International Journal of Linguistics

ISSN 1948-5425
2015, Vol. 7, No. 5

Critical Review of the Models of Reading

Comprehension with a Focus on Situation Models

Mohammad Davoudi (Corresponding author)

Assistant professor of TEFL, Department of English Language and Literature, Hakim
Sabzevari University, Sabzevar, Iran
E-mail: davoudi2100@gmail.com

Hamid Reza Hashemi Moghadam

Ph.D Student of TEFL, Department of English Language and Literature, Hakim
Sabzevari University, Sabzevar, Iran
E-mail: h.hashemimoghadam@gmail.com

Received: September 24, 2015 Accepted: October 12, 2015 Published: October 31, 2015
doi:10.5296/ijl.v7i5.8357 URL: http://dx.doi.org/10.5296/ijl.v7i5.8357

Abstract
Research directed at unveiling the complexities of reading skill burgeoned in conjunction
with the advancements in the field of psycholinguistics. Tremendous effort has been made to
make sense of the complex process of the underlying multi-faceted mechanisms of inference
generation during reading. In this respect, the situation models have recently gained ground.
In studies focused on reading skill, situation models are of top priority, because they
illuminate the interactions among different components including the process of information
network activation, strategic and conscious-based inference as well as the textual and
meta-textual representations in the readers’ memory. This paper reviews prevalent reading
models with a focus on the critical analysis of three foundational situation models including,
Event-Indexing, Construction-Integration and Structure Building models of comprehension.
Unlike the previous research, the orientation of this paper toward situation models is based on
their use in second language reading, which is simplified in tune and comprehensive in
content. Its implications are beneficial for the wide spectrum of SLA theorists, teachers and
students for the reading skill purposes.
Keywords: Situation models, Event-Indexing model, Construction-Integration model

172 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

1. Introduction
In the field of applied linguistics, reading skill, once subordinated to speaking and listening,
has gained momentum due to the multi-dimensional advancements in theory and practice in
recent decades (Lynch & Hudson, 1991; Groundson, 1991). As the result of paradigm shift in
psycholinguistics (Adams, 1990; Bernhardet, 1991; Grabe, 1988), reading is no longer treated
as a passive and receptive skill. Although for a novice reader, reading may seem as a linear,
predictable and clear-cut process, various reading comprehension models claiming to unravel
its complexities attest to the distressingly complex nature of this phenomenon. Consequently,
thus far, multitudes of models of reading skill have been proposed and conceived as the
refinement of the prior ones, in terms of components and their interrelationships.
The present review is a critical analysis of the mainstream models, starting from the rather
simple top down and bottom up to the complex multidimensional models of situation models.
In principal, this paper has struggled to address both advantages and disadvantages of each of
these models from psycholinguists perspectives.
2. Approaches to Reading Process
2.1 Top down Processing
Chronologically speaking, systematic analysis of reading comprehension started with
Goodman (1967) who offered one of the most cited models of reading skill entitled as
top-down or the conceptually driven processing approach. In his conceptualization, reading is
a psycholinguistic guessing game since the readers’ preconceptions and background
knowledge largely impact the lower-level processes such as orthographic and phonological
processing, as well as the word recognition skill. According to Goodman (1967), readers are
not merely the neutral and passive receivers of the information from the text rather, as
Davoudi (2005) asserts, their background knowledge and other interpretive skills
accompanied by the cognitive and metacognitive strategies foster the speed of the lower level
processing during the comprehension stage. For instance, when a linguist reads a text related
to his profession, he skips many terms in the text due to his background knowledge and this
does not distort his understanding from the text while for a non-professional person it is
impossible to skip any part of the text due to the lack of his background knowledge. This
theory which was largely endorsed by many scholars (Barnet, 1989; Carrell & Eisterhold,
1983; Eskey, 1986; Garnham, 1985, to name a few) maintained that the printed information is
just a preliminary step of reading process and there is still a more central component of
reading skill which is called background information. This information is formed as the
propositional, meta-textual and conceptual representations in the learners’ long-term memory.
This approach focuses on what readers add to the comprehension process. In top-down
processing, readers get the information from the text, and then contrast it with their world
knowledge in order to make sense of what is written. According to this model, readers bring
meaning to the text based on their experiential and interpretive prior knowledge. In the view
of Goodman, reading can be deemed as a psycholinguistic guessing game, entailing diverse
stages of textual processing, predicting the content, confirming the true predictions,
correcting the false predictions and finally terminating these processes (understanding the

