CHAPTER 10

THINKING AND
PROBLEMSOLVING

CHAPTER OUTLINE
Solution
Classic Problems and General Methods of
Generate-and-Test Technique
Means-Ends Analysis
Working Backward
Backtracking
Reasoning by Analogy
Blocks to Problem Solving
Mental Set
Using Incomplete or Incorrect Representations
Lack of Problem-Specific Knowledge or Expertise
The Problem Space Hypothesis
Expert Systems
Finding Creative Solutions
Unconscious Processing and Incubation
Everyday Mechanisms
Critical Thinking

his chapter is about different kinds of thinking and problem solving, the kind o
mental work youdo in each of the following tasks:
1 Think of your favorite restaurant. What is its name? Where is it? What are its best disie
What makes ityour favorite?
2. Solve this problem: If 10apples cost $2, how much do 3 apples cost? example
3. Create unusual but appropriate titles for the drawings shown in Figure 10.1:for
"Giant robot head" for Figure 10.1(A).
 Chapter 10: Thinking and Problem
Solving 251
a change in one of
whethera your
consider requirements) would have school's policies (for example,
overall beneficial
distribution or harmful
effects. dropping all
will examine descrip-
We mental
explanations
chapter, for the
done, How did you
just
have tasks? What processes
aOmotsh the
willlook at a
number of
u s e ?We and discuss what
tasks
thinking
eithereasy or hard.
you
id
diterent
thinking
nakes
broad teerm.
Psychologists
is a what look
Thihing
thinking often study thinking
study tasks. Defining
who different
very
a toughjob and one that
ke t o be Thinking has been
requiresthought.
out
tns
beyond the informa-
tself
detfinedas
"going complex
given" (Bruner, 1957), as "a
skil" that "fill[s] up gaps in
ton high-level p. 20), as a
andevidence"(Bartlett, 1958,
the of searching
through a problem Figure 10.1: Ambiguous drawings.
(Newell & Simon, 1972), and as
process
space
do "when we are in doubt about how to act, what to believe, or what to desire"
what we
p. 6).
Baron,2000,
thinking is used to refer to more than one specific activity. This suggests
Clearty,theterm
different types of thinking. One distinction that may prove useful is between
there may be unfocused thinking. Focused thinking begins with a clear starting point and
bcused and
has aspecific
goal.(We will see examples of focused thinking incharacterthis chapter, as well as in
of daydreaming
ch of the material in Chapter 11.)Untocused thinking has the
ideas. We will
wunintentionally calling to mind a number of different and loosely related
oimarily explore focused thinking, especially in the first sections of the chapter that deal
which some have described as
with problem solving. We will then turn to creative thinking, examine how people evaluate the
including aspects of unfocused thinking. Finally, we will ways in which people assess their
products of their thinking. In particular, we will look at the against bias or impulsivity.
ideas, reflect on the implications of their conclusions, and guard
through the use of problems and
YoU may wonder why psychologists studyofthinking thinking that occurs in everyday life (such
puzzies that appear not to mirror the kind what to order in a restaurant, or what route
Wnen you think about what shirt to wear,.
intuition that everyday thinking often
ake to get to work). One reason stems from the hard to study. Moreover, people likely
SSO rapidly and automaticallv that it wouldbear be
much of their backaround knowledae to in their everyday thinking. You pre
and
yyou expect to be doing have
odyChnoose what to wear for the dayon the basis of whatdress. Because people
perhaps on external standards or expectations regarding be nearlyimpossible to devise
varying
aproblem
then, it would
background knowledge and different individuals. By presenting
goals, standardized
that is equal in difficulty for
sets of problems, over the information participants have
have more control throughout the
avaiChaptlablere
and how it isinvestigators Various problems are presented
given to them. phenomena of thinking. Isuggestofthat to maxi-
as
mize the valueOpportunities to work with the time-honored method
observation
of these undertakings, you rely on a
 Laboratory
Outofthe
Psychologyinand discussed in Chapter 1,
Cognitive As

inandtrospecicttlc
introspection. observation of the
252
experimental
psychology: nonjudgmental
and Although
introspectionhas problens contents
in detailed,
concurrent,problem.
summary), itcanatthe very least
the
Simon,
worka detailed
using
Ericsson& andtests mnorethanis
asked
measures The
sciousnessasyou1984,fora moreObjective for: Don't explain key to
or
on howto
justiy
he
When provide
proper
hypotheses providesinstructions
for
techniqueisto
avoid
just
doing
report
thinkingabout; obtain paper
10.1
it. Box pen or pencil you
and a
record your
thoughts as
for

YoUwon't
note
work
taking
or,
be. the problens.
iprntreofsepraecblt,y
A

youread
further,
whichyoucan other quiet place. yourthhoughts-just beyour
recorderinto room orsome
a
showing Notkin
thencaretu obser y
to control
try you can
privacyof your sodon'tcensor describeyour performance,
or
else,
toanyone how wellthe
theories
theories
presented. compare
Toassessdescriptionsofthe
noteswith

Box10.7 InstructionsforIntrospecting
back hunches, guesses, wild
whatever's on your mind. Don't
hold ideas, images
1. Say something atleast once every 5
intentions.
continuously as possible. Say seconds,
even
Speak as involved
drawinga blank." as you become
only,lm voice dropping
Watch outforyour
3 Speakaudibly.
you please. Don't worry about complete sentences and elo-
telegraphically as
4. Speak as normally.
would
quence.
orjustify. Analyze no more thanyou
5. Don'toverexplain events. Get intothe pattern ofsaying| what you're thinking now, not
6. Dontelaboratepast then describingyourthoughts.
while and
ofthinkingfor a

SOURCE: Perkins (1981,p. 33).

into the dlass

similar in at least one respect: They fall
are (you know inne
The problems presented Well-defined problems have a clear goal tostart fror
problems called well defined. solution), present asmall set of information
while yoU 3
diately if you've reached the present a set of rules or quidelines to abide by goals, Stat
and often (but not always)In contrast, ill-defined problems don't have their
working toward a solution. spelled out.
ing information, or steps clearly as folows
be illustrated knowth
ill-defined problems can
The difference between welI- and a purchase, giventhatyoubasic e
the sales tax on taxation, and relatie
Consider the problem of calculating taxable, the rate of be
it is
price of theitem you are buying, whether backgroundI information, itshould probe
of multiplication. If you are armed with this
Contrast this with another dificu
at the taX. 'Conveysa ton
easy for you, a college student,to arrive articulately and sensitively you'restill
often encountered: composing aletter that "Dear Jane" letter to someone thesecases
message (for example, a "Dear John" or any of
clear in
of or aletter to your boss askingfor a promotion). It's not
 Chapter 10: Thinking and
problems what information
of your
qualifications VOu should
and past start from
Pr oblem Solwing 233
ened
many

also not clear hen

be made better?) or you'
year's
whatverules (f any) apply.
(how much of
your
t's
ac omplishments
hOW
natorCanit
reachned the goal (is the do youeducat ion
tell your
on current draft good
they
have focUsed
don't take weeks
Change.It is assumed
gesen,
Mell-defined
or
mont h s probl
to ems for several
that problem solve, they are reasons:
solving for sol ving They are easy
easy to score,
for
tested(Galotti, wel l-def in ed
TO
problem
BSY to
ways
extensively
ginlar
1989). In problems,
on one study, although this
il -defined and they
problems works in
st
demonstratedthat performance
ben
s9) on an ill-defined
etomance
one. well-defined problSchraw, Dunkleassumpt
, and ion has
ems was not Bendixwienth
correlated
PROBLEMS AND GENERAL
aASIC METHODS OF SOLUTION
solve a problem depends, to a great
wayto
istoflyto Los
Theproblem, Angeles, you extent, on the problem. For
might
web pages of relevant airlines or generalvarious airlines or travelinstance,
jor the
call if
surt
your problemiis to travel agents or
websites such as Orbitz or
Tavelocity.Incontrast, if
even

tor.assistancefrom
a balance your
travel agent, but you mightfromcheckbook, you normally would
natask problerm-solving a banker. These
approaches-they
onlybe reviewing a certain
only work for alimited class ofare domain-
class of general, problems.
Hee,I wil
These
methods are stated at a general
wide variety
enough domain-independent
level that, in techniques.
of problems,, not just with problems of a principle, they can be used
wtha certain type or domain.
sENERATE-AND-TESTTECHNIQUE

