lilliliiglillililiiliiiiill \

THch

This is to certify that the
thesis entitled
Psychophysical Testing in Cartography:
An Evaluation of Methodology
presented by

Daniel Gerard Cole

has been accepted towards fulfillment
of the requirements for

 

M.A. degree in Jeogtam

' LIBRARY

Michigan State
University

 

 

MAL/(Mp

Majoproesfs

Date g/g’fli

OVERDUE FINES ARE 25¢ PER DAY
PER ITEM

Return to book drop to remove
this checkout from your record.

 

 

 

PSYCHOPHYSICAL TESTING IN CARTOGRAPHY;
AN EVALUATION OF METHODOLOGY

By

Daniel Gerard Cole

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Geography

1979

ABSTRACT

PSYCHOPHYSICAL TESTING IN CARTOGRAPHY:
AN EVALUATION OF METHODOLOGY

By

Daniel Gerard Cole

Cartographers have employed psychophysical testing to
determine the average map reader's perception of the sizes or
values of circles, dots, gray-tones and patterns. But numerous
variables exist within the testing procedures themselves
which have not been examined. Hence, the author uses
graduated circle maps in analyzing two variables: short
term memory response and task orientation. Ninety-six
subjects were tested on their ability to recall or recognize
a mapped circle pattern given one of three instructional
levels.

Several statistical methods were utilized to evaluate
the data. The results indicate that there is indeed a
significantly large difference between recall and recognition
and a smaller but nonetheless significant difference between

very specific and non-specific instructional sets.

DEDICATION

To my parents for their continual support and without

whom none of this would have been possible.

ii

ACKNOWLEDGMENTS

I gratefully acknowledge the advisement and assistance
received in various forms throughout my program from the
following people: Professor Richard Groop who provided
the original idea for the tOpic of this paper, numerous
suggestions, much patience and an empty red editorial pen;
Professor Richard Smith who provided additional ideas and
guidance as well as the impetus for my attendance at Michigan
State University; Professor Bruce Pigozzi, without whom, the
statistical analysis of my data would have been a disaster;
and Professor Dieter Brunnschweiler who helped maintain my
interest in Remote Sensing such that my dissertation will fall
under that heading. I also wish to thank the Department of
Geography for the teaching assistantship awarded during all
of my attendance.

On the lighter side, many of you probably realize that
graduate students maintain strange hours and eating habits.
In that vein, I must thank Jim Johnson for teaching me that
while researching and writing into the late hours of the
night, one can survive on nothing more than sandwiches of

choke and slide.

iii

CHAPTER

I.

II.

III.

IV.

TABLE OF CONTENTS

INTRODUCTION

Problem Statement
Purpose.
Importance . .
Other Variables.
Summary.

HYPOTHESES AND METHODOLOGY

Introduction
Hypotheses . .
Testing Methodology.

Recall and Replicate
Recognition. . .

DATA ANALYSIS.

Introduction

Frequenqyof Choices-—Recognition .

Mean Deviation Percent Black Per Quadrant
Mean Squared Distance. .
CircularNormal Distribution.

Centroids.

Summary.

DISCUSSION

Conclusions.

Error.

Consistency.
Recommendations.
Further Research

BIBLIOGRAPHY .

iv

71

71

71
72
73
74

76

LIST OF TABLES

Table Page
1. Warrington and Ackroyd's Test of
Orientation Tasks. . . . . . . . . . . . . . ll
2. Absolute Total and (Mean) Deviations. . . . . 48
3. Summary Table of the Mean Error Terms
for all Test Groups. . . . . . . . . . . . . 7O

Figure

9a.

9b.
10.
11.
12
(a- h)
13.
14.
15

(a- C)

16.
(a- C)

LIST OF FIGURES

Results of Tversky's Tests.
Results of Loftus and Loftus' Tests

Hypothesized Difference of Error
Between Recall and Recognition.

Hypothesized Difference of Error
Between Instructional Sets.

Hypothesized Progression of Error Terms

Hypothesized Differences in Consistency
Between Recall and Recognition.

Hypothesized Differences in Consistency
Between Instructional Sets.

Hypothesized Progression of Consistency

Stimulus Map Used in the Recall Tests
(Reduced 10%) . . . . . . . . .

Stimulus Map Used in the Recognition
Tests (Reduced 10%) . . . . . .

Blank Map Used in the Recall Tests
(Reduced 10%) . . . .

An Example of a Reproduced Map:
Subject #4 in the VST (Reeduced 10%).

Distractor Stimuli Used in the Recognition
Tests (Reduced 10%) . . . . . . . . .

Slip of Paper Used for Ranking in the
Recognition Tests .

Mental Processes During Recall and
Recognition (After Loftus and Loftus, 1976)
Rankings for the Stimulus Map

First Order Rankings for all Maps

vi

Page
10
10

18

19
20

20

21
22

26

27

30

31

33—36

37

37
43

45

Figure

17.

18.

19.

20.

21.

22.
(a- f)

23.

24.

25.
26.

27.
(a- 1)

28.
29.

Mean Absolute Deviations of the
Percent Black Per Quadrant.

Test Group Values of the Mean
Squared Distance Before Rotation.

Test Group Values of the Mean
Squared Distance After Rotation

Test Group Values of the Mean Distance
From the Origin (r)

Test Group Values of the Mean Angular
Deviation (s) . . . . . . .

Graphic Representations of the Mean
Distance From the Origin and the Mean
Angular Deviation for All Test Groups

Mean Distance From the Origin for
Each Circle in the Recall Tests

Mean Angular Deviation for Each
Circle in the Recall Tests.

Weighted (r) Values for All Test Groups
Weighted (3) Values for All Test Groups
Response Centroid Locations in
Relation to the Stimulus Centroid

for All Test Groups (Recall x 7;
Recognition x 11) . . .

Mean Vector Lengths for All Test Groups

Weighted Mean Vector Lengths for
All Test Groups . . .

vii

Page

48

50

52

54

55

56

58

58
58
58

60-65
66

67

Chapter I

Introduction

Over the last 100 years, geographers and others have
increasingly relied on thematic maps for the communication

of spatial distributions. The thematic map's ...main
objective is specifically to communicate geographic concepts
such as the distribution of densities, relative magnitudes,
gradients, spatial relationships, movements, and all the
myriad interrelationships and aspects among the distributional
characteristics of the earth's phenomena (Robinson & Sale,
1969, pp. 10-11).” But this objective is not achieved if the
map-reader does not understand the data or misinterprets it.
Thus, in order to effectively communicate information

through maps, cartographers have attempted to standardize
symbols, to increase the accuracy of data bases, and most
importantly, to understand the process through which spatial
information is transmitted from the map to the reader. This
latter area of concern has included many tests of map readers'
perceptions of cartographic symbology such as circle sizes,
gray-tone values, dot densities and pattern correspondence.

The present study is directed toward the methodology involved

in cartographic perceptual testing.

Problem Statement

 

Cartographic perceptual testing methodology usually

1

involves the use of psychophysical techniques. In most
non-cartographic psychophysical testing, investigators
conduct highly controlled experiments whereby only one or
a few variables are scrutinized. This allows an evaluation
of responses to simple stimuli and leads to more certainty
in analyzing results. Cartographic perceptions, however,
are usually much more complex because of the presence of a
complex stimulus—-the map. Experiments are less tightly
controlled and results are more tentative. But most
importantly, the methodology employed in the experiment is
more likely to effect the outcome of the psychophysical test.
One may assume that if the subject's information output
does not equal the cartographer's information input, then
error is being introduced into the testing procedure by
either the map maker or the map reader. In other words,
"the act of testing, per se, appears to alter the behavior
it sets out to measure (Cooper & Monk, 1976, p. 133).”
The variables which induce this error include: (1) the
complexity of the test maps; (2) the map reader's prior
knowledge (environmental perception) of the mapped area and/
or topic; (3) the reader's ability to learn over the course
of successive tastings; (4) the instructional set (task
orientation) assigned to the reader before and during the
test; and (5) the reader's ability to either recall or
recognize (short term memory) patterns or symbols during the
test. To date, little empirical research has been conducted

in cartography to evaluate the effects of these variables on

the results obtained in cartographic psychophysical testing.

Purpose

The focus of the present study is to examine two of the
above variables, task orientation (specificity of instructions)
and short term memory. (Ngtg: Technically, short term
memory is immediate memory; but since all of the variables
pertaining to it are present in this study and since the use
of the subjects' long term memory store was avoided, the
term short term memory will suffice here.) These variables
are scrutinized in the context of pattern recognition and
reproduction on thematic maps.

Four critical questions concerning these variables can
be raised: What are the factors which influence the map
reader's ability to recall or recognize a pattern? How do
these factors affect the reader's short term memory? Will
the changes of specificity in the instructions create
significantly different map reader responses? And will tests
involving recall and recognition produce significantly
different results? The first two questions are addressed
in this chapter while the latter two are dealt with in the

following chapters.

Importance

 

Prior to considering the above questions, the importance
of task orientation and short term memory of map patterns

should be addressed by examining the psychophysical studies

that have investigated and utilized these procedures. Within
cartographic perceptual testing, apparently task orientation
greatly influences the amount and type of information
accessed by a map reader. Most cartographers assign specific
tasks to their test subjects, but few have empirically
studied the effects of the level of instructional set on map
reader responses. Generated responses from various types of
instructions may simulate the information transfer in a
"normal" map reading situation. If a map reader merely
conducts incidental viewing of a map in an article, the
amount of information accessed by him is less than it would
be if the author directed the reader to the map or to parti-
cular portions of the map. In the same way, given non-specific
instructions, the amount of information accessed by a test
subject is less than it would be if the testor very
specifically directed the subject to the pattern or symbols
on the map or indicated a later task to be performed by the
subject.