173 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

meaning of the text). For instance, while a psychologist reads a text, he initially processes the
knowledge from the text which in turn activates his background knowledge. Based on the
background knowledge in the form of formal or content schemata, the reader predicts the
content of the texts. If this prediction is not true, he will correct it, and consequently the
reading process will be terminated. Some researchers (Clarke & Silberstein, 1977; Stanovich,
1986; Garnham, 1985; Rieben & Perfitti, 1991) interrogated the fundamental assumptions of
top-down processing. Studying eye movements during reading proved that even professional
readers focused on most of the words printed on the text. Similarly, studies conducted on
readers with notorious background knowledge but lower processing skill indicated that
despite the predictions of top-down approach, their performance was not as expected. In a
similar line, Grabe (2009) maintains that in top-down models, the comprehension process is
neither mechanical nor linear, but actively controlled by the reader. Thus, the main
mechanisms for the processing of the text are in the mind of readers. From this perspective,
readers recognize letters and words only to confirm their preconceptions with reference to the
meaning of the text. Accordingly, they can successfully decode a passage even if they do not
know the meaning of the new words within the text.
2.2 Bottom up Approach
Another approach toward reading skill was bottom-up processing. The proponents of
bottom-up processing (Stanovich, 1986; Garnham, 1985; Rieben & Perfitti, 1991) put
emphasis on the decisive role of the lower-level recognition skills. In their view, reading is a
hierarchical and step by step process, starting from the perception of single phonemes to
words, clauses, sentences and then the whole piece of discourse. For the adherents of this
strand of research, readers do not skip any part of the text during reading and a great portion
of reading is the outcome of unconscious processes emerging from the text, not the strategic
and conscious processing on the part of the reader. In essence, this approach, as Shahnazari &
Dabaghi (2014) implies, is data-driven and the role of the lower level recognition skills,
including orthographic, semantic, syntactic and phonological processing are crucial. For this
reason, the rapid recognition skill from phonemes to full sentences is of overriding
significance.
According to Barnet, 1989; Carrell & Eisterhold, 1983; Eskey, 1986; Carnham, 1985;
Iran-Nejad, 1987 among many others, during bottom-up processing, the reader takes a step by
step order to process the text and the processing of each component takes place independently;
therefore, it is not feasible to make use of higher-ordered reading skills such as making
inferences, consequently, the reader’s back-ground knowledge plays virtually a very limited
role in driving and interpreting the meaning of a text.
2.3 Interactive Model of Reading
Rumelhurt (1977) withholds that in the interactive model of reading comprehension, meaning
is not bound to the text alone; rather, it is the outcome of co-construction of the information
within the text and the readers’ interpretation. In practice, information picked up by the eyes
is registered visually, and then sent to the pattern synthesizer. At the same time, a wide array
of information about semantic, syntactic and pragmatic concepts is drawn up from the long

174 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

term memory into the working memory (all these happen when reading a text). During this
process, the reader is involved in driving the meaning of the text and making inferences
through constant and simultaneous interactions between the surface structure of the text and
the readers’ background knowledge. Finally, they integrate the activated information from
these two sources into a coherent discourse and derive the meaning of the text. This model, as
Rumelhart confirms, was a refinement over the previous simplistic and linear bottom-up and
top-down models in its orientation. Nevertheless, this model has not been able to propose
operational justifications for the process of unconscious inference making during reading
activity. In a similar line, Alderson (2000) contended that the results of his experimental
study on 25 subjects could not be accounted for by the principles of the interactive model.
Consequently, Alderson questioned the predictability power of this model.
2.4 Stanovichs’ (1980) Interactive-Compensatory Model
One of the central advantages of this model lies in its ability to justify the difference between
the skilled and unskilled readers. Stanovich (1980) model is founded upon the principle that
in reading process, when readers experience problems in one dimention of processing, their
skill in other dimensions can compensate for the possible flaws and deficiencies. To put it
more simply, a deficit in any other parts during reading results in a heavier reliance on other
knowledge sources regardless of their level in hierarchy. So top-down processing, for a
reader with low word recognition, but well at the knowledge of the text topic, may
compensate for this deficit. Grabe (1988) propounded that the interactive model cannot
identify the complex underlying mechanisms of reading, such as the how of the interrelation
between the information networks in the learners’ long- term memory.
3. The Need for Situation Models of Reading Skill
Despite the aforementioned research endeavors, there was still a wide gap to be filled in
explaining the multifaceted nature of reading by reading psychologists. Some psycholinguists
(Kintsch, 1988, 1998; Schank & Abelson, 1977; Johnson-Laird, 1983, Mckoon & Ratcliff,
1992; Zwaan & Radvansky, 1998) introduced situation model as a new reading theory which
is highly applicable to reveal some complexities in decoding the text. Kintsch and Van Djik
(1978) claimed that the comprehension process involves more than merely constructing a
mental representations of the text itself, comprehension is first and foremost the construction
of mental representation of what that text is about which is called situation model. Situation
model draws heavily on inference generation as a key process in language comprehension.
Thus, in this new approach toward comprehension, readers make different types of inferences,
namely bridging and elaborative inferences, while the latter are not based on the text but the
situational representation from the text. Their conceptualization of the comprehension skill
was also informed by the connectionist theories of language learning which confirmed the
simultaneous and parallel processing of semantic, syntactic, phonological, conceptual and
situational components during the interaction between the working memory and these
activated concepts by establishing nodes and links in the readers’ long term memory. The
adoption of the new terms in the analysis of reading comprehension portrayed the complexity
of this multifaceted process. Regarding the invaluable contribution of situation models to our

