The Effects of presenting suboptimal emotional stimuli on subsequent evaluative judgments


Material Information

The Effects of presenting suboptimal emotional stimuli on subsequent evaluative judgments
Physical Description:
vii, 146 leaves : ill. ; 29 cm.
Balzac, Elena Adrian, 1967-
Publication Date:


Subjects / Keywords:
Research   ( mesh )
Facial Expression   ( mesh )
Emotions   ( mesh )
Subliminal Stimulation   ( mesh )
Attitude   ( mesh )
Psychological Tests   ( mesh )
Department of Clinical and Health Psychology thesis Ph.D   ( mesh )
Dissertations, Academic -- College of Health Related Professions -- Department of Clinical and Health Psychology -- UF   ( mesh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1993.
Bibliography: leaves 138-145.
Statement of Responsibility:
by Elena Adrian Balzac.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002337207
oclc - 51624831
notis - ALU0980
System ID:

This item is only available as the following downloads:

Full Text








There are many people who have helped me to

achieve this goal. First, I would like to thank my

dissertation chair, Dawn Bowers, Ph.D., for her advice

and encouragement, and for always being as enthusiastic

about my project as I was. I would also like to thank

my cochair, Bruce Crosson, Ph.D., who gave generously

of his wisdom and support, not only during this

endeavour but throughout our work together. I am also

grateful to my other committee members for their

individual and valuable contributions: Rus Bauer,

Ph.D., for his humor and challenging ideas; Howard

Rothman, Ph.D., for his warmth and support, and for

always making Shabbat morning minyan such a pleasure;

Tiana Leonard, Ph.D., for her help and her interest;

and, Hugh Davis, Ph.D., for helping me to grow during

my years of graduate school, and for teaching me his

love of language and laughter.

I could not have completed this work without the

help of Lyana Doty. She was responsible, reliable and

dedicated, as well as fun to work with, and she saved

me from many sleepless nights. I am very grateful for

everything she so good-naturedly did. I also thank

Barbara Hawes, who graciously shared her mechanical and

computer knowledge (which far exceeds mine) with me.

There are many friends who helped and supported

me, and kept me sane while my dissertation threatened

to consume me. I'd like to thank Larry Freedman, Allen

Silken, and David Greene, who cheerfully agreed to be

guinea pigs. I'd also like to thank Sharon Chase,

Randi Lincoln, Jennifer Oglesby, and Christina

Rodriguez, just for being there when I needed them. I

apologize to my cats, Thelma and Louise, for all the

times I neglected them for this dissertation.

I am also grateful to my family, for the love and

support they have always given me. I'd like to thank

them all for always encouraging me (and I especially

thank my sister, for not finishing school before I


Last, but not least, I'd like to thank everyone

who put on a garbage bag, leaned against a chalkboard

and made funny faces, for the sake of science.




ACKNOWLEDGMENTS ................................... ii

ABSTRACT .......................................... vi


Mechanisms Underlying Implicit vs. Explicit
Memory Processing .......................... 3
Impact of Subliminal Events on Behavior...... 10
Affective Primacy: Emotional Reactions to
Stimuli..................................... 21
Neuroanatomical Evidence ..................... 31
Rationale for Study .......................... 35

STUDY I ........................................... 42

Hypotheses ................................... 42
Methods ...................................... 43
Analyses ..................................... 49
Results ...................................... 50
Discussion ................................... 51

STUDY II .......................................... 62

Hypotheses ................................... 62
Methods ...................................... 62
Analyses ..................................... 64
Results ...................................... 64
Discussion ................................... 65

STUDY III ......................................... 71

Hypotheses ................................... 71
Methods ...................................... 72
Analyses ..................................... 74
Results ...................................... 74
Discussion ................................... 77

STUDY IV .......................................... 88

Hypotheses ................................... 88
Methods ...................................... 89

Analyses ..................................... 96
Results ...................................... 96
Discussion ................................... 100

GENERAL DISCUSSION ................................ 124

REFERENCES ........................................ 138

BIOGRAPHICAL SKETCH ............................... 146


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Elena Adrian Balzac

August, 1993

Chairperson: Dawn Bowers, Ph.D.
Major Department: Clinical and Health Psychology

Four different studies were conducted to determine the

effects of suboptimal affective stimuli, in an attempt to

replicate Murphy's (1990) findings that suboptimal emotional

faces can influence a subsequent liking rating. Right-

handed males of college age served as subjects. In Study I,

subjects saw happy faces, angry faces or neutral polygons as

prime stimuli, followed by a Chinese ideograph serving as a

backward mask and as a target stimulus. Subjects rated how

much they liked the ideograph on a 1-7 scale. Sixteen

subjects saw the primes optimally, for 1000 msec, and

sixteen subjects saw them suboptimally, for approximately 1

msec. Liking ratings did not differ between conditions.

Study II was almost identical to Study I, except that an

increased suboptimal exposure time was used. Ten subjects

saw the primes for approximately 4 msec. There was again no

effect of primes on liking ratings.

In Study III a different set of face primes,

specifically rated for emotional expression and intensity

was used. Neutral faces were added to serve as a neutral

prime stimulus. In the suboptimal condition, sixteen

subjects saw the primes for 4 msec, followed by the target

ideograph. There was a significant effect of prime

condition, such that ideographs preceded by neutral shapes

were preferred to all other conditions. In the optimal

condition, where 16 subjects viewed the primes for 1000

msec, there was also a significant prime effect. Ideographs

preceded by happy faces or neutral shapes were preferred to

ideographs preceded by angry or neutral faces.

In Study IV, three different suboptimal exposure

durations, 4, 16 and 32 msec, were used. Primes consisted

of happy and angry faces rated for emotional expression and

intensity, and neutral polygons. There was no effect of

prime condition at any suboptimal exposure duration. In the

optimal condition, ideographs preceded by angry faces were

liked significantly less than ideographs in all other

conditions. Overall, there was no evidence that the

affective nature of suboptimal stimuli is processed when

presented in this type of format. The implications of these

findings within the field of subliminal and emotional

processing are discussed.



What is consciousness? Attempts to answer this

question have been a continuing theme throughout the

history of humankind. Philosophers and psychologists

have struggled to understand consciousness and the role

it plays in our lives. Is consciousness necessary for

all aspects of cognition, or can there be knowledge

without awareness?

The empiricists and materialists of the 1700s and

early 1800s postulated all cognitive activity and

behavior to be within the domain of conscious

awareness, although they did not agree on the specific

mechanisms involved. Later, structuralists such as

Wundt and Titchener asserted that all mental activity

was accessible by introspection, and thus was

conscious. It was Helmholtz in the 1800s who first

proposed the idea of "unconscious inference," the idea

that some degree of cognitive processing occurred

outside of consciousness. Although this idea was

surrounded by controversy, its popularity quickly


Freud proposed the domain of the unconscious,

containing thoughts, memories, and feelings which are

hidden from conscious awareness because of their

potential emotional impact. According to Freud, this

unconscious knowledge affects conscious behavior

through many mechanisms, including dreams, slips of the

tongue, and jokes.

Ebbinghaus suggested the possibility of cognitive,

rather than simply emotional, knowledge without

conscious awareness. He showed that early learning

trials of nonsense syllables facilitated learning in

later trials, even when he did not have a conscious

memory of what he had learned.

More recently, Claperede (1951) demonstrated what

appeared to be an example of learning without

awareness. He reported on an amnesic patient who

refused to shake his hand after he pricked her with a

hidden pin. When questioned, she had no conscious

memory of having been pricked by the pin, yet her

"avoidance" behavior indicated some level of knowledge

of what had occurred.

These are not isolated examples. Research over

the past two decades has clearly established the

existence of knowledge without awareness in both normal

and neurologically impaired individuals. This has been

empirically approached from two different perspectives.

One approach has focused on findings that normal

individuals and those with specific neuropsychological

disorders have "implicit" knowledge about consciously

experienced events for which they demonstrate failure

on "explicit" tasks. The second approach has examined

the impact of subliminally or "suboptimally" presented

events on subsequent behavior. In the sections which

follow, each of these approaches will be reviewed.

Mechanisms Underlying Implicit vs. Explicit Memory


Explicit memory is the overt recall or recognition

of a previous experience, e.g. recognizing words that

were seen on a previous list. Implicit memory accounts

for those instances when the initial exposure is

"forgotten" (is no longer conscious or explicit), yet

it still exerts an impact on subsequent behavior. For

example, when compared to novel words, subjects can

make lexical judgments about familiar words more

quickly (Forbach, Stanners, & Hochhaus, 1974), can read

them in more degraded form (Jacoby & Dallas, 1981), and

they are more likely to complete word stems with these

words (Tulving, Schacter, & Stark, 1982) in the absence

of conscious recollection of exposure to these words.

In general, explicit memory is tapped by "direct"

tasks, where a conscious reference is made to the

learning situation (e.g. subject is told "complete the

word stem with a word from the list"), while implicit

memory is tapped by "indirect" tasks, where no


reference is made to the initial exposure (e.g. subject

is told "complete the word stem with the first word

that comes to mind"). The single most salient finding

across such studies is that subjects who perform poorly

or even fail explicit/direct memory tasks, nevertheless

show patterns of performance across implicit/indirect

tasks suggesting that their previous exposure to the

stimulus continues to exert some effect on their


Evidence of implicit memory is found in

neurologically impaired populations as well. Amnesics

are as likely as normals to complete word stems with

previously seen words (Warrington & Weiskrantz, 1970)

if no explicit reference to the word list is made (e.g.

the subject is instructed to complete the word stem

with the first word that comes to mind). The amnesic

patient H.M. demonstrated skill learning on procedural

tasks over repeated trials, although he had no

conscious memory of previous exposures to the task

(Milner, Corkin, & Teuber, 1968). In addition,

prosopagnosics, individuals who are no longer able to

recognize or consciously recollect famous or familiar

faces, nevertheless show "implicit" knowledge of faces:

specifically, prosopagnosics demonstrate an affective

preference for faces that they do not have a conscious

recollection of having seen (Greve & Bauer, 1990).

Taken together, these findings suggest that

various underlying "representations" of knowledge may

be intact in these patients, but these representations

are unavailable to conscious (explicit) access. That

is, conscious access to these knowledge structures as

assessed by explicit tasks is disrupted. In both

normals and neurologically impaired patients, there are

"dissociations" between performance on implicit and

explicit memory tasks; performance on one type of task

does not predict or correlate with performance on the


How can such phenomena, knowledge and learning

without conscious awareness, be understood? Within the

domain of memory, there have been numerous theories to

explain how implicit memory works and how it can be

distinguished from explicit memory. Two primary

approaches are that of distinct memory systems

(Schacter, 1987) and transfer-appropriate processing

(Roediger, 1990).

Distinct Memory Systems

According to Schacter (1987), implicit and

explicit memories are separate systems, with each

having qualitatively different rules of operation.

Explicit memory is viewed as a conceptual "top-down"

process. The meaning of the stimulus is extracted

first, and the perceptual features are more completely

analyzed subsequently. Thus a subject may be able to

identify a partially degraded word based on the context

in which it is seen.

