You are on page 1of 19

Journal of Experimental Psychology: General Copyright 2004 by the American Psychological Association

2004, Vol. 133, No. 3, 415– 433 0096-3445/04/$12.00 DOI: 10.1037/0096-3445.133.3.415

Having the Memory of an Elephant: Long-Term Retrieval and the Use of


Analogues in Problem Solving

Zhe Chen Lei Mo


University of California, Davis, and South China Normal University
South China Normal University

Ryan Honomichl
University of California, Davis
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

The authors report 4 experiments exploring long-term analogical transfer from problem solutions in folk
tales participants heard during childhood, many years before encountering the target problems. Substan-
tial culture-specific analogical transfer was found when American and Chinese participants’ performance
was compared on isomorphs of problems solved in European versus Chinese folk tales. There was
evidence of transfer even among participants who did not report being reminded of the source tale while
solving the target problem. Comparisons of different versions of a target problem indicated that similarity
of solution tool affected accessing, mapping, and executing components of problem solving, whereas
similarity of goal object had only a moderate effect on accessing. High school students also evidenced
greater transfer than did middle school students.

Suppose an individual is faced with an insight problem that ago. Although this seems intuitively plausible, little is known
involves discovering the quantity of glass marbles necessary to regarding the circumstances that might facilitate remembering and
equal the volume of a metal statue. Within this scenario, the utilizing such a story after a long period of delay. What difficulties
individual is provided with several ad hoc items, including a large might individuals experience in solving analogous problems after
container filled with liquid, as possible tools for solving the prob- a long delay? Would the likelihood of success in using
lem. The solver might be reminded of the Eureka story regarding Archimedes’ principle differ if the target problem involved a
Archimedes’ discovery of a solution to the problem of verifying similar (crown) or dissimilar goal object (e.g., a metal statue)
the authenticity of the king’s gold crown. The Archimedes story, and/or a similar (a bathtub with water) or dissimilar solution tool
illustrating the principle of water displacement, could function as (e.g., container with another type of liquid)? Would these two
a potential source analogue, even if the solver had heard it years types of object similarity influence problem solving in different
processes and to different degrees?
When facing a novel problem, people often search for relevant
Zhe Chen, Department of Human and Community Development, Uni- circumstances or similar scenarios from long-term memory to
versity of California, Davis, and Center for Studies of Psychological solve it. Although much laboratory research has demonstrated that
Application, South China Normal University, Guangzhou, China; Lei Mo,
people can use information from a source problem in solving
Center for Studies of Psychological Application, South China Normal
University, Guangzhou, China; Ryan Honomichl, Department of Human analogous problems (e.g., Bassok, 1990; Catrambone, 1996; Gick
and Community Development, University of California, Davis. & Holyoak, 1980; Reed, 1987; Ross, 1989), other studies have
The research was partially supported by Grant HD 19011 from the shown that people often fail to use analogous solutions that they
National Institute of Child Health and Human Development. Data collec- are exposed to even immediately prior to being asked to solve a
tion in China was also supported by Grant 01JAXLX001 from the Chinese target problem (e.g., Reed, Ernst, & Banerji, 1974). It has been
Education Ministry. We express our gratitude to Bob Siegler and
suggested that individuals might experience more difficulties in
Myeong-Ho Sohn for invaluable comments on an early version of the
article. Thanks are extended to Jianfang Chang, Hongmin Chen,
solving problems by analogy with longer delays between the
Zhansheng Chen, Xinkui Chen, Ying Leng, Zhiya Liu, Chen Qu, Xuili acquisition of a source analogue and the presentation of a target
Tong, Ruiming Wang, Yanping Xu, and Qing Zeng at South China Normal problem (e.g., Barnett & Ceci, 2002). However, little is known
University and Deloris Abu-Tair, Elizabeth Fallapino, Karen Ho, Dannelle about whether individuals can actually retrieve and use analogues
Larsen-Rife, Denise Lew, and Ngoc Long at the University of California, acquired many years prior to encountering an isomorphic problem.
Davis, for their assistance in data collection and entry. We also thank Susan Because human thinking and problem solving are unquestionably
Goetz for editing an early version of the article and Yuri Rodriguez for her
influenced by experience, it seems reasonable to assume that
skillful creation of the figures.
Correspondence concerning this article should be addressed to Zhe
individuals can use prior examples from long-term memory to
Chen, Department of Human and Community Development, University of solve relevant problems. Therefore, in the present research, we
California, One Shields Avenue, Davis, CA 95616-8523. E-mail: examined three issues: (a) the ability to analogize with a substan-
zhechen@ucdavis.edu tial time gap, (b) factors that influence long-term transfer pro-
415
416 CHEN, MO, AND HONOMICHL

cesses, and (c) developmental differences in long-term transfer investigate in the present research a very remote, challenging type
among middle school versus high school students. of transfer: long-term analogical problem solving across physical
Our primary goal in the present research was to explore whether and functional contexts. Furthermore, demonstration of the com-
and how individuals retrieve and use analogues from long-term mon key processes that are evident in short- and long-term transfer
memory to solve problems across contexts. Almost all laboratory would attest to the generality of the mechanisms that mediate
studies of analogical reasoning have involved presenting source analogical problem solving.
analogues and then observing how participants map source and The second goal of the present study was to explore how
target passages or solve target problems over short delays mea- different types of object similarity, namely, goal object and solu-
sured in minutes (e.g., Bassok, 1990; Chen, 2002; Needham & tion tool similarity, influence different transfer processes. Much
Begg, 1991), hours (e.g., Wharton et al., 1994), days (e.g., Ca- empirical and theoretical research in transfer over the past century
trambone & Holyoak, 1989; Keane, 1987; Schunn & Dunbar, has focused on the effects of similarity shared between source and
1996), weeks (e.g., Catrambone, 2002; Gentner, Rattermann, & target problems (e.g., Anderson, Farrell, & Sauers, 1984; Holyoak
Forbus, 1993; Wharton, Holyoak, & Lange, 1996), or months later
& Koh, 1987; Holyoak & Thagard, 1997; Osgood, 1949; Reed,
(e.g., Chen & Klahr, 1999). Virtually no studies have examined
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1987; Ross, 1987; Salomon & Perkins, 1989; Singley & Anderson,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

whether individuals are capable of using an analogue to solve a


1989; Sternberg & Frensch, 1993; Thorndike & Woodworth,
target problem after years of delay.
1901). Traditionally, similarity between analogous problems has
It also has been typical for researchers in previous studies to
been distinguished as either structural or superficial (e.g., Catram-
present source problems and then test target problem-solving per-
formance in the same setting (i.e., laboratory). However, in a rare bone, 2002; Gentner, 1989; Gentner et al., 1993; Holyoak & Koh,
study examining transfer across context, Spencer and Weisberg 1987; Ross, 1989). Superficial similarity refers to objects and their
(1986) showed that participants experienced difficulty in solving a properties, story protagonists, or story lines common to both
target problem in a setting (laboratory) that was different from the problems, whereas structural similarity refers to the underlying
context (classroom) in which a source problem was learned. The relations among the key objects shared by the source and target
study demonstrated the important role of contextual similarity in problems (Catrambone, 2002; Gentner et al., 1993). Although
analogical transfer. extensive investigation of this dichotomy has yielded fruitful find-
Although it is important to pinpoint the mechanisms involved in ings, theorists have advocated for making finer distinctions of
analogical problem solving in a laboratory setting, rarely in every- types of similarity and their relations to transfer performance (e.g.,
day life does a person learn a concept or solution strategy and then Anderson et al., 1984; Gentner et al., 1993; Klahr & Carver, 1988;
apply it immediately afterward to an analogous situation in the Reder & Klatzky, 1994; Reed, 1987; Ross, 1989).
same setting. Real-life situations often involve transfer of knowl- Previous studies in analogical problem solving have also iden-
edge or strategies from one physical (location) and functional tified three key components involved in transfer. The first process
(activity) context to another with a substantial delay. Thus, it involves accessing a source problem. To solve an analogous prob-
seems obvious that extending previous findings regarding analog- lem, a problem solver needs to retrieve or be reminded of a source
ical transfer to more naturalistic contexts and with more extensive analogue (e.g., Ross, 1989). Spontaneous retrieval of relevant
time gaps between problems would have both theoretical and information from long-term memory may be especially challeng-
practical implications. To the best of our knowledge, the only ing. The second process involves mapping the structural relations
published studies exploring analogical transfer in naturalistic sit- between problems before the source solution can be applied to
uations are those of Dunbar and colleagues (e.g., Dunbar & solve the target problem (e.g., Gentner & Toupin, 1986). Mapping
Blanchette, 2001). These researchers have observed the use of the relations and corresponding objects between source and target
analogies by leading molecular biologists and immunologists dur- problems after a lengthy delay may lead to an obstacle in transfer.
ing laboratory meetings and found that scientists often generate A third process involves executing a solution or strategy (e.g.,
analogies in solving problems during scientific discovery.
Catrambone & Holyoak, 1990; Chen, 2002; Chen & Mo, 2004;
The paucity of studies examining analogical transfer over long-
Reed & Bolstad, 1991). Few studies have explored all three of
term delay in less artificial contexts may be due to methodological
these components simultaneously in the same task context, as was
difficulties. In particular, it does not seem realistic or practical to
done in the present research.
present an analogue to participants and then to test their target
We propose a framework for examining the relationship be-
problem-solving performance after years of delay. To overcome
this methodological limitation, we adopted a novel paradigm that tween different types of object similarity, problem-solving perfor-
involved examining the target problem-solving performance of mance, and transfer processes. Object attributes shared between
participants who had acquired a source problem and its solution in problems have been viewed as superficial similarities that contrib-
various cultural contexts years ago during their childhood. Another ute primarily to the accessing process in transfer. We propose that
group that had not experienced the source analogue served as a finer distinctions between types of object similarity can be made
control group. We hypothesized that differences in target problem- and that different types may influence various processes involved
solving performance between these two groups would reflect the in analogical problem solving. This framework is based on the
effects of the available source analogue on problem solving. The following assumptions: (a) problem goal structure is composed of
effects were also measured by examining participants’ reports of different entities or elements such as a goal, obstacle, solution, and
the target problem reminding them of any story or problem and the outcome; (b) the entities of a problem are typically associated with
helpfulness of what they were reminded of in solving the target objects; (c) these objects associated with the different entities may
problem. Thus, with a novel approach, we were motivated to be similar or dissimilar between analogous problems; and (d)
LONG-TERM ANALOGICAL PROBLEM SOLVING 417

