ATTENTION, EMOTION, AND THE STARTLE REFLEX
SCOTT RICHARD VRANA
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
After working on this project in relative isolation for
a year, it is a pleasure to think back on all those who
contributed to it. First and foremost, Peter Lang's counsel
has been invaluable during this dissertation and all other
aspects of my graduate training. Any future contribution I
may make to psychology will be due in no small part to Dr.
The other members of the dissertation committee have
contributed in many ways to this document and to my
training. Barbara Melamed provided a wonderful model of a
researcher/clinician working in the anxiety disorders.
Russell Bauer was a fine teacher, clinical supervisor, and
research advisor during my years at University of Florida.
I can think of no better model than Rus as I engage in these
activities as a new faculty member. I was lucky to have
Keith Berg and Pat Miller on my committee and as teachers to
share my interest in developmental psychology. Jane
Pendergast provided answers to my statistical questions on
the dissertation and other projects.
I apologize to Bruce Cuthbert for being perpetually
underappreciative of his many talents and kind help. Alas,
there was too much to appreciate, and too little time.
Ellen Spence was a special friend from my first to my last
day in Gainesville. Mark Greenwald, Margaret Bradley, Dan
McNeil, Ed Cook, and David York all provided intellectual
stimulation and friendship. The Lang lab was (and is) an
exciting, stimulating, unique place to work. I thank all
TABLE OF CONTENTS
ACKNOWLEDGEMENTS. .. . . ii
ABSTRACT . . . vi
INTRODUCTION. . . 1
Startle Reflex Response . 1
Startle and Sensory Modality-Specific Attention 2
Startle and Affect. . . 4
Imagery and Affect. . . 6
Processing Task and Affective Text. . 7
Statement of the Experimental Problem .12
METHOD. . . . 17
Subjects. . . .17
Apparatus . . 17
Stimulus Materials. . . 19
Procedure . . 19
Startle Stimuli . . 22
Design. . . .23
Data Reduction and Analysis . .23
RESULTS . . . .26
Analysis Strategy . . .26
Self-report . . 27
Heart Rate . .. ... .27
Startle Reflex: Content Differences . .36
Startle Reflex: Modality-Specific Attention .49
Startle Reflex: Image Vividness . 51
DISCUSSION . . .57
Processing Fear Sentences . .58
Modality-Specific Effects of Sentence Processing. .60
Imagery as a Cognitive Task . .63
Summary and Conclusions . .67
REFERENCES. . . 70
APPENDIX A STIMULI AND SUBJECT INSTRUCTIONS. .75
APPENDIX B DIFFERENCES BETWEEN STARTLE AND NON-STARTLE
EXPERIMENT . . 82
APPENDIX C HEART RATE FIGURES. . 85
APPENDIX D TABLES OF STARTLE REFLEX DATA .90
APPENDIX E ANALYSIS OF VARIANCE TABLES FOR PRIMARY
STATISTICAL ANALYSES. . ... 104
BIOGRAPHICAL SKETCH .................. 117
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
ATTENTION, EMOTION, AND THE STARTLE REFLEX
Scott Richard Vrana
Chairman: Peter J. Lang, Ph.D.
Major Department: Clinical and Health Psychology
The startle reflex response is modulated by the
emotional and attention-engaging properties of ongoing envi-
ronmental stimuli. This study investigated the efficacy of
the startle reflex as a measure of attention and emotion
during internally-generated information processing. Thirty-
six undergraduates memorized a neutral and fearful sentence,
then listened with their eyes closed to a series of tones,
one every six seconds. The tone pitch cued subjects to
either relax or to process the neutral or fearful sentence.
Each sentence was processed for two successive six-second
periods. Depending on subgroup assignment (n=12), subjects
received one of three sets of instructions for the first
period: 1) continue to relax; 2) silently articulate the
words of the sentence; or 3) imagine the sentence. All
subjects, regardless of subgroup, imagined the cued sentence
during the second period. Acoustic probes (50 millesecond,
95 dB white noise) were presented at various times during
sentence processing. Magnitude and latency of eyeblink
response to the probes were measured, as well as heart beat
intervals, recorded continuously from six seconds prior to
sentence processing to the end of the six-second period
immediately following processing.
Heart rate increased more during fear processing rela-
tive to neutral. Startle responses were facilitated (larger
magnitude and shorter onset latency) when probes were pre-
sented during fear sentence processing; probe responses were
significantly smaller during neutral sentence processing and
during the intertrial intervals. The largest startle
responses were observed during fear imagery processing.
Fear imagery also prompted greater heart rate acceleration
than the other tasks.
An imaginal analog of sensory-selective attention
was also investigated. Half the neutral sentences referred
to visual imagery and half referred to auditory imagery.
Startle responses to the acoustic probe were of larger
magnitude when subjects were processing acoustically-
oriented sentences compared with processing visually-
oriented sentences. This effect was greatest in subjects
reporting good imagery ability. Furthermore, startle
responses were attenuated when presented during images rated
as very vivid compared with less vivid images, and self-
reported good imagers had attenuated startle responses over-
all relative to poor imagers. In summary, the startle
response was highly sensitive to emotional and attentional
variables during internally-generated cognitive activity,
paralleling effects found when the same variables were
investigated with externally presented materials.
Modulation of tne startle reflex response has been
found as a function of the attentional and affective proper-
ties of environmental stimuli. A recent theory of affective
imagery (Lang, 1979; 1983; 1985) stresses similarity of
response processes in imaginal and environmental contexts.
This dissertation examines the notion that the startle ref-
lex response is modulated by the same factors in imagery as
in environmental perception. Specifically, it will test if
reflex responses to startle probes are augmented when pre-
sented in the context of fearful, aversive imagery; and
furthermore, if such probe responses are relatively
inhibited when the modality of the probe stimulus and the
modality (auditory or visual) of the image content are not
the same. In order to introduce the empirical portion of
this work, factors modulating the startle reflex are
reviewed, and a conception of imaginal processing is
Startle Reflex Response
The startle reflex response is a motor response to
sudden-onset, intense environmental stimulation in the audi-
tory, visual, or tactile modality. The reflex is found in
various species, and in human beings is found at all devel-
opmental levels. Whereas the motor reflex initially
involves the whole body, experimental investigations typi-
cally measure the eyeblink component of the human startle
response. Although the response is obligatory under certain
stimulus conditions, the magnitude, latency, and probability
of the reflex can be modified by a variety of circumstances.
Most important for the current study, the startle response
is modulated by the sensory modality and affective valence
of ongoing environmental stimuli.
Startle and Sensory Modality-Specific Attention
It is now quite well-established that allocation of
modality-specific attention modulates latency and magnitude
of the startle reflex response (Anthony, 1985). Startle
responses are facilitated when attention is directed to the
sensory modality that the reflex-eliciting stimulus occurs
in and is inhibited when attention is directed to a differ-
ent sensory modality. For example, the startle response is
facilitated when attention is drawn to the startle stimulus
by instructing subjects to judge the duration of the stimu-
lus, whereas the response is inhibited when subjects are
instructed to judge the duration of a co-occurring stimulus
in a different sensory modality (Bohlin & Graham, 1977;
Hackley & Graham, 1984; Silverstein, Graham & Bohlin, 1981).
Modality-specific attention modifies the startle response
using startle probes in the visual, auditory, and tactile
modalities (Anthony, 1985).
Modality-specific attention can modify startle response
in the absence of a task directing attention, when a
foreground stimulus is displayed which captures attention in
a particular modality. Anthony and Graham (1983; 1985)
presented either auditory or visual stimuli to subjects;
then elicited a startle reflex in either the acoustic or
visual modality. The visual and acoustic stimuli were
either interesting (a slide of a human face, a melody) or
dull (a blank slide, a pure tone). Startle responses were
larger and occurred more swiftly when the startle stimulus
was presented in the same modality as the foreground stimu-
lation, relative to conditions in which the startle and
foreground stimuli were presented in different modalities.