Hereisthefirst problemfor you to try. Think of 10 words beginning with the letter cthat
are things to eat or drink. Write down all the things that occur to you, even if they end up
not meetingthe criteria. How are you solving this problem?
Ibred a real-life problem somewhat like this several years ago. Iwas in Minnesota and
had to get 100 Swiss francs to a hotel in Bern, Switzerland, within aweek to hold aroom
reservation. Iwent to the post office to get an international money order but discovered
twould take about a month for the order to make its way tothe hotel. Ideliberated over
how to solve the problem, talked to lotsof friends who've traveled more extensively than
Ihave, and came up with a number of ideas. Icould call people at American Express
and see if they could help. Icould see if by chance any of my friends would be traveling
to Bern in the next week. I could callWestern Union and wire the money. I couldget a
cashier's check from a bank. Icould go to my automobile club, purchase atraveler's
MleK In SwISS francs, and mail it.The first four options,as it turned out,wouldn't Work or
Mele much too expensive. The fifth one met mycriteria of being possible, Working within
dmeeKS time, and being (relativelv) affordable, so that was the one l chose.
example of the generate-and-
The process that I used in solving9 this problemis a goodgenerating
test technique. As the consists of possible solutions (for
name Suggests, it then
example, "Let's call people ifthey can help") and
at American Express and see help me with the following
Testing then (for "Hello, American Express? Can you
possibilities but did for the fifth (it
problem.work,. example,
WOud the ?"). The tests didn't Workfor the first
cost was reasonable, and the money
four
would get there in time).
 PsychologyIn andOutoftheLaboratory
254 (ognitive
used generate
You may have problem of
and-test to solve the
begin with
listing 10 words that
name things to eat
C and that worked on this
or drink. When I Cheerios FAMILY SI}E
names came to
problem, some
if they start
mind that sound as example,
with c but don't (for
Doubla Triscui
ketchup (unless you spell it Mor
W e H

sarsaparilla), and
catsup) and to mind that
Some words came
aren't edible or
start with c but cHiRA
Again,
drinkable (cable, canoe).thinking
was
the process used (generating)
of possible solutions possibili
and then seeing if those(testing). Photo 10.1: Here are someexamples offoodssthat
ties met allthe criteria begin with the
loses its effectiveness leter c.
that very
particular guidance for therapidly
Generate-and-test is atechnique
many possibilities and when there is no
your locker, for instance, the
when thee
generation
you forget the combination to time might exceed your willingness will technique
but your frustration level by that way to keep track of the possibilities
task. Moreover, if you don't have a
to
persevere
yout eventualyNO
ajhaokeve ti
real
along withthe ones you have yet to try, on a Rubik's cube trouble. There's a
you might be in
puzzle sits next
man twists the Rubik's cube into a particular pattern, thrustst0 asigint te
movie in which
UHEblind a blind man working
man. The
correct pattern]?"
of the sighted man, and asks, "ls this it [the The joke is that"No," says the later.
interchange is repeated, rapidly, several times. this method of ie
solving is all but doomed to failure, given the large number of possible configurations and proolen
the lack of any systematic way of trying them.
possibilities i
Generate-and-test can be useful, however, when there aren't a lot of
track of. If you've lost your keys somewhere between the cafeteria and your roomaandyou
made intermediate stops in a classroom, the snack bar, and the bookstore, you can
this technique to help you search.
MEANS-ENDS ANALYSIS

curerty
Suppose you want to visit a friend who lives in Summit, New Jersey, and you are
transportation: walk
residing in Pomona, California. There are several possible means of
ing, bicycling, taking a taxi, taking a bus, taking atotrain, driving a car, or taking aplae
or helicopter. The most practical means might be fly on a commercial airline; is e
toget to the neaes
fastest and fits your budget. However, to board your flight, youhave bicycle, take aas
airport, which is 5 miles east of your residence. Again, youcould walk, airport istodrie
the
and so on. The most efficient and cost-effective means to get to
you are sitting wnenwalkythee
your car. However, the car is parked in the garage, not where choose to
ready to depart, so you have to get to the car. You would probably
(as opposed to, say, calling a cab).
 Chapter 10: Thinking and Problem
Solving 255
techniqueof
problem solving described here is called
the goal
comparing
altomia)thinking (Summit, New Jersey)
of possible ways of overcoming the with means-ends
the (walking,
distance starting analysis.
bicycling, tak-
moves
choosingthe
point (Pomona,
on), andconditions(for best one. The selected option
tme
so
and
prerequisite example, being at the
recondiiONSaren met, then a subgoal is created (for example,
taxi,
(taking a plane) may
airport,
"How with
can ayou get toIf the
the
naa
Certain

aiport?2.Through the creation of subgoals, the task is broken dowWn into

ticket).
manageable
e

a full solution to be constructed.

allow
that their associates studied
steos
Simon (1972) and means-ends analysis while solving certain
Newel and. such as thefollowing:
problems,
arthmetic DONALD
+ GERALD
ROBERT
D=5, determine the values for the other letters.
Giventhat cryptarithmetic problems.) (Problems in which letters
digits are known as
standfor computerprogram, called GPS, or General Problem Solver,
researchers createda
Thesolves
problems in cryptarithmetic and in logic using means-ends analysis. GPS
that
basic strategy. First. it looks at the object it is given (such as the
the following
cryptarithmetic and I
Uses problem) compares it with the desired object (an arithmetic
preceding numbers in place of all letters in which the solution is actually the addition of
with
problemnumbers above the line). By doing so, GPS detects any differences between the
thetwon
actual andthe
desired object.
Next, GPS
considers the operations available to change objects. Here, the available
operationsinclude those that replace certain letters with certain digits, for example, D=
nerations used are chosen with the aim of reducing differences between actual
nd desired obiects. In cases where none of the available operations applies to the actual
biect. GPs tries to modify the actual object so that operations can apply. GPS also tries
tn keep track of various kinds of differences between desired and actual objects and to found,
work on the most difficult differences first. Thus, if several possible operations aredifferent
of ranking the
allof which could applyto an actual object, GPShas some means
operations Such that certain ones are used first.
to both
Newelland Simon (1972) gave several problems in logic and in cryptarithmetic participants
uman participants and GPS and compared the "thinking" of both. Human as you
generated verbal protocols, much like the ones you have been asked to generate operations
subgoals, and the
u nis Gnapter; GPS produced a printout of its goals, its
tapplied as it worked.
concluded that there were
Comparing the protocols generated, Newell and Simon (1972)performance of the Yale stu-
many the performance of GPS and theanalysis, the general heuristic
dents similarities between
who served as participants. Notice that means-ends generate-
(shortcut focused method of solution than
strategy)
and-test: t gives theused by GPS, is a more
problem solver more guidance in choosing what step to take next.
aspects of the problem before
Means-ends
Ssarting to analysis alsoforces the problem
Work on it
goals. Notice here thatand to generate a plan
solver
to solve
to analyze
it. Often this
"blindly" and
requires establishing sub
only after some thought.
the problem solver is acting less
 PsychologyinandOutoftheLaboratory
256 Cognitive
to
the optimal way reach a
sometimes the optimal
analysis is not
wayalways
involves taking a temporary step backward solution,.
howev be
Means-ends
eastern suburb of
imagine you liveinan have Los
firstAngeles
goal. For example, do s0, you to
agreater distance(further west)gotrom your
Angeles to Denver.
the means moving,temporarily, To
that
take a flight from Los to thegoal
Curent distance. Means-ends analysis can make it difficutto see that the
one.
goal isn't always the most direct
path toward a
WORKING BACKWARD
called
Another general problem-solving technique is to
determine thelast step
needed achieve working backward.
it, then the
lyzesthe goal to friend's

to
and so on. In the problem
verylast
her
step
front
is
door
to walk
from
from
the
can get
outside her front

a cab at the
high school
of getting to my door intothe
Manchester, New Hampshire. airport canThe
airport, and So on.
be
Working
house.house, problem
for
next-0ns-tasniea,t
a cab to her house. I functions similarlyto means-ends
analbacykswisa.rdh
involves establishing subgoals, so it
Working backward is a very
important technique for solving
many problems, including the
farnous Towers of Hanoi prob
lem, depicted in Figure 10.2. A
Successful episode of problem Fiqure 10.2: The Towerssof Hanoi problem. Determine as
sequence
transfer the three disks from the firsttto the third peg, moving onlyy
solving might be something like never placing abigger disk on top ofa smaller one.
ot
One isk MOVst
ata timea
the following: "First Ihave toget
the bottom disk moved over. But to do that, I have to move the top two disks.
that if I move the second disk to the spare peg, but to do that. I have to
disk out ofthe way. I could do that by temporarily moving it to the goal peg,
move the t
the second disk to the spare peg, then moving the top disk back to the then moi
moving the bottom disk over." Notice that the solution process usually does not startvit
spare peg, te
the problem solver making a move and seeing what happens. Instead, after onlv al
practice, the usual pattern is to plan moves in advance, setting up many intemet
goals along the way (Egan &Greeno, 1974). Of course, it takes a few trials befoet
problem solver adopts the correct solution; if the puzzle consists of more than three dik
the participants are unlikely to solve it with the minimum number of moves on the fist
trials (Xu &Corkin, 2001).
Working backward is most effective when the backward path is unique, which mals
the process more efficient than working forward. And, as you may have noticed, wot
ing backward shares with means-ends analysis the technique of reducing diferens
between the current state and the goal state.

