13...... ‘ .
.

 

 

 

 

 

 

 

 

-uv--u-

 

 
  

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

Investigating an Imagery Process In Pigeons

presented by

Julie Janelle Neiworth

has been accepted towards fulfillment
of the requirements for

Doctorate degree in Psychology

Major professor

Date August 12, 1982

MS U is an Affirmative Action/Equal Opportunity Institution 042771

 

 

 

 

MSU RETURNING MATERIALS:
Place in book drop to
LIBRARIES remove this Checkout from
“ your record. FINES will

 

 

be charged if book is
returned after the date
stamped below.

 

 

 

INVESTIGATING AN IMAGERY PROCESS IN PIGEONS

BY

Julie Janelle Neiworth

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Department of Psychology

1 987

ABSTRACT

INVEST IGATING AN IMAGERY PROCESS IN PIGEONS

BY

Julie Janelle Neiworth

The structural and functional qualities of an imagery process were investigated in a
memory task with pigeons (Colunba livia). The goal of the investigation was, first, to determine
whether imagery was a memory process used by a nonhuman species, and, second, to define
the characteristics of a representation of movement in such a species. A movement estimation
task involving a rotating clock hand stimulus was employed. Subjects were required to
discriminate constant velocity from aberrations from constant velocity (i.e., acceleration and
deceleration) by a choice response. The memory component of the task was that , on some
trials, subjects had to extrapolate stimulus movement after the stimulus disappeared in order to
judge its final location. In a previous set of studies, this task successfully uncovered a
cognitive estimation strategy in pigeons that supported some of the characteristics of an
imagery process.

Experiment 1 presented various delays to determine to what extent the physical
characteristic of time was represented when movement was represented. Although fairly

accurate in their response to novel delays. pigeons' process for estimating movement did not

always support subtle changes in the delay. Further analyses of subjects' patterns of errors
indicated that fast accelerations overwhelmed the subjects' range of estimation and forced a
shift of strategies from movement estimation to S—R associations. Within a particular low range
of accelerations and decelerations, subjects accurately estimated stimulus movement. but
demonstrated a tendency to confuse acceleration with constant velocity. A phenomenon like
”mental rotation" was observed, in that subjects made more errors as the distance of
extrapolation increased.

In Experiment 2, the speed of visible stimulus movement was increased or decreased in
two test conditions. Resulting performance to these speed changes was accurate, indicating
that subjects could utilize the information about rate presented in the stimulus display to
estimate the location of the moving stimulus. A timing strategy was refuted, and an imagery
process was supported.

The stimulus display of movement was removed in Experiment 3 and its effect on
performance was monitored. The result was a complete shift of strategies from estimation to
SR associations again. This result further supported the hypothesis that imagery was used, for
an imagery process requires a perceptual display for proper representation of movement.

The investigation provided strong support that pigeons use an imagery process to
represent a stimulus as it moves and disappears. The set of studies also revealed some

constraints of representing movement in pigeons.

To Mark,
To Marie,
To my family,

Andto Emma.

ACKNOWLEDGEMENTS

I would like to thank Dr. Mark Rilling for guidance and support throughout this research effort.
Dr. Skip Kendrick, Dr. Thomas Carr, Dr. David Irwin, Dr. Antonio Nunez, Dr. Rose Zacks, Dr. Anthony
Wright, and Dr. Ray Frankmann offered invaluable comments, arguments, and suggestions. Scott
Gillespie, Rick Wood, and Lee Buhler offered technical support. A special thanks to Mark
Rosenberg for providing software, computer, and printing support, and without whom the task of

organizing this dissertation would have been unbearable.

TABLE OF CONTENTS

LIST OF TABLES
LIST OF FIGURES
INTRODUCTION
A Theoretical Description and an Operational Definition of an Image
The Identification of Some of the Characteristics of Imagery: A Brief Review
of Previous Research
Experiment 1
Experiment 2
Experiment 3
Problems With Identifying an Imagery Process in the Pigeon and a Solution:
Quantifying the Representation
The Quantification of the Representation Apart From Verification of Imagery
Plan of Current Research
Experiment One:
The Effects of Changes in the Magnitude of Velocity During the Delay on
the Representational Process
Method
Subjects
Apparatus
Procedure
Results
Test of Transfer

Linearity of Functions

vii

viii

12

15

18

19

23

23

23

25

31

31

35

Test of Parallelism by Multiple Regression
Discussion and Interpretations
Experiment Two:
The Effects of Changes in the Velocity on the Representational Process
Method
Subjects
Apparatus
Procedure
Results
135° Test
180° Test
Comparison of Enor Scores to 135° and 180°
Discussion and Interpretations
Experiment Three:
The Effects of Removing Perceptual Cues for Movement on the
Representational Process
Method
Subjects
Apparatus
Procedure
Results
135° Test
180° Test
Comparison of Error Scores

DISCUSSION

46

53
54
54
55
55
60
62

67

8

72
73
73
73
73
76
76
80
82

85

Summary and Conclusions
APPENDIX A
APPENDIX 8
APPENDIX C

LIST OF REFERENCES

vi

85
90
93
94

96

LIST OF TABLES

Table 1. Rate of acceleration or deceleration during the delay.
Table 2. Three—way analysis of variance with repeated measures:

Trial types (speed) X Order of testing x Location Condition.

Table 3. Mean Errors and Standard Deviations to Speed Conditions at Each Location.

Table 4. Regression Analyses for Each Location.

Table 5. Partial F-tests of the Correlation Coefficients.

Table 6. Predictions of Response Accuracies to the Probes.

Table 7. Analyses of Variance of Speed Conditions and Trained conditions.

Table 8. Analyses of Variance of Test and Trained Conditions: Experiment 3

vii

33

36

45

59

79

LIST OF FIGURES

Figure 1. Trial types for 135° and 180° training, and for the test of transfer at 158°.

Figure 2. Mean percent correct scores from the 158° test after 135° training (top panel)
and after 135°-180° training (bottom panel).

Figure 3. Mean percent correct scores from the 158° test after 135° and 180° training
(Preliminary Experiment 2).

Figure 4. Trial types for 135° and 180° training and for the test of transfer at 202°.

Figure 5. Mean percent correct scores from the 202° test after 135°-180° training.

Figure 6. Error scores from the preliminary experiments showing linearity as a function
of speed.

Figure 7. Trial types in training and testing: Experiment 1

Figure 8. Speed changes and delay conditions of novel violation trials.

Figure 9. Order of testing of the violation trials.

Figure 10. Mean errors plotted for imagery trials (acceleration a: 0) and for violation trials

(all others) for each location tested.

Figure 11 (a—c). Mean enors and best-fitting regression lines for each location tested.

Figure 11a. Mean errors and best-fitting regression line for the 158° test.

Figure 11b. Mean enors and best-fitting regression line for the 180° test.

Figure 110. Mean errors and best-fitting regression line for the 202° test.

Figure 12. Training and test trials employed in speed testing in Experiment 2.

Figure 13. A comparison of average percent correct scores in trained and test
conditions of fast and slow speed testing in Experiment 2.

Figure 14. Percent correct scores to perceptual, imagery and violation trials.

viii

1O

11
13

21

28
29
30

37

40
41

42

56-57

61

Figure 15. Mean percent correct scores to imagery and violation trials within
speed testing with respect to first and second order of testing.
Figure 16. A comparison of mean errors to slow and fast probes at 135° and at 180°.
Figure 17. Training and test trials with removal of perceptual cues for movement.
Figure 18. Mean percent correct scores to trained and test trials in Experiment 3.
Figure 19. Percent correct scores to perceptual, imagery and violation trials.
Figure 20. Mean errors to trained and test trials by speed of movement during the
delay in Experiment 3, compared to errors made in Experiment 1.
Figure 21. Mean errors to trained and test trials plotted by speed of movement during

the delay.

69

75

77

78

83

84

INTRODUCTION

How the results of experience are maintained or "represented“ so as to affect later
behavior has long been of concern to animal Ieaming theorists (Cole, 1907; Thomdike, 1938;
Skinner, 1984) . The popular contention has been that there is a mapping relation between
the features of a “representation" and the features of a stimulus such that the study of the
representation is not necessary because the two worlds (real and representational) are
equivalent (Skinner, 1984; Schwartz, 1984; Carter 8 Werner, 1978; Zentall, Edwards, Moore 8
Hogan, 1981). Only recently, with the advent of the study of cognitive abilities in animals has
the question of the [can of representation in animals arisen (Roitblat, 1982). Imagery is the
most frequently chosen answer to this question (Griffin, 1981 ; Jerison, 1973; Terrace, 1985).
However the answer is fonnulated from the thinking of Darwin, Wundt, and others who argued
that images are omnipresent, or are “the mothers of all internal representations" (Wundt, 1894;
Kosslyn, 1980). While it has been theorized that animals' (i.e., monkeys', pigeons' and rats')
representations of real-world events contain perceptual characteristics of those events and are
thus image-like (see Roitblat, 1982 for a review). there has been no systematic investigation of
the possible identification and use of an imagery process in the cognitive system of any one of
these species.

The function of an Image is to maintain perceptual characteristics of stimuli in memory.
Imagery Is a cognitive process to which neural activity responsble for perception must be
channeled and maintained. The neural mechanisms involved in basic perceptual processes
have been traced in many species, with a special emphasis on avians, cats and monkeys. The
goal of the present investigation is the identification of an imagery process in a nonhuman
species for which biochemical and neural investigations are possible. Because of its ties with
perception, the study of Imagery seems best-suited for an investigation of cognitive and brain

function. The aim is to eventually develop an animal model of perceptual/cognitive processing

2
that can yield descriptions of mind-brain function which can be used to tailor infonnation
processing models. In developing an animal model of such function, it is not assumed that the
neural systems of humans and of nonhuman species are of near-equivalence, nor that their
cognitive systems are identical. It is assumed that some basic similarities exist ( i.e.,
homologous visual pathways and some basic cognitive processes ) that make the systems
compatble. Of central in'portance Is the assunption that an animal model of mind-brain
processing is at least as veridical as is a conputer model of such processing. While conputer
and communications models have been very useful in describing humans' mental or 'mind'
functions, it is clear that the neural processing that accompanies such mental functioning is
much different than the programmed binary interactions that occur in a simulated conputer-nrn
mental operation (Posner, Pea & Volpe, 1982) or that occur in a serial process developed from
communications models (Lachman, Lachman 8 Butterfield, 1979). The neural processing that
occurs with cognitive processing in various species draws a more realistic picture of processing
characteristics because the basic stnrcture of activation and feedback is maintained and the
neural medium through which activity occurs is similar.

In selecting an animal model for cognitive study, two factors are of importance:
1)whether the animal shares the same phylogeny with the species for which the process under
study is to be generalized, and 2)if the animal's cognitive processing and perceptual
processing warrant such study. The goal at present Is to find a simple organism in which neural
and cogntive processing may be studied simultaneously. Therefore the simpler one-celled
organisms used to investigate the chemical and neural correlates of Ieaming (Kandel, 1971 ;
Thomson, 1986) are not the proper species in which to study higher level cognitive and
perceptual processing. While rats have been the simpler mammalian family of interest in neural
investigation, they do not exhibit well-developed abilities in visual information processing. In
fact. I Is suggested that for cognitive processing and visual information processing, the
pigeon's system may be more similar to higher mammals' system (i.e., to humans) than to the

lower mammals' systems (i.e., rats and dogs). While the pigeons hold a position on a different

3
phylogenic tree than do humans, their visual system seems functionally and structurally
equivalent to that of monkeys (Maxwell 8 Grands, 1979). In addition, they are very responsive
to visual stimuli, and their overt behaviors to them are clearly noted in the animal learning
literature. An argument can be made here for parallel evolution, for even though humans and
avians occupy different branches of the evolutionary tree, their visual systems seem to have
evolved sthcturaIIy and functionally as similar. While pigeons are the species of interest in
these studies, it Is probably pmdent that other types of mammals, specifically cats and
monkeys, are studied in this way. There are certainly some qualitative differences between

mammals and avians that warrant cross-species considerations.

A Theoretical Description and an Operational Definition of an Image

Imagery has been defined generally as the processing of visual information not currently
before the organism (Pinker, 1984). The structure of an image remains unclear: some have said
that images are mental "pictures" that are stored (Plato, Bugelski, 1970), or that images are
memories that activate the same neural mechanisms as do percepts (Hebb, 1968; Woodworth
& Schlosberg, 1954), or that images are simply ”picture-like”, with details of their stnrcture left
open-ended (Kosslyn, 1980). From the many studies of the stmcture and function of images, it
seems clear that visual perception and imagery are related. Specifically, it seems that
perceptual detail, orthe 31mm of an object, is maintained In an image as it is perceived.
Moreover, It Is clear that In imaging it is possible to transform or change a representation of an
object. In fact this is described as the active “use" or mnction of an image, insofar as it is easier
to transform a perceptual-Ike display(as in an image) than it is to contrive calculations and
change descriptors (as in abstract codes). And finally it has been demonstrated that perception
actually can mm with an ongoing imagery process in positive or negative ways, i.e.,
percepts can facilitate imaging or they can interfere with imaging. Because of the stnrcture,

function and Interactive nature of imagery, many researchers believe that imagery is a separate

4

and distinct memory process that can be identified from a converging set of evidence which
supports each of these three qualities. Operationally, imagery can be identified if it Is
demonstrated that perceptual characteristics are maintained in the representational process
(the stnrctural assumption), that stimulus transformations are made possible by the process
(the functional assumption), and that perception itself interferes with or facilitates the
representational process (the Interactive assumption).

An organism must maintain a representation of a stimulus after the stimulus has been
presented and before a test is given. An operant task employed to test imagery must require
perceptual information to be maintained and transfomrations to occur during a delay Interval
between the stimulus presentation and the test. The organism's accuracy to respond after the
delay interval can reflect its ability to maintain perceptual lnfomration and to employ mental

transfonnatlons if proper tests are administered.

The Identification of Some of the Characteristics of Imagery:

A Brief Review of Previous Research

Previous research has tested to some extent two of the three characteristics that
describe an image (Neiworth 8 Rilling, 1987). The task employed was a conditional
discrimination involving movement estimation. In it pigeons were required to discriminate a
clock hand that rotated in a continuous fashion from one that did not. figure 1 presents the
various trial types used In training and testing.

There were three trial types presented In training: a perceptual type, an imagery type
and a violation type. On perceptual trials, the rotating clock hand moved at a rate of 90° per 8
along an are from a 12:00 presentation, or a 0° orientation, to a final location. Imagery and
violation trials were similar to perceptual trials, except that the rotating clock hand disappeared
at a 90° orientation, and some time later, reappeared at the final location. On imagery trials, tI

duration of the delay corresponded with the time It would take for the stimulus to be rotated

5
from 90°, its point of disappearance, to its point of reappearance. On violation trials, the
duration of the delay did not correspond with continuous stimulus movement. For example, in
the 135° violation trial (top panel of Figure 1), the stimulus disappeared at a 90° orientation ane
was absent for 1.0 s, or the time it would take it to be rotated to 180°, and then it reappeared at
135°. Perceptual and imagery trials were reinforced with 3-s access to grain for left key
responses. A left key response Indicated that the stimulus moved continuously even if it did
disappear for some time. Pigeons were reinforced for right key responses on violation trials.
The rationale for this methodology was to force the use of an imagery process by requiring
subjects to remember a moving stimulus and to transforrn it through space in its absence. It
should be noted that strategies other than Imagery could be used to accomplish the task:
Specifically, subjects could memorize a series of mles associating delay and location with a
choice response. or subjects could simply count or time the stimulus presentation and respond
appropriately with little need for a visual image. An attean to test these alternatives was made,

and Is pursued further in the dissertation experiments.

Exnenmemmg

Five subjects were first trained with 135° trials (top panel, Figure 1) until criterion
performance of 80 percent correct was met, and then they were exposed to 180° training trials
(middle panel, Figure 1) until criterion was met, and then to 135° and 180° training trials
presented in alternating sessions to criterion. There were two reasons for this approach: first, a
pilot study revealed that pigeons could not learn to respond to all the trial types presented in
the figure In training simultaneously; and second, a goal of the experiment was to determine
what kind of strategy pigeons were using at each step of acquisition of the cognitive process of
estimating the location of the rotating stimulus.

The test given to determine the strategy pigeons were using is described in the bottom
panel of Figure 1. Novel perceptual, imagery and violation trials were presented at 158° twice

each within 135° and 180° training sessions. The amount of transfer, or the accuracy of

Training

Perceptual Imagery Violation

h
W

135° 135° 135°

 

 

 

 

180° 180° 180°

Transfer

Perceptual Imagery Violation Iliolation2

0 O9 0

158° 158° 158°

Figure 1. Trial types for 135° and 180° training, and for the test of transfer at 158°.