similarities of objects associated with different elements in the Experiment 1


problem goal structure may have a distinct impact on performance
and processes. Our primary goal in this experiment was to investigate the
In some of the studies reported here, we manipulated the retrieval and use of source folk tales acquired in naturalistic
source–target similarity of goal object and solution tool and ex- settings (e.g., reading a book or listening to a story) for solving
amined their impact on different transfer processes. We hypothe- analogous problems in an experimental setting after a long delay.
sized that similarity in goal object would influence the accessing To determine the effects of culturally specific experience on long-
process. In contrast, similarity in solution tool was hypothesized to term transfer across contexts, we chose folk tales from two differ-
affect not only the accessing process but also the mapping and ent cultures, China and the United States, as source analogues.
executing components. When a common solution tool is shared by Target problems, which were isomorphic to the source tales, were
created.
problems, the tool can be readily chosen to solve a problem by
One target problem described a scenario in which a chief of a
mapping it to the source analogue. However, when solution tools
riverside village needed to find a way to measure an amount of
are similar in function but differ in object attributes, participants
gold commensurate in weight to a statue without the benefit of a
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

may experience difficulty in matching the source and target tools.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

conventional balance scale. This problem was designed to be


The third and final goal of the present study was to examine
isomorphic to a well-known Chinese tale called “Weigh the Ele-
developmental differences in long-term analogical transfer. Very phant.” The solution involved a principle of equalizing the weights
young children display a wide variety of transfer competencies of smaller items and a large object by compressing a boat to the
(e.g., Brown & Kane, 1988; Chen & Siegler, 2000; Holyoak, Junn, same degree (see Appendix A). Another target problem was the
& Billman, 1984), but analogical problem solving is nevertheless cave problem, which was created to be analogous to the tale
thought to develop in efficiency and sophistication during child- “Hansel and Gretel” (H&G), by the Brothers Grimm, which is
hood (e.g., Brown, 1989; Chen & Klahr, 1999; Gentner & Toupin, commonly read or heard by children in the United States (see
1986). Yet few studies have examined age differences in the Appendix A).
mastery of all three components or in long-term analogical trans- Although the target problems and the source analogues involved
fer. With the present research, we explored several pivotal aspects similar solutions, they shared few contextual and superficial fea-
of developmental differences in long-term analogical transfer. tures such as similar objects and characters. Chinese college stu-
More specifically, how does object similarity affect long-term dents, who had experienced the elephant tale, were predicted to be
transfer at different ages? If adults are successful in solving an more likely to come up with the compression solution to the statue
insight problem over a long delay, even with a dissimilar goal problem than were U.S. students, who had never heard the source
object and solution tool, would children and adolescents also be tale. Similarly, U.S. students, who had heard the H&G story, were
capable of analogizing when the source problem was acquired predicted to be more likely to solve the cave problem by suggest-
years ago? Furthermore, if obstacles associated with various trans- ing that a trail of small objects be created than were Chinese
fer processes are evident in adults, it would be important to explore students.
developmental differences in difficulties associated with these To assess the availability of the source analogues, we conducted
processes and whether similarities in goal object and solution tool a preliminary study with 60 U.S. students and 30 Chinese students.
would have comparable effects on problem solving in younger The Chinese students were asked to write down the elephant tale.
participants. These were the main issues we addressed in the Despite the fact that most reported hearing the story many years
present research. ago, they remembered the key elements of the story. Among these
We conjectured that college students and high school students 30 students, 97% remembered the goal of the problem, and 93%
were more likely to solve an analogical problem after a long delay wrote down the correct solution. The U.S. participants were asked
than middle school students were. Furthermore, developmental to write down the gist of the H&G story and then answer the
question, “How did Hansel and Gretel find their way out of the
differences were hypothesized to reflect the difficulties that par-
forest?” Seventy-three percent recalled some portion of the tale
ticipants experience in accomplishing each cognitive component at
and 52% remembered the strategy involving creating a trail of
different ages. Younger students, compared with older students,
small pebbles or bread crumbs. Despite the differences between
were predicted to have greater difficulty in spontaneously retriev-
these two samples in recalling the source tales, these results
ing a source problem, in mapping the key objects between prob-
pointed to the availability of the source analogues and their solu-
lems, and in implementing the solution. tions, a prerequisite for examining whether and how participants
In Experiment 1, we explored whether Chinese (or U.S.) par- use analogy after long-term delay.
ticipants who had experienced a source analogue would be more Four additional insight problems were chosen as neutral tasks,
likely to solve a target problem than their peers in the United States because there were no known analogues to these problems in either
(or China) who lacked exposure to the analogue. In Experiment 2, culture: the radiation problem (Duncker, 1945; Gick & Holyoak,
we examined the effects of similarity in goal object and solution 1980), the string problem (Maier, 1931), the water jar problem
tool on transfer performance and processes. In Experiment 3, we (Luchins, 1942), and the pool problem (Gardner, 1978). These
examined the accessing process without the potential interference control problems were used to test the hypothesis that performance
of problem-solving activities. In Experiment 4, we explored on insight problem solving would be equivalent between the two
whether middle school students experienced more difficulties in cultures. These problems were selected on the basis of several
solving the target problem than high school students did, especially criteria: (a) They are novel to students in both cultures, (b) they are
in the dissimilar object conditions. insight problems that are similar in nature to the two target prob-
418 CHEN, MO, AND HONOMICHL

lems, and (c) the difficulty level in generating solutions varies the control problems, the participants were asked to indicate if and where
among the problems. they had heard of the problem before. After attempting to solve all six
problems, the participants were presented with the two target problems
Method again and asked whether each of the problems reminded them of any story,
event, or problem. If either or both did, they were asked to write down the
Participants. A total of 270 undergraduate students participated in this gist of what they were reminded of. The six problems were presented to
experiment: 152 undergraduate students from either psychology or educa- both samples in one of two orders. The order for presenting the target
tion classes at South China Normal University, Guangzhou, China, and 118 problems was counterbalanced. In one order, the statue problem was
students from psychology or human development classes at the University presented first and the cave problem last; in the other, the order was
of California, Davis. Although the absolute academic standards of the U.S. reversed.
university may well be higher than those of the Chinese university, these
institutions are major universities within their respective higher education
Results
systems. Although a lower percentage of the Chinese population attends
college, this percentage has increased substantially over the past 2 decades. Responses to each problem were scored as correct or incorrect.
Overall, the two samples appeared to be comparable for the purposes of the The performance of 140 randomly selected participants was coded
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

present experiments. Approximately equal numbers of women and men in independently by two coders, and agreement on problem perfor-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

each sample participated in the study. The scores of an additional 10


mance and reminding was 96% on average, ranging from 93% to
participants in the United States and 12 in China were excluded because
they reported hearing of one or more of the control problems before
98% among the six problems.
participating in the experiment. Participants received course credit for Accurate identification of a source analogue to the target prob-
completing the experiment. lem (i.e., the cave or statue problem) was scored as 1. In the cave
Materials. The materials consisted of six insight problems involving problem, participants were given credit if they reported being
different types of solutions with varying levels of complexity. Two of these reminded of H&G or a story involving dropping bread crumbs or
problems, the statue and cave problems, were target tasks. The other other small objects to make a trail to follow home. In the statue
four problems were control tasks because there are no known relevant problem, reminding credit was given if a participant mentioned
source analogues in either culture. The target problems are provided in aspects of the original elephant tale. Agreement between the two
Appendix B. observers was 94% and 99% for the cave and statue problems,
The statue problem described a scenario in which a chief of a village by
respectively. Preliminary analyses revealed no gender differences
a river needed to use the weight of a stone statue to approximate the
amount of gold coins that he must collect in taxes from a neighboring
in any measure (all ps ⬎ .20). Data were analyzed to address the
village. Because a conventional balance scale was not available, the chief issues of whether Chinese and U.S. students solve the statue and
faced a problem of matching the weights of the statue and the gold coins. cave problems differently and whether they solve the control
The correct solution involved putting the statue in a tub (which was problems similarly. All reported effects are significant beyond the
mentioned in the problem) and placing it in the river to show how much the .05 level, unless otherwise noted.
tub was compressed in the water. Then, after the statue was removed, gold Performance on the statue problem. As Figure 1 shows, 69%
coins were placed in the tub until the water reached the same level. Using of Chinese students and 8% of U.S. students successfully solved
a boat instead of a tub was also considered appropriate. This problem was the statue problem, ␹2(1, N ⫽ 270) ⫽ 120.83. It is worth noting
designed to be isomorphic to the elephant tale.
that the U.S. sample contained 16 Chinese American students, 1 of
The cave problem described a scenario in which a treasure hunter needed
to travel into a cave and then find his way out again, without the benefit of
a map or compass. Successful solution of the problem required the solver
to note that the man could leave a trail of small objects, such as pebbles or
sand, while traveling through the cave and then follow this trail out to exit.
This problem was designed to be isomorphic to elements of the H&G tale,
in which a brother and sister tried to find their way out of a forest by
creating a trail with pebbles and bread crumbs. Again, few contextual and
superficial features were shared by the target problem and source tale.
Four control problems were adopted from some classic insight problems,
including the radiation problem, the string problem, the water jar problem,
and the pool problem. The radiation problem involved using a number of
less intense rays from different directions to converge on a tumor such that
the rays effectively destroy the tumor without damaging the surrounding
healthy tissue. The string problem involved grabbing two strings hanging
apart from the ceiling by tying an available object to one string and
swinging it. The water jar problem involved creating an accurate equation
of measurement using three available containers to achieve a desired
amount of water. The pool problem involved devising a procedure to
engineer a bridge to a platform in the middle of a pool filled with water
given only a small number of flat boards but no nails, glue, or rope.
Procedure. Participants in all groups were given a booklet, which
included a consent form, the insight problems, and space to write down
solutions. The instructions, which were read aloud to the participants,
asked them to read the six problems and generate written solutions for
each. For each problem, participants were given 5 min to read the story Figure 1. Percentages of U.S. and Chinese participants solving the target
problem and then write down one or more solutions. After solving each of tasks (the cave and statue problems) and control problems in Experiment 1.
LONG-TERM ANALOGICAL PROBLEM SOLVING 419

whom (6%) solved the statue problem, indicating that there were Performance on the control problems. The four control prob-
no differences in solving the problem between Chinese American lems varied in difficulty level. As Figure 1 depicts, participants in
and the rest of the U.S. sample. both the Chinese and the U.S. samples were more likely to solve
The problem-solving performance was consistent with re- the string and water jar problems than the pool and radiation
sponses to the reminding questions. A chi-square analysis indi- problems. The problem-solving patterns between the U.S. and
cated that Chinese students (67%) were more likely to report being Chinese students were nearly identical. A chi-square analysis
reminded of the elephant tale than were U.S. students (1%), ␹2(1, performed on each problem revealed no significant differences
N ⫽ 270) ⫽ 121.61. Note that in the U.S. sample, 1 student who between the two samples (all ps ⬎ .10). A t test performed on the
was reminded of the elephant tale was a Chinese American who mean number of control problems solved by each sample also
learned about the tale during his childhood in China. revealed no significant differences (0.92 out of 4 problems for
The two left bars of Figure 2 reveal that Chinese students who Chinese students and 1.02 out of 4 problems for U.S. students),
were reminded of the elephant tale outperformed those who were t(268) ⫽ 1.07, p ⫽ .29.
not. To determine the relation between reminding and problem-
solving performance, we performed a chi-square analysis, which
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Discussion
This document is copyrighted by the American Psychological Association or one of its allied publishers.