The importance of recognition and recall can be seen in
most cartographic psychophysical tests since the comparison
and discrimination between individual symbols or patterns
invariably involves the use of one of these techniques
(Flannery, 1956; Williams, 1956; Muehrcke, 1969; Olson, 1970).
While recognition has been widely employed in cartographic
testing, recall has been limited to test subjects producing

cognitive maps of their environment and a single test in

 

cartography in which the subjects attempted to reproduce

a stimulus pattern (Downs & Stea, 1973; 1977; Steinke, 1975).
But recall is conceptually more important then recognition:
In a map reading situation, the reader seldom refers back

to a previously viewed map; instead, he probably relies

upon his recollection of the map when viewing a secondary
stimulus map. Thus, unless the maps are adjacent to one
another, recall may assume the dominant position in common
map reading tasks.

Next, one may well ask: Why examine pattern recognition
or recall and not that of symbols? The dichotomy between
the study of map patterns and symbols is expressed in the
following passage:

Maps are communicative devices designed to

display spatial information in a two dimensional

format. To some map makers and map users these

displays are considered to be aerial data banks

or storehouses of a myriad of separate and

isolated facts. Other map makers and map

users turn to the map as a communicative

device because the two dimensional format

allows them to display, and see, the new

information which derives from the juxtaposition

of sets of symbols (Jenks, 1975, p. 311).
The latter type of map information transfer, i.e., the map
reader's ability to recall or recognize the pattern as a
whole is the most important in thematic mapping. In fact,
the purpose of the thematic map has been defined as that
which communicates concepts, not data—-the map conveys a
pattern, not the components of the pattern (Gerlach, 1971,

p. 194). And the map reader can perceive the pattern as

a simultaneous whole on the basis of the interrelation of
all of its parts held together in one immediate representation
(Blumenthal, 1977, p. 71)."

For one to examine patterns, and not the individual
symbols within the patterns can be seen in that the map
reader seldom searches for value-size relationships between
symbols. One might also reason that the map reader tends to
group the data into simpler patterns (or rejects it entirely
if it is too complex) so that it may be more easily remembered
for future use. This mental generalization could be due to
the reader's indifference to the task of enumerating symbols
or to the lack of time spent by the reader on the map.

Recent eye movement studies have tended to confirm the
idea that map readers do not ordinarily evaluate individual
symbols. For instance, Dobson (1977) noted that the reader
spent little time looking back and forth between the legend
and the various circles on his test map to check the values.
This is not to say that cartographers do not have to worry
about the perceptual accuracy of individual symbols; on the
contrary, one must still present perceptually accurate symbols
because only they will result in perceptually accurate
patterns. Fortunately, numerous studies have been conducted
in which map readers were tested on their ability to discrim-
inate and assign values to individual circles (Flannery, 1956;
Meiheofer, 1979) or to individual gray-tones (Williams, 1956;

Kimmerling, 1975). Even so, one must remember that the

reader will mentally generalize; that is, forget or coalesce
part of the pattern or combine some of the symbols.

Hence, a number of cartographers have considered pattern
analysis an important part of map communication. Castner
(1964), Jenks (1975) and Steinke (1975, 1979) have attempted
to evaluate the visual comparison and reproduction of
graduated circle patterns. Still other cartographers have
examined the characteristics of patterns on chorOpleth maps.
Several of these people (Olson, 1970, 1972; Monmonier, 1975;
Lloyd and Steinke, 1977) set out to measure the effects of
class interval systems on the visual correlation (pattern
recognition and comparison) of choropleth maps. Other studies
(Muehrcke, 1969; Monmonier, 1974; Olson, 1975; Muller, 1976)
have been solely concerned with the visual analysis of
choroplethic pattern complexity. All of the above studies
examined either the subjects' ability to discriminate, compare
or reproduce patterns. As a unit, these studies illustrate
the manner in which map patterns are processed in the human

perceptual system.

Other Variables

 

Pattern analysis aside, a number of uncontrollable
factors are present within the test subjects which may
influence their recall or recognition capabilities as
illustrated in the following passage:

We have asked the students to produce external

representations as sketch maps. Can we judge
similarity by comparing data taken from the

sketch maps with similar data from a carto-
graphic map? If this is an attempt to solve

the accuracy question, the answer is no.

People vary widely in simple graphic abilities.
Age affects basic manual skills involving eye-
hand coordination. Both the young and the old
differ from our college students. Even
discounting this age factor, we have the problem
of differential training in both the rationale
for and the manual reproduction of sketch maps.
Some people, notably artists, architects and
geographers are trained in technical graphic
skills. One would expect that the mechanical
production and accuracy of the sketch maps would
reflect this training...Styles of training and
thinking affect representations, even if both
people are the same age, experience, skill,
training, the question of the similarity of
cognitive maps cannot be answered definitively.
The nearest that we can come to such a goal is as
follows: Parts of our cognitive maps are common
to all or most members of a large group of people,
parts are common to a subgroup of people, while
still other parts are unique to each person
(Downs & Stea, 1977, pp. 100-103).

Equally troublesome is the problem of attitude on the
subject's part. Since volunteer subjects may bias the
results, "non-volunteers" are usually rendered from classroom
settings, however, no one is ever forced to participate.

The use of non-volunteers, however, probably involves the
use of some students who care little about the experiment.
Since this study is only concerned with the responses of the
"average" map reader, the effects of the above variables
upon the individual reader's expression of short term memory
may be offset by the reactions of other subjects in the same
test group. Hence, while being aware of the aforementioned
uncontrollable variables, they are relegated to a position

of minor importance within this study.

Of greater importance, however, are the variables that
are intimately related to task orientation and short term
memory and their effects on the visual comparison or repro-
duction of patterns on maps. This relationship may be
transformed into the following cause and effect perceptual
processes: (1) the test results are dependent upon short
term memory and its transmittors (eyesight and/or fine motor
skills); (2) short term memory is dependent upon the amount
of information received; (3) the amount of information
received is dependent upon eye movements; and (4) the eye
movements are in part dependent upon the task orientation
(instructions) as initially given to the subject, and in part
dependent upon the pattern's complexity, the amount of time
spent viewing the map, the length of time delay after
viewing the map before starting the recognition or reproduction
task, and the subject's prior knowledge of the mapped topic
and/or area.

These processes may be addressed individually. First,

a number of studies have established that the type of
instructions given in a psychophysical test do indeed affect
the subject's performance. Tversky (1973) noted that if the
subject was previously informed of the type of task to be
performed, be it recognition or recall, he would perform
better than one who was not informed or incorrectly informed.
In all cases, however, the recognition tasks had a higher

response probability than the recall tasks (Figure 1). Even

though Loftus and Loftus (1976) agreed with this statement,
they found that the difference between incidental (nonspecific)
and intentional (specific) types of directions is negligible
in recognition tests while the difference is significant

in recall tasks (Figure 2).

%r
009 0/1/00

9299‘

Response Probabflny

 

 

l
Rec09nmon Recml
Insnucfions

Figure 1 Results of Tversky's Tests

 

 

Recognition
E‘
E
(U
.D
8
CL
a: \
E’ 9ch
O
Q
(I)
<1)
(I
l l
IncMentm knenflonm
Insnucflons

Figure 2 Results of Loftus and Loftus' Tests

11

Warrington and Ackroyd (1975) tested three tasks:
(1) no orientation, (2) relevant orientation and (3) non—
relevant orientation. They found no significant difference
between the no orientation and the non—relevant orientation
tasks but the relevant orientation task produced a signifi—

cantly better performance on the subjects' part (Table 1).

Table l

Warrington and Ackroyd's Test of Orientation Tasks

 

(Mean error)

Non-relevant None Relevant
Words 9.90 12.60 4.75
Faces 10.00 11.10 7.00

DeLucia also tested three different tasks, but he
evaluated the directions in terms of eye movement patterns
across a map. His tasks included: (a) no orientation,

(b) general orientation, and (c) specific orientation.

In the no orientation task, the subjects primarily conducted
free scans of the map. In the general orientation task,

the subject was "only told what to look for but not where
and therefore could not narrow down his search area on the
basis of the assigned task prior to commencing his scan,”
i.e., the subject conducted a free scan until he found what
he thought he was looking for. The specific orientation

task was one of comparison: the subject was told what to

12

look for and where to look on the map and therefore the areas
not fixated upon ”are avoided by the viewer because they

do not contain information relevant to his problem at the
moment (1974, pp. 239 and 241).” Unfortunately, DeLucia did
not conduct any quantitative analysis of his data.

Steinke conducted a test whereby the subjects knew
about the eye movement recording part of his experiment
before it began because of an earlier class presentation
and a brief introduction when they first arrived for the
experiment, but they were not aware that they would have to
reconstruct the target map body later. That is, Steinke
purposely designed his test as non-specific because he
wanted to define what people do under a free look situation
even though relatively little real map reading occurs in
this way. Likewise, telling the subjects before or during
the experiment that they would have to reproduce the map
body later would no doubt have increased motivation but at
the same time would have resulted in very different map
reading activity since few people read maps with the idea
of reproducing them later (correspondence with Steinke,
March 1979).

On the other hand, one might reason that for a specific
orientation task, the test subject will probably scan a test
map in much the same way that the stimulus map was scanned.
In other words, if the search pattern for familiarization is

repeated for recognition, then the subject will have an

l3

easier and faster job of recognizing the correct pattern
(Norton and Stark, 1971; Whiteside, 1978). Hence, in
preparing a test, one should assign one of three tasks for
the subjects: non-specific, somewhat specific and very
specific. The specificity of the instructions will at
least partly determine the test results.