BACKTRACKING
Try this next problem. Imagine there are five women: Cathy, Debbie, Judy, Linda,ad
Sonya. Each of the five women owns a different breed of dog (a Bernese mountaind
a golden retriever, aLabrador retriever, an Irish setter, or a Shetland sheepdog
As
each has a different occupation (clerk, executive, lawyer, surgeon, teacher),
 Chapter 10: Thinking and
Problem Solving
A9Chhasa different number of children (zero, one, twO, three, or four). Given the infor- 257
how
10.2, figure out
Box many children the person
in who owns the
sheepdog
naton h a s .
Shetland
nsovingaproblem,you often need to make certain provisional assumptions. Sometimes
they
turnOutto be wrong and need to be "unmade." In thOse instances, it is useful to
hRvesonme wayto keep track of when and which assumptions were made so you can
backupto certain points of choice and start over, a process known as backtracking.
Thewomen, dogs, children, and jobs problem in Box 10.2 is a case in point. Many people
soNesuchproblerms by setting up a chart like the one shown in Figure 10.3. The chart is
corresponding to the chart of
someone who has read only the first 12 lines of
ncomplete,
point, aproblem Solver can
this determine thatortheLinda.
goldenTheretriever'
Box
who10.2.
At
is an executive with four children, is either Debbie problems owner,
solver
it's Debbie, only to find out when he reads the
termporarily assumne that
might
Box 10.2that
Debbie owns the Bernese mountain dog. He would enter that 13th line in
chart. But if the
problem solver backed up to the point at
which
infomation
he
incorrect
intohis assumption (that is, knew that either Debbie or Linda was the golden retriever- made the
mother-of-four executive). he would now know it is Linda, and
this information
owning, necessary to solve the rest of the problem. The key to
WOuldbe keep close track of
backtracking, then, is
problem solver to choice points-places where she made a
forthe
provisional assumption-So that, if subsequent work leads to a dead end, she can "back
that. choice point and make a different assumption.
up"to

Box10.2

The Women, Dogs, Children, and Jobs Problem

the owner of the Shetland
From the following information, determine how many children
sheepdog has.
There are five women: Cathy, Debbie, Judy, Linda, and Sonya.
surgeon.
There are five occupations: clerk, executive, lawyer, teacher, and
Everyone has a different number of children: 0, 1,2, 3, or 4.
Cathy owns the Irish setter.
The teacher has no children.
Ihe owner of the Labrador retriever is a surgeon.
Linda does not own the Shetland sheepdog.
Sonya is a lawyer.
The Owner of the Shetland sheepdog does not havethree children.
The owner ofthe golden retriever has four children.
Judy has one child.
The
executive
Owns a golden retriever.
Debbie owns the Bernese mountain dog.
Cathy is a clerk.
 Laboratory
258 (ognitive Psychology In and Outofthe

Woman Cathy Debbie Judy Linda

Dog
Irish setter
Sonya
Number of 1
Children

Occupation
Lawyer
Golden Labrador Shetland
retriever
retriever sheepdog
4 3

Executive Surgeon
Teacher
Figure 10.3: Partial solution to the women, doqgs, children, and jobs problem.

REASONING BY ANALOGY

The next problem, known as "the tumor problem," is famous in the literate.

Given a human being with an inoperable stomach tumor and rays that
ficient intensity, by what procedure can onefree him ofthe tumor by thesedestroy organiat tc tissueetsa
rays and
avoid destroying the healthy tissue that surrounds it? same tinthe
Originally posed to participants by Duncker (1945, p. 1), the problem is nhee.
ficult chalenge. Duncker argued from studying the
Box 10.3 presents an example protocol) that problem performance of several particigt
trial and error; rather, it involves a deep solving is not a matter of Ni
understanding
and their relationships. To find a solution, the solver mustof the elements of the bnh
the solution" (p. 4) first and then arrange the grasp the "functional val1ed
specific details.
problem is to send weak rays of radiation (weak enough so The solution to the tugy
inflict damage) from several angles, such that all rays that no individual ray wl
Although the radiation from any one ray will not be strongconverge
enough
at the site of the tuna
to destroy the tumor
the healthy tissue in its path), the
convergence of rays will be strong enougn.
Gick and Holyoak (1980) presented participants with
person had read a story Duncker's tumor problem after
such as the one in Box
10.4. Although the story appeared t
dissimilar to the tumor problem, the underlying method of solution was the same
and Holyoak found that participants who had
told that it contained arelevant hint were moreread the story of the general
likely tosolve the
were participants who simply read the story but did not have the tumor
analogy proole
between th
problems explicitly pointed out. The former group of participants are said to be usingt
problem-solving technique of reasoning by analogy.
 U: Thinking and
Problem Solving 259
Bo
1 r0 3

Protocol From Oneeof

the esophagus.
Dunckers(1945) Subjects
raysthrough
Send
healthytissues by means of a
1. Desensitizethe
chemical injection.
tumor by operating.
2 the
Expose
3.
this
oughtto
decrease the
work?-turn
intensityof the
the rays on at full strength rays on their
only after way; forhas been reached.
the tumor
One
(Experimenter:False analogy; no injection is in question.)
examplewould
Oneshouldswallow something inorganic(w
(which would notallow passage
protectthe healthystomach-walls. (E: It is not merely the stomach of thearerays)
walls which to tobe
5.
protected.)
must enter the body or
the tumor must come out.
Eitherthe rays tumor-but Perhaps one could alter
6 location ofthe how? Through pressure? No.
the
Introducea cannula.-(E:-What, in general, does one do when, with any agent, one wishes
7. toproduceinnaspecific place an effect which he wishes to avoid on the way tothat place?)
(Reply:)One neutralizesthe effectton the way. But that is what I have been attempting all
8.
the time.
around to the
exterior. (Compare 6.) (The Erepeats the problem and
9. Move the tumor
emphasizes" which destroy at sufficient intensity.")
(Compare 4.)
10 The intensity ought to be variable.
Adantation of the healthy tissues by previous weak application of the rays. (E: How canit
be brought about that the rays destroy only the region of the tumor?)
make the rays
12 (Reply:) Isee no more than two possibilities: either to protect the body or to
harmless. (E: How could one decrease the intensity of the rays en route? [Compare 4.))

13. (Reply:) Somehow divert...diffuse rays... disperse... lies stop! Send a broad and weak
the focal point and thus
bundle of rays through a lens in such a way that the tumor at
receives intensive radiation. (Total duration about half an hour.)

SOURCE: Duncker (1945, pp. 2-3).

surface features but

The tumor problem and the problenm of the general differ in their least roughly with
share an Components of one correspond at
underlying structure. The of enemy
the analogous to therays, the capturingfortress
Components of the other: The army is soldiers at the to
forces to the destruction of the tumor, and the convergence ofanalogy, participants must
the Convergence of rays at the site of the tumor. To use the beyond the
by Duncker, moving
engage in the "principle-finding" analysis described problem. Gick and Holyoak (1980)
details and the of
on the relevant structuresabstract the ways
referred into thisfocusing
process as theinduction of an
schema(usingthe term in construct
who
evidence that participantsproblems.
defSuchineda Chapters 6 and 7). They presented
from work on
analogous
representation are more likely to benefit
 PsychologyIn andOutofthe Laboratory
260 Cognitive

Box 10.4
TheStoryoftheGeneral
from a strong fortress by a dictator. The fortress was
ruled
Asmall country was surrounded by farms and villages. Manyroads led tothe
middle of thecountry, general vowed tocapturethefortress..The general
thecountryside. Arebel capture thefortress. He gathered his army at theknew
by his entirearmy wouldfull-scale direct attack.
roads, ready to launcha
However, the general then learned that the dictator had
small bodies of men could
planted
pass over
head
mines on each
or
forttheast hu atal
an

them safely, since the toadk

The mines were set so that from the fortress

would detonate the mines. Not

and workersto and
tor needed to move his troops only would this blow up the road, but However, arthgee dfoitrac
it wouldany
many neighboring villages. It therefore seemedimpossible to capture the fortress.
plan. He divided his army into small
also destr
However, the general devised a simple
patched each groupto the head of a
different road. When all was ready he
gavegrthoupse andi,
road thsefigonartlreands
marched down a different road. Each group continued down its
each group fortress at tthe same time. In thisS to
so that the entire army arrived together at the way, genera
captured the fortress and overthrew the dictator. the
SOURCE: Gick and Holyoak (1980, pp. 351-353).