indicated that more students who were reminded of the elephant


tale successfully solved the problem (88%) than did those who The results illustrate that U.S. participants were more likely to
were not (31%), ␹2(1, N ⫽ 152) ⫽ 51.09. be reminded of the H&G story and to solve the cave problem than
Performance on the cave problem. Figure 1 shows that 75% of Chinese students were. In contrast, Chinese participants outper-
American students and 25% of Chinese students successfully formed U.S. students in solving the statue problem and in being
solved the cave problem, ␹2(1, N ⫽ 270) ⫽ 65.60. Consistent with reminded of the elephant tale. Within each sample, those who were
the problem-solving performance, a chi-square analysis revealed reminded of the source tale were better in solving the target
that U.S. students (37%) were also more likely than Chinese problem than were those who were not reminded. The equivalent
students (7%) to be reminded of the H&G tale, ␹2(1, N ⫽ 270) ⫽ performance in solving the control problems between cultures in
36.98. Note that although no Chinese participants specifically turn indicates that the performance differences in solving the target
reported the H&G tale, 7% mentioned a Chinese story involving a problems were not due to the cultural differences in general ability
trail of grain being created to guide the way out of a village. to solve insight problems. Instead, the mirror pattern of perfor-
To examine the relation between reminding and problem- mance on the statue and cave problems between the two samples
solving performance, we performed a chi-square analysis, which provides clear and compelling evidence that participants are capa-
indicated that students who were reminded of the source analogue ble of drawing on culturally specific experience in solving analo-
successfully were more likely to solve the cave problem (89%) gous problems.
than were those who were not (66%), ␹2(1, N ⫽ 118) ⫽ 7.32. The The results also suggest both explicit and implicit retrieval and
two right bars of Figure 2 illustrate this relation. use of remote analogy in solving problems. Two thirds of Chinese
and over one third of U.S. participants explicitly reported that the
target problem reminded them of the source tale. They were
consciously aware of the relevance between the source analogue
and the target problem. Yet even those who were not aware of the
source analogue in one culture were more likely to solve the target
problem than were their peers in the other culture, where the
analogue was not available. Sixty-six percent of U.S. students who
were not reminded of the H&G story solved the cave problem,
compared with 25% of Chinese students. Likewise, Chinese stu-
dents who were not reminded of the elephant tale outperformed
U.S. students in solving the statue problem (31% vs. 8%, respec-
tively). These results suggest that source analogues also influence
problem solving in an implicit way.
It is important to note that the elephant analogy for Chinese
participants and the H&G analogy for U.S. students are not en-
tirely comparable. First, the elephant tale seems to be better known
in Chinese culture than the H&G story is in U.S. culture. Second,
the solution is the single most important component of the elephant
tale, whereas the solution in the H&G story is much less central.
Third, the H&G tale is substantially longer than the elephant tale.
The structural components in the elephant tale and the statue
problem are more parallel than are those in the H&G tale and the
cave problem. Last, the solution in the elephant tale involves a
procedure that is more complex, whereas the H&G solution in-
volves a more straightforward, one-step action. Given the differ-
Figure 2. Target problem-solving performances of participants who did ences in the levels of availability, accessibility, and applicability
or did not report being reminded of the source analogue in Experiment 1. between these two analogies, it was not surprising that the likeli-
420 CHEN, MO, AND HONOMICHL

hood of being reminded of the source analogue and the perfor- team of scientists on an exploration site who needed to figure out the
mance in solving these two target problems were somewhat dif- weight of a large object (either an elephant or an asteroid) but did not have
ferent. Despite the differences in these two analogies, this access to a scale large enough. After the presentation of the story, a set of
experiment nevertheless reveals parallel effects of source tales on illustrated items as potential objects for solving the problem was shown,
analogical transfer in both cultures. including logs, boxes, rocks, trees, tools, rope, and buckets, that could be
chosen in combination with a solution tool (either a boat or a spring
platform) that could be used to weigh the goal object. The items were
Experiment 2 intentionally created with proper proportions so that the elephant or aster-
oid would appear to fit in the boat or on the spring platform in the pictures
The goal of this experiment was to explore the effects of (see Appendix C). Across conditions, the illustrated items were exactly the
different types of object similarity on long-term analogical transfer same, except for the solution tool, which was required for successfully
processes. Given that the elephant tale appeared to be a more solving the problem. Thus, the target problems differed from the original
effective analogue for exploring transfer, only this source tale was Chinese tale in several superficial features (e.g., scientists vs. emperor) and
adopted in the remaining experiments. Several versions of target story line but were similar or dissimilar in either the goal object (elephant
problems were created, all sharing the same goal structure (rela- vs. asteroid) and/or the solution tool necessary to weigh the goal object
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

(boat vs. spring platform). Although the spring platform differs in appear-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

tions among the objects) and solution with the elephant tale. Two
types of object similarity were manipulated without altering either ance from the source solution tool, a boat, both can serve as compression
devices and thus share the same function. The asteroid differs from the
the goal structure or the story theme across different versions of
source goal object, an elephant, yet both represent a large object to be
the target problem. The manipulation resulted in four types of
weighed.
relations between the source and target problems: similar goal Procedure. Participants were randomly assigned to one of four condi-
object–similar solution tool (SG-ST), similar goal object– tions: SG-ST, SG-DT, DG-ST, and DG-DT. All groups were given a
dissimilar solution tool (SG-DT), dissimilar goal object–similar booklet including a consent form, the story text and illustrations, blank
solution tool (DG-ST), and dissimilar goal object– dissimilar so- space in which to write down their solutions, and a page containing
lution tool (DG-DT). posttask questions. Chinese participants were also given a blank page for
Unlike Experiment 1, where participants were asked to generate free recall of the Chinese tale at the end of the experiment.
possible solutions freely, the present paradigm involved partici- Participants were read aloud instructions that explained that they were
pants reading a target problem and then viewing a set of illustrated being asked to read a story and solve a problem with a pencil and paper.
objects that could potentially be used to solve the problem. After Participants were first given 80 s to read the story problem. After reading
attempting to solve the problem, participants were asked to answer the story problem, they were instructed to view the illustrated items and
then to use any as they wished to help them solve the problem. Participants
several posttask questions to indicate what the target problem
were given 5 min to write down any appropriate solutions. They could
reminded them of, and, if it did remind them of something, they
come up with more than one solution with any items illustrated. After
were to indicate how helpful this memory was in solving the target generating solutions, participants were asked to turn to the next page in the
problem. Finally, Chinese participants were asked explicitly to packet and complete the posttask questions.
write down the elephant tale. Participants’ problem-solving per- A series of questions was designed to assess when or if participants had
formance was observed and the processes involved in analogical heard about the elephant tale in the past and whether they were aware of its
transfer were examined. The accessing component was assessed by relevance to the target problem. Specifically, participants were asked to
whether participants reported being reminded of the tale. The answer the following questions: (a) if the target problem reminded them of
mapping process was measured by participants’ selection of both any story and/or event they had heard of or read before and, if so, what they
the goal object and the solution tool during problem solving. were reminded of; (b) on a scale of 1 (not helpful at all) to 5 (extremely
Finally, the executing component was measured by the partici- helpful), how helpful the story or event they were reminded of was in
pants’ accurate use of the weight compression and equivalence generating a solution to the target problem; (c) when they had first heard
about the story or event that they were reminded of (10 years, 5 years, 3
principles with the solution tool.
years, or 1 year ago); and (d) when they heard about it last (10 years, 5
It was predicted that Chinese participants in the SG-ST condi-
years, 3 years, or 1 year ago). Last, the Chinese participants were asked to
tion would outperform their peers in other conditions, especially in write down their recollection of the elephant tale with at least 100 Chinese
the DG-DT condition. The two types of object similarity were also characters on a blank sheet.
hypothesized to differentially impact the accessing, mapping, and Measures. There were five dependent measures.
executing processes. In contrast, no differences were predicted 1. The first measure involved recollection of the source tale. Availability
across the four conditions for participants in the United States. of the source analogue was assessed by Chinese participants’ recall of the
elephant tale at the end of the experiment. Source story recall was mea-
sured with a 3-point scale: 2 points were awarded for reporting both the
Method goal and the solution, 1 point for either, and 0 points for neither. Perfor-
Participants. A total of 256 university students participated in this mance of 50 randomly selected participants was scored independently by
experiment: 117 undergraduate students from either psychology or educa- two observers, and agreement was 99%.
tion classes at South China Normal University and 139 students from 2. The second measure involved accessing the source story. When asked
psychology or human development classes at the University of California, what previous stories or problems, if any, they were reminded of by the
Davis. Each group was composed of approximately equal numbers of men target problem, participants’ indication of the elephant tale was considered
and women. Participating students received course credit for completing to reflect the retrieval of the source story and scored as 1 point.
the experiment. 3. The third measure involved mapping the key objects between prob-
Design and materials. Instead of using different source stories, we lems. Participants’ mapping of the key objects was measured by their use
manipulated the target problems to create four versions that were all of both the goal object and the solution tool during their target problem
analogous to the original elephant tale. Each target problem described a solving, regardless of whether the problem was ultimately solved. Because
LONG-TERM ANALOGICAL PROBLEM SOLVING 421

almost all the participants mentioned the goal object, this measure mainly Recalling the source story. Figure 3A shows the recall scores
reflected participants’ selection of the solution tool to map onto the source for Chinese participants in the four conditions. A one-way analysis
solution tool when they attempted to solve the problem. of variance (ANOVA) revealed no differences across conditions,
4. The fourth measure involved executing the solution. Participants’ F(1, 113) ⫽ 0.59. These results, consistent with the preliminary
execution of the solution was assessed with two aspects: strategies and
survey of 30 participants who were asked to recall the source story
solution score. Three main problem-solving strategies were identified. The
without attempting to solve the target problem, indicated that the
first approach was the division strategy, which involved cutting or breaking
the goal object into pieces and then weighing the pieces separately. The majority of Chinese college students remembered the source tale
second was the balance strategy, which involved the weight equivalence even many years after learning it. Participants in the United States
concept, such as balancing the goal object with a set of smaller items so the were not asked to recall the Chinese tale.
total weight of the smaller items would be the same as the weight of the Accessing the source story. Figure 3B shows Chinese partic-
goal object. This strategy was not considered an ideal approach because it ipants’ accessing scores across conditions. An overall chi-square
did not involve a reasonably accurate balancing method (e.g., putting the analysis revealed significant differences across condition, ␹2(3,
asteroid on one end of a log or spring and a set of smaller objects on the N ⫽ 117) ⫽ 12.71. Planned paired comparisons revealed that
other). It also often involved selecting items that were not in the illustrated participants in the DG-DT condition were less likely to access the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

set (e.g., a pulley system). The third approach was the compression
This document is copyrighted by the American Psychological Association or one of its allied publishers.

