal goal—it is highly likely that henceforth
into her mental model for the Rhombus Maker, she she would see that knowledge as being applicable and
was able to infer that the Rhombus Maker could not useful. (In fact, NL used this and other properties
make Shape C. with increasing regularity in her subsequent work.)
This activity encouraged and enabled NL to prog- Second, because NL constructed her property-based
ress from a spatial structuring that was incomplete, reasoning from her already-existing cognitive struc-
imprecise, and did not explain the Rhombus Maker’s tures, the newly constructed knowledge was “well-con-
movement (van Hiele Levels 0 and 1) to a property- nected” in the sense that it was firmly anchored in her
based geometric structuring that did explain that knowledge web, making it more likely not only to be
movement (van Hiele Level 2). A combination of two applied in problem solving but used in further acts of
factors may have supported NL’s progress. First, dur- knowledge construction. Third, and finally, it is highly
ing the episode, NL abstracted the “all sides equal” likely that this episode increased NL’s overall appre-
property to a sufficiently high (interiorized) level that ciation for the power of formal geometric reasoning
enabled her to apply it in various situations. Previous in understanding the environment, again making it
episodes indicated that NL had already concluded more likely that, in the future, NL would seek to use
that squares have all sides the same length; in Episode geometry in understanding the world.
2 she raised this notion to a higher level of abstraction. Episode 3. The task was to determine which of
Second, it is highly likely that NL’s conclusions about Shapes 1–7 could be made by the Rectangle Maker
the Rhombus Maker came about partly because she (explaining and justifying each conclusion; see Fig-
had previously made a square with it; she viewed the ure 19.9.) Fifth graders M and T predicted that the
Rhombus Maker as a transformed square. I conjecture Rectangle Maker could make shapes 1–3, but not 4–6.
that NL’s transformation of the Rhombus Maker into They are now checking and discussing their results.
a square, a shape that she conceptualized as having (The entire episode took about 25 minutes.)
all sides equal in length, made a critical connection
that eventually enabled her to transfer the concept of A B
Length(AB) = 123 pixels Angle(A) = 90°
equal side lengths to the new shape of Rhombus Mak- Length(BC) = 67 pixels Angle(B) = 90°
er (via reaching a higher level of abstraction). In a Length(CD) = 123 pixels Angle(C) = 90°
Length(AD) = 67 pixels Angle(D) = 90°
subsequent section, I will discuss how the transforma- D C
tion facility of the Shape Makers is a major component
of their instructional efficacy. 2
1
Episode 2 also illustrates how students using the 3 4
Shape Makers move toward more sophisticated, proper- 6
ty-based conceptions of shapes because of the inherent 5
power these conceptions give to their analyses of spatial
phenomena. In the current situation, NL developed a Figure 19.9 Which of Shapes 1–7 can be made by the
property-based conception of the Rhombus Maker be- Rectangle Maker (Battista, 1998a)?
cause it enabled her to understand why the Rhombus
Maker could not make Shape C—something that truly After using the Rectangle Maker to check Shapes
puzzled her. Only when NL conceptualized the move- 1–3, M and T move on to Shape 4.
ment of the Rhombus Maker in terms of a geometric
structuring did she feel that she really understood its T: I’m positive it can’t do this one.
movement. NL acquired this property-based concep- M: It [the Rectangle Maker] has no slants. We
tion not because someone told her to learn it, but be- had enough experience with Number 3, that
cause it helped her achieve a goal that she was trying it can’t make a slant.
876 ■ STUDENTS AND LEARNING

T: Yes it can, it has a slant in the other one [Shape and abstracted that the Rectangle Maker always has
3]. It [the Rectangle Maker] has a slant right four 90° or right angles. Furthermore, he abstracted
now. . . . this property sufficiently so that he was able to use it
Tchr: What do you mean by slant? to analyze the differences in Shapes 3 and 4. He sub-
M: Like this. See how this is shaped like a paral- sequently conceptualized that the Rectangle Maker
lelogram [motioning along the perimeter of could not make Shapes 4, 5, and 6 because they do
Shape 4]. not have four right angles. In fact, by the end of the
T: This is in a slant right now [points at the class period, the boys saw that the spatial relationship
Rectangle Maker, which is rotated from the they were attending to could be described in terms of
horizontal]. [Note that slant means nonper- the formal mathematical concept of right angle. They
pendicular sides for M, but rotated from the had constructed a geometric structuring, a van Hiele
horizontal for T.] . . . Level 2 analysis, that enabled them to solve the prob-
M: It can’t make that kind of a shape [pointing lem they had embraced.
to Shape 4]. . . . It can’t make something that This episode illustrates how difficult it can be for
has a slant at the top and stuff. students to reformulate unrefined spatial structurings
T: Do you mean it has to have a straight line right of Shape Maker movements in terms of traditional
here [pointing to Shape 6], like coming across? geometric concepts.23 Only after much guidance, re-
[T now uses straight to mean horizontal.] flection, and experimentation can students construct
M: I know they are straight, but they are at a slant formal, property-based geometric conceptualizations
and it [the Rectangle Maker] always has lines of Shape Maker movement constraints. However,
that aren’t at a slant. . . . once students meaningfully move from their initial
Tchr: [After several failed attempts to unobtrusively vague spatial structurings to appropriate geometric
get students to reconceptualize the slant con- structurings, they embrace the latter because of their
ception in terms of angles] Keep on really inherent power.
looking at what makes these different. And
maybe some of the information on the screen Similar Results
will help you. Watch the numbers up there
It is instructive to compare Shape Makers results to
and see if that will help you. . . .
results from other environments. First, note that M
T: [Manipulating the Rectangle Maker after the
and T’s reasoning was similar to, but more advanced,
teacher leaves] Oh, this [the Rectangle Mak-
than that of JN in Logo. All three students were strug-
er] always has to be a 90° angle. And that one
gling with conceptualizing the notion of perpen-
[Shape 4] does not have 90° angles. And so
dicularity. In fact, Berthelot and Salin (1998) found
this one [Shape 3] has to have a 90° angle,
that only 50% of 10–11-year-olds used right angles
because we made this with that [Rectangle
to locate the legs of a rectangular bench if moved.
Maker]. So there is one thing different. A 90°
The other students explicitly used only information
angle is a right angle, and this [Shape 4] does
about length. Also, consistent with the Battista results,
not have any right angles.
Sutherland, Godwin, Olivero, and Peel (2002) found
that use of DGE constructions similar to the Shape
M and T were trying to find a way to conceptualize
Makers helped elementary students learn about prop-
and describe the spatial relationship that sophisticat-
erties of shapes.
ed users of geometry would describe as perpendicular.
In so doing, and typical of students at van Hiele Level
1, they developed several vague and incomplete spa- DGEs at the Middle and High Levels
tial structurings, all of which were inadequate. They In examining instructional DGE use by older stu-
used familiar terminology and concepts like slanted dents, besides having students investigate geometric
and straight that inadequately described the idea with shapes by manipulating them on the computer screen,
which they were grappling. Progress was finally made, two additional major emphases occur. First, students
after the teacher left the boys, when T manipulated are asked to use DGE tools to construct draggable fig-
the Rectangle Maker and focused on its measure- ures. Second, students are asked to prove conjectures
ments. Through this manipulation, he discovered that arise out of exploration.

23
The difficulty students have with this process points out just how deficient traditional cursory memorization-based approaches to this
topic are. It takes a great deal of thought and appropriate experience for students to use the formal geometric conception-based system
with genuine understanding.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 877

To promote understanding of quadrilaterals in lytic. They reformulated their idea of reflections when
junior high students in the UK, Jones (2000) gave a they saw that the mirror line was formed by joining
series of DGE tasks in which students were asked to midpoints of corresponding points. The “analytic”
construct draggable figures. For instance, students were part of this new conceptualization is the geomet-
asked to construct a draggable square and explain why ric description of the reflection concept in terms of
the shape is a square. As students progressed through constituent parts (corresponding points, segments,
the instructional sequence, Jones found that, initially, midpoints). Also, C and M, like Emily, did not arrive
students emphasized description rather than explana- at the traditional view of reflections—they made no
tion, relied on perception rather than mathematical mention of orthogonality, so their midpoints notion
reasoning, and lacked precise mathematical language. was incomplete. It would have been useful to see how
At an interim period in the instruction, students’ ex- further instruction could have extended the girls’ no-
planations became more mathematically precise but tion to include orthogonality. It would also have been
were intertwined with the operation of DGE software. enlightening to better understand how C and M made
At the end of the instructional unit, students’ explana- transitions through their successive conceptualiza-
tions were entirely mathematical. tions. For instance, we might conjecture that as C and
To illustrate, when students were asked to explain M focused on a pair of corresponding points while
why all squares are rectangles, early in instruction one manipulating the flags with the mirror line drawn, the
pair wrote: “You can make a rectangle into a square by two points formed a perceptual unit, bisected by the
dragging one side shorter . . . until the sides become mirror line, which may have activated an interiorized
equal” (Jones, 2000, p. 76). Later in instruction, stu- version of the concept of segment midpoint (either
dents were asked to construct a trapezium that could formal, or situated within Cabri).
be modified to make a parallelogram and explain why
all parallelograms are trapeziums. The same pair of Artifacts of DGEs versus Geometric Properties
students wrote, “Trapeziums have one set of paral- A fundamental open question is if and how stu-
lel lines and parallelograms have two sets of parallel dents “distinguish fundamental characteristics of
lines” (Jones, 2000, p. 76). The apparent argument in geometry from features that are the result of the par-
the first example is couched in terms of the DGE—“via ticular design of the DGE” (Jones, 2000, p. 59). For
dragging, you can see that a square is a special type of instance, similar to Holzl, Healy, Hoyles, and Noss
rectangle.” The second argument is apparently more (1994), Jones (2000) argued that students need to
traditional property-based—“parallelograms have all understand functional DGE dependencies, such as in
the properties needed for trapeziums.” What we do Cabri, although basic points, points on objects, and
not know, however, is how students made the transi- points of intersection look identical on the screen, ba-
tion from DGE-based to formal-geometric conceptu- sic points and points on objects are draggable, whereas
alizations. More research is needed that investigates intersection points are not. Apparently, students have
this transition. difficulty developing such understanding (Holzl et al.,
In another study, conducted by Noss and Hoyles 1994; Jones, 2000). In fact, students often develop non-
(1996), two female students, C and M (age 14), were mathematical conceptualizations of the draggability of
given two flags in Cabri—one the reflection of the figures. For example, asked why her constructed figure
other—and asked to find the mirror line. After a short could not be “messed up,” one student replied, “They
period of dragging points on the flags and noticing stay together because. . . . It just glued them together”
the effects on the images, C and M visually located the (Jones, 2000, p. 71). Thus, as Holzl et al., Jones, and
position of the mirror line. In an effort to be more others (Goldenberg & Cuoco, 1998) have argued, espe-
precise about this location, C had the idea of dragging cially in curricula that make significant use of student
corresponding points on the flags—preimage and re- exploration, research must investigate how students
flection image—so that they met. The girls claimed interpret the construction and behavior of DGE draw-
that such points lie on the mirror line. Subsequently, ings. In essence, students working in DGEs need to see
C and M used two pairs of these overlapping corre- beyond the operational syntax of the environment; the
sponding points to draw the mirror line. With the DGE representations must become “transparent.” But
mirror line drawn, the girls then focused on pairs of researchers do not yet know how this happens.
corresponding points, which led them to see that the
mirror line could be constructed by joining the mid- What Happens When Constructing a Draggable
points of segments between corresponding points. Dynamic Figure?
Similar to Emily in Logo, work with Cabri made Many instructional treatments involving DGEs in-
C and M’s conceptualization of reflections more ana- troduce “a specific criterion of validation for the solu-
878 ■ STUDENTS AND LEARNING

tion of a construction problem: a solution is valid if construction, they did not mention them. Second, if
and only if it is not possible to ‘mess it up’ by dragging” the students already knew the properties of perpen-
(Jones, 2000, p. 58). So let us more carefully examine dicularity, parallelism, and congruence of opposite
the process of constructing a draggable figure. For in- sides, what geometry did they learn by working on this
stance, two (Year 8) students were attempting to make problem in Cabri? Perhaps no new knowledge was ac-
a draggable rectangle in Cabri (Holzl et al., 1994). quired, but instead, the students’ knowledge and rea-
After drawing segment AB (see Figure 19.10), they soning were deepened and enriched. The process of
constructed a perpendicular to this segment, pass- applying known concepts in a new situation requires
ing through B, put point C on the perpendicular and interiorization of the concepts, making them more
constructed another perpendicular through C, put a powerful. Or perhaps connections between proper-
point D on that perpendicular and constructed an- ties were newly constructed or extended. Also, if the
other perpendicular. The students joined points A, B, students noticed parallelism and equality of opposite
C, and D and hid the other lines, forming a trapezoid, sides in the constructed figure, they might have seen
then dragged A so that the trapezoid looked like a that perpendicularity implies parallelism and equality
rectangle. The students realized that they had not cre- of sides (especially if a teacher helped them notice it),
ated a draggable rectangle, but got bogged down in moving them toward or reinforcing van Hiele Level 3
Cabri operations when trying to fix their error. For in- thinking. On the flip side of the argument, however,
stance, they attempted to change the nature of point one might ask what the students really understood
A to be a “point on object [DX].” Finally, when the about the perpendicularity of sides. For instance, the
students were told that this could not be done, they students could have constructed the rectangle visually,
completed their rectangle by creating an intersection using perpendicularity as a procedural tool for mak-
point on DX and AB. ing lines meet in a certain way. That is, they knew per-
pendicularity visually as making things “square,” not a
geometrically as involving right angles.
The claim by Holzl et al. (1994) that students
working in DGEs “ran up against some fundamental
C 2nd perpendicular D
aspects of its [Cabri] geometry” (p. 11) raises additional
issues. First, does Cabri really have its own geometry?
1st perpendicular 3rd perpendicular
If so, what is it and how is it related to traditional math-
B A ematical geometry? For instance, as has already been
segment AB
mentioned, unlike mathematical geometry, Cabri “ge-
ometry” has different kinds of points that behave dif-
X
ferently. As another example, the idea of “point on
object,” as implemented in a DGE, may promote stu-
Figure 19.10 Students attempting to make a draggable dent conceptualizations that are very different from
rectangle in Cabri. the corresponding mathematical concept. In Birkoff’s
formulation of geometry (Moise, 1963), the statement,
A number of questions must be asked about ep- “let X be a point on segment AB” means that X is any
isodes such as this, which are common in students’ of the points that are elements of the point-set that
DGE constructions. First, Holzl et al. claimed that the is segment AB. However, the way the point concept
students “were clearly aware of the properties of rect- is implemented in a DGE might lead some students
angles and the need to construct perpendicular lines to conceptualize a point on a segment as an object
to obtain right angles” (1994, p. 9). But exactly which “on top of” or separate from the segment, not as part
properties of rectangles did the students know? For in- of the segment. In this case, the DGE concept and
stance, was their construction a direct result of know- mathematical concept would be inconsistent. In con-
ing that the sides of a rectangle are perpendicular? trast, in Hilbert’s formulation of geometry (Wallace &
Or, did they know that rectangles have right angles, West, 1992), point, line, and on are undefined terms,
and connect the concepts of right angle and perpen- and there is no requirement that individual points
dicularity to make their construction (perhaps making be elements of sets of points called lines. In this case,
this connection for the first time)? Also, there was no there seems to be no inconsistency between the DGE
explicit evidence in their construction that they knew and mathematical concepts. This recalls the question,
that opposite sides were equal and parallel. Of course, What geometric conceptualizations do students de-
they may have known these properties, but because velop in DGEs, and how are these conceptualizations
perpendicularity is sufficient to complete the Cabri related to various formal mathematical concepts?
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 879

Moving Toward Justification However, several questions can be raised about

At the secondary level, many researchers see the these claims. First, does use of DGEs motivate and
primary purpose for DGE construction activities as a promote the learning of proof better than traditional
bridge to justification and proof: instructional contexts? Second, if students already be-
lieve that a proposition is true because of DGE empiri-
The [instructional] approach to the construction cal evidence, what precisely is added to students’ rea-
problem within the Cabri environment is expected soning and learning by formally proving it? Although
to introduce first, the need to justify a solution, then researchers can see much that might be added, what
the need to negotiate the adequacy of this justifica-
tion. The dragging test has a fundamental role in this
actually happens with students? This is an empirical
process of justification. The fact that students can see question. Third, exactly what are the processes by
for themselves when and which geometrical proper- which DGEs promote the development of proof?
ties remain constant under dragging can help them to A number of researchers have addressed the lat-
understand when they have produced something that ter question. For instance, Mariotti (2001) studied stu-
merits justification. . . . In the subsequent mathemati-
dents working on DGE instructional activities as part
cal discussions, it will become necessary to negotiate
and introduce explicit criteria for the acceptability of of the regular curriculum. In the first problem, ninth-
the justification itself. (Mariotti, 2001, p. 262) [Pre- grade students were asked to construct the bisector of
sumably “acceptable” means formal proof.] an angle then describe and justify their solutions. In
Alex and Gio’s first solution attempt (Figure 19.11a),
Consistent with this claim, but counter to sugges- they wrote,
tions that students who have explored a conjecture
extensively in computer environments often feel no We transferred a segment AB . . . [to] segment r1
need for further justification, de Villiers (1998) re- (AB = AC) [despite what they wrote, the students
ported that he has been able to encourage deductive made the two segments AB and AC equal “by eye,”
thought by asking why DGE results are true. A some- so the transference was not draggable]; we drew two
what different perspective was taken by Hershkowitz circles . . . center in C and point A and center in B
et al. (2002) who argued that and point A. We joined A and [circle intersection]
D. . . . [after testing their construction by dragging]
Failed! (p. 265)
the dragging operation on a geometrical object en-
ables students to apprehend a whole class of shapes in
which the conjectured attribute is invariant, and hence For their second attempt (Figure 19.11b), Alex
to convince themselves of its truth . . . The role of proof and Gio wrote,
is then to provide the means to state the conjecture
as a theorem . . . to explain why it is true, and to enable We drew a circle . . . [that] gave us the segments AB
further generalizations. (p. 676)
and AC belonging to r1 and r2, which are equal be-
cause they are radii of the same circle. We drew two
So formal proof (a) helps students to symbolically circles (centre B and C point A); using the intersec-
and precisely encapsulate their empirical discoveries tion . . . of the two circles, we found the point D that
and (b) offers corroboration for and explanation of we joined with A, determining the angle bisector. (p.
the validity of the discovery. 265, 266)

r1 r1
C

B B
B
A A A O
D C D

r2
r2 D

Problem 1, Attempt 1 Problem 1, Attempt 2 Problem 2

(a) (b) (c)