Support for this position comes from levels of

processing (LOP) research, which shows that conscious

recall and recognition of words are affected by the

elaborateness of processing utilized at encoding:

memory is best for semantic processing, then phonemic,

finally orthographic processing (Craik & Tulving,

1975). This suggests that explicit memory is improved

by access to the meaning of a stimulus, and only

secondarily by the physical features.

In contrast, implicit memory is seen as a

data-driven "bottom-up" process. The reverse of

explicit memory processes, this implies that analysis

of perceptual features is dominant. For example,

performance on implicit memory tasks is better if

during the initial exposure the subject views the

target word (e.g. the subject sees the word HOT and is

later tested on that word), than if she must generate

it after seeing a semantically related word (e.g. the

subject sees COLD, says HOT, and is later tested on

HOT). The former relies on perceptual analysis, the

latter on semantic knowledge. Explicit memory is

better in the generation condition (Jacoby, 1983),

which involves more elaborative processing. Implicit

memory, however, does not show a benefit based on

deeper levels of processing. These differences suggest

separate systems, and thus distinct processes, for

implicit and explicit memory.

Transfer-Appropriate Processing

According to the processing approach, there is

only one memory system. The accuracy of both explicit

and implicit memory depends on the similarity of

conditions between "study" (initial exposure to the

stimulus) and "test" (the retrieval of the information)

(Roediger, 1990). It is the type of processing, rather

than the type of memory, that is important. Thus, if

both study and test require a focus on the meaning of

the stimuli, a benefit can still be shown on a test

that does not tap conscious recollection (implicit

memory). According to Roediger (1990), implicit memory

can be conceptual and explicit memory can be

data-driven, as long as there is correspondence between

study and test.

In order to test this hypothesis, Blaxton (1985)

(cf Roediger, Weldon & Challis, 1989) presented

subjects with a word list either visually or

auditorily, then designed four separate tasks to

measure memory for the list: a conceptually-driven

explicit test (free recall of the word list), a

data-driven explicit test (graphemic cued recall, in

which the cue looked and sounded like the target word

but was not semantically similar), a conceptually

driven implicit test (measuring the likelihood that

subjects would use words from the list on a later

question and answer task) and a data-driven implicit

test (completing word stems). Blaxton found that

memory was affected by correspondence of the type of

processing between study and test (e.g. when the word

list was presented visually and the memory task was

data-driven), rather than correspondence between memory

systems (e.g. when the word list was presented visually

and the memory task was implicit). These findings

suggest that the implicit/explicit memory distinction

is a reflection of the different ways in which

information is processed, rather than indicating

functionally distinct memory abilities.

To date, there has been no resolution of the

controversy between those who advocate a multiple

systems view vs. those who propose a single system

view. Richardson-Klavehn & Bjork (1988), in their

review of the literature, reclassify the multiple

memory systems model and the processing account as

abstractionist and non-abstractionist positions,


According to the abstractionist view, implicit

memory represents a modification of abstract knowledge

structures, while explicit memory depends on the

ability to form and retrieve specific memory traces.

In contrast, non-abstractionists view abstract

representations and specific memory traces as

inseparable. Proponents of this view argue that

different types of memory reflect different operations

employed at study and test, thereby aligning themselves

with the single system view of memory. Although there

are exceptions, in general dissociations between

explicit and implicit memory are considered to support

abstractionist positions, while parallel effects

(similar results on implicit and explicit tasks

measuring the same stimuli) tend to bolster the

non-abstractionist argument.

Although each position is able to explain a

substantial proportion of the findings, neither theory

alone has been able to account completely for the

current pattern of dissociations and parallel effects.

Therefore, Richardson-Klavehn and Bjork (1988) suggest

combining them both. They propose that implicit memory

be viewed as containing two elements: one abstract and

one connected to contextual cues. This would explain

some seeming inconsistencies (e.g. sensitivity to

study-test correspondence without memory of the study

session), while still maintaining the explicit/implicit

distinction. This suggests that implicit memory is a

more complex process than previously thought, and

Richardson-Klavehn and Bjork acknowledge that their

"hybrid" position does not satisfactorily resolve all

inconsistencies either.

Regardless of the specific mechanisms by which

implicit and explicit memory phenomena are explained

and integrated, there is clear evidence for knowledge

without awareness. Normals and amnesics display

patterns of performance that reflect knowledge or

memory about events to which they have been exposed but

do not consciously recollect. This is often in the

absence of an ability to demonstrate that same

knowledge explicitly.

Impact of Subliminal Events on Behavior

The previous section dealt with situations in

which a stimulus was initially perceived consciously

and there was an effect of presentation in the absence

of conscious recall. What about when the initial

exposure itself is outside of conscious awareness? The

field of subliminal, or suboptimal, perception has

examined the effects of such stimuli.

History of Subliminal Perception

Subliminal perception, the idea that we can be

influenced by stimuli of which we are not aware, gained

favor in the 1940s and early 1950s. Numerous studies

were done purporting to show that behavior could be

substantially altered by subliminal messages (e.g.

flashing the words "Buy popcorn and Coca-Cola"

subliminally during movies was "shown" to produce a mad

rush for the concessions stand). In particular, the

field of advertising tried to capitalize on the effects

of subliminal perception.

However, the exaggerated claims and poor

methodology used in these studies led to widespread

criticism of not only studies demonstrating subliminal

perception, but also the existence of the phenomenon

itself. As a result, the idea that information could

be perceived outside of awareness fell into disfavor.

This skepticism continues today. A recent report by

the National Research Council of the National Academy

of Sciences (Psychological Science, 1990, as reported

in The Gainesville Sun, September 25, 1991) declared

that there was no evidence that subliminal messages

used in self-help tapes had any effect on behavior.

Criticism of subliminal perception research has

focused primarily on studies claiming to demonstrate

substantive and long-lasting behavioral changes on the

basis of subliminally presented stimuli. However,

there is other research which has simply examined if

subliminal stimuli can be perceived at some

non-conscious level, and if so, whether this

information has any influence on subsequent judgments

or behaviors, without attempting to demonstrate major

behavioral alterations. This body of literature

suggests that subliminally presented stimuli can be

processed non-consciously.

Dichotic Listening Studies

A number of more rigorous studies in the 1970s

using dichotic listening tasks suggested that

information in the unattended ear (channel) could

affect performance. In such studies, subjects are

required to monitor or repeat (shadow) an auditory

message heard in the attended ear while receiving

simultaneous input in the unattended ear. Lewis (1970)

found that reaction time to repeat a specific attended

word was affected by its semantic similarity to the

unattended word presented at the same time. This

occurred even though subjects failed to consciously

identify the unattended words.

In subsequent studies, Corteen and colleagues

(Corteen & Wood, 1972; Corteen & Dunn, 1974) paired

consciously heard city names with electric shocks.

These shock-conditioned words were then presented to

the unattended channel during a dichotic listening

shadowing task. Subjects responded with increased GSR

to all city names, even though they denied having heard

them during this phase of the task.

Although these studies suggest that nonattended

information affects performance, it is questionable to

what extent these findings reflect processing outside

of conscious awareness. The stimuli could have been

consciously processed had subjects been attending to

them, and thus may not be an example of suboptimal

presentation and perception (Hollender, 1986; Nielson &

Sarason, 1981). Thus, although these findings are

provocative, the methodological flaws of dichotic

listening studies prevent any firm conclusions about

processing without consciousness.

Semantic and Lexical Processing

A second avenue used to explore implicit

processing has involved visually presenting words and

pictures under conditions designed to prevent the

possibility of conscious awareness. This has been

done by presenting the stimulus at very brief

durations, followed immediately by a pattern mask.

The modern era of research in the field of

subliminal perception was largely initiated by Marcel

(1978, 1980, 1981), who conducted a series of seminal

experiments dealing with the impact of suboptimally

presented words on various cognitive judgments. To

prevent conscious processing, Marcel exposed stimuli

for exceptionally brief durations (at or below chance

detection levels, ranging from 20-110 msec), followed


immediately by a dichoptic backward pattern mask. This

was accomplished by presenting the stimulus to the

non-dominant eye, followed by presentation of a

nonsense pattern to the dominant eye.

In one experiment, Marcel showed normal subjects a

word for gradually decreasing durations and asked them

to make presence/absence judgments (was a word

presented), orthographic judgments (which of two words

is graphically similar to the target word), or semantic

judgments (which of two words is semantically similar

to the target word). At durations for which subjects

performed at chance levels in making presence/absence

judgments, these same subjects were able to make

graphic and semantic judgments above chance. Subjects

were able to make semantic judgments at significantly

lower exposure durations than either of the other two


A weakness of this experiment is that subjects

were asked to make decisions about stimuli they were

not aware of having seen, and many were hesitant to do

this. This reluctance may have influenced their

responses. Therefore, in subsequent studies, Marcel

measured the effects of a suboptimal prime on subjects'

reactions times to an optimally presented target



This was examined in a series of lexical decision

tasks in which subjects had to detect whether a letter

sequence was a word (e.g. CAT) or a nonword (e.g. CIT).

These target letter strings were preceded by words

which were semantically related or unrelated to the

target. Marcel found that lexical decisions were

faster when the primes were semantically related to the

target (semantic priming effect). Of critical

importance, this semantic priming effect occurred even

when the primes were presented below threshold (e.g.

subjects did not know they had seen anything).

Marcel also demonstrated a qualitative difference

between the effects of consciously and non-consciously

presented stimuli. Semantic priming effects differed

depending on whether the prime was suboptimally or

optimally (consciously) presented. Specifically, when

the prime was non-conscious, the subject appeared able

to access multiple meanings of polysemous (more than

one meaning) words. However, during conscious

processing subjects only analyzed one interpretation at

a time.

Marcel's findings were extremely significant for

the field of subliminal perception. He reliably

established the existence of non-conscious processing

and demonstrated that semantic processing of

suboptimally presented stimuli could occur, and was

more complete than graphemic or perceptual analyses.

Marcel validated the creation of a qualitative

distinction between conscious and non-conscious

mechanisms of analysis, by demonstrating different

semantic processing effects between the two. In

addition, he proposed a reliable methodology for

measuring non-conscious perception.

Following Marcel, many studies have examined

whether suboptimally presented words and pictures can

have a semantic priming effect on subsequent cognitive

decisions. Fowler and colleagues (Fowler, Wolford,

Slade, & Tassinary, 1981) replicated and extended

Marcel's work using similar methodology and tasks. At

durations at which subjects could not reliably make

presence/absence, graphic or phonetic judgments (10-25

msec), they did make accurate semantic judgments.

Reaction time to target words in lexical decision

tasks is shorter when suboptimal primes are

semantically similar (Balota, 1983; Cheesman & Merikle,

1986) or of a congruent emotional valence (Greenwald,

Klinger, & Liu, 1989). In addition, other types of

judgments such as position decisions (e.g. the subject

is asked on what side of the screen was the prime) are

more affected by the word content (LEFT vs. RIGHT) than

word location (left vs. right side) (Greenwald,

Klinger, & Liu, 1989).

McCauley and colleagues (McCauley, Parmelee,

Sperber & Carr, 1980) presented subjects with a

suboptimal picture prime, followed by a semantically

related or unrelated target picture, which the subject

was required to identify. The suboptimally presented

pictures produced semantic priming effects, decreasing

the reaction time required to name semantically related

targets. They concluded that it is possible to derive

meaning from suboptimally-presented pictures, and that

this process is separable from conscious

identification. Thus, semantic processing of stimuli

presented outside of awareness has been demonstrated

with different types of stimuli (images and words).

