Professional Documents
Culture Documents
Lab in Cognition
Northeastern University
ANALOGICAL PROBLEM SOLVING 2
Abstract
problem solving. In this type of problem solving, individuals use an abstract understanding of a
simple problem to find a valid solution to a problem whose subject matter is extensively
unfamiliar. Successful analogical problem solving occurs in various steps. Our research
investigates the importance of how we learn the components of the source problem and its
subsequent effects. More specifically, we seek to establish the effect of imagery presence and
contextual relevance on successful analogical problem solving. We use the Dual-Coding Theory
and the Levels of Processing Theory to substantiate this. In order to view this interaction, we
tested a group of college students to view a source problem that was either relevant to college or
not. Furthermore, we supplemented imagery to these same source problems so that we may
understand their influence. Based on previous theory, we understand semantic relevance and
processing information in dual modalities to enhance encoding processes and later recall. We
assume these theories concerning how we encode information to have related applications.
Similar to their effect on memory, we reason that encoding information in two modalities and
with contextual relevance will enhance problems solving. We hypothesize that induction of
contextual/semantic relevance and the presence of an image will cause individuals to rate the
analogous solution as most valid, as compared to other, valid solutions. In our results, we find a
significant effect of imagery on analogical problem solving, while semantic relevance does not
present any effect. We also examine their influence on cognitive flexibility to create a substantial
simple domain to find a valid solution to a difficult problem. When thinking about this method of
problem solving, we begin to reason about how we conceptualize new information and
subsequently apply it. Learning is a process of schema development; it requires assimilation and
accommodation. The use of an analogy simplifies these otherwise, difficult processes. This is
comparing old information to new information and thus the use analogies in everyday life is
preexisting network of other information. We assume our means of comprehension and problem
cognitive processes. In this paper, we review current literature and assemble an experiment with
the intention to find exactly what these factors are. This type of problem allows us more insight
into cognitive reasoning, schema development, and our ability to make crucial connections of
preexisting knowledge to novel information. We take our preceding knowledge of the Dual
Coding Theory and the Deep Processing Theory and apply them to the domain of analogical
problem solving. In this interaction of theories in cognition, we attach a new emphasis and
Past researchers have investigated analogical problem solving and effectively provide us
with our understanding of analogical problem solving. Gick and Holyoak (1980) presented an
information from one domain (a source problem) to help solve a problem from a disparate
domain (a target problem). A source problem is characterized as a problem solved in the past,
ANALOGICAL PROBLEM SOLVING 4
which contains simple subject matter. Generally, the answer to this problem is provided. A target
problem is a problem that is to be answered using the source problem. The subject matter is
difficult and unfamiliar to an individuals domain of general knowledge. Gick and Holyoak
(1980) completed this by using the Fortress Problem (among others) as the source problem and
the Tumor Problem as the target problem. They stressed that the semantic consistencies of each
problem are essential to ensuring a legitimate and apparent analogy between the two problems.
By asking the participants to give their own answers they recorded the participants ability to
recognize the analogy, connect the analogy to the target problem, and then successfully apply it.
In their main conclusion, they resolved that solutions to complex problems can be developed
using analogous problems from different domains, especially when hinted that an analogy
For the purposes of our study, we present the ways in which individuals mentally
represent information and how these perceived representations vary by circumstance. They
further detailed the factors needed to reason about difficult problems when using preceding
information. Gick and Holyoak suggest that understanding the abstraction of the source problem
is essential to noticing the analogy. This comprehension is necessary to guide the problem-
solving process. Sternberg (1977) investigates this idea of having to understand the given
solving. This is the part of problem solving that our experiment plans to capitalize on, guiding
the process of schema development and seeing its effects on problems solving. Kintsch and Van
Dijks (1978) further add to this by arguing that analogies have multiple levels of abstraction. The
type of abstraction most suitable for analogical problem solving are named, Macrostructures
which are a broad set of concepts that allow the application of one solution to other problems.
ANALOGICAL PROBLEM SOLVING 5
In review of our study, we assume the importance of how we obtain this most ideal type
of representation. To investigate the procedures needed to obtain this, we regard two applicable,
cognitive theories of how we encode information: The Levels of Processing Theory and the
Dual-Coding Theory. Craik and Tulving (1975) investigated how the different ways in which we
process information can affect how we perceive and therefore later apply that information.