7
response to these novel trials, was observed for 12 sessions. The first important feature to
notice about the test trials is that a 158° orientation was a novel location never before
presented but located halfway between the two trained orientations (135° and 180°).
According to Shepard and Cooper (1982), imagery Is an analog process. If the subjects'
representations are in analog form, then they most likely transforrn a representation of the clock
hand from 90° to 135° and to 180°, and thus the intermediate location of 158° should also be
represented. The result of the use of an analog imagery process in this task, then, would be
immediate and positive transfer to the intermediate location.

The second feature to note in the test trials is that the 158° imagery trial introduced a
novel delay (0.75 s) for which the subjects could have no prior associations. A rule-based
strategy predicts chance perionnance to first presentations of the 158° position. According to a
rule-based strategy, some practice with these trials is necessary for learning new delay -
location mles. However, an imagery strategy predicts immediate positive transfer, because this
strategy Is based on the assumption of continuous rotation applied to a novel example.

Two tests of transfer were conducted in Experiment 1, one test series after 135°
training, and one after 135° and 180° training. The top panel of Figure 2 shows transfer to 158°
trials after 135° training. Transfer was not complete. The strategy indicated by this pattern of
results was that pigeons were following a simple timing rule: If there was no delay or if there was
a short delay, subjects peeked the left key ; but if there was a long delay, subjects peeked the
right key.

The bottom panel of Figure 2 shows mean percent correct scores to 158° trials after 135‘
and 180° training. The result was positive and complete transfer. The mle-based strategy
adopted after 135° training was not used here. If it had been used, subjects would have applied
conflicting rules for responses associated with the trained delays and would have responded
at chance levels to these delays In the 158° test condition (i.e., to the violation trials). In
addition, there was no prior stimulus-response mle for the delay and location presented In the

158° Imagery trials, and chance responding should have been observed to these trials. The

 

PERCEPTUALI]
IMAGERY E
VIOLATION lllllll

 

 

 

roo -
90- 2
r

70..
60-
50-
4o-
30-
20—

ml

100 -
90— 2
80— ~ ,
70-
60-
50—
40-
30—
20-
10-

  
  

 

PERCENT CORRECT

 

 

 

 

 

135° l58°

    

PERCENT CORRECT

 

 

 

 

158° 150°

Figure 2. Mean percent correct scores from the 158° test after' 135° training (top panel)

and after 135°-180° training (bottom panel).

9
fact that subjects responded accurately to these trials suggested that they used some strategy
of estimating the stimulus location during the delay. Imagery is a likely candidate here, but a
timing strategy would also result in accurate estimation. In such a timing strategy, left key
responses would be associated with 0-s delays or short delays, and right key responses would
be associated with longer delays within each location condition. Such a strategy requires
memorized associations between delays, locations and responses. It does not require an
updated memory of stimulus movement.

Nevertheless, this experiment supported the notion that pigeons were retaining
accurate perceptual characteristics of movement during stimulus absence, and that their
representations for movement could be transformed or changed to accommodate a novel
example. The structural and functional characteristics of imagery are supported in part by this

result.

Experimentlm

Next, three naive pigeons were trained with 135° trials, then with 180° trials, and finally
with alternating sessions of 135° and 180° trials until criterion performance was met. Only one
test of transfer at 158° was given In this experiment to put to a test the argument that the
subjects showed positive transfer in Experiment 1 because of the practice they obtained with
the novel trials In the two tests of transfer. Figure 3 shows mean percent correct scores
obtained from the three subjects to 158° trials after 135° and 180° training. Positive and
conplete transfer to these con'pletely novel 158° trials was obtained. Thus practice with the
novel trials was not likely to be responsible for the positive transfer. Exposure to two exemplars

(135° and 180°) in training was necessary for subjects to use an estimation process.

W
All subjects (I.e., the five subjects from Experiment 1 and the three from Experiment 2)

were retrained to accurately respond to 135° and 180° alternating training sessions , and then

 

PERCEPTUALI]
IMAGERY
VIOLATION lllllll

 

 

 

100 -
90...

so-
70..
60—
50—
40-
30-
2o-
10—

PERCENT CORRECT

 

 

 

 

 

 

 

Figure 3. Mean percent correct scores from the 158° test after 135° and 180° training

(Preliminary Experiment 2).

Training

Perceptual Imagery Violation

. h
‘1
135° 35° 135°

180° 180° 180°

0;
-47

V

II

 

 

 

 

Transfer
Perceptual Imagery Violation Violation2

0 0303'

202° 202° 202° 202°

 

Figure 4. Trial types for 135° and 180° training and for the test of transfer at 202°.

1 2
the amount of transfer to a location outside the boundaries of training which was conpletely
novel was tested. Figure 4 shows 135° and 180° training trials, and 202° transfer trials. The
same testing procedure was used here as was used in 158° testing (i.e., each of the four 202°
trials was presented twice within single sessions of 135° and 180° training and this procedure
was followed for 12 sessions). Figure 5 shows mean percent correct scores for the 8 subjects
from 202° testing within 135° and 180° training sessions. Positive and complete transfer was
obtained again. Together with the results from Experiments 1 and 2, these results show that
pigeons can accurately represent rotation when the moving stimulus is absent. and that they
can estimate stimulus movement sufficiently to determine whether novel test locations are
appropriate, even when those locations lie outside training conditions. These results indicate
that the strategy pigeons are using is not based on SR mles, practice, or exposure to test
locations. The results support the stnrctural assumption of imagery, that the process maintains
perceptual characteristics, and the functional assumption, that the representations are
transforrnable.

Note that the interactive quality of imagery, that perception can interact and interfere with
an Imagery process, was not tested in this set of experiments, and thus a critical quality that
defines an image was not investigated. Further, a timing strategy cannot be mled out as an
alternative to imagery, for It can support the results of ExperimeMs 1 and 2 easily, and the
results of Experiment 3, although less strongly. A more detailed description of such a timing

strategy is provided in the next section.

Problems with Identifying an Imagery Process in the Pigeon

And a Solution: Quantifying the Representation

It is clear from the work described above that pigeons in these experiments are

enploying some cognitive process to accurately estimate the location of an object that moves

 

PERCEPTUALI]
IMAGERY
VIOLATION IIIIIII

 

 

l00 -
90-

80..
70..
60-
501
40—
30-
20-
IO

 

PERCENT CORRECT

 

 

 

 

 

 

135° 202° 180°

Figure 5. Mean percent correct scores from the 202° test after 135°-180° training.

1 4
and disappears. It Is also evident that certain characteristics of an Imagery process are present
in this cognitive process. However, there is not enough detail about the representation that is
quantified to conclude that Imagery is at work. For Instance, the structural and functional
characteristics of a moving stimulrs can be maintained by calculations or by a timing strategy,
and not necessarily by the use of a transformable image.

A timing/counting strategy accounts for the previously collected data as follows: First,
subjects determine a time or a terminal count for the perceptual trial In which movement Is
displayed to each location. Then subjects compare these times with a terminal count to the test
location In imagery and violation trials. If the times match, the left key is selected, and if the times
mismatch, the right key is selected. This strategy could be acquired in training, and applied
easily to novel locations that fall within the training trials (i.e., 158°) because the 158° perceptua
trials in which a terminal count is acquired is already embedded in the perceptual training trials.
The timing strategy would have to be stretched to the 202° location, for the terminal count/time
for this location must be estimated or generalized from training exarrrples since it does not lie
within training examples. Still, it does not seem that unlikely that subjects could apply a timing
rule to a location 20° beyond the location trained, and the transfer to 202° was a bit weaker (as
evidenced by less accurate responding than to 158° transfer) which could indicate an
application of a learned strategy to a novel example. This strategy can account for all of the
positive results of the previous set of experiments. The timing strategy represents an
alternative to the imagery hypothesis, despite the results of the previous set of experiments.

A critical difference between a timing strategy and an imagery process is that timing is
tied to a single dimension of time with paired associations for location, while Imagery uses both
space and time actively. Thus, a timing strategy does not require a display of movement, but
Imagery does. As long as the speed of movement remains constant, only the start location and
thefinal location are necessarylcrthetimlng strategy, and a count can be accurately made

between the two. Then, too, a timing strategy requires much practice along the dimension of

15
time If the speed of movement is changed. Many trials would be required for the subject to
learn the new tenninal counts which would have to be fonnulated for the test locations based
on a new speed.

The imagery process applies perceptual information about a stimulus in the stimulus'
absence in order to estimate its location. In this case, then, a perceptual display of movement
would facilitate mental transformations of the stimulus. Application of an imagery process
should produce lass accurate results if the perceptual display were removed, or if only the start
location and the final location were presented.

However, the imagery process could accommodate changes in speeds of movement
readily, as long as those speed changes were presented perceptually. The task of the imagery
process Is to use information about the stimulus that is presented and then to transform the
stimulus in accordance with these perceptual characteristics. The application of this process to

changes in the speed of movement should yield positive results with little need for training.

The Quantification of the Representation Apart From Verification of Imagery

The formulation of the characteristics of a timing strategy and an imagery process lead to
experiments designed to detennlne if imagery Is the representational process used. The
specific manipulations which differentiate these two processes are: changing the speed of
stimulus movement to be represented on probe trials, and removing the perceptual display for
movement on probe trials. The results of these manipulations identify which strategy is most
likely errployed. They also test the functional assumption of imagery, that an image can be
transformed In ways defined by the perceptual display, and also the Interactive assumption of
Imagery, that the percept of movement Influences the imagery process, for its removal is
proposed to cause deleterious effects while its change (i.e., a change in the speed) is
suggested to be Incorporated in the image.

1 6

But another goal is realized in this research: The identification of the characteristics
and/or the constraints of pigeons' representation of movement without regard for tests of
imagery but with an aim to describe pigeons' 'intemal model" of movement. Much work has
been acconplished in humans to investigate similarities and differences between mentally
extrapolated motion (i.e., motion represented after a physical moving stimulus disappears) and
the properties of actual physical motion. One finding Is that cognitive processes resemble
physical processes. An example of such a correspondence Is that humans imagine or
extrapolate the rotation of rigid objects In the way that objects actually rotate (Shepard 8
Cooper, 1982). Another example of correspondence is that humans represent momentum
after moving objects stop (Freyd 8 Finke, 1984). These results suggest that humans' cognitive
system has 'intemalized" many of the laws governing physical motion (Shepard, 1984).

Yet, some cognitive scientists have discovered a host of constraints and discrepancies
in representing or ”extrapolating“ movement in humans. Specifically, after a stimulus moving at
constant velocity disappears, humans tend to speed up or accelerate the representation of
stirrulus movemerl and, later, slow it to constant velocity again (Jagacinski, Johnson 8 Miller,
1983; Runeson, 1974). Humans exhibit a special problem with accelerating stimuli: They are
undale to accurately predict paths of stimuli accelerating at speeds greater than 10° of visual
angle/82 once the stlrnulus disappears (Gottsdanker, 1952; Rosenbaum,1975). The typical
methodology for these studies is that subjects observe a stimulus moving and after it
disappears, subjects press a response key to indicate when the stimulus that disappeared
would have reached a fixed visible marker.

Other studies have shown that people demonstrate erroneous beliefs about motion
when asked to predict the trajectories of objects (McCloskey 8 Kohl, 1983; Kaiser, Proffilt 8
Anderson, 1985). Specifically, people select as predictive of movement a trajectory that Is
Inconsistent with the laws of classical physics ( and also with natural observable laws of motion).

Specifically, adults select a trajectory that implies continual movement along an object's initial

1 7
motion path with gradual slowing to a stop. While their applied theory (the ”Impetus" theory)
works for linearly moving objects, it Is completely Inadequate to predict paths of curvilinearly
moving objects. There are debates concerning whether this nonpredictive strategy is ”learned“
because children express less of a tendency to follow it (Kaiser, Proffilt, 8 Anderson, 1985);
however, these studies identify a nonveridical estimation of stimulus movement whose
'irrpetus' is yet to be determined. A similar investigation of the ”internal model“ of movement
In pigeons would add to the base of knowledge that describes movement representation and
its constraints. As an avian species with a highly developed visual system, the investigation of
movement representation in pigeons initiates an interface between studies of perception and
studies of memory. For example, it provides a methodology for scientists to begin to label
neural networks involved during movement perception and during movement representation.
And It could identify deficits In pigeons' memory processing of movement which could be
avoided In future examinations of movement representation. With regard to humans' abilities, it
might be Interesting to note whether a "simpler" nonverbal species represents movement
more accurately than a more complex verbal species.

In sum, there are two Issues under Investigation in the present work. One is the imagery
hypothesis, and the question of interest is: Does a nonhuman species (such as a pigeon) use
imagery to remerrber objects and predict outcomes of change? To test for the existence of
Imagery, a movement task was selected so that subjects would be forced to maintain a
perceptual-like representation of the stimulus to respond. But the use of a movement task
brings forth a second issue to investigate: Do these subjects accurately represent movement?
What are the limiting factors of movement representation in pigeons? Within what constraints
does accurate representation of movement exist?

A theoretical description of an organism's cognitive system for extrapolating movement
is limited by the measurement of the structure of the representation across space and time.

While the set of experiments designed to test Imagery also reveals some characteristics of the

1 8

process for representing movement in pigeons, It is by no means definitive in describing the
structure of cognitive extrapolation of movement in pigeons. However, the experiments test in
a gross fashion specific variables which have constrained the process for representing
movement in other species: Specifically, changes in velocity, acceleration and deceleration.

While both the Issues of imagery and movement representation are investigated here ,
it should be made clear that one does not preclude or necessitate the other. Whether or not
imagery is found to be the controlling memory process, movement must still be represented in
order to estimate stimulus location In the task and its constraints within the task can be noted.
And, if imagery Is demonstrated to be the process in use, the constraints discovered of
representations of movement may not bear directly on the Imagery process, for they may have
more to do with the subjects' ability to accurately perceive movement and to detect movement

changes. For these reasons, the topics are investigated separately within these experiments.

Plan of Current Research

Three experiments were conducted to detennine if Imagery is the representational
process used to estimate stimulus location, and also to reveal certain characteristics of the
process of representing movement In pigeons. The first experiment identifies the nonveridical
nature of representations of movement In the pigeon, and also helps to test the stnrctural
assunption of imagery: That a full perceptual description of the stimulus is maintained In
memory If imagery is being used. The purpose of the experiment was to examine percent
correct scores to violation trials in which novel delays are introduced. Positive transfer, or
accurate performance across trained and test trials, Indicates that Imagery was used because it
verifies that subtle changes In rate are noted in the representation of the stimulus. In addition,
the pattern of errors obtained from different tests of violations in the moverrrent estimation task

Indicate a constraint of representing movement. The error rates to these novel violation

1 9
conditions reveal the degree to which space-time coordinates are updated during the process
of representing stimulus movement. The results of this experiment help to draw a clearer
picture of movement as it Is represented and possibly “imaged“ in the pigeon.

In the second experiment, the speed of initial rotation was manipulated within the same
task. This experiment tests the amount of transfer of the representational process to novel
movement conditions. In the description of this experiment, certain predictions are made
concerning response accuracies to the novel speeds of movement if a timing strategy Is used,
or If an Imagery process is used. The results of this experiment indicate which strategy is being
used.

Finally, Experiment 3 examines the Influence of the perceptual display oi rotation from 0°
to 90° on the cognitive process of representing movement. It identifies the triggering feature
for representing movement as either the start location, or the start location and some display of
movement. In this experiment, perceptual cues for movement are removed In test trials and
only static stimulus displays separated by delays are provided. Movement is necessary for an
Imagery process, and so If Imagery is errployed, then removing the movement display should
produce a decrement in response accuracy. Alternatively, if a timing strategy Is employed
successfully between 0° and 90°, then no decrement In perionnance should be obtained.

In sum, these three experiments help to detemiine II Imagery is the process at work and
quantify certain characteristics of the representational process: 1) its sensitivity to movement
changes which occur in the stimulus' absence, 2) its responsiveness to changes In visible

stimulus movement, and 3) Its dependence on perceptual cues for movement.

Experiment One:
The Effects of Changes In Rate During the Delay
This experiment examines one quality of imagery, namely the amount of detail of timing

and rate of a moving stimulus retained in memory . The previous set of experiments determined

20
whether the stnrcture of the representation of the moving stimulus was accurate and flexible
enough to accommodate novel locations. In this experiment, the duration of the delay is
manipulated to detennlne how finely tuned the stmcture of the representation is to time.
Accurate responding to novel delays further verifies that the stmcture of physical movement
was aptly represented, and this result also indicates that Imagery was used, in that another
physical component of movement is transformed in the representation.