For example, response to the auditory startle was larger
when subjects were listening to, relative to when they were
viewing, a stimulus. Furthermore, this modality-matching
effect was greater for interesting foreground stimuli:
Response to the auditory startle was largest when subjects
were listening to a melody and smallest when viewing a slide
of a human face. Conversely, response to the visual startle
was largest when viewing a slide of a human face and smal-
lest when listening to a melody. These effects were found
in college students and in four-month-old infants.
Simons and Zelson (1985) investigated response to
acoustic startle probes during high-interest and low-
interest visual events. Slides of nude models (high inter-
est) or of a plain wicker basket (low interest) were shown
for six seconds each, and an acoustic startle-eliciting
probe was presented during slide viewing. Viewing high
interest slides inhibited startle magnitude and latency
relative to low interest slides, suggesting again that
attention directed towards a stimulus can be indexed by the
response to a startle probe.
Startle and Affect
Use of the startle reflex to probe classically-
conditioned fear has produced results which are conceptually
quite different from those described above. Several studies
used rats as subjects (Brown, Kalish & Farber, 1951; Davis &
Astrachan, 1978; Kurtz & Seigel, 1966). In these studies a
light was paired with shock for a number of trials. Subse-
quently, an acoustic startle stimulus presented while the
light was on produced a greater startle reflex relative to
an acoustic startle alone. Appropriate control conditions
ascertained that startle augmentation was a function of the
fear conditioning produced by explicit pairing of the light
and shock. Two studies (Ross, 1961; Spence & Runquist,
1958) found similar results with human subjects. Light was
paired with electric shock. For some subjects onset of the
light signaled shock (fear conditioning). For others, the
shock occurred prior to light onset (backward conditioning).
On test trials, a blink-eliciting airpuff followed light
onset. Blink magnitude was greater for subjects who had
undergone fear conditioning than for subjects who had under-
gone backward conditioning. More recently, Greenwald, Hamm,
Bradley, and Lang (1988) found greater startle eyeblink
magnitude to an acoustic probe presented while subjects were
watching a slide previously paired with electric shock,
relative to a probe presented while subjects viewed a slide
which had not been associated with shock.
Startle facilitation in the context of an aversive
stimulus in these studies occurred despite the fact that the
foreground stimulus (light or slides) and startle probe
(acoustic or tactile) were presented in different modal-
ities. In addition, the signal properties of the condi-
tioned stimulus suggest that attention would most likely be
directed toward the visual modality. Since neither modality
matching nor attentional involvement explain startle facili-
tation under these circumstances, this result can be inter-
preted as indicating that fear facilitates response to the
startle stimulus (Berg & Davis, 1984).
A recent study (Vrana, Spence & Lang, 1988) integrated
these disparate lines of research. The study followed pro-
cedures used to investigate selective attention (Anthony &
Graham, 1985; Simons & Zelson, 1985) in that subjects viewed
slides while startle response to a co-occurring acoustic
probe was assessed. Each slide was viewed for six seconds,
during which time a startle probe was presented. Subjects
viewed twelve slides in each of three categories:
positive/interesting (nudes, food, babies), neutral/dull
(household objects) and negative/interesting (mutilated
bodies and faces, spiders, snakes). Negative slides pro-
duced the largest and shortest-onset startle response, whereas
the startle reflex magnitude was smallest during positive
slides. Further research (Bradley, Cuthbert & Lang, 1988;
Cook, Spence, Gray, & Davis, 1988) have replicated the find-
ing that startle probe amplitude varies significantly with
the emotional valence of foreground visual material.
Startle facilitation in the context of negative visual mate-
rial was interpreted as indicating a matching of response
elements to the aversive slide context and the aversive
startle stimulus: "The startle reaction is construed as an
aversion response, sharing with the response to negative
slides psychological, and perhaps, neurophysiological com-
ponents of avoidance and escape behavior." (Vrana et al.,
1988, page 490)
Imagery and Affect
Lang (1983; 1985) has emphasized the key role of
response processes both in emotion and imagery. His theory
follows other psychological models of memory representation
(Anderson & Bower, 1973) in holding that information is
stored in memory in conceptual units that are interconnected
in associative networks, and that activation spreads through
the network when individual concepts are cued. Lang (1979;
1983) has added to this framework the notion that informa-
tion about one's own responding is part of the network code
in emotion. These response concepts include information
about overt motor and verbal behavior, perceptual adjust-
ments, and autonomic nervous system support for the gross
motor movements. When activated, this information directs
actual context-appropriate responding.
Emotional networks are usually activated by the target
environmental context. However, they can also be accessed
by symbolic stimuli--descriptive text, slides or movies.
When emotional memories are evoked by these media represent-
ations, the overt action may be gated out (subjects report
only feelings of fear or anger). However, the physiological
support for the action (increased heart rate, respiration
increases, skin sweat activity) is still activated, albeit
at a reduced level. The experimental literature supports
this general view. A large number of studies (reviewed by
Cuthbert, Vrana & Bradley, in press) confirms that text-
prompted imagery of an emotional event prompts changes in
report of emotion, and also heart rate, palmar sweat acti-
vity, respiration, and facial expression which mirror the
changes found when people are exposed to the actual emo-
tional context. Memory representations activated in an
actual affective context are also activated during imagery
of the event, and these include codes defining the
Processing Task and Affective Text
Imagery can determine the pattern of response activa-
tion. However, it is not clear that these responses are
obligatory, that is to say, the same text describing an
emotional event could perhaps be dealt with in different
ways. Some modes of processing might not engage associated
response concepts, i.e., if subjects were asked to do a
grammatical analysis or simply count the number of words
presented. This section reviews work on the physiological
correlates of instructional strategies for processing affec-
tive text. The studies use variants of an experimental
procedure, originated by Schwartz (1971), in which subjects
memorize text in advance, then retrieve it from memory cued
by a series of tones occurring at regular intervals. May
(1977a) applied this paradigm to compare imagery and
"thinking" fear material. Snake phobics memorized short
sentences describing touching a snake or reading a magazine.
Subjects were instructed to "think" the sentence and then
imagine the sentence in successive tone-cued ten-second
periods. Heart rate and respiration amplitude were greatest
when subjects imagined the snake sentence, compared to
"thinking" the snake sentence or thinking or imagining the
relaxing sentence. A follow-up study (May, 1977b) found
that imagery of a fearful sentence produced greater
physiological response than thinking the sentence, hearing
the sentence, or seeing a slide depicting the same
situation. Jones and Johnson (1978; 1980) used the same
procedure with unselected subjects and found greater heart
rate, muscle tension, and respiratory activity during im-
agery of "high activity" sentences (short sentences with
information about behavioral response) than imagery of "low
activity" relaxing sentences. Physiological differences
between high and low activity sentences were not apparent
when subjects were "thinking" the sentences.
Vrana, Cuthbert and Lang (1986) refined the methodology
of the above studies to support a clearer interpretation of
the imagery effect. They argued that the subject's task was
unclear when told to "think" a sentence (as opposed to
"image" it). They proposed instead to begin by comparing
imagery with the straightforward instruction to silently
articulate the words of the sentence. Half of the subjects
articulated the material for 10 seconds and imagined the
same sentence in the subsequent 10 second period. The
remaining subjects imagined the sentence first and silently
articulated it in the following period, thus controlling the
order of processing modes. Regardless of the order of proc-
essing, imagery of fearful material resulted in greater heart
rate increase than did silent repetition of fear material.
In these studies, thinking or silent rehearsal of
arousing material resulted in slightly, but not signifi-
cantly, greater physiological response than thinking or
rehearsal of neutral material, suggesting that response code
is contacted to some extent when text containing response
concepts is rehearsed in working memory. In these studies
subjects were cued as to which material to process prior to
the neutral baseline phase of the trial, allowing memory
retrieval of the material prior to the specified cue and
possibly contaminating baseline measurement. In fact, heart
rate during baseline periods differed depending on whether
the material to be processed was arousing or relaxing.
(Vrana et al., 1986). A new paradigm was developed to
control and examine the possibility of pre-instructed memory
retrieval. An initial experiment using this paradigm is
described in some detail here, as the same paradigm was
employed in the empirical portion of this dissertation.