175 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

understanding from the text comprehension, the present review intends to analyze the most
important models of reading and their fundamental theories. It is clear that the results of this
study are of central concern for a vast array of people from learners to teachers, practitioners
and curricular developers.
3.1 The Event Indexing Model
Before elaborating on the principles of Event-Indexing model, it is necessary to consider
some prominent aspects of situation models which are pertinent to this concept. As
mentioned in the previous section, the comprehension of a text is not bound to text-based
representation. Theorists (Zwaan & Radvansky, 1998; Glenberg Meyer, & Linden, 1987;
Morrow, Brower, & Greenspan, 1989) believe that the construction of a representation of
what a text is about requires the construction of a situation model which is the mental
representation of a text contents. While reading a text, we unconsciously build a situational
imagination from that text. For instance, when we read a story about ancient Persia, we have
some imaginations and conceptualizations of the events, their time and place as well as the
relationship between them that form the infrastructure of comprehending the text.
Most texts are composed of multiple events. The model considers predicate or verb as the
central building block of the text for connecting the propositional information and
establishing coherence in the readers’ memory. In the Event-Indexing model, ‘learners
understanding from the text is influenced by their conscious and unconscious tendency to
connect events in terms of six important indexes including protagonists, causality, goal,
motivation, temporal and spatial relationship. The information included in the text act as
clues to the situational representation of the text in the mind of the learners. For instance,
according to Event-Indexing model, when one reads the sentence “Ali went to teacher to ask
for an extension”, after reading the word ‘Ali’, the reader builds a mental representation of a
male person, and by confronting to more linguistic cues in the text, the reader updates the
situational representation of the text (has more exact imagination from the events). In the
above sentence, when we continue reading, we discover that Ali is a student; so our
situational representation from Ali is updated (becomes more complete). This level of
conceptualization is called the integrated level of the situation model. The complete model is
accessed while the linguistic cues are finished and the reader has comprehended the text.
Based on the above description, the main goal of this model is to broaden the scope of
situation model; consequently, it is different from the other situation models including
Resonance model and Construction-Integration model. According to this model, the more
indexes the events in a proposition share, the higher is the possibility of storing them in the
long-term memory. As a proof, in the sentence “I went to class to see the professor” in
comparison with the sentence “I went to class to see chairs”, the possibility of establishing
connection between two events of going to class is more with visiting the professor than
seeing the chairs based on the goal index. It should be noted that, in this model, text-based
information as well as the background knowledge and situational representation build up a
highly intricate and complex relationship which become activated while they are triggered by
the contents of the short-term memory. According to Zewan & Madden (2004), in the

176 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

Event-Indexing model, the events, objects, people, goals and their relationship are considered
as the backbones of the text rather than words, phrases or sentences. Generally, situation
models theorists believe that comprehenders are influenced by different aspects of situations
which have been pointed out in the text; consequently, the structure of the text alone is not
what the readers comprehend from it. To put it differently, when we imagine ourselves in a
situational context, there are important indexes, consisting of time, goals, causation,
protagonists, and objects and space which play a crucial role in our conceptualization from
the linguistic input. To better understand the issue, we must elaborate on the indexes of time,
causality, space and protagonist and their importance for understanding a text.
3.2 Time
Our conception of time as an index is based on our real life experiences. We know that events
take place in a linear and chronological order. But it should be taken into account that in the
world of written or the spoken discourse, this might not be the case. For example, we can say:
“Before Ali wrote the article, he changed his idea about the title”. In this sentence, the act of
writing is reported first, even though it was the last of the two events that had occurred. If we
construct a situation model from this sentence, this sentence should be more difficult to
process than its chronological counterpart. (The same sentence but beginning with after). In
real life, events follow each other in a linear way; however, narratives can have temporal
discontinuity which means that they are not written as they are experienced or performed in
the real world. Therefore, while comprehenders read the sentences that violate the concept of
time continuity, their reading time decreases in comparison with the sentences that do not
violate time continuity. All other things being equal, events that happen just recently are more
accessible to us than events that happened while ago.
3.3 Goals and Causation
We might have experienced that in many cases, there are some unfinished works that make
our mind highly engaged. According to the Event-Indexing model, these types of events or
goals remain in our mind longer than the goals that have already been accomplished. For
example, passing courses is currently more active in the mind of the many hard-working
students than listening to music, because passing courses as a duty has not been fulfilled yet
by the learners. Thus, if a protagonist (the main character) has a goal that has not been
accomplished, that goal should be more accessible to the comprehender than a goal that was
just one logical relationship between the events in the world. Consider the following
examples:
1) When I went to university, I saw Hassan playing football.
2) When I went to university, I saw Hassan buying a ticket.
Studies indicate that the time for processing sentence 1 is longer because the reader is not
able to establish causal relationship between going to university and buying ticket.