Conditions that demand elaborative processing (or deep level processing) encouraged better
recall as compared to conditions that demanded structural processing (or shallow level
processing). By perceiving words that demand more effortful encoding and by using words
intrinsically familiar to ones pre-existing semantic network, the participants displayed greater
processing information in dual modalities. The Dual-Coding Theory proposes that we represent
visual information and verbal information in separate, yet related systems. When observing
information in both verbal and visual contexts, participants tend to recall that same information
significantly more often when the information is encoded in a single modality. Assessing
information in both modalities construct more comprehensive schemas, needed to later apply that
information. Specifically, in the case of Clark and Paivio (1991), they apply the theory to
educational practices. They describe the Dual-Coding Theory to be necessary in creating models
of learning. They argue that processing content in multiple modalities is essential to the
Does the inclusion of semantic relevance in a source problem improve a participant's ability to
problem solve analogously as compared to a participant provided with source problem that is
analogical problem sets as compared to individuals who do not perceive any imagery? Is there an
problem solving?
The proposed research questions are created with the intention of understanding
analogical problem solving and the mechanisms that influence it. We use Gick and Holyoak
(1980) to reproduce their most basic outline to administer analogical problem sets. To support
our research goals, we assume the contingency between Dual Coding Theory and the Levels of
Processing Theory to problem solving processes. With this, we hope to illustrate the effects of
imagery and contextual relevance. With the inclination of dual-coding and deep level processing
to enhance memory, we assume these effects will extend to other cognitive processes. By
translating the non-college relevant fortress problem into the college-relevant problem, we
maximize the relevancy of the domain. By adding images, we increase ones capacity for finding
abstractions from the given information. These manipulations decrease for these reasons, we
hypothesize that the relevant condition will result in higher validity ratings to the analogous
solutions as compared to the ratings of the non-relevant conditions. We support this prediction
with the findings of Craik and Tulving (1975). We also hypothesize that the participants of the
conditions that include supplementary imagery will rate the analogous solution higher than the
conditions that do not include imagery. We support this prediction with Clark and Paivio (1991).
We suppose that with an interaction between the two variables, the relevant condition that also
has supplementary imagery will rate the analogous solution as most valid as compared to the
Our experiment uses a method similar to Gick and Holyoak (1980) to address our
presented research questions. The source problem is presented to the participant before the target
problem. In our study, we present a distraction task between the two problems to ensure the
contingency between the problems is not easily recognized. Our variables of visual
representations and semantic relevance is used to manipulate the context in which we present the
source problem, and hence, how the participants mentally represent the information of the source
problem. Past research has not attended to the potential effect this may of on analogical problem
solving. We attend to this concept by including imagery and contextual relevance in light of the
Fortress Problem (Gick and Holyoak - 1980) as our source problem and the Tumor/Radiation
Problem (Gick and Holyoak - 1980) as our target problem. To test for relevance, we created a
relevant problem. This problem was built off of the exact framework of the Fortress problem, but
made relevant to our college participant sample. To test the effect of imagery, we created images
as to match the setting of the Fortress Problem and the School Problem. In a cross comparison of
these variables, we tested the four conditions: Fortress Problem (non-relevant, no image),
College Problem (relevant, no image), Visual Fortress Problem (non-relevant, image present),
and Visual College Problem (relevant, image present). Past research relied on the participants to
produce their own answers to the target problem. Instead, we provided the answers and asked the
By viewing the validity rating as opposed to free response, we are able to assess
our conclusion, we conclude the effects of visual representations and contextual relevance on
Method
Participants
32 undergraduate students of Northeastern University participated in this study. The ages of these
participants ranged from 18 to 24 years of age. 2 of these participants were male. 30 of these
participants were female. 31 of the participants were native English speakers. We recruited
race were taken into consideration. After completion, the participants were offered a cupcake for
their participation.
