‘ . fit»

“an

an ,
H... .
«t. IV."
3 .PA.

a ‘5

(u
a. 3...;
1. u... 45‘
a. #4933? .

I
in.

33!. .1
32.30018:
1 l .

, . . I.
, I :
1% u
‘ 14??)
§

(9.

winnrwdrdfl;

ic’fiiatfx‘ll

p.19.) .. .\§
iii-:0... bit:

‘sp'lnvl.’«um; (‘1
v.) . its}

 

 

. ,Efifiafifii

)1 .t:

 

«Hrgmfifiwk.

 

qov’v

This is to certify that the

dissertation entitled

Changes in lateral bias associated with
changes in task difficultyfor the perception

of ch1mer1c faces: Which cognitive processes
are respon81b1e?

presented by

Timothy J. Carbary

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Psychology

ééiuflmraidZL4a7£¢~NaJ

a'or,professor
ui

Lauren 1us Harris

Date ?/8/Z‘N L

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRClDateDuopes-p. 15

CHANGES IN LATERAL BIAS ASSOCIATED WITH CHANGES IN TASK
DIFFICULTY FOR THE PERCEPTION OF CHIMERIC FACES:

WHICH COGNITIVE PROCESSES ARE RESPONSIBLE?

by

Timothy James Carbary

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

DOCTOR OF PHILOSOPHY

Department of Psychology

2002

ABSTRACT
CHANGES IN LATERAL BIAS ASSOCIATED WITH CHANGES IN TASK
DIFFICULTY FOR THE PERCEPTION OF CHIMERIC FACES:
WHICH COGNITIVE PROCESSES ARE RESPONSIBLE?
By
Timothy J. Carbary

In the flee-viewing chimeric faces test (CFT), adults make judgments about
mirror-image chimeras of human faces. On this test, adults usually rely more on cues in
their left visual hemispace so that their judgments show a left visual hemispace, or LVH,
bias. The bias has been explained as a product of a change in the balance of arousal or
activation between the cerebral hemispheres leading to an attentional shift to the side of
space contralateral to the more aroused hemisphere. The LVH bias thus is seen as a
product of the greater arousal of the right hemisphere.

Like lateral biases in other perceptual tasks, the CFT bias is known to vary in
strength and even direction, for individual stimuli as well as for some individual subjects.
Inspection of variations of lateral biases in studies that use other stimuli and other
stimuhis presentation methods suggests that the variations in CFT biases might be related
to task difficulty. This Task Difi‘iculty Hmothesis was tested in 7 experiments using
methods for assessing the existence of the effect as well as its strength and direction.
Three possrble outcomes were considered: that task difficulty is associated with increased
right hemisphere arousal, resulting in a stronger LVH bias, on the premise that difficulty
invokes the discrimination of relevant from irrelevant information; that it is associated

with increased left hemisphere arousal, resulting in a weaker LVH bias, on the premise

that task difliculty invokes the use of either a feature-search or verbal analysis; or that it
is associated with decreased right hemisphere arousal on the premise that the task
becomes more diflicult as the quality, or “facedness,” of the stimulus is diminished.

Two kinds of chimeric face tests were used. In a 3-face test, subjects judged the
similarity of each pair of chimeric faces to an unaltered target face. In a 2-face test, they
chose the better exemplar of emotion, femininity, or age in pairs of chimeric faces.

Experiments 1- 4 defined parameters for the remaining experiments and showed
that task difficulty does affect the strength of the LVH bias, although the direction of the
effect began to be clear only in Experiment 4, which showed that difficulty was
associated with a weaker LVH bias. Experiment 5 confirmed the effect and its direction,
as did Experiments 6 and 7, which showed that the effect also generalizes across different
stimuli and methods. Overall results thus supported the Task Difliculty Hypothesis and
showed that task difficulty is associated with a weaker LVH bias. With the effect and
direction established, the experiments then considered the question of mechanisms. To
this end, subjects in Experiments 6 and 7 were asked which of several predefined
cognitive strategies they thought they had used during the test; their accounts were
consistent with the model, that is, with the hypothesized changes in hemispheric arousal
associated with an LVH bias weakened by task difficulty. The subjects’ own accounts
I suggest that the change reflects the adoption of a “first-impression” strategy for easy
judgments and a feature-search and/or verbal strategy for difficult judgments. Further
analysis also suggests a contribution of “facedness” to the effect. Clinical implications of
the results are discussed along with suggestions for further research to validate the

underlying assumptions about cortical activity related to the behavioral findings.

DEDICATION

This work is dedicated to my parents, Jim and Nancy Carbary, and to my fiancee, Faye
Berry. Each, in their own way, has generously supported my eflons.

iv

ACKNOWLEDGEMENTS

Many people have helped me in my research and in other ways throughout my
graduate career. My Dissertation Committee Chair, Professor Lauren Julius Harris, has
shown unwavering dedication to his students, holding them to the same high standards
exemplified in his own work. He has supported my work and that of his other students
with matchless enthusiasm and skill. In particular, through the many drafts of this
dissertation, he taught me that to write clearly one must think clearly, and that it is a life-
long endeavor.

My friend and fellow graduate student, Jason Almerigi, has contributed to this
work at all levels, and I deeply appreciate his fiiendship and support. Over the last
several years, I also have had the pleasure of working with several outstanding
undergraduate students. For the experiments comprising my dissertation, I am especially
indebted to Nick Erber and Lori Buhlman. Audrey McNabb, Nathan Besonan, and Todd
Sanders also contributed in significant ways in our laboratory, and their help was very
important.

I also am very grateful to the members of my Dissertation Committee, Professors
Norman Abeles, Tom Carr, and Joel Nigg, for their insight, recommendations, and
guidance in the design of the final experiments.

Without the cooperation of the students who served as subjects in the
experiments, the research could not have been done. These include hundreds of
undergraduates at Michigan State University who participated through the Department of
Psychology subject pool, as well as dozens of high school students from Alpena High
School, Alpena, Michigan. The assistance fiom Melissa Doubek, Psychology Instructor
at Alpena High School, is much appreciated. I am particularly glad for the opportunity to
have helped her integrate parts of the research into her own classroom lectures. I also
want to thank Professors Linda Gerard and Andrew Barclay for inviting their students to
participate in the research and for giving me the opportunity to lecture on the research to
their classes. For all the students, I hope they found their participation valuable.

Without face models — the many fellow students, faculty, stafl', and friends who
lent their mugs for use in this research - the work itself could not have been done. Here’s
looking at you!

The seven experiments described in this dissertation have been reported in a
different form at the poster session of the ninth (1998), eleventh (2000), and twelfih
(2001) annual meetings of TENNET, a conference of experimental and theoretical
neuropsychology, Montreal, Quebec. Accounts also have been published in special
issues of Brain and Cognition (Experiments 1 and 2: Carbary, Almerigi, & Harris, 1999;
Experiments 4 and 5: Carbary, Almerigi, & Harris, 2001; Experiments 6 and 7: Carbary,
Almerigi, & Harris, in press). I am grateful to the anonymous referees for TENNET and
Brain and Cognition for their helpfiil comments and suggestions. For Experiments 6 and
7, I also thank Academic Press and Professor Harry A. Whitaker, Editor of Brain and
Cognition, for permission to alter and use the cartoon faces designed by Ley and Bryden
(1979) for their research published in the same journal.

Finally, I am pleased to acknowledge the contribution of Dr. David Write and his
fellow members of the University Committee for Research Involving Human Subjects
(UCRIHS), the internal review board for the use of human subjects at Michigan State
University. All of the research was conducted with their approval and oversight, and
their reviews of the experimental procedures were invariably timely and helpful.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix
LIST OF FIGURES ............................................................................................................ x
I. INTRODUCTION .......................................................................................................... 1
II. WHY ARE THERE LATERALITY EFFECTS? ......................................................... 3
Nature of Stimulus Processing ............................................................................... 3
Part- Whole Processing ......................................................................... .3
F eature-Configural Processing ............................................................... 4
Local—Global Processing ....................................................................... 4
High — Low Spatial Frequency Processing ............................................ 5
Task Difficulty ........................................................................................................ 6
Individual Differences ............................................................................................ 9
Strength of Lateralization ........................................................................ 10
Tonic State of Hemispheric Arousal ....................................................... 10

What is the Proximate Processing Mechanism for Visual Hemispace biases?
Dynamic Activation-Attention Model ................................................................. 10
III. TESTING THE TASK DIFFICULTY HYPOTHESIS .............................................. 14
Proposed Role of Mechanisms as Applied to Visual-Verbal Stimuli ................... 14
A ttenuation Hwothesis .......................................................................... l4
Normalization Hwothesis ...................................................................... 15
Proposed Role of Mechanisms as Applied to Face Stimuli .................................. 15
Attenuation Hmothesis .......................................................................... 16
Normalization Hypothesis ...................................................................... 17
F eature-Analysis ijothesis .................................................................. 17
Overview of the Current Research ........................................................................ 17
IV. PARAMETRIC STUDIES: EXPERIMENTS 1 — 4 .............................................. 20
Experiment 1 ......................................................................................................... 20
Method ................................................................................................... 21
Results ................................................................................................... 23
Discussion ............................................................................................. 24
Experiment 2 ........................................................................................................ 24
Method .................................................................................................... 24
Results .................................................................................................... 24
Discussion .............................................................................................. 26
Experiment 3 ........................................................................................................ 26
Method .................................................................................................... 26
Results ................................................................................................... 29
Discussion .............................................................................................. 30
Experiment 4 ........................................................................................................ 32
Method .................................................................................................... 32
Results .................................................................................................... 32
Discussion ............................................................................................. 35

vii

V. PRINCIPAL STUDIES: EXPERIMENTS 5 — 7 ......................................................... 36

 

Experiment 5 ........................................................................................................ 38
Method .................................................................................................... 38

Results .................................................................................................... 38

Discussion .............................................................................................. 40

Experiment 6 ........................................................................................................ 41
Method .................................................................................................... 41

Pilot Studies to Establish Task Difliculty ............................................... 42

Results .................................................................................................... 45

Discussion ............................................................................................ .47

Experiment 7 ........................................................................................................ 48
Method ..................................................................................................... 48

Results .................................................................................................... 48

Discussion ............................................................................................. 52

VI. SUMMARY AND GENERAL DISCUSSION ........................................................... 53
Status of Hypotheses ............................................................................................. 53
Normalization ijothesis ........................................................................ 53

F eature-A nalysis Hwothesis ................................................................... 54

Attenuation Hypothesis .......................................................................... 54

Interpretation of Data From Other Studies ......................................................... .55

VII. CONCLUSIONS AND QUESTIONS FOR FUTURE RESEARCH ........................ 58
Difficulty and Cognitive Strategy ......................................................................... 58
Individual Differences .......................................................................................... 58
Attention-Arousal Model ...................................................................................... 60
Accuracy of Self-Report ....................................................................................... 61
Clinical Implications ............................................................................................. 61
APPENDIX A ............................................................................................... 63
REFERENCES .............................................................................................................. 65

viii

LIST OF TABLES

Table l, p. 18 ................. Predicted Effects of Task Difliculty on Hemispheric Arousal and
Visual Hemispace Bias for Visual- Verbal Stimuli and for Faces. Summary of predicted
changes in bias for difficult judgments according to each hypothesis: Attenuation and
Normalization Hypotheses for verbal stimuli, and Attenuation, Normalization, and

F eature-Analysis Hypotheses for faces.

Table 2, p. 51 ................ Percentage of Subjects Reporting Use of Each Strategy on
Dzflicult Judgments for Experiments 6 and Difficult and Easy Judgments for Experiment
7. Summary of data fi'om self-report questionnaire, which asked subjects what cognitive
strategy they thought they used when making difficult judgments in Experiment 6 and
when making difficult and easy judgments in Experiment 7.

ix

LIST OF FIGURES

Figure l, p. 22 .................. Example of a 3-face set. Face A, an L-L chimera, was made by
merging the left halfof the target (top) face and its mirror reversal along the midline.
From the viewer’s perspective, its left side is identical to the left side of the target face.
Face B, an R-R chimera, was made fi'om the right side of the target face. From the
viewer’s perspective, its right side is identical to the right side of the target face.

Figure 2, p. 27 .................... Example of 2-face sets. Example of an HN “happy-neutral”
2-face set is shown on the left. Each chimera was constructed of a smiling half-face and a
neutral half-face (of the same face model) joined along the midline. On a single test page
an HN chimera and its mirror reversal were shown, one above the other. For one of the
HN chimeric faces, the smiling halfappeared in the subject’s lefi visual hemispace
(LVH), the other in the right visual hemispace (RVH). Therefore, if subjects chose as the
happier chimera the one with the smiling halfto their left, that would constitute an LVH
choice. Examples of MP and Y0 face-sets are shown in the middle and to the right.

Figure 3, p. 43 .................. Examples of Realistic and Line-Drawn Cartoon F ace-Sets.
Cartoon face-sets are shown for three levels of judgment difficulty: easy (top), medium
(middle), and difficult (bottom), for realistic cartoon chimeras (left) and for line-drawn
chimeras (right).

Figure 4, p. 46 .................. Strength of L VH biases for easy through difi‘icult judgments for
photographic and cartoon face-sets for Experiment 6. The figure shows the relation
between strength of LVH bias and level of difficulty. Non-overlapping standard error

bars indicate statistical reliability.

Figure 5, p. 50 .................. Strength of L VH biases for easy through difi'icult judgments for
photographic and cartoon face-sets for Experiment 7. The figure shows the relation
between strength of LVH bias and level of difficulty. Non-overlapping standard error

bars indicate statistical reliability.

Figure 6, p. 56 ................... Strength of L VH biases for photographic, realistic cartoon,
and line-drawn cartoon face-sets for Experiments 6 and 7. The data are collapsed over
difliculty level, showing the potential relation between strength of LVH bias and degree
of “facedness.”

I. INTRODUCTION

Of all the principles about the hurmn brain, perhaps the most familiar is that the
cerebral hemispheres play different roles in cognitive fimctioning, with the left
hemisphere taking the lead for speech and language, the right hemisphere for a variety of
non-verbal tasks. This generalization has found support for well over a century in studies
of firnctioml deficits in brain-injured patients, since the 19503 in behavioral studies of
normal persons and callosotomy patients, and most recently in functional radiological
imaging studies of patients as well as normal persons. Thus, when speech and language
deficits result fiom brain disease, the associated lesions are typically in the left
hemisphere, and when patients show deficits in perception and recognition of faces,
objects, and other visual-spatial forms, the lesions are typically in the right hemisphere
(see Benson & Zaidel, 1985; Hellige, 1993; Harris, 1999a, for reviews).

Likewise, in studies of normal persons, when complex visual images are shown
tachistoscopically in the right or left visual fields (RVF, LVF), letters and words are
usually processed more quickly and accurately in the RVF (e.g., Hellige & Webster,
1979; Rizzolatti, Umilta, & Berlucchi, 1971; Umilta, Sava, & Salmaso, 1980; Wagner &
Harris, 1994; Zurif & Bryden, 1969), whereas non-verbal targets, including faces,
arrangements of dots, and orientations of lines, are usually processed more quickly and
accurately in the LVF (e.g., Boles, 1994; Rizzolati & Buchtel, 1977) Similarly, for free-
viewing tasks using chimeric faces as targets, most persons show left visual hemispace
(LVH) biases consistent with the left visual field (LVF) biases formd in tachistoscopic,
divided visual field studies (e.g., Carlson & Harris, 1985; Hoptrnan & Levy, 1988;

Jaynes, 1976; Luh, Redl, & Levy, 1994; Luh, Rueckert, & Levy, 1991).

Finally, in flyfRI experiments, verbal tasks have been associated with greater left-
than right—hemisphere activation (Puce, McCarthy, Allison, & Gore., 1996; Breier et al.,
1999), whereas non-verbal tasks, including nonverbal tasks including several of those
listed earlier, have been associated with greater right than left hemisphere activation
(Haxby, Horwitz, Ungerleider et al., 1994; McCarthy, Puce, Gore, & Allison, 1997;
Nakamura, Kawashima, Ito, et al, 1999; Puce et al., 1996). Similar effects have been
reported in PET studies (Haxby, Grady, Horwitz, et al., 1997; Sergent, Ohta, &
McDonald, 1992) and in studies using surface electrode recordings (e.g., Bentin, Allison,

Puce, & McCarthy, 1996).

1]. WHY ARE THERE LATERALITY EFFECTS?

Practically as soon as the facts of lateral specialization were established,
hypotheses were proposed to account for the phenomenon, that is, to explain why the
hemispheres are differently specialized for language and spatial functions. For at least
the last two decades, it has been acknowledged that explanations that focus on the
verbal/spatial classification are inadequate primarily because not all lateralized functions
have obvious verbal or spatial content. Two examples are praxis and perception of
melody, which are lateralized predominately to the left and right hemispheres,
respectively.

Nature of Stimulus Processing

The focus of most of the new hypotheses has been less on the content of the
lateralized function than on how the cerebral hemispheres process the information, on the
premise that differences in content will be compatible with different stimulus-processing
modes. Over the years, a number of versions of this basic idea have been proposed (for
an early review, see Bradshaw and Nettleton, 1981; for a more recent review, see Hellige,
1993). The fan examples that follow are only some of the versions proposed. None of
them, ftu‘thermore, are intended to dichotomize the brain but, rather, to suggest a
continuum of function between the hemispheres. Their focus, like the focus of this
dissertation, also is on perceptual, especially visual, ftmctions rather than on output, or
motor fimctions.

Part- Whole Processing
On the premise that all visual stimuli require some degree of decomposition into

their constituent parts for effective processing, Farah and colleagues (Farah, 1994; Farah,

Levinson, & Klein, 1995; Farah, Tanaka, & Drain, 1995; Farah, Wilson, Drain, &
Tanaka, 1998) have proposed that the poles of the dimension of specialization are “part-
analysis” and “whole-analysis” represented by left and right hemisphere processes,
respectively. Using printed texts and faces as examples, the hypothesis is that texts,
which require extensive decomposition (into letters, syllables, words, phrases) for
analysis, are normally processed more effectively by the left hemisphere, whereas faces,
whose parts must be analyzed together, are normally processed more effectively by the
right hemisphere. It follows that a face is recognized as a face or as a face expressing a
certain emotion not so much fiom any individual part but from all the parts seen together.
For certain emotional expressions, some parts also may be more important than others or
may play a special role (e.g., contraction of the obicularis occuli muscles, which is
present in a spontaneous, or Duchenne, smile, but absent in a so-called “false” smile).
Feature-Confrgural Processing

Gauthier and colleagues (Gauthier, Behrrnan, & Tarr, 1999; Gauthier & Tarr,
1997; Gauthier, Williams, Tarr and Tanaka, 1998) have described the essential
processing difference between the hemispheres not in terms of parts and wholes but in
terms of stimulus features and stimulus configurations. Although not committed to seeing
these processes as lateralized, they do not object to this characterization (personal
communication with M. Tarr, September 2, 1999).
Local - Global Processing

Robertson, Lamb, and Knight (1988) have proposed that the dimension is best
characterized as local-global. They based this proposal on their finding that for letter-

stirnuli composed of smaller letters, patients with left-hemisphere damage had difficulty

perceiving the smaller letters that form the larger letter (“local” perception) but not the
larger letter itself (“global” perception). Conversely, patients with right-hemisphere
damage had difficulty perceiving the larger letter but not the smaller letters. From this,
they concluded that perception of the global configuration (larger letter made of smaller
letters) is deficient in right-hemisphere patients whereas perception of local features
(small letters that form the larger letter) is deficient in left-hemisphere patients.
High - Low Spatial Frequency Processing

Save for variations in the names assigned to the poles of the dimension, the three
models described above characterize the difference between the hemispheres in terms of
the analysis of parts and the integration of those parts into a whole. Others have
proposed that the more fundamental difference lies at the level of sensory information
independent of any such “part” — “whole” classification. One such model, by Sergent
(1991) as well as Robertson and Delis (1986), focuses on the spatial frequency of the
stimulus and proposes that higher frequency information is better processed by the left
hemisphere and lower frequency information by the right. In a further development, Irvy
and Robertson (1998) have proposed a two-stage model according to which an initial
analysis is made in terms of spatial fiequency, followed by a second spatial fiequency
analysis dictated by the demands of the perceptual task (e.g., deploying attention to
constituent parts, or to their configuration, as is indicated to be required after the first
analysis). They call this the “double filtering by frequency” model.

By their nature, the parts of a whole, or a configuration, are physically smaller
than the whole itselfand therefore subtend a smaller visual angle. They therefore occur

more fi'equently and require a higher frequency analysis than the configuration of those

parts; the configuration, by definition, occurs Iessfrequently and requires a relatively
lower frequency analysis. That said, the three previously described models and the
spatial frequency model are not incompatible; the spatial frequency model may come
closer to describing how stimulus characteristics are better suited for analysis of parts
versus the configuration of those parts.
Task Dificulty

If the strength and even the direction of the laterality effect are influenced by the
kinds of processing called for by the task, there is indirect evidence that they also are
influenced by the overall difficulty of the task, as indexed by accuracy and reaction time.
From a clinical perspective, this is important because task difliculty is progressively
scaled in most neuropsychological tasks. Failures on more dificult tasks are less reliable
indicators of a disease process than are failures on easier tasks. In clinical practice, it is
generally assumed that when tasks are easy, so that most patients perform correctly, they
perform the tasks in a similar way. When tasks are more difficult, however, and patients
begin to fail, they are more likely to fail for different reasons. For example, failures may
reflect sub-standard premorbid achievement (i.e., never having had the requisite level of
skill in the first place). They also rmy reflect the patient’s inability to change or shift
cognitive strategy to solve difficult tasks in a different way, either because the alternative
strategy itself is dysfunctional or because the ability to change strategies is dysfirnctional.
The proposal that task dificulty is related to changes in cognitive processing strategies
may also help generate explanations for the unusual co-occurrence of low-end errors with
high-end success on some psychometric instruments. In such cases, the strategy for

difficult items may be the only strategy available to the patient, and after initial low-end

errors, then a change is made to process the task differently, making for higher-end
successes.

The evidence for difficulty effects on laterality tasks comes fi'om several of the
research paradigms reviewed in Part I. All the evidence, however, is indirect. For
example, for naming tachistoscopically-projected letters, when the letters are made more
complex or when their font is degraded, presmnably making the task more difficult, the
usual RVF bias has been found to weaken or even change to an LVF bias (Bryden &
Allard, 1976; Hellige & Webster, 1979; Wagner & Harris, 1994). Bryden and Allard
(1976) suggested that this change reflects a change in processing style and that complex
or visually-degraded stimuli must be “normalized” and perceived as an exemplar of a
known stimulus category (e.g., letters must first be recognized as letters) before they can
be separated into relevant and irrelevant components for the usual left-hemisphere
processing.

For face perception, on divided visual field (tachistoscopic) tasks, the usual LVF
bias has been formd to change when faces are inverted (Leehey, Carey, Diamond &
Calm, 1978; Hillis, Hiscock, & Rexer, 1995; Luh, 1997), when subjects are cued to attend
to facial features rather than to the whole face (Rhodes, 1985), when the task is to
discriminate between schematic faces that differ on a single feature (Patterson and
Bradshaw, 197 5; Bradslmw and Sherlock, 1982; F airweather, Brizzolara, Tabossi, &
Umilta, 1982), and when subjects are shown two different views of a face and asked
whether the faces are the same or different (Bertelson,Vanhaelan, & Morais, 1979). Each
such condition presumably could make the task more difficult. In these studies difficulty

had been operationalized as lowered task accuracy and/or increased response latency.

Given the premises of the stimulus processing hypotheses, the laterality data also
suggest that more and less difficult tasks are performed differently. The nature of the
difference is less clear. Broadly, across stimulus classes, does a more difficult task
require a broad configural analysis to identify a stimulus class, or does it require a
detailed feature analysis to do so, or is there an interaction between task content (e.g.,
visual, motor, or verbal) and processing style? Furthermore, will the answers to these
questions be the same for all stimulus categories, or will the dynamics be different for
different stimulus classes?

Whatever may be the basis for the hypothesized difficulty-related changes in
laterality effects, over time and as experience with the stimulus material increases, tasks
also normally become less difficult as subjects develop facility or familiarity with them.
On this view, it lms been suggested that laterality efl‘ects therefore depend, at least in part,
on the subject’s level of expertise (Diamond & Carey, 1986; Gauthier et al., 1998;
Gauthier & Tarr, 1997; Bruyer & Crispeels, 1992; Rhodes & McLean, 1990). For visual-
spatial material, some investigators have proposed that the hallnmk of expertise is
configural processing of the sort for which the right hemisphere appears to be specialized
(Gauthier et al., 1999), whereas non-experts focus on stimulus features, a left hemisphere
function, without regard for the relationship of the featm‘es to one another.

Gautier and colleagues studied the role of expertise in MRI experiments using
“nonsense” visual stimuli before and after subjects were allowed to develop expertise in
discrimination (Gauthier, Anderson, Tarr, et al., 1997; Gauthier, Tarr, Moylan, et al.,
2000). As already noted, laterality effects were not the main focus, but in these studies,

right hemisphere activation appeared to take the lead (M. Tarr, personal communication,

September 2, 1999). For faces, that is to say human faces, however, it seems safe to say
that virtually all normal adults are “experts.” Right hemisphere, configural processing
therefore presumably would be modal, and, for that reason, so would LVF or LVH biases
on face perception tasks, which is what the literature indicates.
Collectively, the tachistoscopic and free-viewing studies suggest a relation between what
I have taken task difficulty to be task difficulty and the direction and strength of lateral
biases. This suggested relation hereafter will be referred to as the Task Difliculty
Hypothesis. What the studies do not clearly show is whether or why the effect should be
in one direction or another. In the context of the models of cerebral specialimtion
described earlier, one could say, for example, that where increased task difficulty is
associated with a decreased LVF or LVH bias, the change is toward local feature
analysis, whereas when it is associated with an increased LVF or LVH bias, the change is
toward holistic, configural processing. That is to say that task difliculty is associated with
a change in processing style and in turn a change in processing style leads to a change in
VH bias. These examples obviously do not exhaust the range of possibilities, and the
predictions may be different for verbal and non-verbal tasks.
Individual Differences

The proposed role of expertise raises the more general issue of individual
difl'erencesandhowtheymight figure inthe assessment ofthe Task Difficulty
Hypothesis. At least two sources, or kinds, of difference can be envisioned.
Strength of Lateralization

One kind is the strength of lateralization. Laterality effects of the kind described

here are generally stronger for right-handers than for left lenders. The difference may

reflect the greater heterogeneity in unselected samples of left-handers along with the less
strictly lateralized function in these individuals (see review in Harris, 1992). Modest sex
differences also have been reported, with females showing weaker laterality effects for
verbal material but also stronger effects for the perception of emotion represented in
faces (Bmton & Levy, 1989; Natale, Gur, & Gur, 1983).
Tonic State of Hemispheric Arousal
Another kind of difference is in characteristic hemispheric arousal. Levy and

colleagues have presented evidence that individual differences on lateralized tasks reflect
differences in tonic state of left versus right hemisphere arousal (Levy, Heller, Banich, &
Burton, 1983; Hoptrmn & Levy, 1988). Some persons, therefore, may have a
characteristic, or default, hemispheric style, whereas for other persons, the style may be
more malleable and therefore more affected by their subjective experience of task
difficulty.

What is the Proximate Processing Mechanism for Visual Hemispace Biases?

The Dynamic Activation-Attention Model
I£ according to the Task Difficulty Hypothesis, as has been suggested, task

difficulty, moderated by individual differences, affects the strength and even the direction
of the lateral bias in visual tasks, what is the proximate mechanism underlying the effect?
For the tachistoscopic divided visual field paradigm, the conventional interpretation of
lateral biases, in consideration of the anatomy of the visual system, is that they are direct
products of structural-anatomical connections and the resultant representation of the
visual fields in contralateral visual cortex. Thus, for faces and other non-verbal targets,

the usual left visual field (LVF) bias is seen as a product of hemisphere-specific

IO

information accessing only the preferred (right) hemisphere, accompanied by signal

decay prior to any interbemispheric transfer of information. That explanation, however,
cannot obviously account for the left visual hemispace (LVH) bias for the flee-viewing of
chimeric faces or for the presence of changes in strength and even direction of the bias in
both tachistoscopic and fi'ee viewing tasks. For these reasons, a strict structural-anatomic
explanation has largely been replaced by more dynamic models.

The particular model adopted for the current series of experiments was first
proposed by Trevarthan (1972) and elaborated by Kinsbourne (1970, 1973, 1974) to
account for laterath efi‘ects on tachistoscopic as well as auditory (dichotic listening)
tasks, and then further developed by Levy and colleagues and applied to free-viewing
tasks with verbal as well as non-verbal material (Levy et al., 1983; Levy & Kueck, 1986).
The model, which may called a dynamic attention-activation model, assumes that the
levels of arousal in the cerebral hemispheres are generally in reciprocal balance, leading
to a fixation of attention straight-ahead in extra-personal space. It then assmnes that the
balance will be disturbed by stimulation fi'om one or the other side of space or by
endogenous unilateral arousal caused by the expectation of a particular kind of stimulus.
Finally, it assumes that the imbalance leads to a change in orientation and attention
contralateral to the more aroused hemisphere and therefore to a bias in readiness to accept
and process input from the visual hemispace on that side. (In keeping with the premises
of the model, the general term, visual hemispace, or VH, and the directional terms LVH
and RVH, will be used to refer to the lateral biases on tachistoscopic divided visual field
tasks as well as on free-viewing tasks.) In other words, the hemisphere specialized for

processing particular stimuli becomes differentially aroused, or “primed,” when that

11

stimulus class is presented; this arousal is associated with a shift of attention to the
opposite side of space. For a flee-judgment task, whether the flees are presented
tachistoscopically or for free-viewing, to the extent that the right hemisphere is aroused
more than the left, attention will be driven to the LVH, enhancing the salience of “face”
information to the viewer’s left side and giving a processing bias to information on that
side. 1

Two assumptions thus underlie the use of the dynamic activation model to explain
the VH biases for judgments of flees and visual-verbal material. The first is that there is
greater activation of one hemisphere in response to the stimulus or task - more right-
hemisphere activation for flees, more left-hemisphere activation for verbal material. The
second is that this activation causes attention to be directed to the contralateral side of
space.

Direct support for the first assumption comes fiom data cited earlier from
firnctional imaging and EEG studies. Indirect support comes from clinical data, also
cited earlier, showing deficits in flee perception and visual-verbal comprehension
associated with right and left brain injury, respectively.

Direct support for the second assumption is harder to obtain, and it depends on
how attention is defined. There is, however, indirect support. For visual tasks, attention
is usually associated with eye movements towards the visual target, so that subjects

showing an LVH bias would be expected to look more or longer to the part of the object

appearing in the left hemispace. As Posner and his colleagues have shown, however,

 

' Kinsbourne’s model has not gone unchallenged. For a relatively recent test of the dynamic activation-
attention model, see the series of 5 experiments and discussion in Reuter-Lorenz, Kinsbourne, &
Moscovitch (1990), wherein they demonstrate three principles of the model: I) that attention is biased to

12

attention can be directed without eye movements (Posner, Walker, Freidrich, & Rafal,
1983). In these experiments, however, subjects were trained to keep their eyes still while
attending to peripheral stimuli. In a more naturalistic setting, deployment of attention is
normally associated with changes in eye movements, as in a typical orienting response.
For this reason Kinsbourne (1970) proposed that eye movements might be a reliable
index of direction of attentional deployment, so that even though attention and

eye movements need not be synchronized, under normal circrunstances in a flee-viewing
flee perception test, eye movements would be expected to follow the locus of attention.
The Task Difficulty Hypothesis therefore predicts changes in lateral deployment of

attention related to the perceived difliculty of a visual-spatial task. 2

 

the direction contralateral to the stimulated hemisphere, 2) that biases are not task dependent, and 3) that in
the caseof a directional conflict, the rightward bias (RVH bias) is stronger.
2 Research by Harris, Pierce, and Henderson (unpublished data, L. J. Han'is personal communication,
l999b) suggests that the relation between eye movements and performance on the lice-viewing tasks
dissertation is very complex. The majority of their subjects did move their eyes from the vertical midline
and did show the usual LVH bias on a face-judgment task, but eye movement and performance were not
directly related in any obvious way.

13

III. TESTING THE TASK DIFFICULTY HYPOTHESIS

As already noted, variations in stimulus and test parameters have been associated
with changes in the strength or even direction of the VB bias for verbal as well as non-
verbal material, including flees. The main premis of this dissertation, as expressed in the
Task Difficulty Hypothesis, is that the changes are directly related to task difficulty. If,
as the hypothesis suggests, task difficulty affects the bias by affecting how the task is
performed, that might help to account for the individual differences and stimulus-related
variations in the strength of the LVH bias on flee perception tasks. Within the context of
a dynamic activation-attention model, however, the hypothesis makes different
predictions for verbal and non-verbal material. As will be explained below, the
hypothesis cannot differentiate between two possible explanations for VH bias changes
for verbal material, and there is a third possibility that can be tested only in a non-verbal
task, which is why, for the current work, the hypothesis was tested with flees.

Proposed Role of Mechanisms as Applied to Visual-Verbal Stimuli

Based on the dynamic activation-attention model, there are at least two ways that
task difficulty could weaken the LVH bias for visual-verbal material - by attenuation of
the character of the stimulus and by normalization of the stimulus.
Attenuation Hypothesis

Assume that the verbal stimulus is a sequence of letters. Whether it is made more
complex (e.g., by adding ornate strokes and curls) or is degraded (e.g., by lowering the
intensityordiflirsingthefont), itisassmnedthattheletterswillbelesslikelyto arouse
left-hemisphere “letter processing” areas to the extent that they look less like letters, that

is, to the extent that their “letterness” has been attenuated. By default, that would leave

14

the right hemisphere more aroused than the left, causing the perceptual bias to change
from the RVH to the LVH.
Normalization Hypothesis

Alternatively, the right hemisphere might become more active not relatively but
absolutely, and not by deflult but directly, if the more complex or degraded stimulus
directly activates the right hemisphere instead of the left hemisphere where the stimulus
is usually processed. Bryden and Allard (1976) suggested that under these
circumstances, what occurs is stimulus normalization, a pre-processing of the stimulus if
it is not innnediately recognized as a letter. They described normalization as a separation
of relevant from irrelevant parts, or signal from noise, and they assumed that it is
primarily a right-hemisphere process. For degraded or unusually complex verbal stimuli,
they further assumed that this process is activated to make sense enough of the stimulus
to determine whether it is verbal (letter or word) or something else. The resulting right-
hemisphere arousal thus manifests as an LVH bias.

In surmnary, to the extent that making the visual-verbal task more difficult is
associated with a change in perceptual bias fiom RVH toward LVH, the change may be
due either to an attenuation of the stimulus character and therefore to lowered arousal of
the left hemisphere, or to a normalization of the stimulus category, thereby recruiting a
different cognitive process fimctionally represented in the right hemisphere. In either
case, the result is a shift of the arousal ratio to the right.

Proposed Role of Mechanisms as Applied to Face Stimuli
For verbal material, both the Attenuation and Normalization Hypotheses predict

the same change in bias for difficult tasks - fiom RVH toward LVH. For non-verbal

15

material, such as flees, they make different predictions. For flees, there also is a third
way for task difficulty to affect the bias.
Attenuation Hypothesis

Just like letter recognition can be made more dificult to recognize by making the
letters look less like letters, so flees can be made more dificult to recognize or to
discriminate by making them look less like flees, that is by attenuating their “flcedness.”
One way to accomplish this is to invert them. Inversion may attenuate flcedness in at
least two ways. First, it may hinder processing at the entry level, therefore requiring at
least additional time to determine whether the stimulus is or is not a flee. Second, even
after an inverted flee is determined to be a flee, inverting it or inverting some of its
features may hinder quick and accurate judgments offlcc qualities such as its emotion
(Thompson, 1980). To this extent, it’s less flee-like. Some inversion studies have shown
that the usual LVH bias weakens or even changes to an RVH bias (Rhodes, Brake, &
Allison, 1993), although other studies have not found any change (Kolb, Milner, &
Taylor, 1983). Where a change has been found, it Ins been suggested that the inverted
flee fails to arouse those right-hemisphere processes nornnlly activated for upright flees
(Rossion, Delvene, Debatisse, Gofl‘aux, et aL, 2001). In terms of the dynamic activation
model, the assumption is that the reduction in right-hemisphere arousal shifts the balance
to the left hemisphere, driving the viewer’s attention to the RVH.

If, for flees, attenuation of flcedness proves not to be associated with a change in
VB bias, then two hypotheses remain - the Normalization Hypothesis and the Feature-

Analysis Hypothesis (i.e., analysis of features that comprise the flcial configuration).

16

Normalization Hypothesis

Just as it does for letter processing, making flee diseriminations difficult by
degrading them or by making them highly complex through the addition of irrelevant
features would require separation of task-relevant from task-irrelevant information.
Unlike letter processing, however, if, by this normalization process, right-hemisphere
mechanisms are recruited, then the result should be to strengthen, not weaken, the usual
LVH bias.

F eature-Analysis ijothesis

For faces, as noted above, there is a third way for difficulty to affect the bias.
Making the task more difficult could precipitate a feature-based analysis. As suggested
above, when a face is decomposed and parsed for its features (e.g., nose, mouth), a left
hemisphere analysis is most likely to be used. Therefore, in contrast to the Normalization
Hypothesis, which predicts that the LVH bias will be strengthened, a F eature-Analysis
Hypothesis predicts that it will be weakened.

In summary, each of the hypothesized mechanisms — Attenuation and
Normalization for verbal stimuli, and Attention, Normalization, and Feature-Analysis for
faces — makes a prediction about changes in the VH bias when perceptual tasks are made
more difficult. These predictions are summarized in Table 1.

Overview of the Current Research

The current research consists of a series of seven experimental tests of the Task
Difficulty Hypothesis for flee-viewing chimeric flee tasks. Experiments 1- 4 were
designed to test the hypothesis and at the same time to establish stimulus and

measurement parameters. With the parameters established, Experiments 5 — 7 were

17

Table 1. Predicted Effects of Task Difficulty on Hemispheric Arousal and Visual
Hemispace Bias for Visual-Verbal Stimuli and for F aces

 

Visual-Verbal Stimuli Predicted Effects

hmthesis hemispheric aroufl visual hemis ace bias
Attenuation J, L — R T LVH l RVH
Normalization — L T R T LVH I RVH

Face Stimuli Predicted Efl'ects

hypothesis hemispheric ampfl visual hemis ace ias
Attenuation - L l R 1 LVH T RVH
Normalization 1 L T R T LVH I RVH
Feature-Analysis T L J, R I LVH T RVH

 

L and R are left and right hemispheres; LVH and RVH are left and right visual hemispace
biases; the dash (—) represents no hypothesized clunge, and arrows represent predicted
increases or decreases in hemispheric arousal and visual hemispace bias.

l8

designed to test the hypothesis more precisely. Experiments 6 and 7, however,
incorporated three important changes. First, instead on relying on the subjects’ own
judgments of task difliculty, difficulty levels were established by another group,
independently of the experimental subjects. Second, in order to achieve experimental
control over the level of difficulty for the judgments, specially designed cartoon flees
were used in addition to photographic flees. Finally, to identify the cognitive strategies
used by the subjects, subjects were questioned about the strategies they used while
making their judgments for the tasks they found to be difficult and easy.

All seven experiments also were designed to assess different possible
explanations of the Task Difliculty Effect by 1) ruling out the Normalization Hypothesis
or by 2) ruling out the Attenuation and Feature-Analysis Hypotheses. For flees, it was
supposed that a weakened LVH bias in association with increased task dificulty will
falsify the Normalization Hypothesis, leaving the Attenuation and the F eature-Analysis
Hypotheses, whereas it was supposed that a strengthened LVH bias will fllsify the
Attenuation and the F eature-Analysis Hypotheses, leaving the Normalization Hypothesis.
Given an outcome where the Attenuation and Feature-Analysis explanations remain,

further studies would be needed to test their relative contributions.

19

IV. PARAMETRIC STUDIES: EXPERIMENTS l — 4

Before we can address the question which, if any, of the hypothesized
explanations underlies the task difficulty efl‘ect, we must establish that the effect exists in
the first place and determine whether it is associated with a stronger LVH bias, consistent
with the Normalization Hypothesis, or a weaker LVH bias, consistent with the
Attenuation and F eature—Analysis Hypotheses. To do this, stimulus, subject response,
and efl‘ect-size parameters must be defined. As just noted, this was the purpose of
Experiments 1 — 4. These experiments used two versions of a flee-viewing chimeric flce
test that has been found to yield a reliable LVH bias. Experiments 1 and 2 used 3-flce
sets, and, based on these results, Experiments 3 and 4 directly compared 3-face and 2-
face sets. Based on these results, Experiments 5 -— 7, comprising the principal part of this
dissertation, used 2-flee sets exclusively. The results of Experiments 1 — 3 also led to
changes in the measurement of task difliculty. All three experiments used an indirect
measure oftask difficulty, whereas Experiment 4 — 7 used a direct measure, as will be
explained below.

Experiment 1

In Experiment 1, with 3-face sets, subjects were asked to compare two chimeric
flees to an unaltered target flee. The target face was located at the top of the stimulus
display with the chimeric faces below it. One chimeric flee, an “L-L chimera,” was
made from the left side of the target flee and the mirror image of that side; the other, an
“R-R chimera,” was made from the right side of the target flee and the mirror irmge of
that side. After a set of L-L and RR chimeras was made, the flees were cropped to a

uniform size and oval shape, primarily to exclude extra-flee cues including the outline of

20

the model’s hair. (See Kowner, 1995, for a discussion.) Chimeras were rmde fiom black
and white photographs of real flees, digitally scanned and manipulated in AdobeO
Photoshop. Face models were volunteers from the academic community, including
faculty, staff, and students. An example of a 3-flce set is shown in Figure 1. For each
flee-set, subjects performed a two-stage task, first judging how similar the L-L and R-R
chimeras were to each other, and then judging which chimera looked more like the target
flee. The assumption was that the more similar the chimeras were to each other, the
more difficult they would be to discriminate fi'om each other or item the target flee,
thereby making the task more difficult.

The subjects’ own ratings of the similarity of the L-L and R—R chimeras thus were
used as an indirect measure of task difficulty. The effect of task difficulty on the VH
bias, therefore, was assessed by examining the relation between the subjects’ L-L and
R-R similarity ratings and the strength and direction of their VH bias. An LVH bias
meant that the L-L chimera was reliably chosen over clmnce. Strength of the bias was
measured by the overall number of L-L choices.

Method

Subjecm. Subjects were 80 undergraduates (72 right-handers, 8 left handers, 46
females, 34 males).

Procedure. Subjects were tested in a classroom in groups of five to ten. Five 3-
face sets were used. Each flee-set was projected onto the classroom screen via overhead
projector. Subjects first were shown a sample flee-set with the chimeric flees visible but
with the target flee covered. They were asked to rate how different the chimeras were

from each other on a 7-point scale. After 10 seconds, the target face was exposed and the

21

 

 

 

 

Figure 1. Example of a 3-face set. Face A, an L-L chimera, was made by
merging the left half of the target (top) flee and its mirror reversal along
the midline. From the viewer’s perspective, its left side is identical to the
left side of the target flee. Face B, an R-R chimera, was made fiom the
right side of the target flee. From the viewer’s perspective, its right side is
identical to the right side of the target flee.

22

 

subjects were asked which chimeric face looked more like the target flee. After this,
testing continued in the same manner with the five 3-face sets.
Results

For each of the five 3-flce sets, a one sample t-test was performed to compare the
proportion of subjects showing a VH bias in either direction to the proportion expected
by chance (.50). Two of the flee-sets showed a reliable LVH bias, two showed a reliable
RVH bias, and one showed no bias in either direction. There was no detectable relation
between task difficulty based on the L-L and RR similarity ratings and strength of VH
bias, and there were no reliable effects of sex or handedness.

Afterwards, it became clear that a different test could be used to assess the
relation between the chirneras’ difference-ratings and the strength of VH bias across
face-sets. To that end, flee-sets were rank-ordered in two ways: by mean strength of
LVH bias and by mean rated differences between the L-L and R-R chimeras. Rankings
were such that a positive correlation would mean that an increasing LVH bias was
associated with a decreasing difference between the chimeras and/or that an increasing
RVH bias was associated with an increasing difference between the chimeras. Although
the Pearson r correlation between the two variables was an encouraging +.69, it was not
reliable and had a reasonable probability (p = .2) of being a random efl‘ect.

Discussion

Although reliable VH biases were found, they were not consistemly in the same

direction across flee-sets, and no relation was found between the VH biases and the

similarity ratings. In contrast, however, to the planned correlational analysis where the N

23

was 80, the post-hoe analysis yielded a promising but statistically unreliable relation, but
because the N was only 5, the statistical power of the analysis was greatly reduced.
Experiment 2

For Experiment 2, instead of a post-hoe analysis like the one in Experiment 1, a
planned comparison was made. To determine the number of flee-sets necessary to
achieve acceptable statistical power for the analysis, a power calculation was performed
based on the results of Experiment 1. The results indicated that twenty flee-sets would
give a power of 75% to detect a correlation of .50 in the predicted direction.

Method

Subjects. Subjects were 59 undergraduates (53 right-handers, 9 left handers, 46
females, 34 males).

Procedure. Subjects were tested in a classroom in groups of five to ten. The
procedure was identical to Experiment 1 but with twenty flee-sets instead of five.
Result

Of the twenty flee-sets, six showed a reliable LVH bias, six a reliable RVH bias,
and eight showed no bias in either direction. Again, there were no reliable effects of sex
or handedness. Of the twelve flee-sets showing a reliable LVH or RVH bias, for only
two was the bias strength related to task difficulty, one a stronger LVH bias, the other a
stronger R VH bias.

All twenty flee-sets were rank-ordered by mean strength of LVH bias and by
mean rated differences between the chimeras. The Pearson r correlation between the two

variables was .01 (p<.98), again non-reliable. The same correlational analysis then was

24

performed for only those twelve flee-sets that elicited a reliable LVH or RVH bias. The
result was a non-reliable Pearson r of .16 (p<.62).
Discussion

Although Experiment 1 showed a high but unreliablecorrelation fiom an
unplanned post-hoe test, consistent in direction with the Normalization Hypothesis, the
results of Experiment 2, with better statistical power, were non-reliable, with no obvious
directional trend. Both experiments, however, also yielded overall weaker LVH biases
than obtained previously in our laboratory when face-sets were presented in booklet form
rather than via overhead projection, as in Experiments 1 and 2, and this may have
contrrhuted to the absence of a reliable relation between task difficulty and VB bias.3 If
so, then establishing a reliable LVH bias must be the first requirement for testing the
Task Dificulty Hypothesis. To help meet this requirement in the subsequent
experiments, all test stimuli therefore were presented in booklet form. Along with this
change, flee-sets were scrutinized for whether or not they produced a reliable VH bias in
Experiment 2. Twelve did, and one nearly did. This left seven 3-flce sets that, for
unknown reasons, showed little propensity to elicit a bias. Those seven were culled,
leaving thirteen flee-sets for continued use. Other studies in our laboratory also have
found larger overall VH biases with 2-flce than with 3—face sets.4 If the overall VH bias
is more reliable with one format than with another, the more reliable format should boost

statistical power due to reduced dispersion of the data. However, if a more reliable LVH

 

3 In Experiments 1 and 2, strength of VH bias for flee perception also may have been influenced by what
Boles (1994) described as input variables (e.g., stimulus characteristics like luminescence and size),
possibly by the large size of the projected flee-set or by the contrast of a dim testing room with an
illuminated flee-set.
‘ There may be several reasons for the difference in the strength, reliability, and direction of the VH bias
between the 2-flce and 3-flce sets. That discussion, however, is beyond the scope of this paper.

25

bias is confirmed with 2-flce sets, a different measure of task difficulty also would have
to be developed. These considerations guided the design of Experiment 3.
Experiment 3

In Experiment 3, along with the thirteen 3-face sets from Experiment 2, fourteen
2-flce sets were used, consisting of eight “happy-neutral” (HN) flee-sets, three same-sex
young-old (YO) flee-sets, and three male-female (MF) flee-sets. The HN chimeras were
made from a smiling half-face merged along the vertical midline with a neutral half-face.
Both half-flees were of the same flce model, whereas the Y0 and MF flee-sets were
necessarily made from different flee models. All 2-flce sets consisted of a chimera and
its mirror reversal, one above the other. For the HN flee-sets, subjects were asked which
flee looked happier; for the YO flee-sets, which flee looked younger; and for the MF
flee-sets, which flee looked more feminine. Examples of each kind appear in Figure 2.
For each kind, choosing the chimera with the relevant eue(s) (happier half, younger half,
more feminine half) to the subject’s left would constitute an LVH choice.

The 2-flce sets were used so that the strength, direction and reliability of LVH
bias could be compared for the 3-flce and 2-flce formats. As already noted, if one
format is more reliable, its use should boost statistical power due to reduced dispersion of
the data.

Method

Subjects. Subjects were 103 undergraduates (96 right-handers, 7 left-handers; 88
females, 15 males).

Procedure. Two test booklets were prepared to counterbalance the left — right

position of chimera for the 3-face sets and the top — bottom position of chimeras for the

26

 

 

 

 

 

 

 

Figure 2. Examples of 2-face sets: happy-neutral (HN) on the left, masculine-feminine
(ME) in the middle, young-old (YO) on the right. All chimeras were constructed in the
same way except that for the HN chimeras the two halves came from the same face
model. Thus, each HN chimera consisted of a smiling half-flee and a neutral half-flee of
the same flee model joined along the midline. On a single test page an HN chimera and
its mirror reversal were shown, one above the other. For one of the chimeras, the smiling
halfappeared in the subject’s left visual hemispace (LVH), the other in the right visual
hemispace (RVH). Therefore, if subjects chose Face A as the happier chimera, that is,
the one with the smiling halfto their left, that would constitute an LVH choice.

Similarly, for the MF and Y0 chimeras, choice of flees B and A as, respectively, the
more feminine and older flees would constitute LVH choices.

27

2-flce sets on the stimulus page. In each booklet, the thirteen 3-flce sets came first,
followed by the eight 2-flee sets, for a total of 21 sets. Similarity ratings (to be used as
the index of task difliculty) were only rmde for the 3—flce sets because the chimeras in
the 2-flce sets are mirror reversals and cannot reasonably be judged for similarity. The
difliculty measure used for 3-flce sets therefore is not compatible for use with a 2-flce
set. All subjects were tested together in a large classroom.

As in Experiments 1 and 2, the subjects were told that they would be asked to
perform a two-stage task. Beginning with a sample 3-flce set, they were asked to
indicate on a 7-point scale how similar the two flees (L-L and R-R chimeras) were to
each other, and then to choose which flee was more similar to the third, target flee. In
contrast to Experiments 1 and 2, to help subjects understand the scoring method and to
help reduce variability by standardizing responses, subjects were shown a flee-set of
identical L—L and R-R chimeras and were told that they were examples of two identical
flees that should be scored 1, or “identical,” on their response sheet. Then they were
shown very dissimilar L-L and RR chimeras and were told they should be scored 7, “low
similarity?” After that, an entire 3-flce set was shown.‘5 The L-L and R-R chimeric
flees were pointed out at the bottom of the flee-set, and the subjects were asked to rate
their similarity on a scale fi'om l to 7 and to record their answer on their response sheets
by circling one of the points on a numbered line.

After they completed the thirteen 3-flce judgments, subjects made judgments for

the eight 2-flce HN chimeras, three YO chimeras, and three MF chimeras for a total of

 

5 The chimeras used for the “identical” demonstration were identical. The chimeras used for the “low
similarity” demonstration were the L-L and R-R chimeras from Experiment 1 rated least similar to one
another. Both faces used for these demonstrations were then retired from the bank of test flees.

28

fourteen 2-flce judgments. As with the 3-flce judgments, subjects first were shown via
overhead projection a sample 2-flee set for HN chimeras. The flees were labeled A and
B and were placed one above the other. Subjects were told that some of the test items
would look like this example and that, depending on the flee-set, they would be asked to
judge which flee was happier, which flee was older, or which flee was more feminine.
(See again the examples in Figure 2.) Then, for a single practice trial, they were asked to
judge which flee was happier and to indicate their choice by circling A or B on their
response sheet. Subjects then were guided. through their stimulus booklet, item by item,
with spoken instructions fi'om the test administrator. They were instructed to turn the
pages together and to wait for instructions before doing so. For example, for the 3-flce
sets, the test administrator said, “T urn the page. Item six, how different are the two flees
at the bottom of the page? (pause) OK, which flee at the bottom of the page looks most
like the one at the top?” Subjects were allowed ten seconds to mark their response.
Results

3-Face Sets. Of the thirteen 3—flce sets, six showed a reliable LVH bias, three
showed a reliable RVH bias, and four showed no bias in either direction. This analysis
had a probability of .77 to detect an effect of .15 in the predicted direction. Again, there
were no reliable effects of subject sex or handedness.

For each of the thirteen 3-flee sets, Pearson rs were calculated to assess the
relation between the difference-rating for the chimeras and the strength of the VB bias for
that flee-set. Only one correlation was reliable (r = - .25, p<.05), for one of the three 3-

flce sets that yielded an RVH bias. In other words, for none of the thirteen flee-sets was

 

6 Overhead projection was used only for the sample items only, so that the entire group received the same
instructional examples.
29

the strength of the difference-rating between the L-L and R-R chimeras directly related to
the strength of the LVH bias, but for one flee-set, a greater difference was related to a
stronger R VH bias. This analysis, with an n of 103, had an 87% probability to detect a
correlation of .3 in either direction.

Correlatiorfll analysis yielded the same results for men and women when their
data were analyzed separately. There was, however, a slight sex difference for rated
similarity of LL and R-R chimeras (on a 7-point scale, men’s mean = 4.52, women’s
mean = 4.05, t = 2.09, 101 df, p<.04).

As planned, all thirteen 3-flee sets were rank-ordered by mean strength of LVH
biases and also by the mean difference-rating between their chimeras. The Pearson r
correlation between the two variables was -.39 (p = .18), again non-reliable. This test,
however, lfld power of only 56% to detect a correlation of .50 in the predicted direction.

The same correlational analysis then was performed for only those nine 3-face
sets that elicited a reliable LVH or RVH bias. The result was a non-reliable Pearson r of
-.54 (p = .13).

Finally, for these same nine 3-flce sets, the mean chimera difference-ratings were
compared for the six LVH-sets and the three RVH-sets. The comparison fliled to show
any difference (t = 1.9, 7de p = .10). This analysis, however, with seven degrees of
freedom, had very low statistical power.

2-Face-Sets. Unlike the 3-flce sets, all fourteen 2-flce sets (HN, YO, MF)
showed overall strong, reliable LVH biases. Of the eight HN flee-sets, subjects made
LVH choices 65% of the time (t = 5.3, 102de p<.001), with the difference reliable for six

face-sets and borderline for the remaining two (p = .06 and p = .09). Of the three YO

30

flee-sets, subjects made LVH choices 74% of the time, with the bias reliable for all three
flee-sets (t = 7.8, 102df, p<.001). Of the three MF face-sets, subjects made LVH choices
70% of the time , again reliable for all three flee-sets (t = 6.8, 102df; p<.001).

Discussion

As expected, the booklet administration yielded VH biases for the majority (nine
of thirteen) of 3-flce sets, with the majority being LVH. Again, however, the results did
not support the Normalization Hypothesis. Instead, as in Experiments 1 and 2, of the 3-
face sets that showed a VH bias, difficulty was related more often to a weaker rather than
a stronger LVH bias, the reverse of what is predicted by the Normalization Hypothesis
but consistent with the Attenuation and F eature-Analysis Hypotheses (see Table 1).

As expected, the overall LVH bias also was stronger for 2-flce than for 3-flce
sets: twelve of the fourteen 2-flce sets but only 6 of the 13 flee-sets showed reliable
LVH biases, and the individual biases for the 2-flee sets were stronger. Based on these
findings, it was decided to test the Task Difficulty Hypothesis with the 2-face as well as
the 3-flce task. As already noted, this would require a new measure of task difficulty,
because the measurement developed for the 3-flce task (assessment of similarity of LL
and R-R chimeras) would not be possible for the 2-flce task. In Experiment 4, therefore,
a more direct measure, nflgnitude estimation, was used for both the 3- and 2-flee sets.
Subjects performed two tasks for each flee-set (3-face and 2-flee sets): first, they made
the appropriate flee judgment; then, using magnitude estimation, they rated the difficulty
of that judgment. Use of this dimculty measure allowed for data analysis similar to that

used in Experiments 1 -— 3 but now for both kinds of flee-sets.

31

Experiment 4

Method

Subjects. Subjects were 121 high school students in introductory and advanced
psychology courses (115 right-handers, 6 left-handers; 78 females, 43 males).7

Procedure. Subjects were tested in much the same way as in Experiment 3.
However, after seeing a sanrple 3—flce set and judging which chimera looked more like
the target flee, subjects were asked how they would rate the difficulty of their judgment
(A or B) if the average difficulty were given a rating of twenty points, and to record that
number on their response sheet.

After the 3-flce task, subjects were shown a sample 2-flce set for HN chimeras.
After choosing the happier flee, they were given the parameters noted above and asked to
rate the difficulty of that judgment. For the next fourteen 2-flce sets, subjects were
allowed ten seconds per judgment, asked to rate the dimculty, then prompted to turn to
the next flee-set in unison. All subjects were tested in their classrooms, and their teacher,
after training in our laboratory, administered the test.
Results

3-Face Sets. For each of the thirteen 3-flce sets, a one sample t-test was used to
compare the proportion of subjects showing a VH bias in either direction to the
proportion expected by chance (.50). Five flee-sets showed a reliable LVH bias, three
showed a reliable RVH bias, and five showed no bias in either direction. This test had

73% power to detect a difference of. 15 in either direction. No sex or handedness

 

7 At the time Experiment 4 was planned, the oppommity presented itself to test high school students. There
was no a priori reason to suppose that the results would difl‘er because of the approximately 3-year age
difference between them and the undergraduates used in the other experiments and in light of evidence that
hemispace bias does not change within this age range (Levine & Levy, 1986).

32

differences were found; however, the test for a handedness effect was weak due to the
limited number of left-handers.

For each of the 3-flee sets, Pearson rs were calculated to assess the relation
between the subjects’ estimated difficulty (now measured by magnitude estimation) of
each judgment and the strength of the VH bias for each flee-set. Of the thirteen
correlations, three were reliable (p<.05), two of them, however, for face-sets that did not
elicit a reliable left or right VH bias. In other words, for only one of the thirteen flee-sets
was difficulty related to a weaker LVH (r = -.21, p< .02). This test had statistical power
of 92% to detect a correlation of .30 in either direction. In Experiment 4, unlike
Experiment 3, there was no difference between the men’s and women’s overall
estimations of task difficulty.

As planned, all thirteen flee—sets were rank-ordered by mean strength of LVH
bias and by the mean rated differences between their chimeras. The Pearson r correlation
between the two variables was -.34 (p = .25), again non-reliable. However, this test had
statistical power of only 44% to detect a correlation of .50 in either direction. The same
correlational analysis on only those eight flee-sets that elicited a reliable LVH or RVH
bias now yielded a non-reliable Pearson r of -.36 (p = .38).

Finally, for these same eight flee-sets, the mean chimera difference-ratings were
compared for the five LVH-sets and the three RVH-sets. The comparison fliled to show
any difference (t = 1.1, 6df, p = .31). This analysis, with 6 degrees offi'eedom, had very
low statistical power.

2-Face Sets. Overall, the fourteen 2-face sets (HN, Y0, and MF) again showed

strong and reliable LVH biases. For the 8 HN flee-sets, subjects chose the chimera with

33

the smile in their LVH 67% of the time (t = 6.9, 120df, p<.001), with seven of the eight
HN face-sets eliciting reliable LVH biases. The 3 YO face-sets elicited an LVH bias
79% of the time (t = 11.2, 120df, p<.001). The 3 MP face-sets elicited an LVH bias 75%
of the time (t = 10.4, 120d£ p<.001). As in Experiment 3, every Y0 and MF face-set
elicited an LVH bias. Only for the MF face-sets, however, were there sex differences in
VB bias strength, with women showing a stronger LVH bias for the MF 2-face sets
(men’s mean = 71% LVH choices, women’s mean = 84% LVH choices, t = 2.55, 118df,
p<.012).

Unlike Experiments 1 — 3, the subjects rated the fourteen 2-face sets as well as the
3-face sets for perceived task difliculty. For each 2-face set (HN, Y0, and MF), Pearson
rs were calculated to assess the relation between the subjects’ estimated dimculty of each
judgment and the strength of the VH bias for that judgment. The relationship was
reliable for only one face-set (r = -.23, p<.013), indicating that for that face-set, the more
dimcult the judgment, the weaker the LVH bias. This analysis had statistical power of
92% to detect an effect size of r = .30 in either direction. The pattern of correlations
between VH bias and task difficulty was the same for men as it was for the entire group.
No face-set, however, showed such a relationship when women’s data were analyzed
separately.

When all fourteen 2-face sets were rank-ordered by mean strength of LVH bias
and mean estimated difficulty, the Pearson r correlation between the two variables was
-.32 (p = .27), again non-reliable. This test had power of only 47% to detect a

correlation of .50 in either direction.

34

Discussion

As in Experiments 1 - 3, the results failed to support the Normalization
Hypothesis for either 2- or 3-face sets and instead showed a consistent but infrequently
reliable relationship in the direction predicted by the Feature-Analysis and Attenuation
Hypotheses. For the 3-face sets, one correlation was consistent with those hypotheses.
The results also showed that the magnitude estimation method was successful. For the 2-
face sets, the relation between task difliculty and VH bias strength was found to be in the
same direction as for the majority of 3-face sets in Experiment 3. Again, only one of the
correlations, now for the 2-face sets, was reliable and again consistent with the F eature-

Analysis and Attenuation Hypotheses.

35

V. PRINCIPAL STUDIES: EXPERIMENTS 5 - 7

So far, none of the experiments support the Normalization Hypothesis. That is,
they fail to show that increased task difficulty is associated with a strengthened LVH
bias. Before either the Feature-Analysis or Attenuation Hypotheses can be accepted as
alternatives, however, we must be confident that the Normalization Hypothesis can be
definitively rejected. To find out will require further experiments. These experiments
can be designed with stimuli, parameters, and estirmted effect-sizes based on findings
fi'om Experiments 1 — 4, but, as noted in the Overview of Current Research in Part 111,
they incorporated four important changes based on the following considerations.

First, is the relation between LVH bias and task difficulty peculiar to these
methods of assessing VH bias and task difficulty, or does it generalize to other methods?
For example, could the use of self-ratings for task difficulty play a role because subjects
perform the task expecting some items to be easier and some to be more difficult? To
find out, difficulty ratings would have to be standardized prior to the experiment, and
subjects would have to be kept naive to variation in task dificulty.

Second, there are at least two possible explanations of the relation between task
dificulty and a weaker LVH bias that involve increased left-hemisphere activity. One is
that the weaker LVH bias reflects increasing left hemisphere arousal because of the
introduction of a feature-based analysis; another is that it reflects the introduction of a
verbal analysis. In the context of the dynamic attention-activation model, the activation
of left-hemisphere language strategies may accormt for the same data as the Attenuation
and Feature-Analysis Hypotheses, which means that a left-hemisphere, language

mediated strategy will have to be considered as well. To better understand the nature of

36

the processing used and to find out whether one of the remaining strategies was, in fact,
being used, subjects in Experiments 6 — 7 were asked directly to describe the strategy, or
style of processing, used for their judgments,

Third, in Experiment 5, as in Experiments 1 - 4, the chimeric faces were made
from photographs of real faces. For those faces, the quality of the displayed emotion
varied in unknown and uncontrolled ways. In Experiments 6 —- 7, to achieve a greater
measure of control over the displayed emotion, two different kinds of specially designed
drawings of faces were used in addition to the photographic chimeras.

Finally, in Experiments 5 — 7, only the HN 2-face sets were used. The reason was
that in Experiment 4, HN face-sets yielded LVH biases around 70%, compared to 75%
for MF and 79% for YO biases. Use of the HN face-sets therefore would allow for a
clearer expression of the effect of difficulty on the LVH bias by lessening the possibility
of a ceiling effect. In Experiment 4, HN face-sets, unlike MP face-sets, also did not elicit
sex differences. Using only the HN chimeras in Experiment 5 —7 therefore made it more
likely that the men’s and women’s scores could be pooled for greater statistical power.
Although this would come at the expense of potentially finding a sex difi‘erence, it was
decided that this was outweighed by the advantage of pooling the scores.

In sum, Experiment 5 was designed to replicate the consensus finding of
Experiments 1 - 4 and therefore to confidently rule out the Normlization Hypothesis;
Experiments 6 and 7 were designed to test whether the task difficulty efl‘ect generalized
beyond the established experimental protocol; and Experiment 7 was designed to better

understand the nature of the processing strategies used by the subjects.

37

Experiment 5
Method

Subjects. Subjects were 358 university students (281 right-handers, 57 lefi-
handers, 13 arnbidexters; 256 females, 95 males).

Procedure. Thirty-four HN 2-face sets were prepared in the manner previously
described. Each face-set was used only once. Two test booklets were prepared in order to
counterbalance the position of face-sets on the stimulus page. (See the example of HN
face-sets in Figure 2. See also Appendix A for examples of photographic HN face-sets
cormnon to Experiments 5 -7.) Subjects were tested in groups of five to ten in a
classroom. They were shown a sample 2-face set and then were told that for all test
items, they would be asked to judge which face was happier. After making the judgment
for the sample, they were asked how they would rate the difficulty of their judgment if
the average difficulty were given a rating of 20 points, and to record that number on their
response sheet.

Resulm

L VH Bias Efl'ects. Over all thirty-four face-sets, the mean LVH bias was 68% (t
= 12.1, 351df, p<.001). There was no sex difference in the size ofthe bias (women
68.3%, men 67.8%; t = .14, 349df, p<.89), but there was a handedness effect (t = 2.50,
336df, p<.01). Lefi-handers made LVH choices 60% of the time (different from chance t
= 2.37, 56df , p<.02); right-handers made LVH choices 70% of the time (different fiom
chance, t = 2.50, 336df, p<.01).

Diflicully Ratings. On the difficulty ratings, there was no handedness difference

(right-handers’ mean = 21, lefl-handers’ mean = 22, t = .34, 336df, p<.79), but there was

38

a sex difference, with men rating the task as more difficult than women (men’s mean =
24, women’s mean = 21, t = 2.8, 349df, p<.006). '

L VH Bias Efl'ects by Level ofDifficulty. The overall correlation between
subjects’ mean difficulty ratings for the thirty-four face-sets and mean LVH bias scores
for those face-sets was r = -.12 (p<.02). This analysis had statistical power of 81% to
detect an effect size of r = .15 in either direction.

For each of the thirty-four face-sets, Pearson rs were calculated to assess the
relation between the subjects’ estimated difliculty for each of the thirty-four judgments
and the strength of the VH bias for each face-set. Thirteen of the thirty-four correlations
were reliable (p<.05), and six more approached significance (p<.10). These nineteen
correlation coeflicients were all negative and ranged fi'om -.O9 to -. 19, indicating that the
more difficult the judgment, the weaker the LVH bias, or, said differently, the greater the
change toward an RVH bias. Data for the right-handed subjects yielded the same results
as for the entire group (i.e., thirteen of thirty-four correlations were reliable). When the
same analysis was conducted on only the lefi-handers’ scores, only two of the thirty-four
correlations were reliable. Finally, when the men’s and women’s scores were analyzed
separately (collapsed over handedness), three of the thirty-four correlations were reliable
for men, nine of the thirty-four correlations for women.

As planned, all thirty-four 2-face sets were rank-ordered according to their mean
strength of LVH bias and their mean estimated difliculty. The Pearson r correlation
between the two variables was -.22 (p = .21) but was non-reliable. This test had a one-
tailed power of only 54% to detect a correlation of .30. When analyzed separately, lett-

handers showed a non-reliable correlation (r = -.14, p<.30), and right-handers showed a

39

reliable correlation (r = -.13, p<.03). When analyzed separately neither men nor women
showed reliable correlations (men: r = -.20 , p< .27; women: r = -.24 , p< .17 ).
Discussion

Again, no support was found for the Normalization Hypothesis for 2-face sets;
instead the relation between task difliculty and VH bias was in the reverse direction. For
about halfthe stimulus face-sets (across all subjects), there were reliable relationships
between subject-perceived task difficulty and a weakening of the LVH bias usually
associated with this type of chimeric face—set. Furthermore, across all face-sets, only
right-handers showed this relationship. In sum, the results showed a small but reliable
relation between self-rated task difficulty and a weakening of the LVH bias.

There may be several reasons for the presence of a reliable effect in Experiment 5
and its absence in Experiments 1 and 2 and the mixed results in Experiments 3 and 4.
Most likely it was due to the relatively low power in the earlier experiments to detect a
small effect, and, secondly, to the use of 3-face sets in Experiments 1 -— 3 with an indirect
measure of task difficulty. In retrospect, the (non-reliable) post hoc positive correlation
(r = + .69) between task difficulty and strength of LVH bias in Experiment 1 was
misleading. Even with their shortcomings, Experiments 2 — 4 frequently showed a
negative relation between LVH bias and task difliculty. Indeed, the reliable results of
Experiment 5 are in the same direction as the majority of statistically reliable results in
the earlier experiments.

So far, task difficulty has not been independently manipulated; it was based on the
judgments of the same subjects who had made the face judgments. In the last two

experiments, therefore, task difficulty was manipulated and assessed independently, prior

40

to administration of the protocol. As already noted, it was hypothesized that the change
in the LVH bias reflected the incorporation of a feature-based analysis, but another
possibility must be considered — that more difficult items engender a strategy
incorporating verbal analysis of the stimuli. To better understand the nature of the
processing strategies used, subjects were asked about the style of processing they
employed while making their judgments. Finally, in Experiments 6 and 7 two different
kinds of specially designed face cartoons were used in addition to the chimeras made
from photographs of real faces in an effort to achieve a greater measure of control over
the displayed emotion, and, therefore, task difficulty.

Experiment 6
Method

Subjects. Subjects were 212 tmiversity undergraduates (212 right-handers; 158
females, 54 males).

Test Faces. Test stimuli were twenty-tln'ee pairs of chimeric faces. Two
different kinds of faces were used to make the chimeras. Eight of the twenty-three face-
sets were made fiom photographs of real faces. (See again the example in Figure 2.)
They were drawn fiom a pool of thirty-four face-sets, each of which had been
independently rated for difficulty to judge by the 358 subjects fi'om Experiment 5. The
eight chosen for Experiment 6 were the four rated “easiest” and the four rated “most
diflicult” to judge (shown in Appendix A). The remaining fifteen face-sets were made
fiom five different cartoon faces. Three were from a set of realistic faces originally

designed by Ley and Bryden (1979) to depict different levels of emotion ranging from

4l

extremely negative to extremely positive. For the current study, only the three faces
depicting the neutral, mildly positive, and extremely positive expressions were used. The
remaining two cartoon faces were more schematic than the Ley and Bryden faces. They
were designed, however, to depict the same three levels of emotion depicted in the Ley
and Bryden faces, that is, neutral, mildly positive, and extremely positive. For each of
the total of five cartoon faces, three face-sets were made such that the hemi—srnile
increased in magnitude incrementally, creating what was intended to be three levels of
judgment difiiculty — easy, medium, and difficult. The three pairings for one of the Ley
and Bryden faces and for one of the line-drawn cartoon faces are shown in Figure 3; all
cartoon face-sets are shown in Appendix A.

Pilot Studies to Establish Task Difficulty. For the cartoon faces, two pilot
studies were conducted to confirm the three levels of difficulty created by the design of
the face-sets. In Pilot Study 1, eighteen subjects were shown three pairs of chimeric
faces made from one of the Ley and Bryden faces and were asked to choose the happier
face of each pair and then to rank order the three sets according to the dimculty of the
judgments. The results were in almost perfect agreement (Pearson r = .98, p<.001) with
the intended difficulty ranking. In Pilot Study 2, eighteen other subjects were asked to
choose the happier face of each pair of all fifteen cartoon face-sets and immediately afler
‘ each judgment to rate each judgment as “easy,” “medium,” or “difficult.” The face-sets
were presented one after the other, in pseudo-random order. These ratings yielded only
two levels of difliculty instead of three as in Pilot Study 1. In Pilot Study 2, the “easy”
and “medium” face-sets received comparable ratings (F = 7.08, l4df, p<.009). Each,

however, was rated as easier to judge than the “difficult” face-sets (LSD post-hoe

42

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3. Examples of Realistic and Line-Drawn Cartoon Chimera Face-Sets.

Cartoon face-sets are shown for three levels of judgment difficulty: easy (top), medium
(middle), and difficult (bottom), for realistic cartoon chimeras (left) and line-drawn
chimeras (right).

43

difference between “easy” and “diflicult”, p<.003; “medium” and “difficult”, p<.028;
“easy” and “medium”, p<.254). The results, nonetheless, were in close agreement
(Pearson r = .72, p<.001) with the intended order of difficulty.

Procedure. All subjects were tested together in a large classroom. They first
were shown a sample face-set (labeled face A and face B) via overhead projection, asked
to judge which face was happier, and then told that they would be asked to make the
same kind of judgment for other pairs of faces depicted in their test booklets. The
booklets contained the twenty-three face-sets, each on a separate page. The subjects then
were guided through the twenty-three face-sets, one at a time, by the test administrator.
They were allowed ten seconds per face-set and were asked to mark their choice (face A
or face B) on their response sheet.

Following the twenty-three judgments, the subjects were directed to another part
of their response sheet and were asked three self-report questions about their more
difficult judgments. They were instructed as follows: “ For the judgments you just made,
you probably found that some were quite easy to make and some were more difficult.
For the judgments that were more diflicult to make, did you: I) talk to yourself about
how to decide; 2) study small features of the face while trying to decide; or, 3) decide in
some other way?” Subjects were told that “talking to themselves” did not necessarily
mean talking aloud and that any sub-vocal, or interml, verbalization should be counted.
They also were asked to mark all answers that applied and to briefly elaborate if they
marked “decided in some other way.” Subsequently, the three self-report categories will

be referred to as “feature-search,” “self-talk,” and “some other way,” respectively.

Results

L VH Bias Efl‘ects. For all twenty-three faces-sets, the mean LVH bias was 0.66
(t = 11.73, 211df; p<.001). Biases for the eight photographic and fifteen cartoon face-
sets were nearly identical, 0.67 for the photographic face-sets, and 0.65 for the cartoon
face-sets. Both means were different fiom chance (t = 9.55, 211df, p<.001 and t = 11.13,
211df, p<.001, respectively) but not fi'om each other (t = 1.54, 211df, p = .13).

L VHBias Efl'ects by Level of Difficulty. For the photographic face-sets, the
LVH bias was 0.71 for the four “easy” sets (t = 10.10, 211df, p<.001) and 0.64 for the
‘ four “difficult” sets (t = 7.04, 211df, p<.001), with both means different fi'om chance. A
one-way AN OVA indicated that the biases were reliably different fi'om each other
(difference = .064; F = 4.90, 423df; p = .027).

For the cartoon face-sets, the LVH bias was 0.67 for the five “easy” sets (t = 8.95,
211df; p<.001), 0.68 for the five “medium” sets (t = 10.16, 211df, p<.001), and 0.60 for
the five “difficult” sets (t = 6.35, 211df, p<.001), with all three means different from
chance. An AN OVA showed that these biases were different fiom each other (F = 6.27,
634df; p = .002). LSD post-hoe tests also revealed reliable differences between the
biases for easy and diflicult face-sets (difference = .07; p = .009) and between the
medium and difficult face-sets (difference = .08; p = .001) but not between the easy and
medium face-sets (difference = .01, p = .459). The bias strengths are shown in Figure 4.

A 2-tailed Pearson correlation for the photographic face-sets showed an inverse
relation of r = -.11 between the two levels of task difficulty and LVH bias (n = 424; p =
.027). Likewise, for the cartoon face-set for which the pilot studies indicated only two

levels of difficulty, the correlation was r = -.14 (n = 636; p<.001).

45

0.75 —

   
   

 

 

 

 

 

 

g 0.70 4
'6
.r:
0
g 0.65 -
.l
I.-
o
5 0.60 .
t
o
o.
2
0- 0.55 ,
0.50 '- ”mt
easy difficul easy medium difficul
photographic fame» cartoon face-sets

 

Figure 4. Strength of LVH bias for easy through diflicult judgments for photographic
and cartoon face-sets for Experiment 6. The figure shows the relation between strength
of LVH bias and level of difficulty. Non-overlapping standard error bars indicate
statistical reliability.

46

Self-Report Questions. For the more difficult judgments, 47% of the subjects
checked “self-talk,” 90% checked “feature-search,” and 23% checked “some other way.”
Several subjects who checked “some other way” and explained that they had made their
choice based simply on their first impression. See Table 2 (page 52).

Discussion

In sum, when task difficulty was defined and validated independently, it had the
same effect on the VH bias as it did in Experiment 4 and earlier experiments. The results
thus further support the Task Difficulty Hypothesis and allow the confident rejection of
the Normalization Hypothesis.

Based on the self-report data, the results also suggest that the change in the VH
bias fiom left to right is more closely related to the use of a feature-search style of
analysis than to the use of verbal mediation. The self-report data, however, were
incomplete because subjects were asked only about the judgments they found to be
diflicult, not the ones they found to be easy. Before the more difficult judgments can be
confidently associated with any particular strategy or strategies, self-report data are
needed for both kinds of judgments, easy as well as diflicult. The results also showed
that subjects who checked “some other way” often noted that their choice was based on
their first impression. If “first-impression” had been included among the categories, it
therefore might have increased the validity of individual choices.

With these considerations in mind, Experiment 7 was designed with two
procedural changes. Self-report questions were asked for the easy as well as the diflicult

judgments, and “first impression” was added as a 4til category.

47

Experiment 7
Method

Subjects. Subjects were 82 university undergraduates (82 right-handed; 66
females, 16 males).

Test Faces. Test booklets and face-sets were the same as those used in
Experiment 6.

Procedure. The testing procedure also was the same except that, as already
noted, in the self-report period, subjects were asked about their easy as well as their
diflicult judgments and were given “first impression” as a 4th category.

Results

LVHBias Efl'ects. For all twenty-three face-sets, the overall LVH bias was 0.67,
virtually identical to the score in Experiment 6 and, again, reliably different from chance
(t = 7.83, 79df; p<.001). Likewise, the biases for the eight photographic and filteen
cartoon chimeras were very similar to those in Experiment 6 and, again, very similar to
each other, 0.70 for the photographic chimeras and 0.65 for the cartoon chimeras, with
both biases again different fiom chance (t = 6.59, 79df, p<.001 and t = 7.12, 82df,
p<.001, respectively) but not fiom each other (t = 1.30, 79df, p = .20).

L VH Bias Efl'ects by Level of Difficulty. For the photographic face-sets, the
LVH bias was 0.71 for the four “easy” face-sets (p<.001; t = 6.01, 79df) and 0.70 for the
four “difficult” face-sets (t = 7.04, 79df; p<.001) (For these and some remaining analyses,
the degrees of freedom are discrepant due to casewise deletions for missing data points.)
This time, a one—way AN OVA indicated that the biases for the easy and dificult face-sets

were not reliably different fiom each other (difference = .017; F = 1.50, 158df; p = .223),

48

although the direction of the efl‘ect was the same as in Experiment 6.

For the cartoon face-sets, the LVH bias was 0.71 for the five “easy” face-sets (t =
6.859, 79df, p<.001), 0.67 for the five “medium” face-sets (t = 5.372, 79df, p<.001), and
0.58 for the five “dificult” face-sets (t = 3.497, 79df; p = .001). A one-way ANOVA
showed reliable difl‘erences between the LVH biases per difficulty level (F = 5.05, 239df;
p = .005). LSD post-hoe tests revealed reliable differences between the easy and difficult
face-sets (difference = .129; p = .001) as well as between the medium and difficult face-
sets (difierence = .087; p = .029). As in Experiment 1, an LSD post-hoe test showed no
difference between the biases of the easy and medium difficulty groups (difference =
.040; p = .291). These results are shown in Figure 5.

For the photographic face-sets, a one-tailed Pearson correlation showed a non-
reliable relation, r = -.028, between task difficulty and LVH bias (n = 160; p = .364). For
the cartoon face-sets, a one-tailed Pearson correlation for all three levels of difficulty
showed a reliable relation between task difficulty levels and LVH bias (n = 240) of r = -
.21 (p = .001).

Self-Report Questions. For the judgments that the subjects said were more
difficult, 51% checked “self-talk,” 92% checked “feature-search,” 17% checked “some
other way,” and 66% checked “first impression.” For easier judgments, 22% checked
“self-talk,” 28% checked “feature-search,” 2% checked “some other way,” and 93%
checked “first impression.” Table 2 lists the percentage of subjects reporting use of each

strategy.

49

0.75 .

 

 

 

 

 

 

 

 

 

§ 0.70 - ‘_

'6

.C

o .L.

I 0.65 —~

3

“5

.g 0.60 —

E

o

8

E 0.55 ‘1

easy difficult easy medum difficdt
”'0me facgsets cartoon face-sets

 

Figure 5. Strength of LVH bias for easy through diflicult judgments for photographic
and cartoon face-sets for Experiment 7. The figure shows the relation between strength
of LVH bias and level of difficulty. Non-overlapping standard error bars indicate
statistical reliability.

50

Table 2. Percentage of Subjects Reporting Use of Each Strategy on Difficult Judgments
for Experiment 6 and on Difficult and Easy Judgments for Experiment 7

 

 

 

 

 

 

 

 

 

 

 

 

 

Experiment 6 Experiment 7
n = 212 n = 82
self-talk strategy diflicult 470/, 51 °/.
easy 22%
feature search strategy diflicult 90% 92%
easy 28%
some other way difficult 23% 17%
easy 02%
first impression diflicult 66%
easy 93%

 

51

 

Discussion

The main results further confirm the task difliculty effect in Experiment 6. They
also confirm the finding indicating the predominance of the “feature-search” over the
“self-talk” strategy for difficult tasks and show that both strategies, but especially
“feature-search,” are associated more closely with difficult than with easy judgments.
Finally, they show that for easy judgments, “first impression” was the overwhelming
strategy reported. In combination, the results suggest that when the judgment is
suficiently easy, subjects can rely only on first impression, precluding the need for any
other strategy.

It has been long established that feature-oriented processing is associated with the
left hemisphere and that configuration-oriented processing is associated with the right
hemisphere. This series of experiments shows tlmt task difficulty is associated with
chsanges from one hemisphere and processing style to another. Still, however, a very
interesting question remains, namely, which leads which? Does task dificulty precipitate
a change in hemisphere and style of processing, or does a change in hemisphere and style
of processing precipitate the experience of difficulty? The current experiments were not

designed to address this question, but addressing it is a very interesting, logical next step.

52

VI. SUMMARY AND GENERAL DISCUSSION

Overall, the results support the Task Difficulty Hypothesis, that is, that task
difliculty is related to the strength of the VH bias for the judgment of emotion in chimeric
faces. After testing the hypothesis while at the same time establishing stimulus and
measurement parameters in Experiments 1 - 3, a task difliculty efi‘ect emerged in a
consistent direction by Experiment 4, and in Experiments 5 - 7 its strength and direction
were established and replicated, with the effect each time being a weakening ofthe LVH
bias as task difliculty increased. In Experiment 5, the same subjects made the lace
judgments who assessed task difliculty, and in Experiments 6 and 7, difficulty was
independently assessed and subjects also were asked to identify the cognitive strategies
they thought they had used to perform the task for diflicult judgments in Experiments 6
and 7 and for difficult and easy judgments in Experiment 7.

Status of Hypotheses

With the task difliculty effect established for different kinds of face-sets and
different ways of assessing difficulty, the question is how do the results fit the difl‘erent
hypothesized explanations? They clearly rule out the Normalization Hypothesis, lean
strongly toward the Feature-Analysis Hypothesis, and possibly suggest a role for the
Attenuation Hypothesis.
Normalization Hypothesis

By the Normalization Hypothesis, more diflicult face-sets, like the more difficult
letters and fonts tested by Bryden and Allard (1976) and by Wagner and Harris (1994),
would have to be perceptually “normalized” so that relevant and irrelevant information

could be distinguished and so that a stimulus class or category could be established to

53

assist firrther processing. The current experiments, however, repeatedly showed that the
LVH bias, instead of being strengthened, as predicted by the Normalization Hypothesis,
was weakened.
F eature-Analysis Hypothesis

The direction of the effect thus supports the Feature-Analysis Hypothesis,
consistent also with the subjects’ self-reports, in Experiments 6 and 7, of their
predominant use of a feature-search strategy and, secondarily, of linguistic cueing for the
dimcult judgments. Therefore, if a weakening of the LVH bias reflects an increase in
left-hemisphere arousal, these results suggest that the change in bias mostly reflects the
introduction of a feature-search strategy.
Attenuation Hypothesis

Although the results show that task difliculty is related to the use of feature
search, the Attenuation Hypothesis has not been ruled out. That is, a weakening of the
LVH bias with increasing task difficulty might still be accounted for by a reduction of
right hemisphere arousal resulting fi'om diminished “facedness” of the target stimulus.
Although a direct test of this possibility was not planned in the designing of the current
experiments, a rough kind of test is possible. Recall that in Experiments 1 — 5, the
chimeric faces were made from photographs of real faces, whereas in Experiments 6 and
7, in addition to the photographic fiaces, two kinds of specially designed cartoon faces
were used: the more realistic Ley and Bryden cartoons and the more schematic line-
drawn cartoons. After designing the cartoon chimeras, it became clear that the
photographic chimeras looked more like real faces than either kind of cartoon face, with

the line-drawn chimeras showing the least resemblance. This permitted comparison of

54

the overall LVH bias (collapsed over task difficulty) across the three levels of
“facedness.”

The results, summarized in Figure 6, suggest that the LVH bias was stronger for
photographic face-sets and realistic cartoon face-sets than for line-drawn cartoon face-
sets. For the combined data fiom Experiments 6 (n = 224) and Experiment 7 (n = 89),
the overall LVH bias was 0.67 for photographic face-sets, 0.66 for more realistic cartoon
face-sets, and 0.62 for line-drawn face-sets. Results fiom a one-way ANOVA indicated
that the biases were reliably different fi'om each other (F = 3.13, 937df; p = .044), and
LSD post-hoc analysis showed reliable differences in the strength of the LVH bias
between line-drawn and photographic face-sets (p<.023) as well as between line—drawn
and realistic cartoon face-sets (p<.043), but not between photographic and realistic face-
SCtS.

One possible interpretation of the difference in strength of LVH bias between the
line-drawn faces and other faces is that the usual right-hemisphere mechanisms were less
aroused for the less face-like faces. Unfortunately, although each kind of face-set was
collapsed over difliculty level for this analysis, there is no way to tell whether the line-
drawn face-sets generally were more difficult to judge because task dificulty cannot be
co-varied out of the analysis. The results nonetheless are suggestive and can be used in
designing new studies of the independent contribution of facedness to the difficulty
effect.

Interpretation of Data from Other Studies
How well do these hypotheses explain results fiom other face-perception studies that

show a change toward (or reversal to) an RVH bias? The F eature-Analysis Hypothesis

55

0.75 1

 

 

 

 

 

 

3
.2 0.70 4
o
.C
0
g 0.65 -
d T
o r
.5 0.60 —
1:
o
o.
2 0.55 ~
0.
0.50 -,_, ' ‘ ‘ ' fl
photographic realistic line-drawn
faces cartoon faces cartoon faces

Experiments 6 and 7

 

Figure 6. Strength of LVH biases for photographic, realistic cartoon, and line-drawn
cartoon face-sets for Experiments 6 and 7. The data collapsed over difficulty level,
showing potential relation between strength of LVH bias and degree of “facedness.”
Non-overlapping standard error bars indicate statistical reliability.

accounts for results that show this change when subjects are cued to attend to facial
features rather than to the whole face (Rhodes, 1985) and when subjects must
discriminate between schematic faces that differ on a single feature (Patterson and
Bradshaw, I975; Bradshaw and Sherlock, 1982; Fairweather, Brizzolara, Tabossi, &
Umilta, 1982). It is less clear how it accounts for the effect when subjects are asked to
make same-different judgments of two different views of the same face (Bertelson,
Vanhaelan, & Morais, 1979). The Attenuation Hypothesis, on the other band, would
seem to help to explain'the weakening of the LVH bias when faces are inverted and
therefore look less like a face (Leehey, Carey, Diamond & Cahn, 1978; Hillis, Hiscock,
& Rexer, 1995; Luh, 1997). As noted in the Introduction, however, none of these studies
determined whether the weakened or reversed LVH bias was associated with an increase

in subjective difficulty.

57

VII. CONCLUSIONS AND FUTURE QUESTIONS
Diffictu and Cognitive Strategy

Overall, the results suggest that when a visual judgment task becomes more
difficult, subjects rely less on right-hemisphere strategies and more on left-hemisphere
strategies. A major aspect of the change appears to be in the increased reliance on
stimulus features and, to a lesser extent, on the use of a language-based strategy. Further
studies are needed to assess the contribution of stimulus character (e.g., fitcedness). At
this time, it also remains to be seen whether and to what degree these strategies are
interrelated.

Individual Differences

Questions also remain about what the individual brings to the task as opposed to
what the task itself elicits. For example, as discussed previously, individual differences
in lateral specialization as indexed by handedness and, possibly, by the sex of the subjects
my be related to lateralized hemispheric arousal and, perhaps, in processing style as
well. The current experiments were not specifically designed to address these questions,
and, with respect at least to the handedness variable, the number of lefi-handers was
usually too small to permit meaningful analysis. Experiment 5 (with 281 right-handers
and 57 left-handers) showed a difference in the overall strength of the bias, with right-
handers showing the stronger LVH bias, which is consistent with the literature. In that
Experiment, right-handers and left-handers showed virtually the same relation between
bias strength and task difliculty, although the latter was not statistically reliable.

Inspection of scores across the seven experiments, however, did not reveal any other

58

obvious trends in the data. The possible role of handedness in the further elucidation of
the task difficulty effect thus remains for further research.

By contrast, for comparisons of the sexes, the Ns were adequate. Only in
Experiment 4, however, did men and women differ on LVH bias strength, with women
showing a stronger overall LVH bias for MF face-sets. As noted previously, this
difference was not pursued in the subsequent experiments so that data for men and
women could be pooled for a more powerful analysis of the relation between task
dificulty and LVH bias strength There was also an unexpected finding of a sex
difference: for an indirect measure of task difficulty in Experiment 3 using 3-face sets
and then for a direct measure of task difliculty in Experiment 5 using 2-face sets. In
both cases, men rated the CFT as an overall more difficult task than did women. In
neither case, however, were there sex differences in the relation between task difficulty
and the strength of the bias.

As discussed previously but not addressed directly in these experiments, there
also is reason to believe tlmt a further role may be played by individual difi‘erences in
characteristic style of hemispheric arousal independent of lateralized motor or perceptual
dominance. Finally, individual differences in experience with any stimulus class or
particular kind of test as well as learning history for the use of one or another processing
style might be expected to contribute to a task being “easier” and therefore less likely to
evoke a difficulty-related feature-search strategy. Like the other individual difi‘erence
variables, the possrhle contributions of characteristic style of hemispheric arousal and

experience/expertise remain for further research.

59

Attention-Arousal Model

Assuming that attention is lateralized to the hemispace contralateral to the more
aroused cortical hemisphere, a frmdamental tenet of the attention-arousal model, we can
now characterize the direction of changes in hemispheric arousal when task difliculty
changes. At this point, however, we can only speculate about changes in arousal within
each hemisphere. For example, if there is a significant increase in left-hemisphere
activity is introduced with increased task difficulty, are those newly aroused areas the
homologues of the right-hemisphere areas, or are they different altogether? To the extent
that left-hemisphere processing is associated with a different style of cognitive
processing, the greater likelihood is that the areas would be different. For example, one
would expect to find activation of left fiontal eye fields and of left parietal and
subcortical structures associated with attentional control as well as some degree of
activation of left-hemisphere fi'ontal and temporal language areas. Likewise, most
normal adults being “face experts” so that all flee judgments are normally easy, one
would nearly all face judgments to be processed predominantly in the right hemisphere,
with the right posterior fusiform gyrus taking the lead, and with an LVH bias as a result.
The firsiform gyms has been implicated for both expert processing and for flee
processing (Gauthier, et al., 1997; Kanwisher, McDermott, & Chun, 1997; Puce et al.,
1996), and it presumably will be active for any flee perception task — easy or dificult.
The introduction of task difficulty is unlikely to recruit left-hemisphere mechanisms
peculiar to an alternative strategy to the exclusion of fusiform activation. Therefore, to
the degree that the introduction of a feature-search strategy has increased left-hemisphere

arousal, it would not be expected to replace arousal in areas usually associated with flce

60

perception in normal subjects; instead it should activate areas that could introduce
complementary processing strategies. One way to address all such questions and
possibilities would be through real-time neuroirnaging studies of the concurrence of VH
bias, difliculty level, and local arousal.
Accuracy of Self-Report

Functional imaging also could help to determine whether and to what degree
subjects’ post-hoe reports of their strategies coincide with physiological changes found to
be associated with the task difliculty effect. For instance, would subjects who reported
talking to themselves for more difficult judgments show activation of known language
areas, and would subjects reporting use of feature analysis show activation of areas
known to be involved in feature searching (e.g., eye fields directing visual scanning,
frontal areas associated with sequencing for point-to-point comparisons), or would their
reports be unrelated to the strategies implied by actual patterns of arousal?

Clinical Imflications

Finally, the results have at least three kinds of implications for clinical practice.
First, consider the interpretation of neruropsychological test results. As noted previously,
insofar as task dimculty is progressively sealed in most neuropsychological tests, then
flilures on more difficult items may reflect either a low initial level of skill or, based in
part on the current research, the patient’s inability to shift strategies. Second, the
possibility that task difficulty is related to cognitive processing strategy may help
generate alternative hypotheses for the unusual co-occurrence of low-end errors with
high-end success on some psychometric instruments (a pattern sometimes associated with

test dissimulation). After an injury, a patient may show deficits for processing test items

61

in the usual way, while the ability to shift strategies remains intact. Being unable to
process the easy items in the usual way, low-end errors are made fi'om the outset, but then
a shift in strategy may occur, and fiom that point on test items are approached with the
cognitive strategy usually reserved for more difficult items. Third, research in
rehabilitation may help to determine, following an injury that has compromised a

patient’s ability to shift strategies, whether shifting cognitive strategies for difficult tasks

can be trained or cued.

62

APPENDIX A

 

 

 

 

 

 

Photographic flee-sets fi'om booklets used for Experiments 6 and 7. In order from top
left and fiom left to right, flee-sets shown are fiom pages I through 8, booklet A.
Booklet B showed the same flee-sets in the same order but with positions A and B
reversed. These eight face-sets are fi'om the pool of 34 photographic face-sets used in
Experiment 5 and represent the four rated most diflicult (Experiment 6 and 7; pages 1, 3,
6, and 8) and the four rated easiest (Experiment 6 and 7; pages 2, 4, 5, and 7).

63

APPENDIX A (continued)

   
   

M
\ .,
I I
Z"
,_ .
Si
and
J
A I n
/. ~l I 1.;
A A
. 193 9%. it?
‘ ‘ '\é‘; \ /®_ 8 I‘ . \ "I ll
”VIA
A I A I A I
A I A I
A I
‘ ' A I
A I A I

Realistic and line-drawn flee-sets from booklets used for Experiments 6 and 7.
Continuing in order from top left, and fi'om left to right, shown are face-sets from pages 9
through 23 fi'om booklet A. Booklet B showed the same flee-sets in the same order, but

with positions A and B reversed.

REFERENCES

Benson, F. D., & Zaidel, E. (Eds) (1985). The Dual Brain, Hemispheric Specialization
in Humans. New York: Guilford Press.

Bentin, 8., Allison, T., Puce, A., McCarthy, G. (1996). Electrophysiological studies of
flce perception in humans. Journal of Cognitive Neuroscience, 8(6), 551-565.

Bertelson, P., Vanhaelen, H., & Morais, J. (1979). Left hemifield superiority and the
extraction of physiognornic information. In I. S. Russel, M. W. Van Hof, & G.
Berlucchi (Eds), Structure and Function of Cerebral Commissures. University
Park Press: Baltimore. pp. 400-410. (First publication 1979, by London:
Macmillan.)

Boles, D. B., (1979). Laterally biased attention with comment verbal load: multiple
failures to replicate. Neuropsychologia, I 7, 353-361.

Boles, D. B., (1987). Reaction time asymmetry through bilateral versus unilateral
stimulus display presentation. Brain and Cognition, 6, 321 -333.

Boles, D. B., (1989). Do visual field asymmetries intercorrelate? Neuropsychologia,
27(5), 697-704.

Boles, D. B., (1990). What bilateral displays do. Brain and Cognition, 12, 205-228.

Boles, D. B., (1994). An experimental comparison of stimulus type, display type, and
input variable contributions to visual field asymmetry. Brain and Cognition, 24,
184-197.

Bradshaw, J. L., & Nettleton, N. C. (1981). The nature of hemispheric specialization in

man. The Behavioral and Brain Sciences, 4(1), 51-63.

65

Bradshaw, J. L., & Sherlock, D. (1982). Bugs and faces in the two visual fields: The
analytic/holistic processing dichotomy and task sequencing. Cortex, 18, 211-226.

Breier, J. 1., Simos, P. G., Zouridakis, G., Papanicolaou, A. C. (1999). Temporal course
of regional brain activation associated with phonological decoding. Journal of
Clinical and Experimental Neuropsychology, 21 (4): 465-476.

- Bruyer, R., & Crispeels, G. (1992). Expertise in person recognition. Bulletin of the
Psychonomic Society, 30, 501-504.

Bryden, M. P. & Allard, F. (1976). Visual hemifield differences depend on typeface.
Brain and Language, 3, 191-200.

Burton, L. A., & Levy, J. (1989). Sex differences in the lateralized processing of flcial
emotion. Brain and Cognition, 11(2), 210-228.

Carbary, T. J., Almerigi, J. B., & Harris, L. J. (1999). The left visual hemispace bias for
the perception of composite flees: A test of the difliculty-of-discrimination
hypothesis Brain and Cognition, 40, 1, 71—75.

Carbary, T. J., Almerigi, J. B., & Harris, L. J. (2001). The left visual hemispace bias for
the perception of chimeric flees: A further test of the difliculty-of-discrimination
hypothesis Brain and Cognition, 46, 47-51.

Carbary, T. J., Almerigi, J. B., & Harris, L. J. (2001). Task difficulty is related to lateral
shifts in visual hemispace bias for judgments of emotion in chimeric flces Brain

and Cognition. (in press).

66

Carlson, D. F ., & Harris, L. J. (1985). Perception of positive and negative emotion in
fi'ee viewing of asymmetrical faces: Sex and handedness differences. Presented at
the 13th Annual Meeting of the International Neuropsychological Society,
February 6-9, 1985, San Diego, California.

Diamond, R., & Carey, S. (1986). Why flees are and are not special: An effect of
expertise. Journal of Experimental Psychology: General, 115(2), 107-1 17.

Fairweather, H., Brizzolara, D., Tabossi, P., & Umilta, C. (1982). Functional cerebral
lateralization: dichotomy or plurality? Cortex 18, 51-66.

Farah, M. J. (1994). Beyond PET methodologies to converging evidence. Trends In
Neurosciences, 1 7(12), 514-515.

Farah, M. J., Levinson, K. L., & Klein, K. L. (1995). Face perception and within
category discrimination in prosopagnosia. Neuropsychologia, 33(6), 661 -674.

Farah, M. J., Wilson, K. D., Drain, M., & Tanaka, J. N. (1998). What is “special” about
flee perception? Psychological Review, 105(3), 482-498.

Farah, M. J., Tanaka, J. N., & Drain, M. (1995). What causes the flee inversion effect?
Journal of Experimental Psychology: Human Perception and Performance, 21 (3),
628-634.

Gauthier, I., Tarr, M. J., Moylan, J., Anderson, A., & Gore, J. C. (2000). Does the
visual subordinate-level categorization engage the functionally defined firsiform
flee area? Cognitive Neuropsychology, 1 7 (1-3), 143-163.

Gauthier, I., & Tarr, M. J. (1997). Becoming a “greeble” expert: Exploring the flee

recognition mechanism. Vision Research, 3 7, 1673-1682.

67

Gauthier, I., Behrman, M., & Tarr, M. J. (1999). Can flee recognition really be
dissociated fi’om flee recognition? Journal of Cognitive Neuroscience, 11(4),
349-370.

Gauthier, 1., Williams, P., Tarr, M. J., & Tanaka, J. W. (1998). Training “greeble”
experts: A fi'amework for studying expert object recognition processes. Vision
Research, 38, 2401-2428.

Gauthier, 1., Anderson, A. W., Tarr, M. J., Skudgarski, P., & Gore, J. C. (1997). Levels
of categorizations in visual objects studied with fimctional MRI. Current
Biology, 7, 645-651.

Harris, L. J. (1992). Left-handedness. In I. Rapin & S. J. Segalowitz (Eds) Handbook
of Neuropsychology, Vol. 6. (pp. 145-208). Amsterdam, Netherlands: Elsevier
Science Publishing Co, Inc.

Harris L. J. (1999a). Early theory and research on hemispheric specialization.
Schizophrenia Bulletin, 25 (1), 11-39.

Harris, L. J. (1999b). Personal communication, September, 1, 1999. Data from work by
L. J. Harris, S. Pierce, & J. Henderson.

Haxby, J. V., Grady, C. L., Horwitz, B., Salerno, J., Ungerleider, L. G., Mishkin, M., &
Sehapiro, M. B. (1993). Dissociation of object and spatial visual processing
pathways in human extrastriate cortex. In B. Gulyas, D. Ottoson, & P. E. Roland
(Eds.), Functional Organization of the Human Visual Cortex (pp. 329-340).

Oxford: Pergamon Press.

68

Haxby, J. V., Horwitz, B., Ungerleider, L.G., Masisog, J. M., Pietrini, P., & Grady, C. L.
(1994). The functional organization of human extrastriate cortex: A PET-RCBF
study of selective attention to flees and locations. Journal of Neuroscience, 1 4,
6336-6353.

Hellige, J. B., & Webster, R. (1979). Right hemisphere superiority for initial stages of
letter processing. Neuropsychologia, 1 7, 653-660.

Hellige, J .B. (1993). Hemispheric Asymmetry: What ’s Right and What ’s Left.
Cambridge; London: Harvard University Press.

Hillis, S. K., Hiscock, M., & Rexer, J. L. (1995). Dual-task interference patterns reveal
differential processing of upright and inverted flees. Brain and Cognition, 28,

1 55-1 72.

Hoptman, M. J ., & Levy, J. (1988). Perceptual asymmetries in left- and right-handers for
cartoon and real flees. Brain and Cognition, 8, 178-188.

Irvy R.B., & Robertson, L. C. (1998). The Two Sides of Perception. Cambridge, MA:
M.I.T. Press.

Jaynes, J. (1976). The Origin of Consciousnes in the Breakdown of the Bicameral Mind.
Boston: Houghton Miftlin Co.

Kanwisher N., McDermott J., Chun M. M. (1997). The fusiform flee area: A module in
human extrastriate cortex specialized for flee perception. Journal Of
Neuroscience, 1 7 (11), 4302-431 1.

Kinsboume, M. (1970). The cerebral basis of lateral asymmetries in attention. Acta
Psychologica, Attention and Performance 111 (AF Saunders, ed.) pp. 193-201.

North-Holland: Amsterdam.

69

Kinsbourne, M. (1973). The control of attention by interaction between the cerebral
hemispheres. Attention and Performance IV. (S. Kornblurn, ed.) pp.239-256.
New York and London: Academic Press.

Kinsbourne, M. (1974). Lateral interactions in the brain. In M. Kinsbourne & W. L.
Smith (Eds) Hemispheric Disconnection and Cerebral Function. pp. 239-259.
Springfield, Illinois: Charles C. Thomas.

Kolb, B., Milner, B., & Taylor, L. (1983). Perception of flees by patients with localized
cortical excisions. Canadian Journal of Psychology, 3 7, 8-18.

Kowner, R. (1995). Laterality in flcial expressions and its effects on attributions of
emotions and personality: A reconsideration. Neuropsychologia, 33 (5), 539-559.

Leehey, 8., Carey, S., Diamond, R., & Cahn, A. (1978). Upright and inverted flees: The
right hemisphere knows the difference. Cortex, 14, 411-419.

Levine S. C., & Levy J. (1986). Perceptual asymmetry for chimeric flees across the life-
span. Brain and Cognition, 5 (3), 291-306.

Ley, R. G., & Bryden, M. P. (1979). Hemispheric differences in processing emotions and
flees. Brain and Language, 7, 127-138.

Levy, J., & Kueck, L. (1986). A right hemifield advantage on a verbal flee-vision task.
Brain and Language, 27, 24-3 7.

Levy, J., Heller, W., Banich, M.T., & Burton, L. A. (1983). Are variations among right-
handed individuals in perceptual asymmetries caused by characteristic arousal
differences between hemispheres? Journal of Experimental Psychology: Human

Perception and Performance, 9(3), 329-3 5 9.

70

Luh, K. (1997). Effect of inversion for perceptual bias for chimeric flees. Poster
presented at the Ninth Annual Meeting of TENNET, 16-19 June, Montreal,
(kuuukh.

Luh, K. E., Rueckert, L. M., & Levy, J. (1991). Perceptual asymmetries for the free-
viewing of several types of chimeric stimuli. Brain and Cognition, 16, 83-103.

Luh, K. E., Redl, J. & Levy, J. (1994). Left- and right-handers see people differently:
F rec-vision perceptual asyrmnetries for chimeric stimuli. Brain and Cognition,
25, 141-160.

McCarthy, G., Puce, A., Gore, J. C., & Allison, T. (1997). Face-specific processing in
the human fusiform gyrus. Journal of Cognitive Neuroscience, 9(5), 605-610.

Natale, M., Gur, R., E., & Gur, R. L. (1983). Hemispheric asymmetries in processing
flcial expressions. Neuropsychologia, 21 , 555-566.

Nakarnura, K., Kawashirna, K, Ito, K., Sugiura, M., Kat, T., Nakamma, A., Hatano, K.,
Nagumo, S., Kubota, K., Fukuda, H., & Kojima, S. (1999). Activation of the
right inferior frontal cortex during assessment of flcial emotion. Journal of
Neurophysiology, 82 (3), 1610-1614.

Patterson, K., & Bradshaw, J. L. (1975). Differential hemispheric mediation of
nonveflxflrnsualsfinnfli JburnalQflhquninunualIhychohmgy:Lhanan
Perception and Performance, 1(3), 246-252.

Posner, M. 1., Walker, J. A., Friedrich, F. J., & Rafal, R. D. (1983). Effects of parietal
lobe injury on covert orienting of visual attention. Journal of Neuroscience, 4,

1863-1874.

71

Puce, A., Allison, T., Gore, J. C., & McCarthy, G. (1996). Face sensitivity regions in
human extrastriate cortex studied by functional MRI. Journal of
Neurophysiology, 74, 1192-1199.

Renter-Lorenz, P. A., Kinsbourne, M., & Moscovitch, M. (1990). Hemispheric control of
spatial attention. Brain and Cognition, 12, 240-266.

Rhodes, G., & McLean, I. G., (1990). Distinctiveness and expert effects with
homogeneous stimuli: Towards a model of configural coding. Perception, 19,
773-794.

Rhodes, G. (1985). Perceptual asymmetries in flee recognition. Brain and Cognition,
4(2), 197-218.

Rhodes, G., Brake, S., & Allison, A. P. ( 1993). What’s lost in inverted flees?
Cognition, 47, 25-57.

Rizzolatti, G, Buchtel, H. A. (1977). Hemispheric superiority to reaction time to flees: A
sex difference. Cortex. 13, 300-305.

Rizzolatti, G., Umilta, C., & Berlucchi, G. (1971). Opposite superiorities of the right and
left cerebral hemispheres in discriminative reaction time to physiognomical and
alphabetical material. Brain, 94, 431-442.

Robertson, L. C., & Delis, D. C. (1986). “Part-whole” processing in unilateral brain
damaged patients: Dysfirnction of hierarchical organizations. Neuropsychologia,
24, 363-370.

Robertson, L., Lamb, M. R., & Knight, R. T. (1988). Effects of lesions of temporal-
parietal junction on perceptual and attentional processing in humans. Journal of

Neuroscience, 8(10), 3757-3769.

72

Rossion, B., Delvenne, J. -F., Dibatisse, Goffaux, V., Bruyer, R., Crormnelinck, M., &
Guérit, J. -M. (2001). Spatio-temporal localization of the flee inversion effect: an
even-related potentials study. Biological Psychology, 5 0, 173-189.

Sergent, J., Ohta. S., & MacDonald, B. (1992). Functional neuroanatomy of flee and
object processing - a positron emission tomography study. Brain, 115, 15-36.

Sergent, J. (1991). Processing of spatial relations within and between the disconnected
cerebral hemispheres. Brain, 114, 1025-1043.

Tarr, M. (1999). Personal communication, September 2.\

Thompson, P. (1980). Margaret Thatcher: a new illusion. Perception, 9, 483-484.

Trevarthen, C. (1972). Brain bisyrmnetry and the role of the corpus callosum in behavior
and conscious experience. In I. J. Cernaeek & F. Podivinsky (Eds) Cerebral
Interhemispheric Relations. pp. 319-333. Bratislava: Vydavatelstvo Slovenskej
Akadémie Vied.

Umilta, C., Sava, D., & Sahnaso, D. (1980). Hemispheric asymmetries in a letter
classification task with different typefaces. Brain and Language, 9, 171-181.

Wagner, N. M., & Harris, L. J. (1994). Effects of typeflce characteristics on visual field
asymmetries for letter identification in children and adults. Brain and Language,
46, 41-58.

Zurif, E. G., & Bryden, M. P. (1969). Familial handedness and left-right differences in

auditory and visual perception. Neuropsychologia, 7, 179-187.

73

I14

   
 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII

lillillllllilllllMlllliill

3 1293 02334 0031