Aside from the instructions given, a number of other
variables affect the subject's short term memory. First,
the complexity of the pattern must be considered. Phillips
(1974) discovered that as the complexity of the pattern
increases, the error in short term memory increases. The
error may be due to the fact that as pattern complexity
increases, the short term memory store becomes overloaded
whereby parts or all of the pattern may be forgotten or
may not even be initially absorbed (Herriott, 1974). In fact,
Kaufman, e3 al.,(1949) provided evidence which indicates that
a subject cannot perceive the pattern as a whole beyond
approximately eight elements in the pattern. But French
(1954) stated that target recognition improved with an
increase in target complexity (possibly due to the "uniqueness"
of the more complex patterns) and became worse with an
increase in visual noise. French's data suggested, however,
that this function is negatively accelerated.

The recognition of pattern types seems to involve some
additional influences of complexity. For example, Fitts 32 a1.

studied the effects of redundancy, i e., an inverse measure

14

of complexity, and their results indicated that ”there is

no simple relationship between the redundancy of figures

and pattern recognition. The introduction of redundancy

may either facilitate or hinder pattern recognition,

depending on the way in which it is introduced (1956, p. 10).”
Fitts, 3; al. also found that "random figures were

recognized more rapidly than were constrained figures.

Symmetrical and vertically oriented figures were recognized

more rapidly than were single or double assymmetrical

figures or horizontally oriented figures of the same

complexity (1956, p. 10).” And during the time that the

subject is viewing the stimulus map and during the delay

time between the initial exposure and the recall or recognition

task, it appears that the pattern is organized into a coding

scheme along horizontal and vertical axes within the human

visual system (Dodwell, 1970, pp. 112-13). Given the

above, perhaps the short term memory coding scheme is stronger

along the vertical axis than it is along the horizontal axis.
Obviously, the construction and arrangement of circle

patterns influences one's ability to recognize that pattern.

Consequently, in constructing a test map, one should try to

create a pattern of medium" complexity, i.e., a pattern
that has a clearly recognizable shape but has no symmetry
and does not contain too few or too many circles, while at
the same time keeping that pattern ”realistic” in appearance.

The amount of time that the subject spends looking at

the map may also affect the amount of information that is

15

being stored. Dobson observed that "the speed at which
information was accessed varied from subject to subject
(1977, p. 53).” And Steinke found that there was ”little
relationship between how much time a person looks at a map
and his ability to reproduce it (1975, p. 220)." (Nggg:

the amount of time that the subjects were exposed to the
stimulus map in each test varied from 12.99 to 32.65 seconds
and 12.6 to 40.3 seconds, respectively.)

The passage of time between the end of the presentation
of the stimulus map and the start of the recall or recognition
test may also be a factor influencing what is retained in
the short term memory store. ”In general, short term memory
is observed as a temporal constraint on recall capacity and
not as a constraint on recognition capacity, recognition being
a distinct and powerful long term memory ability. An object
briefly seen can be totally unavailable to recall, yet days
later its recurrence may be recognized immediately (Blumenthal,
1977, p. 72)." Apparently, the specific impressions of a
circle pattern disintegrates with the passage of time and
one should therefore minimize the delay time in the testing
process.

Taken one step further, "the last few items presented
(or looked at) tend to be recalled first (Baddeley, 1976,

p. 103).” But ”elapsed time, per se, does not affect
retention at all, and that the retrievability of memory is

solely a function of how much interference has occurred

16

at the time of recall (Blumenthal, 1977, p. 72).” Like time,
interference is less likely to affect recognition than it
will recall (Wicklegren & Norman, 1966, p. 346). But neither
recognition nor recall are seriously affected by interference
in the short term memory store (Herriot, 1974, p. 6).

Mandler (1972) noted that the proportion of information
(pattern) which was presented but not recalled had not been
organized by some people; and he stated that some of the
information recalled was not originally presented. Had his
tests included maps, this "new” information was probably
derived from past (long term) memory or environmental
perception of the mapped area or topic and organized into
the stimulus map pattern. Finally, Tulving and Pearlstone
(1966) discovered that there was much more memorized material
available at the time of recall than can actually be retrieved.
This discrepancy may be due to and widened by the use of fine
motor skills and eyesight (each of which may or may not be
well developed), in the reproduction and recognition of

circle patterns, respectively.

Summary

Given that this thesis is focused upon two important test
situation variables, task orientation and short term memory,
one finds that not only do these two variables affect test
results but a number of other variables also influence the
results either directly or indirectly through the above two

variables. Many of these variables can be controlled or

17

at least held constant while others cannot and one must
therefore be aware of their presence when evaluating the

test data. Overall, cartographic psychophysical test results
are not merely influenced by the instructional set or the
type of memory response being elicited; instead, a myriad

of complex variables are present which effect both each

other and final test results.

Chapter II
Hypotheses and Methodology

Introduction

This section specifically addresses the manner in which
three levels of task orientation (non-specific, NST; somewhat
specific, SST; and very specific task, VST) should hypothetically
affect the test results with either the goal of pattern
recognition or pattern reproduction of a graduated circle map.
A detailed explanation of the testing methodology will follow
since, after all, one of the primary foci of this thesis is to

”test the test.”

Hypotheses

Based upon the considerations noted in Chapter I, one
may make the following general hypotheses:

I (a) Error will be less for recognition tasks than

for recall tasks (Figure 3).

Error

 

l
RecaH Recogmflon

Figure 3 Hypothesized Difference of Error Between Recall
and Recognition.

18

19

(b) Error will decrease from a non-specific to a
somewhat specific to a very specific instruc-
tional set (Figure 4).

Error

 

l l
NST SST VST

Speaficny

 

Figure 4 Hypothesized Difference of Error Between
Instructional Sets.

(c) Therefore:

(1) The greatest amount of error by the test
subjects will come from those who are given
a non-specific task requiring a reproduction
skill.

(2) The least amount of error by the test subjects
will come from those who are given a very
specific task requiring a recognition skill

(Figure 5).

20

 

 

 

 

 

 

 

RecaH Gmamm Enor (3 (4)
(5)
RecognMon (& (2) Lwfl er
(1)
NST SST VST

Figure 5 Hypothesized Progression of Error Terms.

11 (a) The stimulus pattern will always be more

consistently recognized than it will be

reproduced (Figure 6).

Consmtency

 

1
RecaH Recognmon

Figure 6 Hypothesized Differences in Consistency Between

Recall and Recognition.

21

(b) The consistency with which the stimulus pattern
is recognized or reproduced will increase as

the task becomes more specific (Figure 7).

Consmtency

 

l l l
NST SST VST

Specnmny

Figure 7 Hypothesized Differences in Consistency Between
Instructional Sets.

(c) Therefore:
(1) The test subjects will exhibit the least
consistency for a non—specific task requiring
a reproduction skill.
(2) The test subjects will exhibit the greatest
consistency for a very specific task

requiring a recognition skill (Figure 8).

22

 

 

 

 

 

 

 

Recall Least Consistency (2) (3)
H)
Recogrnfion (4) (5) GmamstConmmmmy
(W
NST SST vsr

Figure 8 Hypothesized Progression of Consistency.

These hypotheses may be addressed individually. Based
upon the results of Tversky (1973) and Loftus and Loftus (1976),
one would expect a higher response probability, or a lower
error, for all recognition tests than for any recall test.
This difference in error terms between recall and recognition
is probably due to the sheer difficulty of recalling and
reproducing a mapped circle pattern. And if the test map
is to be in any way realistic in appearance, the number of
circles will probably exceed the eight element limit that
Kaufman, 3; a1. (1949) defined as that which can be
efficiently recalled or recognized.

Concerning the instructional set, the conclusions of
Loftus and Loftus (1976) and Warrington and Ackroyd (1975)
indicate that the subject should perform significantly better

given a specific task than if they were given a non—specific

 

23

task. And Loftus and Loftus found that this difference was
much more pronounced in recall than in recognition tests.

The difference between either the NST or the VST and the SST,
however, may not be significant. Even though this difference
is not pronounced, the results, nonetheless, should show a
progression in the error terms as hypothesized.

The reader may note that the slope of the lines
illustrating the differences in mean error between recall
and recognition (Figure 4) is steeper than that between the
different levels of task orientation (Figure 5). Given the
above prior research, and due to the fact that reproduction
of a circle pattern is more difficult than the recognition
of that pattern, the different instructional levels will not
produce as much error as the two memory tasks.

In regard to the subjects' consistency of response
within each test group, one would expect an inverse
relationship to that of the error terms (Figures 7 & 8).

As with error, a significant difference in consistency
between recall and recognition should be evident in the data
analysis. A somewhat different reason may account for the
discrepancy in consistency, however. In a recall test,
given an infinite number of subjects, an infinite variety

of patterns could be reproduced; but in a recognition test,
the subjects are constrained to whatever choices the
researcher gives them.

Neither the memory task nor the instructional set appear

to have been quantitatively examined in terms of the subjects'

24

consistency. But a descriptive measure of consistency was
provided by DeLucia in his tests on task orientation: He
recorded the subjects' eye movements after reciting a
particular level of instruction; the narrowness of the area
scanned decreased as the specificity increased (1974, p. 239;
241). In other words, one may assume that the greater
consistency in eye movements over the stimulus map, the
greater the consistency of the test subjects in recognizing
or reproducing the target map. Finally, in following this
line of reasoning, like error, the slope of the lines
illustrating the differences of the subjects' consistency
is steeper between the levels of memory task as opposed to

that between the levels of instructional set.

Testing Methodology

 

Six different map oriented tasks were administered to
six different groups of sixteen people yielding a total of
96 subjects. The subjects consisted of a variety of under—
graduate and graduate geography students (majors and non—
majors) at Michigan State University. None of the subjects
were tested more than once. Given the range of students
tested, the sample population appeared to approximate the
"average” map reader, thereby reducing the effects of the
problems addressed by Downs and Stea (1977, pp. 100-03).