Method
The entirety of the experiment was compiled into paper surveys. There were four
different packets for each experimental group: non-college relevant fortress Problem, non-
college relevant fortress problem with image, college-relevant problem, and college-relevant
problem with image. The packets that we used to examine the participants had very similar
contents and followed a very similar order. Each part of our materials was used on separate
Informed Consent. We asked every participant to complete the informed consent form.
This detailed their voluntary commitment to the study, being over eighteen years of age, as
Source Problem. In all conditions, the answer to the source problem was provided. At the
top of the page of the source problem, the direction read, Please take time to review the
story provided below. If the problem also had an image to accompany it on the next page,
the directions stated, Please take time to review the story below. This page behind this
ANALOGICAL PROBLEM SOLVING 9
includes an image that accompanies the story. Please review this image as well. Below the
direction, we included the Fortress Problem (Figure 1) or the College Problem (Figure 2),
depending on condition.
Image. The images were present in only two of the four conditions. The visual of the
Fortress Problem may be seen in Figure 3 and the visual for the School Problem may be
viewed in Figure 4. The images presented display of a top-down view of the setting related
to the given problem. Both images were colored and had a key to point out what each
Distraction Task. The distraction task was a demographics worksheet. The worksheet
specified for the participant to provide information to each question asked. It further
specified that this information is held anonymously. The participants gave their answers on
the lines provided next to each question. The questions asked the participants to provide
their age, gender, enrollment status in NU, major, minor, graduation date, participation in
Target Problem. All experimenter packets had the tumor problem. This problem be viewed
in Figure 5. The problem outlines the that a ray would destroy a patients tumor. However,
the ray is so powerful that using it would destroy healthy, surrounding tissue. The problem
Validity Rating Sheet. At the beginning of the page, the directions stated, For the last part
of the study, we ask you to consider the (tumor) problem above. Please provide a whole
number rating to each answer provided. Each answer corresponds to how valid you feel the
answer is to the problem above. Please rate from 0 (not valid) to 100 (most valid) in the
boxes provided. Kindly let the experimenter know when you are finished. Below the
ANALOGICAL PROBLEM SOLVING 10
instructions were three acceptable answers to the Tumor Problem (Figure 6). Next to each
Design
the contextual relevance of the source problem (relevant vs non-relevant) and the use of imagery
paired with the source problem (image present vs no image). For our dependent variable, we
measured the participants subjective validity rating of three different provided answers to the
The experiment followed a very simple outline. As it is for all problems in analogical
problem solving, the participant first receives the source problem, and then the participant
receives the target problem. Figure 7 shows the order in which the study was presented along
with the four experimental groups used. These groups included the Fortress Problem (non-
relevant context, no image), College Problem (relevant context, no image), Visual Fortress
Problem (non-relevant context, fortress image present), and Visual College Problem (relevant
context, college image present). Contextual relevance and imagery are our independent variables.
Contextual relevance refers to the two different types of source problems we used. Each
differed by how relevant the story was to our college participant sample. Gick and Holyoak
(1980) used the Fortress Problem (Figure 1) and we do the same. We consider this problem to be
a part of our non-relevant condition since it presents no relationship to the college students who
are to use it as an analogy. The problem proposes a situation where as a general is unable to take
over a fort due to a given constraint. The answer to this problem is that the fortress must be
attacked by converging on multiple roads. To make the context more relevant to our intended
ANALOGICAL PROBLEM SOLVING 11
participants, we created a new story that had almost complete resemblance to the Fortress
Problem. The problem has the same constraints and answer as the Fortress Problem, but is
related to the context of college, name the college relevant problem. The abstraction that one is
to represent from the Fortress problem is to be the same as the abstraction found in the College
Problem. In the college Problem, a certain restraint causes a student president and other
students to march on to campus using multiple roads. The differences in the problems almost
exclusively occur in the word usage. It is the semantic differences that we alter in the story, not
the actual structure or theme of the story. We find this difference to be essential. By changing the
inherent relevance of the words, we change the context of the story while upholding the integrity
of the required analogy. In these specific problems, the source problems are provided with
answer; the participants never had to provide their answers to the source problem.