It is interesting that participants often had to be explicitly told to use the

general to solve the tumor problem. Only 30% of participants spontaneously story ofte
analogy, although 75% solved the problem if told that the story of the
general
useful in constructing the solution (for comparison, only about 10% solved
noticed te
without the story). Thisis similarto afinding reported by Reed, Ernst, and the proben
Participants' performance was facilitated by their previous work on an Baneri 974,
lem, but only if the analogy was pointed out to them. analogous prct-
In later work, Gick and Holyoak (1983) found that they could do
if they gave two analogous stories rather than one. away with explict hit
Participants read the story of tha
general and a story about a fire chief's puttingout a fire by havinga circle
surround it and allthrow buckets of water at once. Participants were told theof frefghtes
was about story comprehension and were asked to write experiment
a comparison of the two before being given the summaries of each story and
tumor problem to solve. The authos
proposed that providing multiple examples helps participants to form an abstract
(inthis case, what the authors called a "convergence" schema),
schara
new, analogous problens. Catrambone and Holyoak which they later appy
(1989) further
participants were explicitly asked to compare stories, they did notsuggested that Unes
schema with which to solve the problem. form the netess

BLOCKS TO PROBLEM SOLVING

Aproblem, by definition, is something that can't be solved in a single, obvious step.Fr
instance, We don't cOunt combing our hair as an instance of f problem solving becauseth
step of using acomb does not require much thought and no particular obstacles needt
be overcome.
 Chapter 10: Thinking and Problem
oen
SOnO,in contrast, carries the meaning of a goal with some barriers or con-
Solving 261
Sometimesthe barriers and
interfere With,
successtul constraints
are so strong that
solution.,
In this they or at prevent,
apparently
that.
hinder solving a variety of
problems. wesection,
wil review some
atos

MNA
SLET

10.4 presents a
O0
problemS
mberor obtaining
theme:
same
te
amountof water, Jar B
gienthreedifferent-size Jar A
easuring jugs. Before Jar
each
on, work on
order given
inthe
goblem
the time
Capacity Capacity
dOwncomplete Problem of jar A of jar B
Capacity Desired
ondwrite to of jar C amount

takesyou recordany
1 21 127
2 100
one.Also,
14 163
Aach relative 3 18
25 99
aboutthe 43 10
Moughts problems., 42 6 21
dficultyofthe 5 20
23
59 4 31

actualy worked the 6

7 18
49
48
3 20
FOU
goblems, you probably 8 14
22
36 8 6
following: The IFigure 10.4: The water jar problem.
tund the relatively
fist one took a faster and faster at solving the subsequent problems; your
time, but you were
long
ooblem-solving speed corresponded to the number of problems that you had previously
ined. Youalso probably noticeda common pattern to the problems: All could be solved
huthe formula B-A-2C. Didyou use this formula to solve the second-to-last problem?
Fvou did, that is interesting because an apparently more direct solution would be A+C.
The very last problem is also interesting in that it does not fit the first formula at all but is
qickly solved with a very easy formula, A-C. Did it take you some time to realize this? If
s0, your performance might be characterized as being constrained by mental set.
Mental set is the tendency to adopt a certain framework, strategy, or procedure or, more
generaly, to see things in a certain way instead of in other, equally plausible ways. Mental
in a
$et is analogous to perceptual set, the tendency to perceive an object or pattern
aain way on the basis of your immediate perceptual experience. Like perceptual set,
several
Melal set seems to be induced by even short amounts of practice. Working on
iaer jug problems in a row that follow a common pattern makes it easy to apply tne
fomula but tharder to see new relationships among thethree terms.
Luchins (1942) reported on experiments in which problems such as those in Figure 10.4
Were given to university students. After solving the first four problems using theformula
B-A- 2C, all the students solved the fifth problem using this method instead of the
more direct A+ C method. Even more striking, whenthe BA - 2C solution wouldn't
WOrk, the more obvious A+ C
stuwhich suffering from mental set were unable even to see
dents would have
sOItion, worked!
Mertal setof often causes people to make certain unwarranted assumptions without being
asked
aN•re making them. Figure 10.5 gives a famous example. Most people, when
262 Cognitive Psychology In and Out of the Laboratory

to solve the nine-dot prob

lem, make the assumption
that the four lines must stay
within the "borders" of the
dots. This constraint makes
it impossible to come up
with the solution, shown in
Figure 10.6.
Another problem relevant to
mental set is borrowed from
Perkins (1981). Iwill describe
a situation, and you deter
mine what the situation is:
There isa man at home.
That man is wearinga mask.
There is a man coming home.
What is happening? Figure 10.5: The nine-dot problem.
Because Ican't interact with
you, I'll report on the ques
tions (constrained to be of
the yes/no type) that my
students ask when Ipresent
this. Is the man at home at his
own home? (Yes.) Does the
man at home know the other
man? (Yes.) Does the man at
home expect the other man?
(Yes.) Is the mask a disguise?
(No.) ls the man at home in a
living room? (No.) Is the man
at home in the kitchen? (No.)
Part of where my students
start to go wrong is in mak
ing assumptions about the
home in the situation. Many
equate home with house,
although the answer to the
problem is a baseball game. Figure 10.6: Solution to the nine-dot problem.
Perkins (1981) would argue
that the asSumptions people make in interpreting the problem are a kind of menta s
and that this mental set hinders problem solving.
literatue
Another example of mental set is illustrated in another famous problem inthethaths
known as thetwo-string problem(Maier, 1930, 1931). You are shown to a room youcand
two strings attached to the ceiling. The strings are spaced so far apart thatSomehow. N
hold on to both at the same time. Your task is to tie these strings together
 Chapter 10: Thinking and
i nthe
room With you
are a
table, a book of Problem Solving 263
cotton. What can you do?
have
ou
of
pieces
many people have
matches, a
screwdri ver, and afew
strings intodifficulty
solution, which
The to
weight
make one of the
grabit. wait for the pendulum
a discovering, is to use the
pendulum.
to
swing Swing this string, walkscrewdri ver as
a tOgether Fewer than 40% of
and
the toward you, to the other
string
strings hint. One sOurce of
withouta for a screwdriver; they failed to
participants
difficulty seemed in Maier's grab it, and tie the two
to be their experiment solved this
functions
well as for its internded
notice that the
function. This unwillingness to think of other
screwdriver could be used as
weight
ness.
aS
Itappears to be an instance of
has apparently adopted a rigid mental set,phenomenon
in
is called
that person functional
a fixed-
a
fixedness mental set toward an subject
object.
to
functional
INCOMPLETE OR INCORRECT REPRESENTATIONs
USINGG
difficulty in problem solving has to do
Arelatedd with the initial
the
problemn is misunderstood or if the
lem.Ifdisadvantage. The checkerboard wrong interpretation
information
of the
is focused on, the prob
isata problem illustrates this block to solver
problem solving.
The
problem is depicted in
10.7 which shows
Figure
astandard checkerboard
opposite
with two diagonallyut off. Next
corner squares are a
to the checkerboard such
number of dominoes oftakes
dimensions that each
up exactly two checker
board squares. Intact check
erboards, you'll recall, have
64 squares. This one has 62.
Is there a way to arrange 31
dominoes such that every
checkerboard square is cov
ered by a domino?
The key to the solution is
to realize that whatever the
arrangement, each domino
will cover exactly one black
Square and one red square,
given the way checkerboards
are arranged. But now notice
that the two
are the sameexcised squares
color. Because
a domino
must cover twO problem. an 31 dominoes be arranged to cover
the
diferently colored squares, Figure 10.7: The mutilated checkerboard exactly two squares.
covers
there isno way to arrange 31 remaining checkerboard squares? Each domino
dominoes to COver the mutilated checkerboard.
 ofthe Laboratory
264 Cognitive PsychologyIn and Out
they
people have with this problem is that of fall to