3.4 People and Objects

177 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

In every narration, there is the possibility of having one or more protagonists in the text.
While one reads a text, one builds a situational representation from the protagonist in their
mind. Comprehenders are also quick to make inferences about protagonists, in their attempt
to construct a more complete situation model. Consider, for example, what happens after the
subjects read the sentence.
As soon as he got a job, he went to see his old parents.
With respect to the above-mentioned points in the Event-Indexing model as addressed by
some scholars (Zwaan, Langston, & Graesser, 1995; Sanford & Garrod, 1981),
comprehenders parse clauses of text into events. During comprehension, they connect these
events based on the five different situational dimensions: time, space, causation, motivation,
and protagonist. If the event that is currently processed overlaps with the events in the
working memory on a particular dimension, then a link between these events is established
and stored in the long-term memory. An overlap between the sentences is determined based
on whether or not the two events share an index (time, place, protagonist, cause or goal).
It should be taken into account that the more situational indexes shared between the current
model and the integrated model, the easier the updating will be. It must be noted that there
are three basic assumptions behind the Event-Indexing model:
1. Events are the central units of situation models.
2. Events can be linked on five dimensions.
3. Events are related or not related on a particular dimension.
It is explicit that the Event-Indexing model has made inspirational contribution to our
realization of the situation model and a wide range of inferences made during reading
narrative texts. Although the term ‘situational representation’ was introduced prior to Zwaan
by Kintsch (1983), it expanded the scope of comprehension theories by drawing on the six
concepts of (time, causation, time, protagonist, goal and motivation) to better indicate the
relationship among the events in the text. Considering the verb as the building block for the
arguments overlaps (despite Kintsch’s model) for making relationships among different
events in the text, was another major contribution of this model to the field of
psycholinguistics. However, one of the serious flaws of this model, according to Gernbacher
(1990), is its limited focus on the metacognitive strategies as well as the retrieval-based and
text-based inferences of the readers for comprehension. Therefore, Event-Indexing model is
more facilitative for the analysis of narrative texts than expository texts.
4. Construction-Integration Model
Kintsch (1983, 1988) introduced this model which was an improvement over the previous
models of the reading comprehension. Proponents of Construction-Integration model
(Johnson-Laired, 1983; Rumelhart, 1977; Schank & Abelson, 1977; Van den Broek &
Gustafson, 1999) maintained that C-I model was different from the schema-based theories in
the sense that it involved the complex process of mapping the incoming information to
information in the long-term memory along with using different strategies during the

178 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

comprehension. Kintsch (1983) believes that merely emphasizing the role of schemata is not
sufficient to account for reading process. It is also important to identify how different sources
of information are represented in the memory of learners and how the concepts propositional
and situational information are interrelated and how these finally lead to the understanding of
the piece of discourse. It is significant to know how the iterative process in mapping the
current discourse input to the prior discourse context plays a decisive role in comprehension
process. Consequently, in the updated model of Kintsch (1983), comprehension is more than
the relationships between the explicitly mentioned discourse constituents. Rather, it involves
generating inferences that lead to the incorporation of the relevant background knowledge
into the mental representation. These constructs conveyed that deep comprehension reflects
an understanding of the referenced and implied situations, rather than representing the
explicit content in the text. The two key processes involved in Construction-Integration
model, construction and integration, are described below.
4.1 Construction
Before elaborating on the construction stage, we need to know that the understanding of the
meaning of the text is a complicated and multilayered process. Every phrase, word or
sentence the readers process from the text, activates different nodes and links in their long
term memory. The nodes include propositions, concepts, words and their meanings. The
contents of the short memory activate this complex network of nodes and links. For each
cycle of input during construction, there are four potential sources of activation. These
sources include the current input (sentence proposition that is being processed from the text),
the previous sentence or proposition, related knowledge and potentially reinstatements from
the prior text.
In other words, during the construction phase, different levels of representations including
text-based knowledge, background information and situation model (through inferences) are
activated simultaneously. In this process, the linguistic cues in the text guide the reader in
forming a situation model. In the construction phase, propositions and concepts from the long
term memory network are added to the text representation under construction (words,
sentences). As a result of text-based and knowledge-based construction, a set of N+M
elements is obtained. As Kintsch (1983) maintains, N elements from the text, including the
words, phrases, units, concepts, propositions or model elements, plus knowledge propositions
which have been selected from the long term memory by associative activation process, lead
to the comprehension process.
4.2 Integration
During the integration phase, the constructed networks of (semantic, syntactic, propositional
and situational information) are linked together based on the level of their association. Words
are linked to phrases they are component of, phrase to sentences and so on (Kintsch, et. al,
1990). In order to establish connection between the propositions they are connected through
argument overlap. The knowledge propositions that were activated associatively are linked to
text elements through which they were selected in the first place.