Jacoby and colleagues (Jacoby, Toth, Lindsay,

Debner, 1992; Jacoby, Waloshyn & Kelley, 1989) have

suggested an alternative definition for determining

unconscious perception, namely, anything which produces

an unintentional effect on behavior. Jacoby asserts

that if an individual is not aware of a stimulus, and

yet behaves as if she or he is being influenced by it,

it has been unconsciously perceived. Jacoby et al.

have devised tasks which discriminate between the

effects of consciously vs. nonconsciously perceived

stimuli and have shown that stimuli outside of

conscious awareness can influence responding (Jacoby,

Toth, Lindsay, Debner, 1992).


Jacoby further suggests that when such unconscious

influences take effect, subjects are likely to

misattribute the causes of their own behavior. In

other words, they mistakenly think they know the

reasons behind their actions, when in fact they are

unaware of them. Jacoby, Woloshyn & Kelley (1989)

demonstrated this phenomenon in a test of indirect


In the "false fame" paradigm, subjects were first

given a list of nonfamous names to read. They were

then presented with these names along with famous and

new nonfamous names, and asked to make fame judgments.

Subjects were correctly told that names from the first

list were not famous. In addition, the first phase was

done in one of two conditions, full or divided


According to the results, names from the first

list were more likely to be misidentified as famous if

the subject had been in the divided attention

condition. Here was a case of a subject clearly

misattributing his sense of familiarity with the names,

due to incomplete awareness and conscious recall.

Subsequent studies (e.g. Jacoby, et al, 1992) have

yielded similar findings concerning the effects of

stimuli perceived outside of conscious awareness, and

Jacoby considers indirect memory and unconscious

perception to be parallel processes. These findings

indicate that not only can subjects' behavior be

influenced by stimuli outside of conscious awareness,

but that subjects will misattribute the real causes of

their behavior, due to the nature of the influence.

Methodological Concerns

A number of investigators (Hollender, 1986;

Purcell, Stewart, & Stanovich, 1983) have questioned

findings purporting to show non-conscious processing

based on both methodological and conceptual grounds.

One concern is that suboptimally/subliminally presented

stimuli were, in fact, consciously perceived. Findings

from studies using subliminal stimuli rest on the

assumption that primes were not consciously perceived.

Suboptimal presentation of stimuli has been based on

threshold setting trials which have been conducted in

one of two ways: objective threshold or subjective

threshold. The subject's objective/conscious threshold

has been defined as that point for which detection or

recognition judgments are at or below chance.

The subjective threshold is less stringent. Based

on this criterion, recognition judgments may be above

chance, but the subjects report not being aware of what

they saw; in other words, they feel that they are just

"guessing" on recognition judgments. The use of

subjective threshold criteria for presenting


"subliminal" stimuli has been criticized on the ground

that, since overt recognition is above chance, some

conscious perception may have occurred. This is true

regardless of the subject's beliefs about what they


The validity of findings based on objective

thresholds has been questioned as well. In a number of

suboptimal priming studies, subjects were more

light-adapted during priming trials than during

threshold-setting trials (e.g. McCauley et al., 1980).

This occurred because of the additional light produced

by presentation of the target during "priming trials".

In brief, this serves to decrease the effectiveness of

the mask, because a threshold determined under

conditions of dark adaptation may not be sufficient

under conditions of greater light adaptation

(Hollender, 1986; Purcell et al., 1983).

Even after methodological failures have been

accounted for, however, there are conceptual questions

as well. Hollender (1986) suggests that studies

purporting to show implicit processing may instead

reflect the existence of very brief, fragile memories

(e.g. iconic). In his view, suboptimally presented

stimuli are consciously processed, but because of their

brief exposure, one's conscious memory of having seen

these stimuli is so fragile that it is lost by the time

of subsequent recognition tests. He proposes testing

for recognition after each trial, but acknowledges that

even this may not eliminate this problem, as there is

no way of knowing how brief the memories may be. In

addition, competing task demands may interfere with a

subject's ability to report consciously perceived


In response to these criticisms, subsequent

studies have attempted to control for these flaws.

Degree of light-adaptation has been strictly controlled

(e.g. Dagenbach & Carr, 1985, cf Hollender, 1986;

Fowler et al, 1981) and the thresholds used have been

very conservative, frequently well below objective

threshold (e.g. Greenwald, Klinger, & Liu, 1989).

Studies employing these stringent criteria have

replicated the previous findings demonstrating semantic

priming of suboptimally presented stimuli. Taken

together, these findings continue to validate the view

that non-consciously perceived events impact and

influence various cognitive-semantic judgments.

Affective Primacy: Emotional Reactions to Stimuli

The focus of the previously described studies was

to establish the existence of non-conscious processing,

particularly the ability to make semantic/cognitive

judgments based on subliminally presented information.

Not only does subliminally presented information appear

to influence one's cognitive judgments about

subsequently presented stimuli, it also appears to

affect one's affective preferences or liking (Zajonc,

1980, 1984) about subsequent "target" stimuli.

Subjects can develop an emotional reaction (preference)

to a stimulus they cannot identify cognitively

(recognize): an affective reaction develops not only

independent of but also prior to a cognitive response.

Zajonc describes this as "affective primacy".

According to Zajonc, affective preference depends not

on recognition but on "mere exposure"; that is,

exposure to a stimulus, not explicit memory of that

exposure, is the determining factor.

In a seminal study, Kunst-Wilson and Zajonc (1980)

exposed normal subjects to irregular octagons for 1

msec durations. This was immediately followed by

either a recognition memory task or a preference task.

In both, subjects were shown the target octagon and a

foil. On recognition trials, subjects were asked which

of the two octagons they had seen; on preference

trials, they were asked which of the two octagons they

liked or preferred.

Kunst-Wilson and Zajonc (1980) found that their

subjects performed no better than chance at recognizing

which of the shapes had been previously presented. In

contrast, subjects reliably preferred those octagons


that had been previously presented. These findings led

Kunst-Wilson and Zajonc to conclude that individuals

can develop emotional reactions and preferences for

events that they cannot consciously identify (i.e.

recognize). This affective preference effect has been

subsequently replicated by other investigators using

shapes (e.g. Seamon, Brody & Kauff, 1983; Seamon, Marsh

& Brody, 1984; Tassinary, Orr, Wolford, Napps, &

Lanzetta, 1984) and faces (e.g. Greve & Bauer, 1990).

Nevertheless, Zajonc' affective primacy hypothesis

is controversial. He is asserting that an affective

reaction to a stimulus can develop in the complete

absence of any cognitive activity; emotional reactivity

is faster than cognitive reactivity, according to

affective primacy. Lazarus (1982, 1991) argues that

emotion always results from cognitive activity, which

is the opposite of Zajonc' proposal.

According to Lazarus, cognitive appraisal, an

evaluative judgment about a stimulus and its relative

meaning to the observer, is necessary before one can

produce an affective reaction to the stimulus. In the

absence of such appraisal, there can be no emotion. A

complete cognitive analysis is not necessary; an

affective reaction can be produced based on incomplete

information. However, enough cognitive appraisal must

occur for a sense of the stimulus' significance to be

extracted before there can be emotion. Without this

minimal information, the observer does not know enough

to have an affective reaction: there is no sense of

the self's relation to the stimulus.

This initial cognitive processing need not be

deliberate or conscious. Thus, an immediate affective

reaction can result from cognitive activity which is

automatic and almost instantaneous, while more studied

judgments can be made later, on the basis of slower,

more complex and deliberate cognitive processing. In a

sense, Zajonc and Lazarus seem to be describing the

same process, but they label it differently. Both

assert that an initial evaluation of the stimulus based

on partial information must occur. However, Zajonc

considers this to be affective while Lazarus views

evaluation as a cognitive process. The specific label

used is less important than the phenomenon itself,

namely, the finding that subjects can develop an

affective preference for something they cannot

consciously identify.

Mandler and colleagues (Mandler, Yakamura, & Van

Zandt, 1987) have questioned whether the influence of

suboptimally presented stimuli is in fact restricted to

affect, or is a more general reaction to a familiar

stimulus. They have shown that when faced with a

familiar and a novel item, a subject will choose the

familiar item on perceptual variables, such as

lightness and darkness, as well. Mandler et al. (1987)

used methodology and stimuli equivalent to Zajonc' but

changed the subsequent judgment required. They argue

that exposure to a stimulus, even at a preconscious

level, biases the subject toward that object on a

variety of objective measure, and is not just limited

to one's "liking" or affective preference.

Regardless of whether cognition precedes affect or

vice versa, there is substantial evidence that

suboptimal presentation of geometric neutral stimuli

can bias one's subsequent liking for them. Inherent in

this idea is that "familiarity" as defined by previous

exposure to a stimulus increases one's liking for that

same stimulus, even when one is unaware of the

exposure. Repetition priming may play a role in this

effect: repeated exposure to a stimulus lowers one's

threshold of responsivity to it (e.g. Forbach et al.,

1974). However, more recent studies in which affective

liking is biased by the emotional content of suboptimal

primes cannot be accounted for by this phenomenon.

Analysis of Affective Content of Stimuli

In a series of experiments, Murphy (1990) gave

subjects a 4 msec exposure to either happy or angry

faces, immediately followed by a Chinese ideograph

(shown for 2000 msec), and then asked subjects to rate


how much they liked the ideograph. Subjects rated the

same target ideograph significantly higher than

baseline when preceded by a happy face and

significantly lower than baseline when preceded by an

angry face. These findings cannot be attributed to

mere exposure as subjects were rating novel stimuli.

Instead, they suggest that the emotional content of a

subliminal stimulus dramatically influences one's

affective liking for a subsequent "neutral" target.

In contrast, there was no difference in liking

ratings of the ideographs when the face primes were

optimally presented (1000 msec). Murphy suggests that

in the optimal condition subjects resisted having their

preferences "manipulated." According to Murphy it is

also possible that when subjects had more time to

analyze the stimuli they were less likely to base their

judgments on gross characteristics (i.e. smile vs.

scowl), and a more thorough "cognitive" examination of

the other features of the stimulus may have diluted the

effects of the emotional features.

These findings are consistent with Jacoby's

attribution theory. Subjects who developed a positive

feeling after viewing the happy face would be prone to

misattribute the true cause of this feeling, because

they were unaware, consciously, of the face. Instead,

they reported preferring those ideographs, believing


that must be the explanation. In the optimal condition

subjects, knowing they saw a happy face, could

correctly identify the source of their increased

positive feeling, and would not need to mistakenly

attribute this to the ideograph itself. (Negative

reaction to the angry face would work similarly.)

Bornstein (1992) reports that preference judgments, as

seen in the mere exposure effect, are stronger when the

initial stimulus presentation is outside of conscious

awareness, rather than being consciously perceived.