This experiment also identifies a constraint of the representational process by an
analysis of patterns of errors to imagery and violation trials in which novel delays are presented.
It was detennlned from the previous set of experiments (Neiworth 8 Rilling, 1987) that subjects
made more errors to one type of violation trial than to any other violation trial or to any imagery
trial. The specific increases in errors observed were not tied to a particular location or to a
particular delay duration. Rather, a different view of these violation trials based on the type of
stimulus movement that could occur during the delay to predict the location revealed a
common element within each of these trials : momentum was represented In the image.

Figure 6 shows mean enors to the imagery trials (acceleration - 0) and to the violation
trials (all others) In the test locations presented in the previous experiments (i.e., to 158° and
202°). The horizontal axis of this graph depicts the type of movement required of the stimulus,
once It had disappeared, to reach its test location by the end of the delay. In imagery trials, the
type of movement was continued constant velocity and so errors to imagery trials are plotted at
the 0-point of the horizontal axis (meaning 0 acceleration during the delay). Those violation
trials which presented a delay too short for constant velocity movement to the test location
limited that the stimulus had to accelerate during the delay, hence errors to these violation
trials are graphed on the positive side of the horizontal axis. Similarly, violation trials which
presented a delay too long for the stimulus to move at constant velocity to reach the test
location Inplied that deceleration occurred, and errors to these trials are plotted on the

negative side.

Mean Errors

21

 

 

40 .
30 .
-fl- pilot 158°
+ pilot 202°
20 q
10 l r T I I I
-100 0 100 200

Rate of Deceleration/Acceleration

Figure 6. Mean errors plotted as a function of speeds tested In
the imagery and violation trials In Preliminary Experiments 1,2, and 3.

22
Errors are a linear function of deceleration/acceleration In this graph. The linearity of the
functions indicates that pigeons confuse acceleration with constant velocity. A strategy that
would produce such a function Is It subjects "speed up" their representation of movement
once the stimulus disappears, thus stimuli that actually accelerate airing the delay are more
often confused with a stimulus supposedly moving at “constant velocity“.

Note also that the functions are parallel. This Indicates that errors may increase by some
constant as the distance through which a representation must be rotated is increased ( i.e.,
from 158° to 202° ). This increase In errors is not due to the length of time in which the process
ls applied, for delay durations were counterbalanced across location conditions. Rather, errors
increased as the stimulus required transformation through farther distances.

The functions In Figure 6 suggest two conclusions, the first, of concern to the imagery
hypothesis. and the second, with regard to possible constraints of movement representation.
The conclusion concerning Imagery is that tests of certain delays (or of accelerating stimuli, in
general) produce less transfer, and this implies that the timing and rate are mt finely tuned in
the representation or Image. With respect to movement representation, the error patterns in
the preliminary data reveal the 'inlemal model“ of a displayed movement trajectory that Is
constrained by certain characteristics of position as a function of time.

This experiment tests different points on these two functions and Introduces a third
function. The acceleration rates tested in the previous experiments were 44°ls2, 184°/s2 , and
360°lsz. The only acceleration rate that did not produce Increased errors was 360°lsz. This
resul may Indicate that pigeons begin to clearly discem that a stimulus is accelerating in its
absence when It accelerates at a rate of 360°Is2 or higher. Rates below this may be confused
more often with constant velocity. Or the lack of enors to the 360°Is2 rate of acceleration may
be a result of practice. since this trial was one of the original training trials and had been

observed thousands of times by the subjects within the previous experiments.

23

A test of Imagery in this experiment is a test of transfer (or a mairtenance of accurate
performance) across trained examples and examples with novel delays. In general,
performance to perceptual, imagery and violation test trials should be similar to perionnance to
trained perceptual, imagery and violation trials if imagery Is used.

A test of the constraint of movement representation Is a test of linear change of enors to
each of the violation test trials. These two tests, the imagery test and the test of constraints,
may seem opponent, since one is verified by similar performance across trials while the other is
verified by differentiated performance across trials. However, the difference In the two tests is
that in the first, similarity must exist between average training and test perionnance for It to be
verified, while In the second, differences must exist between the various test trials for

verification.

Method

We

Eight White Cameaux pigeons served as subjects. The eight subjects participated In the
three experiments reviewed in the introduction, and were trained to differentiate between
constant velocity and violation conditions with a rotating bar stimulus. The subjects were
maintained at 80 percent +/- 20 grams of their free-feeding weights. They were individually
housed In a temperature-controlled, constantly illuminated colony room wherein water and grit

wasalwaysavallable.

Apparatus
Two Life Sciences Associates secondary video monitors (CAT#509—M01000-190) with
12 cm x 9.5 cm screens were fixed to the stimulus panels of two Lehigh Valley Electronics

standard pigeon chambers. These chambers were used in the three experiments described in

24
the introduction. Five of the eight subjects were tested in one chamber, and the other three
were tested In the second chamber, complying with the previous experiments. A brief
descr'ption of Chamber 1 follows, and then a description of the differences between it and
Chamber 2 Is provided.

Chairman. The secondary monitor was fixed to the stimulus panel so that It was
positioned 6 cm from the top of the chamber and was centered from the sides. The secondary
monlor was wired directly to the video display of a TBS-80 Mlll microcomputer. The stimuli were
constructed using conputer graphics. Electromechanical equipment controlled the chamber
operanda, and these devices were interfaced to the computer through an ALPHA 800
Interface.

Three standard 2.54-cm diameter clear plastic keys were mounted In front of the video
monitor so that they were 13 cm from the top of the chamber, 2 cm apart from each other, and
22 cm from the floor. The keys were connected to a sheet of plexiglass that covered the
monitor. They required a force of 15 g. (0.15 N) to operate. A houselight (28V d.c., GE757)
was placed at the top center of a wall adjacent to the stimulus panel. The grain magazine was 5
cm X 5 cm and was located 9 cm from the bottom of the charmer on the stimulus panel and
directly below the middle response key. The chamber and monitor were housed in an insulated
chamber equipped with a fan to provide ventilation and mask noise. The microcomputer
collected trial-by-trial response data and averaged percent correct scores.

mm This apparatus is almost identical to Chamber 1. The stimulrs panel was
slightly different: It contained an array of 7 keys. The middle three were used in this study, and
the two on either side of these were taped and covered. The dimensions and placements of
the keys, magazine and video monitor, and the size of the chamber and the stimulus were
Identical to those of Chamber 1.

W The stimulus was a clock hand constnrcled of pixels, and it

measured 4.5 cm x 1 cm. It was presented with these exact dimensions at a 0° position, a 90°

25
position and a 180° position. Seven positions, constructed at 13° increments, were presented
between these 90° changes. The stimulus at these positions had "stair step“ sides due to the
low resolution of the computer graphics. The clock hand stimuli were presented statically for 70
ms, and were separated by 70 ms delays. By this method of presentation, the clock hand
”rotated“ about an axis at the center of the screen at a rate of 90° per 5 with the same kind of
progressive change of movement used in Freyd and Finke's work (Freyd 8 Finke, 1984).
Pigeons exhibit a wide range of perceived “apparent movement" from 2 cps to 64 cps when
the object in question is apparently proving at a rate of 20 cm/s vertically, horizontally, and at
angles (Siegel, 1970). They also exhibit this range of cps to rotary movement although the
speed was not conveyed (Siegel, 1971 ). Still, 7 cps falls within the range of apparent
movement by cycles described in the previous work. It should be noted here that humans who
have perceived this movement describe it as “progressively moving" but not as ”real”.

With a completely black background, and strongly Illuminated frequency gratings,
pigeons' threshold for discriminating flicker is 150 cycles per 8 (Dodd 8 Wirth, 1953; Powell,
1967; Powell 8 Smith, 1968). Although the background was dimly illuminated In this
experiment, the contrast of the stimuli was still high in the experimental chanber and the stirruli
were presented at 7 cycles per second, so it is likely that the pigeons perceived these stimuli as
separate flashes, not “fused” together. The test locations used In this experiment were 158°,

180°, and 202°.

Procedure

The procedure was a probe testing technique that tested specific acceleration and
deceleration rates. Two questions were of interest: 1) whether positive transfer would be
obtained to novel delay durations thereby verifying maintenance of the time-based
characteristics of the image, and 2) whether there were patterns of errors that revealed certain

characteristics of the lntemal model of representing movement. The specific patterns of

26
Interest were increased enors to increases in acceleration, and parallel error functions across
location.

The specific deceleration/acceleration rates tested were: «30°52, -10°/s2, 44°52,
80°Is2, 184°/sz, and 360°/s2. The location conditions in which all of these rates were tested
were: 158°, 180° and 202°. A 180° location was selected for probe testing because it was
located between 158° and 202°, and so the error function to 180° would lie between the two
functions shown in Figure 6 if there was a general effect of increased errors to increases in
location. The rates of movement were selected so that specific areas of the functions were
tested: A variety of points between 0 and 360°/s2 to illuminate the acceleration problem and a
few deceleration points to detennlne the shape of the functions there. Most of the rates
suggested for testing had been probe-tested in the previous experiments in one of the three
locations, and their points lie on one of the two functions drawn in Figure 6. The two
deceleration rates were selected so that an area of the function very near constant velocity was
tested. Increased error rates might be expected here since a slight deceleration Is not that
physically different from constant velocity. However, the preliminary error functions drawn In
Figure 6 suggest that there Is in confusion between constant velocity and deceleration, no
matter how slight the deceleration. The only novel component of the probe trials was the delay
durations, for new durations were presented in order to test the novel accelerations and
decelerations during stimulus absence.

Experiment 1 consisted of training and testing phases. First, the eight subjects were
trained to respond accurately to 135° and 180° training trials. Each session in training consistet
of 64 trials of perceptual, imagery and violation trial types. In perceptual trials, the clock hand
stimukrs was present throughout the trial and could be observed as it rotated to Its test location.
In imagery trials, the clock hand rotated from a 0° position to a 90° position, and reappeared aft
a delay at the position to which It would have rotated had It moved with constant velocity. There

were two violation trials presented in training: 1) a 135° trial in which the stimulus decelerated a'

27
a rate of 90°/$2 during the delay or was absent for 1.0 s, and 2) a 180° trial In which the stimult
accelerated 360°/s2 during the delay, or was absent for 0.5 s. In both cases, the clock hand
rotated at 90°/s from 0° to a 90° orientation, and then disappeared during a delay and
reappeared at a location Inconsistent with constant velocity rotation during the delay.
Perceptual and imagery trials were reinforced with 3 s of access to mixed grain for left key
responses indicating the stimulus moved at constant velocity, and violation trials were
reinforced for right key responses indicating the stimulus did not move at constant velocity
during the delay. Trial progression occurred as follows: the stimulus was presented statically at
0° for 5 seconds, then a VR-12 on the middle key of the three-key array initiated stimulus
rotation, then the stimulus either rotated to the test location (perceptual trials) or rotated to a
90° orientation, disappeared for a certain delay interval, and reappeared at the test location
(imagery and violation trials). Correct choice responses on the left or right key produced
reinforcement, and incorrect responses were followed by a 15-s intertrial Interval (ITI). After
reinforcement, all trials terminated with a 15-s ITI. Only one location, either 135° or 180°, was
presented per session. The training trials were displayed in Figure 1 of the introduction and are
shown again here (Figure 7). Subjects remained In the training phase and alternated daily
between 135° and 180° training until criterion performance, or 80 percent correct responding,
was obtained for two consecutive days.

Next, probe trials at 158°, 180°, and 202° were presented infrequently within
training sessions as a generalization test. Figure 8 Identifies the various probe trials and the
lengths of the delays presented.

Each testing session consisted of 72 trials, which included a random mix of 48 trials of
135° or 180° training and 24 trials of probes at a specific location (either 158°, 180°, or 202°).
The probes consisted of six presentations each of two of the six violation trial types specified in
Figure 8 for the test location, and six presentations each of the Imagery and perceptual trial at

that location. In other words, all the violation trial types for a specific location were not

28

Training
Perceptual Imagery Violation
~ h h
V ‘1

135° 135° 135°

C

180° 180° 130°

 

 

 

 

Figure 7. The training trials employed in Experiment 1. The dashed arcs indicate stimulus

movement during stimulus absence. The solid arcs represent visible stimulus movement.

29

Locations and Delays of Imagery and Violation Trials in the Test

Test Location

158°
158°
158°
158°
158°
158°
158°

180°
180°
180°
180°
180°
180°
180°

202°
202°
202°
202°
202°
202°
202°

-30°/s2
-1 0°/s2
0°/s2

44°ls2

80°/s2
1i34°/s2
360°Is2

-30°/s2
-1 0°/s2
0°Is2
44°/82
iio°ls2
184°/s2
360°132

-30°/s2
-1 0°/32
0°/s2
44°Is2
80°/s2
184°/s2
360°/82

Rate of Deceleration/Acceleration

Delay ( in s.)

0.89
0.79
0.75
0.65
0.60
0.50
0.41

1.27
1.06
1.00
0.83
0.75
0.61
0.50

1.76
1.34
1.23
1.00
0.89
0.72
0.58

Figure 8. Speed changes and delay conditions of the novel violation trials.

Training Sessions

135°
180°
135°
180°
135°
180°

135°
180°
135°
180°
135°
180°

135°
180°
135°
180°
135°
180°

30

TEST SERIES

Probe Test

1 58°
1 58°
1 58°
1 58°
1 58°
1 58°

180°
180°
180°
180°
180°
180°

202°
202°
202°
202°
202°
202°

Violation Trials

-30°/52, 80°/s2
44°/s2, 184°/52
-10°/s2, 44°/s2
80°/s2, 184°Is2
-30°/32, 360°ls2
-1 0°/s2, 360°ls2

80°/s2, 184°Is2
-30°/52, 80°/s2
-30°/sz, 360°Is2
44°/s2, 184°Is2
-ro°/s2. 44°/32
-10°/s2, 360°Is2

44°/32, 1 84°/52
-30°/s2 -10°/s2
184°ls ,:360°/s2
44°/s2, ~30°Is2
-10°/s2. sows2
80°Is2. 360°/32

Test Series Progression for each Subject

158,180,202:#25&#10
180,202,158:# 2 &# 8
202,158,180:# 4 &# 7
180,158,202:# 9 &#14

Figure 9. Order of testing of the violation trials.

3 1
probe-tested within a single session of training, rather only one-third of the violation trial types
were tested per session. Reinforcement contingencies were the same in these trials as they
were in training. Each location (158°, 180°, and 202°) was tested for six sessions. Each violatio
type was presented a total of 12 times in testing. The exposure rate to violation trials was kept
low to avoid any effects due to practice with these trials. Training trials were also presented
within testing , and they consisted of 12 perceptual, 12 imagery, and 24 violation trials at either
135° or 180°. The sequence of testing days is outlined in Figure 9. By this plan, testing was

acconplished in 18 sessions.

Results

Three hypotheses required testing: 1) whether positive transfer to the novel test trials
was obtained, 2) whether enors to the test trials showed a systematic relationship with
accelerations and decelerations, and 3) whether the errors to the test trials plotted for each
location condition were parallel. An analysis of the shape of error functions (#2) reveals more
about the acceleration biases Inherent in movement representation in pigeons. Evidence of
parallelism (#3) Implies that “mental rotation” is Involved, insofar as enors increase with
increases in represented space through which transformations must occur. And, transfer (#1)
Is the more general result sought to determine if an imagery strategy is successfully applied to
novel cases. Positive transfer is obtained if there are no differences between response
accuracy to novel and to trained examples. No transfer is obtained If chance responding is
observed to the novel examples.

Insimltanslfl

A two-factor repeated measures analysis of variance (ANOVA) was conmcted to

compare percent correct scores across location (trained 135° and 180° trials, and test 158°,

180° and 202° trials) and across trial type (perceptual, imagery and violation trials). Data ente

32
were mean percent correct scores for each subject collapsed across test sessions for
perceptual, Imagery and violation trials at each trained and test location. Scores to the various
violation test trials were collapsed into one violation score for each subject at each test location.
If the subjects responded to the novel violation trials introduced In this experiment with the
same proficiency as they did the trained trials, then the prediction Is a lack of significance to trial
types and to location, as well as nonsignificant effects of their interactions. An imagery strategy
predicts complete positive transfer, or consistently accurate responding, across familiar and
novel examples.