The stimulus map was designed for simplicity, employing
standard cartographic principles. The circle pattern on the

map consisted of a selected assortment of 20 different—sized,

25

non-overlapping circles. The map as such presents a pattern
that is not easy for the subjects to recall while the task
required is not impossible. Overlapping circles were not
used because the element of overlap introduces additional
unwanted variables into the test (GrOOp & Cole, 1978). Nor
were county borders, a north arrow or a scale used on the map.
Whereas county borders may provide a locational impetus, their
presence creates additional background noise——the influence
of which is hard to measure; on the other hand, the presence
of the latter two items, being merely incidental to the map's
message, adds only clutter to the map. However, a source

and a legend (adjusted to Flannery's constant) were present
on the map because not only are they commonplace but they
also provide necessary information to the reader.

The stimulus map used in the tests, which appears in
Figures 9a and 9b, covered an area and topic which presumably
most students from the state of Michigan would be unfamiliar
with, i e., ”State Park Attendance in Arkansas.” Supposedly,
a map of this sort reduces the effects of previous learning
or experience on the test results. Hence, only in those
areas of the map that the subject did not look at (pgppg

incognita), would one expect great variations in the components

 

of the pattern.
In each test, the subjects were given two envelopes:
A and B. First, they were told to remove the map from

envelope A. Then, one of the levels of instruction was

 

26

ANS 8385 88a. :88 of E was: as: magnum

om ohswflm

 

Aooo—xv
mco:m_> Co 52:52

8W:fl-. _ .-

 

    

Rmp .omcaE_< mmmcuxi Hoot—sow

 

ucomou

 

 

snap
mUZ<Dth< ¥m<m w._.<.rm
m<m2<¥m<

 

 

 

27

ANOH poospomv muon coauficwooom o£u ca pom: amz mDHDEHum no mhswflm

 

 

R9 .omcmE_< mmmcmxz 850m
800.. xv
weogw; .o 52:52

     

 

9:33

 

 

Raw
mQZ<DZwt< v. ¢<m wh<kw
w<wz<¥¢<

 

 

28

delivered to the subjects:

(1) NST -— ”Look at the map.”
(2) SST —— ”Look at the pattern of circles on the map."
(3) VST -- ”Study the pattern of circles on the map;

a) later, you will be asked to rank a set

of maps from most similar to least similar

in relation to the one that you are looking
at now (recognition) or (b) later, you will
be asked to reproduce the circle pattern that

you are looking at now (recall).”

After the initial instructions were read, the subjects
were given 30 seconds to View the stimulus map. At the end
of that time, they were told to ”put the map back in envelope
A and place it off to the side.” The final tasks were then
read to the subjects. Clarity apd expediency were required
of the testor at this point because, as Phillips noted, short
term memory decays rapidly after 10-20 seconds (1974, p. 284).
Granted, the reading of the final task may be deemed inter—
ference by Brown (1958) or Winklegren and Norman (1966), but
the instructions will be held constant for each of the recall
tests and each of the recognition tests and thus the verbal
instruction's effects should be uniform for each group.
Regardless, as Herriot noted, interference does not
seriously affect recall or recognition in the short term

memory store.

29

The final tasks themselves can be broken down into

two types:

Recall and Replicate

The subjects were instructed that "Envelope B contains
a blank map (Figure 10), a slip of paper with a set of
stick—on circles on it, and a knife. Please remove them
carefully. From what you remember of the map that you saw
earlier, very carefully remove all of the circles from the
piece of paper and arrange them on the map until the circle
pattern duplicates the pattern of the first map. You may
rearrange the circles at your discretion.”

The above instructions follow Brown's (1976) definition
of recall in that the testor asked his subjects to generate
a target(s) which matched the initial stimulus to the best
of his memory. The response accuracy depended upon how
complete the subject's information (determined in part by
the instructional set) was of the target stimulus. Once the
map is complete, the subject then decides whether or not he
recognizes the pattern; if not, he may then rearrange some
or all of the circles. Figure 11 illustrates a reconstructed
pattern produced by one of the subjects in VST.

The use of stick-on circles in this study is an outgrowth
of the work conducted by Steinke (1975). In his test,
Steinke presented his subjects with a stimulus map containing

39 circles divided into seven classes. Later, the subjects

30

.Axon woosemmv momma Hamoom «so as some as: xemHm OH spawns

 

R9 .omcmE.< £8:me ”outaom

82: 5
Eozfl>eotooEaz

 

   

 

cacao;

 

Raw
mOZ<Dth< ¥m<m m._.<._.m
w<m2<¥m<

 

 

 

31

.ANOH poospomv

Hm> ofiu Ca qfi uoomflsm new: poosponaom m mo chmem G< HH ohswflm

 

809 5

 

Sill:

tcomod

 

9.0wa Lo conEsz

the .omcmEZ £3:me “8.16m

 

 

 

hhmw
wOZ<DZm._L.< ¥m<m mk<hm
m<m2<¥m<

 

 

32

were asked to reproduce the pattern to the best of their memory
using a set of adhesive-backed circles. Unfortunately, he
encountered a problem in evaluating the reproduced patterns
because he had let the subjects use more circles than were

in the stimulus pattern. This procedure may have inadvertently
allowed his subjects to inject circles into the pattern that
were not originally present. Because different numbers and
sizes of circles may have been used on the target maps as
opposed to those circles on the stimulus map, a quantitative
comparison between the two was difficult to make. A constraint
was therefore imposed on the subjects in the present test,
i.e., the same number and sizes of circles were placed on

the slip of paper in envelope B as were on the stimulus map.

In this way, the ”movement” of each circle and the pattern

as a whole could be quantitatively measured.

Recognition

The subjects were instructed that "envelope B contains
a set of eight maps similar to the one that you just saw,
plus a slip of paper numbered 1—8 (Figures 12a through 12h
and 13, respectively). Please remove them. From what you
remember of the map you saw earlier, arrange the maps from
most similar to least similar. Record the symbols located
in the lower right hand corner of the maps in the appropriate

blanks on the slip of paper.”

 

33

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

  

Legend

 

_____ 100
Number of Visitors
(x 1000)

 

Source Arkansas Almanac. 1977

 

 

 

ARKANSAS
STATE PAR K ATTENDANCE
1977

 

100
Number 01 Visitors
(x1000)

Source Arkansas Almanac. 1977

 

 

 

Tests (Reduced 10%).

/b

Figure 12 Distractor Stimuli Used in the Recognition

 

34

 

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

 

Legend

   

 

00
Number of Visitors
(x1000)

Source Arkansas Almanac,1977

 

 

 

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

1
Number of Visitors
(x100m

Source Arkansas Almanac. 1977

 

 

 

35

 

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

  
 
 

Legend

    

1

Number oquisitors
(x1000)

Source Arkansas Almanac 1977

 

 

 

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

 

00
Number of Visitors
(x1000)

Source Arkansas Almanac. 1977

 

 

 

36

 

 

ARKANSAS
STATE PARK ATTENDANCE
1977

 

 

 

00
Number of Visitors
(x1000)

Source Arkansas Almanac, 1977

 

 

 

 

ARKANSAS
STATE PAR K ATTENDANCE
1977

 

Legend

 

100
Number of Visitors
(x1000)

Source: Arkansas Almanac. 1977

 

 

37

1. Most similar

8. Least similar

Figure 13 Slip of Paper Used for Ranking in the
Recognition Tests.

Search Search hAennory (possible rehearsan
Coniponenls Choose Panern
1 Recall
lsthe Panern Recognued?———-No
.. Processes

Recogmnon

Yes
Coniponents 1

Output

 

Figure 14 Mental Processes During Recall and Recognition
(After Loftus and Loftus, 1976).

 

38

These directions were similar to Brown's (1976) and
Loftus and Loftus' (1976) definitions of recognition in
which one or more potential targets were presented to the
subject who searched his memory, selected a pattern, and
then decided whether or not that pattern most closely
resembled the initial stimulus that he remembered (Figure 14).
The response accuracy depended upon whether or not the

II

subjects had as much information (determined by the
instructional set) as is necessary to discriminate the
target stimulus from the distractor stimuli (Brown, p. 87)."
This type of test involves two additional subtasks on
the subject's part: First, the subject conducts a
discrimination task where he ”judges whether the stimulus
pattern seen is different from some other one or from some

1

other set of patterns.‘ Next, a judgmental task is performed
where the subject ”is required to assign the presented pattern(s)
some value on a scale of judgment, such as size, complexity,
attention value” or similarity (Hake, 1966, p. 146).

Since a rank-ordering of the patterns took place, two
decisions were made before the test was given: (1) Symbols,
and not numbers of letters, were used to prevent a biased
rank-ordering of the pattern. (2) Because Neisser (1967)
had found that test subjects did a faster and better job
of recognition for dissimilar target arrays than for similar
target arrays, eight similar target arrays were presented.

While the use of eight similar patterns probably increases

 

39

the distances between mean error terms for recognition and
recall tests, one should expect that said presentation is
a more realistically difficult assignment than the use of
dissimilar patterns.

At this point, one must also note that there are two
features in recognition tests which are likely to encourage
the use of recall:

(a) if the recognition test requires the subject to
search for a single target amongst many choice
items (in this case, eight), his best strategy
may be to recall the target first (curing the
search memory phase of Figure 14) and then to
search for it.

(b) in a short term memory experiment, the subject
may still be rehearsing the material (also
during the search memory phase) at the time the
test was administered. Since rehearsal involves
repeated recall, recall inevitably mediates
recognition in the test (Brown, 1976, pp. 2—3).

But rehearsal is dependent upon the specificity of the
instructions, i.e., the more specific the task, the more
likely active rehearsal will occur.