Figure 19.11 Alex and Gio’s attempts to construct the bisector of an angle (Mariotti, 2001).
880 ■ STUDENTS AND LEARNING

Note that, other than the one statement, “which distant from the sides of the given angle; or was their
are equal because they are radii of the same circle,” approach strictly visual—they simply tried to find a
the justification provided by the students is an expla- way to put a point D “in the middle” of the angle?
nation of the DGE constructions that they used. Most Additional research needs to delve more deeply into
of what the students wrote is a description of process, students’ thinking processes.
not a justification, or even an explanation for why they In contrast to the Alex and Gio episodes in which
did what they did. constructing draggable figures seemed to support stu-
On a later problem—constructing a perpendicu- dents’ reasoning, Mariotti (2001) also presented data
lar to a given line through a given point—Alex and suggesting a divergence between students’ construc-
Gio drew two circles having centers on the line and tion of draggable figures and their proofs. For instance,
passing through the given point C (Figure 19.11c). Cathy was unable to make a draggable construction
Their perpendicular was the line through the inter- for a perpendicular to line though a point P not on
sections of the two circles. To prove the validity of the line. However, she was still able to state and cor-
their construction, they used the facts that AD and AC rectly prove how to make such a construction. She
are radii of the same circle as well as that BD and BC proved that if one chooses points S and Q on the line
are radii of the same circle (and the triangle congru- so that SP = QP, then the bisector of angle SPQ is per-
ence theorems) to show that angles AOC and AOD pendicular to line SQ. She derived this proposition
are congruent and thus right angles. In this case, the by visually manipulating P and observing angle mea-
students seemed to link the structure of their proof to sure changes—her premises, however, were not “con-
their DGE construction; in their proof, they took as structed” into a draggable drawing. Thus, students
given that which was produced by their constructions. can construct valid theorems and proofs even when
So here is an example of using a DGE construction to
they cannot construct draggable diagrams. Although
formulate a proof.
this finding is not consistent with claims of the cen-
It seems that in Alex and Gio’s case, DGE was be- tral role that draggability plays in students’ work with
ing used as a problem context—it provided exercises proof, it is consistent with the view that DGEs provide
for enriching the proof skills of students who had al- an exploratory medium that makes students’ geomet-
ready been taught formal proof. It helped students
ric ideas more analytically manipulable and thus more
learn to use and appreciate geometric proof. It was
accessible to reflection and refinement.
not being used to evolve students’ informal justifica-
tions into formal ones (cf., Borrow, 2000). Research by Healy and Hoyles (2001) provides ad-
ditional insights into some of these issues. They con-
However, several issues arise in considering Alex
ducted a study with above average 14–15-year-olds in
and Gio’s work. First, how did students move from jus-
England who were studying the national curriculum,
tifying only a small part of their construction in the
first problem to the much more sophisticated solution which emphasizes proof in a range of contexts, not
they provided in the second problem? (In order to just geometry. However, despite this emphasis, the re-
prove these ideas, students must have already learned searchers reported that “only a few students [out of
about formal proof.) Second, precisely how were stu- close to 2500 in a nationwide survey] could actually
dents given, and how did they make sense of, a formal construct a valid deductive argument” (p. 239). In the
theoretical framework for proof? For instance, was the experimental treatment, which lasted 6 to 8 weeks,
need to prove the validity of constructions generated after using Cabri to explore congruent triangles and
by students—did they personally feel a need to remove being introduced to formal proof, students were given
uncertainty—or was it socially motivated by either the tasks that the researchers hoped would integrate DGE
classroom culture or demands of the teacher? and proof experiences. (Before participating in this
A third issue involves the nature of students’ geo- study, none of the students had used Cabri nor had
metric reasoning and how it was supported and fur- they created sequences of logically justified geomet-
thered by the DGE. For example, what conceptualiza- rical statements.) For example, in one task, students
tions and reasoning (about geometry and Cabri) led were asked to (a) make and describe a Cabri con-
students to use equal radii circles in constructing the struction of a rectangle, (b) identify the properties
angle bisector? Were circles seen procedurally as a way of the rectangle that they considered given in their
of laying off equal segments, or as loci of points in the construction, (c) manipulate their constructed figure
plane equidistant from a given point? There seems to discover new properties, and (d) prove one of their
to be evidence of both. Was the students’ approach discovered properties.
visual or analytic? That is, did the students use the In response to this task, one student wrote the fol-
circles analytically as a way of making point D equi- lowing (p. 243).
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 881

(a) Constructed line between 2 points and put per- vertices for further empirical validation of their con-
pendicular line through it. Created a parallel line jecture. Then, with the angle bisectors intersecting at
to the perpendicular line and then a parallel line to right angles, the boys used Cabri’s “check-property
the first line. (b) All angles equal 90°. Opposite sides tool” to test whether BA and CD were parallel. The
parallel. (c) Opposite sides equal. (d) Used a triangle
boys were “disappointed and puzzled that the prop-
congruence proof to prove that opposite sides are
equal. [For the student’s proof that the properties in erty that the two lines were parallel was declared not to
(b) imply the property in (c), see Figure 19.12.] be true in general” (p. 244). However, because in the
counterexamples presented by Cabri’s check-property
A C tool, the angle bisectors did not intersect at right an-
gles, the boys decided that the parallel property had
been declared invalid because the condition of 90°
between the angle bisectors had not been retained by
the construction.
B D
m ∠BEC = 90° B
All angles equal 90° C
∠ABD = ∠ACD . Given.
∠ACB = ∠CBD . Alternate angles equal. E
∠ABC = ∠DCB. Alternate angles equal.
∆ABC ≈ ∆BCD . All angles equal.
BC is common to both triangles. A

[ ∴ ]Both triangles congruent → AB = CD.

Figure 19.12 Student’s proof. D

This student was able to construct a draggable

rectangle and to use a triangle congruence proof to Figure 19.13 Tim and Richard’s quadrilateral with angle
justify his statement that the opposite sides of a rect- bisectors of 2 adjacent angles intersecting at right angles.
angle are congruent. However, consider this student’s
creation of a draggable rectangle in more detail. After Note that the boys’ realization required sophis-
creating the first right angle by using a perpendicular, ticated knowledge about the nature of Cabri con-
it is unclear how he conceptualized what he was do- structions and, more importantly, the function of
ing. Did he create parallels because he wanted to give the check-property tool. This latter knowledge is
the figure the “opposite sides parallel” property about knowledge about Cabri, not geometry. So they had
which he wrote? How did he know, or did he discover, to possess a fairly sophisticated understanding of the
that this would produce right angles (which he did nongeometric functioning of Cabri to deal with the
not prove, but assumed in his proof)? Was there any instructional task presented to them. Students who
explicit connection between the student’s DGE con- do not understand this technology sufficiently would
struction and his proof (for instance, did his use of have their geometric investigation thwarted.
parallelism in constructing the rectangle cause him to Tim and Richard continued investigating their
focus on parallelism in his proof)? conjecture by constructing a line parallel to CD pass-
In the final activity of the Healy and Hoyles teach- ing through B. They found that whenever point A was
ing experiment, students were asked to construct a dragged onto this line, the two angle bisectors inter-
quadrilateral in which the angle bisectors of two ad- sected at right angles, but if A was not on this line, the
jacent angles intersect at right angles. They were to angle bisectors were no longer perpendicular. They
discover, then prove, other properties of this quadri- formulated the conjecture, “Whenever the lines are
lateral. Tim and Richard created quadrilateral ABCD, parallel the angle bisectors are perpendicular.” In
then the bisectors of angles ABC and BCD (see Figure searching for a proof for their conjecture, the boys
19.13). They measured the angle at the intersection returned to manipulating the quadrilateral, with A
of the two bisectors and dragged the vertices of the always on the line parallel to segment CD. They mea-
quadrilateral until this angle measured 90°. They then sured angles and looked for angles that remained
noticed that BA was parallel to CD and conjectured equal when dragging vertices. The boys noticed that
that whenever the two angle bisectors were at right the alternate interior angles where the bisector of an-
angles, BA would be parallel to CD. They dragged the gle ABC intersected segments BA and CD were always
882 ■ STUDENTS AND LEARNING

“the same,” then realized that proof of this equality tions proved too difficult for students. Because they
would show that the two line segments are parallel. could not use the Cabri tools available to them to con-
However, before Tim and Richard moved from struct the quadrilateral, they became frustrated and
their Cabri construction to a strictly formal analysis of failed to solve the problem.
the situation, they used their construction to further The experimenters’ solution to the second diffi-
explore their conjectures. But this was complicated by culty was that a more flexible, programmable environ-
the fact that the angle bisectors did not always intersect ment is needed.
the sides of the quadrilateral, so the relevant angles
could not always be measured. The issue was resolved From all our observations of students’ interaction
when the teacher suggested that they construct lines with Cabri, we have become increasingly aware that
through the sides of the quadrilateral, which allowed the mediation of students’ activities by the software
the boys to measure the relevant angles in all drag- is not necessarily positive for their engagement and
for their learning. . . . As we have shown in the work
gable positions. At this point, the boys stopped ma-
of our less successful students, learners can . . . find
nipulating the figure. The researchers claimed that themselves in a position where they are unable to
the boys had “used the figure they had generated with use the tools they have in mind, even if they are con-
its specific measurements as a generic example, to vinced that their use would make sense mathemati-
help them to formulate a general argument” (Healy cally. (Healy & Hoyles, 2001, p. 252)
& Hoyles, 2001, p. 246).
The DGE construction played two important One can draw several conclusions from the Healy
roles in Tim and Richard’s work. First, it provided and Hoyles research, though not necessarily those
them with a representation that allowed them to in- drawn by the authors. First, the DGE apparently
vestigate the situation empirically so that they could helped some students understand and solve posed
formulate a valid conjecture about a geometric prop- problems. (Recall, however, that the students who did
erty. Second, the boys’ manipulations of their DGE this seemed to have well-developed proof skills and
representation, including the accurate measurements understanding of associated geometric ideas.) Sec-
of angles, helped guide their construction of a proof. ond, the DGE evidently helped some students develop
This measurement process was not only more accu- appropriate proofs.
rate, it was more general than that which could have Just as evident is that other students were stymied
been conducted with paper and pencil because it was in their geometric problem solving. However, it can
maintained during dragging. Thus, in this case, for be argued that the difficulty lies not with the DGE, as
these students, Cabri seemed to form a bridge be- claimed by Healy and Hoyles, but with the design of
tween empirical exploration and proof. Dynamic ge- the instructional tasks. For instance, in the last activity,
ometry helped these students find reasons why their the problem was not chosen so that students could use
empirical discovery was true. There was no evidence Cabri to construct, in a reasonably easy way, a drag-
that the DGE discouraged proof activity. gable figure satisfying the problem hypotheses, even
However, on this same task, many students failed though the researchers, and many students, expected
to see the parallel relationship between the sides of students to do so. If the point of the activity was for
the resulting quadrilateral. Also, making a draggable students to discover and explain a geometric relation-
figure with the “givens” was not the route successful ship, there was no reason to expect them to create, a
students took. Like the two students above, all students priori, a draggable figure that maintained that rela-
who were successful on this problem constructed a tionship. It is much more reasonable to expect stu-
general quadrilateral with adjacent bisectors and ma- dents to construct a DGE figure that allowed them
nipulated it to find when the bisectors were parallel. to reflect on this relationship (as did Tim and Rich-
In fact, students who attempted to construct a quadri- ard). Alternatively, why not give students a draggable
lateral with the given bisector-perpendicularity prop- figure with the desired properties and let them use
erty—as the researchers had intended—encountered it to explore the problem? Perhaps tasks that do not
two problems. The first was reasoning that the given ask students to make their own constructions are bet-
properties were satisfied by a particular quadrilateral, ter for some students at some times. In general, then,
like a parallelogram. Students using this reasoning researchers should be asking what students learn by
either became confused about which properties were using specific DGE tools, and what knowledge—both
given and which were to be deduced or did not see geometric and technological—is required to produc-
that the solution they derived for a special case was in- tively use these tools.
complete. The second difficulty was that constructing Comparisons of student learning with various task
a general quadrilateral that satisfied the given condi- designs could be very revealing. For instance, given
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 883

the problem of investigating quadrilaterals in which a proof. It is as if there was no interaction between
the angle bisectors of two adjacent angles intersect at visualisation and proving” (Laborde, 2001, p. 303).
right angles, researchers might compare the work of
students in four different task environments: (a) no Final Comments on DGE
DGE, (b) making a draggable figure with the given Several major issues and areas for further research
property, (c) making a draggable figure like Tim and have arisen from the above discussion.
Richard’s, and (d) using a preconstructed draggable
Issue 1. There are two major theoretical perspec-
figure with the given property. The research should
tives concerning the use of DGEs to promote the de-
compare not only student success rates, but their strat-
velopment of students’ geometric thinking. In the
egies, required starting knowledge for success, and
first perspective, dynamic figures, especially draggable
what they learn from the experience. (Even though
the Healy and Hoyles research shows that environ- ones, are seen as generating numerous examples. For
ment (b) is very difficult for students, it would still be example, Marrades and Gutiérrez claimed that the
interesting to better understand why it is so difficult.) main advantage of DGEs “is that students can con-
Related to this suggested research, Arzarello, Miche- struct complex figures and can easily perform in real
letti, Olivero, Paola, and Gallirio (1998) attempted to time a very wide range of transformations on those fig-
analyze the types of DGE dragging explorations of stu- ures, so students have access to a variety of examples
dents during different phases investigation of a geo- that can hardly be matched by non-computational or
metric problem. Specifically, they investigated types of static computational environments” (2000, p. 95).
dragging during empirical activity aimed at trying to In the second theoretical perspective, draggable
understand a problem or generate and explore a con- drawings are seen as interesting, manipulable, visual-
jecture (their “ascending” phase), and activity aimed mechanical objects that have movement constraints
at building a deductive proof (their “descending” that can be conceptualized and analyzed geometri-
phase). They suggested that different kinds of drag- cally (Battista, in press b). This approach is taken in
ging occur in different phases. Battista’s Shape Makers environment and Laborde’s
In another study of the use of DGE to support “black-box” activities in which students are given a
the development of students’ proof skills, Marrades DGE figure without being told how it was constructed
and Gutiérrez (2000) described differing levels of so- and are asked to construct a draggable figure that be-
phistication in students’ discovery and justification of haves in the same way (Laborde, 1998). In these drag-
a theorem about the concurrence of perpendicular gable drawings, geometric properties are visually and
bisectors of the sides of a quadrilateral. They stated mechanically manifested as constraints and allowanc-
that “the process of getting conjectures was grounded es of movement and therefore not only can be seen,
on the observation of drawings and regularity in the but can be felt. Thus, rather than examining a set of
measures of angles” (p. 105). Without prompts for figures to conceptualize how they are the same, with
proof, students seemed convinced by empirical valida- Shape Makers or black-box constructions, students in-
tive actions of DGE. However, with teacher prompt- vestigate how an interesting manipulable “geometric”
ing, students seemed able to use DGE in concert with object works (which can, with proper instructional
proof. When students arrived at a sequence of con- guidance, support the construction of meaningful
structions that enabled them to conceptualize why conceptualizations of geometric properties). Seem-
the theorem was true and their dragging/measuring ingly consistent with this view, some researchers argue
actions convinced them of the validity of their ideas, that DGE should not be conceptualized as a device to
they were able to prove the theorem in a way consis- merely “enable us to do what we already do faster, bet-
tent with their construction-based conceptualizations. ter or to a larger extent or degree,” but as an environ-
Thus this research supports the claim that DGE explo- ment requiring a fundamentally different cognitive
ration can provide insight and scaffolding for proof, view than has been taken in traditional approaches to
as well as confidence to proceed with a given line of geometry instruction (Dörfler, 1993, p. 161).
justification (as claimed by de Villiers, 1998, when he In considering these theoretical perspectives,
cited Polya, “When you have satisfied yourself that the which guide both research and curriculum develop-
theorem is true, you start proving it”). ment, several questions arise. First, it is important
Unfortunately, many teachers apparently are to test the theories and determine empirically how
not taking advantage of DGEs’ capability to support students conceptualize DGE draggable figures. Do
the formulation of a proof: “When the students were students see the various configurations of a dynamic
asked to justify, the teachers did not mention the pos- drawing as examples of a class of geometric shapes, or
sibility of using Cabri to find a reason or to elaborate do they see a dynamic drawing as a geometrically in-
884 ■ STUDENTS AND LEARNING