Table 1 shows the results of the ANOVA. Significant differences in percent correct

scores were obtained to location (F4, 105 . 7.368) and to trial type (F1105 :- 117.5) as well as

to the Interaction between the two (F8. 105 - 3.405). The significant trial type effect was a

result of highly accurate performance to perceptual trials (mean = 93.97) as conpared to
imagery (mean - 71.156) and violation (mean - 62.372) trials. The difference between imagery
and violation trials was not significant by a Tukey comparison (Q Imagery-violation - 2.79, p >
0.05) bill the difference between perceptual and imagery and violation trials was significant (Q
perceptual-imagery - 7.25, p < 0.05; O perceptual-violation a 10.051, p < 0.05). Since the
significant effect was caused by men accurate responding, not less, it is not regarded as
evidence against the imagery hypothesis. Average performance across all trial types was
accurate. This is the relevant result, for if response accuracy to imagery and violation trials is still
above chance, then the subjects have shown the minimum necessary perionnance to
entertain the Imagery hypothesis.

The significant difference found to location and to the interaction between bastion and
trial type Implies that transfer was not obtained from training to testing. However these
difference results are not conclusive, for they could be a result of significant variation within the
test locations and trial types or within the trained locations and trial types, and not between the

two (i.e., between trained and test examples). To determine whether performance to test trials

33

Table 1. Analysis of Variance of Percent Correct Scores to Trained and Test Conditions.

SOURCE

Location
Trial Type
L X TT
Error

05

Sums of Squares Mean Square
2330.304 582.576
18589.801 9294.901
2154.069 269.259
8302.497 79.071

F-test P value

7.368 0.0001
1 17.551 0.0001

3.405 0.0016

Tukey Tests: Perceptual, Imagery and Violation Trained and Test Conditions

Trial Type

Perceptual

Imagery

Violation

Comparison

180TR - 158
180TR - 180
180TR - 202
135TR - 158
135TR - 180
135TR - 202

180TR -158
180TR - 180
180TR - 202
135TR -158
135TR - 180
135TR - 202

180TR -158
180TR - 180
180TR - 202
135TR -158
135TR - 180
135TR - 202

Q value (=4.039,table)

0.02767
0.02735
1 .02134
-0.3403
-0.3406
0.653

3.516
0.865
2.3149
4.443
1.793
1.387

2.6477
0.0229
2.543
1.454
3.124
6.645

Significance

not significant
not significant
not significant
not significant
not significant
not significant

not significant
not significant
not significant
significant at p < 0.05
not significant
not significant

not significant
not significant
not significant
not significant
not significant
significant at p < 0.01

34
was significantly lower than perionnance to training trials, a series of Tukey tests were
conducted (bottom panel, Table 1). In these, comparisons were made between average
percent correct scores to trained (135° and 180°) and test (158°, 180° and 202°) locations withi
each trial type. For example, perionnance to trained 135° imagery trials was compared to
perionnance to test Imagery trials (i.e., 158°, 180° and 202° imagery trials). Altogether, 18
comparisons were necessary, and the results are displayed In Table 1.

There were no significant differences in response accuracy to trained and tested
perceptual trials, as one would expect even without an imagery strategy in use. There was only
one significant difference found in response accuracy between trained and tested imagery
trials, but this difference indicated mere accurate responding to the 158° imagery test trials than
to trained trials, not less. In other words. positive and complete transfer was observed between
trained and test perceptual and imagery trials. This is not surprising since these trials were
tested before and positive transfer was obtained.

The critical test was whether response accuracy was significame less to test violation
trials (that introduced novel delays) than to trained violation trials. It is apparent that for the three
locations tested, there was only one significant differences in response accuracy between
trained and test violations. Average response accuracy to 202° vidlation trials was significantly
less than that to 135° violation trials. However, positive transfer was evident to 202° violations
well, as response accuracy to these was not distinguishable from response accuracy to trained
180° violations. Proof of transfer, that accuracy was maintained from training to testing, was 1
great degree obtained to the violation trials, since 5 of the 6 relevant comparisons produced
nonsignificant differences.

MW

Subjects demonstrated fairly accurate average response accuracy to the novel violation

trials, and their estimation strategy could accommodate the new delays fairly well. However,

specific test violation trials were met with increased enors, and a closer examination of these

35
increases might reveal a quality of the representation of movement. Another ANOVA was
concluded in which errors to imagery and violation test trials were compared directly, and as a
function of location. The ANOVA was a three-way repeated measures ANOVA examining
errors to each tested trial type (6 violations and 1 imagery) ,to the order of testing (n=4) and to
each tested location (158°, 180° and 202°).
Table 2 presents the relevant statistics. Two main effects were statistically significant in

the ANOVA: a significant difference between error scores to the various trial types tested

(F628 - 7.687, p < 0.0001) and a significant difference between error scores to the three

locations (F256 = 9.035, p < 0.0004). The order of testing effect, as well as all interactions

between speed, order and location were not significant.

Figure 10 shows the error functions plotted for the imagery and violation trials at each
location, and Table 3 presents the means and standard enors for each point. It seems apparent
that the functions linearly increase as a function of increases in acceleration during the delay.
And It is also suggested that the functions are parallel, both because they appear parallel in the
graph and also because there was no, interaction obtained between speed and location in the
ANOVA. Certainly, a lack of interaction does not prove that the functions are parallel, but further
regression analysies are employed to determine this.

The significant effects found in the ANOVA indicate that the tested aberrations in
movement caused significant changes in the subjects' responding. Since it Is clear that
response accuracy was affected both by novel violation conditions as well as by the location of
the test stimulus, a closer look at the shape and relationship of the error rates plotted as a
function of these variables is necessary. Gained by this pursuit Is a fuller description of the
representation of movement as it changes overtime and distance.

A few characteristics are obvious from the graph of mean errors (collapsed across

subjects) plotted in Figure 10: 1) the functions appear to be parallel with respect to location

Source df: Sum of Squares: Mean Square: F-test:
Speed 6 34035.669 5672.612 7.687
Order 3 1482.661 494.220 0.670
Speed X Order 18 15911.061 883.948 1.198
Subjects Within Groups 28 2662.187 737.935
Location (repeated measure) 2 14095.342 7047.671 9.035
Location X Speed 12 6312.846 526.070 0.674
Location X Order 6 1915.538 319.256 0.409
Location X Speed X Order 36 25713.200 714.256 0.916
Location X Subjects W/in Groups 56 43683.909 780.070
Total 221 67 (M)

Calculated Degrees of Freedom:

36

Table 2. Three-Way Analysis of Variance with Repeated Measures

Trial Types (speed) X Order of Testing X Location Condition

Speed: 6 violations + 1 imagery . 7 (M = 6)
Order: 4 orders (n-1 - 3)
Speed X Order: 63 . 18
Subjects Within ( 7 speeds * 4 orders - 28)
Location: 3 locations (n—1 - 2)
Location x Speed: 6 ‘ 2 . 12
Location X Order: 3 ‘ 2 - 6
Location X Speed X Order: 6 " 3 ‘ 2 - 36
Subjects Within :8 subjects ‘ 7 speeds ' 3 locations - 168.

p-value:

0.0001“'“
0.5778
0.3258

00004:“:
0.7683
0.8698
0.6053

For repeated measures, divide Within by repeated measure: 168 I3 = 56.

Mean Errors

70

60_

50

40

30

20_

10_

37

 

 

 

 

 

 

 

 

I T I ' I ' I ' I
-100 0 100 200 I300 400

Rate of Deceleration/Acceleration

Figure 10. Mean enors plotted for imagery trials (acceleration = 0) and for
violation trials (all others) for each location tested.

38

Table 3. Mean Errors and Standard Deviations to Speed Conditions at Each Location

Location

180°

202°

Speed

-30
-10
0
44
80
184
360

-30
-10
0
44
80
184
360

-30
-10
0
44
80
184
360

Mean:

10.416
14.583
18.403
29.167
37.5

40.625
51.042

20.833
16.667
28.819
37.5

46.875
57.292
62.222

28.125
26.042
35.069
42.708
51 .042
67.708
68.75

Standard Deviation

13.437
21.837
16.213
23.96

16.667
27.195
22.335

22.361
13.608
19.369
14.272
16.351
17.18

36.983

18.972
17.18

16.574
21.055
16.632
23.936
29.737

Standard Error

3.359
5.459
2.34

5.99

4.167
6.799
5.584

5.59

3.402
2.796
3.568
4.088
4.295
9.549

4.743
4.295
2.392
5.264
4.158
5.984
7.434

39
condition with the farthest orieMation (202°) producing the greatest amount of errors, and
2) each function seems to increase as a function of increases in acceleration, with a flattening of
each function occurring at higher rates of acceleration. These error functions were tested for
linearity and curvilinearity by simple and polynomial regression analyses, and then for
parallelism by a multiple regression analysis.

Figures 11a, 11b, and 11c display the best-fitting regression lines for each location
tested, as well as the polynomial equations and correlations. Quadratic functions best fit the
data for all three location conditions, even though the data for the 180° and 202° conditions
appear in Figure 10 to be cubic (I.e., because of the curves at both ends of the functions).

Table 4 reveals the statistics of the trend analysis from Which quadratic equations were

selected as best-fitting. The test of each location condition spanned sin'ple regression
(y=a+8x) to third-order or cubic regression (y=a+8x + l31x2 + 82x3 ). The correlations of each

function (r) are 0.55, 0.585, and 0.602, respectively. The R2 terms which indicate the amount
of variance accounted for by the functions remain between 30 and 36 percent. Although this is
acceptable for regression analysis, it should be noted here that the R2 terms are being held
artificially low because both within- and between-subject variances were combined as error
variance in these analyses. (The alternative was to enter means for each subject for each
violation condition, but then within-subject variance would be completely ignored by this
analysis.)

There was an insignificant increase of less than 1 percent of the variance gained from
quadratic to cubic functions (which is probably due to the change of the statistics alone), but
there was a gain of as much as 5 percent of the variance from linear to quadratic, indicating that
quadratic functions might better fit the data. In addition, partial-F tests which examine the
Independent effects of the coefficients of "x" In the equations indicated that quadratic
functions best fit the data, for the partial-F scores In the quadratic equations were highly

significant while they were not for cubic equations (see Table 5). Statistical evaluation of the R2

40

Figures 11 (a - 0). Mean errors plotted for imagery and violation trials for each
location condition, and the best-fitting regression line for each.

Mean Errors

41

60 y = 18.38 + 0.209x - 3.3609-4x"2 R02 =0.302

50..

40..

-El- 158°

30..

20..

 

10

 

 

*‘l ' l ' l 1 l ' l
-100 0 100 200 300 400

Rate of Deceleration/Acceleration

Figure 11a. Mean errors and the best-lilting regression line for the 158° test.

42

y 2.- 27.08 + 0.254x - 4.3666-4x"2 R02 = 0.343

70-

60..

50..

40..

Mean Errors

30..

20...

 

1O

 

 

j I I 1 I 1 T j I 1
-100 O 100 200 300 400

Rate of Deceleration/Acceleratlon

Figure 11b. Mean errors and the best-fitting regression line for the 180° test.

43

y = 33.81 + 0.259x - 4.4656-4x"2 R"2 a 0.362

80..

70.: q

60-

 

50 ' .3. 202°

Mean Errors

40-

30..

 

 

20 l'l'r'l'n

-100 0 100 200 300 400

Rate of Deceleration/Acceleration

Figure 116. Mean errors and the best-filling regression line for the 202° test.

Location

158°

180°

202°

Trend

linear
quadratic
cubic

linear
quadratic
cubic

linear
quadratic
cubic

p < 0.0001.

44

Table 4. Regression Analyses for Each Location

R

0.52
0.55
0.56

0.54
0.58
0.59

0.56
0.60
0.60

R2

0.27
0.30
0.31

0.29
0.34
0.34

0.31
0.36
0.36

F-test

52.524‘“
30.507‘“
21 .063‘“

59.837‘“
36.746“'
24.415’“

64.515“'
40.088‘“
26.654’“

X

0.101
0.209
0.303

0.114
0.254
0.285

0.115
0.259
0.225

x2

-0.000336
-0.001

-0.000436
-0.001

-0.000446

x3

.000002

7.3456-7

-0.0000006 7.968-7

y-Intercept

19.45
18.38
18.94

28.476
27.081
27.266

35.235
33.808
33.607

45

Table 5. Partial F- tests of the Correlation Coefficients

Location Trend x x2 x3 Fp (x) Fp (x2) Fp (x3)
180° linear 0.101 52.524‘“

quadratic 0.209 -0.000336 21.973'“ 6.467“

cubic 0.303 0001 .000002 13469” 3.076~ 1.821 n.s.
180° linear 0.114 59.837‘“

quadratic 0.254 -0.000436 29.627‘“ 9.904'“

cubic 0.285 -0.001 7.345e-7 10.767'“ 0.881 n.s. 0.179 n.s.
202° linear 0.115 64.515‘“

quadratic 0.259 -0.000446 32.716‘“ 11.081'“

cubic 0.225 -0.0000006 7.96e-7 7.151‘~ 0.007 n.s. 0.225 n.s.
“' p<0.001
“ p<0.005
*~ p<0.01

~ p=0.05

46
terms and of the partial-F tests support the observation that quadratic functions best fit the
data.

An F-test at each location (shown in Table 4) was conducted to detennlne if a large and
significant portion of the variance of enor scores (y) was explained by regression on speed (x).
In each test, the F-score obtained was significant at the 0.0001 level. These F-tests indicate
that for each regression, the predicted slope was significantly different from 0, and was
positively increasing.

In summary, the regression analysis for each function and the F-tests of each support
the hypothesis that the error functions increased with respect to increases in acceleration.
Because quadratic functions best fit the data, the interpretation is that there is a decline In the

rate at which enors increase at the higher rates of acceleration tested.

W

A closer review of the three best-fitting regression equations reveals the parallel nature

of these functions. They are:
y - 18.38 + 0.209x + -0.000336x2 for 158°,
y . 27.08 + 0.254x + -0.000436x2 for 180°, and
y - 33.81 + 0.259x + -0.000446x2 for 202°.

The coefficients of x (speed) describe the slope of these functions and they are virtually
identical (0.209, 0.254 and 0.259) and are not significantly different by a Tukey-Kramer test of
coefficients (Sakai 8 Rohlf, 1981). In addition, the coefficients of x2 (speed 2) describe the
curvature of the functions, and these coefficients are also almost identical (0000336,
0000436, and 0000446) and are not significantly different by the Tukey-Kramer test. In
short, the functions are virtually the same, with the important exception of the y-Intercept. The
y-intercept values increase as the location of testing increases. By plotting 95% confidence

intervals around each y-intercept as a function of its standard enor, it was detemlined that none

47

of the confidence intervals overlapped, and even the highest value of each interval did not
overlap with the lowest value of the next successive y-lntercept as a function of increased
location. The confidence intervals were as follows: for 158°, 18.403 +/- 4.167; for 180°, 28.819
+/- 2.796; and for 202°, 35.069 +/- 2.392. The y-intercepts were distinctly different from each
other. From these observations, the functions seem parallel and separate.

To statistically evaluate the parallel nature of these functions, a nonforced stepwise

multiple regression was conducted. The following equation was entered into the analysis:
y=a+ 80x1+ 81x12+ 82x2+ 133x1x2+ 84x12x2

where a = y-intercept, Bi - beta- coefficient,

x1 a speed (testing for linearity of the functions while ignoring location),

x12- speed2 (testing curvilinearity, ignoring location),

x2 . location (testing linearity by location, Ignoring changes In speed),

x1x2 - speed ‘ location (testing parallel linearity), and

x12x2 - speed2 ‘ location (testing the parallel W of the functions).

A nonforced stepwise regression was employed so that those variables of this set of
live that account for most of the variance would be selected In combination, while the others
would be ignored. Order of variable entry is not manipulable in a nonforced stepwise
regression, for the test selects the variables to enter on the basis of the amount of variance
accounted for or gained by the variables.

The equation produced by the multiple regression that best fit these functions was:

y - 1.089 + 0.284x2 + 0.0003x1x2 + -0.0000004436x12x2
where x2 - location,

x1x2 - speed ‘ location, (parallel linearity) and

48

x12x2 = speedz'locatlon (parallel curvilinearity).

The R2 term for this equation was 0.379, and the partial- F test for each B—coefficient was highly

significant (see Appendix A for details). The variable first selected that accounted for most of
the variance in the three functions was x1x2, or the variable accounting for parallelism of the
linear functions (R2 . 0.295). The second variable selected was location (R2 - 0.339), and the

third variable, x12x2, was the variable accounting for parallelism of the quadratic functions (R2 =

0.379). The variance accounted for by speed (x1) and by speed2 (x12), after taking all

functions into consideration and parcelling out the effects of the other three variables, was
minimal and nonsignificant by partial-F tests (see Appendix A for details). In short, the functions
are clearly parallel.