At the conclusion of each test, the subjects were asked
informally, ”Did you remember any circles in particular?”
Most of the subjects indicated that they remembered the
locations of the larger circles and groups of circles better
than smaller and isolated circles. They were also asked,
”Did you remember a shape to the pattern?” This question
elicited a variety of responses including the shape of a C,

an obtuse triangle, a hook, and so forth. The responses to

the first question qualitatively verifies the impression that

40

relatively large amounts of blackness attract the eye while
the responses to the second question merely indicates the
variety of memory patterns (organization) present in

different people.

Chapter III
Data Analysis

Introduction

In order to quantitatively evaluate the results, the
x,y coordinates of all the circles in each pattern reproduced
in the recall test or chosen in the recognition test, plus
the stimulus pattern, were plotted on graph paper. Since
the maps in the recognition tests were smaller than those in
the recall tests, the x,y grids were adjusted accordingly
so that relevant comparisons could be made between the groups.
From this base, a series of statistical analyses were
conducted in order that the data could be meaningfully
interpreted.

These analyses included: (1) frequency of pattern
choice in the recognition test; (2) the mean squared distance
between the point locations of the stimulus circles and the
chosen or reproduced circles in both tests; (3) the difference
between homologous circles after adjusting for orientation;
(4) the mean deviation of the percent black per x,y quadrant;
(5) the use of the circular normal distribution to obtain
mean distance from the origin and the mean angular deviations;
(6) a weighting factor (area of the circles) applied to 5;

(7) centroids of the patterns and their associated vectors
from the stimulus centroid plotted; and (8) a weighting

41

 

 

42

factor applied to each of the centroids in 7.

A logical question at this point of the study might be:
Why use all of the aforementioned statistical procedures?
The answer is simply because no one statistic adequately
explains all of the variations in the data. While all are
related, each analysis describes a different aspect of the
data; thus, each is used to compliment one another so that
any conclusions that are made have sufficient quantitative
backing. A discussion of the use of these techniques and
interpretations of the results follows.

Before proceeding, a brief examination of the hypotheses
is in order. If one attaches significance levels to the
difference in means for each group using a two-tailed Student's

3 test, then one might realistically formulate the following

specific hypotheses: (1) for all test groups, one can predict
a significance level of a = .01 (t = 2.75) between the mean
error terms of recall and recognition. (2) Based upon

earlier research and given the instructions that were read

to the subjects, there should not be a significant difference
(although a progression should exist) between the mean error
terms of the non-specific and somewhat specific and between
the somewhat specific and very specific instructions. (3)
But a significance level of a = .05 (t = 2.042) should exist
between the non—specific and the very specific instructional

sets .

43

Frequency of Choices--Recognition

First, an examination of the relative rank that the
subjects assigned to the stimulus map and the distractor
stimuli is now in order. This particular measure was
employed to assess the frequency of responses for each of the
test maps. Figures 15a, 15b and 15c show that the stimulus
map was ranked no ”worse" than third by any of the three
instructional groups. Clearly evident is that very little

consistency exists within that range for the NST group.

16 16 16
14 14 14
> 12 12 12
S 10 10 10
a;
g 8 8 8
If; 6 6 6
4 4 4
2 2 2
1 2 3 1 2 3 1 2 3
Rank (NST) Rank (SST) Rank (VST)
a b c

Figure 15 Rankings for the Stimulus Map.

Comparing Figures 15a and 15b, one notes that as the
task becomes more specific, the number of subjects assigning
a first order ranking to the stimulus map increased at the
expense of the third order rankings. Further, a distinctive
downward trend now appears for the SST group as opposed to
the rather even spread of rankings for the stimulus map

within the NST group. These rank frequencies indicate that

 

44

not only do somewhat specific instructions generate more
consistent results than do non—specific instructions, but
they also convey the idea that progressively fewer people
assign the stimulus map to the less similar rankings.

Such is not the case when comparing Figures 15b and 15c.
Whereas the frequency of first order choices in the VST is
greater than that of the SST, the difference between the two
is not as great as expected; in fact, it is less than the
difference between NST and SST. And the progression through
the rankings does not get successively lower for VST.

Before postulating as to the reasons why, one should
look at Figures 16a, 16b and 16c which illustrate the first
order rankings for all maps. No more than five of the eight
maps given were chosen in all of the test groups (Figures 12a-
12e). One may assume that the patterns of the three maps not
chosen must be, in the visual sense, sufficiently different
from the stimulus map. Conversely, one might also assume
that the patterns of the maps that were assigned to the first
order are visually ”close” to the stimulus map pattern. The
term "visual” qualifies the last two statements because, as
discussed later, visual rankings are not necessarily

equivalent to mathematical rankings.

45

16 16 16
14 14 14
> 12 12 12
g 10 10 10
‘5’ 8 8 8
E 5 6 6
4 4 4
2 2 2
< l x A» * < l m < x A
NST SST VST
a b c

Figure 16 First Order Rankings for all Maps.

Figure 16 illustrates a variety of responses (non—
consistent) to NST. On the other hand, in response to SST
and VST, not only do a greater number of subjects choose the
stimulus map over all others combined, but fewer of the
distractor stimuli are chosen for these two tasks as well.
Like Figure 15, however, Figure 16 also shows a greater
difference in responses between NST and SST than between
SST and VST. One might expect, though, that the error
terms calculated for VST are lower than those for SST. But
the patterns chosen by the subjects in VST dictate otherwise.
Because two of the subjects in VST chose the map symbolized
with a triangle (Figure 12d), most statistical measures
will calculate the error in that group to be greater than
the error in SST. In other words, the progression from
high to low mean error terms should be NST-VST—SST,
respectively. Had the test sample been larger, the expected

progression of error terms might have materialized. These

46

are important points to remember later on when the output

of the other statistical procedures are analysed.

Mean Deviation—~Percent Black Per Quadrant

 

In this particular statistical test, an x,y grid was
laid down on top of each of the reproduced and chosen maps
and the percent black per quadrant was calculated for all
maps. The absolute deviation of each quadrant of the maps
was derived from the difference between the percent black

perquadrant of the stimulus map and all other maps (e g.,

mean deviation of quadrant 1: Q1 = /E(X81_Xil/ ) The
/ l6 /
mean absolute deviation was then calculated for each test
B .
group (—£%£Z), In effect, what this statistic measures is

only the percent black per quadrant; it tells us little
about the positions of the individual circles or the patterns.
Since a number of small circles in one quadrant can equal
one large circle in another quadrant, the interchange of the
two would probably result in little or no change in the
respective deviations. While a reproduced pattern may look
very different in comparison to the stimulus map, the percent
black per quadrant could conceivably be equal.

The data for each quadrant indicates where the average
map reader in a particular test group remembers clustering
or blackness to occur and the statistic itself reflects that
memory, i.e., the larger the mean deviation, the less the

subject remembers about the overall pattern. First, Table 2

 

 

 

 

47

illustrates that while both the absolute total and mean
deviations are less for recognition than for recall, only
relatively small variations from the stimulus map occur in

any one quadrant for any one test. Overall, Table 2 indicates
that the subjects approximately remember the relative amounts
of black in each quadrant, regardless of the instructional
set. The absolute mean deviation (in parentheses) points out
the fluctuations among the different task levels for each
quadrant. These fluctuations are rather hard to interpret;
however if one examines Figure 17, the overall error for each
instructional set is clearly outlined. As expected, the
differences between NST and SST for both recall and recognition
are less than the differences between SST and VST. But the
difference of means test revealed that NST and VST for both
recall and recognition are not significant at the a = .05
level, although the difference between recall and recognition

are significant at the .01 level.

 

48

Table 2 Absolute Total and (Mean) Deviations

Quadrant 1 2 3 4
Stimulus 17.7 7.1 33.9 31.0
Recall NST 22.2 (5.74) 9.6 (5.14) 26.9 (7.86) 31.7 (3.24)
SST 20.9 (5.33) 8.4 (3.76) 30.2 (5.46) 30.4 (6.97)
VST 19.7 (3.98) 8.9 (4.47) 34.9 (2.61) 27.0 (5.90)
Recog. NST 16.8 (1.80) 8.7 (1.70) 33.1 (1 80) 32.9 (2.15)
SST 18.5 (1.06) 7.3 (0.90) 32.1 (1.80) 33.2 (2.20)
VST 16.4 (1.26) 8.2 (1.16) 34.2 (0.60) 31.6 (0.70)
6.0 —
‘5 U
'5: S 38 —
a g t
c ‘35 2.4 —
8 m _
2 °\° 12 _ Necognition

 

 

1 l l
NST ssr VST
Spedficny

Figure 17 Mean Absolute Deviations of the Percent Black
Per Quadrant.

49

Mean Squared Distance

 

Two related measures of error were on techniques
suggested by Sneath (1967). Those measures were used to
produce similarity indices, based upon the movement of the
individual circles, between the resultant patterns and the
stimulus. More specifically, the first of these techniques
measured the sum of the squared distances between the circles
in the stimulus pattern and the respective circles in the
recognized or recalled patterns. In essence, this program
performed a comparison between those patterns and the stimulus
pattern and yielded the sum of the squared distances which
are interpreted as dissimilarity indices. The mean index for
each test group was then calculated thereby producing a
measure of similarity between the test groups.

The large differences in error terms between the recall
and recognition groups can be easily seen in Figure 18. As
hypothesized, the difference between NST and SST is less than
that between SST and VST for recall. With recognition, on
the other hand, VST registered a higher degree of error than
SST for reasons already explained. A significant difference
is present between the means of recall and recognition at
the .01 level. NST and VST exhibited a significant difference
for recall at the .1 level while no significant difference
could be detected between any of the instructional sets for

recognition.