teresting manipulable object? How are students’ views contrast, when quantitative results indicate that one
on this matter affected by instruction? instructional treatment is more effective than anoth-
Second, because much of geometry is about cat- er, qualitative methods should be used to determine
egories of objects, do students who think about DGE why such results were obtained.
figures as manipulable objects naturally move to Issue 4: The allure of draggable figures. Investigat-
thinking about classes of shapes? If not, how can such ing draggable figures in DGE seems to strongly attract
movement be engendered? Does students’ reasoning the interest not only of students but of researchers in
about DGE manipulable objects transfer to thinking geometry learning. Why is investigating such figures
about geometric categories? Battista’s Shape Makers so appealing? One suggestion for the appeal of drag-
research suggests that instruction can promote such gable figures comes from the theory of psychological
transfer. Are DGE figures viewed differently by stu- essentialism. This theory conjectures that people act
dents who have already attained strong conceptual- as if things have essences or underlying natures that
izations of the underlying geometric concepts before make them what they are, and that when an essence
using a DGE, as opposed to students who initially is unknown, it may motivate the search for new mean-
construct meaning for these concepts in a DGE? That ings (Gelman & Diesendruck, 1999). Thus, one rea-
is, for an expert, a dynamic rectangle maker might son that DGE draggable figures may be instructionally
indeed be a representation of the class of rectangles, useful is that students naturally attempt to determine
whereas for the novice it might be simply an interest- the essences of these objects—how do they move, why
ing dynamic object whose movement is constrained do they move the way they do.
in particular ways. Issue 5: Draggability and invariance. A number of
Issue 2. How do students conceptualize particu- researchers have claimed that one of the reasons that
lar DGE constructions? For instance, what is involved DGE draggable figures are useful for helping students
conceptually when a student learns to construct a par- develop understanding of geometric properties is that
allel to a given line through a point not on that line? such properties are invariant under dragging move-
What does the construction mean to the student? Is it ments. For instance, Laborde claimed that the spatial
a procedure for constructing a visual configuration, properties of draggable drawings “may emerge as an
or is it a representation of a geometric concept? How invariant in the movement whereas this might not be
does using the construction change, supplement, or noticeable in one static drawing” (Laborde, 1998, p.
supplant a student’s conceptualization of parallel- 117). Similarly, Battista (in press b) suggested that
ism? That is, learning changes a student’s network of properties of shapes are more noticeable as invari-
cognitive structures—how does learning to use DGE ants of draggable movements than as commonalities
construction tools change students’ related concep- across static examples.
tual networks? Conversely, what kinds of conceptual Several clues in the psychological literature point
networks enable students to productively employ DGE toward a possible explanation of this phenomenon.
construction tools? Is the nature of conceptualizations First, the mind’s attentional system is especially alert
created via use of DGE construction tools the same as and sensitive to change (Ornstein, 1991). Thus, when
those learned through verbal definitions or drawing? viewing a figure as it is dragged, because the figure
Issue 3. In general, much additional qualitative maintains its identity as a single entity, changes in its
and quantitative research on the use of DGE in geom- shape are naturally attended to. Second, “[the] search
etry learning is needed. Qualitative research is needed for constancy, the tendency toward certain invari-
to investigate numerous unanswered questions about ants, constitutes a characteristic feature and imma-
the nature of students’ construction of geometric nent function of perception” (Cassirer, 1944, p. 21).
knowledge in DGE. Quantitative studies are needed Third, comparing a set of static figures and compar-
to determine whether use of DGE produces greater ing various configurations of a dragged figure should
learning than use of paper and pencil. However, ge- both induce structure-mappings (Gentner & Medina,
ometry research, and especially its application to 1998) that can make properties noticeable. However,
classroom practice, is best served by integrated com- transforming draggable figures should facilitate this
binations of qualitative and quantitative studies. For structural analysis because the continuity of the trans-
instance, as has been shown previously in this chapter, formation establishes a structural-mapping correspon-
qualitative research results indicate that high quality dence between components of compared figures,
geometry learning can occur with use of DGE. But whereas when comparing static figures, the perceiver
quantitative research is needed to determine the gen- must establish the correspondence (which may not be
eralizability of these results and whether use of DGE easy). Of course, if students do not attend to the en-
is “better” than use of paper-and-pencil techniques. In tire dragging transformation, the original figure and
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 885

the figure that results from dragging may actually be 1. In what ways are the figures alike?
viewed as two separate entities, losing this facilitative
effect. Fourth, a considerable amount of research has
shown that objects that move together are seen as part
of the same object (e.g. Driver & Baylis, 1989; Regan
& Hamstra, 1991). So, dragging motions might cre-
ate “visual chunking” that facilitates the perception
of visual regularities indicative of properties. For in- 2. Explain how shape N is different from shapes P and Q.
stance, because in the Parallelogram Maker the oppo-
site sides move together in a way that maintains their
parallelism and congruence, such movement might
encourage the formation of two “opposite-sides-par- N Q
P
allel” composites or chunks. These composites are a
visual manifestation of the geometric properties of
Figure 19.14 NAEP tasks.
opposites sides parallel and congruent (although,
in this form, the two properties might not be sepa-
rable, which is consistent with the observation that the geometry strand that is consistent with van Hiele-
children’s initial conceptualization of parallel often based research. However, the property focused on in
includes congruence). Thus these composites might Task 2 is not, as has been previously discussed, a good
form the spatial structuring needed later for students’ indicator of van Hiele Level 2; number of sides does
geometric structuring. not involve relationships between parts of shapes. In
contrast, the properties that discriminate the shapes in
Task 1 focus on relationships between parts of shapes
and thus are good indicators of van Hiele Level 2. Un-
RESEARCH ON OTHER GEOMETRIC TOPICS fortunately, and inconsistent with research on geom-
AND NONCOMPUTER ENVIRONMENTS etry learning, NAEP administered the low-level Task 2
across the grades (4, 8, 12) and the richer Task 1 only
at Grade 4 where one would expect to see only mini-
The National Assessment of Education
mal development. Furthermore, the scoring guide for
Progress (NAEP)
Task 1 is not based on a research-derived classifica-
Traditionally, the NAEP has been a useful tool for tion of justifications. For instance, the scoring guide
getting an overall picture of U.S. students’ mathematics weighs the following properties equally: having 4 sides
skills. To illustrate, recent administrations of the NAEP or angles, having two sets of equal side lengths, hav-
in mathematics have shown some improvement in ge- ing 4 right angles, having equal heights, having equal
ometry, but students’ performance in many areas of ge- areas, and “one is more slanted.” So informal proper-
ometry and spatial sense is still low (Kloosterman et al., ties, properties dealing with numbers of components,
in press; Sowder, Wearne, Martin, & Strutchens, 2004). and relational properties are given equal weights. But
However, it is difficult to reliably integrate recent NAEP research suggests that these properties vary consider-
results into the research corpus on geometry learning ably in sophistication. Thus, because the NAEP items
because of two major problems with the NAEP pro- and scoring are only loosely connected with research
gram. First, in recent years, NAEP assessment tasks have on students’ learning of geometry, they provide inad-
not been released to researchers for analysis, making equate information about students’ geometry learn-
the results almost impossible to interpret seriously. For ing for teaching as well as for curriculum evaluation
instance, it is impossible to make much sense out of a and development.
vague NAEP item description such as, “Determine the
shape formed under certain conditions.”
Proof and Justification in Geometry24
Second, many NAEP items and scoring rubrics
are substantively disconnected from current research In 1992, Clements and Battista (1992) concluded
on geometry learning. For instance, consider the two that, in geometry, students are extremely unsuccess-
tasks shown in Figure 19.14. ful with formal proof and that they are deficient in
Both tasks are intended to assess students’ ability their ability to establish mathematical truth. More re-
to use properties to compare shapes, a critical goal in cently, Hershkowitz et al. (2002) drew a similar con-

24
See (Harel & Sowder, 1998) for a discussion of the proof schemes students use throughout mathematics, not just in geometry.
886 ■ STUDENTS AND LEARNING

clusion: “The teaching of mathematical proof appears item that asked students to rewrite a conjecture in “if-
to be a failure in almost all countries” (p. 675). In a then” format and specify the givens and what was to be
large study in the United Kingdom, Healy and Hoyles proved (33–53%). These researchers also constructed
(1998) reported that (a) high-achieving Year 10 stu- a questionnaire that probed students’ beliefs about
dents showed poor performance in constructing essential elements of proofs. About 78% of students
proofs, with empirical verification the most popular agreed with “the principle that one purpose of proof is
form of justification; (b) students’ proof performance to provide insight into why a statement is true.” About
was considerably better in algebra than in geometry; 67% of students agreed with a statement indicating
(c) students recognized that once a statement has that “knowing that a general statement is true implies
been proved, no further work is necessary to check its that the statement is true for all specific instances.” In
validity; and (d) students were better at recognizing a contrast, only 22% of students found the logical flaw
valid mathematical argument than constructing one in a fairly simple two-column proof. On a purported
themselves. Hoyles and Jones (1998) reported that proof that the sum of the angles in any triangle is 180
(a) students do not distinguish between empirical and degrees, students were presented with a table of angle
deductive arguments and prefer empirical arguments; measurements for four scalene triangles—3 obtuse, 1
(b) for many students, deductive proof only adds evi- acute. About half the students claimed that this table
dence for validity; and (c) most students regard proof constituted a proof, with many students saying so be-
as an irrelevant add-on and do not understand its pur- cause a range of types of triangles had been tested.
pose. Similarly, Hershkowitz et al. (2002) stated that Some students claimed that the table did not consti-
students rarely see the point of proving. Battista and tute a proof because it omitted right angles.
Clements (1995) argued that although proof is criti- Chazan’s (1993a) investigation of high school stu-
cally important to mathematicians because it enables dents’ views of geometry proof elaborates and clarifies
them to establish the validity of their mathematical some long-standing issues in the research reported
thought, students do not conceive of proof as estab- above.25 In interviews, he found that 5 out of 17 high
lishing validity, but see it instead as conforming to a school geometry students believed that measuring ex-
set of formal rules that are unconnected to their per- amples “proved” assertions, 7 did not believe that such
sonal mathematical activity. empirical evidence proved assertions, and 5 were un-
Consistent with earlier research, Koedinger clear or changed their minds about this issue during
(1998) reported that it was rare for high school stu- the interview. Some students who believed that em-
dents to think that evidence beyond examples was pirical evidence is proof focused on the number of ex-
needed to support geometric conjectures. He also stat- amples examined, whereas others focused on the type
ed that many students, when considering conjectures, of examples examined. As an example of the latter ap-
first said that the one example they drew was enough proach, in considering a proposition about triangles,
evidence, then, after prompting, that many examples one student said, “Right, obtuse, acute and like isos-
were needed. Most students had to be explicitly asked celes or something like that . . . it’s pretty much the
to write a proof “like they had seen in class” before same effect as having every one” (p. 369).
they attempted to produce one. Also, most students Of the students who believed that empirical evi-
had difficulty formulating proofs. In fact, Koedinger dence does not constitute proof, some cited the fact
reported that only 10% of students successfully for- that there may be counterexamples because one can-
mulated proof problems and began to work on them. not examine all examples. Other students cited the
Interestingly, students found it especially difficult to inexactness of measurement, a point that raises an
state the givens. important issue. On the one hand, side lengths can
McCrone and Martin (2004) found that high be considered as the result of a measuring process
school geometry students’ performance was low on (that is, they are obtained by the act of physically
almost all items on a proof-construction assessment. measuring and, as such, are only approximations).
Scores were lowest on items in which students had to On the other hand, side lengths can be considered
construct their own proofs, even on items that pro- theoretical (that is, they represent the actual length
vided a proof outline (averages ranged from 6–34% measures of the segments; as such they are exact and
correct). Results were a bit better for a fill-in proof without error). In the first case, skepticism about jus-
(44–50%). Scores were highest, but still low, on an tifications based on lengths is warranted; in the latter

25
First, use of the Geometric Supposer was an integral part of Chazan’s students’ courses. But because I focus only on what the research
indicates about proof, I discuss this research in this section rather than in the section on the Supposer. Second, some students were
interviewed in January, some in March—so students had not completed their geometry courses.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 887

case, skepticism is unwarranted. It is important for to (a) a lack of understanding of the logic of proof,
researchers to determine precisely not only how stu- (b) a healthy critical attitude toward mathematical
dents think about measures given in proof problems justification (so students are skeptical not about the
(e.g. approximation versus exact), but how instruc- logic of proof but about particular implementations),
tion and computer geometry programs deal with the or (c) not being exposed to proofs that help them un-
measure/approximation issue and how this affects derstand why a theorem is true (that is, they do not
students’ conceptualizations. believe what does not make sense to them).
In contrast to students who believed that empiri- According to Chazan (1993a), some students be-
cal evidence was proof, some students thought that lieve that deductive proofs apply only to single dia-
proof was merely evidence (Chazan, 1993a). For in- grams or to single types of shapes. In the latter case,
stance, some students believed that counterexamples a student might believe that a two-column proof had
were still possible, even if a deductive proof in a text- to be written, say, for an acute, obtuse, and a right tri-
book was given. However, Chazan’s finding could be angle (and possibly isosceles and equilateral triangles
due to students not understanding the logic of proof, as well). Chazan noted that the notion of multiple
or it could be a reasonable reaction to the possibility proofs for multiple examples might be a carryover
of an incorrectly implemented proof (especially if the from thinking about proof by examples—students
proof is written by a student26). This brings up an im- want to see multiple examples to convince them-
portant issue. Theorems presented in school mathe- selves.28 But there may be more to it than this. Pro-
matics textbooks, and their accompanying proofs, are gressing from reasoning about a specific example to
generally assumed by teachers and students to be cor- reasoning about a deductive proof is similar in some
rect because they appear in the book, which is taken sense to moving from reasoning about specific num-
as authoritative, rather than because the proofs are bers to reasoning about variables. For instance, for
critically examined by students and teachers. This is a specific diagram, one might say that each of two
very different from how mathematicians work—usu- sides measures 50 so they are equal in length; in a
ally, proposed new theorems and proofs are greeted deductive proof of a general proposition, however,
with skepticism, and often with attempts to find coun- we might say that two sides are congruent—that is,
terexamples.27 So, for some students, skepticism about algebraically, the measure of both sides both equal
deductive proofs should not be viewed as a weakness some indeterminate number x. Moving from specific
but rather as justified and healthy. to variable thinking requires a significant jump in ab-
Another reason that students might be skeptical straction. Thus, the multiple proofs for multiple ex-
about the generality of a deductive proof occurs when amples conception might indicate a transitional stage
the proof does not really provide insight into why a in abstraction in moving from empirical to deductive
proposition is true (recall the McCrone and Martin’s, thought. In any event, it is clear that students possess-
2004 finding that 78% of students agreed that proofs ing the multiple proofs for multiple examples view do
should provide insight into why statements are true). not genuinely understand the concept of deductive
For instance, one can read a proof and believe in its proof. However, this should not be surprising, given
validity because each step in its argument is valid but that most instructional treatments do not spend sig-
still not, in the words of Polya, “see it at a glance.” In nificant time specifically helping students understand
this case, the proof does not provide enough insight the nature of mathematical justification and proof.
to genuinely make sense of the proposition, so one Instead, these treatments focus mostly on having stu-
might still be skeptical. In such instances, it is math- dents master formal methods of proof.
ematically healthy and reasonable to further explore In summary, researchers are still grappling with
the proof by examining examples, perhaps reviewing the basic question: Why do students have so much
the proof for some of these examples, even develop- difficulty with geometric proof? Is it because they see
ing alternative proofs (think of how many proofs have little need for proof, because proof is so abstract that
been given for the Pythagorean theorem). Thus, in- it is very difficult to learn, or is it a combination of
stead of assuming that student skepticism about proof the two—that because students see proof as difficult
must be a negative finding, it is important for research- and not useful, they do not expend sufficient intellec-
ers to investigate whether students’ skepticism is due tual effort to master it? What components of proof are