Discussion and Interpretations

In this experiment, novel delays were presented in violation trials in order to detennlne if
time-based characteristics such as rate were represented as an analog perceptual
characteristic of movement. A basic quality of an Image is that It maintains the physical stnrcture
of the entity that it represents. In this case, timing, or rate of change, was tested as a critical
structure In the image. The tested violation trials were met with approximately the same level of
percentage of correct responses as were the trained trials, as confirmed by an analysis of
variance of the percent correct scores. The conclusion from this analysis was that subjects'
acquired cognitive strategy for estimating movement could not be perfectly applied to these
novel conditions, but it was applied well enough for a general result of positive transfer to
occur. Percent correct scores did vary widely within the tested violation trials, indicating that the

representational process was not finely tuned to subtle changes in rate during the delay. 80,

49

the characteristic of movement representation responsible for the significant response
changes was sought. Errors, i.e., failures to discriminate, have been used to infer the nature of
the psychological dimensions along which objects or events are perceived to vary (Treisman,
1986). The goal of the analysis of errors in this experiment was to reveal the nature of the
psychological dimensions along which objects and events are represented as varying.

One characteristic of movement represention identified from pilot testing that resulted in
response changes was a linear relationship between errors as a function of increases in the
acceleration during the delay. The relationship, when fully described and analyzed In this
experiment, is quadratic: Errors increase with increases in speed, but the amount of Increase
decreases at the higher speed conditions tested. Once errors exceed 50 percent over all
responses. the functions start to flatten. There Is no indication In these analyses that the
functions eventually drop back down, forming an Inverted-U. Since this was not Indicated, the

inference that subjects confuse acceleration with constant velocity until W

 

unsupported. Curiously, the data indicate that pigeons confusion of acceleration with constant
velocity Increases with increases In acceleration until It reaches its limit (somewhere between
70 and 100 percent)! If this were the case, It Is hard to believe that the pigeon would have
survived flight, or could have survived predators (i.e., incoming hawks ). This strategy does not
hamess the sensitivity of the highly developed movement detection system in the pigeon's
visual cortex either (Granda 8 Maxwell, 1979).

It is clear In this study that high rates of acceleration produced inaccurate responding.
What is inplied In the data Is that the higher rates of acceleration pmdlced so much Inaccuracy
that subjects abandoned their trained discrimination altogether. The explanation that fits the
data is that In the case of very short delays indicating fast accelerations (i.e., >5O°/82), pigeons

simply responded on the left key (the constant velocity key) . In fact, as acceleration is

50

increased. the tendency to respond by this simple discrimination appeared to increase,
producing an increasing function of errors to increases In acceleration. This makes intuitive
sense if subjects were following a strategy such as this: if there is no delay (or a very short
delay), respond on the constant velocity key.

While the explanation regarding a change In the pigeons' strategy to a slnpler
discrimination accounts for the general shape of the functions, it does not sufficiently account
for error rates to accelerations depicted early in each function (i.e., to accelerations less than
44°Is2 ), for in these subjects are responding mstuftan to the right key to increases In
acceleration. At this level, it seems that subjects are using the strategy trained, and that was to
discriminate constant velocity (left key responses) from aberrations from constarrl velocity
(deceleration OR acceleration, right key responses). Even with an acceleration of 44°ls2,
subjects still respond on the right key 70 percent of the time in the 158° condition, 64 percent
of the time in the 180° condition, and 57 percent of the time in the 202° condition.

The shape of the functions in this area of speeds indicates that subjects tend to confuse
acceleration with constant velocity more often than they confuse it with deceleration. There Is
an asymmetric distribution of errors around constant velocity with more errors to acceleration
than to deceleration. This distribution could be normalized if some acceleration constant (of
between 10°/s2 to 30°/s2) were applied to the data . With an adjustment of an acceleration
constant, all represented movement Is Increased, making true deceleration seem exaggerated
and true acceleration seem more like ”constant velocity' which is represented as increasing.
This kind of adjustment would make deceleration discriminations quite easy, even for slight
aberrations from constant velocity (I.e., ~10°) but it would make acceleration discriminations
more difficult.

But this kind of adjustment would also make representations of true constant velocity

nonveridical, because the appropriate orientation at which the stimulus appears on imagery

51

trials should appear (an. In other words, if acceleration resembles represented constant
velocity, then the test locations of Imagery trials imply a slight deceleration. Why subjects
respond to these trials as if the stimulus moved at constant velocity is not explained by the
acceleration constant notion.

Another explanation is that there is a category of "constant velocity“ with some
accelerations lying within its boundaries. If an acceleration constant is psychologically applied to
movement represented, then some decelerations lie within these boundaries as well --
because with an acceleration constant applied, real constant velocity appears as a slight
deceleration. The psychological boundaries that define deceleration seem more distinctive
than those defining acceleration, for in no case was a true deceleration confused with constant
velocity, but in many cases, a true acceleration was confused with constant velocity.

To large increases in accelerations, subjects abandon the estimation strategy and
respond more often on the left key ( and incorrectly ) to greater increases in acceleration. This
strategy Is a simple generalization in response to very short delay and no delay conditions. A
model which describes subjects' overall strategy in this experiment is twofold: a constant
velocity-aberrations discrimination with a tendency to confuse acceleration with constant
velocity is applied in the tested area around 0 acceleration, and a simple delay discrimination is
used In the area testing greatly increased accelerations (>50°/82).

A constraint on the representational process of accelerations has been Identified. It
should be noted here that the error results do not necessarily imply that an Image was used, for
a category for constant velocity could exist out of context of imagery. However, other
behavioral evidence of "mental rotation” further strengthens the argument that Imagery Is
employed in the task.

Cooper and Shepard (1973), Metzler and Shepard (1974), and Shepard (1975) have

viewed the process of mental rotation as an lntemal analog of the process that occurs when the

52

rotation of an external object is perceived. They reported as evidence of mental relation that
the time required by human subjects to detennlne whether two visual stimuli are the same in
shape or mirror images Increases linearly with increases in the angular difference between the
orientations of the visual stimuli. Hollard and Delius (1982) reported that reaction times of
pigeons In the same task do not increase with increases in angular difference, rather they
remain the same across the tested orientations. They concluded that pigeons may not use a
strategy like mental rotation to evaluate the location of a represented stimulus, but that pigeons
have some parallel processing strategy for making such estimations that does not Involve
step-wise lntemal processing. There are problems with the conclusions of this study, due in
part to the fact that reaction time has not been extensively collected and accepted as a
dependent variable reflecting cognitive processing In pigeons (Herrnstein, 1985), and In part
because the design of the experiment assumed tacit knowledge of static rotations of stlrnuli
with which their subjects may not have been equipped. It should also be noted that Cerella
(1987) could not obtain evidence of rotational Invariance in pigeons at all, nor has Herrnstein
(see Cerella, 1987), and Hollard and Delius's results are currently being disputed.

The experiment reported here displays simulated rotations of stimuli, and requires
subjects to extrapolate the location of the rotating stimulus in its absence to accomplish the
task. Increases In angular difference, the variable of interest in Shepard and his colleagues'
studies as well as In Honard and Delius 's (1982) and Cerella 's studies (1987), was manipulated
here by changing the test location to which subjects must extrapolate rotation from 158° to
180°, and to 202°. The dependent variable which was systematically affected by changes In
angular difference was subjects' errors. The test of parallelism of the error data indicates that
some constant error rate is added to error functions as the test location Is moved from 158° to
202°. Note that this error Increase is not a result of increases in the length of time during whicf

subjects must trandonn the stimulus from 90° to the test location, for the delays used were

53

roughly counterbalanced across location conditions. Rather the error increase Is a result of an
increase In the distance through which subjects must transform the stimulus, irrespective of
time for processing. This result provides indirect evidence of mental rotation in pigeons: for If
pigeons are mentally rotating the stimuli during stimulus absence, then the farther they must
rotate the stimuli, the more errors they make.

The evidence of “mental rotation” provided by the significant test of parallelism further
strengthens the argument that imagery is the mechanism used in extrapolating the trajectory of
movement in this task. Such evidence clearly indicates that distance is being represented, and
that the process of estimating stimulus location requires analog transformations through which
increased errors are made with increases In space for transformations. This evidence supports
the conclusion made by Hollard 8 Delius (1982) that pigeons do have the capacity for rotational
transformations, and it refutes the conclusions drawn by Cerella and others (Cerella, 1987;
Herrnstein,1985 ) that rotational transformations are not possible by pigeons. Rotational
transformations are obtained if pigeons are trained to recognize rotated versions of stimuli as
rotated, and evidence of mental rotation as a process ( not obtained by Hollard and Delius,

1982) Is possible if errors are the variable of interest.

Experiment Two:

The Effects of Changes In Velocity on the Representational Process

The task of the second experiment was to further test the notion that imagery Is the
controlling process for movement estimation, and to examine other possible constraints of the
representational process. In this experiment, the speed of visible stimulus relation was
manipulated. Accurate responding to these novel movement conditions demonstrates that the

representational process has considerable plasticity, in that it can be applied to different

54

speeds. It has already been demonstrated that transfer from trained trials to test trials was , in
general, positive to different delays (Experiment 1, this plan of study) and to different locations
(Experiments 1, 2 and 3 of the Introduction). However, errors occurred within testing in
Experiment 1 that revealed a constraint of accurate representation of time and a constraint on
representing movement through farther distances of represented space. As In Experiment 1,
Imagery Is Investigated here by a test of transfer from a trained speed to novel speeds, and the
internal model of movement representation is investigated here by an examination of enor
patterns to the tested speeds.

The Imagery hypothesis assumes a dynarnlc representation of movement during
stimulus absence. An imagery strategy predicts positive transfer from training to the test cases
in which faster or slower perceptual displays of movement are presented. Alternatively, a timing
strategy would not result in positive transfer to the novel movement conditions. The
experiment was designed so that a timing strategy based on terminal counts associated with
specific locations predicts negative transfer, or would result In below chance responding to the
novel movement conditions. The assumption underlying the description of imagery In this
paper is that an imagery process applies perceptual information about movement during
stimulus absence. Using this strategy, transfer to the novel movement conditions should be

positive, or there should be rapid acquisition.

Method

alarm
The eight subjects from Experiment 1 participated in this experiment. They were

maintained under the same conditions as were described in Experiment 1.

Apparatus

The same two chambers described in Experiment 1 were used In this experiment.

55

W

The procedure of this experiment was to probe-test new fast and slow velocity
conditions within standard training sessions. The procedure was the same as had been
employed in all of the experiments: first, subjects were trained to respond accurately to 135°
and 180° location conditions, and then novel trials were presented infrequently within these
training sessions and response accuracy to the novel trials was noted. The novel trials
introdlced two new rates of constant velocity, 45°/s and 180°/s, in the familiar 135° and 180°
conditions.

Figure 12 depicts the trial types employed in Experiment 2. The top panel of this figure
shows the 135° and the 180° trials presented in training. The delays printed to the right of each
stimulus represent the time it took the stimulus to be rotated. The delays with arrows represent
the times during which the stimulus was absent. In training, the stimulus rotated at a rate of
90°/s, and so it took 1.55 for the stimulus to be rotated from 0° to 135° and it took 2.08 for the
stimulus to be rotated from 0° to 180°. Imagery trials used the same durations of time to press
movement as did the perceptual trials. The only difference was that in Imagery trials, the
stimulus rotated to a 90° orientation in 1.0 s, and disappeared for the remainder of the time it
took it to reach its final destination (another 0.5810! 135° and another 1.0s for 180°). In the
violation trials, the durations of delays during which the stimulus was absent violated the
constant velocity conditions maintained in these trial types. For example, in the 135° violation
trial, the stimulus disappeared for 1.0 s and then reappeared at 135° Instead of at 180°. The tria
progression for these types was the same as that described in Experiment 1, and right key
responses were reinforced In violation trials, while left key responses were reinforced in
perceptual and imagery trials. Locations for training (135° and 180°) were alternated daily. Eacr
session consisted of 64 trials, Including 16 perceptual, 16 imagery and 32 violation trials, to

insure that equal amounts of right key and left key responses were required per session.

56

Figure 12. Training and test trials employed in speed testing in Experiment 2.

57

hnagery
"(
‘l
t:

J

[as
61.47
[as
{-0.529

 

 

VHOHNJon
\
§_
/

 

 

ll .L .4
a
n N \/
w c D
c C
P
e
D.
Amxooov
@558...

 

 

 

mm m
a \le
W C
e>

 

Perceptual

e@

3?va
5.5.6200 >>o_m

 

 

 

 

a
a”. am
n 06>
r _
a \/
m C
V
a
.0.
W“
m... 9
a
m.©
.Mm 0. I.
a»
Pl
9
p C

$.02 v
556:8 and“.

58

Once subjects responded with 80 percent accuracy to both 135° and 180° training trials,
the slow and fast conditions were introduced within training sessions. One location was
presented per day. so 135° slow or fast trials were presented as probes only within 135° trainin
sessions, while 180° slow or fast trials were probed only within 180° training sessions. This
constraint on testing allowed for the manipulation of one specific variable: speed of rotation.

Probe-testing was accomplished in blocks of testing sessions. The slow and fast
conditions at 135° and 180° were tested within training sessions for 12 sessions each. In these
blocks of testing, 135° training and 180° training sessions were alternated daily, just as in
training. Slow test trials at 135° and 180° were Introduced within their respective training
sessions. The middle panel of Figure 12 displays the slow trials at 135° and 180°. In these, the
rate of movement was 45°/s. It took 3.0 s for the stimulus to be rotated to 135° at this speed,
and it took 4.0 s for it to be rotated to 180°. Again, the imagery trials supported the constant
velocity condition of 45°/s movement by presenting the appropriate delays for movement to
these two locations. But In the violation trials, the stimulus was absent for a period of time that
was too short for It to be rotated constame and for it to reach its final destination. In both
violation trial types in the slow condition, the stimulus actually increased Its speed during the
delay, or moved at 90°/s instead of 45°ls during the delay.

Note also that the delays used in the slow condition trials provided counterbalancing for
the response-delay associations acquired in training. For example, the delay presented while
the stimulus was absert in the 135° imagery training trial was 0.5 s, and a left key response was
required for reinforcement. In the test for the slow condition, a 0.5 s delay was presented while
the stimulus was absent in the 135° yigIathn trial. The test locations in both trials were exactly
the same (135°) and the delays during which the stimulus was absent were the same (0.5 s),
but the 135° imagery training trial provided reinforcement for left key responses and the 135°
slow condition violation trial provided reinforcement for right key responses. The critical

difference In the two trials was that the initial movement presented between 0 and 90° in the

59

first case was 90°/s and In the second case was 45°ls, and because of this difference, the trial
require opposite choice responses for reinforcement. lf subjects applied a strategy timed to a
90°/s speed of rotation directly and ignored the speed of initial rotation in the new movement
oonditlon, then they would respond incorrectly to the imagery and violation trials at 135° and at
180°. If, however, they adjusted their strategy appropriately to take into account the new speed
of rotation (45°/s), then accurate responding to all trials would be expected.

The bottom panel of Figure 12 shows the test trials to 135° and 180° fast conditions, or
when a rate of 180°/s was used. These were presented infrequently within same-location
training sessions within an altematlng training series. Note again that the delays used in the
violations during stimulus absence were of the same duration as those used in the original
Imagery trials (top panel). If a strategy tied to 90°/s rotation was used, inaccurate responding to
these trials is expected. The following table Identifies each trial type, the predicted pattern of
enors if speed of rotation is noted and accounted for in the representational process, and the
predicted pattern of errors if the speed of rotation is Ignored and the response-delay

associations acquired In training are applied to the novel trials.

Table 6. Predictions of Response Accuracies to Probes
135° SLOW imagery accurate responding below chance responding
violation accurate responding below chance responding
FAST Imagery accurate responding chance responding
violation accurate responding below chance responding
180° SLOW imagery accurate responding chance responding
violation accurate responding below chance responding
FAST imagery accurate responding below chance responding

violation accurate responding below chance responding

60

Those trials in which below chance responding was expected present a delay associated with a

choice response from training which is opposite the response reinforced in the novel
movement conditions. Those in which chance responding was expected present a novel delay
to which no prior response association could have been acquired.