50

— T—T“\\\jmw
8—

Mean Squared Distance
Before Rotaflon (x1000)
l

2 — .\\\\\~—"’——Recogmfion

1 1 1L,

NST SST VST
Spedficny

 

 

Figure 18 Test Group Values of the Mean Squared Distance
Before Rotation.

The second of these techniques was attempted to bypass
several difficulties associated with the above measure and
other analyses conducted by Jenks (1975) and Steinke (1975).
In both of these studies, subjects were asked to visually
compare mapped circle patterns and to assign a value of
similarity to them (1—7 and 1-5, very dissimilar to very
similar, respectively). While their results may have been
overtly subjective, they provided a basis for comparison
between perceptual and mathematical pattern similarity.
Jenks also tried to evaluate the similarities between circle
patterns by means of a correlation grid. Unfortunately, the
correlation between the stimulus pattern and the resultant
pattern was either very low or zero. This problem may also

exist to a lesser extent in the analysis conducted below.

51

In an attempt to solve this problem, while at the same
time maintaining an objective measure of pattern comparison,
the following steps were performed: the x,y locations of
all the circles, stimulus and resultant, were transformed
into standard deviation units. Each resultant pattern was
then rotated over the stimulus pattern until the ”best fit"
between the patterns was achieved as measured by the least
squared distances between homologous circles. The mean
deviation of each group was also calculated to compare test
groups.

While compensating for minor rotational errors on the
subjects' reconstructed patterns, this technique introduced
a different problem: it rotated the patterns and examined
them out of the context of the map. For example, if a
subject had chosen the map symbolized by a star (Figure 12h)
as the most similar pattern, the index between it and the
stimulus pattern was similar to the stimulus pattern (with
minor variations when rotated 180 degrees). In other words,
this technique overcompensates for the subjects' mental
errors in orientation and would subsequently assign a
similarity index between patterns that may be significantly
different from an index created by a visual comparison of
the patterns. That problem aside, when the mean squared
distances after rotation were plotted for each test group
(Figure 19), surprisingly, the shape of the plots was
basically the same as those on Figure 18. The similarity

between Figures 18 and 19 may indicate that the average test

 

52

subject performed very little mental rotation of the

stimulus pattern.

40 -
a) _
Q
g 32 '— \Eeca”
‘27; c ._
5 9
1:: E 24 '—
93 o
m o:
g 5 16a
m t _
8 8 — Recognition
2 ’- \/

 

 

l l l
NST SST VST
Specfimny

Figure 19 Test Group Values of the Mean Squared Distance
After Rotation.

A significant difference at the .01 level was found
between recall and recognition. The relative significant
difference between NST and VST for recall was strengthened,
in comparison to the mean squared distance before rotation,
in that their difference was significant at the .05 level.
No significant differences were noted between any of the

recognition tasks.

Circular Normal Distribution

 

Based on research conducted by Reyment (1971) and
Mardia (1972), several measures of positional error of the
reproduced or recognized circles were made. The goal of

both of the previous studies was to determine whether or not

53

the data contained in a circular distribution was normal.
Of primary interest here, however, are the statistics used
to arrive at a determination of normality: the distance
from the origin which reveals how close to the stimulus
the circles in the resultant patterns come, and the angular
deviation which indicates a test group's angular consistency
in the positioning and locating of the circles. The
distance (r) from the origin (the point of origin is
arbitrary, but it must be the same for all circles) is
found by the following formula:

r = (22 + Y2) where R = E (COS ai) and Y = E (sin ai)

N N

(Reyment, 1971, p. 23). This statistic establishes the length
of the mean vector for a circle. The length is a unit
measurement, i.e., as E approaches 1.0, the mean circle
location approaches that of the stimulus. The mean vectors
for the patterns were also calculated and plotted in
Figure 20, illustrating the differences between recognition
and recall and between the instructional sets for E- This
graph also points out the inverse relationship between E
and error.

The mean angular deviation(s) is essentially the
standard deviation of a test group measured along the
circumference of a circle instead of a straight line. It
is defined as: s = 2 (l—r) (Reyment, 1971, p. 27). This
computation yields the value of g in radians. By converting

radians to degrees, one arrives at the mean angular deviation,

54

1.0
E /\Recognition
.8 —
a, .6~
g '- R ll
cu ._ 8C8
2 4
.2 -

 

 

1 1 1
NST SST VST
Spedficny

Figure 20 Test Group Values of the Mean Distance
From the Origin (r).

i.e.. the angle which defines one standard deviation unit to
either side of the mean vector. As p approaches zero, the
mean circle location approaches that of the stimulus. The
mean angular deviation for all circles in each test group
was also computed and plotted in Figure 21 which illustrates
the consistency of each group's ability to recall or
recognize the stimulus pattern.

As anticipated, the mean differences between the
recall and recognition test groups for both E and g were
significant at the .01 level. But no significant difference
at the .05 level could be found between any of the instruc-
tional sets in either memory group for both E and p.
The relationships between the mean vectors and the spread

of the observations within one standard deviation unit

55

80c
,___‘___fmm
64~

(S)

 

 

48—
: _
8 32.. Recognition
2 ’— \/.
16—
1 a 1
NST SST VST
Specificity

Figure 21 Test Group Values of the Mean Angular
Deviation (s).

underlying those vectors are shown in Figures 22a through
22f. The mean directions of the vectors were determined from
the mean x,y positions. Again, the differences between

the instructional levels is not as noticeable as that

between recall and recognition.

56

.mmsouo omoH Ham pom scamme>om swaswc< coo: osu
pcw GHwHHo emu Eosm oocmumflo Cmoz map mo mCOHuMDComouaom oflfimmpo NN ouswflm

hw> kmw hmz

 

ZOFEOOUmm

no

44<Umm

 

57

A weighting factor was also applied to E and g;

the formulas were the same as the above except now:

X = E(cos a - w) and Y = E(sin a ~ v0
Ew Ew

where w equals the area of a circle. The rationale for
using a weighting system of this sort has been set forth

in eye movement research (Steinke, 1975; Dobson, 1977)
which showed the greater visual ”importance" of large black
figures (circles) on a map versus smaller figures. And the
subjects in the present study verbally indicated a better
memory of the larger circles as opposed to the smaller ones.
This concept is supported by Figures 23 and 24. These two
figures graphically represent the mean 3 and p values,
respectively, for each circle in the recall test before

the weighting factor was added. Note that error increases
dramatically at first, and then levels off as circle size
decreases.

The use of the weighting factor produced unexpected
results (Figures 25 and 26) where the weighted E and g
values for the test groups, while not being mirror images of,
are nonetheless the reverse of those seen in Figures 20 and
21. But the reasons for this reversal are not very apparent

and do not appear to be forthcoming.

Centroids
Another measure of positional error was calculated for

the response patterns in the tests. Unlike most of the

Mean (0

58

 

 

1.0-

.8-

.6»

4— ' ,'

.2» ' ' . '~" °
11111111111111111111
1 5 10 15 20
lamem smflhst

Chde She

Figure 23 Mean Distance From
the Origin for Each Circle in
the Recall Tests.

(r) Weighted

hAean

_a

 

 

0 r—
.8 -

— New“
.6 _

_ \I‘jecognition
4 —
.2

1 1 1
NST SST VST
Spedficny

Figure 25 Weighted (r)
Values for All Test Groups.

 

 

80-
72- . . ' :.. °.
$64” .0
r: - ° '
CU
saw .
48c
401111111111111111111L
5 10 15 20
Ewen . ’ smdbm
Cnde Sue

Figure 24 Mean Angular

Deviation for Each Circle

in the Recall Tests.

 

 

80 —
E _
E 64 r Recognition
‘5' * "—7—.————h n
; 48E em
51 32 _
C _
g 16
1 1 1
NST SST VST
Spedficny
Figure 26 Weighted (5)

Values for All Test Groups.

 

59

other statistical analyses used, centroids are closer to
defining error for an entire pattern rather than the
accumulation of error terms for individual circles. This
statistic essentially defines the mean center of the response

patterns in relation to that of the stimulus pattern:

Ex By
20 , 20 . Like 5 and g, a weighting factor (area of a
circle) was also computed: Exw Exw where w equals

 

Ew , Ew
the area of a circle. The response centroid and weighted
centroid locations and their distances from the stimulus
centroid are presented in figures 27a through 1.

The centroids plotted for the recognition tests
indicate the number of choices for each location. Because
of the limited number of choices made in the recognition
tests, these vectors do not appear to show as much of an
orientation as do the locations of the patterns in the
recall test groups. And in viewing the unweighted recall
vectors, the reader may note that while the VST vectors are
closer to the stimulus than either the NST or SST vectors,
a more unimodal distribution of the centroids appears in
the NST. When the weighting factor is added on, the
orientation of the vectors shifts counterclockwise, from
beneath the stimulus to the right of it. Such a shift may
indicate that the larger circles, which were primarily
concentrated on the left and upper sides of the stimulus
map, were positioned by the subjects towards the center and

the right side of their maps.

6O

.AHH x Goauflcwoowm an x Hamummv maDOHU umoH Ham How
pflowuceo wDHDEHum esp ou cowumaem Ca mCOHONUOJ pHOHucoo wmcoowmm mm whswwh

\

44<Umm.n ZOELZOOUmm m c

PmZ

61

.3<Umm ,U

Hmm

 

ZO:._ZOOUME .o

62

j<omm g

  

ZOF_ZOOQmm o

Fm>

63

   

.:<owm .r. ZO_P_ZOOUmm .m

82995
.52

 

64

uu<omm._

 

 

ZOE._ZOOUmm

poEQoB
Hmm

m

o—

65

44<Umm

ZO_.EZOOomm .v.