probleim.ncludeThusthe
The difficulty most intheir initial
informationSimilarly,
representation
the at home)
resentation
pieces of is incomplete.
crucial in the baseball game(man
the problem in terms of a person sitting in a house
earlier, representing would be a case of using an
incorrect
down the wrong path. It presented in the problem and not correc
included information not
The choice of representation
studying problems such as the
can often make a great difference. S. H.
women-dogs-children-jobs repres talton-ong
problem of Box
the one in Figure 10.3were much m
solvingwho
ful inpeople
that constructed charts like
the problems than people who merely wrote down names, dogs,
soforth with arrowS or lines connecting them (for example, Cathy-Irish
retriever-four children).
020
more suCCAS
of a case in
Here's another well-known example
can make a
which the form of the representation t's called
problem either very easy or very hard. of each 6 7
objective
the "numbers game," and the of
player is to choose three digits from a set
digits such that the digits chosen total exactly
15.Two players are given a sheet of numbers, 1
23456 789. They take turns crossingone of 1 5
the digits off the list and adding it to their own
list. The first player to have three digits totaling
15(for example, 4, 5,6 or 1, 6, 8) wins.
Ifyou were to play this game, what would your 8 3 4
strategy be? lIf you played first, which digit would
you cho0se? What if you played second and
your opponent had first chosen a 5? The first
IFigure 10.8: A
time or two you play this game, you might find numbers game. tic-tac-toe board representation fte
it surprisingly challenging. Now look at Figure
10.8, which presents an alternative way of representing this game. Notice that, derict
this way, the difficult "numbers game" is actually the game of tic-tac-toe in disaua
Rendered as in Figure 10.8, the game is easy, but without this representation, the problen
is much harder to solve.

LACK OF PROBLEM-SPECIFICKNOWLEDGE OR EXPERTISE

Until now, we have been discussing general problem-solving abilities in terms of proÝ
lems that have a puzzlelike character. The assumption is that most of these problerns #
about equally unfamiliar to everyone and that people go about solving them in Dâsa
the same way. Other kinds of problems-for example, those in chess or oner
games; textbook problems in physics, geometry, or electronics; computer programmig
and problems in diagnosis- seem to be different in kind from the puzzles we have been
talking about. In particular, experts and novices approach most such problems different)
(Chi, Glaser, &Farr, 1988).
We saw in Chapter 3 that experts and novices differ in their perceptual abilties, wt
Efec's
experts able to "pick up on" more perceptual information than a novice WOuld.omain d
of expertise are not limited to perceptual with a
abilities, however. Familiarity
 the
Chapter 10: Thinking and
o w e d gseems
e
tO ch£nge
compare
way
the one solves Problem Solving 265
is to ability of probl ems within that frame of
gofessorS

ooblems
to design experiments.
connectedto
relevant
the Typi
task. Their
from the irelevant
cally, underg raduat
professors
experience in
earepsychol
much ogy
better majors
at
reference.
and their
tne
outt obe noticed. Experience also informat ion designing experimentssoving
and call to mind lets the
ootneed
t h enumber of participants to be provides
used,
a
number of variOrules
Us to them
hat
Astimating
pertormed,
the duration of the
experiment, the kinds
and so on. of shortcut
statistical situations
use in
be
problem
with alimited knowledge base are
analyses
clearly at a Problem solvers who come tocana that
Study of expert-novice
CassiCthe thinking processes of differences
both
disadvantage.
was carried out by
master players considered about thechess masters and
A
PXamined de Groot
thatthe best move more easily. Chase
same number of weaker players, (1965). He
the
chose moreexpertise a chess
player
and Simon
(1973), in a but finding
possibilities somehow
had, the more
thatthe
exposuresto chessboards set up to
brief master and chess beginner are both reflect an information replication
ongoing chess
he
extracted study, found
from even
shown a game. That is, when a
chess
Mmaster
reemember more about wherethe chessboard
to depict a possible chess game. pieces are placed-but only
will
configured
for 5
seconds, the chess
if the
pieces are
Simon (1996) examined the
Gobet anad
Professional Chess Association world sophistication of play of
champion, he played
as Gary
Kasparov, a
againstfour to eight opponents who were all
chess masters. His simultaneous games
allowed3 mi minutes per move (on opponents were each
average), while Kasparov was allowed one
eighth that amount of time for each move fourth to
simultaneously). Despite the tremendous time (because
one he was playing
constraints, multiple games
as
well he did under tournament Kasparov played
Conditions, where he wouldface only one almost as
Wee times as much time to think through and plan his opponentand
concluded that Kasparov's superiority came from his ability moves. Gobet and Simon
nfrom his ability to plan future moves. to recognize patterns more
Ihey based this conclusion on the fact that the
ine nressure of simultaneous games would
severely hamper Kasparov's ability to think
ahead, yet the overallquality of his play did not suffer.
Lesgold et al. (1988) compared the performance of five
frst-, second-,third-, and fourth-year medical residents asexpert radiologists with that of
they diagnosed X-ray pictures.
They found the experts noted more specific properties of the X-ray films,
more causes and more effects, and hypothesized
clustered more symptoms together than did any of
the nonexpert groups of medical residents.
Giaser and Chi (1988), reviewing this and other studies of expert-novice differences,
i er ownseveral qualitative distinctions between the two groups. First, experts excel
domains; that is, their knowledae is domain specific. Agrand master chess
Payer, Tor example, would not be expected to solve chemistry problems as well as a
st would. We have already noted in Chapter 3 that experts perceive larger meaning
Palterns in their domain of expertise than novices do. Experts are taster than noVces
at perforning skills in their domain of expertise, and they show greater memorial abilities
1or information
within that domain.
In
morproblem solving,levelexperts domain at a deeper
in their information and
e principled than do
see novices, a problem
who tend
and represent to represent superficially
Chi, Feltovich, &Glaser, 1981). For example, when solving physics problems, experts
 the Laboratory
266 Cognitive Psychology In and Out of
such
the problems in terms of physics principles, as Newton'sf
tend to organize tend to focus on the objects mentioned in the
motion; novices instead
an inclined plane or
process
who are ofmore
problem
africtionless

to
a problem,
surface.
trying
Experts
to
qualitatively analyzing plunge in and start looking for
likely solving,
also
spend proportionatelv
grasp or understand it, relative
solutions.
Finally,in their
experts are more likely to check for errors
oroblern,
Such

Expertise by itself
2000).ofPatient
study
is not always enough tor problem solving,
architect withalesion to the
right
as shown
prefrontal cortex
PF was a 57-year-old architect who suffered a grand mai(Goel
an experienced
tdramatihcraoluygiad
seizure

treated for a stroke.

to the prefrontal Subsequent
cortex,
scans showed a
MRIbrain
a part of the
predominanthy
&Gr
rignt-hemisphere
previouslyimplicated in deficits in the
and solve problems. Goel and Grafman asked PF. (and a control architect, matched
dtea
for
lab to develop anew design their lab space. Botth
and education) to come to the task as relatively easy.
the controlparticipant regarded
this PF. a
sophisticated architectural knowledge
PF. wasobservedto have "his
itquite skilfully during the problem
structuring phase" (Goel &
phase to the problem-solving
Grafman,base.2000,
phase. was unable
. intact,
p. 415)a.ndhe
s
PF. was unable to move from this through thetwo-hour session, and to generdea
preliminary design util twothirds of the way created:
was never developed or detailed. These authors an erai,
and minimal preliminary design that
that
the beginning of the designs
the preliminary represented il-structured problem solving (of the
chapter) and that PF's lesion "has resulted in a selective impaiment type descrcOnicbuedate
Computations" (o 421
neural system that supports ill-structured representations and

THE PROBLEM SPACE HYPOTHESIS

Researchers studying problem solving often

think about the processes in terms of mentally Initial state
searching a problem space (Baron, 2008;
Lesgold, 1988; Newell, 1980; Newell &Simon,
1972). The main idea behind this problem
space hypothesis is that every possitble state
of affairs within a problem corresponds to a
node in a mental graph. The entire set of nodes
OCcupies some mental area, and this area,
together with the graph, is the problem space.
Figure 10.9 presents a schematic diagram of a
generic problem space. Each circle, or node,
Corresponds to a certain state of affairs at some Goal states
point during the problem-solving process. If the
Figure 10.9: Ageneric problem space.
problem is to win a chess game, for example,
each node corresponds to a possible chess 'Corresponds
node labeled "initial state"
board configuration at each point in the game. The beforethe
chessboard|solved-tr
a
to the conditions at the beginning of a problem-for example, problem is
ths
first move. The goal states correspond to conditions when thestates (unlabeledin
example, configurations in which a game is Won. Intermediate
diagram) are depicted by the other nodes.
 Chapter 10: Thinking and
ntIsDOSSible mOve
from one state to another Problem Solving 267
to operation. that mOve can be
ymeans ofsomepaper-and-pencil representa-
inany space by a line Connect- Initial state
problem
nodes. Any sequence of "mental
the
of
on
two
shown as a
sequence of moves from
moves"s
another. Any sequence of moves
node
to initial state and ending at
a t the
one
naginning constitutes a path through
goalstate
final Figure 10.10 depicts a
heproblemspace.
the solution
path; Figure 10.11 depicts a
gonernic
problem spaCe for the Towers of Goal states
the
of
Hanoi
problem.
pat