The slow condition and the fast condition were tested in separate blocks of 12 training
sessions per block. A typical test session included 64 training trials and 8 probe trials, for a total
of 72 trials per session. The 64 training trials were comprised of 16 perceptual, 16 Imagery and
32 violation trials in which the stimulus moved at 90°/s. The 8 probe trials were comprised of 2
perceptual, 2 imagery and 4 violation test trials in which a new speed was introduced. The
number of trials per session was planned in this way so that there were equal numbers of left
key and right key responses required for each session. Subjects were divided In half
pseudorandomly and half received fast speed testing first, followed by slow speed testing;
while the other half received slow speed testing first, followed by fast speed testing. In
between the two blocks of testing, subjects participated in 135°-180° standard training until 80
percent correct performance was achieved.

The data of interest were average percent correct scores to the novel speed conditions.
Positive transfer, or accurate and consistent responding across speed conditions, indicates
that the subjects are adjusting their representational process to account for the changes in the

speed of rotation.
Results
The question of interest In this experiment was whether positive or negative transfer was

obtained to novel movement conditions in the movement estimation task, for this would

determine to some extent whether an imagery or a timing strategy was employed. Figure 13

PERCENT CORRECT

G)

8

10

PERCENT EORRECT
o

61

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

80 , 80 -
60 " 60 q
40 - 4O -
20 " 20 'l
0 0
135° 135° 135° 135°
trained slow trained fast
0 80 a
so .
40 d
O 20 -
o -,.‘ . 7 3 0 :1: . .
180° 180° 180° 180°
trained slow trained fast

Figure 13. Mean percent correct scores to the trained and novel speed conditions at
135° and 180°.

62

shows the mean percentage of correct scores from all subjects to the trained and tested speed
conditions. Separate bars are drawn for response accuracies to two trained conditions: one in
which the trained speed ( 90°/s ) was presented within slow speed testing, and one in which
the trained speed was presented within fast speed testing. It is apparent from the figure that
positive transfer was obtained for all conditions: the mean percent correct scores ranged from
67% to 75% across speed conditions and were clearly greater than chance, or 50%. All trials
produced above-chance responding, and the prediction of negative transfer made by a timing
strategy (Table 6) was not supported by the data. Figure 14 shows the percent correct scores
for each condition to perceptual, imagery and violation trials.

Two 8-way repeated measures analysis of variance (ANOVA) were conducted to test for
positive transfer. The differences examined were: speed of rotation (45°ls vs. 180°/s vs. two
trained speed conditions (90°/s) , one paired with fast speed testing and one paired with slow
speed testing), order of testing (fast/slow vs. slow/fast) and sessions (n26, to examine
acquisition) within each location tested (one ANOVA for 135° and one for 180°). Positive
transfer was obtained if there were no significant differences between percent correct scores
to the various novel and trained speed conditions.

BMW

For 135° testing, the ANOVA showed a nonsignificant effect of speed (F3508 - 2.169,

p>0.10) indicating that there were no significant differences between responses to novel and

trained speed conditions. In addition, the session effect was not significant in this ANOVA
(F536 . 0.745, p > 0.60). These findings are consistent with the argument that transfer to the

novel speeds was positive and immediate in 135° testing. In addition, the interactions of these
variables, as well as with order of testing, were not significant (see Table 7 for the ANOVA

statistics in detail).
A surprising result of this ANOVA was a significant effect of order of testing (F136 =

6.756, p < 0.02). Upon closer examination of the relevant data, it was clear that this was the

63

 

Mean Percent Correct
.5
O
l

 

 

 

 

 

 

Training 135° Fast 135° Slow

 

Mean Percent Correct
8

 

 

 

 

 

 

 

 

Training 180° Fast 180° Slow

 

 

El Percent Correct: Imagery
I Percent Correct: Violations

E] Percent Correct: Imagery
I Percent Correct: Violations

Figure 14. Mean percent correct scores to perceptual, imagery and violation trials.

64

Table 7. AI'IOVB ‘EDIO 10" "IO 135° T081 and "‘0 180° T081

Anova table for the 135° Test

Source: df: Sum of Squares: Mean Square: F-test: P value:
esting (A) 1 832. 832.
session 459.052 91.81
AB 4 0.992 94.198
subjects w. 4437.307 123.259
482. 160.686

343.1 114.383
1081. 2.12
1104. 1 3.

8 .6 4.

 

Anova Table for the 180° Test

Source: df: Sum of Squares: Mean Square: F-test: P value:
11 . 233.1
of .111 .111
572.34 114.468
w. groups 36 6 54. 187.628
) 1 .113 363.3

1067.43 71.162
318. 106.069
639.433 42.629
9 . 44 89.868

 

65

result of a change of response accuracy to the slow condition, depending upon whether it was
tested first or second. if if was tested first, the slow condition in 135° testing yielded an average
(taken across training and testing) of 69.77 percentage correct responses, but if it was tested
second (i.e., after the fast speed had been tested) it was met with an average 0176.01
percentage correct responding. Fast speed testing produced an average of 76.44 percentage
correct responses both when it was tested first and second.

Tukey comparisons of response accuracies to the trained and test conditions by order of
testing revealed that the significant effect of order was a result of a significant change in
responses to test trials and to training trials when the slow speed was first introduced.
Response accuracies to trained trials during slow speed testing were 61.54 and 66.42 in
correspondence with the slow speed test occurring first or second, respectively. There was a

significant reduction in response accuracy to trained trials when slow speed testing was
Introduced first ( q4’362 4.885, p <0.05 ). Average response accuracies to the slow speed
probes were 65.92 and 71.98 to first and second tests, and were significantly different by a
Tukey comparison (q 4,36 = 6.052, p < 0.05). All other con'parisons of probe data by order of

testing, as well as of training data by order of testing were not significant.

Figure 15 illustrates the relationships between training and test mean percent correct
scores by order of testing. It is evident in this illustration that the significant order effects found
in the Tukey comparisons were caused by response differences in the violation trials, but not in
the Imagery trials. In training and in testing, the order of testing did not differentiate percent
correct scores to imagery trials. However, the mean percent correct scores were quite different
with respect to the order of testing in violation trials.

In sum, the fast and slow speeds tested at the 135° location produced accurate
responding with no differences in general found between trained and tested conditions. By a
significant order of testing effect It was revealed that responses to trained and test trials within

slow speed testing were not quite as accurate as were those to the other speed conditions, but

Mean Percent Correct Scores

66

 

 

 

 

 

 

 

90
80 _
-[3- 45° probes: First
70 _ -I- 45° probes: Second
I
'0- 90° training: First
1 _| + 90° training: Second
60 q
50
IMAGERY VIOLATION

Trial Type

Figure 15. Mean percent correct scores to imagery and violation trials within
slow speed testing to compare first and second order of testing.

67

this effect was evident only when the slow speed was the first novel speed introduced, and,
only to the violation trials. These results imply that the pigeons could respond accurately to
novel movement conditions, but that there was more initial interference to trained and test
violation trials when a slower speed was introduced first.
Wat

The results of the ANOVA for the 180° location were slightly different than those obtained

in the ANOVA for the 135° location (see Table 7 for statistical detail). First, the order of testing

effect at 180° was not significant (F136 - 3.012, p > 0.05). But the speed effect was

significant (F3303 :- 4.043, p < 0.01), indicating that response accuracies to the fast, slow and

the trained speeds in 180° testing were different. This result works against the hypothesis of
positive transfer to novel speed conditions, since there was a significant difference found
between response accuracies to the trained and test speed conditions. Tukey tests of all the
meaningful comparisons of this effect were conducted (i.e., to response accuracies to trained
and test trials in the slow condition, to trained and test trials in the fast condition, as well as to
slow vs. fast test trials, and to the two trained conditions). All of these comparisons produced
nonsignificant calculated q values. After some post-hoc unplanned comparisons were made, It
was detennlned that the significant effect found in the ANOVA was the result of a significant
difference between a test and trained condition that were not meaningful to compare.
Specifically, a significance at the 0.005 level was obtained with a Tukey test of average
response accuracy to Wing (the bar on the far left of the 180°

histograms in Figure 13) and average response accuracy to W (the bar to the far right

in Figure 13, q 4’ 48 - 4.799). Because of this, the significant effect is spurious in nature.

The session effect in the ANOVA at thef80° location was nonsignificant (F536 - 1.243, p

> 0.30) and all the Interaction effects of speed, order of testing, and of sessions were

nonsignificant in this ANOVA, just as in 135° testing.

68

W

The enor data from Experiment 1 revealed that there was a curvilinear relation between
enors and rate of acceleration or deceleration presented during the delay in the movement
estimation task. Specifically, the error functions were increasing with a quadratic trend with
respect to increases in acceleration, and they were parallel with respect to location of testing,
with farther distances producing more enors.

The plotting of the enor data from Experiment 2 along a deceleratiorVacceleration axis
reveals any similarities between estimating movement at the rate of a trained speed and at
novel speeds. An obvious change in enor rates to violation trials as a function of location Is
displayed in Figure 16. Here, the mean enors to the imagery trials across new speeds are very
similar, but the error scores to the violation trials are quite diverse. One important implication is
that there are increased errors to tested decelerations with novel speeds. The functions seem
more symmetric with respect to discriminations of acceleration and deceleration than they did in
Experiment 1 with the trained speed. A relation which seemed to pop out of the graph Is shown
by the arrows drawn between the pairs of functions. These arrows point out a change in the
mean error rate as the location of testing is increased from 135° to 180° within speed testing.
While the functions no longer appear parallel with a constant enor rate separating them with
location increases as they did In Experiment 1, they still increase with increases in the distance
through which processing must occur. The new conjecture to add to the description of
movement representation is that with a novel speed, the errors increase and the shape of the

function changes with increases in location.

Discussion and Interpretations

The positive transfer to novel speeds in Experiment 2 indicates that pigeons'

representational process for estimating movement is quite flexible and fairly accurate. The

Mean Error Scores

 

69

 

38 q
36 _
34
fi
32 _
30 _
-EI- Errors to 180° Slow
. FAST SLOW
28 + Errors to 180° Fast
7 -|- Errors to 135° Slow
' + Errors to 135° Fast
26
I l I l I I I I I I I ‘l
-400 —300 -200 -100 0 100 200

Rate of DeceleratiorVAcceIeration

Figure 16. A comparison of mean enors to fast and slow probes at
135° and 180°.

70

positive transfer obtained supports two of the criteria of the definition of Imagery put forth at the
beginning of this paper: 1) the sthcturaI assumption , that pigeons maintain an accurate
perceptual description of the mov ing stimulus in memory , and 2) the functional assumption,
that pigeons can mentally transform the stimulus accurately even at the newly perceived rates.
The interactive assun'ption, that perception Interacts with imaging was not tested in this
experiment. However, the experiment did reveal that perception had a positive influence on
the imaging process, for perceived changes in velocity were immediately incorporated In the
image. Still, a timing strategy tuned to the perceptual display could account for the positive
results obtained.

Are the results of Experiment 2 consistent with the description of pigeons' representation
of movement of Experiment 1, or do they require a modification of the description of movement
representation developed so far?

The asymmetric discrimination between constant velocity, deceleration and acceleration
found in Experiment 1 was not supported by the pattern of errors in Experiment 2. Rather,
increased enors to increases in deceleration appeared as a new characteristic. However, the
data are insufficient to calculate a full function to describe this effect, so this finding is only
mentioned in passing.

The general finding that enors increase with increases in location is consistent across the
speed tests of Experiment 2, and this finding is consistent with the notion that an analog
process of estimating movement is employed. However, when new speeds were introduced In
Experiment 2, the enor functions increased and changed shape qualitatively with increases in
location, so the parallel-like functions (i.e., the constant increase in enors to increases in
represented distance) were not obtained.

Some qualitative changes in the process of representing movement occur when the rate
of movement is varied. With one trained speed (i.e., in Experiment 1), there Is an asymmetric

discrimination of constant velocity from acceleration and deceleration. There are increased

7 1
errors to increases in acceleration, and there Is a constant increase of enors to increases in the
distance through which processing must occur. With new speeds, the error functions become
more symmetric with respect to the discrimination between constant velocity and acceleration
and deceleration. There are still Increases in enors to increases in the distance of extrapolation,
although the shape of the functions also changes with increases In the distance of
extrapolation for novel speeds. Positive transfer was obtained to the novel speeds, or, in other
words, responding remained accurate to the novel speeds. However, the differences in the
functions produced by the novel speeds reveals complications in the discrimination resulting
from the novel speeds.

A direction of research extrapolated from studies of the psychophysics of movement in
humans may help resolve the complication In error rates caused by the introduction of novel
speeds. When adapted to a specific velocity (usually with moving sinusoidal gratings), people
perceive test velocities that are slower than the adapted rate as slower than their actual rates,
and test velocities that are faster than the adapted rate as faster than their actual rates (Pantle 8
Sekuler,1968; Carlson, 1962; Rapaport, 1964; Thompson, 1981; and for a review, see Anstis,
1986). Obviously, this adaptation effect canmt be applied directly to the present study, since
threshold detection methodologies were not employed and a full 15-s IT I Intenupted the
presentation of the ”adapted“ or trained velocity and the test faster and slower velocities.
Nevertheless, the speculation here Is that this adaptation effect may also occur at a higher
cognitive level; for example, an organism's judgment of movement may be a relative one, or
one that is affected by other observed movements in context. This type of effect could result in
qualitative shifts In discrimination efficiency to infrequent presentations of slow velocities or fast
, velocities within an ”adapted” velocity session. Systematic research Is needed that tests many
points along the deceleratiorVacceleration axis along with changes in velocity for this
speculation to be tested. It is offered here as a possible explanation for the qualitative shift

observed to changes in velocity In Experiment 2.

Experiment Three:
The Effects of Removing Perceptual Cues for Movement on the

Representational Process

The aim of this experiment was to test whether the perceptual display of stimulus
rotation was necessary for accurate responding in the task. On certain probe trials within typical
training sessions, the perceptual display of movement was removed from the stimulus
presentation. On these test trials, subjects perceived a static presentation of the stimulus at its
"start” location (0°), followed by an auditory cue for movement and a delay, and then by a test
location. Response accuracies to these trials demonstrated whether there was any influence of
the perceptual cues for movement on the representational process. The characteristic of the
process quantified by this experiment is its triggering feature. The results help to detennlne
whether the start location and a cue for movement are all that is needed, or if the start location
plus a perceptual display of movement are required.

In addition, this experiment attempted to test imagery again as the controlling process by
manipulating the type of information available for representation. A timing strategy (even one
tuned to the perceptual display, as in Experiment 2) would not necessarily be adversely
affected by the removal of the perceptual display of movement, as long as boundary start and
stop locations were presented. An imagery process would be affected, because subjects'
representation of the rotating stimulus would be restricted to a static display.

The same subjects were employed in this experiment as were used in the previous
experiment. To avoid the possibility that subjects would assume different speeds of rotation (as
was tested in Experiment 2) during the test condition in which perceptual cues for movement
were removed, responses to the trained speed were re-trained until criterion performance was

met, and the trained speed was the only speed presented in training and testing in Experiment

72

73
3. In addition, an auditory cue was added to the beginning of each trial in the training phase of
Experiment 3 to signal stimulus movement. The only kind of movement paired with the auditory
cue was the trained speed. This auditory cue was also presented in the test trials In which the
perceptual display for movement was removed, and it served the purpose of indicating when
the stimulus started its movement. The only type of movement that should have been cued by
the auditory cue was the trained speed. Thus the novel speeds tested in Experiment 2 should

not interfere with this task.

Method

Subjects
The same 8 subjects used in Experiments 1 and 2 participated in this study.

Apparatus
The two chambers used in Experiments 1 and 2 served as experimental chan'bers in this

study. Subjects tested In Chamber 1 in previous experiments continued to receive training
and testing in Chamber 1, and subjects trained In Chamber 2 continued In the experiment in
Chanber 2.

Emedum

The variable of Interest in this experiment was response accuracy to trials in which the
perceptual display of stimulus rotation was removed. The procedure was very similar to that
followed in all the other experiments: first subjects were trained to respond accurately to 135°
and 180° training trials, and then test trials were presented infrequently within training sessions
and response accuracy to these trials was noted. However, the trial progression in training and

testing was slightly altered. The standard trial progression had been: a static display at 0° is

74
presented for 5 s, and following a VR-12 to 15 on the middle key, stimulus movement was
presented, followed by a delay and/or the presentation of the test location. In this experiment,
a 2000 Hz tone was presented for 500 ms after the response requirement was met to the static
presentation at 0° and before stimulus movement was initiated. Then, in the test trials, the tone
was presented after the response requirement was met, but following it, no perceptual display
of movement was presented. The auditory cue was added to insure that subjects had a clear
and unvarying cue for the start of stimulus movement (i.e., the response requirement was
variable) and for the type of stimulus movement employed. It was added to training and testing
so that the only variable manipulated in testing was the display of perceptual cues for
movement.