In this study (Vrana, Cuthbert & Lang, in press) sub-
jects memorized a neutral and fearful sentence, and then
heard a series of tones, one every six seconds. The sub-
ject's task was to repeat the word "one" silently and relax
at this tone. A higher or lower frequency tone cued memory
retrieval of either the neutral or fearful sentence. Sen-
tence processing occurred for two consecutive six-second
periods following retrieval. All subjects imagined the
material during the second period, but differed as to ini-
tial cognitive task. Processing instruction for the first
period was a between-subjects variable. One group imagined
the sentence immediately upon hearing the memory retrieval
signal, another group silently articulated the words of the
sentence upon memory retrieval, and a third group continued
to repeat "one" and relax at the first signal tone (called
"null task"). After the second period all subjects returned
to relaxing and repeating the word "one" until the next
higher or lower tone. Figure 1 diagrams the structure of a
single trial for each group.1
1 Further details about the procedure and data reduction can
be found in the Method section, and differences between the
two studies are described in Appendix B.
Group 2 "one" "one" ARTICUL. IMAGE "one" "one"
ba n p peiods
base processing periods
period t 1 t 2
sentence cue tones
data collection interval 18 sec.
- -_J -- >
Figure 1. Diagram of events that occurred during a single
trial for each experimental group. Pulse trains of 500
msec tones (one every six seconds) were presented to
all subjects. On hearing the initial non-signal tone
the subject said the number "one" silently. The pulses
with angled tops represent higher- or lower-pitched
tones that cued neutral or fear sentence processing.
Heart rate data was collected during the time period
represented in bold face. Processing instructions for
successive periods in the trial are written within the
Fearful sentences were rated as less pleasant, more
arousing, and involving less dominance than neutral sen-
tences. Table 1 lists mean heart rate change for Periods
one and two. Period one (Null task, Articulation, or
Imagery) data are considered first. Overall, processing
fearful sentences resulted in greater heart rate increase
than did processing neutral sentences (F(1,27)=15.00,
<.0006). Imagery resulted in greater heart rate increase
overall than did the null or articulation task
(F(2,27)=16.4, p<.02). There was a tendency for text proc-
essing instruction to determine the degree to which fear
sentences occasioned heart rate increase (Task X Content
F(2,27)=2.79, p<.08): Imagery of fear sentences resulted in
greater heart rate increase than either silent articulation
or null processing of the fear material (Task effect for
fear material F(2,27)=3.73, p<.04). During the second
period, when all subjects were imagining the material, heart
rate was equivalently greater for fear relative to neutral
imagery, regardless of group membership (F(2,27)=19.33,
<.0002). Results therefore show that this paradigm is
sensitive to the affective properties of processed sentences
during imagery and other processing tasks.
Statement of the Experimental Problem
The present experiment uses the same sentence imagery
paradigm (Vrana et al., in press) just described. Subjects
will learn fearful and neutral sentences which are subse-
quently retrieved according to the frequency of the cue
Mean heart rate change and standard deviation (in paren-
theses) over six seconds for each group during processing of
neutral and fearful sentences. T-tests are for neutral-fear
comparisons within groups.
Note: The data are presented separately for Period one (null
task, articulation, imagery) and Period two (imagery for all
groups). T-values have 9 degrees of freedom.
tones. Different groups will process the material diffe-
rently in an initial period (null task, articulation, im-
agery); in a second period all subjects will do the imagery
task. In addition, this research for the first time employs
an acoustic startle probe methodology to measure mental
imagery. Probe stimuli will be presented during all proc-
essing tasks. Three primary questions will be addressed.
1. Is the Startle Reflex Augmented During the Processing of
a Fear Image?
Heart rate will be recorded, and will be considered the
criterion for response element activation in memory. It is
expected that heart rate results from the previous study
will be replicated: Greater heart rate increase will be
evident during fear relative to neutral processing, with
this difference being greater during imagery than silent
articulation or null processing.
A number of studies have found that viewing fear-
eliciting material (slides depicting negatively-valent stim-
uli, a light signaling electric shock) facilitated the
startle response to an acoustic or tactile probe. This
effect was interpreted by Vrana et al. (1988) as a function
of response matching: The aversion response elicited by the
startle probe matched in affective valence the response
elements accessed from emotional memory by perceptual proc-
essing of the unpleasant slides. It is predicted that proc-
essing fear material in imagery will prompt a similar match
and facilitate the reflex (larger magnitude and shorter
latency), relative to the response during neutral processing
or no-task control periods.
The response matching hypothesis implies that startle
facilitation, like heart rate, will be greater to the extent
that a response set is activated during cognitive proc-
essing. Therefore, greater startle facilitation during fear
relative to neutral material should occur during imagery
than during silent articulation or under instructions to
refrain from processing the sentences.
2. Are Images Modality Specific? Is the Startle Reflex
Augmented When the Sensory Content of Imagery Matches the
Modality of the Probe?
Imagery is hypothesized to activate the same response
disposition as that activated by the represented environmen-
tal situation. Previous work has shown augmentation of
startle probe responses when probe stimuli are in the same
sensory modality as the stimulus foreground to which sub-
jects are attending. Isomorphism between imagery and per-
ception implies that a parallel augmentation should occur if
the startle probe modality matches the modality of the
imaged stimulus material. This is tested in the current
experiment. Half of the neutral material is designed to
prompt primarily visual memory processing; the remaining
neutral sentences prompt processing in the auditory moda-
lity. All startle stimuli are presented auditorally. If
cognitive processing leads to similar modality specific
response dispositions as perceiving environmental events,
then augmented startle responses will be elicited during
auditory-oriented sentence processing, relative to visually-
3. Is Image Vividness Related to Startle Modulation?
Vivid imagery is associated with a disposition to
become absorbed in experience (Sheehan, McConkey & Law,
1978). In information processing terms, when cognitive
capacity is committed to an imaginal production, less capa-
city is available for processing other, external stimuli.
It may be deduced from this assumption that (a) more vivid
imagery will be associated with a reduced response to con-
current environmental input, i.e., less capacity is avail-
able to process the startle probe. A corollary hypothesis
is (b) that individuals who profess to be generally good
imagers (as defined by questionnaire, Sheehan, 1967) are
likely to show reduced responses to the startle probe rela-
tive to poor imagers.
Nested within this overall argument are two deductions
concerning the effects of match/mismatch between the sensory
modality of the startle probe and the dominant modality of
the image: (c) if imagery and the sensory intake differ-
entially activate the same perceptual processing sub-systems
(see number 2 above), then facilitation of acoustic probe
responses during auditory imagery (and relative inhibition
with visual imagery) should be greater when imagery is more
vivid; furthermore, (d) this modality-specific effect should
be larger for good than for poor imagers.
Subjects were 36 normal volunteers recruited from the
pool of students attending an introductory psychology course
at the University of Florida. The sample was randomly
divided into three subgroups of six males and six females
for this experiment.
Subjects sat in a comfortable reclining chair in a dimly
lit room adjacent to the equipment room. The timing of
events and collection of data were accomplished under the
control of a PDP-11/23 computer. All auditory stimuli were
presented to the subject binaurally through Pioneer SE-205
stereo headphones. Tones were generated using a Coulbourn
Voltage Controlled Oscillator with a Selectable Envelope
Rise/Fall Gate set for 80 msec rise/fall time. The low,
medium, and high tones were 800, 1100, and 1500 Hz, and were
measured monaurally at 71, 72.5, and 73.0 dB (A), respec-
tively. The acoustic startle-producing stimulus was a 50
msec burst of white noise (20-20,000 Hz) with a monaural
intensity of 95 dB (A) and instantaneous rise time. All
sound pressure level measurements were made using a Bruel
and Kjaer Type 2203 Precision Sound Level meter with a half-
inch Type 4134 condenser microphone and a Type 4153 artifi-
Lead I EKG was obtained using Beckman standard Ag-AgCl
electrodes filled with Beckman electrode electrolyte and
placed on each inner forearm. The signal was filtered
through a Coulbourn Instruments Hi Gain Bioamplifier/
Coupler, and a Schmitt trigger interrupted the computer each
time it detected a cardiac R-wave.