By adding imagery to the source problem, we tested the effect of visual representations in
analogical problem solving. The imagery was made to fit the context of each problem. The
Fortress Problem was supplemented with an image that had the fortress, roads that converged
onto the fortress and surrounding buildings (Figure 3). The College-Relevant Problem was
supplemented with an image that had the college, the roads that converged onto the college, and
surrounding buildings (Figure 4). Both images had a key to ensure the participants could fully
comprehend the figures in each image. These images represent a top-down view of source
problems setting. We made the top-down view for the purposes of stressing the importance of
the roads which converge onto the central object (fortress or college). The idea of convergence
from different angles is essential to this case of analogical problem solving. This image is given
to provide basic framework of the problem, not to add more details, otherwise not given in the
problem.
ANALOGICAL PROBLEM SOLVING 12
Before handing the participants the target problem, we give them a distraction task guised
distraction. This ensures that the participants do not find an obvious reason to believe that the
For our target problem, we use the Gick and Holyoak (1980) once more. We used the
tumor problem (Figure 5). The problem considers that a patient is in need of destroying a tumor
in his/her stomach. However, the ray needed to destroy the tumor is so powerful that it would
destroy healthy tissue. The problem asked how you how you may get around this constraint. Our
experiment provides three different solutions to this problem. One of these answers provides an
analogous solution to the problem. The analogous solution states, Apply low-intensity rays from
several different directions so they simultaneously converge at the tumor. This solution is
analogous because the solution of the source problem matches the solution to the target problem.
The army/student body cannot attack the fortress/college from one road just like the ray cannot
be directed at the tumor at one angle. In this way, the parallels between the source problem and
the target problem emerge. The Tumor Problem must be solved in the same way the source
problems are, by an attack from converging sides. We provided two other solutions from Gick
and Holyoak (1980). These were our distraction solutions. They provided valid answers to the
constraints of the problem but were not analogous to the source problem. One solution stated,
Send high-intensity rays down the esophagus so they strike the tumor. The other distraction
solution stated, Insert a tube through the healthy tissue to the tumor, and then send high-
The study asked the participants to rate the validity of each individual solution. The
subjective scale ranged from 0 (not valid) to 100 (most valid). The rating scale provided us with
the ability to understand the data in a different way, since it is not so much related to past studies.
The rating provides us the understanding if someone applies the analogy or not and tell us more
about the role of our independent variables. If having used a method of producing solutions or
using a yes/no method, our results would not have been as manageable. As you will later see, the
rating scale allows us to use different ways to analyze our results since we are able to observe the
Design
The participants used the website, Doodle, to sign up for the participation of the study.
Each participant showed up at the experiment location at their designated time. The study took
place in a quiet room of the university library. The participants completed the study individually.
participant did not receive the whole packet at one time. Instead, the experimenter presented each
page of the packet at different times. First the participant received the informed consent form.
After signing the informed consent form, the participant returned the form to the experimenter.
The participant was then presented with either just the source problem or the source problem
with the related image. This was based on the participants experimental group assignment. The
participants were asked to read the instructions. This asked the participant to read the story and
view the related image. The participants notified the experimenter once they were done reading
the story and looking at the image and key (if provided). The experimenter then collected these
pages from the participant. The experimenter then provided the participant with the distraction
ANALOGICAL PROBLEM SOLVING 14
task/demographics worksheet. The participants recorded their answers and completed the
worksheet. After completion of the demographics worksheet, the experimenter provided the
participants with the tumor problem and the answer sheet. The participants were instructed to not
look the next page (the answer sheet) until done reading the problem presented. Using verbal
instruction, the experimenter recited to the participants, Please take your time to read the next
problem. Once you are done reading the problem, please flip to the next page and read/follow the
directions provided. You may reference the (tumor) problem again if need be. Once the
participants completed reading the problem, the participant turned to the next page and read the
instructions. After reviewing each of the provided answers, the participants rated their validity on
a scale of 0 to 100. The participant turned in the last two pages to the experimenter.
The participants were thanked for their participant and offered a cupcake as
compensation or their time. The experimenter then debriefed the participant by stating what the
study was about and what the goal of the study was. The study lasted approximately 10 minutes.
Results
Scoring
Our dependent variable was a validity rating score given to all three-possible given tumor
problem solutions. These scores ranged from 0 (not valid) to 100 (most valid). All rating scores
were accounted for and averaged across each experimental condition. We measured and
compared these means in three different ways. Our results include an analysis of analogous
validity, analogous endorsement, and cognitive flexibility. We define analogous validity to be the
mean rating the participants gave to the low-intensity answer. This only used the means of the
most analogous answer. Analogous endorsement is a comparison of the means for each answer
ANALOGICAL PROBLEM SOLVING 15
and thus utilized the validity scores from each answer. Given the participant rated the analogous
answer to be more valid than the other two answers, we term the participant to have endorsed
participants did not endorse the solution, we marked that participant with a 0. Cognitive
flexibility is regarded as the case in which a participant rates two or more answers to be valid. In
this study, we assume a valid score to be equal to or more than 70. The mean scores for each
answer were taken when reviewing cognitive flexibility. The participants who answered the
questions with cognitive flexibility were marked with a 2. The participants who rated only one
answer as valid was marked with 1. No participants were marked with a 0 since all
participants rated at least one answer as valid. We judged the significance threshold (p-value) to
Analysis
Based on previous theories we expected that there should be a significant effect of image
presence and a significant effect of relevance on validity scores. We also anticipated that there
would be an interaction between the two in which college-relevance and image presence would
Our analysis is divided into three different approaches. First, we ran analysis on
agreement/belief of validity for the low intensity answer. Figure 8 shows the descriptive mean
comparison for relevance by image. Results show that there were higher ratings for the low-
intensity answer for the non-relevant (fortress) problem with an image and without an image
compared to the college-relevant problem. Participants also produced higher ratings on both
To examine the effects for analogous validity scores we ran a 2 (relevance) x 2 (image)
ANOVA on agreement/belief of validity for the Low Intensity Answer. There was a significant
main effect for image, F=1.877, p<.2. There was no significant main effect of relevance,
F=.106, p>.2. Additionally, there was no significant interaction of image and relevance, F=.138,
p>.2. To determine the main effect of image, an independent samples t-test investigating the
effect of presence an image on endorsement of low intensity solution. There was a significant
Our results supported our predictions for the effect of imagery. We predicted that that
imagery added to the presentation of the source problem would result in increased validity scores
as compared to source problems that are presented without imagery. Imagery did cause a
significant increase in the validity rating scores. However, our findings did not support our
predictions for the effect of contextual relevance. We predicted that the relevant context of the
source problem would increase the validity ratings as compared to non-relevant condition. No
compared to the distraction solutions. Figure 9 shows the descriptive mean comparison for
relevance by image for analogous endorsement. Results show that there was better performance
for rating of analogous validity of the low intensity answer with an image present for the college-
relevant problem. However, there was no change in scoring for the non-relevant (fortress)
problem when an image was present. Participants also performed better on the Non-college-
relevant fortress problem than the college-relevant problem when no image was present.
ANOVA on endorsement of low intensity rays over the other two options. There was a trending
ANALOGICAL PROBLEM SOLVING 17
significant effect of image, F=1.068, p=.310. There was no significant effect of relevance,
F=.119, p=.733. Additionally, there was a trending significant interaction of image and
relevance, F=1.068, p=.310. To determine the effect of image we then conducted an independent
sample t-test looking at the effect of the presence of image on endorsing the low intensity
solution over the other solutions. It shows that there is a slight trending significance for the
conducted two independent samples t-test of image presence. These data are organized by
relevance condition. For college relevance, the t-test shows a slight trending significance for the
presence of an image. For non-college relevance, the t-test shows no significance. Ultimately,
these results indicate that the presence of an image did nothing for participants abilities.
one solutions validity greater than or equal to 70. Figure 10 shows the descriptive mean
comparison for relevance by image for cognitive flexibility scoring. For college-relevant
problems, results show greater cognitive flexibility in the absence of an image. Results also
indicate greater performance overall on cognitive flexibility for the non-college relevant problem
across image presence. For non-college relevant problems, participants had greater cognitive
flexibility scoring in the presence of an image compared to the absence of an image. These
findings suggest that participants show more cognitive flexibility in the non-relevance and image
absent domains.