problemsolvingis thought
to
be the cre- Figure 10.10:A solution path through the
problem space.
Good efficient paths: ones that are as short
ationof as few
pOssible and take
AS possible between the
detours as It
and the goal state.
Start
I state are
assumed the best paths
nitial
through see:arching, with
found being more
thorough searchessolutions.
kely to turn up
arti
Researchers in the field of
created
focial intelligence have
diferent search algorithms to
search through problem spaces
Nisson, 1998; Winston, 1992).
One is depth-first search, which
goes as far down a graph as
tcan to search for a goal state Stop: Note that this
before backing up to examine moves the bottom disk
aternatives. Another is breadth to the wrong peg.
first search, which examines all
nodes at a given level to search
for a goal state before delving
deper into the graph. Different
algorithms have different prob
ablfties of success
depending
Ue nature of the graph, of
on
cOurse.
Astudy by Burns and
R002) yielded some Vollmeyer
Finish

lve findings
U relevant to nonintui
the idea
Searching
spaces through Figure 10.11: Apart of the problem space for the Towers of Manoi problem, showing
to generate problem the solution.
These authors solutions.
mance. Moreover,believed
they believed, exploration
that exploration of a was
problemmorespacelikely when would yield better perfor-
the process was not
Curtailed by a person's eagerness to achieve aspecific goal.
 PsychologyInandOutoftheLaboratory
268 Cognitive
(2002) used the task depicted in Figure 10.12,
asked toimagine that
Burns and Vollmeyer they worked in alaboratory and that they were trying

howto control
by
various
changing
certain water-quality effect.
inputs to achieve a lime) and
inputs (salt, carbon, temperature). observing
what PTahteiyc patcsoud
iwhathnapputesnethe
task
at this (oxygenation, chlorine concentration, In
outputs example, a change in
the reality,
the o
the figure-for salt
relationshipsshowing in concentration--but participants were not
toldinput had
increase in the chlorine
ship between
in the figure).
input and output was (that is, they
weren't shown
the values produced
on the
they
Allparticipants were told to
would eventually be asked Salt
achieve a certain goal in terms +6
of specific values for the
outputs. Some of the partici
pants (called "specific goal"
OnygenatoN
participants) were given the Carbon
-0.5
specific goal at the start of the
task, but they were told they
wouldn't have to achieve this +2 +4
conNcentrale
goal until after an exploration
period; others (called "non Lime
specific goal" participants)
weren't told what the goal Temperatue
was until after the exploration Figure 10.12: The water-tank system(note that CII
concentrate is h concetck.
period. After the exploration output).
phase, all participants were
thehioine
given a diagram similar to Figure 10.12 but without any of the links
to draw links between inputs and outputs, placing directions andshown weights and onwereth aske
ifthey thought they knew them. From this, the researchers deriveda "structure s
to calculate a participant's degree of knowledge about the correct values of the linke
directions and weights.
Nonspecific goal participants received higher structure scores than specific goal partc
pants. Both groups performed equivalently when asked to achieve
the outputs, but nonspecific goal participants performed better specific goal values tu
with new goal values than did specific goal participants. Burns and on a transfer problen
follow-up study in which participants in both conditions were asked Vollmeyer
to think
(2002) dida
aloud asthej
performed the task. These results indicated that nonspecific goal
likely to test specific hypotheses during the exploration participants were mat
phase.
cificgoal can cut down on the amount of effort devoted to Presumably, having asy
searching the problem
Newell and Simon's terminology, although this can have its costs, depending spate,task
onthe
The problem space hypothesis can also be
to problem solving work. Searches that fail recruited to help
to explore partsusof understand n0w d
the space (because
mental set, for example) can block problem solving when the solution lies in apartofthe
space that isn't searched. Incomplete or incorrect representations are likely toresult
the construction of an incomplete or incorrect problem space, which in turn alsohats
problem solving.
 Chapter 10: Thinking and
Tme
nof
acquisitton expertiseis
another way to Problem Soving 269
develop hetter hunchesexplore
what aboutprobl whicehm areas
people to
tusefulto explore
and in
order spaces.of Expertise
l bemost

PERT
SYSTEMS
exploration is most liktheely toproblyieldempresum
space
results.
space hypothesis has been
h problem
e used to create
to model the judgments of
designed one or more expert systems, computer pro
amssystemscontain a knowledge base, which human experts in a
4
Fpert
bypicalyalso
s0archengine
contain a set of
that the program
inference
uses to
rules
search the
stores facts
(of the form
knowledge base using
relevant particular
within
the that
field.
"fXis true, then Yis field.
inference
They
nles,andsomeinterface, or means of interacting with ahuman user who has aquestion true"),or
for which he
or sheisconsulting expert system(Benfer, Brent, &
the
poblem
Furbee,
Oneexampleof an expert systemiss MUckraker, an expert system designed to 1991).
give advice
oinvestigativereporters regarding the best way to approach people for interviews, to
prepareforinterviewS, and to examine public documents while investigating an issue
Benferetal., 1991). Table 10.1 presents Some (simplified) rules MUckraker uses to give
adiceonhow
to approach a person for an interview.
Theformat of the rules used includes several antecedents, or conditions. Rule 2, for
example,hasthree antecedents: (a) The probable sOurce will not talk by telephone with
crucial; and (c) there are more than 6days in which to
reporter;(b)theinterview is
theinterview. antecedents get
Each of these specifies a condition that must be met for the
the activated.
letobe
conseguents are
pies also have a consequent part, indicated by the word THEN, These
actonsto betaken if the rule is
applied. For example, the action of Rule 2is to set avariable
certain value (namely, 80). Some rules alsoinclude an explanation or
(send by_mail2)to a by mail1,"
istfcation, preceded by the word BECAUSE. Notice the references to "send program. Rules
"send by mail2," and so forth. These are names of variables used by the
1through 4assign values to send_by_mail1 through send_by_mail4, respectively. Rule 5
variables have been assigned a value greater than
checks to see whether any of these four
interviewee a request by mail.
79. If so, Rule 5 directs the reporter to send the potential
Typically, one or more human experts
Ureating expert systems is a complex undertaking.They are often asked to generate a verbal
ntne domain are interviewed, often repeatedly.instances or solve problems (Stefik, 1995).
classify
ine protocol, thinkingaloud as they it is difficult for any expert to state all or nis
rat of the difficultycomes from the fact that
or her knowledge. Suppose
studying for academicexams, right?
For
rexample, you are probably "expert" at
pertainsto the activity of studyingfor
that
Isimply asked you to state all your knowledgedevelopers themselves
therefore often find "dotheir
an exam. Hard to do, isn't it? Expert system asthey
They follow experts around et a., 1991).
adopting | techniques from anthropologists. thinking as it happens(Benfer
thing" often asking themto elaborate ontheir are able to formulate rules such as those
Though repeated interviews, the developers
shown in Table 10.1.
270 Cognitive Psycnolo
MUckraker
Examples ofRulesFrom
Table 10.1 Simplified
Rule 4: Older_sources
Rule 1: Prefer_mail IF the age of the sourcee is >49
whethersource will talk AND the interview is critical years
IFunknown telephone AND there are >6 days to get the
with reporteron Critical
is interview
AND the interview
to get the
AND there are > 6 days THEN (send-by-mail), request 90
interview BECAUSE older individuals respond
request = 60
THEN (send _by_mail), more positively to written requests.
40
ELSE telephone request = with Rule 5:Combine_Send-by_mail
BECAUSE may get the interview
IF maximum of (send_by_maili > 79
a formal, written request.
prefer_mail THEN send written request and ASK:
Rule 2: Definitely
nottalk with Do you want tosee a sample letter?
IFprobable source will
reporter on telephone ELSE telephoning worth a try
AND the interview is critical BECAUSE most sources willtalk to a
AND there are > 6davs to get the reporter on the telephone.
interview

THEN (send_by_mail), request = 80

BECAUSE see Rule 1.
Rule 3: Telephone_anyway
IF probable source will not talk with
reporter on telephone
AND if the interview is not-critical
OR there are <6 days to get the
interview
THEN (send_by_mail), request = 10
BECAUSE there isn't time for mail
AND telephoning worth atry.
SOURCE: Benfer et al. (1991, p.6).
Why would anyone want to develop computerized expert systems and use them in pla
of human experts? One reason may be that the supply of highly trained human exoerss
limited in many domains. Not every city, for example, has an expert in every domain, a
ifthe knowledge possessed by experts can be distributed through software, the weat
is spread.
Asecond argument willbe further elaborated in Chapter 11. There we willsee that hut
decision making is often, if not always, tainted by biases, some quite insidious. Espaci
when a problem is complex, with many factors, the cognitive load placed on te pesu
facing the problem can quickly become overwhelming. Having an expert systei
gaining a handle on the complexity and preventing it from crushing the procesSs offindi
the best solution.