The test trials in this experiment were 135° and 180° imagery and violation trials modifie
from training. The modification was that In the test trials, a static presentation of the clock hand
was presented at 0°, and after the middle key response requirement and the tons, a delay
during which the stimulus was absent was initiated until the test location was presented. The
simulated rotation from 0° to 90° that was standard in the training trials was removed in these
test trials. Figure 17 diagrams the trial types for training and testing.

First, the eight subjects in this experiment were exposed to alternating sessions of 135°
and 180° training until the criterion performance of 80 percent correct for two consecutive days
was met. In these training sessions, reinforcement contingencies were the same as were
described in Experiments 1 and 2 for training. Sessions included 64 trials, consisting of 16
perceptual trials, 16 imagery trials, and 32 violation trials.

Next, a series of test sessions was presented for 12 days. These test sessions were
corrprlsed chiefly of 135° or 180° training trials (i.e., 64 out of 72 total trials were training trials)
The remaining 8 trials at either location were comprised of imagery (2) and violation test trials (4)
that did not present stimulus rotation, and 2 perceptual trials. The imagery and violation test

trials progressed In the following manner: a static presentation of the clock hand was presented

75

Training Trials

Perceptual Imagery Violation

1.5, 1.0: La!
{-0.53 {-1.43
2.49 La La,
6!-a9 60-5,
Test Trials
Perceptual Imagery Violation

._
3 MC)...

Figure 1 7. Training and test trials used in test of perceptual cues
for movement in Experiment 3.

 

 

U

 

 

 

:7

   

 

 

 

76
at 0° for 5 s, and following responses on a VR-12 on the center key, a 2000 Hz tone was
presented for 500 ms. Next, the stimulus disappeared during a delay (i.e., one which included
the time taken for stimulus rotation from 0° to 90° and rotation to the test location) and then the
stimulus reappeared at a test location. These trials were reinforced under the same
contingencies as those in training: imagery trials were reinforced for left key responses and
violation trials were reinforced for right key responses.

During testing, 135° and 180° training were alternated daily. Test trials at 135° were
presented in 135° training sessions, and test trials at 180° were presented in 180° training
sessions. A total of 24 exposures to imagery and perceptual trials, and 48 to violation trials was
planned in the test series. The data of interest were percent correct scores to the novel test

trials, and enor scores to those trials.

Results

Figure 18 shows the mean percent correct scores to trained and test trials at the 135°
location and at the 180° location, and Figure 19 displays the means to perceptual, Imagery and
violation trials within these conditions. The means to trained trials appear greater than those to
test trials, although all means appear lower than criterion (i.e., less than 80 percent conect
responding).

A two-way repeated measures analysis of variance (ANOVA) was conducted to compare
percent conect scores from training and testing, and across sessions to each location
condition. A significant difference in response accuracy between training and testing, with test
accuracy significantly lower than accuracy to trained trials, supports the hypothesis that the
removal of perceptual cues for movement debilitated performance. The analysis for each

location is discussed in separate sections and is presented in Table 8.

Percent Conect Scores

Percent Correct Scores

77

 

 

 

 

 

 

 

 

80 -
60 u
40 - El 135° train
‘ I 135° test
20 s
0
Visible Static
stimulus Displays
movement
so 4
60 -
40 "l
. CI 180° train
I 180° test
20 .-
1
o
Visible Static
stimulus Displays
movement

Figure 18. Mean percent correct scores to the trained and test conditions
in Experiment 3.

78

135° Scores

I Perceptual

‘ 5' Violations

Mean Percent Conect

 

 

Training Test: Static Displays

10° ‘ 180° Scores

I Perceptual

Violations

Mean Percent Correct

 

 

Training Test: Static Displays

Figure 19. Percent correct scores to perceptual, imagery and violation trials.

79

Table 8. Analyses of Variance of Percent Conect Scores in Experiment 3.

Analysis of Variance of Percent Conect Scores to Trained and Test Conditions to 135° Trials

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Source: df: Sum of Squares: Mean Square: F-test: P value:
SESSION (A) 5 1261.069 252.214 1.17 33977
data w. groups 42 9050.762 215.494
Trained vs.JTest (s) 1 53.417 53.417 .423 .5188
AB 760.86 152.172 1.206 .3229
B xdata w. groups 42 5298.556 126.156

Analysis of Variance of Percent Conect Scores to Trained and Test Conditions to 180°

Source: df: Sum of Squares: Mean Square: F-test: P value:
SESSION (A) 5 1046.246 209.249 .979 .4416
data w. groups 42 8974.183 213.671
Trained vs. Test (8) 1 1147.821 1147.821 6.465 .0148
"H3 848.239 169.648 .956 .4558
B x data w. groups 42 7456.815 1fi.543

 

 

 

80
10113521861
In the ANOVA at 135°, a significant difference was not obtained between trained and

test mean response accuracies (F132 - 0.423, p > 0.50). This result indicated that the removal

of the perceptual cues for movement had no effect on response accuracy. The mean percent
conect scores to trained and test trials at 135° were 63.679 and 62.187, respectively. Since
response accuracy remained between 75 and 80 percent conect before testing, these scores
indicate that there was a general suppression effect of conect perionnance during testing. To
evaluate this, two two-tailed t-tests contrasting the trained and test mean percent correct

scores with 50 percent conect responding (I.e., the level of chance) revealed that perionnance

to trained trials was significamly higher than chance (t8 - 4.133, p < 0.005) but that

performance to test trials was not significantly different from chance (is .. 2.1878, p >0.05).
Although the deleterious effect of the removal of perceptual cues for movement was not
revealed in the ANOVA, these t-tests reveal that performance to trials with the perceptual cues
for movement removed was worse than was performance to trained trials.

The effects of sessions, as well as the Interaction between trained vs. test trials and

sessions were not significant in the ANOVA at 135°. Also there were no significant differences
found In an analysis of variance testing for subject differences in training (F7'4o - 1.28, p >
0.20) or in testing (F140 - 0.848, p > 0.50).
W

The repeated measures ANOVA at 180° revealed a significant difference between
percent conect scores to trained and to test trials (F1,42 - 6.46, p < 0.02). The mean score to

trained trials was 67.74, and the mean score to test trials was lower: 60.825. The result is
consistent with the hypothesis that the removal of perceptual cues for movement in test trials
debilitated performance. The sessions effect was not significant, nor was the interaction.

At this location of testing, both trained and test means were significantly greater than

8 1
chance level responding would predict (18(train) . 2.948, p < 0-013‘8(test) a 3.347, p < 0.005).
Still, the result of the ANOVA clearly indicated that percent correct scores to test trials were
significame lower than those to training trials.
An analysis of variance conducted to examine subject differences in training and testing

revealed significant differences between subjects' training scores (F7’4o = 2.52, p < 0.05) and

between subjects' test scores (F140 = 3.003, p < 0.02). These effects were brought on by the

extremely low percent conect scores of two subjects: labelled #25 and #4. The trained means
for these two subjects were 53.98 and 52.32, respectively, while the means to trained trials of
all other subjects ranged between 60.00 and 73.78 percent conect. In addition, the test means
of these two subjects were 51.19 and 65.47 respectively, while all others' test scores ranged
between 62.00 and 81.78 percent correct. These subjects were clearly responding at chance
levels to the trained trials, and entered artificially low values into the test data in the ANOVA.
For the purpose of testing the amount of transfer to the test condition as a mean effect
of the experiment (i.e., despite the chance performance of two subjects), another two-way
repeated measures ANOVA was conducted with the data from these two subjects removed.

The results of this ANOVA were virtually the same as the results of the first: a significant
difference was obtained to percent conect scores from trained and test trials (F130 a 7.29, p <

0.02) and nonsignificant differences were obtained to sessions and to the interaction. The
mean scores to trained and test trials were increased from the first test: the trained mean for the

6 remaining subjects was 70.876 and the test mean was 63.382. These scores were also
significame different from chance (16(train) - 3.299, p < 0.01; t6(test) - 4.55, p < 0.005).
However, the subject effect found significant in the first analyses of variance disappeared after
removal of the two subjects (F530 (trained) a 1.94, p > 0.10; F530 (test? 2.14, p > 0.08).

Regardless of the disparate data of two subjects, the significant effects found at 180°

hold: a significantly lower response accuracy to test trials than to training trials. The addition of

82
the data from the two subjects merely produced a significant subject effect, but did not occlude
the deleterious effect of removal of perceptual cues for movement on performance.
WM

Next, mean errors for all eight subjects were plotted as a function of the speed of
rotation evident in the delay for each trial type. Mean enors to imagery trials were plotted at the
0-point of the declaration/acceleration axis seen In Figure 20. The mean enors to the violation
trials were graphed on either side of the 0-point. When matched with the enor results of
Experiment 1 , the mean errors from the trained and test conditions of Experiment 3 appear
quite different. Specifically, errors to the imagery trials seem inflated and the mean errors to the
violation trials are less, not more, than the enors to the imagery trials. These functions clearly do
not possess the parallel and linearly increasing characteristics of the functions of Experiment 1.

On closer examination (see Figure 21), there is a pattern to the enors seen in
Experiment 3 that describes the strategy subjects followed In testing. With the introduction of
the new delays inadvertently caused by the removal of all perceptual cues for movement,
subjects adopted a simpler, albeit less accurate, strategy of response: If there is not a delay,
peck the left key; but if there is a delay, peck the right key. This strategy explains the inflated
error rates to imagery trials, for in training and testing, subjects responded on the right key (and
inconectly) more often than on the left to any delay. This strategy also explains the more
accurate results (and fewer errors) to the violation trials for which right key responses were
reinforced.

It is important to note that this strategy was not followed before the removal of perceptual
cues for movement, for subjects were responding between 75 and 80 percent correct to 135°
and 180° training trials before testing. The removal of perceptual cues for movement on probe
trials was so dismptive that subjects abandoned their cognitive strategy for estimating
movement and adopted an S-R strategy of response similar to that first applied in preliminary
Experiment 1.

Mean Errors

 

83

80

{l- 135° test
+ 135° trained

-|- 158°: Exp. 1

 

.0. 180°: Exp. 1
-I- 180° test
{1- 180° trained

 

 

I I I 1 T I I I T j
-100 O 100 200 300 400

Rate of Deceleration/Acceleration

Figure 20. Mean errors to trained and test trials by speed of movement during the
delay in Experiment 3, compared to errors made in Experiment 1.

Mean Errors

84

 
 
 
 

 

 

80
70_
I {i- 135° test: no movement cues
60 + 135° trained
" -|- 180° test: no movement cues
'0' 180° trained
50
4o_
30_
20
' l ' I 1 I T I ' l
-100 o 100 200 300 400

Rate of Deceleration/Acceleration

Figure 21. Mean errors to trained and test trials plotted by speed of movement during
the delay.

85

Discussion

Experiment 3 was devised to examine the influence of perceptual cues for movement
on the representational process. Removal of the perceptual cues in probe trials resulted in a
complete shift of the strategy from cognitive estimation to a less efficient S-R timing strategy.
Percentage of correct response scores appeared low for trained and test trials, but the most
debilitating effect was to responses to the test trials. The error data revealed the SR strategy
more clearly, due to the inflated enor rates to Imagery trials and the more accurate response
patterns to all violations.

When perceptual cues for movement are present, subjects estimate the location of the
stimulus by imaging its movement In its absence. When perceptual cues are
absent, a simple strategy is employed in which right key responses are emitted following any
delay.

Perceptual cues for movement are a necessary ingredient for representing movement.
Further, static stimuli, even when timed accurately to convey movement, are not salient cues
for movement in pigeons. Transformational studies in which static stimuli are employed would

most likely fail, and poor results have been obtained in almost every case (Cerella,1987).

Summary and Conclusions

A goal of this Investigation was to put to a test the characteristics of an imagery process.
The structural assumption, that imagery maintains the perceptual characteristics of the stimulus
held in memory, was examined in Experiment 1 by a manipulation of the delay intervals which
supported the timing component of remembered stimulus movement. Positive transfer, or

accurate responding to these delays, was found, but the finding was weakened somewhat by

86
the disparate data within the tested violation trials. This experiment was a fine-grained analysis
of the structure of movement in the representational process by a manipulation of delays, and It
revealed that the representational process did not always accurately Incorporate subtle
changes in timing. Imagery was implicated as the process used in Experiment 1 however, by a
general effect of transfer across training and test trials.

The functional assumption, that imagery allows for appropriate mental transformations of
stimuli, was tested in Experiment 2 by an observation of the effects of new velocities on the
process. Positive transfer was observed to fast and slow speeds at 135° and at 180° . The
representational process accommodated new transformations at different speeds quite readily.
The functional assunption of Imagery was supported. In addition, the timing strategy was
refuted as an alternative which accounted for pigeons' performance thus far, since the
predicted negative transfer of the timing strategy was not realized in the data.

Finally, the interactive assunption, that perception interacts with imaging, was not
tested in any of the experiments. The Influence of a perceptual display on the representational
process was tested in Experiments 2 and 3, by a manipulation of the speed presented in the
perceptual display (Experiment 2) and by removal of the perceptual dynamic display
(Experiment 3). It was found In Experiment 2 that the representational process used could pick
up new information about movement in the perceptual display and use It immediately and
readily in the memory process. It was thought that this evidence supported the imagery
hypothesis better than a timing strategy because in Imagery, perceptual input of the dynamics
of movement would be represented realistically, while In timing only particular cues and times
would be incorporated. Still, a timing strategy tuned to the perceptual dynamic display could
account for the transfer results. The imagery process was again tested in Experiment 3, by a
measurement of the effect of removing perceptual cues for movement on the process. The
result of this manipulation was a conplete shift In the pigeons' strategy from an estimation

strategy to an S-R strategy that was not as accurate. Clearly, removal of the perceptual cues

87
had devastating results on the representational process. It appears as if the representational
process pigeons enployed required a perceptual display for movement for accurate
estimation, and since timing would not necessarily require a dynamic display of movement,
imagery is thought to be at work here.

Imagery had been defined operationally and two criteria to identify its existence were
tested. It Is possible that a timing strategy tuned to the perceptual display and requiring
perceptual input for timing calibration could account for the data, thus Imagery Is not Isolated as
the process by the data. The more parsimonious explanation supported here is that imagery is
the process used to represent stimulus movement because the timing strategy does not
support all the changes in enor rates very easily and because It should not be necessary for a
timing strategy to require perceptual dynamic input If timing is preserved in the static input
enployed. Still, the question is empirical, and begs a return to the Interactive assumption, that
perception Interacts with imaging, for its answer. An experiment In which perceptual Input is
given while subjects are imaging can reveal whether perceptual input can facilitate or debilitate
the representational process. A timing strategy might be influenced by perceptual input during
timing, but it should not be as debilitated as would an imagery strategy it that Is what subjects
are using.

Another goal of this Investigation was to attenpt to quantify certain characteristics of
movement as it Is represented in pigeons. The question of interest in this investigation was to
what extent pigeons' representation of movement complied with the laws and characteristics of
real physical movement. The results of the error analyses in the three experiments delineate
certain characteristics of the representational process. A description of pigeons' ability to
represent movement, as reflected by the results of these experiments, follows.

To estimate movement at all, the subjects needed a perceptual display for movement
(Experiment 3). If one was not presented, subjects responded on the basis of the presence or

absence of a delay. After a perceptual display was provided and once the moving stimulus

88
disappeared, subjects transformed a representation of the stimulus through space to its test
location, and they made increasingly more enors as the test location was increased
(Experiments 1 and 2). To tested rates above 44°ls2, or to very short delays, subjects adopted
a simple delay strategy (Experiment 1). Within their bounds for testing accelerations, subjects
have a category for constant velocity in which some accelerations and few decelerations fit.
Finally, if a change in visible stimulus movement is noted, this change is accurately represented
in the stimulus movement (Experiment 2). And, subjects still make more errors as they
transform the stimulus at the new rate through farther distances, but there is a qualitative shift in
the increase in response to novel speeds.

This description of movement representation is suggested by the data from the three
experiments. A full description of movement as it is represented requires that more is
understood about pigeons' perception of movement and movement changes. In addition, a
more systematic evaluation of represented movement as it changes over time and distance
would test the description offered; certainly it would identify whether an acceleration constant
is applied in the process, or whether there is some category for movements applied.