In addition to using emotional primes Murphy

employed nonaffective "cognitive" primes as well. She

presented target ideographs which were preceded by

suboptimal exposures to large vs. small shapes (squares

and circles), or symmetrical vs. asymmetrical designs,

and measured subjects' size and shape judgments of the

targets. As with the face primes, each ideograph was

preceded once by a small shape and once by a large

shape (or once by a symmetrical and once by an

asymmetrical design). Subjects were either asked how

big an object was the ideograph supposed to represent

(size judgment) or how symmetrical was the ideograph

itself (shape judgment), depending on which primes they

had received. These suboptimal primes did not affect

subjects' ratings of the ideographs. Thus, in contrast

to suboptimal emotional primes, "perceptual" priming

did not appear to have an effect on subsequent


However, optimal primes did impact subjects'

judgments of the target in these conditions, in the

expected direction. Murphy proposes that such physical

perceptual features require more time for analysis, and

thus have an effect at optimal but not suboptimal

exposures. This shows a clear dissociation between the

processing of emotional and nonemotional "cognitive"


Murphy also found that suboptimal male vs. female

face primes had no impact on subjects' ratings of

whether the ideograph represented a male or female

object. Optimal primes, however, resulted in

significantly higher femininity ratings for targets

preceded by female primes than for those preceded by

male primes. In all Murphy's studies, affective primes

impacted judgments after suboptimal exposures, whereas

"cognitive" primes had an effect only after exposure

durations at levels of conscious awareness. This

suggests that this latter type of information requires

more time to analyze and process than emotional


By presenting stimuli for such brief durations,

Murphy avoided many of the methodological criticisms

that have been raised (e.g. Hollender, 1986; Purcell et

al., 1983). Exposure duration was not based on the

subject's objective or subjective threshold, and

therefore there is no question of inadequate threshold

setting or incomplete dark adaptation in determining a

threshold. Four milliseconds is well below conscious


According to Murphy's findings, emotional content

is extracted from stimuli at exposures too brief to

allow recognition, and further, this affective

information has an influence which information about

the physical characteristics of the stimulus does not.

Thus, it is not clear if the analysis of physical

attributes can occur as rapidly or completely as

emotional analysis, which supports Marcel's (1978)


If in fact processing of affective information

occurs more rapidly than processing of perceptual

information, this may explain some inconsistent results

in the research. Studies measuring the effects of

primes based on semantic content or physical attributes

have reported null effects with exposure times below a

certain length (generally less than 20 msec). Studies

tapping emotional reactivity have yielded positive

results with exposures of 1-4 msec (e.g. Murphy, 1990;

Silverman & Weinberger, 1985). This difference, rather

than being due to methodological differences, may


instead be a function of the emotional vs. nonemotional

nature of the task.

Other studies have looked at the impact of

emotional suboptimal primes on subsequent decisions or

behavior. Such primes have been shown to affect

ratings of neutral stimuli (Smith, Spence & Klein,

1959), levels of anxiety (Silverman & Weinberger,

1985), and subsequent judgments and decisions (Bargh &

Pietromonaco, 1982): positive emotional stimuli

produced decreased anxiety and more favorable ratings

and judgments, while negative emotional stimuli

produced the opposite effects. These findings have

been in response to words and phrases presented as

briefly as 4 msec.

Viewed as a whole, this body of literature yields

support for the assertion that suboptimal primes can

impact affective processing. This can be done in two

ways. Repeated exposure to neutral stimuli can produce

an affective preference, and novel emotional stimuli

can influence subsequent decisions and behaviors, in

the direction of the previously presented emotion (e.g.

a happy stimulus leads to a more positive rating).

Both these events occur in the absence of conscious

awareness of exposure to the stimulus. This indicates

the presence of affective priming of suboptimally

presented stimuli.

Neuroanatomical Evidence

Now that these findings have been demonstrated

behaviorally, a second question is, what is the

neurobehavioral basis for affective priming? There is

a large literature regarding the neurobiological bases

for emotional behavior (Papez, 1937; Macchi, 1989;

MacLean, 1986). Much of this has centered on the role

of various limbic structures, particularly the

amygdala, hypothalamus, and connected regions. Until

recently, it was thought that the limbic system did the

majority of the work in emotion, and was responsible

for orchestrating the body's responses to emotional

situations. It is now known, however, that the cortex

also plays an integral part in emotional perception and

communication, as well as in modulating some of the

core limbic structures (Heilman, Bowers, & Valenstein,


There has been recent speculation by LeDoux and

colleagues (LeDoux, 1989, 1986, 1984; Iwata, LeDoux,

Meeley, Arneric, & Reis, 1986) that one's immediate

unconscious emotional reactions to various stimuli are

mediated by a neural circuitry that is distinct from

that which mediates one's conscious cognitive appraisal

of emotional stimuli. Specifically, they propose that

there are direct projections from the thalamus to the

amygdala (Iwata et al., 1986), the processing of which

is extremely rapid, although based on incomplete data.

The affective system, which in LeDoux's framework is

centered to a great extent in the amygdala, can be

accessed extremely rapidly because of the direct

pathways from sensory input areas. Although this

process may not be sufficient for complex stimuli or

involved affective reactions, LeDoux proposes that it

is adequate for initial responsivity to simple


LeDoux contrasts these subcortical connections

(sensory input to thalamus to amygdala) with cortical

circuitry (sensory input to thalamus to cortex to

limbic system). He views these latter connections as

more sophisticated, allowing for the passage of more

information and enabling more complex judgments.

However, because this information is more complex, it

also requires additional processing, thus taking more

time. Thus according to LeDoux, initial affective

analyses can occur before cognitive or perceptual

processing. It is after the initial stages of

processing that the systems become interactive. Later

decisions are a joint function of both affect and


It is important to note that LeDoux draws his

conclusions from the emotional reactivity of rats in

response to shock-associated acoustic stimuli. He


examined the effects of lesions in sites he considered

essential to affective processing (amygdala,

caudate-putamen, and medial geniculate body). He found

that lesions interrupting projections from the medial

geniculate body of the thalamus to the amygdala

diminished rats' physiological reactions to the shock,

leading LeDoux to suggest that these projections are

important in the development of emotional responses.

In contrast, lesions which interrupted connections

between the thalamus and the auditory cortex did not

have an impact on affective reactivity. This finding

bolstered LeDoux's contention that subcortical, but not

cortical, circuitry is involved in initial affective


Although largely speculative, LeDoux's formulation

is extremely provocative and interesting. It certainly

raises questions as to whether the affective

preferences that humans display in response to

suboptimal emotional stimuli (e.g. 4 msec presentation

of affective faces, as in the Murphy study) might

involve this thalamo-amygdala circuitry. More

experimental work is clearly needed to replicate

LeDoux's basic findings and to see if they generalize

to primates and other response systems.

A second question concerning the neuroanatomical

bases of affective processing involves the unique

impact that faces may have in the perception and

evaluation of emotion. There is evidence (Leonard,

Rolls, Wilson, & Baylis, 1985; Baylis, Rolls, &

Leonard, 1985; Rolls, Baylis, & Leonard, 1985; Perrett

et al., 1988) that in monkeys, there are neurons in the

superior temporal sulcus and the amygdala which respond

differentially to faces. A small subset of neurons in

these areas respond more intensely to faces than to

visual non-faces, other sensory stimuli, and other

arousing stimuli. In addition, many of these cells

respond differentially to different faces or to

specific parts of faces. In a few instances, neurons

respond more strongly to emotional faces (Leonard et

al., 1985) than to neutral faces. Thus, analysis of

faces and facial emotions seem to occur at a basic

"hard-wired" level.

It is possible that this is one reason Murphy

found such robust effects with her primes: subjects

were reacting not only to the emotions in the stimuli

but to the nature of the stimuli (faces) as well, both

of which affect the human processing system at a very

basic level. One argument against this hypothesis is

that Leonard et al. (1985) found that neurons evidenced

a response latency of 110-200 msec, and terminated with

the offset of the stimulus. With four msec

presentations, it is not clear if these neurons will

have had the chance to be activated.

Rationale for Study

As reviewed in the above literature, there is a

great deal of controversy in the field of suboptimal

perception. The phenomenon has been called into

question, and studies purporting to show its existence

have been challenged on methodological grounds (e.g.

Hollender, 1986). However, there is a substantial body

of literature, largely resistant to methodological

criticisms, which has consistently found evidence for

suboptimal perception (e.g. Marcel, 1978; Dagenbach &

Carr, 1985, cf Hollender, 1986; Cheesman & Merikle,

1986). These robust studies have employed pattern

masking of briefly presented stimuli to ensure that

primes do not reach conscious attention.

In a subset of this field, emotional processing of

suboptimal stimuli has been examined. This appears to

occur even more quickly than semantic, lexical, or

perceptual judgments. Studies using emotional stimuli

or examining emotional reactions to stimuli have

generally found significant results at much briefer

durations than studies tapping cognitive processes

(e.g. Greenwald, Klinger & Liu, 1989; Silverman &

Weinberger, 1985; Kunst-Wilson & Zajonc, 1980).


Zajonc explains the rapidity of this phenomenon as

affective primacy, the idea that affective processing

occurs independently of and prior to cognitive

processing. Murphy (1990), in an expansion of the

affective primacy hypothesis, demonstrated that a four

millisecond exposure to masked emotional faces

influenced preference judgments of a subsequently

presented figure, while cognitive evaluations were not

influenced by suboptimal presentation of figures

requiring perceptual or objective analysis. Optimal

presentations of the faces had no subsequent effect on

affective preference. Murphy interpreted this as the

result of an increased ability to process more

"cognitive" information about the prime with increased

time, thus decreasing the impact of the affect.

One confound in Murphy's work was the use of

polygons, rather than faces, as neutral stimuli. The

different nature as well as the different valence of

the primes may have affected her results. Subjects may

have responded to the polygons as they did because the

polygons were neutral, and not affective (which is

Murphy's hypothesis), or because they were geometric

designs rather than faces. It is possible that any

face, regardless of expression, will produce an

emotional reaction in a viewer. As Leonard et al.

(1985) demonstrated, there is something hard-wired into

the primate brain, such that there are some neurons

which specifically respond to faces and no other

stimuli. Thus, Murphy's comparison is not simply

between affective and neutral stimuli, but also between

face and non-face stimuli. It is not clear which

difference, or both, can be implicated in her results.

Therefore, it is important to discover if her findings

would exist if faces were used as neutral stimuli. If

this exists it would suggest that it is the affective

nature of the stimuli, not the fact that they are

faces, which influences subjects.

The following studies will attempt to replicate

and expand Murphy's results. Her studies were the

first to reveal the influence of a subliminal affective

stimulus using this type of methodology, namely, the

effect of an affective stimulus on a subsequent

affective judgment. If in fact this finding is robust

and generalizable, the ramifications are far-reaching.

For example, current controversy exists concerning the

differential lateralization of emotional processing.

One question, at least as it pertains to affective

priming, is whether the two hemispheres differ in terms

of their responsivity to suboptimal affective stimuli.

The research on hemispheric asymmetries in emotional

processing has yielded three predominant and somewhat

conflicting theories of cerebral lateralization of


According to the global right hemisphere model,

the right hemisphere is dominant for the perception and

expression of both positive and negative affective

signals. This model is based on a variety of studies

with both normals and neurologically impaired

individuals, showing that emotional stimuli are

processed more accurately by the right hemisphere, and

that individuals with right hemisphere damage are

impaired in processing this information (Ley & Bryden,

1979; Schwartz, Davidson, & Maer, 1975; Borod,

Vingiano, & Cytryn, 1988; Borod, Kent, Koff, Martin,

and Alpert, 1988; David, 1989; Heilman, Scholes, &

Watson, 1975; Bowers, Coslett, Bauer, Speedies, &

Heilman, 1987; Blonder, Bowers, & Heilman, 1991).

