TOWARD A PSYCHOLOGY OF MEMORY ACCURACY

Annu. Rev. Psychol. 2000. 51:481–537
Copyright
2000 by Annual Reviews. All rights reserved
TOWARD A PSYCHOLOGY OF MEMORY
ACCURACY
Asher Koriat, Morris Goldsmith, and Ainat Pansky
Department of Psychology, University of Haifa, Haifa, Israel, 31905; e-mail:
akoriat@psy.haifa.ac.il, mgold@psy.haifa.ac.il, pansky@psy.haifa.ac.il
Key Words
memory correspondence, false memory, memory distortion, memory
illusions, memory metaphors
Abstract
There has been unprecedented interest in recent years in questions
pertaining to accuracy and distortion in memory. This interest, catalyzed in part by
real-life problems, marks a signi?cant departure from the quantity-oriented approach
that has characterized much of traditional memory research. We outline a correspon-
dence metaphor of memory underlying accuracy-oriented research, and show how the
features of this metaphor are manifested across the disparate bodies of research
reviewed here. These include work in the Gestalt tradition, spatial memory, memory
for gist, schema theory, source monitoring, ?uency misattributions, false recall and
recognition, postevent misinformation, false memories, eyewitness research, and auto-
biographical memory. In examining the dynamics of memory accuracy, we highlight
the importance of metacognitive monitoring and control processes. We end by dis-
cussing some of the methodological, theoretical, and metatheoretical issues inherent
in accuracy-oriented research, attempting to prepare the groundwork for a more coher-
ent psychology of memory accuracy.
CONTENTS
Introduction ...................................................................................... 482
Basic Characteristics of the Accuracy-Oriented Approach to Memory ............... 483
Expressions of the Correspondence Conception in Accuracy-Oriented
Memory Research ........................................................................... 485
Accuracy-Oriented Research: How and Why Memory Can Go Wrong......... 487
The Gestalt Approach to Memory Changes Over Time .................................. 488
Spatial Memory and Distortion ............................................................. 489
Memory for Gist versus Detail .............................................................. 491
Schema-Based Effects on Memory Accuracy .............................................. 493
Source Monitoring............................................................................. 496
Illusions Stemming from Fluency Misattributions ........................................ 498
False Recall and Recognition................................................................ 500
Misleading Postevent Information........................................................... 503
Real-Life False Memories and Their Creation ............................................ 505
Eyewitness Memory ........................................................................... 507
Autobiographical Memory.................................................................... 510
0084–6570/00/0201–0481$12.00
481

---

482
KORIAT
GOLDSMITH
PANSKY
Metacognitive Processes and Accuracy .................................................. 513
Monitoring the Correctness of One’s Own Knowledge: Metamemory Illusions ..... 514
The Strategic Regulation of Memory Accuracy ........................................... 515
Toward a Psychology of Memory Accuracy: Methodological,
Theoretical, and Metatheoretical Issues............................................... 516
Correspondence-Oriented Research: Phenomena, Questions, and Theories ......... 517
Experimental Paradigms and Assessment Procedures ................................... 519
Memory Accuracy and Error Within a Broader Functional Perspective .............. 521
INTRODUCTION
Despite the enormous amount of research and theorizing on memory in the past
century, there is still no consensual conceptual framework for thinking about
memory. In our view, this state of affairs re?ects the multifarious nature of mem-
ory itself, calling for a pluralism of approaches to the study of memory (Koriat
& Goldsmith 1996b, 1997).
One approach that has dominated the experimental study of memory during
the past century has followed Ebbinghaus (1895) in adopting a quantity-oriented
conception. In this conception, memory is seen as a storehouse into which discrete
items of information are initially deposited and then later retrieved (Marshall &
Fryer 1978, Roediger 1980). Memory is then evaluated in terms of the number
of items that can be recovered over some retention interval. This approach to
memory underlies the traditional list-learning paradigm that continues to produce
much of the data that appear in scienti?c journals.
More recently, however, a very different approach to memory has been gaining
impetus, inspired by real-life memory phenomena. In this accuracy-oriented
approach, which may be traced to the seminal work of Bartlett (1932) among
others, memory is viewed as a representation or reconstruction of past experience.
Hence, memory is evaluated in terms of its correspondence or ?t with past events,
rather than in terms of the mere number of input items that can be recovered.
The vast amount of recent work on memory accuracy and distortion has pro-
duced many new ?ndings and also a search for theoretical frameworks that can
accommodate them. Yet, only very recently has there been an emerging recog-
nition that the new wave of accuracy-oriented research calls for the development
of a metatheoretical foundation that can help in organizing the data and in moti-
vating speci?cally accuracy-oriented memory theories (Koriat & Goldsmith
1996b, Payne & Blackwell 1998, Roediger 1996, Schacter et al 1998). In this
chapter, we present a selective review and analysis of accuracy-oriented memory
research. We ?rst outline the basic characteristics of the accuracy-oriented
approach in terms of a correspondence conception of memory. We then survey
some of the main accuracy-oriented research areas, in the attempt to bring out

---

PSYCHOLOGY OF MEMORY ACCURACY
483
the common features and issues inherent in the study of memory accuracy and
error. We go on to emphasize the role that metacognitive processes have come to
play in current treatments of memory accuracy. Finally, we discuss some of the
theoretical, metatheoretical, and methodological issues that must be faced on the
road to a psychology of memory accuracy.
Basic Characteristics of the Accuracy-Oriented Approach to
Memory
In order to appreciate the unique features of the accuracy-oriented approach to
memory, it is helpful to contrast it with the traditional, quantity-oriented approach.
These two approaches appear to re?ect two fundamentally different conceptions
or metaphors of memory—the storehouse and correspondence metaphors, respec-
tively (Koriat & Goldsmith 1996a,b).
The
Quantity-Oriented
Storehouse
Conception
The
quantity-oriented
approach to memory, inherent in the storehouse metaphor, is well illustrated by
the standard list-learning paradigm, perhaps the hallmark of traditional memory
research (Neisser 1991). This paradigm essentially simulates the course of events
presumed to take place when memory items are initially deposited into and then
subsequently retrieved from a memory store. The contents of the store are
assumed to consist of discrete, elementary units (items), whose basic character-
istic is countability: Measures of memory can be based simply on the number of
recovered elements. Moreover, memory is assessed in an input-bound manner:
One begins with the input and asks how much of it was recovered in the output.
In scoring free-recall performance as percent correct, for instance, commission
errors are essentially ignored (Roediger et al 1997). Forgetting, then is conceived
as simple item loss. Moreover, the items are completely interchangeable as far as
the total memory score is concerned: The content of the recollected and forgotten
items is immaterial. What matters is not what is remembered, but rather, how
much.
These aspects of the list-learning paradigm characterize an approach to mem-
ory in which memory is studied primarily in terms of its amount (Schacter 1989).
This emphasis guides not only the way in which memory is assessed, but also
the phenomena investigated, the questions asked, and the methods and theories
developed to answer them. Until recently, the dominance of this approach was
virtually unrivalled (Payne & Blackwell 1998, Roediger 1980).
The Accuracy-Oriented Correspondence Conception
The accuracy-oriented
approach, in contrast, can be illustrated by a memory paradigm common in eye-
witness research, in which subjects ?rst observe a staged event and are later asked
to recount the event, or are questioned about speci?c details (e.g. Belli & Loftus
1996, Fisher et al 1994). This paradigm embodies a different way of thinking
about memory, one in which the focus is on the correspondence between what

---

484
KORIAT
GOLDSMITH
PANSKY
the person reports and what actually occurred (see Winograd 1994, Payne &
Blackwell 1998). Indeed, much of the recent work inspired by real-life memory
phenomena discloses a keen preoccupation with the reliability, accuracy, or faith-
fulness of memory that has no parallel in the traditional, quantity-oriented
approach. In order to capture the essential features of this alternative view, Koriat
& Goldsmith (1996b) explicated a correspondence metaphor of memory in terms
of the following interrelated attributes:
1. Aboutness: Memory is considered to be about past events (Conway 1991).
Thus, memory reports are treated as descriptions, consisting of propositional
statements that have truth value, rather than as mere collections of recovered
items.
2. Focus on accuracy: Interest lies primarily in the extent to which the memory
report is reliable, trustworthy, and accurate, i.e. the extent to which it accords
with reality (or some other criterion) (see Kruglanski 1989).
3. Forgetting: Forgetting is conceived as a loss of correspondence between the
memory report and the actual event, that is, as a deviation from veridicality.
Thus, in addition to a concern with information loss, this view leads to a focus
on the many different types of qualitative memory distortions (e.g. Bartlett
1932, Schacter 1995)—simpli?cation, fabrication, confabulation, and the like.
4. Content: Unlike the quantity-oriented approach, in which interest focuses on
how much is remembered, in the correspondence-oriented approach (and vir-
tually all real-life memory situations), it matters a great deal what is remem-
bered and misremembered (Conway 1991).
5. Output-boundedness: The assessment of memory correspondence is inherently
output bound. Unlike the storehouse approach, which leads one to begin with
the input and ask how much of it is represented in the output, in a correspon-
dence view of memory it is more natural to focus on the output (e.g. an eye-
witness report) and examine to what extent it accords with the input (e.g. a
witnessed event). In general, accuracy can be measured only for what a person
reports, not for what is omitted.
6. Memory as the perception of the past: The correspondence view of memory
has much in common with the way we think about perception. In perception,
interest lies in the correspondence between what we perceive and what is out
there (i.e. veridicality), and in the various ways in which percepts may deviate
from reality (e.g. illusions). Likewise, under the correspondence metaphor,
memory may be conceived as the perception of the past, and the question then
becomes, To what extent is this perception veridical or illusory (Roediger
1996)?
Collectively, these ingredients of the correspondence conception characterize
an accuracy-oriented approach to memory. This way of treating memory has
become increasingly salient in memory research and theorizing, particularly in
work prompted by real-life memory phenomena (e.g. Intons-Peterson & Best

---

PSYCHOLOGY OF MEMORY ACCURACY
485
1998, Lynn & McConkey 1998, Ross et al 1994, Schacter 1995, Winograd &
Neisser 1992).
Expressions of the Correspondence Conception in
Accuracy-Oriented Memory Research
Even a cursory survey of the recent wave of accuracy-oriented memory research
reveals fundamental differences from traditional quantity-oriented research. In
this section, we point out some of these differences as a backdrop for considering
the speci?c research areas that will be reviewed later.
First, the assumptions underlying the traditional use of the list-learning para-
digm provide quantity-oriented research with memory measures of very broad
applicability. The standard measures of percent recall and recognition, based on
the assumption of item interchangeability, provide a common denominator that
allows a broad spectrum of quantity-oriented research ?ndings to be compared
and integrated. In contrast, accuracy-oriented research has yielded a plethora of
paradigm-speci?c dependent measures that allow less cross-talk between different
areas at the level of memory assessment. This situation derives from features of
the correspondence conception. Because the study of memory accuracy is con-
cerned with the content of the information reported, it is less amenable to global
memory measures. Moreover, the many qualitative ways in which memory of the
past can deviate from veridicality call for memory measures that are tailored to
individual dimensions of miscorrespondence. The focus of quantity-oriented
research on only one dimension of miscorrespondence, omission, helped circum-
vent the many serious methodological and metatheoretical issues facing the study
of memory correspondence (see Methodological, Theoretical, and Metatheoretical
Issues).
Second, the focus on memory accuracy has led to a far more extensive analysis
of the memory output than has been customary in traditional quantity-oriented
research (e.g. Bartlett 1932; Brewer 1988a,b; Neisser 1981). The dramatic
increase of interest in commission errors and false memories in recent years epit-
omizes the departure from the input-bound storehouse conception of forgetting
as loss of studied items. That conception has particular dif?culty accommodating
the idea that memory can be supplemental, i.e. that some of the changes that occur
between study and test involve ‘‘memory’’ for information that was not contained
in the input. As noted by Roediger et al (1998), false recall and false recognition
responses have generally been considered a mere methodological nuisance in the
study of memory, rather than an object of interest in their own right.
Third, the treatment of memory reports as propositional descriptions that have
truth value brings to the fore relational aspects of correspondence and miscorres-
pondence that cannot easily be accommodated within a conception of memory as
a store of elementary units. The treatment of list-learning memory responses as
propositional-relational statements played an insigni?cant role in traditional mem-
ory theorizing and was essentially optional. In contrast, propositional relations

---

486
KORIAT
GOLDSMITH
PANSKY
have ?gured prominently in the study of semantic memory (e.g. ‘‘a canary is a
bird’’) and now they constitute a core of interest in the accuracy-oriented study
of episodic and autobiographical memory as well. Such relational judgments as
when or where an event took place (e.g. Winograd & Neisser 1992), whether the
source of a memory is perception or imagination (e.g. Johnson 1997), and so
forth, are now integral both to memory assessment procedures and to theorizing
about underlying memory mechanisms. Indeed, de?cits in binding together the
various features of complex events have been proposed to underlie such varied
memory errors as the effects of postevent misinformation and confabulations
resulting from frontal lobe pathologies (see Schacter et al 1998).
Fourth, the af?nity between memory and perception inherent in the corre-
spondence metaphor is apparent in much research and theorizing about memory
accuracy and distortion. Several researchers have stressed the similarity between
memory illusions and perceptual illusions (Roediger 1996), proposing that ‘‘we
should consider the study of sensing and perceiving as a model for studying
remembering’’ (Roediger et al 1998:238). Such an af?nity is perhaps most clear
in the study of visual and spatial memory, in which principles governing percep-
tion have often been extended to apply to memory (e.g. Shepard 1978). However,
it is also evident in other research areas, for example in the application of Gestalt
principles to describe changes in memory for depicted events (Allport & Postman
1945), and in cognitive social-psychological research on person perception and
person memory (e.g. Wyer & Srull 1989). The af?nity between perception and
memory is incorporated in the perception/reperception framework (Payne &
Blackwell 1998), which applies similar concepts to the analysis of perception and
memory, as well as in the study of ‘‘memory psychophysics’’ (Algom 1992). In
addition, the attributional approach of Jacoby and his associates (e.g. Kelley &
Jacoby 1998), affords an analysis of perceptual and memory illusions within the
same conceptual framework.
Fifth, perhaps also part of the legacy of perception in the study of memory
correspondence, is the increased interest in the phenomenal qualities of recollec-
tive experience. Experiential, subjective qualities attracted little interest in tradi-
tional quantity-oriented memory research. In contrast, many current studies of
memory accuracy incorporate measures of various subjective characteristics simi-
lar to those used in imagery research, such as vividness and perceptual-contextual
detail (Conway et al 1996, Johnson 1997, Lampinen et al 1998). Also included
are measures of the state of awareness accompanying remembering (the know-
remember distinction) (Tulving 1985; see also Gardiner & Java 1993, Gardiner
et al 1998), as well as metacognitive feelings like the sense of familiarity, the
feeling of knowing, the feeling of recall imminence, and subjective con?dence
(Benjamin & Bjork 1996, Koriat & Levy-Sadot 1999, Schwartz 1998). Such
subjective measures have been examined in connection with reality and source
monitoring (Suengas & Johnson 1988), autobiographical memories (Brewer
1992), false recall (Payne et al 1997, Roediger & McDermott 1995, Schacter et
al 1996), postevent misinformation (Zaragoza & Mitchell 1996), ?ashbulb mem-

---

PSYCHOLOGY OF MEMORY ACCURACY
487
ories (Conway 1995), and eyewitness testimony (Fruzzetti et al 1992). No longer
mere epiphenomena, experiential qualities have been treated as an integral com-
ponent of the process of remembering (e.g. Johnson 1997, Norman & Schacter
1996) and in particular, as diagnostic clues used by both rememberers and observ-
ers in the attempt to distinguish genuine from false memories (e.g. Conway et al
1996, Koriat 1995, Ross 1997, Schwartz 1998). The assumption is that the quality
of phenomenal experience may be critical in leading the rememberer to accept a
memory as true.
Finally, the conception of memory as being about something has spawned a
departure from the passive storehouse conception toward a more active view, in
which remembering is an intentional, goal-directed ‘‘effort after meaning’’ (Bart-
lett 1932:20). This, of course, is the hallmark of Bartlett’s reconstructive
approach, in which ‘‘remembering is not the re-excitation of innumerable ?xed,
lifeless and fragmentary traces. It is an imaginative reconstruction or construc-
tion’’ (1932:213). Thus, a vast amount of accuracy-oriented research has been
devoted to examining the consequences of the assumption that information is not
simply deposited into a memory store, but is assimilated and integrated into cog-
nitive structures (e.g. schemas) and later recreated from those structures. More
recently, the active role of the rememberer has also been gaining prominence in
the expanded notion of retrieval processes (e.g. Norman & Schacter 1996) and
in work emphasizing the metacognitive processes of monitoring and control that
mediate accurate memory performance (Goldsmith & Koriat 1999, Koriat &
Goldsmith 1996c). Many authors have emphasized complex evaluative and deci-
sional processes used to avoid memory errors or to escape illusions of familiarity
(e.g. Burgess & Shallice 1996, Kelley & Jacoby 1996, Koriat 2000, Schacter et
al 1998). The operation of these processes is particularly crucial in real-life sit-
uations (e.g. eyewitness testimony) in which a premium is placed on accurate
reporting.
The preceding list represents a rough attempt to characterize some of the
unique features of the correspondence-oriented study of memory. We now turn
to an examination of how these features manifest themselves in speci?c research
areas.
ACCURACY-ORIENTED RESEARCH: HOW AND WHY
MEMORY CAN GO WRONG
The following survey brings together and examines somewhat disparate lines of
accuracy-oriented memory research. Because one of our aims is to demonstrate
the broad scope and diversity of accuracy-oriented research, it is simply not pos-
sible to be comprehensive. Instead, this survey is both selective and deliberately
biased to highlight the issues, experimental paradigms, and phenomena that are
distinctive of the study of memory correspondence. The sections have been orga-

---

488
KORIAT
GOLDSMITH
PANSKY
nized to preserve as much as possible the coherence of different bodies of
research. Following this survey, we go on to an analysis of some of the common
issues and challenges facing these various lines of research.
The Gestalt Approach to Memory Changes Over Time
Although Bartlett (1932) is generally credited as being the founder of the quali-
tative, accuracy-oriented approach to memory, many facets of this approach were
already apparent in the study of memory by Gestalt psychologists (see Koffka
1935, Riley 1962). Rejecting the Ebbinghaus-type focus on the number of remem-
bered nonsense syllables, they revived the emphasis on qualitative aspects of
memory that had been pioneered as early as the late nineteenth century by several
students of memory for visual form (see Estes 1997, Woodworth 1938, for
reviews). These latter researchers found, for example, that observers’ reproduc-
tions from memory were characterized not only by loss of detail, but also by
substitution of new detail, and object assimilation—a tendency of reproductions
to shift toward the typical form of familiar objects.
Gestalt psychologists, extending the Gestalt principles from perception to
memory (see Koffka 1935, Riley 1962), explained these distortions as resulting
from autochthonous cortical forces that transform perceptual traces into more
regular, symmetrical, and simple memory forms (Pra¨gnanz). Wulf (1922), who
had subjects draw geometrical ?gures from memory, identi?ed two opposite types
of changes: ‘‘sharpening,’’ which involves the exaggeration of selected charac-
teristics of the original ?gure, and ‘‘leveling,’’ which entails a weakening of one
or more features. These changes were assumed to be progressive, such that later
reproductions tend to exaggerate the deviations of the previous ones. Based on
these results, Wulf postulated three causal factors underlying both leveling and
sharpening: ‘‘normalizing,’’ which refers to changes toward a well-known or con-
ventional ?gure, ‘‘pointing,’’ which refers to changes that emphasize a feature of
the stimulus, and ‘‘autonomous changes,’’ which re?ect systematic self-governed
modi?cations of the memory trace toward simpler and more regular patterns
(‘‘good form’’). It is the postulation of autonomous, intrinsic changes operating
on the memory trace that is unique to Wulf’s Gestalt perspective. According to
this perspective, the memory engram ‘‘cannot be regarded as an immutable
impression which can only become blurred with time, similar to a drawing carved
on a brick. Rather this engram undergoes changes by virtue of gestalt laws’’ (Wulf
1922:370).
Goldmeier’s stress model (1982) speci?es the conditions under which distor-
tions in the direction of ‘‘good ?gure’’ will occur, conceptualizing Pra¨gnanz in
terms of the notion of ‘‘singularity.’’ Singular features (e.g. a full circle) contain
no stress and should remain stable and accurate over time. By contrast, nonsin-
gular features, such as ambiguous or poorly integrated material, produce unstable
and imprecise traces that gradually lose information. It is the nearly singular
features (e.g. an almost closed circle), those having the strongest stress, that

---

PSYCHOLOGY OF MEMORY ACCURACY
489
elicit the tendency to shift toward singularity and therefore should exhibit pro-
gressive distortions over time. The nearly singular traces originally have the struc-
ture of schema-plus-correction, and the gradual distortion in memory eliminates
the correction but retains the schema. This change toward increased self-
consistency is adaptive, because it achieves maximal compactness within the trace
system while suffering only a minimal loss of information.
Rhodes’ (1996) more recent work on distortions in face recognition can also
be seen as an example of the operation of sharpening. Using a computer-
implemented caricature generator to manipulate the distinctiveness of facial fea-
tures, line drawings of faces were distorted by either exaggerating the metric
differences between each target face and a norm (i.e. sharpening), thus creating
a caricature of the original face, or, conversely, by reducing these differences (i.e.
leveling), creating anticaricatures. Several studies (see Rhodes 1996 for a review)
indicated that when subjects learned to associate a name with a face, naming the
caricature version of the face was faster than naming its anticaricature version,
suggesting that sharpening is less disruptive to recognition than leveling (Rhodes
et al 1987, 1997). Furthermore, the recognition of the caricature versions was as
good or even better than that of the original face (Benson & Perrett 1994, Rhodes
et al 1987). Rhodes concluded, ‘‘In some cases caricatures are even superpor-
traits, with the paradoxical quality of being more like the face than the face itself.’’
(Rhodes 1996:1). She proposed that the effectiveness of caricatures in recognition
may derive from the fact that the representations stored in long term memory are
‘‘schematized so as to emphasize the distinctive properties of what is being rep-
resented.’’ (Rhodes et al 1987:474).
The Gestalt idea of distortion toward ‘‘better form’’ has also been very in?u-
ential in social cognition, primarily in cognitive consistency theory and cognitive
balance theory, following from the work of Fritz Heider (1958; see Gilbert et al
1998). This research too illustrates the continuity between memory and percep-
tion, and in fact in this type of research, the distinction between memory (i.e.
person memory) and perception (i.e. person perception) is generally blurred.
Spatial Memory and Distortion
The study of spatial memory also brings to the fore various features of the cor-
respondence metaphor. First, it discloses an explicit interest in the nature and
basis of the correspondence between memory representations and their spatial
referents. Second, it invites the application of assessment procedures that depart
greatly from those that follow from the storehouse metaphor. Third, it highlights
some inherent similarities between the study of memory and of perception, allow-
ing both to be analyzed in terms of the same theoretical constructs.
Early studies were primarily concerned with demonstrating an isomorphic
mapping between spatial layouts and their memory representations. Kosslyn et al
(1978), for example, found that the time it took to scan between two points on a
mental image of a memorized map increased with the actual distance between the

---

490
KORIAT
GOLDSMITH
PANSKY
points. A similar isomorphism was demonstrated in studies comparing perfor-
mance in the presence and in the absence of spatial maps (e.g. Kerst & Howard
1978, Thorndyke 1981). These studies laid the groundwork for exploring various
aspects of miscorrespondence.
Memory psychophysics (see Algom 1992) embodies the view of memory as
the perception of the past, bringing perceptual issues and techniques to bear on
the study of memory. For example, Kerst & Howard (1978) found that perceptual
and memorial estimates of distance were related to the actual distances by similar
power functions, but the exponent of the memorial function was equal to the
square of the exponent of the perceptual function. They proposed a ‘‘re-perceptual
hypothesis’’: The same psychophysical transformation that operates on the sen-
sory input to produce a perceptual representation is reapplied to the perceptual
representation to produce the memorial estimates (but see Radvansky et al 1995
for alternative accounts). Thus, memorial judgments are performed via ‘‘internal
psychophysics’’ (Moyer 1973), causing a magni?cation of perceptual distortions.
Thorndyke (1981) found that for perceptual as well as memorial tasks, esti-
mated distances increased as a linear function of the number of intervening points
(i.e. ‘‘clutter’’) along the route. Thus, although distance estimations made from
memory were not entirely faithful to the actual distances in the external environ-
ment, they were faithful to the information that perception delivered to memory.
In fact, Thorndyke, as well as others, implied that such memory distortions as the
clutter effect actually stemmed from misperception: perceptual biases or illusions.
However, accumulating evidence of systematic distortions in spatial memory
has motivated alternative accounts, attributing spatial distortion to error-prone
reconstructive heuristics. For example, Byrne (1979) found an overestimation of
distance for routes containing bends as opposed to linear routes, and for routes
within the town center as opposed to peripheral routes, as well as a tendency to
normalize the angles between urban roads to 90o. Byrne proposed that spatial
representations do not preserve the exact metrics of the spatial environment (e.g.
veridical distances or angles); rather, subjects base their estimates on heuristics
(e.g. ‘‘the more locations that are remembered along a route, the longer the route
must be’’) that are generally adequate, but are sometimes prone to bias or
inaccuracy.
Additional types of errors in distance estimation also challenge the assumption
that internal representations preserve metric spatial information. For example, the
?nding that landmarks produce asymmetric distance estimates—nonlandmarks
judged as closer to landmarks than vice versa (e.g. McNamara & Diwadkar
1997)—clearly violates the symmetry of Euclidean distances. This bias too has
been explained in terms of the reconstructive view, which suggests that spatial
properties ‘‘are not retrieved from long-term memory and reported in pure form,
but rather, are interpreted and scaled by the context in which the retrieval takes
place.’’ (McNamara & Diwadkar 1997:188). This approach can also account for
the ‘‘perspective effect’’: Subjects who imagined themselves in New York judged
the distance between New York and Pittsburgh to be longer than those who imag-

---