An investigation of movement representation allows for the development of models of
perceptual-cognitive function in a direction so far Ignored in animal work. In pigeons, as well as
in many studied species (macaque monkeys, cats, other avians such as owls), much is known
about the neural networks and brain structures involved in movement perception (Granda &
Maxwell, 1979). The stmcturalist view of perception , that percepts are merely additive
combinations of color, light, edges, bars, and movement, controls this kind of Investigation.
While this point of view lends itself to the technique of neural investigation of percepts In intact
organisms possmle at this time (primarily intracellular recordings) , it may not appropriately
describe the perceptual experience of the organism (Hochberg, 1980). It is clear from
experiments of Illusions and of gestalt psychology with humans that higher level rules of

perceptual organization are involved that allow for object constancy across changes in size,

89
orientation, and movement. Object constancy despite change In the environment is a
perceptual phenomenon that has clear survival value and there is some evidence that it exists
in species other than humans, and specifically in pigeons. An investigation of a moving object
as it Is perceived and as It Is represented begins to explore these issues because In such an
investigation, transformations of an object that maintain object constancy are identified. Again,
as in the investigation of imagery in pigeons, the goal is to build a model of

perceptual-cognitive function.

APPENDIX A

A test of parallelism by nonforced multiple regression analysis.

Stepwise Regression Y1 :Error Scores 5 x variables

 

F to Enter

F to Remove
Number of Steps
Variables Entered
Variables Forced .0

Residual Inionnation Table
SS[e(i)-e(i-1)]: e 2 0: e < 0: DW test:

575263.306 [223 [209 [1.597 [

 

 

 

 

predictor

 

 

 

 

Stepwise Regression Y1:Error Scores 5 x variables

 

 

 

 

 

STEP NO. 1 VARIABLE ENTERED: x4: parallel.acc‘loc
R: R-squared: Adj. R-squared: Std. Error:
[.543 [.295 [.294 [21.328 [
Analysis of Variance Table

Source DF: Sum Squares: Mean Square: F-test:
LREGRESSION [1 [81974.108 [81974.108 [180.213
BESIDUAL [430 [195595.167 [454.872 [
[TOTAL [431 [277569.275 [ |

 

STEP NO. 1 Stepwise Regression Y1:Error Scores 5 x variables
Variables in Equation

 

 

Parameter: Value: Std. En.: Std. Value: F to Remove:
FNTERCEPT [ 27.593 [ [ |
[ parallel.acc'loc [.001 [.0000928 [.543 | 180.213

 

Variables Not in Equation
Parameter: Par. Corr: F to Enter:

SPEED -.046 .914
-.22 21.899

location .25 28.654
-.249 28.444

 

90

91

Stepwise Regression Y1:Error Scores 5 x variables

 

 

 

 

 

STEP No.2 VARIABLE ENTERED: x3 : location
R: R-squared: Adj. R-squared: Std. Error:
[.583 [.339 [.336 [20.673 ]
Analysis of Variance Table

Source DF: Sum Squares: Mean Square: F-test:
[REGRESSION [2 [94220.323 [47110.162 [110.228
[ RESIDUAL [429 [ 183348.952 [427.387 |
[ TOTAL [431 [277569.275 [ [

 

STEP NO. 2‘ Stepwise Regression Y1 :Error Scores 5 x variables

Variables in Equation

 

 

 

 

 

 

Parameter: Value: Std. En.: Std. Value: F to Remove:
[INTERCEPT [1.089 I I I
[ location [.298 [ .056 [ .211 [28.654
[ parallel.acc*loc | .001 [.00009054 | .519 [173.008
Variables Not in Equation
Parameter: Par. Corr. F to Enter: f
[ SPEED [.092 [3.654 [
[speed sqrd [-.161 [11.412 [

 

[quad.accz‘loc [-.245 [27.309 [

(L881 Step) STEP NO. 3

92

Stepwise Regression Vi :Error Scores

VARIABLE ENTERED:

5 X variables

x5 : quad.acc2*loc

 

 

 

 

 

R: R-squared: Adj. R-squared: Std. Error:
[.616 [.379 [.375 [20.067 ]
Analysis of Variance Table
Source DF: Sum Squares: Mean Square: F-test:
[REGRESSION [3 [105217.548 [35072.516 [87.095
| RESIDUAL [ 428 [172351.728 [402.691 |
[ TOTAL [ 431 [277569.275 [ [

 

STEP NO. 3 Stepwise Regression Y1 :Error Scores

Variables In Equation

5 X variables

 

 

 

 

 

 

 

 

 

 

 

 

Parameter: Value: Std. Err.: Std. Value: F to Remove:
INTERCEPT 1 .089
location .284 .054 .201 27.518
parallel.acc‘loc .003 .0002863 1.141 83.417
quad.accz'loc -.000004436 8.488E-7 -.651 27.309
Variables Not in Equation

Parameter: Par. Corr: F to Enter.

[SPEED [.095 [3.88 [

[speed sqrd [.084 [3.004 |

 

 

APPENDIX B

Cumulative Record of Sessions

in Experiments 1, 2, and 3

93
APPENDIX 8

Cumulative Records of Sessions In Experiments 1, 2, and 3

BIRD# 7 8 9 25 2 14 4 10
Sessions to 10 24 11 12 14 21 16 32
criterion
Exp1 18 18 18 18 18 18 18 18
Sessions to 4 10 16 5 7 6 12 16
criterion
Exp2(first) 12 12 12 12 12 12 12 12
Sessions to 6 9 2 12 3 12 11 6
criterion
Exp2(sec.) 12 12 12 12 12 12 12 12
Sessions to 12 3 6 19 8 13 20 17
criterion
Exp3 12 12 12 12 12 12 12 12

TOTAL 86 100 89 102 86 106 125 113

APPENDIX C

SUBJECT TESTS

Experiment 1: ANOVA of subject differences and location differences.

 

 

 

 

 

Source: df: Sum of Squares: Mean Square: F-test: P value:
subject (A) ‘7 3355.864 479.409 .335 .9341
data w. groups 48 68735.713 1431.994
location (B) 2 14095.342 7047.671 10.455 .0001
AB 14 12915.33? 922.524 1.369 .1837
B x data w. groups 96 64710.155 674.064

 

 

 

 

 

 

 

 

Note: Subject differences are not significant (p - 0.9341).

Experiment 2: ANOVA of subject differences and speed condition (135°)

 

 

 

 

 

 

Source: df: Sum of §laresz Mean Square: F-test: P value:
Subject (A) 7 1 . 223.454 1.59 1063')
data w. groups 40 5353.652 133.841
Repeated Measure (B) 3 577.625 192.541 2.592 .063
AB 21 2193.279 104.442 1.406 .1051
Bx data w. groups 120 8914 74.283

 

 

 

 

 

 

 

 

Note: Subject difference Is not significant (p - 0.2003).

Experiment 2: ANOVA of subject differences by speed condition (180°)

Source: df: Sum of Squares: Mean Square: F-test: P value:
2948.844 421.264 2.
data w. 4 8101.99 2.
3 482. 1 . 2.36

21 2386. 113. 1.6 4
120 8145. 6 .881

 

Note: Subject difference Is not significant (p - .1044).

94

95

Experiment 3: ANOVA of subject differences and training vs. testing (135°)

P value:
.3043

F-test:
1 .24

:Mean
262. 09
211.822
53.41 .384
71. 71 . 1
139.03

Sumof
1838.
4 84 .866
1 53.417
4 .1

5 1.218

df:

      

Source:

         

TA w.

AB
x w.

4
Note: Subject differences are not significant (p a 0.3043).
Experiment 3: ANOVA of subject differences and training vs. testing (180°)

 

 

 

 

 

Source: df: Sum of Squares: Mean Square: F-test: P value:

‘ SUBJECT (A) 7 4835.245 690.749 5.329 .0002

. data w. roups 40 5185.184 129.63
Test Condition (B) 1 1147.821 1147.821 6.23 .0168
AB 7 934.99 133.57 .725 .6517
B x data w. groups 40 7370.064 184.252

 

 

 

 

 

 

 

 

Note: Subject difference IS significant (p = 0.0002).
Analyses reported in dissertation are one-way ANOVA's for training and testing.
These are presented next, and then the differences when #25 and #4 are removed

are shown.

One-Way ANOVA testing subject differences at 180° (n - 8).

 

 

 

 

 

   
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Source: DF: Sum Squares: Mean Square: F-test: f
3 Between roujig [1415.326 [202.189 [3.003 [
é Within roups I40 [2693.029 [67.326 Ip-.0125 I
F ITotal [47 [4108.355 I I I
Source: DF: Sum Squares: Mean Square: Host:
8 Between rouds7 [4354.909 [622.13 [2.523 [
,2 Within roups [40 [9862.219 [246.555 Ip-.O301 |
[47 [14217.128 I I I
One-Way ANOVA testing subject differences at 180° (n - 6).
8, Source: DF: Sum Squares: Mean Square: F-test:
:1: I4 5 .112 I1.93_8____1
g I1441 6 -.1174
I1goz.zs I
o r ; DF: m _. al': ° Mean ua ' F- es:
“tween oorou 326.691 465.338 2.142
‘r- I 3_5 9.815411

 

 

LIST OF REFERENCES

LIST OF REFERENCES

Anstis, S. (1986). Motion perception inthefrontal plane Sensory aspects. (Chapter 16). In

KR Boll. L Kaufman8-JP ThomastEdSI BandbcclmLEemsmnnandhuman
Bannnnanga, New York: John Wiley 8 Sons, Inc.

Bugelski, BR. (1970). Words and things and images. Amanganfimnglggm 1002-1012.

Carlson, V.R. (1962). Adaptation in the perception of visual velocity. .igumamLExpanmamal
2620001002. 55. 192-197.

Carter, D. E., 8 Werner, T .J. (1978). Complex Ieaming and information processing by pigeons:
A critical analysis WIS) 565- 601

Cerella, J. (1987). Shape constancy In the pigeon: The perspectivetransfomtations
deconposed. In M. L. Commons, S. M. Kosslyn, 8 R..J Herrnstein (Eds.,) Eafiam
. . . = . . - ,Canbridge, MA: MIT

 

Cole, R. W. (1907). Conceming the intelligence of raccoons. W
W. 11. 211-261.

Cooper, LA. 8 Shepard, RN. (1973). The time required to prepare for a rotated stimulus.
Winn. 1. 246-250.

Dodd, E., 8 Wirth, A (1953). Differentiation between rods and cones by flicker

electroratinocraphy in pigeon and guinea Pig MSW
80—89.

Freyd, J. J., 8 FII'IKO, R A. (1984). Representational momerrtun. W
W. 10. 126132.

Granda. AM. 8 Maxwell. JH- (1979). Wu. New York:
Plenum Press.

Griffin, DR. (1981). W New York, NY: The Rockefeller
University Press.

Hebb, DO. (1968). Concerning imagery. WM 466-477.

Hennstein, R. J. (1984). Objects, categories and discriminative stimuli. In H.L. Roitblat, T.G.
Bever, 8 HS. Terrace (Eds.),AnimaLQani1ign, Hillsdale, NJ: Lawrence Erlbaum
Associates.

Herrnstein, R. J. (1985). Riddles of natural categorization. In AnimaLiniaiiiganga, New York:
Oxford University Press.

96

97

Hollard, V.D. 8 Delius, JD. (1982). Rotational Invariance in visual pattern recognition by
pigeons and humans. Saianga, 218, 802-804.

Jagacinski, R ..,J Johnson, W. W. 8 Miller, R. A. (1983). Quantifying the cognitive trajectories of
axlrapOIated movements. Jammnmemalflsxcmmmmanflramm

W 43-57
Jerison, H.J. (1973). W New York, NY: Academic Press.

Kaiser, M. K., Proffitt, D. R., 8 Anderson, K. (1985). Judgments of natural and anomalous
trajectories in the presence and absence of motion. W
W 1.1. (4) 795603

Kosslyn, SM. (1980). [magaancudm Cambridge, MA: Harvard University Press.

Lachman, R., Lachman, J.L. 8 Butterfield, E.C. (1979).mgni1iya_Esygm|ggy_ang_inignnaiign
Engaging, Hillsdale, NJ: Lawrence Erlbaum Associates.

McCloskey, M., 8 Kohl, D. (1983). Naive physics. The curvilinear impetus principle and its role

in interactions with moving cbIects .lournalgLExarirnataLEsxcanIauarning.
Mama. 2. 146-156

Metzler, J. 8 Shepard, R. N. (1974). Transformational studies of the internal representation of

three-dimensional cbiects In R L Solsc. Ed W
W Potomac, MD. Lawrence Erlbaum Associates.

Neiworth, J..J 8 Rilling, M. E. (1987). Amethod for studying Imagery in animals. .[gumaLgi
; ‘ - ;. 11(3), 203-214.

 

Pantle, A.J. 8 Sekuler, R. (1968). Contrast response of human visual mechanisms sensitive to
orientation and direction of motion. W 2, 397-406.

Pinker, S. (1984). Visual cognition: An introduction. W163.

Posner, M. E., Pea, D., 8 Volpe, B. (1982). Cognitive-neuroscience: Developments toward a
science ofsynthesis. an. Mehler, M. Ganett8 E. Walker (Eds) W
lI-ll- :‘0‘ :0'.t.l. ." tilt: II- -I0 l-tt‘t- 0 0 OOII 3.0.0;-

WHiilsdale, NJ: Lawrence Erlbaum Associates, pp. 251-276.

Powell, R.W. (1967). The pulse-to-cycle fraction as a detenninant of critical flicker fusion in the
pigeon. WISt-ISO.

Powell, R. W. 8 Smith, J. C. (1968). Critical flicker-fusion thresholds as a function of very small
pulse-to-cycle fractions. W 35-40.

Rapaport, J. (1964). Adaptation In the perception of rotary motion. W
W 61. 263-267.

Regan, D.R., Kaufman, L. 8 Lincoln, J. (1986). Motion in depth and visual acceleration.
(Chapter 19). In K.R. Boff, L. Kaufman 8 JP. Thomas (Edam
angfiumanfiangnnanga, New York: John Wiley 8 Sons. Inc.

98

Roitblat, H. L. (1982). The meaning of representation in animal memory. Ibafiahayigraland
Brainficisnasj. 353406

Rosenbaum, D. (1975). A perception and extrapolation of velocity and acceleration. ..[gumaLgi
.: . . . . : 1, 305-403.

 

Runeson, S. (1974). Constant velocity: Not perceived as such. Eambglggjaalflaxim 31,
3-23.

Runeson, S. (1975). Visual prediction of collision with natural and nonnatural motion functions.

W 18. 261 266.
Schwartz, B. (1984). EsygnglggygLLaaminganiBanayign New York, NY: WW. Norton 8 Co.,

Inc.

Shepard, R. N. (1975). Form, formation, and transfomlation of lntemal representations. In R.

SoISO. Ed. WWW. Hillsdale.
N. J.: Lawrence Erlbaum Associates.

Shepard, RN. (1984). Ecological constraints on internal representation: Resonant kinematics
of perceiving, imagining, thinking, and dreaming. EsygnglggigaLBayiaw, 91, 417-447.

Shepard. FIN. & Cooper. LA. (1982).MataL1rnaga_andJ]Jsirlranst0rrnatigns Cambridge.
MA: MIT Press.

Skinner, BE. (1984). Canonical papers of BF. Skinner. WW2,
473-764.

Sokal, RR. 8 Rohlf, F.J. (1981). fiigmany (2nd edition), San Francisco, CA: Freeman 8 Co.

Tenace, H. S. (1985). Animal cognition: Thinking without language. In L. Weiskrantz (Ed.,)
AnimaLimaiiiganga,Oxford, England: Clarendon Press, 113-128.

Thompson, PG. (1981). Velocity after-effects: The effects of adaptation to moving stimuli on
the perception of subsequently seen moving stimuli. Winn, 21, 337-345

Thorndike, E. L. (1898). Animal Intelligence: An experimental study of the association
processes in animals EsxcnolggiaLBaimMaaracns. 2. (Whole No 8)

Treisman, A. (1986). Properties, parts and objects. (Chapter 35). In K. R. Boff, L. Kaufman 8

J P Thomas (Eds )hangbggkgtflercsntignangilumanflsrfgrmana. New York:
John Wley 8 Sons, Inc.

Woodworth, RS. 8 Schlosberg, H. (1959). WWMw York: Holt, Rinehart
8 Winston.

Wundt. W. (1894). LaturagnhumanandAnimaLEsxcnglggy. Translated by 813- Creighton
8 EB. T‘Ici'lner. New York, NY: Macrnillian.

Zentall, TR, Edwards, C.A., Moore, B..,S 8Hogan, DE. (1981). Identity: Thebasisfor both

matching and oddity learning in pigeons .lgumaLgLExanmenmLEsycnglmnimal
W22. 70-86

 
 
    

   

  

MMMMMMM

Iiiijiliil Ililillilllijii‘liljiil

293 03

RIES

ill