source story than were those in the SG-ST, SG-DT, and DG-ST
strategy, which involved an effective way to measure the degree of com-
conditions, ␹2(1, N ⫽ 60) ⫽ 7.68, ␹2(1, N ⫽ 59) ⫽ 5.19, and ␹2(1,
pression of the boat or spring platform. This approach could lead to an
effective and complete solution, but it might also involve an incomplete N ⫽ 58) ⫽ 4.92, respectively.
solution (e.g., simply putting the goal object on the boat or spring and The participants who accessed the source analogue reported that
estimating the weight by observing how far it compressed). A score of 1 they first learned about the source story 9.4 years ago on average
was assigned when a participant used one of these strategies, regardless of and the last time they heard about the story was 7.4 years ago. No
whether the problem was ultimately solved. differences between conditions were found. Note that these num-
In addition to these three strategies, some Chinese participants in the two bers do not precisely reflect the time elapsed because the scale
dissimilar solution tool conditions (the spring platform versions) cited a involved indicating the interval of over 1, 3, 5, or 10 years. None
boat as a solution tool, even though it was not among the possible solution of the participants in the United States reported being reminded of
items (creating the boat strategy). The use of this approach was considered
the elephant tale.
to reflect negative transfer, because it interfered with the selection of the
Mapping the tools. Figure 3C shows differences in Chinese
appropriate solution tool.
The second measure of the execution component involved a 3-point participants’ use of both the goal object (either the elephant or the
solution scale. A score of 2 was assigned when a complete and accurate asteroid) and the solution tool (either the boat or the spring) across
solution was generated: that is, using the boat (in the boat versions) or conditions. Given that most participants mentioned the goal object,
using the spring platform (in the spring versions) as a compression tool to the mapping measure mainly reflected whether the solution tool
equalize the weight of the elephant or asteroid and smaller objects and then was used in their solution attempts. A chi-square analysis revealed
weighing the smaller objects separately. A score of 1 was assigned when that Chinese participants were more likely to use the solution tool
the solution was correct only at a general level but was neither complete than were participants in the United States, ␹2(1, N ⫽ 256) ⫽
nor precise, such as generating an idea of equalizing the weight of the goal 26.38. An overall chi-square analysis revealed significant differ-
object with the summed weights of smaller objects without specifying how
ences across conditions in the Chinese sample, ␹2(3, N ⫽
or with an inaccurate procedure (e.g., using the log as a balancing board).
Participants who did not come up with any appropriate solution received a
117) ⫽ 15.78. Planned paired comparisons revealed that par-
score of 0. Problem-solving performance of 80 randomly selected partic- ticipants in the SG-ST condition were more likely to use the
ipants (from both China and the United States) was scored independently solution tool than were those in the DG-DT and SG-DT con-
by two observers, and agreement between the two observers was 97%. ditions, ␹2(1, N ⫽ 60) ⫽ 13.42 and ␹2(1, N ⫽ 59) ⫽ 8.04,
5. The fifth measure involved evaluating the usefulness of the source respectively. The DG-ST condition also differed from the
story. The participants were asked to indicate how much the source DG-DT condition, ␹2(1, N ⫽ 58) ⫽ 5.90, suggesting that a
analogue had helped them solve the target problem on a scale of 1 (not common solution tool facilitated mapping when participants
helpful at all) to 5 (extremely helpful). The data were relevant only to those attempted to solve the target problem.
who reported the elephant tale. In contrast, an overall chi-square analysis revealed no signifi-
It is important to note that the actual order of assessing the participants’
cant differences across conditions in the U.S. sample, ␹2(3, N ⫽
problem-solving activities does not correspond to the order in which these
five measures are presented. As described in the Procedure section, par-
139) ⫽ 3.97, p ⫽ .26. Presumably because U.S. participants were
ticipants were first asked to attempt to solve the target problem before less likely to use the ineffective division strategy (i.e., cut the
being asked to report any potentially relevant source stories, to evaluate the object into pieces and then weigh them separately) for an animate
usefulness of the reported source analogue, and to recall the source tale. object (elephant) than for an inanimate object (asteroid), their
For theoretical reasons, data analyses are presented in the order of recall- performance in selecting the solution tool was somewhat better in
ing, accessing, mapping, executing, and evaluating, which is the sequence the similar goal object (elephant) conditions than in the different
in which these processes were hypothesized to occur. goal object (asteroid) conditions.
Executing the solution. Figure 4 reveals the percentage of
Results participants using the division, balance, and compression strategies
in each condition. Participants in the United States were more
Preliminary analyses revealed no gender differences in any likely to use the division strategy than Chinese participants were.
measure (all ps ⬎ .30). The results are reported in five sections. All The analyses were focused on the use of the appropriate approach,
reported effects are significant beyond the .05 level, unless other- the compression strategy. An overall chi-square analysis revealed
wise noted. significant differences across condition in Chinese participants,
422 CHEN, MO, AND HONOMICHL
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 3. U.S. and Chinese participants’ performance in recalling the source tale (A), accessing the analogue (B),
mapping the key objects (C), executing the source solution (D), and evaluating the helpfulness of the analogue (E) in
Experiment 2. SG-ST ⫽ similar goal object, similar solution tool; SG-DT ⫽ similar goal object, dissimilar solution
tool; DG-ST ⫽ dissimilar goal object, similar solution tool; DG-DT ⫽ dissimilar goal object, dissimilar solution tool.

␹2(3, N ⫽ 117) ⫽ 19.57. Planned paired comparisons revealed that ␹2(1, N ⫽ 57) ⫽ 4.51 and ␹2(1, N ⫽ 58) ⫽ 4.08, respectively. In
participants in the SG-ST condition were more likely to use this contrast, a substantially lower percentage of participants in the
strategy than were those in the DG-ST condition, ␹2(1, N ⫽ 58) ⫽ United States used this strategy, and an overall chi-square analysis
4.47, which differed from the SG-DT and DG-DT conditions, revealed no significant differences across conditions.
LONG-TERM ANALOGICAL PROBLEM SOLVING 423
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 4. Participants’ problem-solving strategies (division [A], balance [B], and compression [C]) used in
Experiment 2. SG-ST ⫽ similar goal object, similar solution tool; SG-DT ⫽ similar goal object, dissimilar solution
tool; DG-ST ⫽ dissimilar goal object, similar solution tool; DG-DT ⫽ dissimilar goal object, dissimilar solution tool.

In addition, 14% and 7% of the Chinese participants used the ative transfer occurred when solution tools differed in attributes
creating the boat strategy (which was considered incorrect because between problems.
a boat was not in the list of possible tools) in the SG-DT and Participants’ problem-solving performance across conditions
DG-DT conditions, respectively. These results suggested that neg- was examined (see Figure 3D). To determine if there were differ-
424 CHEN, MO, AND HONOMICHL

ences in the performance scores between the two populations and “Did this problem remind you of any story or event that you have heard of
across conditions, we performed a 2 (population: China vs. United before?”
States) ⫻ 4 (condition) between-subject ANOVA. This analysis
revealed significant main effects for population, F(1, 248) ⫽ Results
118.16, MSE ⫽ 51.64, ␩2 ⫽ .32, and for condition, F(3, 248) ⫽
8.96, MSE ⫽ 3.91, ␩2 ⫽ .10. The interaction between population The accessing measure was the same as that used in Experiment
and condition was also significant, F(3, 248) ⫽ 6.37, MSE ⫽ 2.79, 2. (Unless otherwise noted, all reported effects are significant
␩2 ⫽ .07. Fisher’s protected least significant difference (PLSD) beyond the .05 level.) One hundred percent (33 out of 33), 97% (28
revealed that Chinese participants’ performance in the SG-ST out of 29), 97% (34 out of 35), and 80% (24 out of 30) of the
condition was better than that in the SG-DT, DG-ST ( p ⫽ .057), participants in the SG-ST, SG-DT, DG-ST, and DG-DT condi-
and DG-DT conditions. Performance in the DG-ST condition was tions, respectively, reported the source tale. An overall chi-square
also better than that in the SG-DT and DG-DT conditions. In analysis revealed significant differences across condition, ␹2(3,
contrast, no differences were found in U.S. participants’ perfor- N ⫽ 127) ⫽ 12.86. Planned paired comparisons revealed that
mance among conditions, F(3, 135) ⫽ 1.17, p ⫽ .33. participants in the DG-DT condition were less likely to retrieve the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

source story than those in the SG-ST, SG-DT, and DG-ST condi-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Evaluating the helpfulness of the source story. A one-way