26
In this research, it is not always clear if the students’ statements are about textbook proofs or about student proofs.
27
In fact, analyzing why counterexamples cannot be found often provides insight into why a proposition is true.
28
Of course, there are proofs in mathematics texts that are broken down in terms of cases. But students in school would not have been
exposed to many such proofs.
888 ■ STUDENTS AND LEARNING

difficult for students, and why? How can proof skills suggested that construction of a formal mathematical
best be developed in students? More research is re- formulation for the standard abstract angle concept
quired on students’ development of geometric proof of two rays with common endpoint can be considered
skills and their understanding and beliefs about the the fourth stage of angle concept development. They
nature of proof. As was discussed in the section on further argued that, when students are forming their
DGE, use of DGE has shown promise in encouraging first abstract angle concept, they may be in the process
and supporting the development of students’ geomet- of forming additional contextual angle concepts. Fur-
ric proof. To better understand the nature of the fa- thermore, they claimed that angle concepts are con-
cilitative effect of DGE on proof, however, qualitative stantly generalized as new situations and contexts are
research comparing students’ proof processes in DGE recognized as similar.
and paper-and-pencil environments is needed. Similar to several other researchers in geometry
Apparent contradictions in the research also need learning, Mitchelmore and White use a theory of ab-
to be resolved. For instance, Healy and Hoyles (1998) straction to analyze students’ thinking.
reported that students recognized that once a state-
ment has been proved, no further work is necessary For example, students often demonstrate the similar-
to check its validity, but Chazan (1993a) reported that ity between a tile and a ramp by physically moving a
corner of the tile to fit into the space between the
some students believed that counterexamples were
ramp and the horizontal base. . . . The recognition
still possible, even if a deductive proof in a textbook of similarities between different angle contexts is
was given. As another instance, Koedinger (1998) therefore a constructive process requiring reflective
reported that students found it especially difficult to abstraction. (2000, p. 216)
state the givens, whereas McCrone and Martin (2004)
reported that scores were highest, but still low, on an Could it be that the critical component of this reflec-
item that asked students to rewrite a conjecture in if- tive abstraction is the mental operation of matching
then format and specify the givens and what was to corresponding parts of angle contexts, and that this
be proved. It is likely that such contradictions are due matching is the same type of structural mapping dis-
to the fact that, in many cases, small numbers of as- cussed by Gentner and Medina (1998)?
sessment tasks were examined to draw conclusions, Mitchelmore and White (2000) also suggested
and results are task specific. That is, although there that an essential part of the process of mentally con-
are general processes applicable to most geometric structing a general conceptualization of angle is link-
proofs, success on most proofs also depends on un- ing together different angle concepts. This returns
derstanding specific concepts and situations. Research the discussion to the issue of abstraction as disembed-
in this area must balance investigating general versus ding/decontextualizing versus linking. In fact, the de-
task-specific components of proof. velopmental sequence described by Mitchelmore and
White provides a nice example of how the sequence
Understanding Angles of abstractions that lead to abstract angle concepts in-
volves both disembedding and linking. For example, to
In an investigation of the development of angle compare the angle concepts embodied by the corner
conceptualizations in elementary school students, of a tile and the space between a ramp requires the
Mitchelmore and White (2000) describe three stages disembedding operation of interiorization because
of abstraction. In Stage 1, similarities between angle-re- the original abstractions of these embodiments are
lated situations that look alike, involve similar actions, applied outside of their initial contexts. But the reflec-
and are experienced in similar circumstances lead to tive abstraction that compares the tile-and ramp-angle
abstraction of situated angle concepts (conceptualizations concepts can result in a new abstraction that is linked
in my language). For instance, situations involving hills to both ramps and tiles.
would be a single situated angle concept. In Stage 2, Mitchelmore and White (2000) reported three
similarities between different situated angle concepts abstract angle concepts: point is sometimes used by
are abstracted to form contextual angle concepts. For in- young children to relate corners and intersections; a
stance, students might abstract the similarity between single sloping line relates sloping and turning objects;
hills and roofs into a contextual angle concept of and, most common, two inclined lines meeting at a point,
“slope.” In Stage 3, similarities between different con- which they call the standard angle concept, can be used
textual angle concepts are abstracted to form abstract to relate all physical angle contexts. Mitchelmore
angle concepts. For instance, students might abstract and White found that more than 80% of students in
the similarity between intersections of lines, corners, Grades 2–6 could see similarities between a standard
bends in paths, and slopes. Mitchelmore and White angle (as represented by a folding straw) and physical
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 889

contexts of scissors, a folding hand fan, a diagram of of structural angle components, which leads to failure
a road intersection, a four-sided tile, and two inter- in establishing appropriate structural mappings.
secting walls. Students had more difficulty relating the Because of the difficulty that students have in see-
standard angle to a wheel or door (turning) and a hill ing standard angles in turning contexts, Mitchelmore
(slope). Less than 30% of students in each of Grades and White (2000) claim that defining an angle as an
4, 6, and 8 represented the slope of a hill using a stan- amount of turning (about a point) from one line to
dard angle between the hill and the horizontal, with another is inappropriate for beginning learners. But
many, instead, representing the slope as the angle be- the notion that understanding angles in terms of
tween the vertical and hill. However, this finding must turns is necessarily bad for children is contraindicated
be tempered by the fact that in the representation of by some (but not all) Logo research. Indeed, Hof-
the hill used in this study, the angle with the vertical fer claimed that “turns or rotations are natural for
was much more salient (see Figure 19.15). young children . . . this dynamic way of working with
angles should be started early and carried throughout
the students’ work” (1988, p. 251). In support of this
claim, Clements and Battista’s research on elemen-
tary students’ learning in the Logo Geometry curricu-
lum shows that learning about angles in the context of
turtle turning and paths can be effective for elemen-
tary students, if the relationship between amount of turn
and measure of angle produced is explicitly and carefully
taught (Clements & Battista, 2001). However, contin-
ued research is needed on how best to instructionally
Figure 19.15 Representation of slope of a hill
develop general abstract conceptualizations of angle
(Mitchelmore & White, 2000, p. 221).
that are adequate for the many manifestations of this
concept in mathematics.
Most importantly, the difficulties that students
had relating the standard angle concept to various an-
gle contexts seemed to be directly dependent on the Coordinate Systems and Locations
visual availability or salience of structural angle com- Locating objects in space is a critical process in
ponents (the two sides and vertex) in the contexts. geometry and everyday life. Geometry formalizes this
For instance, a large portion of students could not process through the use of coordinate systems. In this
identify the two lines that make up a standard angle in section, I briefly describe some research issues related
the contexts of a wheel or door, and few students saw to the use of coordinate systems.
the standard angle as slope plus horizontal in the hill
context illustrated above. Thus, the availability and Types of Coordinate Systems
salience of these structural components seemed to Students are exposed to two different types of “co-
determine students’ success in performing the struc- ordinate-system” conceptualizations, maps and Carte-
tural mappings required to see similarities between sian coordinate systems. Maps utilize nonmetric ordinal
examples from different contexts. coordinates to specify locations: Labels on axes signify
The findings that students have difficulties relat- ordered locations without metric relationships. To lo-
ing the standard angle concept to inclination and cate a point in 2d space, one must find specified lo-
turning are important not only in their own right but cations on the axes, then move orthogonally to both
because these two contexts are extremely important axes to find the intersection. The key idea here is not
in postelementary mathematics such as algebra (incli- distance, but coordinating order on two intersecting
nation as slope) and trigonometry (turning). Mitchel- lines.29 On geographic maps, “coordinates” (which
more and White (2000) also suggested that the diffi- can be letters or numbers) often specify intervals,
culty that students have in learning to use a protractor not points. Of course, the difficulty with this concep-
might stem from the fact that, on a protractor, sev- tualization is dealing with “in-between” locations. In
eral lines can be chosen for the initial side of an angle contrast, Cartesian coordinate systems are based on
but the terminal side must be imagined. So again, the both axis-orthogonality and uniform metric axis scales,
cause of students’ difficulties seems to be the absence which ensures that relationships between points are

29
Although axes are usually perpendicular with equally spaced labels, this concept works even if the lines are not orthogonal and labels are
not equally spaced.
890 ■ STUDENTS AND LEARNING

uniform throughout the coordinate system and allows of coordinates were action based was the finding that
the system to be used to represent many important some students who did not usually reverse x and y co-
geometric properties. ordinates would do so when the first coordinate was
Research is needed to determine which of these 0. Such students moved horizontally for the second
two conceptualizations students are developing in coordinate (e.g., go over 6) as the first step in locating
their classroom work with coordinate systems. Such the point, needing a nonzero movement to start their
research must account for the fact that mathemati- plotting procedure.
cal coordinate axes have multiple purposes. One use Another difficulty observed by Sarama et al. (2003)
of coordinate axes is simply to locate points in space. was students not seeing that coordinates were given
This use does not require metric properties of num- with respect to a particular reference point, the origin
ber lines, only order properties. However, full under- (0, 0). For instance, some students, after having plotted
standing of mathematical coordinate systems affords one point, plotted the next point relative to that point,
students more than the ability to locate points. It not the origin. Of course, this error may have occurred
also allows them to reason about distances between because students were working in a Logo environment
points using coordinates. This is a critical property of in which many commands are relative.
Cartesian coordinate systems and one of their major So what does it take to construct a fully developed
sources of conceptual/reasoning power. To construct conceptualization of a mathematical coordinate sys-
a full understanding of Cartesian coordinate systems, tem? Sarama et al. (2003) suggest that students must
merely having students locate points is not enough; be able to “distributively coordinate” two orthogonal
students must also analyze distances between points “conceptual rulers.” This idea is similar to students’
and how those distances can be determined from co- conceptualization of 2d arrays of squares (Battista,
ordinates (see, for example, Sarama, Clements, Swam- Clements, Arnoff, Battista, & Van Auken Borrow,
inathan, McMillen, & González Gómez, 2003). How- 1998). To construct a properly structured mental
ever, researchers and teachers must be alert to the fact model of a 2d array, students must iterate a row com-
that students’ initial conceptualizations of coordinate posite along the elements of a column. To construct
systems may be skewed by early map-like uses. a properly structured mental model of a coordinate
plane, one must continuously sweep one coordinate
An Example of Research on Student Learning axis (conceptualized as a fully metricized number
of Coordinate Systems line) along the orthogonal axis. However, the coor-
Sarama et al. (2003) found that, when initially dinate axis situation is more complicated than the
working with numerically labeled axes, some fourth array situation because arrays are discrete, but coor-
graders ignored the labels. For instance, to locate the dinate axes are continuous. So one cause of students’
point for (8, 9), one student, TA, counted 8 lines over problems with coordinates axes may be the difficult
and 9 lines up, even though the grid lines were labeled shift from discrete to continuous objects. Another dif-
with numbers (so counting was unnecessary). The ficulty that Sarama et al. (2003) observed is that some
authors argued that such students did not take the students fail to interrelate coordinates and locations
numeric grid labels as symbols or curtailments of the of multiple points. For instance, asked to plot (20,
actual counting process. They conjectured that this is 35) and (25, 35), less sophisticated students plotted
similar to students not being able to count on—they them separately. In contrast, more advanced students
cannot operate on the first number in the count as saw immediately that the second point was 5 units to
representing a given number of counting acts—they the right of the first point. The advanced students had
must produce the counting acts. Such students have constructed second-level interiorized versions of plot-
not raised their counting process to the symbolic, sec- ting schemes and properly structured mental models
ond level of interiorization. The occurrence of the un- of frames of reference that allowed them to interrelate
necessary counting behavior on axes is also consistent locations and coordinates without running through
with the conjecture that such students possess an ac- the process of locating each point.
tion-oriented meaning for coordinates—coordinates
are considered inputs for a counting- or movement- Large- and Small-Scale Spaces
based procedure for locating an object. Such a pro- Mathematical coordinate systems function the
ceduralized conception of coordinates would be dif- same whether one is locating a point on a sheet of
ficult to apply to nondecade coordinates on grids with paper or oneself in a city. However, research in psy-
only multiples of ten labeled (a difficulty that these chology suggests that the cognitive processes used in
researchers observed). Another piece of evidence dealing with these situations differ, which might have
suggesting that students’ initial conceptualizations important implications for research on the learning
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 891

and teaching of coordinate systems. For instance, ac- [Level 5] subjects do not begin with a single mea-
cording to Siegel (1981), large-scale spaces (e.g., a surement only to discover the need for a second di-
university campus) cannot be perceived from a single mension after a period of trial-and-error. From the
outset, they realize the logical necessity to take both
viewpoint, and their structure is mentally constructed
dimensions into account and the measurements are
from a number of observations over time. In contrast,
straightaway coordinated so as to be at right angles
small-scale spaces (e.g., a sheet of paper) can be men- to one another. . . . [This] behavior is operational in
tally constructed from a single vantage point. One nature. Before the facts are given in actual experi-
acts in large-scale spaces, but on small-scale spaces. ence, they are engendered by structured anticipa-
Consequently, it might be difficult for students to tory schema. (p. 169)
transfer their thinking about mapping a classroom or
playground to coordinate systems shown on a page or Presumably it is at this level that relations of order
computer screen. and distance between objects are replaced by similar
In Siegel’s (1981) developmental model for the relations between the positions themselves. It is as if
formation of cognitive maps of large-scale spatial en- a space were emptied of objects so as to organize the
vironments, in the first phase, as one moves about in space itself. In fact, Piaget argued that
an environment, landmarks are noticed and remem-
bered. In the second phase, routes linking landmarks a reference frame is not simply a network composed
are formed (and with further experience metricized). of relations of order between the various objects them-
In the third phase, as coordinated frames of reference selves. It applies equally to positions within the network
are developed, routes are integrated into overall con- as to objects occupying any of these positions and en-
ables the relations between them to be maintained in-
figurational frameworks or survey maps.
variant, independent of potential displacement of the
In contrast, for small-scale spaces, Piaget’s work objects. (Piaget & Inhelder, 1956, p. 376)
on students’ construction of coordinate systems is still
relevant (Piaget, Inhelder, & Szeminska, 1960). For Despite differences in the mental development
instance, when students were asked to draw a point of large- and small-scale spaces, there are two major
on one rectangular sheet that was in exactly the same commonalities. First, coordination (of actions and
place as a point on another sheet, several levels were ob- objects) is crucial. Second, a frame of reference must
served. At Level 1, students used global visual estimates be mentally constructed, and actions and objects must
to locate the point; measuring devices were either not be properly integrated into it. Like so many processes
used or used inappropriately. At Level 2, students used in geometry, both the coordination process and the
only one measurement (say from a rectangle side or formation of frames of reference depend critically on
vertex); they were satisfied with one measurement be- creating appropriate mental models.
cause they could not coordinate two. At Level 3, stu-
dents recognized that two measurements are needed.
For instance, they used a single measurement from a
corner and tried to visually preserve the slope of the GEOMETRIC MEASUREMENT
line when using this measurement. Or, they made two
oblique measurements. At Level 4, students successful- The concept of measurement30 is woven throughout
ly used two perpendicular measurement dimensions the fabric of geometric conceptualization, reason-
to locate the point. This perpendicularity served as a ing, and application. From the historical beginnings
conceptual device that enabled students to coordinate of geometry as “earth measure” to Birkhoff’s axioms
the measurement dimensions. However, students’ use for Euclidean geometry, measurement plays a central
of two coordinated measurements was not yet interior- role in reasoning about all aspects of our spatial envi-
ized; they tended to discover this approach using trial ronment. Measurement is critical for understanding
and error instead of immediately using it. Finally, only the structure of shapes, using coordinate systems to
at Level 5, did students construct an interiorized math- determine locations in space, specifying transforma-
ematical coordinate system. tions, and establishing the size of objects. Geometric

30
I use the term geometric measurement in the broad and abstract mathematical sense to refer to the concept of assigning numbers to geometric
entities in accordance with a set of axioms. At the school level, understanding geometric measurement includes (a) conceptualizing that
numbers can be used to quantify the amount of an attribute (e.g., length, area, volume) contained in a geometric object by determining
the number of attribute-units that fit in the object and (b) being able to implement procedures for assigning measurements to objects
(e.g., iterating units, using a ruler, choosing appropriate measurement units). Thus, understanding measurement requires an integration
of procedural and conceptual knowledge.
892 ■ STUDENTS AND LEARNING

measurement is also embedded in the graphic repre- S: It’s 45 packages. And the way I found it is I
sentation of functions and algebraic equations. multiplied how many packages could fit in
However, despite the importance of geomet- the height by the number in the width, which
ric measurement, students’ performance on mea- is 3 times 3 equals 9. Then I took that and
surement tasks is alarmingly low (Kloosterman et multiplied it by the length, which is 5, and
al., 2004; Martin & Strutchens, 2000; Sowder et al., came up with 9 times 5, which is 45.
2004). For instance, according to NAEP results, in Obs: How do you know that is the right answer?
the year 2000, less than 25% of 4th graders and only S: Because the equation for the volume of a box
about 60% of 8th graders were able to determine the is length times width times height.
length of an object pictured above a ruler with its Obs: Do you know why that equation works?
end not aligned to that of the ruler (Kloosterman S: Because you are covering all three dimen-
et al., 2004; Sowder et al., 2004). On a similar item sions, I think. I’m not really sure. I just know
in 1996, the percentages of Grade 4, 8, and 12 stu- the equation.
dents answering this item correctly were 22, 63, and
83, respectively. Thus, almost 1 in 5 seniors in high Because of an inappropriate connection between
her spatial structuring and the numerical procedure,
school seemed to lack understanding of measuring
this student did not understand that the mathematical
length with a ruler. Performance on area and surface
formula she applied was inappropriate for this prob-
area was worse. In 2000, only about 14% of Grade 8
lem—a common problem for students. Indeed, only
students could determine the number of square tiles
38% of the students in her geometry class answered
it takes to cover a region of given dimensions, and the item correctly, despite the fact that all of them had
only about 25% could determine the surface area scored at or above the 95th percentile in mathematics
of a rectangular solid (Sowder et al., 2004). In 1996, on a widely used standardized mathematics test in 5th
only 35% of 12th graders and 27% of 8th graders rec- grade. Battista also found that only 19% of Grade 7
ognized and could explain why a square, and a right and 8 students in a range of classes in the same school,
triangle with the same height and twice the width of but not including this geometry class, answered this
the square, have the same area (Martin & Strutchens, problem correctly (1999b), and that less than 10% of
2000). And on volume, in 1990, only 55% of 12th Grade 3–5 students correctly solved a similar problem
graders and 41% of 8th graders knew that a measure- in which the box was only one layer high (1998b). Si-
ment of 48 cubic inches for a rectangular box repre- mon and Blume (1994) reported preservice elemen-
sented volume (NAEP web site). tary teachers making the same error in an analogous
Strongly implicated in this low performance is area situation. Similarly, in a study by Reynolds and
the research finding that for many students—perhaps Wheatley (1996), fourth grader Kristin was trying to
the majority—there is a basic disconnection between determine the number of 3-by-5 cards needed to cov-
spatial and measure-based numerical reasoning (Bar- er a 15-by-30 rectangle. To solve the problem, she di-
rett & Clements, 2003; Battista, 2001a; Clements et al., vided 15 into 450. However, the researchers found no
evidence that Kristin realized that this quotient would
1997). That is, many students do not properly main-
be correct only if the two dimensions of the small card
tain the connection between numerical measure-
divided, respectively, the two dimensions of the large
ments and the process of unit-measure iteration. For
rectangle. In fact, Kristin seemed suspicious of her
instance, students who incorrectly measure the length
procedure, saying that she needed to verify it by draw-
of an object when one of its ends is not aligned with
ing. Thus, Kristin was struggling to connect the spatial
“zero” on a ruler do not clearly conceptualize how the and numeric aspects of this problem.
ruler’s numerical markings indicate the iteration of In general, students working on these types of
unit lengths. And most students who correctly use the nonstandard measurement problems must perform
formulas for the area of a rectangle or volume of a two critical processes: (a) they must construct a proper
right rectangular prism in standard problem contexts, spatial structuring of the situation; (b) they must co-
neither understand why the formulas work nor apply ordinate their spatial structuring with an appropriate
the formulas appropriately in nonstandard contexts. numerical scheme. Too often, students skip the first
For example, a bright eighth grader who was 3 weeks process and proceed directly to the second. Also, even
from completing a standard course in high school when students recognize that they must perform the
geometry—so she was 2 years ahead of schedule for first process, they often have difficulty doing so. That
college prep students—responded as follows on the is, because many traditional curricula prematurely
problem shown in Figure 19.16 (Battista, 1998b). teach numerical procedures for geometric measure-
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 893

Collin has some packages that each contain two identical cubes. He wants to know how many of these packages
it takes to completely fill the rectangular box below.