\\

poEmBB
Fw>

m—

66

The values of the mean vector lengths of the unweighted
and weighted test groups are plotted in Figures 28 and 29,
respectively. As predicted, the difference between NST
and SST is less than that between SST and VST for recall.
This statistic also shows that the recall error in VST is
sufficiently less than NST to be significant at the .05
level. Recognition, on the other hand, again exhibited a
tendency for the error terms to decrease from NST to SST
and then rise from SST to VST, although the difference at
the .05 level could be detected between any of the
recognition test groups while recall and recognition did

differ significantly at the .01 level.

24 — \Recan

_ RecognMon
1 'T“'l
NST SST VST
Specificity

Mean Vector Length

 

Figure 28 Mean Vector Lengths for All Test Groups.

67

 

 

40-
)-—
.C
:53» 3.2 -
3 ’6 *- Recall
0.:
g E? b
> fa: 1.6 -
c: F
8 8 _ . .
2' \Becognmon
1.
1 1 1
NST SST VST

Spedficny

Figure 29 Weighted Mean Vector Lengths for All Test Groups.

The addition of a weighting factor affected the recall
mean vector lengths in much the same way that the weighted
p and g terms behaved, i.e., the progression of error
terms from NST to VST are reversed from the unweighted terms.
Surprisingly, the weighted error terms for recognition did
not behave in the same fashion. In fact, the progression
resembles that for the mean deviation of the percent black per
quadrant (Figure 17). The relationship between the two
slopes could mean that since the larger circles are given
greater weight in the location of a pattern centroid, their
influence also generates a greater percentage of black for a
particular quadrant; and since, in VST, the stimulus pattern
is chosen more often than in either NST or SST, the ”pull"

of the larger circles to the stimulus centroid is that much

stronger. But no significant difference for either the

68

weighted recall or recognition test groups could be seen
at the .05 level. As in the other analyses, the two memory

tasks differ significantly at the .01 level.

Summary

Table 3 summarizes the relationship between error and
consistency for the recall and recognition test groups given
the above statistical analyses. All of the tests demonstrated
that the error and consistency in recall and recognition are
significantly different at the .01 confidence level. The
difference between the instructional sets, as expected, was
not as great, and in most cases was not statistically
significant. Between NST and VST, however, a significant
difference can be noted at the .05 level for the mean squared
distance after rotation and the mean vector lengths and at
the .1 level for the mean squared distance before rotation
while no significant differences were found between any of the
recognition tasks by any of the above statistical procedures.
Due to the choices made by two of the subjects in the very
specific recognition task, all of the statistical procedures,
with the exceptions of the frequency of choices, the mean
deviation of the percent black per quadrant and the weighted
centroids, showed VST to possess a larger degree of error
and a smaller degree of consistency than SST. But the three
recall tasks exhibited either a nearly linear decrease in
error terms (mean squared distance after rotation, mean 3)

or an accelerated decrease in error terms from NST to SST to

69

VST (mean squared distance before rotation, mean deviation
of the percent black per quadrant, mean vector lengths).
Consistency, on the other hand, decreased linearly

(mean g) or at a decelerated rate (frequency of choices)

from NST to SST to VST.

 

70

NS.
mm.

mm
News.
m.mN
mmmw.
mm.m
N.HQ¢H
new.

Hm>

.Apou5wflo3v Luwcoa Hoboo> coo:
.LuwcoH Hoboo> cmoz
.ApouszoBV coaumw>op smaswcm cmoz
.ApoustoBV cHwHHo ofiu Eonm mucmumflp cwoz
.cowumw>oprumaswcm cmoz
.cwwwno can Eonm oocmumflp Gums
.cOAOmuou Houwm moamumwp poemsvm new:
.COAOmuou ouomon mocmumwp poemsvm awe:
.ucmupmsw Hum xoman ufioouoa can mo SOAOMH>op cmoz
mm. mm. mq.m mm.m
mm. we. om.H ow.N
Ho mm as we
nwmq. some. came. Ammo.
©.¢N N.¢m H.mo m.mo
mmmm. omqw. scum. memm.
wm.m eq.m qw.mm mm.Hm
q.mHm q.oowH H.mawm m.mmmw
qu.H Nmm.H omm.q Hom.m
Hmm Hmz Hm> Hmm
Cowuwflwoomm Hamomm

MH.N
mo.m
Ne
mumm.
o.wo
mowm.
wm.mm
m.mnmw
oom.m

Hmz

 

wmsouw umoH Ham Mom mEuoH Honpm Geo: oLu wo oHan mpmaabm

m

seams

Chapter IV

Discussion

Conclusions

Error

Several questions should be readdressed at this time:
First, did the two different memory tasks of recall and
recognition produce significantly different results? Regardless
of the statistical measure used, there was a highly significant
difference between recall and recognition tasks: recall
always produced greater response error than recognition.
Two basic reasons account for this consistently large
difference in error: (1) In order to make the recognition
test sufficiently difficult, the distractor map patterns had
to be constructed which were similar to the stimulus pattern;
so the error of the recognition tasks was expected to be
small whereas the subjects in the recall tasks could possibly
create a map having close to 100 percent error. (2) The
variation in recognition error was limited to the eight maps
presented as potential targets; in any recall task, on the
other hand, the variance is unlimited because an infinite
number of subjects conceivably could create an infinite
variety of patterns, given the confines of the map.

Second, did changes of specificity in the instructions

create significantly different map reader responses? While
71

72

the statistical analysis has proven that no significant
difference in the mean error terms exists between NST and
SST or between SST and VST, a definite progression from
high to low error does exist from NST to SST to VST. One
may also assert that there is a significant difference in
mean error terms between NST and VST for the recall test
groups. But due to the problems concerning the one map
that was chosen by two of the subjects, no empirical state—
ment, strong or weak, will be recorded about the difference
of error between the recognition test groups.

Based upon the results obtained in the data analysis,
hypothesis Ia, error will be less for recognition tasks than
for recall tasks, is confirmed. And while the error terms
for each level of instruction may not be significantly
different, a trend does exist which supports hypothesis Ib:
error will decrease from a non—specific to a somewhat specific

to a very specific instructional set.

Consistency

Finally, did the subjects consistently remember the
circle patterns to be a particular shape or in a particular
position on the map? From those who responded, apparently
a variety of shapes were consciously or subconsciously
utilized by the subjects as an aid in remembering the pattern
of circles. From the vectors plotted in the circular
distributions (Figure 22) and from the vectors plotted to

the centroids (Figure 29), one might note that the pattern

73

locations, while probably not being significantly consistent,
are at least relatively consistent within a given area.

Most of the circles in each pattern were congregated on the
left side of the map-—where the greatest amount of blackness
was present on the stimulus map. Obviously, the mental
coding and organization of the stimulus pattern varied

from subject to subject while the regurgitation of that
pattern was somewhat consistent within a given test group
because the mean vectors indicate that the level of
specificity does indeed partially influence the position of
the pattern.

Given the frequency of responses in the recognition
tests, the mean angular deviations and the centroid locations
of the patterns for all test groups, one should note that
a trend does exist which supports hypotheses IIa: the
stimulus pattern will always be more consistently recognized
than it will be reproduced; and 11b: the consistency with
which the stimulus pattern is recognized or reproduced will

increase as the task becomes more specific.

Recommendations
Given the variables considered in this study, one may
put forth the following guidelines for cartographic psycho—

physical testing methodology: (1) Even though a strong

 

case cannot be made on the differences between task orientations,
a trend did exist such that one may state that the instructions

in any testing procedure should be as specific as possible.

 

 

74

Indirectly, one might assume that clarity, conciseness and
consistency should also be maintained because otherwise,
when the tasks are read to all segments of the test
population, the level of specificity will either differ
from subject to subject or from sample to sample. Thus,
inasmuch as a very specific task is required to prevent
biased results, it should not be written or verbalized to
theextreme that clarity is sacrificed for it.

(2) Although recall and recognition are rarely

 

intentionally used together in any psychophysical test,

 

one should be aware of the exact type or combination of

 

memory response(s) being elicited from the subjects on a

 

particular test. Even though some degree of overlap is
present between recall and recognition, they are still two
different memory responses and the variables that affect
them react accordingly. And while the use of recognition

is greater than recall in cartographic psychophysical tests,
the latter may have greater applicability to everyday map
usage; but recall is a much harder response to study given

the difficulty of the task for a subject.

Further Research

Additional testing should be conducted on the types of
memory response and instructional set now present in testing
methodologies. Concerning this topic alone, the following
suggestions are put forth: (1) the sample population

should be large enough so that individual variations in

75

the responses do not significantly affect the variations

in the mean response; (2) the use of eye movement technology
as reported by DeLucia (1974) and Steinke (1979) would
present important quantitative and qualitative additions

to the measures already used in this study; (3) additional
variables, i e., the complexity of the test maps,
environmental perception, exposure time, and so forth,

could be added into the testing procedure to see what, if
any, effects their presence would have on map reader responses.
In other words, while recommendations have been made based
upon the research conducted within this study, the

evaluation of cartographic testing methodology is not
complete; and until the effects of all test situation
variables are examined, cartographers should exercise caution

in both their tests and in the conclusions that they reach.

 

 

76

BIBLIOGRAPHY

 

Anastasi, Anne. Psychological Testing. New York: Macmillan,
1976.

Archer. E. James. "Identification of Visual Patterns as a
Function of Information Load,” Journal of Experimental
Psychology, Vol. 48, No. 5 (November 1954), pp. 313—17.

 

Attneave, Fred and Malcolm Arnoult. "The Quantitative Study
of Shape and Pattern Recognition,” in Uhr, Leonard, ed.,
Pattern Recognition. New York: John Wiley & Sons, 1966.