These findings apply to studies using faces, prosody

and gestures as stimuli.

A contrasting theory, the bivalent model, asserts

that the right hemisphere is specialized for processing

negative emotion while the left hemisphere is

specialized for positive emotion. Studies in support

of this model have found that, among post-stroke

patients, those with left-sided lesions are

significantly more likely to be catastrophically

depressed while right hemisphere damaged patients are

more likely to be indifferent or inappropriately

cheerful (Gianotti, 1972; Robinson, Kubos, Starr, Rao,

& Price, 1984).

Studies with normals have demonstrated left

hemisphere superiority in processing positive affect

and right hemisphere superiority in negative affect

(Natale, Gur, & Gur, 1983; Ladavas, Nicoletti, Umilta

and Rizzolatti, 1984; Reuter-Lorenz, Givis and

Moscovitch, 1983; Schiff & Lamon, 1989), as well as EEG

asymmetries, such that there is greater right frontal

activation during negative emotional experience, and

greater left frontal activation during positive

emotions (Davidson, Schaffer, & Saron, 1985; Davidson &

Fox, 1989). Unlike studies which support the right

hemisphere model, the majority of these studies

examined the expression, rather than the perception, of


The variant model is a partial integration of both

of the above models. Support for the bivalent and

global right hemisphere models have yielded seemingly

inconsistent findings. However, when examined more

closely, it becomes apparent that most of the support

for the bivalent model comes from studies examining the

experience of emotion, while studies demonstrating

right hemisphere superiority for emotional processing

have focused on the perception and interpretation of

affective signals. The variant model tries to

reconcile these findings, by proposing that the

anterior regions of the right and left hemispheres are

differentially involved in mediating the expression of

negative and positive emotions, respectively, whereas

right posterior regions are dominant for the

interpretation of nonverbal emotional signals,

regardless of valence. There is evidence from normals

(e.g. Davidson, Schwartz, Saron, Bennett, & Goldman,

1979) and neurologically-impaired patients (e.g.

Bowers, Glantz, Morris, Blonder, & Heilman, 1991) that

emotional processing may occur in this way.

Taken together, findings from these three models

(global right hemisphere, bivalent, and variant) show

clear evidence for hemispheric asymmetries in the

processing of emotional stimuli. The right and left

hemispheres are specialized for different aspects of

affective analysis, although there is still debate over

specifically what roles each fulfills. However, this

work has primarily been done with consciously perceived

stimuli. If affective information is being processed

subliminally, how is it being done? Is there a

lateralization effect at this level as well, and if so,

is it consistent with that found in optimal stimuli?

If Murphy's effect, that suboptimal emotional stimuli

are being processed, can be replicated, then the next

logical question is how is this processing organized

hemispherically. Specifically, how do the right and

left hemispheres process positive and negative

emotional stimuli when presented suboptimally. Such

information will expand knowledge about emotional

processing, and integrate these two major areas,

suboptimal perception and hemispheric lateralization.

The purpose of these studies is to determine if

Murphy's findings are robust enough to be replicated in

another setting and to discover under what conditions

this phenomenon exists. If these results can be

replicated, then the question of hemispheric

differences in suboptimal emotional processing will be

addressed. The methodology used in all four studies is

a modification of the procedures used by Murphy (1990).



This study attempted to replicate Murphy's study,

examining subjects' perception of optimal and

suboptimal affective stimuli. This was determined by

measuring subjects' liking ratings of neutral

ideographs after an affective face prime which varied

according to valence (positive and negative), presented

for 4 msec.

If Murphy's findings are reliable, then ideographs

should be more liked when preceded by suboptimal

positive primes (smiling faces) and less liked when

preceded by suboptimal negative primes (scowling


Murphy (1990) found no effects on preference when

primes were presented optimally. According to her

interpretation, subjects were more able to process

nonaffective information with more time, thus the

impact of the affect was decreased. This suggests that

there will be no effects on liking ratings when primes

are presented optimally.



Subjects were 32 right-handed male University of

Florida and Santa Fe Community College undergraduates.

Handedness was determined by the Edinburgh Inventory

(Oldfield, 1971). Subjects were randomly assigned to

either the optimal condition or suboptimal condition,

resulting in 16 subjects in each condition. Only

right-handed male subjects were used because in the

majority of studies examining emotional perception, the

findings apply primarily to right-handed males.

Results using right-handed females and left-handers

sometimes yield similar findings, but in general are

less consistent and clear-cut, suggesting that there

may be differential emotional processing in these

populations. There is evidence (e.g. Wylie & Goodale,

1988; Natale, Gur, & Gur, 1983) that the lateralization

of cognitive and emotional functions in women and left-

handers may be different than that of right-handed

males, the population most studied.

Stimuli and Equipment

The prime stimuli were the affective faces and

neutral polygons used by Murphy in her study (the

stimuli were obtained directly from Dr. Murphy). The

affective prime stimuli consisted of male and female

faces expressing happiness and anger. As reported by

Murphy (1990), each face was photographed against a

black background with a black cloth covering any

clothing that might otherwise be visible.

Photographs of five different males and five

different females were used. Each individual was

photographed twice, once smiling and once scowling, for

a total of 20 different stimuli, 10 positive and 10

negative. The neutral primes consisted of 20 randomly

chosen polygons. The target stimuli were 45 different

Chinese ideographs, considered to be effectively

neutral, novel and ambiguous, according to Murphy

(1990). As she had already used these stimuli to

obtain her findings, and the purpose of this study was

to replicate her methodology, no further ratings were

done at this time. All stimuli were presented as


Two slide projectors, each outfitted with a

Uniblitz shutter, were used to present the prime and

target stimuli onto a screen approximately 1 meter in

front of the subject. Subjects were seated slightly to

the right of the screen. An IBM-AT micro computer

controlled the advancement of the slide carousels as

well as the sequential activation of the Uniblitz

shutters. The aperture speed of the shutters,

calibrated to be accurate to within ten percent of the

selected shutter speed, was controlled by a BRS Digibit

Logic System, using a model 405 relay device and 453

precision one-shot timing mechanism. Stimuli were

projected at approximately eye-level for subjects. The

prime stimuli were shown for either 4 msec suboptimall

condition) or 1000 msec (optimal condition). There was

an intertrial interval of 2000 msec between the

completion of the subjects' rating of the ideograph and

the next presentation of the fixation point. Luminance

on the screen was approximately 60 cd/m2.


Testing began with the subject sitting comfortably

in a quiet room. Subjects were told the study dealt

with snap judgments of novel stimuli. They were told

that they would be presented with Chinese characters

which they were to rate on a 1-7 scale depending on how

much they liked the ideograph, with "I" meaning they

didn't like the ideograph at all and "7" meaning they

liked the ideograph a great deal. This rating was done

verbally after each presentation, and was recorded by

the examiner. Use of this scale represented a

departure from Murphy's procedure, in which a 5-point

scale was used. Previous research has shown that 7-

point scales tend to be as sensitive, or more

sensitive, than 5-point scales (Cicchetti et al as

cited in Rasmussen, 1989).

Subjects participated in one of two experimental

conditions: the suboptimal exposure condition or the

optimal exposure condition. In both conditions the

experimental trials consisted of the presentation of a

fixation light for 1000 msec, followed immediately by

the onset of the prime stimulus (either a face or a

random polygon). In the suboptimal condition, the

primes were shown for 4 msec. The prime was

immediately followed by presentation of the target

(Chinese ideograph) for 2000 msec. In this way, the

Chinese ideograph served both as a pattern mask and as

the target stimulus.

The optimal exposure condition was identical to

the suboptimal condition except that the prime was

shown for 1000 msec (rather than 4 msec) prior to

presentation of the target ideograph for 2000 msec. As

subjects in this condition were aware of both the prime

and the target, they were told that there would be two

slides presented on each trial. To explain the

presentation of the primes, subjects were told that

there would be "other experimental conditions" in which

they would be asked to make different judgments

involving the primes but in "this condition" they were

only to rate the second slide. Instructions in both

the optimal and suboptimal conditions are those

reported by Murphy (1990).

Both the suboptimal and optimal conditions began

with 5 "no prime" trials in which no prime stimulus was

presented prior to the target stimulus. This was

followed by 40 experimental trials in which positive

(smiling faces), negative (scowling faces) and neutral

(polygons) primes were randomly, but equally often

presented to the subject. In order to allow for valid

comparisons, ten of the target ideographs were

presented twice, once paired with a positive prime

(happy face) and once paired with a negative prime (the

same individual making an angry face). On one

presentation the ideograph was upright and on one

presentation it was sideways. This was done to prevent

a "mere exposure" effect, that is, having the subject

prefer the ideograph the second time because of having

previously seen it (e.g. Zajonc, 1980).

The order of presentation was counterbalanced,

such that 5 of the ideographs were presented upright

first and 5 were seen sideways first. Also, 5 were

presented upright with a happy face and sideways with

an angry face, while the other 5 were presented in the

reverse fashion.

In both the suboptimal and the optimal conditions,

subjects verbally rated each ideograph according to how

much they liked it. This rating was made after each

trial. After subjects completed the experimental

trials, they were asked if they had noticed anything

unusual, and were encouraged to speculate as to the

purpose of the experiment.


In order to determine whether the suboptimal prime

stimuli were in fact suboptimall" and outside

awareness, a forced-choice test of awareness was given

following completion of all the experimental trials.

This forced-choice test was conducted in order to

ascertain whether or not subjects were able to detect

pattern masked stimuli that had been exposed for 4

msec. In this task, subjects were told that a briefly

presented face would be shown and then followed by a

Chinese ideograph. The procedure used was similar to

that used in the experimental task. A fixation point

was presented (1000 msec), followed by a 4 msec

exposure to a face. Primes were immediately followed

by a 2000 msec presentation of an ideograph, which also

served as a backward mask.

Subjects were then given a recognition test. They

were shown two test faces for 2000 msec: one face which

was the actual prime and one face which served as a

foil. Subjects were asked which of the two faces they

thought was the prime. The same faces were used during

the post-test as during the study. This increased the

possibility that subjects would recognize the faces, as

repeated exposure to a degraded stimulus decreases

recognition threshold (e.g. Seamon, Marsh & Brody,

1984). Thus, lack of recognition at this level almost

guarantees that the faces were not consciously seen

during the study.

During the recognition task, faces were vertically

presented, one above the other, rather than side to

side, to prevent any biasing effects of visual field

presentation. On half the trials the actual prime was

on top, and on half the foil was on top. The prime and

the alternate were of the same gender and emotional

expression, as it is possible that affective or

cognitive information may be extracted even though the

specific stimulus is not recognized (e.g. Marcel, 1980;

Murphy, 1990). Each subject participated in 12 forced

choice trials. After these trials, subjects were fully

debriefed and thanked for their participation.


The dependent variable was average liking rating

(1-7) of the target ideographs. The affective valence

of the prime (happy, angry, neutral) was the

independent variable. Data from the suboptimal and

optimal exposure conditions were analyzed separately

using one way repeated measures ANOVAs with one within

subject factor (prime type). To compare conditions a

repeated measures ANOVA with one within subjects factor

(prime type) and one between subjects factor

(condition) was used. T-tests were also used to

compare results, in order to make these the current

data analyses equivalent to Murphy's.