ANOVA was performed to determine whether Chinese partici- tions, ␹2(1, N ⫽ 63) ⫽ 7.29, ␹2(1, N ⫽ 59) ⫽ 3.86, and ␹2(1, N ⫽
pants in the four conditions rated the helpfulness of the source 65) ⫽ 4.94, respectively.
story differently. The analysis revealed an effect for condition,
F(3, 95) ⫽ 4.23, MSE ⫽ 5.16, ␩2 ⫽ .12. Fisher’s PLSD showed Discussion
that helpfulness was rated lower in the DG-DT condition than in
the SG-ST and DG-ST conditions, respectively. If participants’ accessing reports had been influenced by the
target problem-solving activities, the pattern of accessing across
the four conditions would have been different from that in Exper-
Discussion
iment 2. Although the overall performance in accessing the source
The results illustrate differential effects of goal object and tale appeared to be slightly better, the effects of condition on
solution tool similarity in transfer processes. First, although most accessing were similar to those observed in Experiment 2. The
Chinese participants reported spontaneously retrieving the source majority of the participants were reminded of the elephant tale
story even when the key objects differed between the problems, when reading the target problem, even when the target and source
receiving either a common goal object or a common solution tool problems shared little object similarity. Adding a key common
led to a higher proportion of participants successfully accessing the object, either the goal object or the solution tool, significantly
source story. Second, goal object similarity did not play an im- facilitated reminding.
portant role in mapping the entities between the problems. In
contrast, solution tool commonality enhanced the selection of the Experiment 4
solution tool in the target that could be readily mapped onto the
corresponding item in the source problem. Finally, having a The first three experiments demonstrated access to and use of
common solution tool facilitated the executing process; having remote analogy in solving problems across contexts. The results
a common goal object did not. Given the different object attributes, also point to various obstacles to transfer, specifically dissimilar-
participants might not have readily realized the functional similar- ities between the key objects in the source tale versus the target
ity in the solution tools between problems and thus failed to problem. Experiment 4 was designed to address developmental
implement the solution strategy acquired from the source problem. differences and examine possible difficulties that younger students
In sum, a common goal object did not seem to play a critical role might experience, especially when the source and target problems
in postretrieval components, whereas a similar solution tool influ- do not share key objects. Previous studies have demonstrated that
enced all the key transfer processes and greatly improved younger children are less likely to solve an analogous target
performance. problem than older children are (e.g., DeLoache, Miller, & Pier-
routsakos, 1998; Goswami, 1996; Holyoak et al., 1984; Siegler,
Experiment 3 1989). Yet relatively little is known concerning age differences in
various transfer components involved in long-term analogical
In Experiment 2, the accessing component was measured after problem solving. Middle school and high school students from
the participants attempted to solve the target problem; therefore, it China participated in Experiment 4, which had a design similar to
is possible that their problem-solving activities might have af- that of Experiment 2. It was predicted that middle school students
fected their reports concerning what they were reminded of when were more likely to experience difficulties associated with transfer
reading the target story. Experiment 3 was designed to allow us to processes than high school students were.
examine the accessing component without the possible influence
of the problem-solving activities and performance.
Method
Method Participants. A total of 222 seventh (mean age ⫽ 13.1 years) and tenth
grade students (mean age ⫽ 16.3 years) from local schools in southern
A total of 127 Chinese undergraduate students who did not participant in China were included in the study.
any other experiments were assigned to one of the four target problem Design, materials, and procedure. The design and materials were
conditions. Instead of solving the problem, the participants were asked, identical to those in Experiment 2. The procedure was also similar to that
LONG-TERM ANALOGICAL PROBLEM SOLVING 425

in Experiment 2, except that the middle and high school students were compression strategy than did participants in the SG-DT and
given 160 s (instead of 80 s) to read the target story and view the illustrated DG-DT conditions, ␹2(1, N ⫽ 51) ⫽ 6.89 and ␹2(1, N ⫽ 52) ⫽
objects before they attempted to solve the problem. Also, the students were 7.59, respectively. An overall chi-square test for 7th grade partic-
given 8 (instead of 5) min to write down their solutions. ipants revealed differences across conditions, ␹2(3, N ⫽ 115) ⫽
39.73. Planned paired comparisons revealed significant differ-
Results
ences between the SG-ST condition and the SG-DT, ␹2(1, N ⫽
Recalling. Figure 5A shows students’ memory of the source 59) ⫽ 27.08; DG-ST, ␹2(1, N ⫽ 63) ⫽ 9.27; and DG-DT
problem. (All reported effects are significant beyond the .05 level, conditions, ␹2(1, N ⫽ 57) ⫽ 32.73. The differences between the
unless otherwise noted.) A 2 (grade level) ⫻ 4 (condition) DG-ST and DG-DT conditions were also significant, ␹2(1, N ⫽
ANOVA revealed no main effects for grade level, F(1, 212) ⫽ 56) ⫽ 10.60.
0.001, p ⫽ .98, or condition, F(3, 212) ⫽ 0.13, p ⫽ .94. The In addition, 9% (7% and 11% in the SG-DT and DG-DT
interaction was also not significant, F(3, 212) ⫽ 1.53, p ⫽ .21, conditions, respectively) of the 10th graders and 25% (30% and
suggesting that both age groups remembered the source story 20% in the SG-DT and DG-DT conditions, respectively) of the 7th
equally well. graders used the creating the boat strategy, ␹2(1, N ⫽ 222) ⫽ 4.84.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Accessing. Figure 5B shows the percentages of participants in These results suggested that younger students were more likely to
each grade who were reminded of the source problem. A chi- use a tool that no longer existed in the target problem and, thus,
square test revealed that 10th graders (85%) were more likely to they were more susceptible to negative transfer.
access the source story than were 7th graders (68%), ␹2(1, N ⫽ Participants’ solution performance in each grade level across
222) ⫽ 8.38. Overall chi-square tests across conditions were conditions was shown in Figure 5D. A 2 (grade level) ⫻ 4
performed in separate grades and revealed differences across con- (condition) between-subject ANOVA was performed. This analy-
ditions for 10th grade, ␹2(3, N ⫽ 106) ⫽ 8.79, and for 7th grade, sis revealed significant main effects for grade, F(1, 214) ⫽ 8.65,
␹2(3, N ⫽ 115) ⫽ 8.54. Planned paired comparisons revealed that MSE ⫽ 5.23, ␩2 ⫽ .04, and for condition, F(3, 214) ⫽ 21.20,
for the 10th grade sample, participants in the DG-DT condition MSE ⫽ 12.80, ␩2 ⫽ .23; however, no significant interaction was
were less likely to access the source problem than were those in the found, F(3, 214) ⫽ 1.86, MSE ⫽ 1.13, ␩2 ⫽ .03, p ⫽ .14. Fisher’s
DG-ST, ␹2(1, N ⫽ 52) ⫽ 4.39; SG-DT, ␹2(1, N ⫽ 55) ⫽ 5.26; and PLSD revealed that 10th graders’ performance in the SG-ST
SG-ST conditions, ␹2(1, N ⫽ 55) ⫽ 3.56, p ⫽ .059. For the 7th condition was better than that in the SG-DT and DG-DT condi-
grade sample, fewer participants accessed the source story in the tions. Performance in the DG-ST condition was also better than
DG-DT condition than in the SG-ST condition, ␹2(1, N ⫽ 57) ⫽ that in the SG-DT and DG-DT conditions. Similarly, 7th graders’
8.59. No other significant differences in accessing were found performance in the SG-ST condition was better than that in the
among the four conditions. SG-DT, DG-ST, and DG-DT conditions. Performance in the
Mapping. The percentages of participants mapping the key DG-ST condition was also better than that in the SG-DT and
items in each grade are presented in Figure 5C. A chi-square test DG-DT conditions.
revealed that 10th graders (73%) were more likely to map the To examine the relation between accessing and problem-solving
items than were 7th graders (61%), ␹2(1, N ⫽ 222) ⫽ 3.61, p ⫽ performance, we performed a t test for both grades combined in the
.058. An overall chi-square test revealed differences across con- DG-DT condition, where both accessing and executing appeared to
ditions for 10th grade participants, ␹2(3, N ⫽ 107) ⫽ 19.90. be most difficult. Students who accessed the source analogue
Planned paired comparisons revealed that participants in the solved the problem more effectively (earning an average score of
SG-ST condition outperformed those in the SG-DT and DG-DT 1.07 out of a possible score of 3.00) than did those who did not
conditions, ␹2(1, N ⫽ 55) ⫽ 8.59 and ␹2(1, N ⫽ 56) ⫽ 12.60, (mean score ⫽ 0.32), t(51) ⫽ 3.45.
respectively. Students in the DG-ST condition also did better than Evaluating. A 2 (grade level) ⫻ 4 (condition) ANOVA was
those in the SG-DT and DG-DT conditions, ␹2(1, N ⫽ 51) ⫽ 7.03 performed. This analysis revealed a significant main effect for
and ␹2(1, N ⫽ 52) ⫽ 10.53, respectively. For 7th grade partici- condition, F(3, 156) ⫽ 4.45, MSE ⫽ 5.50, ␩2 ⫽ .08, but not for
pants, an overall chi-square test also revealed differences across grade, F(1, 156) ⫽ 1.03, MSE ⫽ 1.28, ␩2 ⫽ .01. An interaction
conditions, ␹2(3, N ⫽ 115) ⫽ 26.50. Planned paired comparisons was also marginally significant, F(3, 156) ⫽ 2.27, MSE ⫽ 2.81,
revealed differences between the SG-ST condition and the SG-DT, ␩2 ⫽ .04, p ⫽ .082. Fisher’s PLSD revealed that 10th graders in
␹2(1, N ⫽ 59) ⫽ 17.36; DG-ST, ␹2(1, N ⫽ 63) ⫽ 8.21; and the SG-ST condition gave a higher rating to the helpfulness of the
DG-DT conditions, ␹2(1, N ⫽ 57) ⫽ 24.08. Differences between source story than did 10th graders in the SG-DT and DG-DT
the DG-ST and DG-DT conditions were also significant, ␹2(1, N ⫽ conditions. The helpfulness rating in the DG-ST condition was
56) ⫽ 5.83. also higher than that in the SG-DT and DG-DT conditions. In
Executing. Participants’ strategies were assessed. The percent- contrast, 7th graders’ rating was not different across conditions,
ages of participants using the compression strategy in each grade ps ⬎ .10.
are presented in Figure 6C. A chi-square test revealed no grade
effects between 7th grade (58%) and 10th grade (66%). An overall Discussion
chi-square test revealed differences across conditions for 10th
grade, ␹2(3, N ⫽ 107) ⫽ 18.68. Planned paired comparisons Both middle and high school students had little difficulty in
revealed that more participants in the SG-ST condition used this recalling the source story. The pattern of accessing the source
strategy than did participants in the SG-DT and DG-DT condi- analogue across the four conditions in high school students was
tions, ␹2(1, N ⫽ 55) ⫽ 10.89 and ␹2(1, N ⫽ 56) ⫽ 11.79, similar to that in college students: Dissimilarity in both goal object
respectively. Also, more students in the DG-ST condition used the and solution tool hindered the accessing process. Furthermore,
426 CHEN, MO, AND HONOMICHL
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 5. High school and middle school students’ performance in recalling the source tale (A), accessing the
analogue (B), mapping the key objects (C), executing the source solution (D), and evaluating the helpfulness of
the analogue (E) in Experiment 4. SG-ST ⫽ similar goal object, similar solution tool; SG-DT ⫽ similar goal
object, dissimilar solution tool; DG-ST ⫽ dissimilar goal object, similar solution tool; DG-DT ⫽ dissimilar goal
object, dissimilar solution tool.
LONG-TERM ANALOGICAL PROBLEM SOLVING 427
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 6. Participants’ problem-solving strategies (division [A], balance [B], and compression [C]) used in
Experiment 4. SG-ST ⫽ similar goal object, similar solution tool; SG-DT ⫽ similar goal object, dissimilar solution
tool; DG-ST ⫽ dissimilar goal object, similar solution tool; DG-DT ⫽ dissimilar goal object, dissimilar solution tool.