FINDING CREATIVE SOLUTIONS

Many of the problems psychologists ask people to solve require insight, a changei
oganted
frame of reference or in the way and
elements of the problem are interpreted
 Chapter 10: Thinking and
which insight occurs is not well Problem Solving 271
called creatunderivistyto.od. Whatever it is, it
is
rolein what
commonl y
originalitynevchol ogis ts
precisely, many
that suits some agree that creativityAlhasthoughto the term is appears
tis, Runco purpose do with difficut
Akett

are
Simonton, 2008; 2004;
mundane,; Conversely. original ideas
&

Si m ont o (
n, H ennessey
2011).
&
Amabi le, 1988;
that do not Appropriate ideas that lack
apprNioprelisaen,te
bizarre. Other cognitive
Ovlty
are
scient ists talk of creativaddress some problem in a
"glreadyhas," iknowlof,edge,
way
recombination, of
etl or
r
hation, knows ity as
or hasintormation, or mental consisting of a com-
musical,scientific,orother
depicted (Dartnal, 2002). re present atio ns- things
Eureka!"experience when the discoveries often seem to share acritical
Gestartistic

presenteda review of prover bial"l ig htb ul b "

gmental
tcomposers,aartists,
scientists, and other
eminent goes on. Many moment,
1981,
something that
experts
some of these). Such begin with "Eureka!"
stories lead to
biographi
storieses
peoplehave less the notion that
peative Work very different
pOcesses in ways (at creative people don't have or that
least while they are being their cognitive
slesscreativepeople. creative) than those
will focus ontwo types of
section, we
nthis creativity as special
describes cognitive explanations
processing one that
and
for creative insight: one
that
cognition, describes it as the result
ofnomal,everyday
UNCONSCIOUS PROCESSING.
GANDINCUBATION

Asa collegejunio, I| took cOurses in calculus, which,

although extremely useful, were
otenextremely frustrating to me. I would work on a homework
Oneofthe problems absolutely unworkable. The problem would nag at
assignment only tofind
me, and try
technique I could think of. In frustration, I would put the problem aside and l'd
go on
woher things. Late that night, sometimes waking from sleep, Iwould see the problem
ina whole new light. Often Ihad discovered the correct solution. (On occasions when l'd
hton another incorrect solution, my feelings of frustration were renewed. On those occa
sions, Iwas known to grab the textbook and fling it against the nearest wall., The book did
not survive the semester)

he experience lam describing is a "textbook case" of unconscious processing, or

incubation. The idea is that while my mind was actively running other cognítive pro
ases, Some other sort of processing was happening in the background. (Those of you
MO IKe computer metaphors might describe this as "batch processing" as opposed to
"interactive processing.") The unconscious processing churned away, even as I slept,
Ue answer was found: then the answer announced itself allat once, evenfit nad
to wake me up. Those who believe in incubation typically believe in the existence of
an unconscious layer of the mind that can process information without giving rise to
COnscious awareness.
S.M. Smith and Blakenship (1989) ofered one empirical demonstration of incubation
efsolved
tects by15 rebuses,
means oftheypicture-word a 16th,rebuses.
puzzleswithcalled
were presented
After the participants had
which had a misleading cue that
nduced fixation on an incorrect interpretation. They were later giventhisandcritical rebus a
stheeCondcue. timControl
e, without
participants
the cue, saw
andthe again
second
were asked to solveofthe
presentation the
puzzle
rebus
alsoto recall
immediately, but
experimental participants received either a 5- or 15-minute"break" from the
puzzle, when
 Laboratory
272 Cognitive Psychology In and Out ofthe
askedto complete a
they either did nothing or wereto work on the rebus demanding
music
surreptitiously). Smith
percepiOn
prevent them from continuing
were given longer "filled" intervals
predicted that those who
(duri
was presented) would be more likelyto forget the misleading cue and ng
rebus. In fact, this pattern of results is exactly the one they reported.
which and Blala
to find positive effects of
Most empirical studies, however, fail during problem
ability toincubat
solving, solveion,
breaks
who take physical and mental rarely show increased and who
incubation,
more opportunity for
thoroughly or more quickly than
participants who work
on steadily
incubation at the
1979). Moreover, participants in another studythought aloud abot
th e r i
effects reporprotdeermd
ore
ing the "break" periods, they surreptitiouSly
another experimental condition who during the breakthe
in incovert thinking about the problem (by having to memorize a
participants
engaging were problem.
showed very few effects of incubation (Browne &Cruse,
1988)
Designing critical tests of the incubation hypothesis is very difficult:
preletet
make sure participants really do cease consciously thinking about the
incubation interval, a challenging task for experimenters who cannot read problem
EVERYDAY MECHANISMS
minds
Does creative insight depend on special cognitive processes, such as
alternative view asserts that it results from ordinary cognitive processes
every person uses in the normal course of life (Perkins, 1981). Perkins's that incubation?
vitu
A

coherent overview of this approach to the study of creativity and will be iddeas
prioidea
reviewed
here. Other authors offer slightly different proposals but share Perkins's iteO
processes leading to creativity are not extraordinary(Langley &Jones, 1988;
1988;Ward, Smith, &Finke, 1999; Weisberg, 1988). Serber,
Perkins (1981) described examples of cognitive processes that underlie normal evsn
functioning as wellas creative invention. One such process is directed remembeto e
is the ability to channelyour memory in order to make conscious some past experiencen
knowledge that meets various constraits. The first task in this chapter, asking youtbot
of foods and drinks whose names begin with c, isa directed-remembering task. Petis
argued that the same process goes on in creative invention. Darwin's construction oteN
lution theory, for instance, had to provide an explanation consistent with existing scert
knowledge. That knowledge constrained the types of explanations he could develop.
Asecond relevant cognitive process is noticing. An important part of creation, artists at
scientists assert, is revising drafts. In revising, one needs to notice where the probent
are. Noticing also plays a role in many "Eureka!" or "Aha!" experiences, accordingu
Perkins, in which creators notice a similaritybetween one problem and anoTiel.
as Some
Contrary recognition, or the ability to recognize objects not for what they are but
thing else, is another important creative process. Seeing a cloud as a castleisafamila
the creatort
example. This ability obviously
move beyond the bounds relatesof towhatanalog
of reality, is, andthinking
inthat it requires
to imagine reality in other ways.
samecogh
This approach to creativity, then, assumes that creative individuals Usethethat asto
tive processes that so-called noncreative people use.
Its proponents arguefollowingthë
of insight" actually occur in aprogressive, step-t-by-stepfashion. Incubation,
 to do with
Chapter 10: Thinking and
argumenthas
olda0proaches that did not work. maki
Notengthat fresh start on
a Problem Salving 273
neof
this the
itnmeans
fwhat
to break a mental set.

relationship between description is solquiteutionsimiprocess,

problemideasolvofinga and the lar to forgetting
ndae

p
the,
i s strong. Both include the
COativityy
that meet various requirements or cont
mental search raryforrecognition descriptions
forconstraints. person'spossibilities approach
and
OVelwillirngnessto search harder and
withaconstitutes longer
creativity, then, are a
solu tions
A
that meet creativity hasthatto aredo
What withstand potentialy long periods
#blityto
creator's own values for originalmul, tiple constraints.
without success,
propOsed accounts of creativity and plans and useful results: an
Many question of whether acts of remain, for the abilities.
ThUS,theremains open, asS researcherscreativity use most part,
investigate creativity (Runco & struggle to
processes untoreregular
special-purpose sted empirically.
dsto develop
should be seen as ideas that can Sakamoto, 1999). Theappropriate cognimeth
tive
guide empi rical
proposals just described,
developedtheories that have survived
rigorous future
then,
testing.investigations rather than as well-
CRITICALTHINKING

Much creativityhinges on the ability to

the wall" or "out of
seem "off touch." generate a number of ideas that might at
evaluated and assessed in terms of Once
a novel idea is
its generated, first
be
really meet allthe objectives and constraints?appropriateness. Does the however, it must
subtleproposed
Are there hidden or
what are the proposal's implications? flaws in thesolution
idea?
Aperson asking these kinds of questions can be
philosophers, and educators call critical described as doing what
exist. Devwey (1933), who called it
nd careful
thinking.
"reflective Many definitions of critical
thinking," thinking
defined it as "active, persistent,
psychologists,
consideration of any beliet or
the arounds that support it and
the further supposed
conclusions
form of knowledge in the light of
istinguished between reflective thought and other to which it tends" (p. 9), Dewey
beliets for which a person has no kinds: random ideas, rote recall, or
evidence.
Wertheimer (1945),a Gestalt
psychologist, presented
several examples that illus Altitude
trate critical thinking
well. One quite
concerns
ing how to find learn Base
a the area of (A) (B)
0 paral
teachleSomeone
logram. Onehowwayto
do this is by
lOrmula, such asteaching
the
a
one from famil
Qeometry: Areahigh= Base
school
Alti ude. Figure 10.13(A)x
presents an
(C) (D)

þarnd aaltitude example

l elogramlabeled.
with the base
of a Figure 10.13:Parallelograms and other geometric hgures.
 Laboratory
274 Cognitive Psychology in and Out of the
If the student memorizes this formula carefully, he will have a "rote"
repetitive" means. of
Wertheimer (1945) called a"mechanically if a student
that
problem with rote solutions, however, is Wertheimer the
forgets solthevinformul
g a.
at a complete loss. A
grasp the "essential
mental issues.
better
structure"
approach,
of the problem-to identify

Consider the parallelogram in Figure 10.13(B),

noting the
argued,
is to teach
and understand
shaded area.
ortetnh,seGiuhderhtuna
the
to the other side, as
is cut off the parallelogram and added familiar and simple
transformation of the object creates a
The formula for finding the area of arectangle is well
known
area to the
geometric Su
objp
shown in Figure
e o
cts
, e
(Base Altitu1de)0.13.0,Te
x
th i s
the transformation has added exactly the same has not right side of the
subtracted from the left. As a result, the total area
"regular" geometric figure, with exactly the same area, has
been changed.
been figure Noieta
created, Instea, ance s
What is the advantage of teaching this method of solution2 For one thing, Fo
generalizable. The method applies not just to parallelograms but to many
irregular figures, such as the one depicted in Figure 10.13(D).
For
shows a deeper understanding of why the formula works. In this another. the
is not simply blindly applied to the problem but rather grows out of the the geometrica
instance, Tosomltua
standing of the nature of a parallelogram as a geometric object.
In a more recent study of critical thinking, David Perkins and his
student 's under
&Hafner, 1983) presented students and adults of various levels ofcolleagues
educational
with various controversial issues. Participants were asked to reasSon aloud
issue. One example was "Would alaw requiring afive-cent deposit on bottlesabout eah
(Perbkaicnksg.yAouleda,
reduce liter?"(Perkins et al., p. 178). The authors measured critical and cars
by
the number of times a participant raised objections or challenges to his Own lookinga thinking
example of what the researchers would consider good critical thinking is the thinking, An
fol owing
The law wants people to return the bottles for the five cents, instead of littering them. Butl dont
five cents is enough nowadays to get people to bother. But wait, it isn't just five cents ata blowbeite
people can accumulate cases of bottles or bags of cans in their basements and take them bakdl
once, so probably they would do that. Still, those probably aren't the bottles and cans that oet ite
anyway; it's the people out on picnics or kids hanging around the street and parks that iter botte:ad
cans, and they wouldn't bother to return them for a nickel. But someone else might. (p. 178)
Notice the structure of the thinking here: Each sentence in some way presents an obje
tion to the previous one.
According to Perkins, good thinking requires a large knowledge base and some mearsd
using itefficiently. Good thinking also requires the kind of objection raising just ilustatt
showing the thinker actively trying to question herself and to construct examples a0
cOunterexamples with respect to her conclusions. What often hampers critical thinking
kind of mental laziness-stopping thinkingwhenever you get any answer at al o
cal thinker might reason about the bottle billas follows: "Well, it'd be nice to try tocase,the
litter, but five cents won't motivate people, so it won't work." Notice that in thisanyofthe
person constructs one mental scenario and then stops, without questioning peah
assumptions or trying to think of any other possibilities. Perkins et al. (1983) urged sers
Superficial
to overcome this tendency to think about an issue only until things make interpretations
and, instead, to search harder and look longer for other possibilities and
 Chapter 10: Thinking and
Problem Soling 275

Thinking,the manipulation
different
Summaryof
information, occurs for a wide
appear be very
to
tasks.
example, Psychologists draw range of what
types of problems (tor between well-defined anddistinctions among
typesoffthinking (for example,
and among
clear,however,
whether the
cognitivefocused versus unfiol-Ocused)
definedIt ones)
. tasks
is not
yet
themselves really differentin kind. An processes used for different
are
looklike
different kinds of alternatfromive possibilfty is that what
thinking really stem
same cognitive processes. different
combinations of
psychologists studying problem
Some solving have discovered
2. strategies(for example, generate-and--test,
means-ends general
ingby
analogy) that they believe analysis,
people use in a wide variety reason-
of situations.
Psychologists have alsO explored different blocks to problem solving
functional|fixedness, incorrect or
set, (mental
incomplete problem representations),.
for
Other psychologists argue the importance of adomain-specific knowledge
and strategies as better predictor of
a whether given person
problem. These will have suc-
cess solving a given investigators point out that problem-
solving strategies often vary with the expertise, or background knowledge,
solver.
of the problem
4
Cinilarities among kinds of thinking can lso be idetified. Some psycholo
gists arguethatthe similarities can be explained byacommon framework: the
that allsorts of instances of thinking (including episodes of problem solv
ing, inventing, and even reasoning and decision making, covered in Chapter
12) are alla kind of mental search (Baron, 2008). This proposal accepts the
oroblem space hypothesis, or something close to it, asa good account of
how people mentally manipulate information. The problenm space hypothesis
views thinkingas finding a path through a "mental graph" of possibilities (the
mental graph being the problem space). Sometimes the search for a path is
very focused and constrained; at other times (for example, during episodes of
daydreaming), it meanders without a definite goal.
5. Expert systems, computer programs designed to mimica human expert in a
specific field,are one kind of instantiation of aproblem space. Expert systems
contain a knowledge base, inference rules, some means of searching through
the knowledge base, and a user interface so that the human user can ask
questions and be queried to provide the program with more intormation.
OPsychologists studying creativity differ over whether there is one general cre
alvy, independent of domain, or whether creativity, like expertise, is specitic
0 a domain. Some arque for special-purpose creative cognitive processesS,
that creativ
uen as incubation and unconscious processing: others believe
such as directed
ity makes use of everyday, ordinary cognitive processes
remembering and contrary recognition.
7 to promote good
Some psychologists have argued Ithat thefactors that seem others. Among
performarnce on one type of thinkingtask also seem to help on possibilities,
these factors are remaining open-minded, exploring avoidunusual
bias, and trying
questioning the first conclusion you come to, trying to
to find new and fresh
approaches.
 and of
Out theLaboratory
In
CognitivePsyhology
276 thinking skills
argue thatsuggestedis can substitute
one
no most
would that good thinkingt0r
a
8.
Although
you get the
anddeep
knowledge of thewhat
outbase, is
knowiedge you have This suggestion,
educators. sk l s brhaedto
anecdotal proposals from philosophers, and
used in all kinds
largely on future research on the processes
gists, awaits pthinskyicnhgo,lo
of
Review Questions
1. Do well-defined and ill-detined processes make use of the same cognitive
cesses? How might psychologists go about trying to answer this question?
Compare and contrast the generate-and-test. the means-ends analysis, and the
problem solving.
2
reasoning-by-analogy approaches to
3. What might the Gick and Holyoak results on reasoning by analogy suggest abot
theoretical principlesinreal-world situations? Explain.
people applying
In what ways is mental set similar to perceptual set (described in Chapter 3 n
4. phenomena dissimilar?
what ways are the two
differences in problem solving
5. Describe some of the expert-novice
Discussthe problem space hypothesis. How might it account for and explain the
6
various blocks to problem solving?
differences between various possible problem:
7. Explore the connections and solving
strategies reviewed in the chapter.
8. What kinds of cognitive processes have been proposed to account for oresk.
ity? Howcan an experimental psychologist test the role of any one of thesa
processes?

Key Terms
backtracking il-defined problem problem space hypothesis
creativity incubation reasoning by analogy
critical thinking inference rule thinking
expert system means-ends analysis unconscious processing
functional fixedness mental set well-defined problem
generate-and-test technique perceptual set working backward
GPS (General Problem Solver)
problem solving