Two subjects in the suboptimal condition performed

significantly better than chance on the post-test

(correctly identifying 9 of the 12 faces). It was felt

that subjects who scored 75% or more of their responses

in any one direction (correct or incorrect) were

potentially being influenced by the subliminal

information, as such a pattern of responding was highly

unlikely, simply by chance. Their data were discarded

as this suggested that they may have perceived the

primes consciously, without realizing it (in other

words, the primes were above their objective, but not

their subjective thresholds). Repeated measures ANOVAs

were performed on the data from the remaining 32

subjects to determine if the average liking rating of

the ideographs differed according to the type of prime

preceding them (positive, negative, neutral, or no


For the 16 subjects in the suboptimal condition,

there was no significant effect of prime valence on

ideograph ratings, F(3,45)=.23, p>.8. (See table 1,

table 2, figure 1). In the optimal condition there

were no significant findings. Prime valence did not

have a significant effect on ideograph ratings,

F(3,45)=.77, p>.5. (See table 3, table 4, figure 2).

There was no significant effect of condition for liking

ratings of ideographs in any condition F(3,90)=.52,

p>.6. (See table 5). There was also no significant

prime by condition interaction F(3,90)=.52, p>.6. (See

table 6).

To ensure that these analyses were comparable to

Murphy's, paired t-tests were also run on the data.

This was done to ensure that differences in findings

were not due to the use of more stringent statistical

techniques. There were no significant effects in

either condition when t-tests were used (see table 7,

table 8).


The purpose of this study was to replicate

Murphy's findings using her methodology and stimuli.

Murphy demonstrated that Chinese ideographs which were

preceded by a subliminal (4 msec) prime of a happy face

were liked significantly more than ideographs preceded

by an angry face prime or a neutral shape. Ideographs

preceded by an angry face received liking ratings

significantly lower than average and lower than other

primes. Ideographs preceded by neutral primes were not

rated significantly different from average. Murphy

found no effects of prime valence when the primes were

presented optimally.

This study was unable to replicate these results

in the suboptimal condition. There were no significant

differences in the liking ratings of ideographs in the

different prime conditions (happy, angry, neutral).

There were also no differences in the optimal

condition, as reported by Murphy. This study found no

difference between liking ratings between the optimal

and suboptimal conditions, further suggesting that

viewing the primes had no effect.

However, there were some procedural differences

between this protocol and the one reported by Murphy.

First, and perhaps most significant, the exposure

duration of the primes in the suboptimal condition

(presumably 4 msec) did not correspond to that reported

by Murphy. Although the Uniblitz shutter used in this

study was set at 4 msec, the duration exposure of the

primes approximated only 1 msec, since apparently it

took the shutter 3 msec to begin opening.

It was a surprise to discover that the shutter,

although set to open for 4 msec, only opened for 1

msec. After Murphy's findings were not replicated, the

Uniblitz manufacturer was contacted, to make sure that

the shutter used in this study was fully operational

and opening properly. It was at this time that the


company engineers explained that it took 3 msec for the

electrical impulse requiring the shutter to open to

actually reach the shutter, due to the slowness of the

mechanical relay system. It then took another 2-3 msec

for the shutter to open completely. After the full

exposure time ended, it then took another 2-3 msec for

the shutter to close. Thus, when the shutter was set

to be open for 4 msec, for the first 3 msec the shutter

itself made no movement. In the remaining 1 msec it

began to open, but did not have time to open

completely, and then closed. According to the

manufacturer, the exact opening and closing times are

unknown, but vary within a few msec. Because of the

opening and closing times, during which the slide is

partially visible, it is in fact impossible to expose

the slides for exactly 4 msec.

Thus, in the suboptimal condition, instead of

opening for 4 msec, the shutter never opened

completely, and was open partially for no more than 1-2

msec. At no time was the slide which the subjects saw

free of movement artifact. It was thought that this

may explain the failure to replicate Murphy's results,

as she reported that her subjects viewed the slides for

a full 4 msec. Therefore, Study 2, identical to Study

1, but with an increased exposure time, was conducted.

In addition, there was one minor methodological

variation from Murphy's protocol which was corrected.

In Study 1, subjects were seated to the right of the

screen, not directly in front, as in Murphy's study.

Therefore, in Study 2, subjects were seated directly in

front of the screen.

There were two other differences from Murphy's

methodology. She employed a red filter over her

shutter, to mask the contrast differences between the

slides. This was not used in this study, as the value

of it was not clear. There was no evidence that

lightness differences affected subjects' perceptions.

According to self-report, none of the subjects were

aware of seeing the faces during the study. During the

post-test some subjects reported seeing a flash of

light, but this did not vary between subjects who

failed the post-test and those who passed it. Thus,

there is no evidence that the primes were seen

optimally, in spite of the evident contrast.

In addition, subjects in this study employed a 7-

point, rather than a 5-point scale to rate the

ideographs. It was felt that a 7-point scale would

increase sensitivity to minor differences, thus

maximizing the likelihood of replicating Murphy's

results. Therefore, use of this scale was maintained,

although simply the fact that it represents a

difference from Murphy's procedure may have played a

role in the failure to replicate her findings.

Interestingly, Murphy (1990) reported that none of

her subjects were aware that certain ideographs were

repeated. This contrasts with subjects in the present

study, many of whom recognized the ideographs the

second time they were shown, even though they had been

turned sideways. This suggests that there may have

been some differences between the subjects used by

Murphy and those used in this study. Also, some

subjects who recognized the ideographs reported trying

to be consistent by giving them the same rating the

second time. It is surprising that none of Murphy's

subjects recognized the ideographs.

A second difference between subjects is that she

reported that none of her subjects failed the

post-test. In this study, two subjects in the

suboptimal condition (who were discarded) and 2

subjects in the optimal condition differed

significantly from chance in their post-test

performances. Given these differences, it is possible

that there were either some subject population or

methodological differences between these two studies

which may have contributed to the discrepancy in the


Table 1: Means and Standard Deviations of Ideograph
Liking Ratings According to Prime Type
Study I Suboptimal Condition

Prime Type Mean Standard Deviation

Happy 4.24 .685

Angry 4.34 .888

Neutral 4.35 .602

No-Prime 4.19 .750

Table 2: ANOVA, Study I Suboptimal Condition

Source Sum of df Mean F prob>F
Squares Square
Prime .31 3 .10 .23 .874

Within 19.97 45 .44


c c
*H 0

4-JC -H

cc -H
:3 J CU

E OC*1
u C-

C D..
0 -0


a to oLL

r- (0 LO) It c) (M 0
SOUTLJl qdej6oepI

Table 3: Means and Standard Deviations of Ideograph
Liking Ratings According to Prime Type
Study I Optimal Condition

Table 4: ANOVA, Study I Optimal Condition

Source Sum of df Mean F prob>F
Squares Square
Prime 1.10 3 .37 .77 .518

Within 21.51 45 .48

Prime Type Mean Standard Deviation

Happy 3.89 .833

Angry 4.17 .833

Neutral 4.10 .598

No-Prime 4.24 .801

N o- to I mo
s6UTIBU qdej6oapi

Cm Ti- 0

: ",..

S '' .' i. s '0 l'.
Si. '. i" ,' .'?' *.2; :

-.. ,., .., ; *. -; ..,' .:

. ..... .'.:>" 1A .^>h i..." ,..


















Table 5: ANOVA, Study I Suboptimal vs. Optimal
Conditions- Main Effect

Source Sum of df Mean F prob>F
Squares Square
Cond 1.08 1 1.08 1.16 .290

Within 27.87 30 .93

Table 6: ANOVA, Study I Suboptimal vs. Optimal
Conditions Interaction Effect

Source Sum of df Mean F prob>F
Squares Square
Cond by .71 3 .24 .52 .672

Prime .69 3 .23 .50 .682

Within 41.48 90 .46

Table 7: Paired T-Test Values
Study I Suboptimal Condition

Paired t-test t value p value

Happy vs. Angry -.69 .5

Happy vs. Neutral -.75 .463

Happy vs. No-prime .20 .846

Angry vs. Neutral -.05 .962

Angry vs. No-Prime .45 .658

Table 8: Paired t-test values
Study I Optimal Condition

Paired t-test t value p value

Happy vs. Angry .88 .391

Happy vs. Neutral -1.17 .259

Happy vs. No-prime -1.30 .214

Angry vs. Neutral .27 .789

Angry vs. No-Prime -.29 .778



The purpose of this study is to correct

methodological differences between Study I and Murphy's

studies (e.g. slide exposure duration, placement of

subject) and once again attempt to replicate Murphy's

findings, with a different set of subjects. The

hypotheses for this study are the same as for Study 1.

Namely, it is expected that, in the suboptimal

condition, ideographs preceded by a positive prime will

receive the highest ratings, ideographs preceded by a

neutral prime will receive average ratings, and

ideographs preceded by angry face primes will receive

ratings significantly lower than ideographs in the

other prime conditions.



Subjects were 10 right-handed males between the

ages of 18-34. Only 10 subjects were run as this study

was done as a type of pilot, to determine if correcting

for certain methodological variations in Study I would

have an impact on results. Had there been evidence

after 10 subjects that there may have been a prime


effect, 6 more subjects would have been run, to achieve

equivalence with the previous studies. There was no

optimal condition as the purpose of this study was to

replicate the subliminal condition as reported by

Murphy. These subjects were different from those used

in study 1. Handedness was determined by the Edinburgh

Inventory (Oldfield, 1971).

Stimuli and Equipment

Materials in Study 2 were identical to those used

in Study 1.


The procedures used in study 2 were almost

identical to those used in study 1, with three

exceptions. The exposure time of the shutter was

increased to 10 msec, so that the slide appeared for at

least 4 msec. During the first 3 msec, there was no

shutter movement at all, until the signal reached the

shutter. The next 2.5-3 msec consisted of the shutter

opening fully. The shutter then remained fully open

for the remaining 4 msec. The slides were partially

exposed during closing for an additional 3-4 msec.

This resulted in a total exposure time of 9.5-11 msec.

In addition, only the suboptimal condition was

conducted. Finally, all subjects sat directly in front

of the screen, approximately 1.5 meters from the

screen, instead of to the side.


A forced choice test of awareness identical to

that done in study 1 was given.


The dependent variable was the average liking

rating (1-7) of the target ideographs. The affective

valence of the prime (happy, angry, neutral) was the

independent variable. A repeated measures ANOVA, with

prime type (happy, angry, neutral, no-prime), as the

within-subjects factor, was conducted to analyze the

results. Paired t-tests were done to make comparisons

with Murphy's results.


A repeated measures ANOVA was performed to compare

the average liking rating between ideographs with

different affective primes. There were no significant

difference between the liking ratings of ideographs

preceded by primes of different valences, F(3,27)=.93,

p>.4. (see table 9, table 10, figure 3). To ensure

equivalency with Murphy, paired t-tests were performed.

These tests also revealed no significant effect of

prime type (see table 11).

In spite of the increased exposure time, there was

no substantial increase in recognition on the

post-test. In this study, the data of subjects who

failed the post-test (n=2) was included in the final

analyses, as the pattern of responses of these subjects

during the actual study were no different than those of

the remaining subjects.