middle school students in the DG-DT condition were less likely to object and a common solution tool. However, they were less likely
be reminded of the source story than were those in the SG-ST to access the source story, to map the key objects, and to execute
condition. Middle school students performed as well as high the solution when one or both objects were different between the
school students when the problems shared both a common goal problems.
428 CHEN, MO, AND HONOMICHL

General Discussion occurs when common objects are shared between problems, the
analogous target problem need not share key object attributes with
Summary of Results the source analogue in long-term memory.
1. U.S. college students were more likely to solve the cave The lack of efficient transfer found in previous laboratory stud-
problem than Chinese students were, given the availability of the ies (e.g., Perkins & Grotzer, 1997; Reed et al., 1974) and the robust
H&G tale in American culture. Likewise, Chinese students out- use of remote analogy evident in the present research seems to
performed U.S. students in solving the statue problem because of present a paradox. Yet, as Dunbar (2001) has pointed out, struc-
their experience with the elephant tale, indicating that participants tural analogies are a frequent rather than a rare phenomenon when
were capable of spontaneously retrieving and using an analogue reasoning is investigated in naturalistic environments. For exam-
acquired as long as 10 years ago to solve a target problem. ple, Dunbar and Blanchette (2001) found that scientists often use
2. Chinese college students in the conditions with at least one structural features and higher order relations in analogizing during
similar object were more likely to solve the target problem than the discovery process. Source analogues may well be encoded in a
those in the DG-DT condition were, suggesting that object simi- rich way in various cultural settings, as in the context of scientific
discovery, so that the acquired information is more durable, ac-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

larity facilitated long-term analogical transfer. U.S. participants


This document is copyrighted by the American Psychological Association or one of its allied publishers.

did not differ in problem-solving performance across conditions. cessible, and applicable.
3. Chinese college students were more likely to access the These experiments also point to both explicit (or conscious) and
source analogue when the target problem involved either a com- implicit (or unconscious) retrieval and use of source analogues
mon goal object or a common solution tool. Yet even in the after a substantial delay. Prior work suggests that analogues may
DG-DT condition, over two thirds of the participants spontane- be used in problem solving either explicitly (e.g., Gentner &
ously accessed the source story. Thus, common structural relations Forbus, 1991; Gick & Holyoak, 1980) or implicitly (e.g., Reed,
seem to play a role in analogue retrieval; however, any additional Willis, & Guarino, 1994; Schunn & Dunbar, 1996; also see dis-
object similarity also facilitated the accessing process. cussion in Markman & Gentner, in press). One might speculate
4. When the solution tool in the target problem was similar to that when source information is presented immediately prior to the
that in the source story, participants were more likely to map it problem-solving task, participants would be likely to be aware of
onto the source tool and execute the solution successfully com- the relevance of the source to the target. After increasingly longer
pared with those in the dissimilar solution tool conditions. In delays, participants might be less likely to explicitly retrieve and
contrast, a similar goal object did not play a significant role in use the source information. Although the present research does not
postaccessing processes. Therefore, the impact of solution tool examine the relationship between distance of transfer and con-
commonality was more robust than that of goal object similarity. sciousness or contrast short-term and long-term transfer, it never-
5. Middle school students were less likely to solve the target theless demonstrates that even after a delay of years, participants
problem than high school students were. Middle school students can be aware of the relevance of a remote source tale, the simi-
were more likely than high school students to experience difficul- larities between problems, and the helpfulness of the source ana-
ties in accessing the source problem, in mapping the key objects logue. Evidence of the implicit use of the source analogue comes
between the problems, and in executing the source solution, espe- from the findings that participants who did not report being re-
cially in the dissimilar solution tool conditions. minded of the source tale outperformed their peers in the culture in
which the source analogue was unavailable (Experiment 1). Fur-
ther exploration of factors that influence explicit and implicit use
Long-Term Transfer
of analogy may prove to be a fruitful avenue for investigating the
The distance of transfer between a source and target problem mechanisms of long-term analogical transfer.
can be determined in a number of aspects. Barnett and Ceci (2002)
recently proposed a taxonomy of the dimensions involved in Processes of Analogical Problem Solving
transfer distance. They suggest that time interval and contextual
similarity between source acquisition and target test are critical We propose a conceptual model that specifies how different
dimensions and that the exploration of long-term analogical prob- types of object similarity influence the ways individuals access
lem solving across contexts has important theoretical and educa- source analogues, map entities between problems, and execute an
tional implications. There is little doubt that accessibility and acquired source solution.
applicability of information in long-term memory decrease over Accessing source information. The initial process of analogi-
time and that degree of transfer depends on the time lapse between cal transfer involves searching for relevant experience, such as
the original learning and the solving of the target problem and on similar events, instances, or stories, in long-term memory at the
the degree of semantic and contextual similarity between prob- appropriate time. One accessing cue involves object reminding.
lems. However, theorizing and modeling in analogical problem The presence of common object attributes between problems in-
solving have been almost entirely based on experimental results of creases the likelihood of retrieving a source analogue. Previous
short-term transfer within the same contexts. work with short-term analogical transfer has yielded consistent
The present research provides compelling evidence for remote findings concerning the effects of superficial similarity on acces-
transfer in that the source and target problems shared few semantic sibility (e.g., Catrambone, 2002; Gentner et al., 1993; Wharton et
and contextual features. Even with a significant time gap between al., 1994). In the present study, sharing either goal object or
source and target problems, solvers can be reminded of a source solution tool alone proved sufficient to increase the accessibility of
analogue and use it effectively when encountering an isomorphic the source tale, suggesting that both play similar roles in remind-
problem in a different context. Although the best performance ing. The present results suggest that memories contain object-
LONG-TERM ANALOGICAL PROBLEM SOLVING 429

specific information of a story or problem situation and, thus, experience an obstacle in implementing a source problem’s solu-
object cues in a target problem may trigger the retrieval of a source tion (e.g., Chen, 2002; Reed & Bolstad, 1991; Ross, 1989). The
analogue. present results further reveal that participants can also suffer from
Another avenue of accessing relevant experience involves struc- an execution deficiency when the solution tool for solving the
tural reminding. Models of artificial intelligence and case-based target problem differs in attributes from that used for solving the
reasoning view the process of human memory retrieval as typically source problem. The execution difficulty might also have been due
driven by higher order causal or goal structures (e.g., Hammond, to some participants’ persistence in using the creating the boat
Seifert, & Gray, 1991; Kolodner, 1993; Schank, 1982), implying strategy, reflecting the occurrence of what Luchins (1942) referred
that individuals are capable of accessing knowledge in long-term to as “mental set,” a form of negative transfer.
memory on the basis of higher order structural similarity. Over two The likely reason why only the solution tool is associated with
thirds of Chinese participants retrieved the elephant tale from both mapping and executing components is that the solution tool
long-term memory, even when the target shared no common object serves as a crucial part of the causal path, whereas the goal object
attributes with the source. These findings extend previous work by is arbitrary. Although any object can be weighed with compres-
demonstrating that common goal structures can play a central role sion, only objects that can be compressed may serve as a solution
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

in guiding the memory accessing involved in long-term analogical tool. Furthermore, the goal object is automatically selected, but
transfer and that individuals can spontaneously access a source there are alternatives from which a solution tool can be chosen and
analogue that shares little superficial similarity with the target. utilized to generate a solution procedure.
Mapping structural elements between problems. Previous This conceptual model addresses the observation that individu-
work on the mapping process has typically involved participants’ als often experience difficulty in long-term analogical transfer and
making soundness evaluations. Mapping of problems has been that participants, especially younger children, often fail to solve
found to be guided by the common relational structures when problems by remote analogy. This model also addresses questions
rating the soundness of an analogy (e.g., Clement & Gentner, of why some analogues are more difficult to use in problem
1991; Gentner et al., 1993). In the present research, mapping was solving than others and why similarity in some objects is more
assessed by the participants’ choice of the goal object and solution useful. The generality and significance of the present object sim-
tool during problem solving. In analogical problem solving, the ilarity framework seems to depend on the assumption that problem
selection of the corresponding target objects ultimately reflects the solving typically involves selecting and using appropriate tools.
mapping of the structures between problems. After the relevant Numerous examples come from early Gestalt research, such as
source story or event is retrieved, the common structural relations Köhler’s (1927) reaching problem, Maier’s (1931) two string
of the entities or objects associated with the entities between the problem, Duncker’s (1945) candle problem, and Luchins’s (1942)
problems serve as guides for matching the problems (e.g., Bassok, water jar problem, as well as more recent work involving proba-
1990; Gentner & Markman, 1997). The present results further bility and physics problems (Bassok, 1990; Catrambone, 1998;
demonstrate that different types of object similarity play distinc- Reed, 1989; Ross, 1989).
tive roles in transfer. Although solution tool similarity played a It is noteworthy that although these components are hypothe-
critical role in the mapping process, goal object commonality did sized to be completed one by one in the order of accessing,
not appear to influence how the key elements are mapped between mapping, and executing, the operation of the components does not
problems. have to be always perfectly serial. An individual might choose the
Once a source analogue has been retrieved, solvers need to tub as a tool for weighing the statue (mapping and executing)
match the corresponding elements or objects between problems. before actually realizing the analogous relation to the source tale
When source and target problems share common object attributes, (accessing). Furthermore, the retrieval and use of analogous source
the objects may be mapped readily. However, when corresponding information do not have to be conscious. Individuals might access
objects share only functional relations but differ in object at- a source tale and use the analogous solution without being aware
tributes, the mapping between problem elements becomes less of the relevance between the source and target problems.
direct, and the corresponding objects would be less likely to be
selected to map onto the source problem. Furthermore, mapping Developmental Differences
between the goal objects may be more or less automatic because
solving the problem requires operating on the object to achieve the Age differences in various processes associated with long-term
goal, regardless of whether similar object attributes are shared. In transfer were also evident in this study. Previous studies of chil-
contrast, mapping the solution tools between problems becomes a dren’s analogical reasoning have typically focused on demonstrat-
more effortful and challenging process. When the solution tools ing the importance of a single component to the overall process
share no object attributes, the common functions of the tools may involved in short-term transfer (e.g., Chen, 1996; Daehler & Chen,
not be immediately apparent, and thus the different attributes of the 1993; Holyoak et al., 1984; Kotovsky & Gentner, 1996). In the
solution tools decrease the likelihood of correct selection from present study, we explored the contributions of all three key
among a set of possible tools. components on a single task of long-term analogical problem
Executing a learned solution. The accessing of an analogue solving. The present results are consistent with Gentner’s rela-
and mapping of the corresponding elements between problem tional shift concept, which posits that development in analogical
structures do not ensure successful execution of a solution re- reasoning involves a shift from reliance more on superficial sim-
trieved from long-term memory. Previous studies have shown that ilarity (e.g., object correspondences) to the mapping of relational
when the source and target problems share only a solution princi- structure (Gentner, 1989). Results also extend previous develop-
ple but differ in specific procedures, solvers are more likely to mental work, further indicating that age differences are associated
430 CHEN, MO, AND HONOMICHL