 

Attneave, Fred. ”Symmetry, Information and Memory for
Patterns,” American Journal of Psychology, Vol. 68,
No. 2 (June 1955), pp. 209—22.

 

Baddeley, Alan D. The Psychology of Memory. New York:
Basic Books, Inc., 1976.

 

 

Blumenthal, Arthur L. The Process of Cogpition. Englewood
Cliffs, New Jersey: Prentice-Hall, Inc., 1977.

Brown, John. ”Some Tests of the Ceday Theory of Immediate
Memory,” Quarterly Journal of Experimental Psychology,
Vol. 10, No. l (1958), PP- 12-21.

 

. "An Analysis of Recognition and Recall and the
Problems of their Comparison," in Brown, John, ed.,
Recall and Recognition. New York: John Wiley & Sons,
1976.

Castner, Henry. ”The Role of Pattern in the Visual Perception
of Graded Dot Symbols in Cartography,” unpublished Ph.D.
dissertation. Madison, Wisconsin: University of Wisconsin.
1964.

Castner, Henry and Arthur Robinson. ”Dot Area Symbols in
Cartography: the Influence of Pattern on Their
Perception,” American Congress on Surveying and Mapping,
Technical Monograph #CA—4, 1969.

Cooper, Aldwin and Andrew Monk. "Learning for Recall and
Learning for Recognition," in Brown, John, ed., Recall
and Recognition. New York: John Wiley & Sons, 1976.

DeLucia, Alan. "The Nap Interpretation Process: Its
Observation and Analysis Through the Technique of Eye
Movement Recording," unpublished Ph.D. dissertation,
University of Washington, 1974.

 

77

Dobson, Michael W. "The Map in the Mind's Eye,” Auto-Carto II,
(1975), PP. 225-32.

 

”The Adaption of an Eye Movement Recording
Technique to a Cartographic Experiment Involving the
Graduated Symbol," unpublished Ph D. dissertation,
University of Kansas, 1977.

 

Dodwell, Peter C. Visual Pattern Recognition. New York:
Holt, Reinhart and Winston, Inc., 1970.

 

Downs, Roger M and David Stea, eds. Image and Environment:
Cognitive Mgpping and Spatial Behavior. Chicago:
Aldine Publishing Co., 19731

 

 

Downs, Roger and David Stea. Maps in Minds: Reflections
on Cognitive Mapping. New YOfk: Harper & Row,’l977.

 

 

Eagle, Morris and Eli Leiter. "Recall and Recognition in
Incidental and Intentional Learning,” Journal of
Experimental Psychology, Vol. 68, No. 1 (January 1964),
pp. 58463.

 

 

Fitts, Paul, Meyer Weinstein, Maurice Rappaport, Nancy
Anderson and J. Alfred Leonard. ”Stimulus Correlation
of Visual Pattern Recognition; A Probability Approach,”
Journal of Experimental Psychology, Vol. 51, No. 1
(January 1956), pp. 1—11.

 

Flannery, James J. "The Graduated Circle: A Description
Analysis and Evaluation of a Quantitative Map Symbol,"

unpublished Ph.D. dissertation, University of Wisconsin,
1956.

French, Robert J. ”Identification of Dot Patterns from
Memory as a Function of Complexity," Journal of
Experimental Psychology, Vol. 47, No. 1 (January 1954),
pp. 22-267

 

 

"Pattern Recognition in the Presence of
Visual Noise, Journal of Experimental Psychology,
Vol. 47, No. 1 (January 1954), pp. 27-31.

 

H

 

Gerlach, Arch C. ”Visual Impact of Thematic Maps as a
Communication Medium,” International Yearbook of
Cartography, Vol. 11 (1971), pp. 169—76.

 

 

Gr00p, Richard E. and Daniel G. Cole. ”Overlapping Graduated
Circles: Magnitude Estimation and Method of Portrayal,"
Canadian Cartographer, Vol. 15, No. 2 (December 1978),
pp. 114-22.

 

78

Hake, Harold W. ”Form Discrimination and the Invariance of
Form," in Uhr, Leonard, ed., Pattern Recognition.
New York: John Wiley & Sons, 1966.

 

Herriot, Peter. Attributes of Memory. London: Methuen and
Co., 1974.

 

Jenks, George F. ”The Evaluation and Prediction of Visual
Clustering in Maps Symbolized with Proportional Circles,"
in Davis, John C. and Michael J. McCullagh, eds.,

Displgy and Analysis of Spatial Data. New York:
John Wiley & Sons, 1975.

 

Kaufman, E.L., M.W. Lord, T.W. Reese and J. WOlkmann.
"The Discrimination of Visual Number," American Journal
of Psychology, Vol. 62, No. 4 (October 1949), pp. 498-
525.

 

Lloyd, Robert and Theodore R. Steinke. "Visual and Statistical
Comparison of Choropleth Maps,” Annals of the Association
of American Geographers, Vol. 67, No. 3 (September 1977),
pp. 429-36.

 

Leewenberg, E.L. and H.F. Buffart, eds. Formal Theories of
Visual Perception. New York: John Wiley & Sons, 1978.

 

 

Loftus, Geoffrey R. and Elizabeth F. Loftus. Human Memory -
The Processing of Information. New York: John Wiley &
Sons,’1976.

 

Mandler, George. "Organization and Recognition," in
Tulving, Endel and Wayne Donaldson, eds., Organization
of Memory. New York: Academic Press, 1972.

 

 

Mardia, K.V. Statistics of Directional Data. New York:
Academic Press, 1972.

 

 

McGeoch, John A. "Forgetting and the Law of Disuse,"
Psychological Review, Vol. 39, No. 4 (July 1932),
pp. 352-70.

Mendenhall, William, Lyman Ott, and Richard Larson. Statistics:

 

A Tool for the Social Sciences. Boston: Duxbury Press,
1974.

 

Monmonier, Mark S. "Measures of Pattern Complexity for
Choropleth Maps,” American Cartographer, Vol. 1, No. 2
(October 1974), pp. 159-69.

 

Muehrcke, Phillip. "Visual Pattern Analysis: A Look at
Maps,” unpublished Ph.D. dissertation, University of
Michigan, 1969.

 

79

Muller, Jean-Claude. ”Objective and Subjective Comparison
in Choropleth Mapping," Cartographic Journal, Vol. 13,
No. 2 (December 1976), pp. 156—66.

 

Neisser, U. Cognitive Psychology. New York: Appleton,
Gentry and Crofts, 1967.

 

Norton, David and Lawrence Stark. "Scanpaths in Saccadic
Eye Movements while Viewing and Recognizing Patterns,"
Vision Research, Vol. 11, No. 9 (September 1971),

929-

Olson, Judy. ”The Effects of Class Interval Systems on
Visual Correlation of Choropleth Maps,” unpublished
PhpD. dissertation, University of Wisconsin, 1970.

"Class Interval Systems on Maps of Observed
Correlated Distributions, “ Canadian Cartographer,
Vol. 9, No. 2 (December 1972), pp. 122- 31.

 

. "Autocorrelation and Visual Map Complexity,”
Annals of the Association of American Geographers,
Vol. 65, No. 2 (June 1975), pp. 189-204.

Pellegrino, James W. , et al. ”Short Term Retention of Pictures
and Words as a Function of Type of Distraction and Length

 

of Delay Interval," Memory and Cognition, Vol. 4, No.1
(January 1976), pp. ll—15.
Phillips, W.A. ”On the Distinction Between Sensory Storage

and Short Term Visual Memory,” Perception and Paypho—
Physics, Vol. 16, No. 2 (October 1974), pp. 283—90.

 

 

Reyment, R.A. Introduction to Quantitative Paleoecology.
New York: Elsevier Publishing Co. 1

Schulman, Arthur I. ”Recognition Memory and the Recall of
Spatial Location,” Memory and Cognition, Vol. 1, No. 3
(June 1973), PP. 256—60.

 

Sneath, P.H. ”Trend Surface Analysis of Transformation Grids,”
Journal of Zoology, Vol. 151 (1967), pp. 65-122.

 

Steinke, Theodore R. ”The Optimal Thematic Map Reading
Procedure: Some Clues Provided by Eye Movement
Recordings," Auto-Carto II, 1975.

Taylor, Peter J. Quantitative Methods in Geography. New
York: Houghton—Mifflin, 1977.

 

Tulving, Endel and Zena Pearlstone. ”Availability vs.
Accessability of Information in Memory for Words,

80

Journal of Verbal Learnin and Verbal Behavior,
Vol. 5, No. 6 (August 196 ), pp. 381—91.

 

Tversky, Barbara. ”Encoding Processes in Recall and
Recognition," Cognitive Psychology, Vol. 5, No. 3
(November 1973), pp. 275—82.

 

. "Eye Fixations in Prediction of
Recognition and Recall,” Memory and Cognition, Vol. 2
No. 2 (April 1974), PP. 275-78.

 

Warrington, Elizabeth K. and Carol Ackroyd. ”The Effects
of Orienting Tasks on Recognition Memory,” Memory and
Cognition, Vol. 3, No. s (March 1975), pp. 140—42.

Wathen-Dunn, Weiant, ed. Models for the Perception of
Speech and Visual Form. Cambridge, Massachusetts:
M I.T. Press, 1967.

 

 

Whiteside, John A. ”Visual Display Recognition and the
Duplication of Inspection Sequences," Perce tion and
Psychophysics, Vol. 23, No. 4 (April 1973), pp. 350-55.

Wicklegren, Wayne A. and Donald A. Norman. "Strength Models
and Serials Position in Short Term Recognition Memory,”

Journal of Mathematical Psychology, Vol. 3, No. 2
(July 1966), pp. 316-47.

 

U

"111111";th

HliHlllfliirliilmilil MI 3
0626

724

 

O