It is not clear if Murphy showed her primes for a

full 4 msec, or for only a fraction of that time.

According to her report, she set the shutter at 4 msec,

as was done in Study 1. Therefore it is possible that

she also showed her slides more briefly than intended.

According to the manufacturers of the shutter, it is

not possible to show a slide for exactly 4 msec with

this shutter. This is because if the shutter does open

for a full 4 msec, there will also be opening and

closing time during which the slide is partially

visible, and the exact length of this time is variable.

However, even with an increased exposure time such

that the primes were fully exposed for 4 msec and

partially exposed for 3-6 msec, it was not possible to

replicate her findings. Thus, the length of exposure

does not appear to explain the failure to replicate

Murphy's findings.

Another possible reason for the failure to

replicate Murphy's findings was that her affective

primes did not accurately reflect anger and happiness.

Murphy reported that she used happy and angry

expressions because the universal recognition of this


affect would decrease "the possibility of idiosyncratic

responses to the primes" (Murphy, 1990). However, she

does not report gathering independent ratings to

validate her judgments of the faces even though there

is a great deal of individual variability concerning

the perception and expression of emotion. Therefore,

as an adjunct to this study, independent ratings were

obtained on her affective face primes.

Twenty right-handed males identified the emotions

on each face. Raters were run individually. Each

rater was shown a slide, one at a time, of each face

used by Murphy and asked to identify the emotion on the

face. The physical set-up was identical to studies I

and II. Each face was shown for at least 1 sec (longer

if the subject requested to see it again) after which

the subject rated it. Of her ten "happy" primes, 9

achieved an 80% agreement criterion (e.g. 80% of the

raters identified the face as happy). Only 2 of her 10

angry primes reached this criterion. Apparently,

subjects in these studies did not view 9 of her 20

primes as effectively salient, and that may be why they

had no effect. Therefore, a third study was done, with

a new set of affective prime stimuli which were chosen

based on affective ratings by independent judges.

In addition, the red filter that Murphy used to

mask the contrast of her slides was added. Although


its value was not clear, it is possible that it played

a role in the subjects' perceptions. Finally, the use

of polygons rather than faces as neutral primes seemed

inappropriate. It may be that subjects would react

differently to a face than to a polygon, regardless of

emotional expression, because of the type of stimulus.

There is no way of knowing if the subject is reacting

to the valence of the stimulus (happy, angry, neutral)

or the nature of it (face vs. inanimate shape). Using

neutral faces for the neutral stimulus would eliminate

that concern and permit comparisons less marked by


Table 9: Means and Standard Deviations of Ideograph
Liking Ratings According to Prime Type
Study II

Prime Type Mean Standard Deviation

Happy 4.09 .576

Angry 3.89 .624

Neutral 4.18 .554

No-Prime 4.28 .761

Table 10: ANOVA Study II

Source Sum of df Mean F prob>F
Squares Square
Prime .82 3 .27 .93 .439

Within 7.91 27 .29
Cell _____






) -,-

> .JI-4

C) 1-)

c 1-1
-. .. .


I- (0 lO t CO C 0
S6UTIeU qdej6oapI

Table 11: Paired T-Test Values
Study II

Paired t-test t value p value

Happy vs. Angry 1.14 .285

Happy vs. Neutral -.99 .347

Happy vs. No-prime -.65 .532

Angry vs. Neutral -1.79 .106

Angry vs. No-Prime -1.21 .256



The hypotheses for this study are identical to the

hypotheses for study 1. If Murphy's findings are

reliable, ideographs should be more liked when preceded

by suboptimal positive primes (smiling faces) and less

liked when preceded by suboptimal negative primes

(scowling faces).

Murphy (1990) found no effects on preference when

primes were presented optimally. According to her

interpretation, subjects were more able to process

nonaffective information with more time, thus the

impact of the affect was decreased. This suggests that

there will be no effects on liking ratings when primes

are presented optimally.

The effect of adding neutral faces as a neutral

prime is unknown. It is expected that there will be no

significant difference between neutral face primes and

polygon primes, and that positive primes will produce a

higher liking rating and negative primes a lower liking

rating than either neutral prime.



Subjects were 32 right-handed males between the

ages of 18-28, different from those used in the

previous two studies. Handedness was determined by the

Edinburgh Inventory (Oldfield, 1971). Subjects were

randomly assigned to either the optimal condition or

suboptimal condition, with half the subjects in each


Stimuli and Equipment

Materials in study 3 were similar to those used in

study 1 except that different primes were used. The

target ideographs and neutral polygons were used as

before. However, a new set of faces were used. As in

Murphy's study, each face was photographed against a

black background with a black cover over any clothing

that might otherwise be visible. Photographs of five

different males and five different females were used.

Each individual was photographed three times, once

smiling, once scowling, and once with no emotional

expression, for a total of 30 different stimuli, 10

positive, 10 negative, and 10 neutral. The 10 neutral

faces were used with 10 polygons as neutral stimuli.

The only faces used were those which achieved an

80% agreement criterion by the raters on their

respective emotions. Thus, positive primes were those


in which at least 80% of the raters considered the face

happy, negative primes were those in which at least 80%

of the raters considered the face angry, and neutral

face primes were those in which at least 80% of the

raters considered the face expressionless.

Ratings were done by 20 right-handed males who

participated in studies I and II. Each face was rated

separately for attractiveness and intensity of emotion

on a 7 point Likert scale by at least 10 and up to 20

males. The attractiveness ratings were obtained on

faces with neutral expressions, so as not to be

influenced by the emotional expression. Typically,

smiling faces are likely to be viewed as more

attractive than scowling faces.

There was no significant difference between

intensity ratings for happy faces (5.24) vs. angry

faces (5.14). Intensity ratings for emotional faces

ranged from 3.9 to 6.77 (on a 7 point scale), while the

average intensity rating for neutral faces was 2.07,

ranging from 1.54 to 2.6 (See table 12). Average

attractiveness ratings ranged from 2.8 to 4.4.

However, there was a significant amount of individual

variation, such that all faces received both high (5-7)

and low (1-2) attractiveness ratings.


Procedures for study III were almost identical to

study 1, with three exceptions. Subjects sat directly

facing the screen, a red filter was used over the slide

projector, to mask contrast, and the exposure duration

of the slides was increased.


As in studies I and II a forced choice test of

awareness was given. Procedures were identical. The

affective primes were those used in Study III.


Repeated measures ANOVAs were used to analyze

differences in average liking ratings, with valence of

prime (happy, angry, neutral faces, neutral shapes, no-

prime) the within subjects factor. To compare results

between conditions a repeated measures ANOVA with one

within subjects factor (prime type) and one between

subjects factor (condition) was done. Paired t-tests

were performed to further specify any differences

detected from the ANOVAs. A Bonferroni t-test

correction was used to prevent an increase in Type I

errors due to the number of t-tests done.


Subjects who failed the post-test and those who

passed it were used. Subjects who answered 75% or more

of their responses in the same direction (correct or

incorrect) were considered to have failed. However,

during the test itself, these subjects did not perform

any differently than subjects who passed the post-test,

and the data did not change when these subjects were

included; therefore, all subjects were used.

Suboptimal Condition

In the suboptimal condition, there was a

significant effect of prime type on ideograph ratings

F(4,60)=3.10, p=.022. (See table 13, table 14, figure

4). Using a Bonferroni t-test correction, ideographs

preceded by neutral shapes (M=4.59, sd=.382) were liked

significantly more than ideographs preceded by angry

faces (M=4.04, sd=.809; t=-3.13, p<.01) and neutral

faces (M=4.11, sd=.660; t=3.16, p<.01). There was a

trend for ideographs preceded by neutral shapes to be

preferred to those preceded by happy faces (M=4.21,

sd=.648; t=-2.75, p>.01).

There was no significant difference (t=-1.73,

p=.105) between ideographs which were preceded by

positive primes (M=4.21, sd=.648) and those preceded by

negative primes (M=4.04, sd=.809). Further, there was

no significant difference (t=-.41, p=.688) between

ideographs preceded by angry vs. neutral faces (M=4.11,

sd=.660) or between ideographs preceded by happy faces

(M=4.21, sd=.648) vs neutral faces (t=.74, p=.473).

Optimal Condition

In the optimal condition, there was a significant

effect of prime type on ideograph rating F(4,60)=5.45,

p=.001) (See table 15, table 16, figure 5). Ideographs

preceded by happy faces (M=3.99, sd=.740) were liked

significantly more than ideographs preceded by angry

faces (M=3.66, sd=.705; t=-2.95, p<.01) or neutral

faces (M=3.72, sd=.582; t=3.49, p<.005). Ideographs

preceded by a neutral shape prime (M=4.06, sd=.704)

were liked significantly more than ideographs preceded

by a neutral face (M=3.72, sd=.582; t=3.50, p<.01), and

there was a trend for them to be preferred to

ideographs preceded by negative primes (M=3.66,

sd=.705, t=-2.89, p>.01).

There were no differences between angry and

neutral face primes (t=-.57, p=.579) or between happy

face primes and neutral shape primes (t=-.56, p=.582).

It is possible that ideographs preceded by neutral

shapes were preferred because there was something

intrinsically more likeable about those ideographs. In

the first two studies, there were 20 ideographs

preceded by polygons. In this study, 10 of these

ideographs were preceded by polygons, and 10 by neutral

faces. To determine if the ideographs preceded by

polygons were more inherently likeable, data from Study

I were re-analyzed. The liking ratings of the

ideographs preceded by polygons vs. neutral faces

(which, in Study I, were all preceded by polygons) were

compared using a paired t-test. There was no

significant difference (t=-.08, p=.938) between liking

ratings of ideographs later used with neutral faces

(M=4.17, sd=.589) and those later used with neutral

shapes (M=4.18, sd=.588).

Optimal vs. Suboptimal Conditions

A repeated measures ANOVA was performed to

determine if ratings for ideographs preceded by primes

of a particular valence varied between optimal and

suboptimal conditions. There were no significant

effect of condition on ideograph liking ratings

F(1,30)=1.62 p>.l (See table 17). There was also no

significant prime by condition interaction effect

F(4,120)=1.62, p>.l (See table 18).


All subjects, including those who failed the post-

test, were used in this study, as there was no

difference between the data with or without these

subjects. Although some subjects failed the post-test,

there was no evidence that this decreased threshold

influenced their responding to the test itself.

This study attempted to replicate Murphy's

methodology using improved stimuli. Affective faces

were rated, and only those consistently (by 80% of


raters) identified as happy, angry or neutral were used

as positive, negative or neutral primes, respectively.

Emotionally neutral faces, as well as polygons, served

as neutral primes, providing a more appropriate

comparison to affective face primes.

Using these procedures, Murphy's findings were

still not replicated. There were, however, some

unexpected findings. In the suboptimal condition,

ideographs preceded by neutral polygons were preferred

to ideographs preceded by neutral and angry faces, and

there was a trend for these ideographs to be preferred

to ideographs preceded by happy faces. In the optimal

condition, ideographs preceded by neutral shapes

received higher liking ratings than ideographs preceded

by neutral faces, and showed a trend toward having

higher liking ratings than ideographs preceded by angry

faces. Based on analysis of the liking ratings from

study I, there is nothing inherently more preferable

about these ideographs. This suggests that subjects

were responding to the primes.