1 cube packages made from 2 cubes

box

Collin knows that he can fit 3 packages He knows that he can fit 3 packages He knows that he can fit 5 packages
along the height of the box. along the width of the box. along the length of the box

Figure 19.16 Package problem.

ment, students have little opportunity to think about

the appropriateness of the numerical procedures THE DEVELOPMENT OF STUDENTS’
they apply, and they have insufficient opportunities REASONING ABOUT LENGTH
to develop skill in spatially structuring arrays of mea-
surement units. In particular, even when students rec-
In many ways, research on linear measurement has
ognize that the volume formula is inappropriate for not answered fundamental questions that were asked
the Collin problem, they have significant difficulty 2 decades ago. To illustrate, in a still widely cited study,
constructing an appropriate spatial structuring of the Hiebert (1981) found that many students who failed
packages because they have had so few opportunities Piagetian tasks of conservation and transitivity of
to develop their structuring skills. length apparently still “learned” measurement strate-
Research also shows that students commonly in- gies. He argued that although
terchange measurement units or computational pro-
cedures (Chappell & Thompson, 1999; Nunes, Light, conservation and transitivity are logical prerequisites
& Mason, 1993; Pesek & Kirshner, 2000; Woodward for completing many measurement tasks, children
& Byrd, 1983), and that most students have difficulty do not seem to use this knowledge when they solve
understanding how the multiplication of two length the tasks. . . . They use an intermediate measurement
to compare two lengths and do not think to ask the
units produces units of area (Battista, 2004; Kordaki
transitivity question; they move a unit to measure the
& Potari, 1998; Nunes et al., 1993). Overall, then, the re- length of an object and do not worry about whether
search indicates that many students’ measurement reasoning the length is being conserved. Simple skills or tech-
is superficial, with poorly understood procedures or formulas niques apparently allow children to bypass the logical
substituting for deep understanding. In fact, the tradition- structure of many measurement tasks. (p. 208)
al premature instructional focus on computational
formulas seems to interfere with students’ concept Hiebert conjectured, however, that students who
development in geometric measurement. possess a particular Piagetian reasoning ability might
894 ■ STUDENTS AND LEARNING

learn related measurement concepts with a deeper expect that measuring consists in counting, but they
level of understanding than those who do not pos- are not consistent in their attention to measuring”
sess this ability—that, for example, students who do (p. 55). At Level 2b, students integrate the iteration
not conserve length learn certain measuring strate- of a single unit of length into a sequence of units,
gies only at a rote level. To partially test this hypoth- but restricted to a single direction. So students might
esis, Hiebert made a crooked path with rods and correctly iterate a unit along the top of a rectangle
asked students to construct a straight path that had but lose track of the unit as they move to an adjacent
just as far to walk on it (the rods used for the straight but perpendicular side. At Level 3, students correctly
path were different in length than those used in the iterate units around complex multidirectional paths,
crooked path). All children failed this task before in- and they operate on units of units and reason about
struction. However, whereas 13 of 16 of the students measurements in the absence of perceptual objects.
who passed both Piagetian tasks were at least partial- Number schemes and spatial schemes are properly
ly successful on this task after instruction, only 2 of coordinated.
16 of the students who failed the Piagetian tasks were Battista (2003a, in press c) attempted to inte-
successful. Thus, it seems that conservation, or more grate previous research and develop a broader char-
likely, the mental operations that enable conserva- acterization of students’ construction of meaning
tion, are necessary for a conceptually sound comple- for length and length measurement. According to
tion of this task. Consistent with Hiebert’s hypoth-
this characterization, there are two fundamentally
esis that simple techniques may substitute for deep
different types of reasoning about length. Nonmea-
understanding, Joram, Subrahmanyam, and Gelman
surement reasoning does not use numbers. Instead it
(1998) claimed that even though young children can
involves visual-spatial inferences based on direct or
learn some simple measurement skills, actual physi-
indirect comparisons, imagined transformations,
cal measurement such as determining lengths with
or geometric properties. Measurement reasoning in-
paper clips is challenging for many middle-school
volves unit-length iteration, that is, determining the
students. They also reported poor performance on
number of fixed unit lengths that fit end-to-end along
linear measurement estimation even for high school
the object, with no gaps or overlaps. Measurement
students and adults. Thus, both old and new research
reasoning includes not only the process of measur-
suggests that students’ knowledge of linear measure-
ing, but reasoning about numerical measurements
ment may be more superficial than it appears in most
(e.g., adding lengths to find the perimeter of a poly-
common assessments.
gon, making inferences about length measurements
based on properties of figures). Although students
Levels of Sophistication in Students’ typically develop nonmeasurement strategies before
Reasoning About Length measurement strategies, nonmeasurement reason-
and Length Measure ing continues to develop in sophistication even after
measurement reasoning appears. Furthermore, the
Several groups of researchers have attempted to
most sophisticated reasoning about length involves
describe levels of sophistication in students’ develop-
the integration of nonmeasurement and measure-
ment of the concepts of length and length measure-
ment reasoning.
ment. Because each group has focused on different
aspects of length and used different sets of tasks, com-
parisons are difficult but also revealing. Nonmeasurement Reasoning
Clements et al. (1997) observed three levels of Nonmeasurement Level 0. Students’ reasoning
sophistication in students’ reasoning about path about length is appearance-based and holistic. They
lengths, which were later extended to four levels by attend to how things look, focusing on whole shapes
Barrett and Clements (2003). At Barrett and Clem- rather than systematically on parts within shapes. They
ents’ Level 1, students use gross visual comparison of often use direct and indirect comparison.
objects. There is a disconnection between numerical Nonmeasurement Level 1. Students systematically
measurement and the extent of line segments; stu- use decomposing/recomposing to compare lengths. At
dents often estimate or guess measurements. At Lev- first, students rearrange (physically, by drawing, or in
el 2a, students connect numbers to iterative move- imagination) some or all path pieces, and directly and
ments along a path but often have difficulty keeping visually compare the rearranged paths as wholes. Later,
track of their unit iterations. “Children at this stage students compare two paths by matching, one-by-one,
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 895

pairs of pieces that they think are the same length— Stephan, McClain, & Gravemeijer, 2001; Piaget, In-
they do not transform one path into another. helder, & Szeminska, 1960). For instance, Joram et
Nonmeasurement Level 2. Students compare path al. (1998) cited research indicating that, in making
lengths by sliding, turning, and flipping shape parts in linear measurement estimates, iterating a unit is the
ways that allow them to infer, based on shape and motion most commonly reported strategy. In one study, 81%
properties, that one transformed shape is congruent to of elementary students reported using unit iteration
another. in estimation, with about one third of these students
physically marking off units, whereas the other two
Measurement Reasoning thirds were observed marking off units with their eyes.
Measurement Level 0. Students use counting to find Furthermore, consistent with the unit-counting strat-
lengths; however, their counting does not represent the egy hypothesis, Joram et al. cited one study in which
iteration of a fixed unit length. For instance, students the time it took adults to estimate the length of line
might recite numbers as they continuously move their segments increased linearly with the length of the seg-
fingers along a path. Or they might count dots along ments and another study in which the times taken to
a path but not as true indicators of unit lengths. (See estimate the number of 1-cm units in a 6-cm segment
Sarama et al., 2003, for similar findings.) were the same as those to estimate the number of 2-
Measurement Level 1. Students attempt to iterate cm segments in a 12-cm segment.
what they consider to be a unit length along an object But how exactly do the conceptualizations of
or path. However, because students (a) do not fully un- length and length measurement develop? What are
derstand what a length unit is, or (b) do not properly the basic cognitive operations required? Joram et al.
coordinate iterated units with each other, their itera- (1998) argued that the mind mentally represents
tions contain gaps, overlaps, or different length units, counting numbers as magnitudes using a continu-
and are incorrect. ous accumulator mechanism (like putting water into
Measurement Level 2. Students not only un- a beaker). This mechanism includes a bidirectional
derstand what a length unit is, when iterating unit- mapping between verbal/written number symbols
lengths, they properly coordinate the position of each and accumulated mental magnitudes. In this view,
unit with the position of the unit that precedes it so counting a set of 5 objects activates the correspond-
that gaps, overlaps, and variations in unit lengths are ing magnitude on a mental number line (via the ac-
eliminated. cumulator), and, inversely, when a quantity on the
Measurement Level 3. Students determine some mental number line is activated, a portion of the
length measurements without explicitly iterating ev- mental number line is subdivided and intervals as-
ery unit length. That is, after iterating some length signed appropriate numbers.
units, students operate on the results of their itera- Joram et al. (1998) discussed how this counting
tions numerically or logically to find other lengths. process can be applied to linear measurement esti-
Measurement Level 4. Students numerically and mation and how the process explains why linear es-
inferentially operate on length measurements without timation is more difficult than counting. According
iterating any unit lengths. They make complex, proper- to the accumulator model, the mental representa-
ty-based visual inferences about measurements using tion activated when counting a set of discrete items
properties of geometric shapes and transformations. is a continuous magnitude. During object counting,
At this highest level of measurement reasoning, stu- this continuous magnitude is made discrete by single
dents fully integrate and apply the processes from bursts “of impulses to the accumulator for each item
Nonmeasurement Level 2 with their measurement in the set” (p. 422). The perceptual discreteness of the
reasoning. counted objects provides the signals for these bursts.
Joram et al. hypothesized that what makes linear mea-
surement estimation more difficult than counting is
Cognitive Processes for the
that, when estimating, there are no perceptual cues
Conceptualization of Length Measurement
to trigger the bursts. Such burst activation requires an
Research suggests that students construct mean- individual to mentally segment the linear object into
ingful understanding of length measurement as units, which requires (a) recalling and maintaining an
they abstract and reflect on the process of iterating image of the length unit, (b) visualizing the iteration
unit lengths (Barrett & Clements, 2003; Carpenter of the unit along the object to be estimated, (c) keep-
& Lewis, 1976; Hiebert, 1981; Kamii, 1995; Cobb, ing track of where the last unit ended and where the
896 ■ STUDENTS AND LEARNING

next one must begin, and (d) maintaining a running marize and interpret. According to this theory, indi-
count of units while performing tasks a–c. Further viduals abstract continuous experiential items as they
complicating matters, the burst interval (interval size) move a finger along an item or scan the item with
changes, depending on the length unit employed their eyes. Because continuous motion is contained
(e.g., inches, feet). in the abstraction of the object, the perceived object
Given that both linear extent and Joram et is experienced as continuous. Furthermore, because
al.’s (1998) hypothesized mental representation of of the way that an individual becomes aware of a con-
counting are based on continuous mental models, tinuous path—the perceptual abstraction contains its
it is interesting that linear measurement estimation endpoints, the motion that produces the path, and the
is more difficult to master than counting (Barrett & duration of the motion—the individual also becomes
Clements, 2003; Piaget et al., 1960). In some sense, aware of the path’s linear extent. “The property of
it would seem that the fact that both length and the a segment that we call length is an awareness of the
counting-accumulator are continuous would make scanning action over an image of the segment along
linear quantification easier than discrete quantifi- with an awareness of the duration of the scanning ac-
cation. And perhaps, at some level, it is. That is, in tion” (p. 3). Extending this analysis, for a segment to
some contexts linear extent seems easier to compre- be used as a unit in iteration, it must be abstracted at
hend than counting (thus, for example, the wide- an even higher level. The scanning motion used to
spread use of bar graphs). However, Joram et al.’s perceive the unit length must be interiorized, then ap-
burst-theory explanation suggests that difficulties in plied in steps a–d above.
applying counting to measuring length reside in the
transition from discrete to continuous quantities;
Cognitive Processes for Conserving Length
individuals must establish an isomorphism between
the discrete counting system and linear quantities. What kinds of cognitive processing are required
The episode below illustrates a student having dif- to conserve length? In dealing with conservation of
ficulty with this isomorphism (Battista, in progress; length tasks, Piaget et al. (1960) found that students
cf., also Clements et al., 1997). at the lowest levels (1 and 2a) of sophistication on
In this episode, when asked how she knew the rect- length conservation failed to simultaneously account
angle she drew was 40 units around, JAK (grade 2) drew for both ends of the sticks, which the researchers saw
irregularly spaced dots along its inside edge, stopping as an indication that the students were not attending
when her count reached 40 (see Figure 19.17). There to the intervals of length between the endpoints. So
was no indication that JAK used dots as indicators of the objects were not compared as continuous entities.
fixed unit lengths. She was unable to properly connect Perhaps the reasoning used by students at this level is
the continuous quantity that was the perimeter with the based on abstraction of experiences with comparisons
discrete items of the verbal counting sequence. in which only one set of endpoints are examined, for
instance, comparing who is taller when two people
are standing next to each other. (Although the logic
of comparing single endpoints in this context seems
obvious to adults, it is complex and may not be at all
clear to young children.)
At the next highest level (2b), coordination of
parts is clearly evident—student responses indicate
“the beginning of a relationship between the two
paired extremities (i.e., the four extremities [of the
two sticks] taken in pairs” (Piaget et al., 1960, p. 100).
This relationship arises as a consequence of students’
abstraction and comparison of two scanning motions,
one for the extent of each stick. In discussing this level,
Piaget et al. made an important observation about op-
erational reasoning. They stated that some students at
Figure 19.17 JAK’s rectangle. Stage 2b are close to operational reasoning (which they
claimed happens at Stage 3) but not quite there. For
A somewhat different description of the process instance, one student noted that the sticks were equal
by which the mind constructs length and length units when aligned, but when the sticks were moved out of
is given by Steffe (in press), which I will liberally sum- alignment, he had to realign them to convince himself
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 897

that they were still equal. Piaget et al. identified this rea- tems is understanding how to meaningfully enumerate
soning as empirical, not operationally reversible. rectangular 2d and 3d arrays of squares and cubes,
Level 3 is achieved only when students see con- respectively. Previous research studies have described
servation of length as logically necessary. Piaget et al. separate levels of sophistication in students’ enumer-
(1960) claimed that this happens when objects are ating squares and cubes in such arrays (Battista, 1999;
viewed in terms of a reference system “which provides Battista & Clements, 1996; Battista et al., 1998). More
a common medium for all objects” (p. 103). Within recently, Battista (2003b, 2004) has integrated the sep-
this medium exists a system of sites that exist indepen- arate models into a more general model for area and
dent of positional changes because they are fixed in volume, which is described below.
relation to an invariant backdrop: “When an object
undergoes a change of position the empty sites which Underlying Mental Processes
have previously been occupied are equivalent to sites
which were previously empty and are now filled, so Battista claims that five basic cognitive processes
that the overall length of the object remains constant” are essential for meaningful enumeration of arrays
(p. 103). That is, students coordinate the scanning of squares and cubes: abstraction, forming and using
motions performed on objects within a system of pos- mental models, spatial structuring, units-locating, and
sible scannings, which is the “space” in which these organizing-by-composites (the first three processes
motions exist (which suggests that the scannings have have already been discussed).32 The units-locating
reached the second level of interiorization). process locates squares and cubes by coordinating
In contrast to the Piagetian view, a simpler expla- their locations along the dimensions that frame an
nation of the mechanism for length conservation is array. For instance, to understand the location of a
possible. As children repeatedly compare abstracted square in a rectangular array, an individual must “see”
scanning motions for nonaligned equal-length ob- the square in terms of a two-dimensional coordinate-
jects by placing them in appropriate end-to-end align- like system—for example, it is in the fourth column
ments, they abstract a general reversible alignment and the second row, or it is the fourth unit to the right
process. This process, coupled with exposure to adults and the second unit down. The organizing-by-com-
who interpret the process as indicating that the length posites process combines an array’s basic spatial units
of an object does not change when it moves, builds (squares or cubes) into more complicated composite
the reasoning structure that supports conservation.31 units33 that can be repeated or iterated to generate
the whole array. For instance, in a 2d array, a student
might mentally unite the squares in a row to form a
spatial composite unit that can be iterated in the di-
THE DEVELOPMENT OF STUDENTS’ rection of a column to generate the array. Students
REASONING ABOUT AREA AND VOLUME can meaningfully enumerate arrays of squares and
cubes only if they have developed properly structured
Genuine understanding of area/volume measure- mental models that enable them to correctly locate
ment requires comprehending (a) what the attribute and organize the squares and cubes.
of area/volume is and how it behaves (i.e., conserv-
ing it as it is moved about and decomposed/recom-
Levels of Sophistication in Students’
posed), (b) how area/volume is measured by iterating
Structuring and Enumeration of Arrays
units of area/volume, (c) how numerical processes
can be used to determine area/volume measures for Level 1: Absence of Units-Locating and Organizing-
special classes of shapes, and (d) how these numerical by-Composites Processes
processes are represented with words and algebra. Students do not organize units into spatial com-
A core idea in developing competence with mea- posites, and, because they do not properly coordinate
suring area and volume in standard measurement sys- spatial information, their mental models are insuffi-

31
Objects actually change length as they are moved from one temperature to another (which was a major problem in establishing stable
units of measure in human history). So, when we say that length is preserved, we are implicitly utilizing one particular simplified model
of how the world works. Thus, although it might seem that abstracting conservation of length is empirically based, it also depends on
social consensus.
32
See Battista (2003b, 2004) for further details. Also, see (Wheatley, 1997) for additional examples of students’ difficulties with these
processes in 3d situations.
33
A composite unit is a cognitive entity (an abstraction) that results from mentally uniting and taking as one thing a collection of perceived
or re-presented objects.
898 ■ STUDENTS AND LEARNING

cient to locate all the units in arrays. Enumeration of Level 7: Numerical Procedures Connected to Spatial
squares or cubes seems almost random. The “double- Structurings, Generalization
counting” error is ubiquitous. Students’ spatial structuring and enumeration
schemes reach a level of abstraction at which they can
Level 2: Beginning Use of the Units-Locating and the be reflected on and analyzed, thus enabling students
Organizing-by-Composites Processes to explicitly understand the connection between an
enumeration strategy and the spatial structuring on
Students not only start to spatially structure arrays in
which it is based. Students’ mental models incorpo-
terms of composite units, their emerging development
rate row-by-column or layer structuring that is abstract
of the units-locating process produces mental models
and general enough to apply to situations in which
sufficient for them to recognize equivalent composites.
the basic units are not cubes (e.g., packages made
For instance, after counting the cubes visible on one from two cubes; e.g., see Figure 19.16).
side of a building, a student infers the number of cubes
on the opposite side. Additional Reasoning
In developing sophisticated reasoning about area
Level 3: Units-Locating Process Sufficiently Coordinated and volume measurement, there are additional types
to Eliminate Double-Counting of reasoning that students must attain, beyond the lev-
Students’ units-locating process coordinates sin- els listed above. For instance, students must generalize
gle-dimension views (e.g., top, side) into a mental the reasoning so that they can reason about fractional
model that is sufficient to recognize the same unit units in area and volume measurement, a move that
from different views, enabling students to eliminate can be difficult for students (Kordaki & Potari, 2002).
double-counting errors. However, coordination is still Students must also generalize their thinking to curved
insufficient to build a mental model that properly lo- or irregular shapes in which it is not easy to visual-
cates interior squares or cubes. ize square or cube tilings; in a sense, students must
be able to think about deformable units. Students
must also generalize their thinking from the small
Level 4: Use of Maximal Composites, But Insufficient
scale spaces in which area/volume are initially stud-
Coordination for Iteration
ied to large scale spaces such as rooms and buildings.
Students structure arrays in terms of maximal Also, consider the ability to spatially structure the de-
composites (rows or columns for area; layers for vol- composition of, say, a rectangular prism into smaller,
ume). But, due to insufficient coordination, the stu- non-square units—as might be needed in carpentry
dent cannot precisely locate these composites, instead (Millroy, 1992). Conceptually, such reasoning may be
estimating their locations. available at Level 7, but spatially, such reasoning can
be extremely demanding and is unlikely to be learned
Level 5: Use of Units-Locating Process Sufficient in traditional school settings.
to Correctly Locate All Units, But Less-Than-Maximal Researchers must carefully investigate students’ at-
Composites Employed tainment of these additional types of reasoning. How,
Students’ mental models correctly locate all for example, are these types of reasoning related to
squares/cubes in an array. However, although stu- the seven levels already described? Students must also
dents sometimes get correct answers, because they be able to interrelate area and volume. For instance,
inefficiently organize arrays into composites, they fre- advanced thinking about volume (e.g., that needed in
calculus) depends on understanding a multiplicative
quently lose their place in enumeration. Furthermore,
relationship between the height of an object and its
students’ structuring and enumeration strategies are
cross-sectional area. How do students make the transi-
not generalizable and are inadequate for large arrays.
tion from a conceptualization based on iterating lay-
ers of cubes to this more sophisticated idea?
Level 6: Complete Development and Coordination
of Both the Units-Locating and the Organizing-by-
Composites Processes Additional Research
Students’ interiorized mental models fully incor- Transitions to Abstract Tools
porate a row-by-column or layer structuring so that The research seems pretty clear that students
they can accurately reflect on and enumerate an array need appropriate preparation before learning about
without perceptual or concrete material for the indi- the use of standard measuring tools such as rulers and
vidual units within composites. the use of formulas. Students need to develop under-
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 899

standing of iteration of appropriate units of length, Understanding, Making, and Drawing Tilings
area, and volume. However, research has not yet clear- A number of researchers have investigated stu-
ly described how this transition takes place. Although dents’ understanding of unit-area tilings of larger
Battista has described levels of sophistication in the shapes and the plane, especially as revealed by stu-
development of students’ reasoning about length, dents’ drawings. For instance, Outhred & Mitchel-
area, and volume, other than for volume, he has not more (2000) found several levels of sophistication in
followed students during extended periods of instruc- elementary students’ drawings of unit-square cover-
tion. More longitudinal research is needed. ings of rectangles. At Level 0, the unit squares drawn
by students did not completely cover rectangles with-
The Connection Between Linear and Area Measurement out gaps or overlaps. At Level 1, unit squares com-
Several studies have indicated that students have pletely covered rectangles without overlap, but the
difficulty properly relating and separating the concepts organization was unsystematic; units varied consider-
of length, area, and volume. For instance, preservice ably in size and shape. At Level 2, drawings of unit
and inservice teachers frequently believe that whenev- squares exhibited correct array structure, with equal
er the perimeter of a figure increases, so does its area numbers of units in rows, and in columns, but the
(Ma, 1999; Tierney, Boyd, & Davis, 1990). Students and size of the units was determined visually, not from the
teachers also often believe that when the lengths of dimensions of the rectangle. At Level 3, rectangles’
the sides of a square or cube double, so does the area dimensions were used to iterate rows; one dimension
and volume (e.g., Tierney, Boyd, & Davis). Nunes et al. was used to find the number of units in a row, while
(1993) found that when 9–10-year-olds were asked to the other dimension was used to determine the num-
compare the amount of paint needed to cover two rect- ber of rows. At Level 4, students did not need to draw
angles, far more students answered correctly when they units; the numbers of units in rows and columns were
used physical squares than when they used rulers. In used to calculate the total number of squares.
fact, 26 out of 29 students who used rulers either added Investigating students’ tilings of different kinds
the measures of the sides or did not know what do to do of shapes (not just squares) has shed additional light
with the rulers to make an area comparison. on the cognitive processes underlying tiling. As a case
At an even more fundamental level, Outhred and in point, Owens and Outhred (1997) suggested that,
Mitchelmore (2000) conjectured that some elemen- to be successful, students have to understand the ge-
tary students do not relate the number of squares in ometry of the tile and how it affects gaps and overlaps
the rows and columns of a rectangle to the lengths in tiling. Students also have to understand the struc-
of the sides. Corroboration for this conjecture comes ture of the pattern of the tiling. Owens and Outhred
from Battista (1996) who found that some fifth grad- found that many students were unable to see or draw
ers, who correctly enumerated the number of cubes appropriate tilings. Wheatley and Reynolds (1996)
in a box when different types of pictures depicting argued that drawing a tessellation of a shape requires
boxes and cubes were shown, were unable to de- students to develop an image of the shape that can
termine the number of cubic centimeters in an un- be used in drawing it then developing a plan for cov-
marked box when they were given a ruler (students ering the plane. Wheatley and Reynolds found that
were shown a cubic centimeter and told that it mea- some students drew shapes one at a time, seemingly
sured 1 cm on each edge). However, this idea has not without understanding the overall structure of the til-
yet been fully and carefully investigated. For instance, ing. Other students used an overall structure, includ-
Outhred and Mitchelmore’s students’ errors may ing the use of composite units, to guide their activity.
have been caused by a lack of spatial structuring of Thus, tiling a shape requires two critical components.
the squares in a rectangular array, or by deficiency in Component 1 consists of constructing a mental mod-
linear measurement. Battista’s students presumably el of the shape that (a) incorporates critical features
were able to structure 3d arrays for cubes correctly, of the shape’s geometry and (b) can be mentally
but he did not separately assess their understanding manipulated. (Some students are able to visualize
of linear measurement, so there is no way of know- exactly where a tiled shape should be placed, while
ing if the cause of the problem for the students could other students must use trial and error with physical
be attributed to their deficiencies in conceptualizing materials.) Component 2 consists of developing an
length. In fact, given the difficulty that many students appropriately structured mental model of the array of
have with the concept of length measurement, it is shapes in the tiling.
not surprising that some of them struggle with build- Research by Clements, Wilson, and Sarama
ing on this concept and properly relating it to area (2004) that describes levels of sophistication in young
and volume measurement. children’s ability to make a larger shape with pat-
900 ■ STUDENTS AND LEARNING

tern blocks helps elaborate these components. For appropriate structure. Furthermore, “the attempt to
instance, in Clements et al.’s scheme of seven levels, represent a covering in a drawing can help children
at the second level, children employ trial and error to examine their experience in new ways and lead to
to make simple given shapes, have limited ability to new insights” (p. 165).
use turns or flips to help them make these shapes, An episode from Battista et al. (1998) corrobo-
view shapes only as wholes, and see few geometric re- rates and extends Outhred and Mitchelmore’s claim
lationships between shapes. At the third level, shapes about material squares pre-structuring arrays. CS was
are chosen using a gestalt configuration or one com- asked to predict how many squares would cover the
ponent such as side length, vertices may be matched rectangle shown in Figure 19.18a (making her predic-
but without a concept of angle size, and turning and tion without drawing). CS pointed and counted as in
flipping are used, usually by trial and error. At the Figure 19.18b, predicting 30. So clearly she was having
fourth level, children combine shapes with growing difficulty properly structuring the array. When check-
intentionality and anticipation, angles are matched by ing her answer with plastic squares, she correctly cov-
size, and turning and flipping are used with anticipa- ered the rectangle with squares. So, consistent with
tion to select and place shapes. Thus, in relation to Outhred and Mitchelmore’s claim, the placement of
Component 1, students gradually come to use the ge- the squares was structured by the physical materials.
ometry of shapes, moving from holistic to parts-based However, CS initially enumerated her correctly placed
reasoning, attending to side length before angle mea- squares as shown in Figure 19.18c, getting 30, then
sure, and using motion first physically then in imag- counted the squares again, first getting 24, then 27.
ery, a progression that harkens back to the van Hiele So, although CS was able to correctly place physical
levels and the previous discussion of mental models. squares to cover the rectangle, she was still unable to
The fact that Clements et al.’s last three levels deal structure the squares so that they could be correctly
with increasing sophistication in the use of composite enumerated—CS had not mentally constructed the
units is consistent with the notion that students de- structuring contained in the physical materials.
velop Component 2, structuring the tiling array, after Let me now address Outhred and Mitchelmore’s
Component 1. claim that drawing tilings can help students examine
Also related to Component 2, Outhred and Mitch- their experience in new ways, leading to new insights
elmore (2000) made an important observation about into the structure of an array. This claim is consistent
the difference between drawing and using physical with the theory that spatial structuring is a matter of
shapes to tile the inside of a larger shape. In the case abstracting one’s actions on an array, not directly ab-
of squares tiling the inside of a rectangle, Outhred stracting the array itself. That is, structuring is a re-
and Mitchelmore argued that because of the way that flective not empirical abstraction. However, to gain
wooden squares fit together, using wooden squares additional insights from one’s drawing, more than
to cover a rectangle “prestructures” the array—“the abstraction of one’s structuring actions is required.
array structure is inherent in the materials and does For example, typical of young students, CS not only
not need to be apprehended by the learner” (p. 146). drew an inadequately structured covering as shown in
(They also cited research suggesting that such use of Figure 19.18d, she failed to recognize the inadequacy
squares may not be effective instructionally.) When of her structuring (Battista et al., 1998). Thus, draw-
drawing, however, students must explicitly create an ing may not promote better structuring than placing

1 2 3 4 5 6 14 13 12 11 10 9

19 18 17
16 27 28 20 21 22 23 8
29 7 15
2026
30
24 16 19 28 29 24 7
15 25 30 8
21 22 23 18 27 26 25
17
14 1 2 3 4 5 6
13 12 11 10 9

(a) (b) (c)

(d)

Figure 19.18 CS’ structuring of arrays of squares (Battista et al., 1998).

 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 901

Figure 19.19 Stage 1 of the process of abstracting unit iterations.

physical squares unless students recognize difficulties what students abstract during these experiences may
in their drawings, creating cognitive perturbations. not match the formal concept. However, during these
With proper instructional guidance, such perturba- or subsequent experiences, students might notice in-
tions can occur and cause refinements in students’ adequacies in their conceptualizations, causing per-
structuring. That is, making drawings and checking turbations to occur. These perturbations may cause
drawings using physical squares can cause students them to refocus their attention and abstract conceptu-
to reflect on and refine the structuring contained in alizations of the attribute that are close enough to the
their mental models of arrays. In fact, an analogous formal attribute that social interactions or problem-
“predict and check” approach was shown to be very solving contexts do not cause further perturbations.
effective in promoting the development of students’
structuring and enumeration of 3d arrays of cubes Abstracting Unit Iterations
(Battista, 1999). This approach is effective because it The process of abstracting unit iterations suffi-
focuses students’ attention and analysis, not on arrays, cient to measure various geometric attributes is ex-
but on their mental models of arrays. tremely complex. But, based on current research, we
might conjecture the following stages in the abstrac-
tion process.34
SYNTHESIS Stage 1, incomplete abstraction of unit iterations.
Because units and iterations are abstracted at most at
the internalized level, iterations are uncoordinated
Cognitive Processes Underlying Geometric and have gaps, overlaps, and nonequivalence of units.
Measurement: The Many Faces of The student work shown in Figure 19.19 illustrates
Abstraction this level for the contexts of length, area, volume, and
two-dimensional tiling.
Several kinds of abstractions are critical in geo- Stage 2, units interiorized and coordinated. To
metric measurement. meaningfully and correctly perform unit iteration,
students must, at the very least, develop an interior-
Abstracting the Attribute ized mental model of the unit. This mental model in-
First, students must abstract the attribute to be corporates relevant elements of the geometry of the
measured (say, length), distinguishing it from all the unit and permits coordinated mental manipulation of
other spatial attributes possessed by an object. This the unit so that gaps, overlaps, and nonequivalence
might happen through reflection initiated by (a) are eliminated in the iteration process.
dealing with various kinds of problematic situations Stage 3, structure of iterations interiorized. Students
(e.g., comparing objects, fitting an object through an develop interiorized mental models of the structure of unit
opening) or (b) interacting with people who already iterations that make it possible (a) to see a particular unit
possess the concept and draw attention to it. Initially, in relation to the iterated sequence of units, enabling

34
Each step described below must be properly coordinated with numerical operations.
902 ■ STUDENTS AND LEARNING

students to understand a unit’s location; (b) to struc- Final Comments on Measurement

ture iterations in terms of composite units; and (c) to
establish a part-whole relationship between the iterated Although students’ measurement difficulties are
units and the whole by reconstituting the whole as the alarming because measure is essential for most real-
sequence of iterations of the unit (Barrett & Clements, world applications of geometry, these difficulties may
2003; Steffe, in press; see also Cobb et al. 2001).35,36 represent only the tip of a huge learning-difficulty ice-
Stage 4, numerical measurements become sym- berg—an iceberg that research has not yet investigat-
bols. Mental models of structured unit iterations are ed. Indeed, given the pervasiveness of measure in geo-
raised to the second level of interiorization so that metric and graphical contexts, poor understanding of
the resulting enumerations become symbols. That is, measure might be a major cause of learning problems
numbers, construed as measurements, act as symbols for numerous advanced mathematical concepts.37 For
for the unactualized process of iteration. This enables instance, students’ learning of high school geometry
students to meaningfully reason about measurements and analytic geometry is endangered by poor under-
without iterating units. standing of geometric measure, especially by misun-
derstandings of length. Furthermore, given students’
Integrating Perspectives on Length, Area, disconnection between numerical measurements and
and Volume Measurement the process of unit iteration, researchers must ques-
tion the depth of students’ understanding of coordi-
Because individual research studies usually have nate systems, function graphing, locus problems, and
focused on only one type of geometric measurement at theorems about lengths or distances.
a time (length, area, or volume), research has not yet
Specific examples of these endangered topics
produced a comprehensive theory of geometric mea-
abound: (a) Comprehending common proofs of the
surement. Battista (2003b, 2004) integrated models for
the development of area and volume measurement but Pythagorean theorem requires understanding length
did not extend this integrated model to length. How- and area and the distinction between the two. (b)
ever, comparing Battista’s levels for length to those for Hollebrands (2003) found high school students un-
area/volume suggests directions and issues for integra- able to generalize their conceptualizations of linear
tion. In particular, the question of how nonmeasure- magnitude in a way that enabled them to construct
ment and measurement reasoning are related must the notions of vectors and translations defined by vec-
be addressed. That is, Battista’s levels for area/volume tors. (c) Almost half of a set of very advanced eighth
address only one component of understanding these graders and more than 80% of above average sev-
concepts—enumerating units. Nonmeasurement ideas enth and eighth graders had great difficulty concep-
are not addressed, nor is the critical relationship be- tualizing common geometric shapes (rectangles and
tween nonmeasurement and measurement reasoning. squares) in terms of measurement-based properties
A first attempt at developing a general theory of geo- (Battista, 1998a). (d) A number of researchers have
metric measurement is shown in Table 19.1 (which, documented students’ difficulties in understanding
at this point, can serve only as a conjecture). In this the connection between linear equations and their
theory, nonmeasurement and measurement reason-
graphs (Dugdale, 1993; Knuth, 2000; Moschkovich,
ing develop in parallel (though not simultaneously).
Schoenfeld, & Arcavi, 1993). Knuth, for instance,
However, in contrast to this parallel-development hy-
found that students failed to recognize that the coor-
pothesis, Kordaki and Potari (2002) argued that first,
students should spatially explore areas; second, they dinates of graph points used to calculate the slope of
should physically or pictorially transform and compare a line are solutions to the equation of that line, a find-
areas without the use of numbers; third, they should ing that could be directly traceable to an insufficiently
use measurement units; and finally they should use developed conceptualization of how coordinates are
area formulas. Much more research is needed to tease related to lengths. That is, students might see coordi-
out the interrelationships between nonmeasurement nates not as lengths but as markers on a map (just like
and measurement reasoning. margin letters on road maps).

35
It is unlikely, however, that a–c occur at the same time.
36
Structuring unit-length iterations increases in complexity when paths are situated in 2d space. For instance, Barrett and Clements (2003)
found that maintaining unit lengths was more difficult along nonstraight than straight paths.
37
The notion of poor understanding of geometric measure endangering later mathematics learning was developed in collaboration with
Jack Smith.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 903

Table 19.1 Reasoning About Measurable Geometric Quantities

Nonmeasurement Reasoning Measurement Reasoning

1. Holistic Visual Comparison of Shapes 1. Use of Numbers Unconnected to Unit Iteration

Students’ reasoning is appearance-based and holistic. It is Students use counting to determine measures; however, their counting
appearance-based because students focus strictly on the does not represent the iteration of an appropriate unit.
appearance, or “how things look,” of objects. It is holistic because
students focus on whole shapes or objects, not parts within shapes. 2. Unit Iteration and Enumeration
Students’ strategies are imprecise and often vague. Conservation Students attempt to iterate and enumerate appropriate units of
and transitivity gradually develop. measure to fill the shape being measured. At first, because students do
not properly coordinate these units with either each other or the whole
2. Visual Comparison of Shapes by Decomposing/ object being measured, their iterations contain gaps, overlaps, or
Recomposing different units—students often lose track of units while enumerating.
Students decompose and recompose spatial regions into smaller Gradually, students overcome these difficulties.
regions either physically or mentally, then compare the recomposed
regions directly. There is no thought about enumeration of units or Cognitive Milestones
measured quantities, so no measurement reasoning occurs. When • Units properly maintained, coordinated, and located
decomposing shapes, students might compare recomposed whole • Units properly organized into composites
shapes, or they might compare shapes region-by-region. • Iterations properly structured
3. Comparison of Shapes by Property-Preserving • Operating on iterations
Transformations/Decompositions • Iteration schemes generalized to fractional units
Students compare shapes by decomposing them then sliding, • Iteration schemes generalized to non-segment/-square/-cube
turning, or flipping shape parts in a way that allows them to infer, units
based on shape or transformation properties, that one transformed
shape is congruent to another. Students do not necessarily explicitly 3. Operating on Numerical Measurements
mention geometric properties (congruence of shapes, segments, Students numerically operate on measurements without iterating units.
angles) or transformations (slide, flip, or turn), but their reasoning At this level, iteration seems to have fallen in the background. It is as if
is consistent with such properties and transformations. iteration is taken as being done.

4. Integrated Measurement and Nonmeasurement Reasoning

Students inferentially operate on measurements without iterating units. They can make complex, property-based visual inferences about
measurements, which are often accomplished with transformations or by using properties of geometric shapes. At this highest level of
measurement reasoning, students fully integrate and apply the processes from Nonmeasurement Level 3 with their measurement reasoning.
The difference between Nonmeasurement Level 3 and Measurement Level 4 is that in level 4 students make inferences about numerical
measurements, not shapes. For instance, students can develop, meaningfully use, and justify procedures or formulas for finding the areas/
volumes of nonrectangular shapes.

reasoning. Because of the major research foci in this

CONCLUSIONS domain, the research has been both heavily phenom-
enological—how students construct meaning for geo-
Despite geometry’s importance in mathematical the- metric concepts—as well as cognitive—what cognitive
ory and application, students continue to have diffi- processes are necessary for students’ construction of
culty learning it with genuine depth. There is much geometric meaning. Far fewer studies have examined
confusion and conceptually shallow, procedural learn- geometry learning from a sociocultural perspective.
ing. The research described in this chapter discusses This perspective would focus on how the geometric
in great detail many student difficulties, often sug- meanings students construct are affected by their par-
gesting avenues for instructional improvements. But ticipation in various communities in which certain
much additional research is needed that delves even mathematical practices, interaction patterns, concep-
more deeply into students’ difficulties and carefully tions, beliefs, and reasoning are shared.
examines the effects and effectiveness of alternate in- The focus of geometry research on cognitive rather
structional approaches. than social processes has been both good and bad for
the field. On the positive side, the cognitive approach
has produced significant understandings of the devel-
Moving Beyond the Cognitive Perspective
opment of students’ geometric reasoning. Although
This chapter has discussed a wide range of re- there is still much to understand about the cognitive
search relevant to understanding students’ geometric processes that underlie geometric reasoning, the cog-
904 ■ STUDENTS AND LEARNING

nitive approach not only is critical for adequately inves- cation? First, many geometry computer environments
tigating these processes, it helps connect the field to are interesting to students. For instance, elementary
important related fields such as psychology. It is critical students often get really excited when first exposed
that this approach continue. However, on the negative to Logo or the GSP Shape Makers. But the pedagogical
side, not enough has been done to understand socio- and research interest in these environments goes well
cultural and affective factors in geometry learning. beyond motivational factors. Environments such as
Researchers do not know nearly enough about how Cabri and GSP are tools that enhance the process of
various components of social practice affect students’ doing geometry for everyone, not just students. That
construction of geometric concepts and reasoning. For is, DGEs allow individuals to explore geometric ideas
instance, what kinds of classroom practice effectively in ways that are very different, and arguably better,
encourage students to personally embrace (rather than explorations with paper and pencil. DGEs signifi-
than mimic) increasingly sophisticated types of justifi- cantly extend the ability to examine large sets of pre-
cation in geometry, and what social processes enable cisely drawn examples and to infuse dynamic motion
such progress? (See Borrow, 2000 and Quinn, 1997 for into investigations. So it is no wonder that researchers
discussions of this issue that combine social and cog- spend so much time investigating the use of this new
nitive perspectives.) We also need additional research and powerful tool in instruction. And, as has been dis-
on how affective factors are interrelated with geometric cussed earlier in this chapter, this research is getting
learning. For instance, how do students “feel” about be- progressively more sophisticated and providing major
ing required to give formal proofs when they under- insights in geometry learning and teaching.
stand neither the need for such proofs nor their logic, However, researchers still need to better under-
and how do these feelings affect students’ cognition stand how these technological enhancements affect
and learning? Of course, in all but classes dedicated students’ learning. For instance, even though numer-
specifically to geometry, classroom practices generally ous studies show that DGEs can encourage and sup-
transcend specific mathematical topics; so investigating port significant learning in students, we do not know
some of these issues from a geometric perspective may the effects of scaling up this use, and we do not know
be difficult. However, such investigations are possible how, in very precise terms, learning in DGEs differs
(e.g., Quinn, 1997). from learning in paper-and-pencil environments.
Another sociocultural aspect that has been ne- Thus we need more comparison studies, both quanti-
glected, but not ignored (e.g., Battista, 1999), in- tative, to investigate generality, and qualitative, to in-
cludes ways that face-to-face interactions in collabora- vestigate differences in cognitive processes.
tive small-group work affect geometry learning. For
instance, how do students make sense of ideas com-
REFERENCES
municated by other students when those ideas are
(a) discrepant from their own views or (b) consistent Arzarello, F., Micheletti, C., Olivero, F., Paola, D., & Gallirio,
with their views but expressed in different language or G. (1998, July). Dragging in Cabri and modalities of transition
reasoning? In class discussions and small-group work, from conjectures to proofs in geometry. Paper presented at the
22nd PME Conference. Stellenbosch, South Africa.
how do vague and inadequate verbal descriptions
Barrett, J. E., & Clements, D. H. (2003). Quantifying length:
of spatial ideas (which are inevitable) affect student Fourth-grade children’s developing abstractions for
learning and participation in the classroom culture? measures of linear quantity. Cognition and Instruction,
What are the social dynamics of small-group work and 21(4), 475–520.
how do they affect students’ geometric thinking and Battista, M. T. (1990). Spatial visualization and gender
knowledge construction? How does classroom dis- differences in high school geometry. Journal for Research in
course affect students’ geometry learning? Thus, al- Mathematics Education, 21, 47–60.
though the current cognitive approach to geometry Battista, M. T. (1994). On Greeno’s environmental/model view
of conceptual domains: A spatial/geometric perspective.
research should continue, numerous important ques- Journal for Research in Mathematics Education, 25(1), 86–94.
tions should be investigated from other perspectives. Battista, M. T. (1996). Unpublished raw data.
Battista, M. T. (1998a). Shape Makers: Developing geometric
reasoning with the Geometer’s Sketchpad. Berkeley, CA: Key
The Attraction of Computer Environments Curriculum Press.
A great deal of modern research on geometry Battista, M. T. (1998b). How many blocks? Mathematics Teaching
in the Middles Grades, 3(6), 404–411.
learning and teaching has focused on the use of com-
Battista, M. T. (1999a). Fifth graders’ enumeration of cubes
puter environments. Why have geometry teaching in 3D arrays: Conceptual progress in an inquiry-based
and research embraced technology more enthusiasti- classroom. Journal for Research in Mathematics Education,
cally perhaps than any other area of mathematics edu- 30(4), 417–448.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 905

Battista, M. T. (1999b, April). Mathematics Education Reform at Carpenter, T. P., & Lewis, R. (1976). The development of the
Lafayette Middle School. Paper presented at the annual concept of a standard unit of measure in young children.
meeting of the American Educational Research Association, Journal for Research in Mathematics Education, 7(1), 53–58.
Montreal, Canada. Cassirer, E. (1944). The concept of group and the theory of
Battista, M. T. (2001a). A research-based perspective on teaching perception. Philosophy and Phenomenological Research, 5(1),
school geometry. In J. Brophy (Ed.), Advances in research 1–35.
on teaching: Subject-specific instructional methods and activities Chappell, M. F., & Thompson, D. R. (1999). Perimeter or area?
(pp. 145–185). New York: JAI Press. Which measure is it? Mathematics Teaching in the Middle
Battista, M. T. (2001b). Shape Makers: A computer environment School, 5(1), 20–23.
that engenders students’ construction of geometric ideas Chazan, D. (1993a). High school students’ justification for
and reasoning. Computers in the Schools, 17(1), 105–120. their views of empirical evidence and mathematical proof.
Battista, M. T. (2003a, July). Levels of sophistication in elementary Educational Studies in Mathematics, 24, 359–387.
students’ reasoning about length. Paper presented at the 27th Chazan, D. (1993b). Instructional implications of students’
annual conference of the International Group for the understandings of the difference between empirical
Psychology of Mathematics Education, Honolulu, HI. verification and mathematical proof. In J. L. Schwartz, M.
Battista, M. T. (2003b). Understanding students’ thinking Yersushalmy & B. Wilson (Eds.), The Geometric Supposer:
about area and volume measurement. In D. H. Clements What is it a case of? (pp. 107–116). Hillsdale, NJ: Erlbaum.
(Ed.), 2003 Yearbook: Learning and teaching measurement (pp. Chazan, D., & Yerushalmy, M. (1998). Charting a course for
122–142). Reston, VA: National Council of Teachers of secondary geometry. In R. Lehrer & D. Chazan (Eds.),
Mathematics. Designing learning environments for developing understanding
Battista, M. T. (2004). Applying cognition-based assessment to of geometry and space (pp. 67–90). Mahwah, NJ: Erlbaum.
elementary school students’ development of understanding Churchland, P. S. (2002). Brain-Wise: Studies in neurophilosophy.
of area and volume measurement. Mathematical Thinking Cambridge, MA: MIT Press.
and Learning, 6(2), 185–204. Clements, D. H. (1999). Teaching length measurement: Research
Battista, M. T. (in press a). Representations and cognitive challenges. School Science and Mathematics, 99(1), 5–11.
objects in modern school geometry. In K. Heid & G. Blume Clements, D. H., & Battista, M. T. (1992). Geometry and spatial
(Eds.), Research on technology in the learning and teaching reasoning. In D. A. Grouws (Ed.), Handbook of research
of mathematics: Syntheses and perspectives. Greenwich, CT: on mathematics teaching and learning (pp. 420–464). New
Information Age Publishing Inc. York, NY: National Council of Teachers of Mathematics/
Battista, M. T. (in press b). Development of the Shape Makers Macmillan Publishing Co.
geometry microworld: Design principles and research. Clements, D. H., & Battista, M. T. (2001). Logo and Geometry.
In K. Heid & G. Blume (Eds.), Research on technology in the Journal for Research in Mathematics Education Monograph.
learning and teaching of mathematics: Syntheses and perspectives. Reston, VA: National Council of Teachers of Mathematics.
Greenwich, CT: Information Age Publishing Inc.
Clements, D. H., Battista, M. T., Sarama, J., Swaminathan, S.,
Battista, M. T. (in press c). A cognitive perspective on formative & McMillen, S. (1997). Students’ development of length
assessment. New England Mathematics Journal. concepts in a Logo-based unit on geometric paths. Journal
Battista, M. T., & Clements, D. H. (1991). Logo geometry. for Research in Mathematics Education, 28(1), 70–95.
Morristown, NJ: Silver Burdett & Ginn. Clements, D. H., Wilson, D. C., & Sarama, J. (2004). Young
Battista, M. T., & Clements, D. H. (1995). Geometry and proof. children’s composition of geometric figures: A learning
Mathematics Teacher, 88(1), 48–54. trajectory. Mathematical Thinking and Learning, 6(2), 163–184.
Battista, M. T., & Clements, D. H. (1996). Students’ understanding Cobb, P., Stephan, M., McClain, K., & Gravemeijer, K. (2001).
of three-dimensional rectangular arrays of cubes. Journal for Participating in classroom mathematical practices. The
Research in Mathematics Education, 27(3), 258–292. Journal of the Learning Sciences, 10(1 & 2), 113–163.
Battista, M. T., Clements, D. H., Arnoff, J., Battista, K., & Van Coren, S., Ward, L. M., & Enns, J. T. (1994). Sensation and
Auken Borrow, C. (1998). Students’ spatial structuring perception. Fort Worth, TX: Harcourt Brace College.
of 2D arrays of squares. Journal for Research in Mathematics de Villiers, M. D. (1994). The role and function of hierarchical
Education, 29(5), 503–532. classification of quadrilaterals. For the Learning of
Becker, H. J. (2000, July). Findings from the Teaching, Learning, and Mathematics, 14(1), 11–18.
Computing Survery: Is Larry Cuban Right? Paper presented de Villiers, M. (1998). An alternative approach to proof in
at the 2000 School Technology Leadership Conference of dynamic geometry. In R. Lehrer & D. Chazan (Eds.),
Council of Chief State School Officers, Washington, D. C. Designing learning environments for developing understanding
Berthelot, R., & Salin, M. H. (1998). The role of pupil’s spatial of geometry and space (pp. 369–394). Mahwah, NJ: Erlbaum.
knowledge in the elementary teaching of geometry. In C. Denis, L. P. (1987). Relationships between stages of cognitive
Mammana & V. Villani (Eds.), Perspectives on the teaching of development and van Hiele level of geometric thought
geometry for the 21st century (pp. 71–78). Dordrecht: Kluwer. among Puerto Rican adolescents. Dissertation Abstracts
Borrow, C. (2000). An investigation of the development of 6th grade International, 48, 859A.
students’ geometric reasoning and conceptualizations of geometric Dorfler, W. (1993). Computer use and views of the mind. In
polygons in a computer microworld. Unpublished doctoral C. Keitel & K. Ruthven (Eds.), Learning from computers:
dissertation, Kent State University. Mathematics education and technology (pp. 161–186).
Calvin, W. H. (1996). How brains think. New York, NY: Basic Grenoble Cedex, France: NATO ASI Series, Computer and
Books. Systems Sciences.
906 ■ STUDENTS AND LEARNING

Driver, J., & Baylis, G. C. (1989). Movement and visual attention: technological environment. Journal of Mathematical Behavior,
The spotlight metaphor breaks down. Journal of Experimental 22, 55–72.
Psychology: Human Perception and Performance, 15, 448–456. Holzl, R., Healy, L., Hoyles, C., & Noss, R. (1994). Geometrical
Dugdale, S. (1993). Functions and graphs: Perspectives on relationships and dependencies in Cabri. Micromath, 10(3),
students’ thinking. In T. A. Romberg, E. Fennema & T. 8–11.
P. Carpenter (Eds.), Integrating research on the graphical Hoyles, C., & Healy, L. (1997). Unfolding meanings for reflective
representation of functions (pp. 101–130). Hillsdale, NJ: symmetry. International Journal of Computers for Mathematical
Erlbaum. Learning, 2, 27–59.
Edelman, G. M. (1992). Bright air, brilliant fire. New York, NY: Hoyles, C., & Jones, K. (1998). Proof in dynamic geometry
Basic Books. contexts. In C. Mammana & V. Villani (Eds.), Perspectives
English, L. D., & Halford, G. S. (1995). Mathematics education: on the teaching of geometry for the 21st century (pp. 121–128).
Models and processes. Mahwah, NJ: Erlbaum. Dordrecht, The Netherlands: Kluwer.
Gelman, S. A., & Diesendruck, G. (1999). What’s in a concept? Jaime, A., & Gutiérrez, A. (1989, July). The learning of plane
Context, variability, and psychological essentialism. In isometries from the viewpoint of the van hiele model. Paper
I. E. Siegel (Ed.), Theoretical perspectives in the concept of presented at the 13th PME Conference. Paris.
representation (pp. 87–111). Mahwah, NJ: Erlbaum. Jaime, A., & Gutiérrez, A. (1995). Guidelines for teaching plane
Gentner, D., & Medina, J. (1998). Similarity and the development isometries in secondary school. Mathematics Teacher, 88(7),
of rules. Cognition, 65(2/3), 263–297. 591–597.
Goldenberg, E. P., & Cuoco, A. A. (1998). What is dynamic
Johnson-Gentile, K., Clements, D. H., & Battista, M. T. (1994).
geometry? In R. Lehrer & D. Chazan (Eds.), Designing
The effects of computer and noncomputer environment
learning environments for developing understanding of geometry
on students’ conceptualizations of geometric motions.
and space (pp. 351–367). Mahwah, NJ: Erlbaum.
Journal of Educational Computing Research, 11, 121–140.
Goldin, G. A., & Kaput, J. J. (1996). A joint perspective on the
Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive
idea of representation in learning and doing mathematics.
science of language, inference, and consciousness. Cambridge,
In L. P. Steffe, P. Nesher, P. Cobb, G. A. Goldin & B. Greer
MA.: Harvard University Press.
(Eds.), Theories of mathematical learning (pp. 397–430).
Mahwah, NJ: Erlbaum. Johnson-Laird, P. N. (1998). Imagery, visualization, and thinking.
In J. Hochberg (Ed.), Perception and cognition at century’s end
Greeno, J. G. (1991). Number sense as situated knowing in a
(pp. 441–467). San Diego, CA: Academic Press.
conceptual domain. Journal for Research in Mathematics
Education, 22(3), 170–218. Jones, K. (2000). Providing a foundation for deductive reasoning:
Gruber, H. E., & Voneche, J. J. (Eds.). (1977). The essential Piaget. Students’ interpretations when using dynamic geometry
New York: Basic Books. software and their evolving mathematical explanations.
Gutiérrez, A. (1992). Exploring the links between van Hiele levels Educational Studies in Mathematics, 44(1/2), 55–85.
and 3-dimensional geometry. Structural Topology, 18, 31–48. Joram, E., Subrahmanyam, K., & Gelman, R. (1998).
Gutiérrez, A., & Jaime, A. (1998). On the assessment of the Measurement estimation: Learning to map the route
van Hiele levels of reasoning. Focus on Learning Problems in from number to quantity and back. Review of Educational
Mathematics, 20(2/3), 27–47. Research, 68(4), 413–449.
Gutiérrez, A., Jaime, A., & Fortuny, J. M. (1991). An alternative Kamii, C. (1995). Why is the use of a ruler so hard? Paper presented
paradigm to evaluate the acquisition of the van Hiele levels. at the Proceedings of the seventeenth annual meeting of
Journal for Research in Mathematics Education, 22(3), 237–251. the North American chapter of the international group
Harel, G., & Sowder, L. (1998). Students’ proof schemes: Results Psychology in Mathematics Education, Columbus, OH.
from exploratory studies. CBMS Issues in Mathematics Kloosterman, P., Warfield, J., Wearne, D., Koc, Y., Martin, G.,
Education, 7, 234–283. & Strutchens, M. (2004). Knowledge of mathematics and
Healy, L., & Hoyles, C. (1998). Technical report on the nationwide perceptions of learning mathematics of fourth-grade
survey: Justifying and proving in school mathematics. London: students. In P. Klooosterman & F. K. Lester (Eds.), Results
Institute of Education, University of London. and interpretations of the 2003 mathematics assessment of the
Healy, L., & Hoyles, C. (2001). Software tools for geometrical National Assessment of Educational Progress. Reston, VA:
problem solving: Potentials and pitfalls. International National Council of Teachers of Mathematics.
Journal of Computers for Mathematical Learning, 6, 235–256. Knauff, M., Fangmeier, T., Ruff, C. C., & Johnson-Laird, P.
Hershkowitz, R., Dreyfus, T., Ben-Zvi, D., Friedlander, A., Hadas, N. (2003). Reasoning, models, and images: Behavioral
N., Resnick, T., et al. (2002). Mathematics curriculum measures and cortical activity. Journal of Cognitive
development for computerized environments: A designer– Neuroscience, 15(4), 559–573.
researcher–teacher–learner activity. In L. D. English (Ed.), Knuth, E. J. (2000). Student understanding of the Cartesian
Handbook of international research in mathematics education connection: An exploratory study. Journal for Research in
(pp. 657–694). Mahwah, NJ: Erlbaum. Mathematics Education, 31(4), 500–508.
Hiebert, J. (1981). Cognitive development and learning linear Koedinger, K. R. (1998). Conjecturing and argumentation
measurement. Journal for Research in Mathematics Education, in high-school geometry students. In R. Lehrer & D.
12, 197–211. Chazan (Eds.), Designing learning environments for developing
Hoffer, A. R. (1988). Geometry and visual thinking. In T. R. understanding of geometry and space (pp. 319–347). Mahwah,
Post (Ed.), Teaching mathematics in grades K–8: Research based NJ: Erlbaum.
methods (pp. 232–261). Boston: Allyn and Bacon. Kordaki, M., & Potari, D. (1998). Children’s approaches to
Hollebrands, K. F. (2003). High school students’ understandings area measurement through different contexts. Journal of
of geometric transformations in the context of a Mathematical Behavior, 17(3), 303–316.
 THE DEVELOPMENT OF GEOMETRIC AND SPATIAL THINKING ■ 907

Kordaki, M., & Potari, D. (2002). The effect of area Mason, M. M. (1989). Geometric understanding and misconceptions
measurement tools on student strategies: The role of a among gifted fourth through eighth graders, Unpublished
computer microworld. International Journal of Computers for doctoral dissertation, Northern Illinois University, IL.
Mathematical Learning, 7, 65–100. Mayberry, J. (1983). The van Hiele levels of geometric thought
Kosslyn, S. M. (1980). Image and mind. Cambridge, MA: Harvard in undergraduate preservice teachers. Journal for Research
University Press. in Mathematics Education, 14, 58–69.
Kosslyn, S. M. (1983). Ghosts in the mind’s machine. New York: W. McCoy, L. P. (1996). Computer-based mathematics learning.
W. Norton. Journal of Research on Computing in Education, 28, p. 438–460.
Kosslyn, S. M., Margolis, J. A., & Barrett, A. M. (1990). Age McCrone, S. M. S., & Martin, T. S. (2004). Assessing high school
differences in imagery abilities. Child Development, 61(4), students’ understanding of geometric proof. Canadian
995–1010. Journal of Science, Mathematics and Technology Education,
Kosslyn, S. M. (1988). Aspects of a cognitive neuroscience of 4(2), 221/240.
mental imagery. Science, 240(4859), 1621–1626. Millroy, W. L. (1992). An ethnographic study of the mathematical
Kosslyn, S. M. (1994). Image and brain. Cambridge, MA: MIT Press. ideas of a group of carpenters, Journal for Research in Mathematics
Education Monograph. Reston, VA: National Council of
Krascum, R. M., & Andrews, S. (1993). Feature-based versus
Teachers of Mathematics.
exemplar-based strategies in preschoolers’ category
Mitchelmore, M. C., & White, P. (2000). Development of angle
learning. Journal of Experimental Child Psychology, 56, 1–48.
concepts by progressive abstraction and generalisation.
Laborde, C. (1992). Solving problems in computer-based Educational Studies in Mathematics, 41(3), 209–238.
geometry environments: The influence of the features of the
Moise, E. E. (1963). Elementary geometry from an advanced
software. Zentralblatt fur didaktick der mathematik (International
standpoint. Reading, MA: Addison-Wesley.
Reviews on Mathematical Education), 4, 128–135.
Moschkovich, J., Schoenfeld, A. H., & Arcavi, A. (1993). Aspects of
Laborde, C. (1993). The computer as part of the learning
understanding: On multiple perspectives and representations
environment: The case of geometry. In C. Keitel & K.
of linear relations and connections among them. In T. A.
Ruthven (Eds.), Learning from computers: Mathematics
Romberg, E. Fennema & T. P. Carpenter (Eds.), Integrating
education and technology (Vol. 121, 48–67). Grenoble Cedex,
research on the graphical representation of functions (pp. 41–68).
France: NATO ASI Series, Computer and Systems Sciences.
Hillsdale, NJ: Erlbaum.
Laborde, C. (1998). Visual phenomena in the teaching/ NAEP web site: http://nces.ed.gov/nationsreportcard/itmrls/.
learning of geometry in a computer-based environment. In
Noss, R., & Hoyles, C. (1996). Windows on mathematical meanings:
C. Mammana & V. Villani (Eds.), Perspectives on the teaching
Learning cultures and computers (Vol. 17). Dordrecht, The
of geometry for the 21st century (pp. 113–121). Dordrecht,
Netherlands: Kluwer Academic.
The Netherlands: Kluwer.
Nunes, T., Light, P., & Mason, J. (1993). Tools for thought: The
Laborde, C. (2001). Integration of technology in the design of
measurement of length and area. Learning and Instruction,
geometry tasks with Cabri-geometry. International Journal of
3, 39–54.
Computers for Mathematical Learning, 6(3), 283–317.
Olive, J. (1991). Logo programming and geometric understand-
Lehrer, R., Jenkins, M., & Osana, H. (1998). Longitudinal study ing: An in-depth study. Journal for Research in Mathematics
of children’s reasoning about space and geometry. In R. Education, 22(2), 90–111.
Lehrer & D. Chazan (Eds.), Designing learning environments
Outhred, L., & Mitchelmore, M. (2000). Yound children’s
for developing understanding of geometry and space (pp. 137–
intuitive understanding of rectangular area measurement.
167). Mahwah, NJ: Erlbaum.
Journal for Research in Mathematics Education, 31(2), 144–167.
Lewellen, H. (1992). Description of van Hiele levels of geometric
Ornstein, R. (1991). The evolution of consciousness. New York:
development with respect to geometric motions. Unpublished
Touchstone.
doctoral dissertation, Kent State University, OH.
Owens, K., & Outhred, L. (1997). Early representations of tiling
Leyton, M. (1992). Symmetry, causality, mind. Cambridge, MA: areas. Proceedings of the 21st Conference of the International
MIT Press. Group for the Psychology in Mathematics Education, Helsinki,
Ma, L. (1999). Knowing and teaching elementary mathematics. 3, 312–319.
Mahwah, NJ: Erlbaum. Pegg, J. (1992, July). Students’ understanding of geometry:
Mariotti, M. A. (2001). Justifying and proving in the Cabri Theoretical perspectives. Paper presented at the 15th Annual
environment. International Journal of Computers for Conference of the Mathematics Education Research
Mathematical Learning, 6, 257–281. Group of Australasia, Sydney.
Markovits, H. (1993). The development of conditional Pegg, J., & Davey, G. (1998). Interpreting student understanding
reasoning: A Piagetian reformulation of mental models in geometry: A synthesis of two models. In R. Lehrer & D.
theory. Merrill-Palmer Quarterly, 39(1), 131–158. Chazan (Eds.), Designing learning environments for developing
Marrades, R., & Gutiérrez, A. (2000). Proofs produced by understanding of geometry and space (pp. 109–133). Mahwah,
secondary school students learning geometry in a dynamic NJ: Erlbaum.
computer environment. Educational Studies in Mathematics, Pegg, J., Gutiérrez, A., & Huerta, P. (1998). Assessing reasoning
44(1/2), 87–125. abilities in geometry. In C. Mammana & V. Villani (Eds.),
Martin, G., & Strutchens, M. (2000). Geometry and measurement. Perspectives on the teaching of geometry for the 21st century (pp.
In E. A. Silver & P. A. Kenney (Eds.), Results from the seventh 275–295). Dordrecht, The Netherlands: Kluwer.
mathematics assessment of the National Assessment of Educational Pesek, D. D., & Kirshner, D. (2000). Interference of instrumental
Progress (pp. 193–234). Reston, VA: National Council of instruction in subsequent relational learning. Journal for
Teachers of Mathematics. Research in Mathematics Education, 31(5), 524–540.
908 ■ STUDENTS AND LEARNING

Piaget, J. (1950). The psychology of intelligence. London: constructive odyssey. The Journal of Mathematical Behavior
Routledge. Monograph.
Piaget, J. (2001). Studies in reflecting abstraction. Philadelphia: Sutherland, R., Godwin, S., Olivero, F., & Peel, P. (2002). Design
Psychology Press. Initiatives for Learning: ICT and Geometry in the Primary
Piaget, J., & Inhelder, B. (1956). The child’s conception of space. School. BERA Conference, University of Exeter, UK.
London: Routledge and Kegan Paul. Tall, D. (1992). The transition to advanced mathematical
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The child’s thinking: Functions, limits, infinity, and proof. In D. A.
conception of geometry. New York: Routledge and Kegan Grouws (Ed.), Handbook of research on mathematics teaching
Paul/Basic Books, Inc. (pp. 495–511). Reston, VA: National Council of Teachers
Pinker, S. (1997). How the mind works. New York: W. W. Norton. of Mathematics/Macmillan.
Posner, M. I., & Raichle, M. E. (1994). Images of mind. New York: Tierney, C., Boyd, C., & Davis, G. (1990, July). Prospective primary
Scientific American Library. teachers’ conceptions of area. Paper presented at the 14th PME
Presmeg, N. C. (1997). Generalization using imagery in Conference, Mexico City, Mexico.
mathematics. In L. D. English (Ed.), Mathematical reasoning van Hiele, P. M. (1986). Structure and insight. Orlando, FL:
(pp. 299–312). Mahwah, NJ: Erlbaum. Academic Press.
Quinn, A. (1997). Justifications, argumentations, and sense making von Glasersfeld, E. (1982). Subitizing: The role of figural
of preservice elementary teachers in a constructivist mathematics patterns in the development of numerical concepts.
classroom. Unpublished doctoral dissertation, Kent State Archives de Psychologie, 50, 191–218.
University, OH. von Glasersfeld, E. (1991). Abstraction, re-presentation, and
Regan, D., & Hamstra, S. (1991). Shape discrimination for reflection: An interpretation of experience and Piaget’s
motion-defined and contrast-defined form: Squareness is approach. In L. P. Steffe (Ed.), Epistemological foundations
special. Perception, 20, 315–336. of mathematical experience (pp. 45–67). New York: Springer-
Reynolds, A. M., & Wheatley, G. H. (1996). Elementary students’ Verlag.
construction and coordination of units in an area setting. von Glasersfeld, E. (1995). Radical constructivism: A way of
Journal for Research in Mathematics Education, 27(5), 564–581. knowing and learning. London: The Falmer Press.
Ross, B. H. (1996). Category representations and the effects Vygotsky, L. (1986). Thought and language. Cambridge, MA: The
of interacting with instances. Journal of Experimental Child MIT Press.
Psychology: Learning, Memory and Cognition, 22(5), 1249–1265. Wallace, E. C., & West, S. F. (1992). Roads to geometry. Englewood
Sarama, J., Clements, D. H., Swaminathan, S., McMillen, Cliffs, NJ: Prentice Hall.
S., & González Gómez, R. M. (2003). Development of Wheatley, G. H. (1997). Reasoning with images in mathematical
mathematical concepts of two-dimensional space in activity. In L. D. English (Ed.), Mathematical reasoning (pp.
grid environments: An exploratory study. Cognition and 281–298). Mahwah, NJ: Erlbaum.
Instruction, 21(3), 285–324. Wheatley, G. H. (1998). Imagery and mathematics learning.
Siegel, A. W. (1981). The externalization of cognitive maps Focus on Learning Problems in Mathematics, 20(2/3), 65–77.
by children and adults: In search of ways to ask better Wheatley, G. H., & Reynolds, A. M. (1996). The construction of
questions. In L. S. Liben, A. H. Patterson & N. Newcombe abstract units in geometric and numeric settings. tiling the
(Eds.), Spatial representation and behavior across the lifespan plane. Educational Studies in Mathematics, 30, 67–83.
(pp. 167–194). New York: Academic Press. Wilson, B. (1993). The Geometric Supposer in the classroom.
Siegler, R. S. (1996). Emerging minds: The process of change in In J. L. Schwartz, M. Yerushalmy & B. Wilson (Eds.), The
children’s thinking. New York: Oxford University Press. Geometric Supposer: What is it a case of? (pp. 17–22). Hillsdale,
Simon, M. A., & Blume, G. W. (1994). Building and NJ: Erlbaum.
understanding multiplicative relationships: A study of Woodward, E., & Byrd, F. (1983). Area: Included topic, neglected
prospective elementary teachers. Journal for Research in concept. School Science and Mathematics, 83(4), 343–347.
Mathematics Education, 25(5), 472–494. Yerushalmy, M. (1993). Generalization in geometry. In J. L. Schwartz,
Smith, E. E. (1995). Concepts and categorization. In D. N. M. Yerushalmy & B. Wilson (Eds.), The Geometric Supposer: What
Osherson & L. R. Gleitman (Eds.), An invitation to cognitive is it a case of? (pp. 57–83). Hillsdale, NJ: Erlbaum.
science. Vol. 3, Thinking. Cambridge, MA: MIT Press. Yerushalmy, M., & Chazan, D. (1993). Overcoming visual
Smith, L. B. (1989). A model of perceptual classification in obstacles with the aid of the Supposer. In J. L. Schwartz,
children and adults. Psychological Review, 96(1), 125–144. M. Yersushalmy & B. Wilson (Eds.), The Geometric Supposer:
Sowder, J. T., Wearne, D., Martin, G., & Strutchens, M. (2004). What is it a case of? (pp. 25–56). Hillsdale, NJ: Erlbaum.
What grade 8 students know about mathematics: Changes
over a decade. In P. Kloosterman & F. Lester (Eds.), The
1990 through 2000 mathematics assessment of the National AUTHOR NOTE
Assessment of Educational Progress: Results and interpretations.
(pp. 105–143). Reston, VA: National Council of Teachers Time to prepare portions of this material was par-
of Mathematics. tially supported by the National Science Foundation
Steffe, L., & Cobb, P. (1988). Construction of arithmetical meanings under Grant Nos. ESI 0099047and 0352898. Opinions,
and strategies. New York: Springer-Verlag. findings, conclusions, or recommendations, however,
Steffe, L. P. (1998, April). Principles of design and use of TIMA
are those of the author and do not necessarily reflect
software. Paper presented at the NCTM Research Presession,
Washington, D.C. the views of the National Science Foundation.
Steffe, L. P. (in press). Articulation of the reorganization hypothesis. I would like to thank Angel Gutiérrez for his care-
In L. P. Steffe (Ed.) Children’s fractional knowledge: A ful and thoughtful reading of this manuscript.