with all the key processes; failure in any of the components can this paradigm may have allowed more accurate and reliable as-
disrupt problem solving. Despite the availability of the source tale sessment of memory retrieval as a component of the overall
in long-term memory, some young participants failed to come up long-term transfer process.
with an appropriate solution because of their inability to access the Likewise, the mapping process was assessed in the same
source story from long-term memory when they encountered the problem-solving context, but it was separated from the final exe-
target problem. Others failed because of difficulties in mapping the cution process. Successful mapping of the elements between prob-
entities of the source story onto the target problem, thus failing to lems was measured by the correct selection of the key objects,
choose a novel tool that matched the boat in function but not in namely, the goal object and the solution tool. Other typical mea-
attributes. Yet others failed to execute the acquired analogous sures of mapping, such as a soundness rating of the relations
solution because they used the selected tool inaccurately. between stories or events (e.g., Gentner, 1989; Gentner et al.,
Despite the fact that high school students had heard of the 1993), might have altered other cognitive processes that we in-
source analogue in the more distant past, they nevertheless were tended to examine in the problem-solving context. We believe that
more likely to access and use it to solve a target problem than the proper selection of these key items reflects the participants’
middle school students were. A likely explanation for this result, attempt to match these items with the corresponding items in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

consistent with Piaget and Inhelder’s (1969) notion of reconstruc- source story.
tive memory, involves an enrichment process in the quality of It is worth noting that although recent studies have revealed that
representation following the original exposure to the source infor- ways of thinking may be associated with general social systems,
mation, with new experiences being incorporated into the child’s cultural values, and cultural practices (e.g., Nisbett, Peng, Choi, &
representation of the analogue. With age and experience, the Norenzayan, 2001), our present findings indicate that differences
representation of the source analogue may become increasingly in strategies and cognitive performance may also be traceable to
rich, accessible, and applicable. specific cultural experiences (e.g., events or stories). The present
unique approach may not be limited to the study of long-term
Methodological Implications analogical problem solving and may prove to be a fruitful avenue
for the investigation of other aspects of higher order cognition
An important feature of this series of experiments was that the across cultural contexts. This paradigm may be adapted to the
source story was introduced in naturalistic settings, such as reading exploration of how specific cultural experience influences cogni-
from a book or listening to a story, many years before the partic- tive processes involved in memory, conceptual understanding,
ipants took part in the present research, in which one of several knowledge representation, similarity comparison, reasoning, and
target problems was presented. The source analogue acquisition human learning in general.
and target problem solving took place in different physical and
functional contexts and at different points in time. Thus, partici- Conclusions
pants would not readily expect the links between the target prob-
lem and a particular experience, making it possible to examine the The present results demonstrate that long-term analogy is a
retrieval and use of the source analogue after a substantial delay. powerful tool for problem solving and transfer. The findings
A second feature of the present research involved examining the provide much needed and robust evidence that individuals are
major transfer processes independently but within the same task capable of solving problems by remote analogy with an extensive
context. The first process was accessing. Unlike prior work that delay after the acquisition of source information. The present
involved presenting a set of source stories and subsequently asking research further suggests that representations of stories or prob-
participants to choose which story that they were reminded of lems in long-term memory appear to contain both abstract struc-
when reading a target story (e.g., Gentner et al., 1993; Wharton et tures and specific object features and that human memory is
al., 1996), we asked participants to provide information about any sensitive to both types of information. Given the nature of these
story or event they were reminded of. The fact that no prior source representations, object similarity plays a critical role in long-term
information was provided by the experimenters helped prevent a analogical transfer. When different types of object similarity fulfill
potential problem of directed reminding or forced accessing different structural roles in source and target problems, different
(Wharton et al., 1996), in which participants might be guided to processes involved in transfer may be affected. Although the
attend to a particular set of stories or problem situations that they accessing process was found to be constrained largely by com-
read minutes or days earlier. Instead, in the present study, we monality in both goal object and solution tool, only similarity in
examined free reminding, which involves retrieval of the most solution tool influenced the mapping and executing processes. This
relevant information from long-term memory. distinction of object similarity type in the processes of analogical
It is also important to note that problem-solving performance transfer thus further confirms the multifaceted nature of similarity.
might not be a precise and reliable measure of retrieval of source The results also reveal developmental differences in remote
information, mainly because solvers may have been reminded of a transfer. Age differences in analogical transfer are not due to the
prior source analogue but nevertheless failed to solve the problem deficiency associated with one single cognitive component. The
(see Catrambone, 2002). Yet, as Hammond et al. (1991) have lower scores of middle school students were partially due to the
pointed out, it is important to observe the processes in the same difficulty experienced in accessing the source story, in mapping
problem-solving tasks to assess them accurately. We attempted to the key components between problems, and in implementing the
measure the accessing process separately, apart from mapping and solution by using the tool correctly and effectively. Perhaps with
executing processes, to examine their unique contributions to age, the more mature cognitive system and richer knowledge base
long-term analogical transfer in the same task context. Therefore, improve the efficiency in these three components.
LONG-TERM ANALOGICAL PROBLEM SOLVING 431

It is evident that overall relations between source and target naturalistic settings yet so difficult in the psychological laboratory. In D.
problems are multidimensional, involving various types of simi- Gentner, K. J. Holyoak, & B. Kokinov (Eds.), Analogy: Perspectives
larity, and that the cognitive activities associated with analogical from cognitive science (pp. 313–334). Cambridge, MA: MIT Press.
transfer are multicomponential. A comprehensive theory or com- Dunbar, K., & Blanchette, I. (2001). The in vivo/in vitro approach to
putational model of analogical problem solving needs to take the cognition: The case of analogy. Trends in Cognitive Sciences, 5, 334 –
different roles of object similarities into consideration, to identify 339.
Duncker, K. (1945). On problem-solving. Psychological Monographs,
key processes revealed in the same task context, and to incorporate
58(5, Whole No. 270).
findings evident in transfer over long-term delays across contexts.
Gardner, M. (1978). Aha! Insight. New York: Freeman.
Gentner, D. (1989). The mechanisms of analogical learning. In S. Vosni-
References adou & A. Ortony (Eds.), Similarity and analogical reasoning (pp.
199 –241). Cambridge, England: Cambridge University Press.
Anderson, J. R., Farrell, R., & Sauers, R. (1984). Learning to program in
Gentner, D., & Forbus, K. D. (1991). MAC/FAC: A model of similarity-
LISP. Cognitive Science, 8, 87–129.
based access and mapping. Proceedings of the Thirteenth Annual Con-
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ference of the Cognitive Science Society (pp. 504 –509). Hillsdale, NJ:
learn? A taxonomy for far transfer. Psychological Bulletin, 128, 612–
This document is copyrighted by the American Psychological Association or one of its allied publishers.