One possible explanation for why ideographs were

preferred when preceded by a neutral shape may have to

do with priming of expectations. The shape and the

figure are similar in that they are both abstract

geometric designs. Seeing the shape may create some

sort of expectation that a similar stimulus will

follow. Seeing a face prime would set up a different

type of expectation. It is possible that seeing an

ideograph after a face was contrary to expectations,

and therefore less pleasing.

This finding was stronger in the suboptimal than

in the optimal condition, suggesting that with less

time to analyze the prime subjects relied more heavily

on their expectations. When subjects had more time to

process the complex features of the prime, they may

have become more aware of the differences between the

polygon and the ideograph, and may have been influenced

less by the overall gestalt of the figure and more by

specific features.

It is also possible that a face stimulus is too

complex to be processed in 4 msec. Thus, subjects were

unable to process the information in the face,

precluding their forming a reaction to it. The

polygon, however, is relatively simple, and may not

require extended exposure.

This study found significant differences between

the liking rating of ideographs preceded by positive

and negative primes in the optimal condition, but not

in the suboptimal condition. The faces used in this

study were chosen based on ratings which identified the

faces as clearly happy or angry. Thus, these faces

seem to be tapping the positive and negative emotions


intended. It is not surprising that subjects would be

affected by seeing these positive or negative stimuli.

This finding was unexpected only because Murphy did not

report it. However, given that her faces were less

representative of happiness and anger, based on

ratings, it is possible that her faces did not evoke

positive and negative emotions strongly enough to

produce an effect. It is in fact intuitively sensible

that subjects' liking ratings would be affected by

seeing a positive or negative stimulus immediately

prior to the ideograph: being exposed to something

unpleasant would tend to bias someone towards seeing a

fairly ambiguous stimulus in an unpleasant light

(projecting that unpleasantness onto the subsequent

stimulus), whereas a pleasant stimulus would tend to

have the reverse effect.

However, there is still the unanswered question of

why it has not been possible to replicate Murphy's

findings of suboptimal affective priming. There are

two significant remaining factors that may have

influenced the results.

The first question concerns the exposure duration

of the primes. The shutters appear to have a great

deal of variability, such that opening time, closing

time, and the delay before the shutter begins to open

must be considered in determining an exposure time. It

is in fact impossible to show a slide for only 4 msec,

because the slide will also be partially exposed while

the shutter opens and closes. Exactly how long these

processes take varies within a window of quite a few

milliseconds. Therefore, it is possible that Murphy

employed greater (or lesser) exposure times than

intended. One possibility then, is that this

phenomenon of suboptimal affective priming exists, but

not at an exposure time as brief as 4 msec.

A second problem with the use of the shutters is

the existence of a movement artifact. Because the

slide is visible during the opening and closing of the

shutter, the subject is seeing changes in the prime,

and has only a brief period in which the slide is

static. Because of the movement of the shutter, and

the difficulty in knowing the exact exposure time, a

tachistoscopic presentation of the primes and

ideographs may be more likely to tap this phenomenon.

This would ensure a definite exposure time, without

movement, and thus seems to be a more appropriate

situation in which to test for the existence of

subliminal affective priming.


Table 12: Intensity of Emotional Expression Ratings for
Face Primes
Study III

Face Type Minimum Maximum Average
Rating Rating Rating

Happy 4.31 6.10 5.24

Angry 3.90 6.77 5.14

Neutral 1.54 2.60 2.07

Table 13: Means and Standard Deviations of Ideograph
Liking Ratings According to Prime Type
Study III Suboptimal Condition

Prime Type Mean Standard Deviation

Happy 4.21 .648

Angry 4.04 .809

Neutral 4.11 .660
Neutral 4.59 .382
No-Prime 4.19 .519

Table 14: ANOVA, Study III Suboptimal Condition

Source Sum of df Mean F prob>F
Squares Square
Prime 2.95 4 .74 3.10 .022

Within 14.27 60 .24

(u t CIO CM J 0
sBuTwuU qdej6oepi

C 0
*H -H

(0 *
cu -H

0) 0

L. M

0i 0


0) 4-J
--I ()


w >





Table 15: Means and Standard Deviations of Ideograph
Liking Ratings According to Prime Type
Study IIT Optimal Condition

Prime Type Mean Standard Deviation

Happy 3.99 .740

Angry 3.66 .705

Neutral 3.72 .582
Neutral 4.06 .704
No-Prime 4.16 .811

Table 16: ANOVA, Study III Optimal Condition

Source Sum of df Mean F prob>F
Squares Square
Prime 3.03 4 .76 5.45 .001

Within 8.35 60 .14
Cell _______


-- I



z ..
0) CL

.0 1-- O
--------------------- ) >~ C

L- *L +

a-. 0


.. LO




co ( lO cn cO C 0
S6UTI.BU qdvj6oapi

Table 17: ANOVA Study III
Suboptimal vs. Optimal Conditions
Main Effect

Source Sum of df Mean F prob>F
Squares Square
Cond 3.84 1 3.84 2.60 .117

Within 44.30 30 1.48
Cell _______

Table 18: ANOVA Study III
Suboptimal vs. Optimal Conditions
Interaction Effect

Source Sum of df Mean F prob>F
Squares Square
Cond by 1.22 4 .30 1.62 .174
Prime 4.77 4 1.19 6.32 .001

Within 22.62 120 .19
Cell _______



There are many different possibilities as to why

Murphy's findings were not replicated by the previous

studies. One is that the shutters used are too inexact

to ensure a precise exposure time, and that, therefore,

there were unintentional differences between the

exposure duration used by Murphy and those used in

these studies. If that is the case, it is expected

that Murphy's results will be replicated by this study,

which will provide increased exposure times, as Murphy

might have done. Specifically, liking ratings of

ideographs should vary depending on the suboptimal

prime preceding them, such that positive primes will

produce higher than average ratings, negative primes

will produce lower than average ratings, and neutral

primes will result in average ratings. It is not known

at which exposure times this finding will occur.

It is possible that the specific exposure time is

not a factor in this phenomenon. Murphy's findings of

the influence of suboptimal perception may simply not

be generalizable or replicable in this setting. If


that is the case, it is expected that there will be no

effect of suboptimal prime presentation.

Optimal primes, however, are expected to influence

liking ratings of ideographs, as in Study III. Happy

face primes should increase ratings, and angry face

primes should decrease ratings. The faces used in this

study were rated as were those used in study III, and

therefore should have a similar effect.



Subjects were 48 right-handed male University of

Florida undergraduates, different from those subjects

used in studies I-III. Handedness was determined by

the Edinburgh Inventory (Oldfield, 1971). Subjects

were randomly but evenly assigned to one of three

exposure conditions (one optimal condition and two

suboptimal conditions), such that there were 16

subjects in each condition.

Stimuli and Equipment

The prime stimuli consisted of random polygons and

happy and angry faces. The polygons serving as neutral

primes were those used by Murphy (1990) and in studies

I-III. Affective primes consisted of faces considered

to be happy or angry by independent raters. The target

ideographs were those used by Murphy, and in studies I-

III, and were considered to be effectively neutral,

novel and ambiguous (Murphy, 1990).

Since subjects viewed primes in two exposure

conditions, they saw twice as many primes. Therefore,

new face primes were required. To ensure consistency

in the ratings of primes, all primes from previous

studies were rerated before being used. In this way,

all primes met the same standards.

Face prime stimuli originated from two sources.

Ten face stimuli came from affective faces used by

Ekman (1972). The remaining 30 were photographed for

this study. Individuals were photographed either

smiling or scowling, against a black background, with a

black cover over any clothing that might be visible.

All faces were then rated by 20 right-handed males.

Only those faces with a 70% agreement rating were used

(e.g. a positive prime face was one in which at least

70% of the raters considered it to be happy). Although

this criterion was less strict than that used in Study

III, it still excluded 6 of Murphy's 20 primes.

All faces used as primes had at least a 70%

agreement rating on emotional expression. Happy faces

had an average intensity rating of 4.48, ranging from

3.42 to 5.84. Angry faces had intensity ratings

ranging from 3.58 to 6.84, with an average of 5.34 (see

table 19).

This procedure yielded a total of 40 different

stimuli, 20 positive and 20 negative. Ten different

males and ten different females were used, with each

individual demonstrating both a happy and an angry

face. Forty polygons were used as neutral stimuli.

Ten polygons were shown twice, turned in different

orientations, to create different shapes. Eighty-five

Chinese ideographs served as target stimuli consisting

of 35 different ideographs (five of the ideographs

served as no-prime controls; the remaining ideographs

were shown more than once, either in the same or a

different orientation).

All primes and target ideographs were made from

photographs into black and white paper copies. All

stimuli were of a size to fit on a G1147 Precision

Stimulus tachistoscope card, and each picture was

pasted onto a card.

All stimuli were shown via tachistoscopic

presentation. A Gerbrands G1135 tachistoscope with

four manually operated visual fields was used. The

subject was seated directly in front of the t-scope


Prime stimuli were presented for two exposure

durations to each subject. In one suboptimal

condition, half the primes were presented for 4 msec

and half for 16 msec. In the second suboptimal

condition, half the primes were shown for 4 msec and

half for 32 msec. During the optimal condition, half

the primes were shown for 500 msec and half for 1000

msec. Target stimuli were always shown for 2000 msec.


Testing began with the subject sitting comfortably

in a quiet room, positioned so that they could easily

lean forward and look through the t-scope viewer.

Subjects were told that this was a study dealing with

the snap judgments of novel stimuli. They were told

that they would see a Chinese ideograph, which they

were to rate on a 1-7 scale, with "1" meaning they did

not like the ideograph at all, and "7" indicating they

liked it a great deal. Because subjects in the optimal

condition were aware of seeing two stimuli, they were

told that the first picture was for "other conditions"

but for this condition they were to focus on rating the

ideograph. Subjects verbally rated each ideograph

after presentation, and the rating was recorded by the


A trial began with a subject seeing a fixation

point lasting for 1000 msec. This was followed by the

prime (happy face, angry face, random polygon), the

duration of which varied depending on the condition.

Primes were randomly but evenly divided such that half

the primes of each valence were shown for one exposure

duration, and half the primes of each valence were

shown for a second exposure time in each of three

experimental condition. In the two suboptimal

conditions, this meant that primes were presented for

either 4 or 16 msec (Condition 1) or 4 or 32 msec

(Condition 2). In the optimal condition, primes were

shown for 500 or 1000 msec.

Immediately following the prime, the target

(Chinese ideograph) was presented for 2000 msec, thus

serving both as a backward pattern mask and as the

target stimulus. The next trial began immediately

following the examiner's recording of the subject's


The first 5 ideographs served as no-prime

controls, so that no prime was presented between the

fixation point and the ideograph. This was followed by

80 experimental trials in which 20 positive (smiling

faces), 20 negative (scowling faces) and 40 neutral

(polygons) primes were randomly presented to the

subject. These 80 trials were evenly divided, such

that half the primes of each valence were presented in

one exposure duration, while half were shown in the

second exposure time.

In order to allow for valid comparisons, ten of

the target ideographs were presented twice at each

exposure time, once paired with a positive prime (happy

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd