You are on page 1of 31

Journal of the Learning Sciences

ISSN: 1050-8406 (Print) 1532-7809 (Online) Journal homepage: http://www.tandfonline.com/loi/hlns20

On the Use of Analogy in Text-Based Memory


and Comprehension: The Interaction Between
Complexity of Within-Domain Encoding and
Between-Domain Processing

Leslie J. Caplan & Carmi Schooler

To cite this article: Leslie J. Caplan & Carmi Schooler (1999) On the Use of Analogy in Text-
Based Memory and Comprehension: The Interaction Between Complexity of Within-Domain
Encoding and Between-Domain Processing, Journal of the Learning Sciences, 8:1, 41-70, DOI:
10.1207/s15327809jls0801_2

To link to this article: http://dx.doi.org/10.1207/s15327809jls0801_2

Published online: 17 Nov 2009.

Submit your article to this journal

Article views: 38

View related articles

Citing articles: 2 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=hlns20

Download by: [Library Services City University London] Date: 18 March 2016, At: 00:54
THE JOURNAL OF THE LEARNING SCIENCES, 8(1), 41-70

On the Use of Analogy in Text-Based


Downloaded by [Library Services City University London] at 00:54 18 March 2016

Memory and Comprehension:


The Interaction Between
Complexity of Within-Domain
Encoding and Between-Domain
Processing
Leslie J. Caplan and Carmi Schooler
National Institute of Mental Health
Bethesda, Maryland

In this article, we examine the role of analogy transfer in text comprehension. We


tested the hypothesis that people would benefit more from manipulations encourag-
ing the induction of common structure from source domains when they had previ-
ously engaged in complex encoding of those domains. Participants read sets of pas-
sages about analogous domains (i.e., topics). Each set included 1 or 2 source passages
followed by a target passage. The independent variables were: (a) degree of complex-
ity of source passage encoding; (b) amount of information about the nature of the
analogy provided with the target passage (no information, a title stating that the do-
mains were analogous, or a title stating that the domains were analogous plus a de-
scription of their common structure); and (c) number of source passages in a set (1 or
2). Participants then were tested on memory and inference for the target passage. For
the memory measure, performance was better following complex encoding than fol-
lowing simple encoding only when the short title had been provided. For the inference
measure, performance was better following complex encoding than following simple
encoding only when 2 source passages had been provided. These findings support the
hypothesis that complex source encoding provides the best support for the use of ex-
plicit analogy. More generally, they suggest that the effects of within-domain encod-

Correspondence and requests for reprints should be sent to Leslie I. Caplan, Section on
Socio-Environmental Studies, National Institute of Mental Health, National Institutes of Health, Fed-
eral Building, Room BIA-14, Bethesda. MD 20892-9005.
42 CAPLAN AND SCHOOLER

ing complexity on analogical transfer depend on whether between-domain process-


ing has been abstraction- or episode-based.

The question of how people bring past experiences to bear on new ones (i.e., trans-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

fer) is an old question in psychology, education, and philosophy. In this article, we


examine possible mechanisms underlying the role of analogy in a
text-comprehension task. In cognitive psychology, such use of analogy has been
studied primarily in the context of analogical transfer. Although researchers and
theorists have some difficulty agreeing on an exact definition of analogical transfer,
most would concur with the claim that it involves transfer from a previously famil-
iar (source) domain, problem, or event to a novel (target) one that has an identical,
or very similar, underlying structure. Most would also agree with Gick and
Holyoak's (1983) claim that, "The essence of analogical thinking is the transfer
from one situation to another by a process of mapping-finding a set of one-to-one
correspondences (often incomplete) between aspects of one body of information
and aspects of another" (p. 2).
In particular, we focus on the roles of two kinds of processing in analogical
transfer: within-domain and between-domain processing (see also Cummins,
1992, and in a related vein, Mayer & Greeno, 1972). We use the term domain as do
other researchers in the field (e.g., Fong & Nisbett, 1991; Gentner, 1989; Holyoak
& Koh, 1987). According to Gentner (1983), "Domains and situations are psycho-
logically viewed as systems of objects, object-attributes and relations between ob-
jects" (p. 156). More generally, domain is usually synonymous with content
domain in the analogical transfer literature. In this article, we use the term domain
to refer to a subject or topic (e.g., algebra rate problems, the geography of Africa, a
computer programming language, the work of a particular novelist), including its
component objects and concepts and the relations among them.
Within-domain processing involves the encoding, representation, and retrieval
of information from a single domain-regardless of its relations with other do-
mains (or problems or events). In this article, we are concerned with one particular
aspect of within-domain processing: encoding complexity. By encoding complex-
ity, we refer to the extent to which the component concepts and interconceptual re-
lations of a domain have been encoded: In simple encoding, relatively few are
encoded, whereas in complex encoding, many, if not all, are incorporated in the
domain's representation.
In contrast, between-domain processing involves the encoding, representation,
and retrieval of information about multiple domains and their interrelations. We
subsequently describe how between-domain processing can involve either epi-
sode-based processing (where each domain is represented relatively holistically
and independently, with little analysis of interdomain similarities), or abstrac-
COMPLEXITY AND ANALOGY 43

tion-based processing (where a common structure underlying two analogous do-


mains has been abstracted and encoded).

WITHIN-DOMAIN PROCESSING: ENCODING


Downloaded by [Library Services City University London] at 00:54 18 March 2016

COMPLEXITY

Until recently, investigators of learning and transfer have paid relatively little atten-
tion to the role of source domain processing and representation in transfer to a new
domain. In the last few years, however, several researchers have reported that
source domain representations that are complex by our definition (i.e., that include
many concepts and interconceptual relations) are superior in eliciting transfer per-
formance. For example, Gentner, Rattermann and Forbus (1993) suggested that
"relational access," which presumably increases the probability of analogical
transfer, may be enhanced by what they call "intensive encoding" (i.e., encoding
that includes many of the lower and higher order interconceptual relations in a do-
main) or by expertise. Similar distinctions occur in the expert-novice literature. For
example, Novick (1988) demonstrated that experts (whose representations of
source problems are likely to be relatively complex) are more likely than novices to
demonstrate analogical transfer in algebra, particularly when the two problems
share relatively few surface characteristics. In developmental work, Brown (e.g.,
Brown, 1989; Brown & Kane, 1988; Brown, Kane, & Echols, 1986) demonstrated
that when children are familiar with the causal or relational structure (i.e., the rela-
tions among concepts and events) of a source domain, they are more likely to dem-
onstrate spontaneous analogical transfer.

BETWEEN-DOMAIN PROCESSING:
ABSTRACTION-BASED VERSUS
EPISODE-BASED PROCESSING

In contrast with the relative paucity of work on the importance of within-domain


processing for analogical transfer, the nature of between-domain processing has
long been critical to theoretical developments in transfer and other related fields in
the psychology of learning and memory. Over the last decade or so, a common the-
oretical dichotomy regarding two possible kinds of between-domain processing
has emerged in many research areas that address transfer. In studies of categoriza-
tion, memory, problem-solving transfer, text comprehension, and perception, re-
searchers have proposed a distinction between what we will here refer to as epi-
sode-based and abstraction-based processing (for a related discussion with respect
to problem solving and categorization, see Medin & Ross, 1989).
44 CAPLAN AND SCHOOLER

Episode-Based Processing

As a general term, episode-based processing involves the representation and re-


trieval of discrete events (or objects or domains) that a person has experienced.
In this type of processing, underlying features or structures that are common
Downloaded by [Library Services City University London] at 00:54 18 March 2016

across events remain largely unanalyzed in the representation, and relatively few
relations among events are explicitly included in the representation. Instead,
people encode and retrieve each event (or object or domain) independently, as a
discrete unit. Such processing is apparent in exemplar models of memory (e.g.,
Hintzman, 1986) and category representation (e.g., Medin & Schaffer, 1978); in
nonanalytic processing in a variety of domains involving perception and catego-
rization (e.g., Brooks, 1987); in textbases as the propositional representations of
particular texts (e.g., Kintsch, 1986; van Dijk & Kintsch, 1983); in the role of re-
minding (e.g., Ross, 1984; Ross & Kennedy, 1990); and in analogy-based trans-
fer (e.g., Holyoak, 1984; Holyoak & Koh, 1987) in problem solving. As applied
to the context of this article (i.e., that of analogical transfer in text comprehen-
sion), this type of processing involves retrieving a familiar domain to apply the
knowledge of (or response to) that domain to a novel one. In other words, repre-
sentations of similar previous domains are retrieved, and people respond to the
new domain as they had to previously experienced individual ones, presumably
by engaging in some mapping process between representations of the source and
target domain.

Abstraction-Based Processing

In general, abstraction-based processing involves the abstraction and subsequent


storage and retrieval of some rule, concept, schema, or feature(s), typically those
common across a number of objects, domains, or events. Abstraction-based pro-
cessing is involved in prototype and feature models of categorization (e.g., Barr
& Caplan, 1987; Rosch, 1973); analytic processing in perception and categoriza-
tion (e.g., Brooks, 1987); schema-based reasoning in problem solving and trans-
fer (e.g., Catrambone & Holyoak, 1989; Gick & Holyoak, 1983; Holyoak,
1984); and in situation models-schematic representations of a domain or situa-
tion described by one or several texts (e.g., Kintsch, 1986; van Dijk & Kintsch,
1983). In the context of analogical transfer of learning, this type of processing
involves recognizing that a new domain shares some underlying, nonobvious
characteristic or structure with a category of familiar domains, and using this
knowledge in responding to the new domain, presumably by engaging in map-
ping processes between the abstraction-based schema and the novel domain (for
a more detailed discussion of the categorization processes involved, see Reed,
1989). In contrast with episode-based processing, in which an individual ac-
COMPLEXITY AND ANALOGY 45

cesses and uses a previously familiar single domain, abstraction-based analogi-


cal transfer involves the abstraction of the common structure that underlies the
analogy, encoding it, and accessing it when a novel domain is subsequently en-
countered.'
Downloaded by [Library Services City University London] at 00:54 18 March 2016

The Relative Roles of Abstraction- and Episode-Based


Processing for Memory and Inference

Interestingly, abstraction-based and episode-based processing appear to enhance


performance on different kinds of tasks. Donnelly and McDaniel(1993) found that
performance on an inference task was better following analogy-based learning than
following literal learning (i.e., learning for which no analogy was provided); the re-
verse was true for memory for basic concepts. Other investigators also have found
that many variables have differential effects on simple retrieval and inference (e.g.,
Camp, Lachman, & Lachman, 1980;Mannes & Kintsch, 1987;Mayer, 1975,1976;
Mayer & Greeno, 1972; Ross, 1989). In general, learning experiences that encour-
age abstraction-based processing seem to enhance inference, problem solving, and
far transfer, whereas those that encourage episode-based processing seem to be best
for simple retrieval tasks.

ON THE INTERACTION BETWEEN WITHIN-DOMAIN


AND BETWEEN-DOMAIN PROCESSING

What is the relation between within- and between-domain processing in transfer?


Some authors have suggested that the latter aspect is more important than the for-
mer. For example, Cummins (1992) found that individuals who compared algebra
problem structures performed better than those who had analyzed the structure of
individual problems on a variety of problem-solving tasks. She concluded that be-
tween-domain processing is more important than within-domain processing in de-
termining the likelihood of transfer. In contrast, we propose that the relation be-

'We recognize that in the computer learning literature, there are many instances of case-based induc-
tion of schemas or principles, resulting from exposure to a single instance. Some researchers in human
learning also have pointed out that "intraproblem" analysis could, logically, allow one to make such in-
ductions (e.g., Cummins, 1992). However, such learning tends to require a substantial knowledge base
(e.g.. Mitchell, Keller, & Kedar-Cabelli, 1986). Furthermore, the literature on human learning has
tended to stress the abstraction of such structures from exposure to multiple instances, and the data sup-
port this emphasis (e.g., Cummins, 1992; Gick & Holyoak, 1983). We will, therefore, restrict our term
abstraction-basedto refer to those abstractive processes resulting from exposure to multiple instances,
although we recognize that, under some circumstances, exposure to a single instance could conceivably
lead to a similar result.
46 CAPLAN AND SCHOOLER

tween within-domain and between-domain processing is better characterized as an


interaction than as a competition. We suggest that the nature of the be-
tween-domain processing involved (i.e., whether an individual is engaging in ab-
straction-based or episode-based processing) can determine which type of
within-domain processing (i.e., degree of encoding complexity) will be most effec-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

tive in eliciting transfer.


Some support for this claim comes from a study by Caplan and Schooler
(1990). In that study, people were trained to use a microcomputer drawing pack-
age. They were trained with or without an analogical model of the package; the
model condition was developed to facilitate abstraction-based between-domain
processing because it encouraged people to compare the new domain to one with
which they were familiar. In contrast, the no-model condition, which made no de-
mands for people to compare this novel domain with any other, was designed to
elicit episode-based between-domain processing. In addition, practice was de-
signed to encourage either simple or complex (within-domain) encoding. The re-
sults indicated that performance was best either when (a) no-model training had
preceded simple encoding or (b) model training had preceded complex encoding.
Neither model condition nor practice complexity condition yielded main effects.
These results are consistent with the claim that between-domain and
within-domain processing interact in their effects on learning and performance.

Abstraction-Based Processing and the Effects of


Encoding Complexity

Why should such interactions occur? Consider first an individual engaging in ab-
straction-based between-domain processing. If source domains have received rela-
tively simple encoding, then using such representations as the basis for abstrac-
tion-based processes such as analogical mapping and schema induction could
easily yield a schema that would be incomplete at best, and inaccurate at worst. In
fact, simple encoding may not produce enough structural constraints to organize
the increased amount of information provided by analogies, which can then lead to
indeterminate or structure-violating mappings (see Holyoak, 1984) when schema
abstraction is attempted. This proposal about the importance of within-domain re-
lations is similar to Gentner's (1983; Gentner & Gentner, 1983) systematicity prin-
ciple. According to this principle, the within-domain relations of a source domain
are most likely to be mapped onto a target domain when those relations are part of a
richly connected system (i.e., a system that includes many interconceptual rela-
tions). Therefore, one might hypothesize that when abstraction-based be-
tween-domain processing is involved, complex source encoding will yield greater
transfer than will simple source encoding.
COMPLEXITY AND ANALOGY 47

Episode-Based Processing and the Effects of


Encoding Complexity

A different picture emerges when episode-based between-domain processing is


Downloaded by [Library Services City University London] at 00:54 18 March 2016

considered. In this case, there is no representation of common information encoun-


tered in multiple domains, problems, or events. Instead, individuals rely on their ac-
cess to particular earlier domains in dealing with new ones. In this situation, either
simple or complex encoding of the domains may provide enough information to en-
able people to retrieve earlier instances of the domain because even the relatively
superficial information that results from simple encoding should be sufficient for
useful retrieval. In fact, the relatively high number of interconceptual links in-
volved in a complex representation of a source domain may prove to be a disadvan-
tage when episode-based retrieval is involved because they might detract from epi-
sode-based retrieval by leading individuals down "garden paths" of associations.
Accordingly, one might hypothesize that when episode-based between-domain
processing is involved, there will be little or no difference in transfer performance
between simple and complex source domain encodings. In fact, performance may
be higher after simple source encoding.

DESIGN AND RATIONALE OF THIS STUDY

In this article, we tested two hypotheses, using a text comprehension and transfer
task. Participants read sets of passages (source passages followed by target pas-
sages) about analogous domains (i.e., topics). The major dependent variables were
derived from performance on multiple-choice questions covering information
from the target passage. Because the literature suggests that episode-based and ab-
straction-based processing are best for simple retrieval and inference, respectively,
we asked people to answer both multiple-choice questions about memory for the
passage and multiple-choice questions about inferences from it.
We manipulated within-domain processing of a source passage by varying the
complexity of the orienting task that participants performed while reading source
passages. Participants in the simple source encoding condition were asked to re-
spond to relatively simple questions about source passages that requested simple
repetition of information presented in the passage. Participants in the complex en-
coding condition were asked to respond to more complex questions, whose correct
answers involved knowledge of interconceptual relations within the domain. It
should be noted that because complexity of source passage encoding, rather than
the existence of a source passage, was the most critical variable in the study, it
made no sense to include a no-source condition, as do most studies of analogous
transfer.
48 CAPLAN AND SCHOOLER

To influence the nature of between-domain processing, we decided to use two


variables designed to affect the likelihood of abstraction-based or episode-based
between-domain processing: single versus multiple source domains, and amount
of analogical information presented. For both variables, manipulations that in-
creased the emphasis on between-domain relations were assumed to increase the
Downloaded by [Library Services City University London] at 00:54 18 March 2016

likelihood of abstraction-based processing; in contrast, the absence of such an em-


phasis was assumed to increase the likelihood of episode-based processing (see
Brooks, 1987; Caplan & Schooler, 1990).
The first between-domain processing variable was number of source domains.
Based on Gick and Holyoak's (1983) claim that exposure to multiple sources in-
creases the likelihood of developing a schematic representation common to the an-
alogical domains, we expected that the two-source-passage condition would
encourage abstraction-based processing and that the one-source-passage condition
would encourage episode-based processing. Gick and Holyoak examined the ef-
fects of multiple sources on transfer in a problem-solving task; the experiment we
present here could be considered, in part, an attempt at extending their findings to a
text comprehension task. Our manipulation differed from Gick and Holyoak's,
however, in another noticeable respect. In their experiments using two sources,
Gick and Holyoak required their participants to describe the similarities between
the two sources. In our own pilot work, we found that performance in the
two-source-passage condition was not affected by whether participants described
the similarity between the two source passages. Therefore, to avoid confounding
the number of source passages presented with the presentation of a comparison
task, we did not ask participants in the two-source-passage condition in this exper-
iment to describe intersource similarities.
The second between-domain processing variable was whether target passage ti-
tles provided information about the analogy between the source and target domains.
Previous work (Caplan & Schooler, 1989) demonstrated that encoding complexity
interacted with whether analogical information was provided: Analogical models
improved learning when combined with complex practice but decreased it when
combined with simple practice. These findings provide support for the hypothesis
that between-domain processing is most effective when combined with complex
within-domain processing (and vice versa). In addition, in a study more closely re-
lated to our study, Caplan and Schooler (1991) found that providing people with ti-
tles indicating that two passages describedanalogous topics increased thelikelihood
of their noticing and reporting deep (i.e., schematic) similarities between the do-
mains described by the passages. In a similar vein, in their study of transfer in prob-
lem solving, Gick and Holyoak (1983) included analog and analog plus principle
conditions. In their experiment, the addition of the principle underlying the analogy
improved transfer performance in the two-source but not in the one-source condi-
tion. From these results, they concluded that exposure to multiple sources increases
the probability that people will be able to induce the structure that underlies aclass of
COMPLEXITY AND ANALOGY 49

analogically related problems and make subsequent use of it in a problem-solving


task. Accordingly, in this study, weextended this approach to text comprehension by
providing the following three conditions: (a) no explicit analogy (no-title condi-
tion); (b) a statement that an analogy existed (short-title condition); and (c) a state-
ment that an analogy existed, along with a brief description of the principles
Downloaded by [Library Services City University London] at 00:54 18 March 2016

underlying the analogy (long-title condition). We expected that both titleconditions


would encourage abstraction-based processing, whereas the no-title condition
would encourage episode-based processing.
In keeping with the hypotheses described previously, we expected that complex
encoding would provide better support for transfer than simple encoding when ab-
straction-based processing had occurred. Therefore, we made the following pre-
dictions:

Prediction 1: When two source passages have been provided, performance


should be better following complex source encoding than following simple
source encoding.

Prediction 2: When a target passage title (either short or long) with informa-
tion about the analogy has been provided, performance should be better fol-
lowing complex source encoding than following simple source encoding.

However, when episode-based processing has been encouraged, we expected


relatively little difference between the simple and complex source-encoding con-
ditions because either type of encoding could support retrieval of particular do-
mains. In fact, if complex encoding resulted in the representation of a large number
of within-domain concepts and conceptual links, less efficient retrieval might re-
sult. This led to the following predictions:

Prediction 3: When one source passage has been provided, there should be no
difference between simple and complex source-encoding conditions; if a dif-
ference is found, it should favor the simple condition.

Prediction 4: When no target passage title has been provided to convey infor-
mation about the analogy, there should be no difference between simple and
complex source encoding conditions; if a difference is found, it should favor
the simple condition.

METHOD

Design

The basic design and procedure of this study are presented in Table 1. There were
three between-subject variables, resulting in a 3 x 2 x 2 design. These variables
50 CAPLAN AND SCHOOLER

TABLE 1
Summary of Design and Procedure Followed for Each Pair of Experimental Passages

Task No. Task ManipulationNariable

1 Participant reads source passage(s) Participant reads either one or two


Downloaded by [Library Services City University London] at 00:54 18 March 2016

source passages (independent variable:


between-subjects)
2 Participant answers simple or complex Questions are simple or complex
short-answerquestions about source (independent variable:
passage after reading each source passage between-subjects)
(passage remains available)
3 Participant reads target passage Target passage has either no title, a
short title, or a long title (independent
variable: between-subjects)
4 Participant answers multiple choice Half of questions test memory, half test
questions about target passage inference (two dependent variables)

were: (a) title condition (no title, short title, long title), (b) number of source pas-
sages (one or two), and (c) encoding complexity (simple or complex). Participants
were randomly assigned to one of the 12 resulting cells of the design.

Participants

Ninety-six participants, all of whom were paid for their participation, were re-
cruited from the community. Sixty-four of the volunteers were women, and all par-
ticipants had at least a high-school level education. The mean age was 33 years
(range = 19-77 years, SD = 13.76 years).

Stimuli

The passages for each topic were subsets of stimuli used in previous work (Caplan
& Schooler, 1989). They consisted of two sets of stimulus passages. (See Appendix
A for a sample set of passages, and a mapping of the underlying analogy; note that
participants never saw the mapping.) Each passage was approximately 1 to 1% dou-
ble-spaced pages long, and was similar in style to passages found in undergraduate
textbooks. In each set, the source passage(s) and the target passage described struc-
turally analogous domains. In this article, we will refer to the sets as the optics and
competition sets, although participants never encountered these general titles. Each
participant received both optics and competition passage sets; the order in which
the two sets were presented was counterbalanced across participants.
The content areas for the two sets of stimulus passages are presented in Table 2.
It should be noted that the two source domains for each set differed in their superfi-
COMPLEXITY AND ANALOGY 51

cia1 similarity to the target domain, although the domain described by both source
passages shared the same structural characteristics with the target domain.
In the one-source-passage condition, half of the participants received one of the
possible source passages for a given target passage, and the other half received the
other source passage. The source passages used were counterbalanced across par-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

ticipants. In the two-source-passage condition, the order of the superficial similar-


ity of the two source passages used was counterbalanced in two ways. First, the
order in which participants encountered the two superficial similarity conditions
TABLE 2
Domains and Titles Used

Title Length Condition One Two

Competition set
Short The black-footed ferret is like the The black-footed ferret is like the
obstetriciadthree-toed sloth three-toed sloth and the obstetrician
Long The black-footed ferret is like the The black-footed ferret is like the
obstetriciadthree-toed sloth: Both three-toed sloth and the obstetrician:
survive by being successful at two All survive by being successful at
kinds of competition: competition two kinds of competition:
with members of their own groups competition with members of their
and competition with members of own groups and competition with
other groups. How much competition members of other groups. How much
occurs depends on the extent to competition occurs depends on the
which the two groups require the extent to which the two groups
same resources. require the same resources.
Optics set
Short The rangefinder camera is like the The rangefinder camera is like the eye
eyelsingle lens reflex camera and the single lens reflex camera
Long The rangefinder camera is like the The rangefinder camera is like the eye
eyelsingle lens reflex camera: Both and the single lens reflex camera: All
work by passing light through an work by passing light through an
opening to a light-sensitive area at opening to a light-sensitive area at
the back of the system. The amount the back of the system. The amount
of light that passes through is of light that passes through is
regulated by a special mechanism, regulated by a special mechanism,
and the image is focused by a lens. and the image is focused by a lens.
How much detail is recorded How much detail is recorded depends
depends on how individually the on how individually the units of
units of light-sensitive material light-sensitive material function.
function.

Nore. In the one-source-passage condition, participants received one of the two domains separated by slashes
above. In the long title condition, participants received the same title as that used in the short title condition. followed
by a description of the principles common to the source and target passages.
was counterbalanced between the first and second passage sets. For example, if a
participant's first passage set had as its first source passage a domain superficially
similar to the target domain, then the second set had as its first source passage a do-
main superficially dissimilar to the target domain. Second, whether the first pas-
sage set's first domain was superficially similar or dissimilar to the target domain
Downloaded by [Library Services City University London] at 00:54 18 March 2016

was counterbalanced across participants.


For each source passage, participants responded to a set of six short-answer
questions about the material presented in the first passage. Questions in the simple
condition asked participants to reiterate information that was stated explicitly in
the passage; questions in the complex condition asked participants to answer ques-
tions about the content of the passage for which the answers were not explicitly
stated in the passage. For each source domain in a passage set, we created two sets
of encoding questions, Set A and Set B (both sets are included in Appendix B). The
questions in Set A were directly analogous for the two source passages, as were the
questions for Set B. Each participant in the two-source-passage condition received
Set A questions for one source passage, and Set B for the other. Each participant in
the one-source passage condition received Set A questions for one topic set (either
Competition or Optics), and Set B for the other. The combinations of question sets
and domains were counterbalanced across participants.
In addition, one third of the participants received target passages that had titles
stating an explicit analogy between the content areas of the passages in a set
(short-title condition) One third of the participants received target passages with the
same titles, followed by brief explanations of the structural similarities between the
source and target domains (long-title condition). One third of the participants re-
ceived target passages without such titles (no-titlecondition). The titles used arepre-
sented in Table 2. In the two-source-passage condition, the title made reference to
both source domains and the target domain. In the one-source-passagecondition, the
title made reference to the one source domain provided and the target domain. It
should be noted that, with the exception of the title manipulation, the target passages
for a given content area (i.e., Optics or Competition) were all identical.
For each of the sets of passages (i.e., Optics and Competition), a set of 12 multi-
ple-choice questions (target questions) was constructed for the target passage; these
questions are provided in Appendix C. To avoid as much as possible people's re-
sponding to the fact retrieval questions by using inference strategies, the six memory
questions were presented first. In these questions, participants were asked about in-
formation that had been presented in the passage. In the last six questions (inference
questions),participants were asked to make inferencesabout the material presented.

Procedure

At the start of the experimental session, participants were asked a few preliminary
questions about their background. Then, they were administered the Vocabulary
COMPLEXITY AND ANALOGY 53

subtest of the Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler,


1981).
A summary of the procedure is presented in Table 1. Participants were in-
structed that they would be reading a series of passages, and that after they had
read each passage, they would be given some written questions to answer about it.
Downloaded by [Library Services City University London] at 00:54 18 March 2016

They were told that the questions would be either short-answer or multiple-choice
format, but that at the time they were given the passage to read, they would not
know what kind of questions would follow it. They were informed that if the ques-
tions were short-answer, they would be able to keep the passage and refer to it as
many times as they needed to answer the questions. If, however, the questions
were multiple-choice, then the passage would be removed before the questions
were given to them. Participants were not informed that there would be any rela-
tion among any of the passages they were to read.
Participants began each of the two experimental passage sets by reading the
first passage at their own pace. When participants indicated that they had com-
pleted reading the passage, they were given the encoding questions for that pas-
sage, and answered the questions in writing, while the passage was still available
to them. After participants answered the encoding questions, the experimenter col-
lected the questions, answers, and first passage. Participants in the
two-source-passage condition repeated this procedure for the second source pas-
sage. Then participants were given the target passage. After they indicated that
they had finished reading the target passage, the passage was removed, and they
were asked to answer the target questions. Participants were tested individually.
They read the passages and answered the questions at their own pace; the task took
most participants approximately 1 hr.

RESULTS

Preliminary analyses indicated that there were no effects or interactions involving


occasion, that is, effects were similar for the first and the second set of passages a
participant encountered. In addition, there were no significant differences in results
between the superficially similar and superficially dissimilar source passages (in
the one-source-passage condition) or between the two orders of superficial similar-
ity (in the two-source-passage condition). Therefore, in the analyses reported sub-
sequently, we collapsed over occasion and superficial similarity.
Separate 3 x 2 x 2 analyses of variance (ANOVAs) were performed on partici-
pants' number of correct responses on the multiple-choice memory and inference
questions. In both analyses, the three between-subject factors were: title condition
(none, short, long), encoding complexity (simple vs. complex), and number of
source passages (one vs. two). However, early analyses revealed that, despite our
random assignment procedures, WAIS-R Vocabulary scores were marginally
54 CAPLAN AND SCHOOLER

higher in the one-source-passage condition than in the two-source-passage condi-


tion @ = .06). Therefore, we also performed an analysis of covariance (ANCOVA)
on each of the two dependent measures (i.e., number of correct responses on the
memory and on the inference questions), using WAIS-R Vocabulary score as the
covariate. The patterns of results from the ANOVAs and ANCOVAs were virtu-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

ally identical; where the two differed, we note the difference subsequently. We re-
port all effects significant at the p < .05 level.

Memory

As reflected in Table 3, the analysis on memory performance revealed a reliable in-


teraction between complexity and title condition, F(2,84) = 4 . 5 1 , ~ = .0138. In the
ANCOVA, this effect was marginally significant, p = .06. Planned comparisons re-
vealed no significant effects of complexity in the no-title or in the long-title condi-
tion; however, in the short-title condition, performance was better following com-
plex rather than simple source passage encoding, F(l,84) = 6.21, p = .Ol. No other
effects, including main effects, were significant ( p > .10 throughout).

Inference

As reflected in Table 4, the analysis on inference performance revealed an interac-


tion between encoding complexity and number of source passages, F(l,84) = 5.99,
p = .02. Planned comparisons revealed no significant effects of encoding complex-
ity in the one-source-passage condition; however, in the two-source-passage con-
dition, performance was better following complex than following simple encoding,
F(l, 84) = 4.80, p = .03.
Investigation of the other effects revealed that performance was marginally
worse in the two-source-passage condition than in the one-source pas-
sage-condition, F(1, 84) = 3.52, p = .06; however, in the parallel ANCOVA, this

TABLE 3
Mean Number Correct and Standard Deviations on Memory Questions as a Function of
Analogical Title Condition (None, Short, Long) and Source Passage Encoding Complexity

Analogical Title Condition

None Short Long


Encoding
Complexity M SD M SD M SD
-

Simple 11.69 0.79 10.94 1.57 1 1.62 0.62


Complex 11.38 1.02 11.81 0.54 11.12 1.09

Note. Means collapsed over conditions involving number of source passages.


COMPLEXITY AND ANALOGY 55

TABLE 4
Mean Number Correct and Standard Deviations on Inference
Questions as a Function of Number of Source Passages and
Source Passage Encoding Complexity

No. of Source Passages


Downloaded by [Library Services City University London] at 00:54 18 March 2016

One Two
Encoding
Complexity M SD M SD

Simple 9.96 1.97 7.75 2.69


Complex 9.04 2.88 9.33 2.48

Note. Means collapsed over title conditions.

effect was far from reliable (F< l), suggesting that the difference was due to inad-
vertent sample differences rather than to experimental treatment. No other effects,
including main effects, were significant @ > .20 for all effects).

Summary of Results

For the memory dependent measure, we found that the title condition interacted
with encoding complexity: There was an advantage for complex encoding in the
short-title condition, but not in the no-title condition. Examination of the pattern of
means for these two title conditions reveals that it generally conformed to predic-
tions: Performance was best either when complex encoding had preceded a
short-title target passage or when simple encoding had preceded a no-title target
passage.
For the inference dependent measure, results also conformed to predictions:
There was an advantage for complex encoding in the two-source, but not the
one-source condition. Examination of the means reveals that performance was
best either when participants engaged in complex encoding of two source passages
or in simple encoding of one.

DISCUSSION

In this article, we tested the hypothesis that a within-domain processing variable,


encoding complexity, would have different effects on analogical transfer, depend-
ing on the nature of the between-domain processing (i.e., episode- or abstrac-
tion-based) that had been devoted to the domains. In particuiar, we predicted that
the complexity with which a source domain had been encoded would affect how
people would make use of analogical information in text comprehension. We hy-
56 CAPLAN AND SCHOOLER

pothesized that complex encoding of earlier source passages would support the ab-
straction-based retrieval and transfer processes that result from exposure to multi-
ple training instances or to statements that directly compare domains. In contrast,
we expected that simple source-passage encoding would not support abstrac-
tion-based processing and would be equal to (or better than) complex
Downloaded by [Library Services City University London] at 00:54 18 March 2016

source-passage encoding when episode-based processing had been encouraged. In


general, these predictions were supported: Variables designed to encourage or dis-
courage abstraction-based processing interacted with source domain encoding
complexity.
In addition, as expected, we found that memory and inference performance
measures were affected by different independent variables. However, in contrast
to the predictions based on the previous literature, we did not find that abstrac-
tion-based processing was always superior for inference and episode-based pro-
cessing for memory. Instead, each of the two variables designed to influence the
likelihood of abstraction-based or episode-based processing interacted with source
domain encoding complexity for one outcome measure: Number of source pas-
sages influenced inference, whereas title condition influenced memory. In a previ-
ous work (Caplan & Schooler, 1989), we obtained similar findings. We found that
an earlier title manipulation had an effect on memory performance similar to that
obtained here, and that similarity of passage structure (where dissimilar structures
were assumed to increase the likelihood of creating a schema or situation model)
interacted with complexity in its effect on inference. Why should memory and in-
ference respond to different variables? The following consideration of the differ-
ence between inference and memory suggests that the first should be affected by
the number of source passages and the second by the title conditions.
In this study, for the inference measure, we found that when people had re-
ceived two source passages, performance was better following complex rather
than following simple encoding of the source passage; however, there was no sig-
nificant effect of encoding complexity when people had received only one source
passage. We suggest, following Kintsch (1986; Perrig & Kintsch, 1985; van Dijk
& Kintsch, 1983), that inferences are based on situation models,that is, on a sche-
matic representation of a domain that is divorced from the representation of the
text itself, the "textbase." Presumably, the comparison processes that operate on
multiple analogous domains allow people to abstract a situation model that is, in
essence, the schema underlying the analogy between the two domains. In this
study, simple encoding of the source domain would have led to scanty or rnislead-
ing situation models, leading to poorer inference performance. In contrast, com-
plex encoding would have led people to correctly represent enough of the concepts
and interconceptual relations of a source domain to abstract a relatively complete
and accurate situation model, which in turn would support inference.
In the case of the memory measure, titles, rather than number of source pas-
sages, comprised the critical variable. This makes sense because titles can serve as
COMPLEXITY AND ANALOGY 57

retrieval or access cues, reminding people of the relevance of source passages. Re-
sults from the no-title and the short-title conditions were as predicted. In the
short-title condition, performance was better following complex encoding than
when following simple encoding. In contrast, there was no difference between the
two encoding conditions in the no-title condition. This pattern provides support for
Downloaded by [Library Services City University London] at 00:54 18 March 2016

our original hypothesis that complex source domain encoding would enhance ab-
straction-based processing and transfer, but that either simple or complex source
encoding could support episode-based processing and transfer. It is also consistent
with previous findings in computer software learning (Caplan & Schooler, 1990)
and text comprehension (Caplan & Schooler, 1989), where we found that complex
encoding improved learning only when analogical models had been provided.
However, performance in the long-title condition, where performance in the
two complexity conditions was not significantly different (and where the direction
of the means was the opposite of prediction), remains somewhat of a mystery. We
note that Gick and Holyoak (1983; Experiment 5) obtained similar results in their
problem-solving transfer task. Although participants in that experiment benefitted
from the presentation of the underlying principle of the analogy before receiving a
hint regarding the analogical relation between the sources and the target, the prin-
ciple provided no additional benefit once a hint had been provided (i.e., in the con-
dition most like our long-title condition). In fact, if anything, participants in that
study were somewhat more likely to produce post-hint problem solutions in the
no-principle condition than in the principle condition. One possible explanation
for our findings may lie in the nature of the description we provided of the overall
principles (i.e., the schema) that governed both the target and the source domains.
This description, in essence, included relations among concepts that held within
each domain, rather than information that would have encouraged schema abstrac-
tion across multiple domains, as originally intended. Another possibility is that the
long-title condition provided information that conflicted with the abstractions or
schemas that people formed spontaneously; such a conflict would have left people
with no consistent schema on which to rely. Finally, we note that the means for the
long-title condition were uniformly high, suggesting the possibility of a ceiling ef-
fect, which could have obscured any effects of encoding complexity. At the mo-
ment, we cannot decide with any certainty which of these possible explanations is
correct; the answers await future empirical work.
Nevertheless, the overall pattern of our results was as predicted. To what extent
might this pattern be due to mechanisms other than those proposed here? First, one
might hypothesize that any constructive activity would increase learning, perhaps
particularly for inference measures. Certainly, such "depth of processing" ac-
counts can go far in explaining similar results from the laboratories of other inves-
tigators. For example, Wiley and Voss (1996) examined the effects of various
tasks and presentation formats on people's understanding and recall of a history
text passage. They suggested that knowledge transformation, or the result of con-
58 CAPLAN AND SCHOOLER

structive activity, improves understanding. However, the results of our study sug-
gest that more than simple constructive activity was involved. If constructive
activity had been the only relevant variable, then we might have expected to see
only main (i.e., additive) effects of encoding complexity, number of source pas-
sages, and title conditions. Instead, we obtained a different pattern: interactions be-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

tween encoding complexity and our between-domain variables. Furthermore, we


obtained interactions only between our between-domain and our within-domain
variables; we never obtained interactions between the two between-domain vari-
ables. A similar argument could be used against the claim that our results are due to
the fact that our simple encoding questions were similar to the memory questions,
and that the complex encoding questions were more similar to the inference ques-
tions. According to this argument, the participants in this experiment were just
demonstrating superior performance on the type of question to which they had be-
come accustomed. By choosing to use different formats for the encoding and target
questions (open-ended vs. multiple choice), we had hoped to minimize such diffi-
culties. If such a materials similarity effect had remained despite these efforts, it
would have resulted in main effects of encoding complexity for both memory and
inference measures; however, such main effects were not obtained (F< 1, for both
memory and inference analyses).
The distinction between within-domain and between-domain processing is not a
new one. Hesse (1966) discussed the differences between vertical relations (i.e.,
among elements of the representation of adomain) and horizontal relations (i.e., be-
tween the representations of two or more domains). Similarly, Mayer (e.g., Mayer,
1984; Mayer & Greeno, 1972)has long maintained that one's representation of ado-
main can vary on two dimensions. Internal connectedness refers to the degree to
which the components of a domain's representation are connected. External
connectedness refers to the degree to which those components are connected to as-
pects of knowledge outside thedomain. In a series of experiments, Mayer found that
performance on tasks requiring complex comprehension or complex problem solv-
ing was generally facilitated by external connectedness, and performance on tasks
requiring fairly simple reiteration of materials or simple solutions to problems was
facilitated by internal connectedness (e.g., Mayer, 1983; Mayer, 1984; Mayer &
Greeno, 1972). As discussed in the introduction, Cummins (1992) also argued that
between-domain processing yields better transfer than within-domain processing.
However, our theoretical approach as well as our complete pattern of results extends
this body of work by proposing that neither of these two types of processing (i.e.,
within-domainlinternally connected and between-domainlexternally connected) is
necessarily beneficial to a particular type of performance; rather, we suggest that
they interact in their effects on performance.
The use of analogy is a common pedagogical device, which is usually assumed
to benefit learners by providing a link between a novel, unfamiliar domain and an
older, more familiar one. In this article, we have suggested that analogy alone may
COMPLEXITY AND ANALOGY 59

not be sufficient to enhance learning-as reflected in either memory for material


or inferences a learner may derive from it. Instead, its effectiveness depends on at
least one other characteristic of learning history, that is, the complexity with which
the source domain has been encoded. More broadly, we have suggested that (a) en-
coding complexity encourages the development of within-domain relations in a
Downloaded by [Library Services City University London] at 00:54 18 March 2016

domain; (b) other variables, like number of training instances, may affect the de-
velopment of between-domain relations; and (c) the beneficial effects of be-
tween-domain processing are most likely to be evident when within-domain
encoding has been complex.

ACKNOWLEDGMENT

Portions of these data were presented at the 1990 meeting of the Psychonomic Soci-
ety, New Orleans, LA.

REFERENCES

Barr, R. A,, & Caplan, L. J. (1987). Category representations and their implications for category struc-
ture. Memory and Cognition, 15, 397418.
Brooks, L. R. (1978). Non-analytic concept formation and memory for instances. In E. Rosch & B.
Lloyd (Eds.), Cognition and categorization (pp. 169-21 1). Hillsdale, NJ: Lawrence Erlbaum Asso-
ciates, Inc.
Brooks, L. R. (1987). Decentralized control of categorization: The role of prior episodes. In U. Neisser
(Ed.), Concepts and conceptual development (pp. 141-174). New York: Cambridge University
Press.
Brown, A. L. (1989). Analogical learning and transfer: What develops? In S. Vosniadou & A. Ortony
(Eds.), Similarityandanalogical rearoning (pp. 369412). New York: Cambridge University Press.
Brown, A. L., & Kane, M. J. (1988). Preschool childrencan learn to transfer: Learning to learn and learn-
ing by example. Cognitive Psychology. 20, 493-523.
Brown, A. L., Kane, M. J., & Echols. C. H. (1986). Youngchildren's mental models determine analogi-
cal transfer across problems with a common goal structure. Cognitive Development, 1. 103-121.
Camp, C. J., Lachman J. L., & Lachman, R. (1980). Evidence for direct-access and inferential retrieval
in question-answering. Journal of Verbal Learning and Verbal Behavior, 19, 583-596.
Caplan, L. J., & Schooler, C. (1989, November). Analogy and complexity of processing in text compre-
hension. Paperpresentedat the Twentieth Annual Meeting of the Psychonomic Society, Atlanta, GA.
Caplan, L. J., & Schooler, C. (1990). Problem-solving by reference to rules or previousepisodes: The ef-
fects of organized training, analogical models, and subsequent complexity of experience. Memory
and Cognition. 18, 21 5-227.
Caplan, L. J., & Schooler, C. (1991, November). Analogies, similarities, and abstraction. Poster pre-
sented at the Twenty-Second Annual Meeting of the Psychonomic Society, San Francisco.
Catrambone, R., & Holyoak, K. J. (1989). Overcoming contextual limitations on problem-solving trans-
fer. Journal OfExperimental Psychology: Learning, Memory, & Cognition, 15. 1147-1 156.
Cummins, D. D. (1992). Role of analogical reasoning in the induction of problem categories. Journal of'
Experimental Psychology: Learning, Memory, & Cognitron, 18, 1103-1 124.
Donnelly. C. M., & McDaniel, M. A. (1993). Use of analogy in learning scientific concepts. Journal of'
Experimental Psychokogy: Learning, Memory, & Cognition. 19, 975-987.
60 CAPLAN AND SCHOOLER

Fong. G. T., & Nisbett, R. E. (1991). Immediate and delayed transfer of training effects in statistical rea-
soning. Journal of Experimental Psychology: General, 120, 34-45.
Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7,
155-170.
Gentner, D. (1989). The mechanisms of analogical learning. In S. Vosniadou & A. Ortony (Eds.). Simi-
larity and analogical reasoning (pp. 199-241). New York: Cambridge University Press.
Downloaded by [Library Services City University London] at 00:54 18 March 2016

Gentner, D., & Gentner, D. R. (1983). Flowing waters or teeming crowds: Mental models of electricity.
In D. Gentner & A. L. Stevens (Eds.), Mental models (pp. 99-129). Hillsdale, NJ: Lawrence
Erlbaum Associates, Inc.
Gentner, D.. Rattermann, M. J., & Forbus, K. D. (1993). The roles of similarity in transfer: Separating
retrievability from inferential soundness. Cognitive Psychology, 25, 524-575.
Gick, M. L., & Holyoak, K. J. (1983). Schemainduction and analogical transfer. Cognitive Psychology,
15, 1-38.
Hesse, M. B. (1966). Modelsandanalogies in science. South Bend, IN: University of Notre Dame Press.
Hintzman, D. L. (1986). "Schema abstraction" in a multiple-trace memory model. Psychological Re-
view, 93. 41 1-428.
Holyoak, K. J. (1984). Analogical thinking and human intelligence. In R. Stemberg (Ed.),Advancesin
thepsychology ($human intelligence (Vol. 2, pp. 199-230). Hillsdale, NJ: LawrenceErlbaum Asso-
ciates, Inc.
Holyoak, K. J., & Koh, K. (1987). Surface and structural similarity in analogical transfer. Memory &
Cognition, 15, 332-340.
Kintsch, W . (1986). Learning from text. Cognition and Instruction, 3, 87-108.
Mannes, S. M., & Kintsch, W. (1987). Knowledge organization and text organization. Cognition andln-
struction, 4, 91-1 15.
Mayer, R. E. (1975). Different problem-solving competencies established in learning computer pro-
gramming with and without meaningful models. JournalcifEducadonal Psychology, 67,725-734.
Mayer, R. E. (1976). Some conditions of meaningful learning for computer programming: Advance or-
ganizers and subject control of frame order. Journal ($Educational Psychology, 68, 143-150.
Mayer, R. E. (1983). Can you repeat that? Qualitative effects of repetition and advance organizers on
learning from science prose. Journal of Educational Psychology, 75, W 9 .
Mayer, R. E. (1984). Aids to text comprehension. Educational Psychology, 19, 3C-42.
Mayer, R. E., & Greeno, J. G. (1972). Structural differences between learning outcomes produced by
different instructional methods. Journal of Educational Psychology, 63, 165-173.
Medin, D. L., &Ross, B. H. (1989). The specific character of abstract thought: Categorization, problem
solving, and induction. In R. J. Sternberg (Ed.).Advances in the psychology of human intelligence
(Vol. 5, pp. 189-223). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Medin, D. L., & Schaffer, M. M. (1978). Context theory of classification learning. Psychological Re-
view, 85. 207-238.
Mitchell, T. M., Keller, R. M., & Kedar-Cabelli, S. T. (1986). Explanation-based generalization: A uni-
fying view. Machine Learning, I, 47-80.
Novick, L. R. (1988). Analogical transfer, problem similarity, and expertise. Journal of Experimental
Psychology: Learning, Memory, & Cognition, 14, 510-520.
Pemg, W., & Kintsch, W. (1985). Propositional and situational representations of a text. Journal of
Memory & Language, 24, 503-5 18.
Reed, S. K. (1989). Constraints on the abstraction of solutions. Journal of Educational Psychology, 4,
532-540.
Rosch. E. (1973). On the internal structure of perceptual and semantic categories. In T. E. Moore (Ed.).
Cognitive development and the acquisition cif language (pp. 111-144). New York: Academic.
Ross, B. H. (1984). Remindings and their effects in learning acognitive skill. Cognitive Psychology, 16,
COMPLEXITY AND ANALOGY 61

Ross, B. H. (1989). Distinguishingtypes of superficial similarities:Different effects on the access and


use of earlier problems. Journal of Experimental Psychology: Learning, Memory, & Cognition, 15,
456-468.
Ross, B. H., & Kennedy, P. T. (1990). Generalizing from the use of earlierexamplesin problem solving.
Journal of Experimental Psychology: Learning, Memory, & Cognition. 16,42-55.
van Dijk, T. A., & Kintsch, W. (1983). Strategies ofdiscourse comprehemion. New York:Academic.
Downloaded by [Library Services City University London] at 00:54 18 March 2016

Wechsler. D. (1 98 1). WAIS-R manual. Wechsler Adult Intelligence Scale-Revised. New York: Psycho-
logical Corporation.
Wiley, J., & Voss, J . F. (1996). The effects of "playing historian" on learningin history.Applied Cogni-
tive Psychology, 10, S63-S72.

APPENDIX A
Sample Source Passage, Sample Target Passage (No-Title
Version), and Sample Mappings From Set 2

(Note that mappings were not provided to participants in the experiment.)

Source Passage

The eye is the basic organ for vision. It works in the following way. The part of the
eye that ultimately receives the image of objects in the environment is called the ret-
ina. The retina is located at the back of the eye. It has two kinds of receptors. The
chemicals in the receptors undergo areaction when light hits them. The two types of
receptors are used under different lighting conditions. Cones are the receptors that
work best in bright light. Because cones work relatively individually, and because
they are packed closely together on the retina, they record most of the detail in the
image that is sent to the retina. Rods are the other kind of receptor; they work best
under relatively dim lighting conditions. Because rods work in clumps, which are
more spread out on the retina, they do not record as much of the detail of the image.
How does the image of the outside world get to the retina? First, light is re-
flected off objects in the world, and enters the eye through the pupil. The pupil is
surrounded by the iris, a muscle that controls the size of the pupil. When the iris
makes the pupil small, relatively little light enters the eye. When it makes the pupil
large, more light enters. Light that has entered through the pupil next goes through
the lens. The lens focuses the image on the retina. Because of the way the lens
works, it turns the image upside-down before it hits the retina. When a lens is, in
effect, thick, it focuses objects that are relatively close to the person. When it is, in
effect, thin, it focuses objects that are farther away.
To focus the eye, the muscles surrounding a lens respond to the conditions on
the retina. The image, which has passed through the lens, hits the retina. If the im-
age is focused properly, it is sharp on the retina. If it is not, it is blurred on the ret-
ina, and the muscles surrounding the lens adjust until the image is sharp.
62 CAPLAN AND SCHOOLER

Target Passage

The range-finder camera is a camera that one focuses by turning the lens until the
two images seen in the viewfinder merge into one. It works in the following way.
Light, which is reflected off objects in the environment, enters the rangefinder carn-
Downloaded by [Library Services City University London] at 00:54 18 March 2016

era through the aperture. The size of the aperture can change, depending on the ac-
tion of the diaphragm, which is the mechanism that surrounds the aperture. The dia-
phragm can act to make the aperture smaller (so that it lets in less light), or to make
the aperture larger (so that it lets in more light). After passing through the aperture,
the light passes through the lens of the rangefinder camera. Once the light has
passed through the lens, it hits the film, located at the back of the interior of the
rangefinder camera. The function of the lens is to focus the image of the object on
the film. When the lens is thick, it focuses objects that are close to the rangefinder
camera. When it is thin, it focuses objects that are farther away.
As the image's light passes through the lens, the image is turned upside-down,
so that is the way it hits the film. When light hits the film, it starts a chemical reac-
tion in the silver bromide crystals of the film. Depending on how much light there
is, different types of film will be used. Under low light conditions, fast film work
best. Because the silver crystals of this type of film work in clumps, spread widely
across the surface of the film, a lot of the detail in the image will be lost. Under
high light levels, slow film works best. Because the crystals of slow film work
more individually, and are spaced closely together on the film's surface, more of
the details of the image get recorded.
To focus the rangefinder camera, the photographer looks through the view-
finder. Two images are directed by a mirror and a prism into the viewfinder. If
the image is focused properly, one image is seen by the photographer. If it is not,
two images are seen, and the photographer adjusts the lens until one image is
seen.

Mappings

EYE RANGEFINDER CAMERA


Lens - Lens
Pupil - Aperture
Retina - Film
Iris - Diaphragm
Rods - Fast film
Cones -
- Slow film
Pupil changes size in - Aperture changes size in response
response to changes to changes in light level
in light level
The iris changes the - The diaphragm changes aperture size
COMPLEXITY AND ANALOGY 63

pupil size
The lens focuses the -
- The lens focuses the image on the film
image on the retina
The lens turns the -
- The lens turns the image upside-down
image upside-down
Downloaded by [Library Services City University London] at 00:54 18 March 2016

When the lens is thick = When the lens is thick it focuses things
it focuses things that that are close
are close

APPENDIX B
Encoding Questions As a Function of Encoding Condition,
Passage Set, Question Set, and Similarity Between Source
and Target Domain

Similarity Between Source and Target Domain


Question
Passage Set Set Superficially Dissimilar Superficially Similar

Simple encoding condition


Competition A 1. What is the typical work 1. What is the habitat of the three-toed
environment of the obstetrician? sloth?
2. What is an occupational role? 2. What is an ecological niche?
3. For the obstetrician, is 3. For the sloth, is inter-specific
between-occupation competition very competition very strong?
strong?
4. What does "between- occupation" 4. What does "inter-specific"
competition mean? competition mean?
5. What is the obstetrician's preferred 5. What is the three-toed sloth's
patient? preferred food?
6. What makes a member of an a 6. What makes an animal a
"generalist? "generalist?"
Competition B 1. What is an occupational system? I . What is an ecosystem?
2. 1s the occupational role of an 2. Is the ecological niche of the
obstetrician narrow or broad? three-toed sloth narrow or broad?
3. How much between-occupation 3. How much inter-specific
competition occurs when two competition occurs when two
occupational roles do not overlap? ecological niches do not overlap?
4. What does "within-occupational 4. What does "intra-specific
competition" mean? competition" mean?
5. What is a work environment? 5. What is a habitat?
6. What makes a member of an 6. What makes an animal a
occupation a "specialist?" "specialist"?
Optics A 1. How is the lens controlled? 1. How is the lens controlled?
2. What is the pupil? 2. What is the aperture?
3. What is the retina? 3. What is film?
64 CAPLAN AND SCHOOLER

4. What shape of lens is best for seeing 4. What shape of lens is best for
objects that are close? photographing objects that are close?
5. What type of retinal receptor works 5. What type of film works best under
best under bright lighting conditions? bright lighting conditions?
6. What is the iris? - -
6. What is the diaphragm?
Optics B 1. What type of receptor works in 1. What type of film has crystals that
Downloaded by [Library Services City University London] at 00:54 18 March 2016

groups or clumps? works in groups or clumps?


2. What does the lens do? 2. What does the lens do?
3. What part of the eye does the image 3. What part of the camera does the
end up on? image end up on?
4. How do retinal receptors work? 4. How does film work?
5. What type of retinal receptor works 5. What type of film works best under
best under low lighting conditions? low lighting conditions?
6. What shape of lens is best for seeing 6. What shape of lens is best for
far away? photographing objects far away?
Complex encoding c:ondition
Competition A 1. Imagine that a new occupation that 1. Imagine that another animal that also
involved providing medical treatment ate Cecropia leaves were to be moved
for pregnant women were developed. to the forests where the three-toed sloth
How would the obstetrician's role relate lives. How would the sloth's niche
to that of the new occupation? What relate to that of the other animal? What
would happen to the obstetrician's would happen to the sloth's
between-occupation and inter-specific and intra-specific
within-occupation competition? Why? competition? Why?
2. Which would be more likely to 2. Which would be more likely to
survive sudden societal changes, an survive a catastrophic event, an animal
occupation with a narrow occupational with a narrow ecological niche, or one
role, or one with a broad role? Why? with a broad niche. Why?
3. Imagine that there was a sudden 3. Imagine that there was a sudden
decrease in the number of obstetricians. decrease in the number of three-toed
What would happen to the sloths. What would happen to the
obstetrician's within-occupation sloth's intra-specific competition?
competition? Why? Why?
4. Imagine that improvements in birth 4. Imagine that many Cecropia trees
control dramatically decreased the were killed by the use of a defoliant.
number of pregnant women. What What would happen to the sloth's
would happen to the obstetrician's intra-specific competition? What would
within-occupation competition? What happen to the number of sloths?
would happen to the number of
obstetricians?
5. Imagine that many obstetricians 5. Imagine that many sloths were killed
stopped practicing as a result of as a result of sloth-hunting by man. and
increased cost of malpractice insurance, that subsequently, sloth-hunting was
and that subsequently. caps were put on outlawed. What would happen to the
insurance costs. What would happen to intra-specific competition of those
the within-occupation competition of remaining sloths? Why?
those remaining obstetricians (assume
that those who left obstetrics did not
return to it)? Why?
COMPLEXITY AND ANALOGY 65

6. Imagine that obstetricians developed 6. Imagine that sloths mutated, so that


new specialties, so that some some sloths ate only the leaves of
obstetricians treated only patients under young Cecropia trees, and the rest ate
30 years of age, and the rest treated the leaves of old ones. What would the
Downloaded by [Library Services City University London] at 00:54 18 March 2016

patients 30 years or older. What would result be for the breadth of the sloth's
the result be for the breadth of the ecological niche, and to levels of
obstetrician's occupational role, and to intra-specific and inter-specific
the levels of within-occupation and competition?
between-occupation competition?
Competition B I . lmagine that there was a sudden I. Imagine that there was a sudden
shortage of medical supplies for use in shortage of Cecropia leaves. What
obstetrical practice. What would be the would be the results in terms of
results in terms of within-occupation intra-specific competition for sloths?
competition for obstetricians?
2. lmagine that many obstetricians got 2. lmagine that many sloths died of a
tired of delivering babies in the middle disease. For those sloths who remained,
of the night. and left the field. For those what would happen to the degree of
obstetricians who remained. what inter-specific competition? Why?
would happen to the degree of
between-occupation competition?
Why?
3. lmagine that there was a sudden 3. lmagine that there was a sudden
increase in the number of obstetricians. increase in the number of sloths. What
What would happen to the would happen to the sloth's
obstetrician's within-occupation intra-specific competition? Why? What
competition? Why? What would be the would be the implications for the
implications for the number of number of sloths who lived in a given
obstetricians who practiced in a given area?
area?
4. Imagine that after a major world 4. lmagine that one of the types of tree
war, there was a significant decrease in in a South American forest was
the world population. In response to basically eliminated because humans
this decrease, the number of pregnant valued it for lumber, and that the
women increased. What would happen Cecropia was therefore able to increase
to the obstetrician's within-occupation in number. What would happen to the
competition? What would happen to the three-toed sloth's intra-specific
number of obstetricians? competition? What would happen to the
number of sloths?
5. lmagine that midwives suddenly 5. lmagine that some animal from the
became more dominant in providing South American forest began to eat
obstetrical care. What would the Cecropia leaves. What would the
implications be for obstetricians' access implications be for sloths' access to
to resources, and subsequent levels of resources, and subsequent levels of
within- and between-occupational intra- and inter-specific competition?
competition?
6. Imagine that the nature of obstetrics 6. lmagine that three-toed sloths
changed, resulting in a change in the mutated, resulting in a change in their
type of patients obstetricians would eating habits. After mutation, they are
treat. After this change, they treated able to live on the leaves of several
women for many different medical trees, some of which are eaten by other
problems. some of which are treated by animals. What would happen to the
Downloaded by [Library Services City University London] at 00:54 18 March 2016

members of other medical professions. breadth of the sloth's ecological niche,


What would happen to the breadth of and to its levels of intra- and
the obstetrician's occupational role, and inter-specific competition. Why?
to hisher levels of within- and between
competition? Why?
Optics A 1. How would the eye behave 1. What would someone working with
differently for conditions in which there a single-lens reflex camera have to do
is relatively little light, and conditions differently to photograph in conditions
in which there is a lot of light? in which there is relatively little light,
and conditions in which there is a lot of
light?
2. If one's retina had only cones, how 2. If one had only slow film, how
would that individual's vision differ would that individual's photography be
from normal? affected?
3. What would happen to someone's 3. What would happen to someone's
vision if his lens could no longer photography if he had only one camera
change shape, and was relatively thin? lens, which was relatively thin?
4. How would an individual's vision be 4. How would an individual's
affected if his lens muscles did not photography be affected if his camera
work? lens could not be changed or adjusted?
5. If a lens became clouded, what 5. If a lens became clouded, what
would happen to a person's vision? would happen to a person's
photography?
6. Imagine that someone had a 6. lmagine that someone's single-lens
damaged iris. What would be the reflex camera had a damaged
effects on that person's vision? diaphragm. What would be the effects
on that person's photography?
Optics B I. lmagine someone wearing I. lmagine a camera with a dark filter
sunglasses. How must the eye adjust to over the lens. What must the
produce good vision? photographer do to produce a good
photograph?
2. If one's retina had only rods, how 2. If one had only fast film, how would
would that individual's vision be that individual's photography be
different from normal? affected?
3. Why is it hard to see detail under 3. If one has no flash, why is it hard to
dim lighting conditions? photograph detail under dim lighting
conditions?
4. How would an individual's vision be 4. How would an individual's
affected if the iris could not change photography be affected if the
shape? diaphragm could not change shape?
5. Imagine someone who can only see 5. Imagine a camera that can only
well those objects that are relatively far photograph those objects that are
away. What is likely to be the source of relatively far away. What is likely to be
the problem? the source of the problem?
COMPLEXITY AND ANALOGY 67

6. Imagine that someone had their 6. Imagine that someone had their
lenses removed during cataract surgery, single lens reflex camera's lens
and that no artificial lenses were used to removed for cleaning. If they tried to
replace them. What would be the take a picture with the single lens reflex
effects on the person's uncorrected (i.e., camera while it had no lens, what
without glasses) vision? would be the effects on the person's
Downloaded by [Library Services City University London] at 00:54 18 March 2016

photography?

APPENDIX C
Multiple-Choice Target Questions

(Note: For each set of 12 questions, questions 1 through 6 are memory questions,
and questions 7 through 12 are inference questions.)

Optics Passage Set

1. Which of the following parts of the camera focuses the image?


a. Lens
b. Film
c. Diaphragm
d. Aperture
2. To focus a rangefinder camera, a photographer:
a. Adjusts the lens until he sees a sharp image
b. Adjusts the lens until colors are clear
c. Adjusts the lens until he sees one image
d. Adjusts the lens until he sees multiple images
3. Fast film should be used:
a. When the subject is moving quickly
b. When the image seen by the photographer is blurred
c. When lighting is low
d. When a thick lens is used
4. Film works because of a chemical reaction in:
a. Silver crystals
b. Gelatin
c. The lens of the camera
d. The viewfinder of the camera
5. The aperture of the camera:
a. Lets light into the camera
b. Focuses the image
c. Controls the film advance
d. Has silver crystals
6. The diaphragm of the rangefinder camera:
a. Works best under low lighting conditions
b. Turns the image upside-down
c. Undergoes a chemical reaction
d. Controls the aperture
7. If an individual had arangefinder camera with adiaphragm that was erratic,
Downloaded by [Library Services City University London] at 00:54 18 March 2016

such that he never could tell what it would do when he set it, which of the
following would be true of his photographs?
a. Some of the pictures would be focused, but some would not
b. Photographs of objects that were nearby would be better focused than
photographs of objects that were farther away
c. Detail would be lost
d. Some pictures would be overexposed, and some would be underex-
posed
8. If one had only slow film available, what could that person do to compen-
sate?
a. Use a thicker lens
b. Use a more open aperture
c. Photograph under conditions of dimmer light
d. Arrange for the diaphragm to close the aperture up
9. Imagine that someone had only one lens, which was relatively thick, for his
rangefinder. How would this affect his photography?
a. Pictures taken of objects far away would be blurry
b. Pictures taken of nearby objects would be blurry
c. He would be unable to use fast film
d. He would be unable to use slow film
10. If one's rangefinder had only one aperture setting, which made the aperture
very small, how would this affect the person's photography?
a. He would have to use slow film
b. His camera would work best under bright lighting conditions
c. He would have difficulty focusing the image
d. He would need to use a thicker lens
11. If the lens of arangefinder was lost, which of the following would happen?
a. The photographer would be unable to adapt to differences in light levels
b. The photographer would be limited to the use of fast film
c. The photographer would be unable to focus images
d. The diaphragm of the single lens reflex camera would become nonfunc-
tional
12. Imagine that someone had bought some slow film, which turned out to be
defective. Although the person wasn't aware of it at the time he used the
film, the film's silver crystals were spread in large clumps across the film's
surface. What would happen to this person's photographs?
a. They would be overexposed
COMPLEXITY AND ANALOGY 69

b. They would be more detailed than expected


c. They would not be well focused
d. Only pictures taken of nearby objects would be in focus

Competition Passage Set


Downloaded by [Library Services City University London] at 00:54 18 March 2016

1. The black-footed ferret:


a. Is a mammal
b. Is found in Asia
c. Has a broad ecological niche
d. Eats insects
2. The black-footed ferret lives in:
a. Trees
b. Nests
c. Underground burrows
d. Caves
3. The black-footed ferret eats:
a. Insects
b. Prairie dogs
c. Grasses and leaves
d. a and b
4. When two ecological niches do not overlap much:
a. There is little inter-specific competition
b. There is a lot of inter-specific competition
c. The two animals involved must live in different areas
d. The niches involved must be broad
5. The usual result of high levels of inter-specific competition is:
a. One of the species dies out
b. The two species adjust so that both can survive
c. The breadth of the ecological niche changes
d. One of the species involved changes its food source
6. Intra-specific competition:
a. Refers to competition among different species for resources
b. Refers to competition between man and animal for control of natural
wildlife areas
c. Refers to competition among individuals of the same species
d. Has little effect on the number of ferrets who live in a region
7. If the ferret were to change its diet such that it began to eat a variety of small
rodents:
a. The ferret's ecological niche would narrow
b. The ferret would become more of a specialist
c. The ferret would be better equipped to survive in another habitat
70 CAPLAN AND SCHOOLER

d. The ferret's niche would be less likely to overlap with the niches of
other animals
8. If the populations of coyotes and bobcats were to decline, due to hunting by
Downloaded by [Library Services City University London] at 00:54 18 March 2016

man, which of the following would happen to the black-footed ferret?


a. Inter-specific competition would increase
b. Intra-specific competition would decrease
c. The ferrets would decline in number
d. a and c
9. Imagine that the hawk was declared an endangered species, and that laws
were instituted to protect it from extinction. Subsequently, the number of
hawks over the plains of North America began to increase. What would
happen to the black-footed ferret?
a. There would be more ferrets
b. Inter-specific competition would decrease
c. There would be an increase in the total amount of food available
d. Intra-specific competition would increase
10. Which of the following would lead to an increase in intra-specific competi-
tion for the ferret?
a. A population explosion of prairie dogs
b. A decrease in the population of coyotes
c. A decrease in the population of ferrets
d. A decrease in the number of available prairie dogs
11. Which of the following would result in increases of both inter-specific and
intra-specific competition for the ferret?
a. Importing a new animal to the North American plains that lived exclu-
sively on prairie dogs
b. Importing a new animal to the North American plains that lived exclu-
sively on coyotes
c. Increases in the number of prairie dogs living on the North American
plains
d. An increase in supply of the food that prairie dogs eat
12. Destruction of large portions of the plains for urban development would
lead to:
a. Increases in intra-specific competition for the ferret
b. Decreases in inter-specific competition for the ferret
c. A narrowing of the ferret's ecological niche
d. A decrease in the extent to which the ferret's ecological niche overlaps
with those of other animals

You might also like