ANOVA. There was no significant main effect of image, F=.156, p=.696. However, there was a
significant effect of relevance, F=3.889, p<.2. Additionally, there was a significant interaction of
ANALOGICAL PROBLEM SOLVING 18
cognitive flexibility. Results show significant higher cognitive flexibility ratings for the non-
college relevant condition, p<.2, compared to the college relevant condition which showed no
significance.
presence of images on cognitive flexibility (broken down by relevance). Results show a non-
significant effect of image in the non-college relevant domain. However, results also indicate a
significant effect of image in the college-relevant domain. The direction of the effect shows that
as the presence of an image goes from present to absent, flexibility scores show a significant
Discussion
Our results supported our predictions for the effect of imagery. We predicted that that
imagery added to the presentation of the source problem would result in increased validity scores
as compared to source problems that are presented without imagery. Imagery did cause a
significant increase in the validity rating scores. However, our findings did not support our
predictions for the effect of contextual relevance. We predicted that the relevant context of the
source problem would increase the validity ratings as compared to non-relevant condition. No
In addition to these results, we also saw in the case of analogy endorsement that an image
aided individuals who received a relevant story. However, the same did not occur for the
presence of non-relevant information. It appears that participants give greater endorsement of the
analogous solution when an image is present and when information is relevant. In contrast to
analogical problem solving we gained insight into cognitive flexibility. Participants were less
ANALOGICAL PROBLEM SOLVING 19
capable of creative thinking when more information was present. For both relevance groups,
Our results are evidence because it followed a pattern based on our research question
supported by previous research. Images aid in analogical problem solving and hinder creative
cognitive flexibility. When comparing to past studies we see a method very similar to Gick and
Holyoak (1980). Although, there was additional manipulations including the presence of an
image and relevance. With that our method also draws on Craik and Tulving (1975). Our
experiment took components of both studies to develop a greater understanding of the context of
Gick and Holyoak (1980) conclude that analogies aid in problem solving. Being the
foundation of our study their theory is supported. Additionally, our study presents information
using an image to represent multiple modalities. Results showing that an image aids in
comprehension of knowledge and application of analogy support Clark and Paivios (1991) Dual
Coding Theory. Alternatively, we did not consistently see the effect of relevance in analogous
validity. Therefore, their theory of levels of processing is not fully supported as it applies to
There are considerable limitations to our experiment that may have changed the outcome
or interpretation of our results. With an increase sample size of more than 32 participants we may
have seen the trending effect of analogous endorsement transition over into a main effect.
Additionally, our source problem manipulation was based in semantic relevance. This
manipulation limits the full scalability of relevance. Greater manipulation of the relevant source
problem may have produced a significant effect of the relevant condition across analogous
validity and endorsement. If participants were introduced a source problem that was contextual
ANALOGICAL PROBLEM SOLVING 20
manipulated to maximize relevance we might see an effect. As an extension to this study we seek
would involve the identifying the threshold for comprehension in levels of processing.
Ultimately, our study examines the effects of contextual relevance and imagery on
analogical problem solving. In studying these effects, we gain greater insight into what aids us in
our problem-solving capabilities. As we see imagery aid our analogical problem solving there are
implications in how we interact with everyday challenges especially in the learning environment.
Whether it be an academic setting or an everyday situation we see that images may help when
applying information from one source to a novel domain. As we continue to research analogical
problem solving and the modalities that can stand to affect them we may see additional aids in
References
Clark, J. C. & Paivio, A. (1991). Dual coding theory and education. Educational Psychology
Craik, & Tulving (1975). Depth of processing and the retention of words in episodic memory.
Gick, M. L. & Holyoak, K. J. (1980). Analogical problem solving. Cognitive Psychology, 12,
306-355.
Kintsch, W., & Van Dijk, T. A. (1978) Toward a model of text comprehension and production.
353-378.
ANALOGICAL PROBLEM SOLVING 22
Figures
60
50
40 Image
30 No Image
20
10
0
College Relevant Non-College Relevant
Source Problem
0.6
0.5
0.4
Image
0.3
No Image
0.2
0.1
0
College Relevant Non-College Relevant
Source Problem
1.2
1
0.8 Image
0.6 No Image
0.4
0.2
0
College Relevant Non-College Relevant
Source Problem