179 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

In other words, for Kintsch (1983), knowledge is represented as an associative network, the
nodes of which are words, concepts and propositions. Comprehension in the CI framework is
the result of interaction between texts and the general stored knowledge and personal
experience that the comprehernder brings to situation. Integration refers to the spreading of
activation across the network until it settles. This process results in greater activation for
peripheral concepts that have fewer connections to other concepts in the mental
representation. Due to the limitations of the working memory, activation is spread through
network and finally leaves only those few concepts and ideas that are connected to many
other concepts, whereas less connected concepts lose activation. To put it more simply, the
concepts that are associated to more concepts remain in the long term memory and the others
are not adopted in the comprehension process. In the integration phase, all the activated
components are linked together based on their level of association. In this stage of
comprehension, text-based knowledge, background information and situational representation
of the text link to one another based on their associations. That is why the principles of this
model are grounded in statistical, connectionist, emergent and constraint-based theories of
storing and retrieving of knowledge.
4.3 Levels of Representation
According to CI model, every sentence that is being read has three levels of representations.
In the surface structure, each word in the text is represented by a node, and the links between
these nodes indicate syntactic relations. It is important to note that the surface structure is
often disregarded in the computational model because it is assumed to have limited impact on
the comprehension.
One of the pioneering contributions of CI approach to the comprehension theory was the
introduction of text-based level representation to the text processing, which is formed in
terms of the propositions. Propositions consist of predicate and arguments. A proposition
basically represents one complete idea. It represents the underlying meaning of the explicit
information in the text, discourse or scenes. According to Kintsch (1983), readers
comprehend discourse based on the complexity of the propositions, not sentences. In his
studies, Kintsch indicated that the processing loads are the result of the propositional
information, not the syntactic relations within a text. In this conceptualization, meaning-based
analysis is preferred to the syntactically-based sentences in understanding discourse. For this
purpose, Kintsch (1990) defined the notion of atomic proposition that includes one predicate
and two arguments and arguments fill slots determined by the predicate. (e.g., agent, object,
instrument, goal). The following example consists of predicate ‘gave’ and three arguments
including an agent (I) object (article) and goal (professor).
-Sentence: I gave the article to the professor
-Propositional representation: gave (article, I, professor)
It should be noted that the explicit information regarding time and place is represented within
complex propositions, which consists of several sub-propositions to a core proposition. There
are many ways to represent complex sentences propositionally.

180 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

When the reader is able to find relationship between arguments (argument overlap), he can
connect the two propositions and understand their meaning. If the arguments do not overlap,
the reader must make inferences to fill the gap. It is worth mentioning that the overlap
between predicates (verbs, modifiers) in the CI model does not result in forming relationship
between them. For example, the two sentences in the examples 1 and 2 would be linked with
an argument overlap (with classroom) and thus would be more cohesive, whereas the two
sentences in the example 3 and 4 would be necessary to be connected with a text-based
inference.
Sentence 1: Yesterday in the classroom I gave the new article to my professor
Sentence 2: Most of the other students in the class smiled.
Sentence 3; Yesterday, in the classroom I gave the new article to the professor.
Sentence 4: She gave it immediately to her friend.
Notice that in sentence 4, there are no explicit arguments that provide overlap between the
sentences. This results in a potential cohesion gap because, although the verb gave occurs in
both sentences, overlap resulting from predicates is not included in the model. Notably, that
is driven solely by argument overlap and not by events or actions (Kintsch, 1983, 1990);
however, based on this model, the reader would be likely to make the inference she was the
student (based on previous sentences) and this inference would connect the two sentences.
These types of textual bridging inferences and knowledge-based inferences widely contribute
to the situational level of representation.
In cases where there is no text-based information, the reader makes inferences that go beyond
the concepts explicitly mentioned in the text. Kintich (1983) classifies the inferences that
contribute to the situational level of representation according to whether they are automatic
versus controlled and also whether they are retrieved or generated. The retrieval-based
inferences are extensively supported by Resonance model offered by a number of scholars
(Albrecht & O’Brien, 1993; Myers & Obrien, 1998; Albrecht & Myers, 1995). In this model,
multiple sources of information are activated and this activation is a subconscious process.
For this reason, this process is called dumb process. According to Kintsch (1983), only those
types of inferences that are not text-based, the elaborative inferences that include the
information outside the text content and involve strategic and effortful processing, are
included in the situational representation.
4.4 Cohesion, Coherence, Situation Model
It is tangible that coherence and cohesion of different text-types are of immense application
in the theories of text processing. A coherent understanding of a text or discourse emerges to
the extent that readers activate related knowledge, integrate that knowledge with the mental
representation, and establish connection between propositions in discourse representation.
Although in many cases these complex processes are automatically-based on the part of the
reader (they do not need conscious effort), it should be noted that some texts are not textually
coherent. In cases that text-based information is not helpful (i.e., cohesion gaps) for a

181 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

coherent understanding of the text, the readers are induced to activate more background
knowledge and engage in effortful inferential process. In this type of inference, the situational
representation of the text become helpful. If the reader can make relatively automatic
connections to the prior discourse, then less prior knowledge will be activated. If gaps are
encountered, then the reader will activate prior knowledge to the extent that it is available. In
cases where learners have more prior information (e.g., in their special fields), this
knowledge leads to multitude of connections in the nodes, consequently, the representation
settles more quickly becomes more stable and results in stronger long-term memory.
Studies done by McNamara & Kintsch (1998) and McNamara (2007) indicate that there is
considerable interrelationship between text cohesion and prior knowledge. According to these
studies, readers with limited background knowledge benefit from greater cohesion in the text,
because they lack the necessary prior knowledge to generate bridging and elaborative
inferences. When the text lacks cohesion, inferences may improve the readers’ text-based
level understanding and those inferences may improve the situation model for individual
sentences. The theoretical explanation for these cohesion effects are grounded in the premise
that comprehension is largely determined by the coherence of the readers’ situation models,
and this is the function of both the cohesion of the text and the inferences generated by reader.
This assumption is generally accepted by most of the computational models of
comprehension.
4.5 Limitations of CI Model
Although the CI model by Kintsch (1983, 1988) was a breakthrough among the theories of
text processing (Grabe, 2009; Alderson, 2000), there were numerous drawbacks within it.
Firstly, the concept of argument overlapping as the only way for connecting propositions to
establish coherence was put to question by some scholars (Rumelhart, 1984; Schnotz, 2002;
Trabasso & Sperry, 1985). Moreover, this model is adopted to work for the analysis of
expository texts and does not take into account the narratives and other genre. In addition,
this model does not explain the implementation of some of the readers’ goals and
metacognitive process. It does not take into account the individual differences in the
understanding of the text.
5. Structure Building Model
Gernbacher (1990) proposed the Structure Building model with the goal of providing a theory
of comprehension regardless of medium (including pictures, films, etc.). The underlying
process of the model includes general cognitive operations that function apart from the
information that is common across different modalities. Thus, the focus of the model was on
identifying and describing the processes that operate during the comprehension of various
media such as texts and pictures.
5.1 Fundamental Assumptions
The Structure Building model defines comprehension based on three central cognitive
processes: (1) laying a foundation for the mental representations of the text or discourse
structure, (2) mapping information onto that foundation, and (3) shifting the new structures

182 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

when new information is not in conformity with the existing structure or it is the beginning of
the new ideas. There are two mechanisms that operate to determine the strength of memory
loads. Enhancement increases activation and suppression lowers activation. The following
section describes the model’s assumptions.
5.2 Laying the Foundations
When a reader confronts with the information in a text, he encodes the initial contents in the
discourse which is of great importance for mapping the subsequent information to the
foundation. To put it more simply, when the reader reads the beginning of a novel, his first
conceptualizations of the topic and the possible up-coming context, act metaphorically like a
foundation of a building on which other parts of it are built. Consequently, in the Structure
Building model, the first stage is laying the foundations which is more resource demanding
(more cognitive load) than the other process such as mapping and enhancement.
The assumption that laying a foundation occurs during comprehension is supported by three
sources of evidence. First, comprehendres show slower reading times during the preliminary
stages of processing a text, for example, when they read the first sentence of paragraph. (e.g.,
Glanzer, Fisher, and Dorfman, 1984) or the first sentence of episode. Readers also process
more slowly the first sentence of the novel (Haberlandt, Berian., & Sandson, 1980). And
finally Haberlandt, Berian & Sandson (1980) maintain that compreheders show an advantage
for first mention where the first mentioned protagonist is more easily accessed to the second
mentioned. For example, in the sentence, Ali watched TV, and Reza went home, the first
protagonist, Ali is more quickly accessed in memory than is Reza after reading both
sentences. Thus, even though Reza is more recent, Ali is more accessible in memory.
5.2.1 Mapping and Shifting
After laying the foundation, the comprehender maps the upcoming information with the
aforementioned foundation (things he predicted about different aspects of the text). The
possibility of effectively mapping new information to the structure is triggered by syntactic,
referential, temporal, and causal relationships (Gernsbacher & Givon, 1995). Overlaping
(relationship between sentences or paragraphs) can be observed in various ways, including
syntactic cues (Gernbacher, 1991), concept repetition (Haviland & Clark, 1974) pronoun
reference, temporal continuity and causal coherence. When comprehenders cannot map to a
structure, then a sub-structure is built, which in turn necessitates laying another foundation or
substructure is created, and the process resumes. The establishment of sub-structure happens
when the foundation built by the reader does not match the subsequent knowledge in the text.
5.2.2 Suppression
As a substantive principle in Structure Building model, enhancement is put to work when the
incoming information from the text is in some aspects pertinent to the current foundation
built by the reader. Comprehension depends on the efficient construction and maintenance of
mental structures, in this case, it is enhanced and incorporated into the mental structure and is
then added to the foundation. Nonetheless, if the new information does not relate to the
current structure, the comprehender may shift to a new mental sub-structure (building a new

183 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

foundation) or suppress new irrelevant information.

Unlike the two previously discussed models, the Structure Building model can account for
the individual differences in comprehension skill. According to the model, skilled and less
skilled comprehenders can be distinguished in terms of the efficiency of suppression process
which determines how quickly the irrelevant meaning of ambiguous words lose activation
(Gernbacher & John, 2000). For skillful readers, identifying the irrelevant information in the
foundation is much easier than less skillful readers. Consequently, As Grenacher (1990)
maintains, the readers endowed with more effective suppression mechanism create fewer
sub-structures because they are able to inhabit the irrelevant information.
6. Conclusion
The arguments of this paper analytically illustrate that the top-down and bottom-up
processing are too general and simplistic to account for the complicated nature of reading
comprehension. Drawing on simple examples, we showed that through situation models, it is
possible to have more profound understanding from reading skill. The
Construction-Integration model was pioneer in conveying how the different levels of
representation are stored in the readers’ memory. This model also incorporates the notorious
concepts of cohesion and coherence in reading process without any recourse to the
Chomskyan and behaviorists’ accounts of language nature. The adoption of connectionist
theory of learning for explaining the construction of text-based knowledge, background
knowledge and situational representation along with their integration was very helpful in this
regard. The model also sheds light on the two types of text-based and knowledge–based
inferences and shows how the bridging and elaborative inferences play major role in filling
the cohesion and coherence gaps in the text. It was also shown that Event-Indexing model is
helpful in presenting a plausible explanation for the situational representation of a narrative
text. This model, by drawing on the five indexes of time, causality, space, protagonist,
motivation, presented a coherent model of situational representation for the first time. In a
similar strand, Structure Building model is helpful in indicating how individual differences
are mirrored in the suppression stage of the irrelevant information in story comprehension. It
goes without saying that if teachers are familiarized with the aforementioned models, the
quality of their reading instruction will outstandingly improve and lead to the more
reading-oriented accomplishments on the part of the learners.
References
Adams, M. (1990). Beginning to read. Cambridge, MA: MIT press.
Albrecht, J. E., & Myers, J. L. (1995).The role of context in accessing distant information
during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21,
1459-1468.
Albrecht, J. E., & O’Brien, E. J. (1993). Updating a mental model: Maintaining both local
and global coherence. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 19, 1061-1070.
Alderson, J. C. (2000). Assessing reading. Cambridge: Cambridge University Press.

184 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

Barnett, M. (1989). Teaching reading strategies: How methodology affects course articulation.
Foreign Language Annals, 21, 109-121.
Bernhardet, E. (1991). Reading development in a second language: Theoretical perspectives.
Norwood, NJ: Albex.
Carrell, P. L., & Eisterhold, J. C. (1983). Schema theory and ESL reading pedagogy. TESOL
Quarterly, 17(4), 553-573.
Clarke, M. & Silberstein, S. (1977). Toward a realization of psycholinguistics principals for
the ESL class. New York: Methuen.
Davoudi, M. (2005). Inference generation skill and text comprehension. The Reading Matrix,
5, 106-124.
Eskey, D. E. (1986). Theoretical foundations. In F. Dublin, D. E. Eskey, & W. Grabe (Eds.),
Teaching second language reading for academic purposes. (pp. 3-23). Reading:
Addison-Wesley.
Garnham, A. (1985). Psycholinguistics: Central topics. New York: Methuen.
Gernbacher, M. A., & John, M (2000). Modeling the mechanism of surprising lexical
accesses. In D. S. Gorfein (Ed). On consequences of meaning. Washington, Dc: APA
Publication.
Gernsbacher, M. A. (1990). Language comprehension as structure building. Hillsdale, NJ:
Erlbaum.
Gernsbacher, M. A. (1991). Comprehending conceptual anaphors. Language and Cognitive
Processes, 6, 81-105.
Gernsbacher, M. A., & Givon, T. (1995). (Eds.). Coherence in spontaneous texts. Amsterdam:
Benjamins.
Glanzer, M, Fisher, B., & Dorfman, D. (1984). Short- term storage in reading. Journal of
verbal learning, 23, 467- 468.
Glenberg, A. M., Meyer, M., & Linden, K. (1987). Mental models contribute to
foregrounding during text comprehension. Journal of Memory and Language, 26, 69-83.
Goodman, K. (1967). Reading: A psycholinguistic guessing game. Journal of the Reading
Specialist, 6(4), 126-135.
Grabe, W. (1988). Reassessing the term interactive. In P. Carrel, J. Devine, & D. Eskey (Eds),
Interactive approaches to second language reading (pp. 56-70). New York: Cambridge
University Press.
Grabe, W. (2009). Reading in a second language: Moving from theory to practice. New York:
Cambridge University Press.
Groundson, L. (1991). ESL literacy instruction: A guide-book to theory and practice.
Englewood Cliffs, NJ: Prentice Hall Regents.

185 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

Haberlandt, K., Berian. C. & Sandson, J. (1980). The episode schema in story & Samson, J.
(1980). Journal of verbal learning and verbal behavior, 19, 635- 650.
Haviland, S. E., & Clark, H. H. (1974). What’s new? Acquiring new information as a process
in comprehension. Journal of Verbal Learning and Verbal Behavior, 13, 512-521.
Iran- Nejad, A. (1987). The schema: A long-term memory structure or a transient structural
phenomenon. In R. J. Tierney, P. L. Andres, & J. N. Mitchell. Understanding readers’
understanding (pp. 109-127). NJ: Lawrence Erlbaum.
Johnson-laired, P. N. (1983). Mental models: Towards a cognitive science of language
inferences, and consciousness. Cambridge, MA: Harvard University press.
Kintcsh, W. (1983). The role of knowledge in discourse comprehension: an event indexed-
model. Psychological Science, 6, 292-297.
Kintsch, E. (1990). Macroprocesses and microprocesses in the development of
summarization skill. Cognition and Instruction, 7, 161-195.
Kintsch, W. (1988). The role of knowledge in discourse comprehension: A
construction-Integration model. Psychological Review, 95, 163-182.
Kintsch, W. (1998). Comprehension: A paradigm for cognition. New York: Cambridge
University Press.
Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and
production. Psychological Review, 85(5), 363-394.
Kintsch, W., Welsch, D., Schmalhofer, F., & Zimmy, S. (1990). Sentence memory: A
theoretical analysis. Journal of Memory and Language, 29(2), 133-159.
Lynch, B., & Hudson, T. (1991). EST Reading. In M. Celcia-Murcia, & L. McIntosh (Ed.),
Teaching English as a Second or Foreign Language (pp. 216-232). Boston, MA: Heinle &
Heinle.
Mckoon, G., & Rattcliff, R. (1992). Inference during reading. Psychological Review, 99,
440-466.
McNamara & Kintsch, W. (1998). Learning from text: effects of prior knowledge and text
coherence. Discourse Process, 22(3), 247- 288.
McNamara, J. (2007). A three pronged-method for studying inference generations in literary
text. Poetics, 20, 193-232.
Morrow, D. J., Bower, G. H., & Greenspan, S. L. (1989). Updating situation models during
narrative comprehension. Journal of Memory and Language, 28(3), 292-312.
Myers, J. L., & O’Brien, E. J. (1998). Accessing the discourse representation during reading.
Discourse Processes, 26, 131-157.
Rieben, L., & Perfetti, C. A. (Eds.) (1991). Learning to read: Basic research and its
implications. NJ: Erlbaum.

186 www.macrothink.org/ijl
 International Journal of Linguistics
ISSN 1948-5425
2015, Vol. 7, No. 5

Rumelhart, D. E. (1977). Toward an integrative model of reading. In S. Dornic (Ed.,

Attention and performance (pp. 573-603.). New York: Academy Press.
Rumelhart, D. E. (1984). Schemata and the cognitive system. In R. S. Wryer. Jr & Tulk, the
world of text. (pp. 76-98). New York: prentice Hall.
Sanford, A. J., & Garrod, S. C. (1981).Understanding written language: Explorations of
comprehension beyond the sentence. New York: Wiley
Schank, R. C., & Abelson, R. (1977). Scripts, Plans, goal, and understanding. Hillside, NJ:
Lawrence Erlbaum Associates.
Schnotz, W. (2002). Commentary—towards an integrated view of learning from text and
visual displays. Educational Psychology Review, 14(1), 101-120.
Shahnazari, M. T., & Dabaghi, A. (2014).A critical overview of models of reading
comprehension with a focus on cognitive aspects. Iranian Journal of Research in English
Language Teaching, 1(3), 7-19.
Stanovich, K. (1980). Toward an interactive-compensatory model of individual differences in
the development of reading fluency. Reading Research Quarterly, 16, 32-71.
Stanovich, K. E. (1986). Matthew effects in reading: Some consequence of individual
differences in the acquisition of literacy. Reading Research Quarterly, 21(4), 360-407.
Trabasso, T., & Sperry, L.L (1985). Causal relatedness and importance of the strong events.
Journal of memory and language, 24, 595-610.
Van den Broek, P., & Gustafson, M. (1999). Comprehension and memory for texts. Three
generations of reading research. Journal of Memory and Cognition, 29, 51-84.
Zewan, R. A., Langston, M. C., & Grasser, A.C. (1995). The construction of situational
models during comprehension. Journal of Memory and Language, 13, 441- 469.
Zwaan, R. A, & Radvansky, G. A. (1998). Situation models in language comprehension and
memory. Psychological Bulletin, 123, 162-185.
Zwaan, R. A., & Madden, C. J. (2004). Updating situation models. Journal of Experimental
Psychology: Learning, Memory, & Cognition. 30(1), 283-288.

Copyright Disclaimer
Copyright for this article is retained by the author(s), with first publication rights granted to
the journal.
This is an open-access article distributed under the terms and conditions of the Creative
Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

187 www.macrothink.org/ijl