637. Erlbaum.
Bassok, M. (1990). Transfer of domain-specific problem-solving proce- Gentner, D., & Markman, A. B. (1997). Structure mapping in analogy and
dures. Journal of Experimental Psychology: Learning, Memory, and similarity. American Psychologist, 52, 45–56.
Cognition, 16, 522–533. Gentner, D., Rattermann, M. J., & Forbus, K. D. (1993). The role of
Brown, A. L. (1989). Analogical learning and transfer: What develops? In similarity in transfer: Separating retrievability from inferential sound-
S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning ness. Cognitive Psychology, 25, 524 –575.
(pp. 369 – 412). Cambridge, England: Cambridge University Press. Gentner, D., & Toupin, C. (1986). Systematicity and surface similarity in
Brown, A. L., & Kane, M. J. (1988). Preschool children can learn to the development of analogy. Cognitive Science, 10, 277–300.
transfer: Learning to learn and learning from example. Cognitive Psy- Gick, M. L., & Holyoak, K. J. (1980). Analogical problem solving.
chology, 20, 493–523. Cognitive Psychology, 12, 306 –335.
Catrambone, R. (1996). Generalizing solution procedures learned from Goswami, U. (1996). Analogical reasoning and cognitive development. In
examples. Journal of Experimental Psychology: Learning, Memory, and H. W. Reese (Ed.), Advances in child development and behavior (pp.
Cognition, 22, 1020 –1031. 92–135). New York: Academic Press.
Catrambone, R. (1998). The subgoal learning model: Creating better ex- Hammond, K. J., Seifert, C. M., & Gray, K. C. (1991). Functionality in
amples so that students can solve novel problems. Journal of Experi- analogical transfer: A hard match is good to find. Journal of Learning
mental Psychology: General, 127, 355–376. Sciences, 1, 111–152.
Catrambone, R. (2002). The effects of surface and structural feature Holyoak, K. J., Junn, E. N., & Billman, D. O. (1984). Developmental
matches on the access of story analogs. Journal of Experimental Psy- analogical problem solving skills. Child Development, 55, 2042–2055.
chology: Learning, Memory, and Cognition, 28, 318 –334. Holyoak, K. J., & Koh, K. (1987). Surface and structural similarity in
Catrambone, R., & Holyoak, K. J. (1989). Overcoming contextual limita- analogical transfer. Memory & Cognition, 15, 332–340.
tions on problem-solving transfer. Journal of Experimental Psychology: Holyoak, K. J., & Thagard, P. (1997). The analogical mind. American
Learning, Memory, and Cognition, 15, 1147–1156. Psychologist, 52, 35– 44.
Catrambone, R., & Holyoak, K. J. (1990). Learning subgoals and methods Keane, M. T. (1987). On retrieving analogues when solving problems.
for solving probability problems. Memory & Cognition, 18, 593– 603. Quarterly Journal of Experimental Psychology: Human Experimental
Chen, Z. (1996). Children’s analogical problem solving: Effects of super- Psychology, 39A, 29 – 41.
ficial, structural, and procedural similarity. Journal of Experimental Klahr, D., & Carver, S. M. (1988). Cognitive objectives in a LOGO
Child Psychology, 62, 410 – 431. debugging curriculum: Instruction, learning, and transfer. Cognitive
Chen, Z. (2002). Analogical problem solving: A hierarchical analysis of
Psychology, 20, 362– 404.
procedural similarity. Journal of Experimental Psychology: Learning,
Köhler, W. (1927). The mentality of apes. New York: Harcourt Brace.
Memory, and Cognition, 28, 81–98.
Kolodner, J. (1993). Case-based reasoning. San Mateo, CA: Morgan
Chen, Z., & Klahr, D. (1999). All other things being equal: Acquisition of
Kaufmann.
the control of variables strategy. Child Development, 70, 1098 –1120.
Kotovsky, L., & Gentner, D. (1996). Comparison and categorization in the
Chen, Z., & Mo, L. (2004). Schema induction in problem solving: A
development of relational similarity. Child Development, 67, 2797–
multidimensional analysis. Journal of Experimental Psychology: Learn-
2822.
ing, Memory, and Cognition, 30, 583– 600.
Chen, Z., & Siegler, R. S. (2000). Across the great divide: Bridging the gap Luchins, A. S. (1942). Mechanization in problem solving: The effect of
between understanding of toddlers’ and older children’s thinking. Mono- Einstellung. Psychological Monographs, 54(6, Whole No. 248).
graphs of the Society for Research in Child Development, 65(2, Serial Maier, N. R. F. (1931). Reasoning in humans: II. The solution of a problem
No. 261). and its appearance in consciousness. Journal of Comparative Psychol-
Clement, C. A., & Gentner, D. (1991). Systematicity as a selectional ogy, 12, 181–194.
constraint in analogical mapping. Cognitive Science, 15, 89 –132. Markman, A. B., & Gentner, D. (in press). Nonintentional similarity
Daehler, M. W., & Chen, Z. (1993). Protagonist, theme, and goal object: processing. In R. Hassin, J. A. Bargh, & J. S. Uleman (Eds.), The new
Effects of surface features on analogical transfer. Cognitive Develop- unconscious. New York: Oxford University Press.
ment, 8, 211–229. Needham, D. R., & Begg, I. M. (1991). Problem-oriented training pro-
DeLoache, J. S., Miller, K. F., & Pierroutsakos, S. L. (1998). Reasoning motes spontaneous analogical transfer: Memory-oriented training
and problem solving. In W. Damon (Series Ed.) & D. Kuhn & R. S. promotes memory for training. Memory & Cognition, 19, 543–557.
Siegler (Vol. Eds.), Handbook of child psychology: Vol. 2. Cognition, Nisbett, R. E., Peng, K., Choi, I., & Norenzayan, A. (2001). Culture and
perception, and language (5th ed., pp. 801– 850). New York: Wiley. systems of thought: Holistic versus analytic cognition. Psychological
Dunbar, K. (2001). The analogical paradox: Why analogy is so easy in Review, 108, 291–310.
432 CHEN, MO, AND HONOMICHL

Osgood, C. E. (1949). The similarity paradox in human learning: A ent effects on the access and use of earlier problems. Journal of Exper-
resolution. American Psychologist, 56, 132–143. imental Psychology: Learning, Memory, and Cognition, 15, 456 – 468.
Perkins, D. N., & Grotzer, T. A. (1997). Teaching intelligence. American Salomon, G., & Perkins, D. N. (1989). The rocky road to transfer: Re-
Psychologist, 52, 1125–1133. thinking mechanisms of a neglected phenomenon. Educational Psychol-
Piaget, J., & Inhelder, B. (1969). The psychology of the child. New York: ogist, 24, 113–142.
Basic Books. Schank, R. C. (1982). Dynamic memory. New York: Cambridge University
Reder, L. M., & Klatzky, R. L. (1994). Transfer: Training for performance. Press.
In D. Druckman & R. A. Bjork (Eds.), Learning, remembering, believ- Schunn, C. D., & Dunbar, K. (1996). Priming, analogy, and awareness in
ing: Enhancing human performance (pp. 25–56). Washington, DC: complex reasoning. Memory & Cognition, 24, 271–284.
National Academy Press. Siegler, R. S. (1989). Mechanisms of cognitive development. Annual
Reed, S. K. (1987). A structure-mapping model for word problems. Jour- Review of Psychology, 40, 353–379.
nal of Experimental Psychology: Learning, Memory, and Cognition, 13,
Singley, M. K., & Anderson, J. R. (1989). The transfer of cognitive skill.
124 –139.
Cambridge, MA: Harvard University Press.
Reed, S. K. (1989). Constraints on the abstraction of solutions. Journal of
Spencer, R. M., & Weisberg, R. W. (1986). Context-dependent effects on
Educational Psychology, 81, 532–540.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

analogical transfer. Memory & Cognition, 14, 442– 449.


Reed, S. K., & Bolstad, C. A. (1991). Use of examples and procedures in
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Sternberg, J. R., & Frensch, P. A. (1993). Mechanisms of transfer. In D.


problem solving. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 17, 753–766. Detterman & R. J. Sternberg (Eds.), Transfer on trial: Intelligence,
Reed, S. K., Ernst, G. W., & Banerji, R. (1974). The role of analogy in cognition, and instruction (pp. 25–38). Westport, CT: Ablex.
transfer between similar problem states. Cognitive Psychology, 6, 436 – Thorndike, E. L., & Woodworth, R. S. (1901). The influence of improve-
450. ment in one mental function upon the efficiency of other functions: II.
Reed, S. K., Willis, D., & Guarino, J. (1994). Selecting examples for The estimation of magnitudes. Psychological Review, 8, 384 –395.
solving word problems. Journal of Educational Psychology, 86, 380 – Wharton, C. M., Holyoak, K. J., Downing, P. E., Lange, T. E., Wickens,
388. T. D., & Melz, E. R. (1994). Below the surface: Analogical similarity
Ross, B. H. (1987). This is like that: The use of earlier problems and the and retrieval competition in reminding. Cognitive Psychology, 26, 64 –
separation of similarity effects. Journal of Experimental Psychology: 101.
Learning, Memory, and Cognition, 13, 629 – 639. Wharton, C. M., Holyoak, K. J., & Lange, T. E. (1996). Remote analogical
Ross, B. H. (1989). Distinguishing types of superficial similarities: Differ- reminding. Memory & Cognition, 24, 629 – 643.

Appendix A

Source Analogues for Experiment 1

A Standard Version of the “Weigh the Elephant” Tale Gist of the “Hansel and Gretel” Tale

Long ago in China, there lived a powerful emperor. Every year, the A woodsman and his wife were very poor and could not afford to feed
rulers of the surrounding countries had to give him jewelry, gold, cloth themselves and their two small children, Hansel and Gretel. As a solution,
the wife suggested that they lead the children out into the deep forest and
materials, and animals as presents. One day, a ruler of a southern country
leave them there. So the next day they took them out to the forest to chop
presented him with an elephant as a gift. The emperor was delighted to see
wood. Hansel, with his pockets full of pebbles, dropped one at regular
the elephant and asked the ruler what the weight of the elephant was. The
intervals to mark a path back home. Later, while the children napped, the
ruler was embarrassed because even the biggest scale he owned was too
parents returned home without them. That night, the children followed the
small to weigh the huge elephant. The emperor’s youngest son, named trail of pebbles that Hansel had left behind all the way home. When they
Chao Chong, came up with an idea: “You could find a boat, put the arrived home, the parents acted happy at their return but immediately
elephant in it, and mark the new water level on the boat. Then, you could planned to take them back into the forest again the next day. Hansel, aware
take the elephant out of the boat and put smaller stones into the boat until of the plan, wanted to go outside that night to get more pebbles. However,
the water level reaches the same mark. Then, you could weigh those stones the wife locked both of the children in their room before he could do so.
separately with a small scale. When you add up all the weights, you would The next morning, she gave both children a piece of bread for lunch.
know how heavy the elephant is.” Everyone was surprised and impressed Instead of saving the bread to eat, Hansel tore pieces off and left them on
by the boy’s solution. the trail to mark the way home. . . .
LONG-TERM ANALOGICAL PROBLEM SOLVING 433

Appendix B

The Target Problems Used in Experiment 1

The Statue Problem scale? Write down any ideas that you think might be appropriate. Any idea
is valid.
In a village by a river, the chief of a tribe had the responsibility of taking
care of the tribe’s sacred stone statue. It was the chief ’s custom every year The Cave Problem
to go down river to the next village to collect taxes. To assess the amount
of taxes, the chief would ask for the statue’s weight in gold coins. His A treasure hunter is going to explore a cave up on a hill near a beach. He
method of measuring this amount was to put the statue in a large tub at one suspected there might be many paths inside the cave so he was afraid that
end of a hanging balance scale and hook the other end of the scale to he might get lost. Obviously, he did not have a map of the cave; all that he
another large tub. This second tub was filled with gold coins until the scale had with him were some common items such as a flashlight and a bag.
balanced the weight of the statue. During a recent trip to collect taxes, the What could he do to make sure he did not get lost trying to get back out of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

chief forgot to bring his balance scale. Now the chief has a problem: How the cave later? Please write down any ways in which the treasure hunter
This document is copyrighted by the American Psychological Association or one of its allied publishers.

can he figure out how much gold to take to match the statue’s weight might be able to get out safely. Write down any ideas that you think might
without the benefit of a balance scale, a pulley system, or a conventional be appropriate. Any idea is valid.

Appendix C

An Example of the Target Stories Used in Experiments 2, 3, and 4

Dissimilar Goal Object, Dissimilar Solution Tool: Asteroid and To gather relevant information about the asteroid, the team needed to
Spring Target Problem measure its size and weight. The tape measure that they had allowed them
to measure the asteroid’s height and length. However, the scale that they
A group of geologists were traveling through a rainforest, on their way had brought could only measure up to 300 pounds of weight. Thus, the
to an excavation site, when they discovered a very large, strange-looking scale was obviously too small to weigh the asteroid. They looked around
rock. None of them were sure, but they thought that perhaps the rock was in a nearby village to see what resources they had to weigh the asteroid.
an asteroid that had fallen down from outer space thousands of years ago. How could they solve the problem?

Figure C1. The resources available to solve the asteroid and spring target problem.

Received May 12, 2003


Revision received February 14, 2004
Accepted February 20, 2004 䡲

You might also like