The startle response was measured as electromyographic
activity at the right obicularis oculi using Med Associates
miniature electrodes filled with Beckman electrode electro-
lyte. The signal was amplified by a Coulbourn S75 series
bioamplifier with high- and low-pass filters set at 100 and
1000 Hz, respectively, then fed through a Coulbourn S76-01
Contour Following Integrator set for a measured time con-
stant of 200 msec. The integrated signal was sampled at
1000 Hz once every msec for 250 msec after the onset of the
startle stimulus. The amplification for the muscle tension
measure was set at 60,000 for each subject at the beginning
of the session, and two calibrating startle stimuli were
presented. If the response to these startles exceeded the
limits of the analog-to-digital (A-D) converter, the
amplification was reduced to 50,000; if the startle response
was small relative to the limits of the A-D converter, then
amplification was increased to 70,000. The amplification
was reduced for four subjects and increased for 12 subjects.
The experimental groups did not differ in average
Stimulus materials were six sentences describing moder-
ately positive, relaxing situations and six describing
common fearful, arousing situations. Each fear sentence
contained at least one reference to autonomic (e.g., "My
heart pounds") or behavioral ("I grip the chair")
responding. Three of the six neutral sentences referred
explicitly to stimuli in the auditory modality and three
contained explicit references to stimuli in the visual
modality. All twelve sentences were presented to the sub-
ject printed on index cards with key phrases in capital
letters; subjects had to repeat only these key phrases
verbatim in order to meet sentence memorization criterion.
The twelve sentences are presented in Appendix A. For each
subject the sentences were randomly grouped into six
After arriving at the laboratory subjects read and
signed an informed consent and filled out Sheehan's (1967)
short form of the Questionnaire Upon Mental Imagery (QMI;
Betts, 1909). Electrodes were then attached and the
instructions read. These instructions read in part: "In a
little while you will memorize two sentences, one fearful
and one neutral in content. You will use these sentences to
create images in your mind using the following procedure.
After you memorize the two sentences and I leave the room,
you will hear a series of short tones, one every six
seconds. Each tone will be at one of three different fre-
quencies. The tone that you will hear most often is the
middle tone. I'll call this the 'normal' tone. Whenever
you hear this tone, just relax and think the word 'one' to
yourself each time you breathe out. This is to help clear
your mind and help you remain relaxed." Subjects were then
told that tones which were higher or lower in pitch compared
to the "normal" tone would be presented every so often, and
that when they were presented they would occur twice in
succession at the usual six-second interval, cuing subjects
to retrieve one of the sentences from memory and process it
for two six-second periods in a row. The pitch of the tone
signaled subjects to retrieve either the fear or the neutral
Instructions regarding sentence processing was a
between-groups variable. One group was told to imagine the
sentence specified immediately upon hearing the memory
retrieval tone, and then to imagine it again during the
second period (Image-Image group). Another group was
instructed to silently repeat the words of the sentence upon
memory retrieval, then imagine it at the second tone
(Articulate-Image group). A third group was told to con-
tinue to think "one" and relax at the first signal tone, and
then to imagine the sentence specified at the second signal
tone (Null task-Image group). Subject instructions are
presented in Appendix A. Thus, all subjects imagined the
material during the second period, but differed as to how
they processed the material immediately upon memory
retrieval. The second period ended after six seconds with
another "normal" tone, at which point all subjects returned
to relaxing and repeating the word 'one' until the next
signal tone. The interval of relaxation between processing
sentences varied randomly in six-second increments from 18
to 30 seconds. The tones occurred such that the subject
processed the neutral and the fearful sentence six times
each, with neutral and fearful material processed in a
pseudo-random order. A schematic of a single trial for each
group is presented in Figure 1.
After completing six neutral and six fear processing
trials, the subject heard two tones one-half second apart, a
signal to open his or her eyes and rate the images of each
sentence along the dimensions of affective valence, arousal,
dominance (Osgood, Suci & Tannenbaum, 1957), and vividness.
Each of these ratings was accomplished by marking a vertical
line through a horizontal line with words anchoring each
end. After making these ratings, subjects memorized another
neutral and fearful sentence, and processed these six times
each in the same manner as the earlier block of trials.
Each subject processed all six neutral and fearful sentences
six times each, for a total of 36 trials of each type.
After the experimental session, subjects rated each fear
sentence on a 1-7 scale for "how frightened you would be if
you were actually in this situation."
Subjects were instructed regarding the startle stimuli
as follows: "At times you will hear loud clicks, like a
finger snapping. These are meant to elicit a response we
wish to measure, but you just need to ignore them and
continue with the task".
Within each block of six trials, three startle probes
were presented during Period one (articulation, null task,
or imagery) and three were presented during Period two
(imagery for all subjects) for neutral and fearful trials,
totaling twelve startle probes during each block of sentence
processing. In addition, three startle probes were pre-
sented during unsignaled "Count 'one'" periods (intertrial
intervals). The three startle probes in each processing
period X affective content cell and in the intertrial
intervals were distributed as follows: One occurred one
second after tone onset (Early), one three seconds after
tone onset (Middle), and one 5.5 seconds after tone onset
The six startle probes presented during sentence proc-
essing for each affective content were distributed in the
following way within a block of trials: One trial contained
a startle in the Early position of Period one and in one of
the three positions of Period two. Two trials contained one
startle during Period one only (one at the Middle and one at
the Late position), and two trials contained one startle
during Period two only (In the two positions not covered by
the only trial containing two startle probes). The
remaining trial contained no startle probe. This arrange-
ment was designed to maximize subject uncertainty regarding
occurrence of startle probes while providing an adequate
number of data points in each condition.
Within each of the three groups, half the subjects proc-
essed the fear material at the high-pitched tone and the
neutral material at the low-pitched tone. This was reversed
for the remainder of the subjects. Each subject partici-
pated in 36 trials within each of the two stimulus contents
(fearful and neutral): six unique sentences with six trials
using each sentence.
Data reduction and analysis
Interbeat intervals were recorded and converted off-line
to heart rate in beats per minute for each half-second from
six seconds before a signal tone to six seconds after sen-
tence processing had ended. Mean heart rate was calculated
for the "Count 'one'" period before the tone signaling
sentence processing (baseline), the first period (null task,
silent articulation of the text, or imagery), and the second
period (imagery). Heart rate for the six second "Count
"one'" period before the signal tone was used as baseline to
create heart rate change scores for the first period and the
second period. Data from all trials of the same content
(fear or neutral) were averaged together. The imagery
ratings of valence, arousal, dominance and vividness were
recorded on a 15 centimeter horizontal line. This line was
divided into 30 half-centimeter sections for scoring pur-
poses and each rating was assigned a number from 0 to 29
based on the section of the line marked by the subject.
Each heart rate measure and each rating was subject to a
univariate Group X Content analysis of variance (ANOVA)
using the BMDP2V AVOVA program, with Content being a
repeated measure within subjects. Heart rate was also sub-
ject to an initial Group X Period (one, two) x Content
ANOVA, with Period and Content involving repeated measures.
The reflex eyeblink data were reduced off-line by a
computer program (Balaban, Losito, Simons, & Graham, 1986)
which eliminated trials with an unstable baseline and scored
each trial of reflex eyeblink data for latency to blink
onset (in milliseconds) and peak amplitude (in arbitrary A-D
units). Trials with an unstable baseline constituted 8.5%
of all trials and were treated as missing data. For trials
in which a startle response was not detected, amplitude was
scored as zero and latency was considered as missing data.
Trials in which no scorable blink occurred and latency was
treated as missing data comprised 8.4% of all remaining
For each block of trials, there was one startle probe
data point in each cell of the Content (Neutral, Fear) X
Period (One, Two) X Startle Probe Time (Early, Middle, or
Late) matrix and one data point at each startle probe time
during the intertrial interval period. The available data
points from each of the six blocks were averaged. An
initial Group (Null task-Image, Articulate-Image, Image-
Image) X Content (Neutral, Fear) X Period (one, two) X
Startle Probe Time (Early, Middle, Late) ANOVA was conducted
on the averaged latency and magnitude data. Group was a
between-subjects factor while Period, Content and Startle
Probe Time were within-subjects factors. The latency and
magnitude data from each period were then subject to sepa-
rate Group X Content (Neutral, Fear, intertrial interval) X
Startle Probe Time ANOVAs. Note that the same intertrial
data were used in the ANOVA for Period one and Period two.
Other analyses are described in the relevant sections of
Results. Tables detailing ANOVA results for all major
analyses are located in Appendix E. For all analyses,
Greenhouse-Geisser (Greenhouse & Geisser, 1959) corrected
degrees of freedom are reported to correct for unequal
correlation among repeated measures. Follow-up t-tests were
conducted for significant ANOVA results.
The present experiment repeats the method used by Vrana
et al. (in press). An assessment of the replication's
success is presented first, based on results for heart rate
and self-report, the measures common to this and the earlier
study. The remainder of the results section will address
the questions about the startle response posed in the state-
ment of the problem. Startle modulation during neutral and
fear processing will be examined first. Image processing of
modality-specific sensory information will be described
next. Finally, the relationship between image vividness and
the startle response will be assessed by analyzing the
startle response as a function of each subject's rated
vividness of the image. Individual differences in imagery
ability will also be examined in these latter two analyses.
For these analyses, only a subset of the subject sample will
be used; namely, those receiving extreme scores on the
Questionnaire Upon Mental Imagery2
2 For all imagery ability analyses, good imagers were
defined as scoring below 75 on the Questionnaire Upon Mental
Imagery; poor imagers were defined as scoring above 95 on
this questionnaire. Subjects scoring between 75 and 95 were
omitted from this analysis. These cutoff scores resulted in
Subject ratings of image valence, arousal, dominance,
and vividness, and fear of being in the situation depicted
in the fear sentences are presented in Table 2. Subjects
felt less happy (F(1,33)= 471.3), more aroused
(F(1,33)=381.5) and less dominant (F(1,33)= 167.9) during
fear than during neutral imagery (all p<.0001). The
Articulate-Image group rated themselves as feeling more
dominant during imagery than did the other two groups
(F(2,33)=3.58, 2<.05). This occurred for both neutral and
fear processing and may have been because silent articula-
tion was the most straightforward task required of subjects
in this study. There were no other significant effects for
valence, arousal, and dominance, and no effects found for
image vividness or fear rating.
A Group X Content ANOVA was conducted to assess heart
rate during the six seconds prior to the signal tone, used
as a baseline measure for the subsequent change scores. No
significant differences in baseline heart rate emerged as a
function of subject group assignment or the content of the
upcoming processing trial.
four subjects per group in each Group X Imagery Ability
cell, with the following exceptions: three subjects were
represented in the Articulate-Image/Good imager cell and
the Image-Image/Poor imager cell, and there were five sub-
jects in the Null Task-Image/Poor imager cell.
Self-reported ratings of image valence, dominance, arousal, and
vividness as a function of processing group and content of
sentence. Standard deviations are in parentheses).
Note: Ratings of fear in the actual context ("fear ratings")
were recorded for the fear sentences only. Standard deviations
are in parentheses. All ratings are on a 0-29 point scale except
for the fear rating, which is 1-7.
Mean heart rate change for Period one (null task,
articulation, imagery) and Period two (imagery) by Group and
Content (neutral, fear) can be seen in Table 3. An initial
analysis including both processing periods was conducted.
Overall, heart rate increase from baseline was greatest on
trials involving retrieval of the fear sentence relative to
neutral trials (F(l,33)=29.9, p<.0001). Heart rate evi-
denced no significant change from Period one to Period two
on neutral trials (t(33)=1.14, 2>.10), while increasing from
Period one to Period two on fear trials (t(33)=2.40, p<.05,
overall Content X Period F(1,33)=18.5, 2<.0001). No other
effects were significant.
Period one (Null task, Articulation, Imagery)
Just as in a previous study (Vrana et al., in press),
processing fearful text resulted in more pronounced heart
rate increase than did processing neutral text (F(1,33)=
15.22, 2<.0004). As can be seen in Table 3, all three proc-
essing modes again elicited significant neutral-fear diffe-
rentiation. Imagery again produced the largest mean heart
rate overall--higher than null processing or articulation of
the fear sentence. However, the overall difference between
processing modes (Group F(2,33)=1.18, 2>.30) and the inter-
action between modes of processing and fearfulness of the
materials (F(2,33)=1.36, p>.25) did not achieve an accep-
table confidence level.
Mean heart rate change over six seconds for each group
during processing of neutral and fearful sentences,
Presented separately for Period one (null task,
articulation, imagery) and Period two (imagery for all
groups). Standard deviations are in parentheses.
FEAR t, 2<
11 degrees of freedom.
Note: T-values have
Period two (Imagery)
During this period all subjects imagined the textual
material. As can be seen in Table 3 (right panel), fear
imagery resulted in greater heart rate increase than did
neutral imagery (F(1,33)=36.0, p<.0001). No other effects
approached significance. Imagery effects found in the ear-
lier study (Vrana et al., in press) were thus replicated in
Combined Study analysis
The current experiment and the earlier, non-startle
study (Vrana et al., in press) separately show strong
differences in heart rate following memory retrieval of fear
and neutral sentences. However, while each study found that
imagery tended to increase the neutral-fear difference in
heart rate relative to the other processing instructions,
neither study individually found conclusive statistical
evidence for this apparent pattern. Two questions were
raised. First, is this pattern of results a chance finding,
or is the lack of statistical support a power problem, e.g.,
too few subjects in each individual study? Second, did the
addition of the startle probe produce reliable differences
in heart rate response, compared to the no-startle
situation? To address these questions, data were combined
across experiments, and mean heart rate change for each
Period was subject to a Study X Group X Content ANOVA.
Figure 2 shows heart rate for the combined sample
during the first processing period and the immediately
preceding "Count 'one'" period, presented on a half-second
basis for each processing instruction for neutral and fear.
Table 4 presents mean heart rate change for the combined
sample in the same manner as each sample was presented sepa-
rately. During Period one, heart rate increase was greater
overall following the tone cueing fear material relative to
the tone cueing neutral material (F(1,60)=30.81, 2<.0001).
All three groups exhibited significantly greater heart rate
increase during fear relative to neutral trials (see t-test
comparisons, Table 4). Imagery resulted in greater heart
rate increase than the other two tasks (F(2,60)=5.44,
2<.007). When the two studies are combined, imagery of fear
sentences prompted greater heart rate increase than did the
null task or silent articulation (Content X Group
F(2,60)=4.16, p<.03). During Period two, fear imagery
resulted in greater heart rate increase than neutral imagery
(F(1,60)=50.22, R<.0001) and no other effects were signifi-
cant. Subjects in the two studies did not differ in heart
rate response, either alone or in interaction with other
variables, for Period one or two (all Fs <1.50).
In summary, the two studies combined produced robust
results indicating greater heart rate increase while proc-
essing fear relative to neutral sentences. Furthermore,
they provided clear statistical evidence that heart rate
increase was greater during fear imagery than silent
articulation or null processing of the fear sentences. In
both studies, all groups rated themselves as feeling less
Mean heart rate change from baseline over six seconds for
each group during processing of neutral and fearful sen-
tences for both studies combined, presented separately for
Period one (null task, articulation, imagery) and Period two
(imagery for all groups). Standard deviations are in
FEAR t, p<
Note: T-values have 21
degrees of freedom.
Figure 2. Continuous heart rate waveform in half-second
intervals for each group for the "Count 'one'" period
and the first sentence processing period for the com-
bined sample (startle and no-startle studies). The
tone cueing retrieval of the neutral or fearful sen-
tence is signified by a vertical line at the six second
mark in each graph.
70 ----- FEAR
pleasant, more aroused, and less dominant during fear rela-
tive to neutral imagery. Because startle response facilita-
tion, like heart rate increase, is assumed here to indicate
more extensive fear processing, the heart rate results pro-
vide an empirical basis for the previous (pp. 14-15) predic-
tions regarding the effect of sentence processing on the
Startle Reflex: Content Differences
Magnitude. Table 5 shows that startles elicited during
fear trials were consistently larger in magnitude relative
to those elicited during neutral trials (overall
F(1,33)=25.5, p<.0001). Table 6 illustrates how the sen-
tence content effect was modulated by the processing task.
The neutral-fear difference was larger in Period two (when
all subjects were imagining the material) than in Period
one, when subjects were performing different processing
tasks (Content X Period F(1,33)=5.99, p<.02). Still,
responses during fear trials were reliably larger than
responses during neutral trials in both Period one and
Period two (see later sections on individual Period
The startle response also differed with the timing of
the probe (Table 5, Probe Time F(1.7,27.5)=5.42, <.02). It
was smaller in magnitude at the Early relative to the Middle
(t(66)=2.28, p<.05) position and was marginally inhibited
Startle reflex magnitude and latency for neutral, fear, and
intertrial interval startles by startle probe times during
periods one and two. Standard deviations are in paren-
theses. These data are presented separately for each exper-
imental group in Tables 7 through 12 of Appendix C.
Startle reflex magnitude and latency for each group separately
for neutral, fear, and intertrial interval startle probes.
Standard deviations are in parentheses.
relative to the Late (t(66)=1.47, .10
Middle and Late did not differ (t(66)=0.80).
Latency. Latency measures produced results generally
consistent with magnitude (see also Tables 5 and 6): Over-
all, startle response was facilitated (shorter onset latency)
during fear relative to neutral trials (F(1,33)=8.49,
p<.007); reflex latency also tended to vary with the timing
of the startle probe (overall F(1.8,29.9)=3.05, p<.06),
reflecting a tendency toward inhibition at the Early rela-
tive to Middle (t(66)=1.55, .10
startles did not differ (t(66)=0.07). Finally, startle
latency was shorter during Period two relative to Period one
Statistical differences for processing groups did not
appear in these omnibus analyses. However, a more focused
test was planned for the first sentence recall period, where
the actual processing task manipulation occurred. For compa-
rison, a parallel analysis was done on the second recall
period, where instructions to groups did not differ (Group X
Content (Neutral, Fear, Intertrial interval) X Time).
Period one (Null Task, Articulation, Imagery)
Magnitude. Consistent with the omnibus analyses,
startles elicited while processing fear material in Period
one were generally of larger magnitude than startles pre-
sented during neutral processing (t(66)=2.37, p<.03, overall
Content F(1.95,64.4)=14.83, p<.0001). Fear sentence
startles were also greater in magnitude than probe reaction
during intertrial intervals (t(66)=4.43, p<.0001), as were
responses elicited during neutral processing (t(66)=2.06,
As is evident in Figure 3, we can also infer differ-
ences in the way groups processed the neutral and fearful
material, i.e., the different sentence recall tasks appear
to have differentially modulated the startle response
(Content X Group, F(3.9,64.4)=2.95, p<.03). Thus, a sepa-
rate analysis of the imagery group confirmed a significant
sentence content effect (F(1.5,16.8)=10.13, p<.003):
Startles elicited during fear imagery were larger than those
elicited during neutral imagery (t(22)=2.44, p<.05) or
intertrial intervals (t(22)=3.64, p<.005), while the latter
probes did not differ from each other (t(22)=1.20, 2>.20).
Startles elicited during silent rehearsal (Articulate-
Image group) also showed a main effect for Content
(F(1.8,20.2)=6.25, p<.01). Probe responses during sentence
processing were larger than startles elicited during inter-
trial periods (neutral t(22)=1.95, .05
t(22)=2.80, 2<.02). However, probe reflexes during fear
material were not significantly larger than neutral sentence
probes (t(22)=0.85, p>.40).
Instructions not to process the sentences (Null Task-
Image group) appeared to further reduce the sentence recall
effect and the overall analysis was not significant (F<1.0).
Nevertheless, an individual t-test suggested that startles
Figure 3. Magnitude of response to the startle probes for
each content (neutral, fear) presented separately for
each group. The bars in the left-hand columns depict
data from Period one (null task, articulation,
imagery), while the bars in the right-hand column
depict data from Period two (imagery). The horizontal
line across each graph represents magnitude of inter-
trial interval startle responses for each group.
NULL TASK IMAGE
elicited during fear trials might yet be greater than those
elicited during neutral trials (t(ll)=3.07, p<.02).
Probe timing and processing task. Startle magnitude
was affected differently at different times in the six-
second period, depending on whether subjects were processing
sentences or in an intertrial period (Content X Time, F(3.6,
118.2)=4.54, p<.003). To explore possible probe time dif-
ferences for the specific Period one processing tasks, a
Content X Group ANOVA was undertaken separately for the
Early, Middle, and Late startle positions. The mean values
tested are presented in Table 5. No significant differences
emerged at the Early startle position (Fs <1.5). In con-
trast, the nature of the Content differences at the Middle
(F(1.9, 63.4)=10.12, p<.0002) and Late (F(1.9, 61.2)=11.57,
p<.0001) positions is that described earlier for the overall
data. Thus, the affective content appeared to have its
greatest impact on startle modulation at the Middle and Late
probe positions; magnitude at the Early position seemed to
be controlled by the temporal relationship between the
startle probe and preceding tone (see Graham, 1975).
Latency. Latency to startle responses elicited after
retrieving fear material from memory was shorter than
latency to startles presented during intertrial intervals
(overall F(1.96, 64.5)=7.27, 2<.002; t(66)=3.08, 2<.005);
and marginally shorter than neutral processing (t(66)=1.70,
fer less from each other (t(66)=1.38, p>.20). The overall
Group X Content effect was not significant. However, given
the hypothesized affective content differences as a function
of mode of processing, individual group analyses (Content X
Startle Probe Time) were conducted for latency.
Startles varied significantly with content for the
Image-Image group (overall F(1.9,20.4)=6.61, p<.007).
Reflexes elicited during intertrial periods were inhibited
(longer in onset latency) relative to startles elicited
during fear imagery (t(22)=2.93, p<.01). Intertrial
startles were less clearly inhibited relative to neutral
imagery (t(22)=1.84, .05<2<.10). Mean startle latency
during fear imagery was shorter than during neutral imagery,
but this result was not significant (t(22)=1.08, p>.20).
Finally, there were no content differences in startle onset
latency either during the null task (F(l.8,19.8)=1.55,
p<.25) or during articulation of sentences
Period Two (Imagery)
Magnitude. As illustrated in Figure 3, startles eli-
cited during fear imagery were larger in magnitude than
startles elicited during neutral (t(66)=4.19, p<.0001, over-
all F(1.8,59.0)=18.90, p<.0001) or intertrial periods
(t(66)=4.52, 2<.0001), while neutral and intertrial interval
startles did not differ in magnitude of response (t=0.32).
As expected, when all subjects were engaged in the same
imagery task, there was no overall group difference.
The relationship between sentence processing and
intertrial periods again differed as a function of the time
of the startle probe within the interval (Content X Time
F(3.2,105.5)=4.30, p<.006). During intertrial intervals,
startle responses were larger at the Early relative to the
Middle (t(132)=3.16, p<.002) or Late (t(132)=3.89, p<.0002)
positions. Conversely, for fear imagery, startle responses
were smaller at the Early relative to Middle probe position
(t(132)=3.79, p<.0005) and the Late probe position
(t(132)=2.74, 2<.01). Like fear imagery, startle responses
during neutral imagery were smaller at Early relative to the
Middle Startle Probe Time (t(132)=2.63, p<.01). Late probe
startles during neutral imagery were also smaller than
Middle probe startles (t(132)=2.11, R<.05). Again, there
were no group effects in Period two, when all subjects
imaged the sentence.
Latency. Figure 4 shows that mean onset latency was
shorter to startles elicited during fear imagery than those
elicited during neutral imagery (t(66)=1.70, .05
which in turn exhibited shorter onset than those elicited
during intertrial periods (t(66)=2.39, p<.02, overall
F(1.96,64.7)=12.80, p<.0001). As for magnitude, the experi-
mental groups did not differ in latency of response in
Figure 4. Latency of response to the startle probes for
each content (neutral, fear) presented separately for
each group. The bars in the left-hand columns depict
data from Period one (null task, articulation,
imagery), while the bars in the right-hand column
depict data from Period two (imagery). In this Figure
smaller bars represent facilitated startles, i.e.,
those with a faster onset latency. The horizontal line
across each graph represents latency of intertrial
interval startle responses for each group.
PERIOD I PERIOD 2
NULL TASK IMAGE
Startle Reflex: Modality-Specific Attention
Of the six neutral sentences, three referred explicitly
to the auditory modality and three referred to the visual
modality. It was hypothesized that modality-specific proc-
essing would have similar effects as attending to
environmental stimuli in a specific modality. Data from
neutral sentence processing trials were subject to a Sensory
Modality (Auditory, Visual) X Processing Period (One, Two) X
Time X Group X Imagery Ability (Good, Poor) ANOVA (Tables 13
through 20 in Appendix D present these data).
Magnitude of response to the acoustic probes was larger
while subjects were processing auditory sentences (mean=248
A-D units) relative to visual sentences (mean=222 A-D units;
Sensory modality F(1,17)=4.67, p<.05). As Figure 5 illus-
trates, this modality-specific modulation was strongest at
the middle probe positions, and was less clear or reversed
when the acoustic probe immediately followed a tone cue
(Early) or preceded the cue to cease sentence processing
(Late in Period two; Sensory Modality X Period X Time,
Latency of response was not significantly facilitated
during auditory relative to visual sentences (F(1,15)=0.90,
p>.30) for all subjects. However, this effect was found
specifically in questionnaire-defined good imagers and was
most pronounced at the Middle startle position (Sensory
EARLY MID LATE EARLY MID
Figure 5. Startle response magnitude during processing of
neutral sentences which refer to the visual or auditory
modality, presented as a function of startle onset time
during Period one and Period two.
modality X Imagery Ability X Time, F(1.8,26.6)=3.58, p<.05).
This effect can be seen in Figure 6. It should also be
noted from this Figure that poor imagers produced startles
with generally shorter onset latency than did good imagers
(35.5 msec vs. 39.3 msec, Imagery Ability, F(1,15)= 5.08,
p<.04). No differences between sentence processing groups
were found for latency or magnitude of startle.
ANOVAs were then conducted individually for Period one
and Period two. Trends in the data were the same in each
period; however, because of the small number of datapoints
per cell statistically significant effects were not found
for either period when analyzed separately.
Mean heart rate was slower during acoustically-oriented
sentences relative to visually-oriented sentences (-0.18 vs.
0.25) beats/minute over Periods one and two). To determine
if this trend was statistically significant, a Group X
Imagery Ability X Sensory Modality X Period ANOVA was con-
ducted for heart rate change, to parallel the analysis of
startle data. None of the resulting F values were signifi-
cant (Modality F(1,17)=1.39, p>.25).
Startle Reflex: Image Vividness
It was hypothesized that overall, more vivid images
would result in attenuated response to the startle probe
relative to less vivid images. In order to test this hypo-
thesis, the six neutral and six fear sentences were sepa-
rately ranked for each subject from least to most vivid
Figure 6. Startle response latency during processing of
neutral sentences which refer to the visual or auditory
modality, presented as a function of startle onset time
for good and poor imagers. Smaller bars indicate faci-
litated startles relative to larger bars.
based on ratings of image vividness for each specific sen-
tence. A Group X Imagery Ability X Content (Neutral, Fear)
X Vividness (six levels nested within each content) ANOVA
was conducted for magnitude and latency of startle probe.
Because ratings were made based on image vividness, only
data from Period two were considered. Only results
revealing a significant linear trend over the six levels of
vividness are reported here.
Figure 7 shows that for both neutral and fear imagery
magnitude of startle response decreased as rated vividness
of the image increased (overall linear trend for vividness,
F(1,17)=16.43, g<.0008). Although the vividness effect was
evident for both good and poor imagers, it was most pro-
nounced for poor imagers (Vividness X Imagery Ability linear
trend, F(1,17)=4.93, p<.05). The good imagers also tended
to show smaller magnitude startles overall than did poor
imagers (F(1,17)=3.87, .05
The results suggest that prior processing task may have
modulated imagery vividness (Vividness X Group linear trend
F(1,17)=6.18, p<.01). The linear trend for rated vividness
was clear for the Image-Image and Null Task-Image groups,
but not the Articulation-Image group.
Startle latency was significantly shorter for poor
relative to good imagers (F(1,17)=4.63, p<.05). However,
latency did not vary consistently as a function of rated
vividness of imagery.
3 250- o
0 1 2 3 4 5 6
VIVIDNESS RANK (1 =MOST VIVID)
Figure 7. Startle response magnitude during Period two
presented as a function of within-subject image
vividness ranking presented separately for neutral and
fear image trials.
Vividness and Sensory Modality
The previous section reported vivid imagery as being
associated overall with inhibited startle response. Earlier
it was predicted that acoustic imagery rated as very vivid
would be associated with facilitated startle responses,
while more vivid visual imagery should be associated with
inhibited acoustic startle responses. To test this hypo-
thesis, each subject's three neutral acoustic sentences and
three neutral visual sentences were separately ranked based
on that subject's ratings of vividness, and a Sensory
Modality X Group X Imagery Ability X Vividness (3 levels
nested within each sensory modality) ANOVA was conducted for
magnitude and latency. Sensory Modality and rated Vividness
did not interact significantly as predicted. Nevertheless,
because the modality effect on the reflex was strongest at
the Middle startle position, the analysis was repeated for
these startle responses only. Figure 8 shows the pattern of
results for this middle stimulus. Startle latency was in-
hibited with increasing vividness during visual sentences
and was facilitated with increasing vividness during acous-
tic sentences (Sensory Modality X Vividness
F(1.9,23.1)=4.11, p<.04). No reliable results were found
for magnitude in the parallel analysis.
32 -9- VISUAL
0 1 2 3
SENTENCE VIVIDNESS RANK (I =MOST VIVID)
Figure 8. Latency of response to Middle startle probes as a
function of sensory modality of the neutral sentence
(auditory or visual) and subject's ranking of the
vividness of that sentence.
Clear evidence appeared for startle reflex augmentation
during processing of fear sentences. Blink responses to the
acoustic probes were larger and faster onset during
cued recall of sentences with fear content than during cued
recall of neutral sentences or during uncued intertrial
periods. This effect tended to occur under all
instructional conditions, but was greatest when subjects
were instructed to imagine the sentence content.
The dominant modality of the image appeared to modulate
the startle response: The blink reflexes to mid-image
acoustic probes were relatively facilitated when subjects
recalled sentences with primarily auditory sensory content
than when recalling sentences with visual sensory content.
Image vividness influenced startle modulation in two
ways. First, images reported to be vivid were generally
associated with smaller blink reflexes. Second, images
rated as vivid were associated with response facilitation
during recall of auditory sentences relative to visual con-
tent sentences. Vivid imagery had a similar effect between-
subjects, as self-rated good imagers showed smaller and
longer onset startle responses relative to poor imagers;
good imagers also exhibited greater reflex facilitation
during auditory relative to visual sentence imagery. These
major results are explicated in the sections below.
Processing Fear Sentences
The activation of fearful, effectively negative mate-
rial in memory resulted in heart rate acceleration and self
report of unpleasantness, high arousal, and low dominance,
all relative to activation of relaxing, moderately positive
material from memory. This result obtained here replicates
previous findings with this (Vrana et al., in press) and
other experimental paradigms (Bauer & Craighead, 1979; Lang,
Kozak, Miller, Levin & McLean, 1980).
The overall results also indicated that processing mode
instruction modulated heart rate: When the sentences were
imagined, response information was more completely acti-
vated, as evidenced by greater heart rate during fear
imagery relative to silent articulation or null processing
of the fear material. However, silent articulation of the
text, and even explicit instructions not to process the
text, also resulted in greater heart rate acceleration fol-
lowing signals to retrieve fearful relative to neutral sen-
tences from memory.
Startle reflex facilitation was expected during fear
sentence retrieval, on the assumption that such negative
material evokes in the organism a general aversive response
disposition, and that this matched the negative response
valence of the startle reflex. Facilitated magnitude and
latency were in fact found while subjects were processing
fear sentences, relative to when subjects were processing
neutral sentences or performing the intertrial "Count 'one'"
task. This result is consistent with findings of reflex
facilitation as a component of the conditioned fear response
to a visual stimulus in man (Greenwald et al., 1988; Ross,
1961; Spence & Runquist, 1958) and animals (Berg & Davis,
1985; Brown et al., 1951); as well as human startle facili-
tation while viewing unpleasant visual material (Bradley et
al., 1988; Cook et al., 1988; Vrana et al., 1988). In the
current experiment, unlike previous work, no external stim-
ulus initiated fear processing. Comparison of these studies
emphasizes the specificity of the startle reflex as a mea-
sure of an aversive response disposition. That is, the
startle response is facilitated regardless of the media
evoking the response disposition aversivee slide content,
conditioned visual signal, or cued recall of fear sen-
tences). Across experiments, it is also independent of the
heart rate response. Heart rate accelerates during fearful
imagery, signaling activation of response code, but decele-
rates while subjects maintain an attentive set to the aver-
sive slides (Vrana et al., 1988). In each case, the reflex
to the aversive startle input is primed.
Imagery of negative material produced startle responses
of larger magnitude and shorter onset latency than neutral
imagery. As for heart rate, null processing and silent
articulation produced smaller differences in the same direc-
tion. Two conclusions are warranted here. First, the
spread of activation from language to affective response
elements in memory (i.e., from processing the words of the
sentence to processing the associated emotional response) is
to a considerable extent automatic. Second, instructions to
image may facilitate this natural process, while other
instructions may involve competing tasks (e.g., speech) which
differently engage the motor domain.
Other recent studies have also found an inability to
inhibit memory-cued material (Wegner, Schneider, Carter &
White, 1987). It is clear that instructional control over
cognitive processing, in affective or nonaffective contexts,
is only partially successful. Factors which may affect
processing include developmental level, affective valence of
the processed material, specific input (e.g., slides, text,
video) and output (physiology, self-report) variables, ima-
gery ability, and the context in which processing occurs.
For example, some of the results found here may be specific
to contexts in which processing occurs immediately upon
memory retrieval, and would not occur following a prepara-
tory period (May, 1977a; 1977b; Vrana et al., in press).
Rather than focus specifically on the instruction to image,
it is recommended that research focus more broadly on speci-
fying the conditions in which affective memory networks,
including response elements, are activated, and the implica-
tion of this process for memory network modification.
Modality-specific Effects of Sentence Processing
Attending to stimulus input in a particular sensory
modality predisposes one to greater responding to a startle
probe in that modality (Anthony, 1985). Sensory modality
effects found with environmental stimuli were replicated
here with processing of modality-specific sentence content.
That is, startle response magnitude was larger when the
sensory modality referred to in the neutral sentence matched
the startle probe modality (acoustic sentence with acoustic
probe), relative to mismatched startle and sentence modali-
ties (visual sentence with acoustic probes). A trend toward
less heart rate acceleration in acoustic relative to visual
neutral sentences suggests this magnitude difference was not
due to greater response activation in the acoustic sentences.
The "modality specific" effect found here cannot be
construed as a competition for attentional resources in a
purely perceptual context. Instead, the reflex was appar-
ently tuned to respond to information in a particular modal-
ity by activation of the event memory. Segal and Fusella
(1970) also found modality specific effects of imagery; in
that study an auditory or visual image disrupted detection
of a threshold stimulus in the same modality. It is unclear
why image activation facilitates a modality-matched reflex
elicitor but attenuates detection of modality-matched sig-
nals at the threshold level. The tasks are quite different:
Segal and Fusella presented subjects with the dual tasks of
image processing and signal detection, relying on subjective
judgment of detection as the performance measure. Imagery
was of discrete objects or sounds. In contrast, the current
study involved imagery of complete events, perhaps engaging
a more broad attentional set than Segal and Fusella, a set
known to generally enhance reflex response (Bohlin, Graham &
Silverstein, 1981). The response measure was the startle
reflex which, unlike the controlled signal detection task,
can be automatically elicited (Anthony, 1985). The auto-
matic response may be more conducive to facilitation by
activation of a particular sensory channel, while the con-
trolled process of signal detection may be more susceptible
to interference when competing with previously engaged sen-
sory pathways. Future work with various image processing
tasks, sensory signals (from threshold to startle), and
response measures (startle, orienting, subjective signal
detection) will doubtlessly shed more light on these re-
sults, and on the relationship between image activation of
sensory channels and perception.
The sentence modality effect was in turn modulated by
two other variables. First, Figures 5 and 6 make clear the
magnitude difference between visual and acoustic startle
content sentences occurred most clearly at the Middle
startle position. A similar effect occurred during the
affective valence modulation (see Table 5), as the startle
magnitude difference between neutral and fear sentence
trials is smallest immediately following the signal tone
(Early probe position). Thus for both the affective valence
and sensory content analyses, the Early startle probe
responses appear the least sensitive to sentence content or
cognitive processing variables. The Early startle probe
occurred approximately 420 msec following offset of the
preceding tone. A change in stimulus energy (onset or off-
set of a stimulus) can inhibit startle reflex responding to
a subsequent probe within this time interval (Graham, 1975).
Future studies can determine if the Early inhibition is due
to inhibition of the startle reflex response by tone offset,
or reflects the time course of the cognitive and affective
The second variable to influence sensory content modu-
lation of the sentences was image vividness. Startle
latency was facilitated with matching sentence and startle
modalities, but only among questionnaire-defined good
imagers (Figure 6). Vivid imagery produced the same effect
within subjects: Modality-specific effects were more evident
in each subject's self-reported most vivid imagery (Figure 8).
This is consistent with other studies finding content-
specific physiological activation to be more pronounced
during imagery by good relative to poor imagers (Miller,
Levin, Kozak, Cook, McLean & Lang, 1987; White, 1978), and
with the theory that the central feature of imagery is acti-
vation of context-specific response disposition (Lang, 1979).
Imagery as a Cognitive Task
Imagery of sentences containing auditory-specific con-
tent facilitated reflex response to the acoustic probe.
However, regardless of content, imagery involves internal
processing and the consequent shutting out of external stim-
uli, or "stimulus rejection" (Lacey, Kagan, Lacey, & Moss,
1963). Such tasks have been theoretically associated with
heart rate acceleration (e.g., Lacey et al., 1963). Stimu-
lus rejection should be associated with decreased response
to that input, i.e., attenuated startle response. It seems
reasonable to assume intra- and inter-individual differences
in ability to shut out environmental stimulation while proc-
essing the sentences, and that this ability might be reflec-
ted in rated image vividness, which is related to one's
ability to become absorbed in an experience (Sheehan et al.,
1978). This was in fact the case. Images ranked as more
vivid by a subject produced an inhibited startle response
relative to less vivid imagery (Figure 7). In addition,
people who rate themselves as good imagers on an imagery
questionnaire reported more vivid images overall, and exhib-
ited less response overall to startle probes than self-rated
Good imagers evidenced a less pronounced relationship
between startle magnitude and vividness than did poor
imagers. Good imagers can create vivid affective images (as
evidenced by physiological activation) to familiar and unfa-
miliar situations, while poor imagers require highly fami-
liar, personally-relevant scenarios in order to create a
vivid image (Miller et al., 1987). The sentences here depi-
cted a range of situations, some extremely familiar to sub-
jects, some not. This different relationship between
startle magnitude and image vividness may be a difference in
range of imaginal experience: Good imagers had consistently
good images (generally attenuating magnitude), while poor
imagers created images of more variable quality.
From the perspective of within-task changes in startle
response as a function of vividness, the internal orienta-
tion of the cognitive productions attenuated the startle
reflex. When compared with response to startle probes
during intertrial periods, however, sentence processing
appeared to have an activating effect on the reflex. During
Period one, sentence articulation of neutral material
resulted in facilitated startle magnitude relative to unsig-
naled intertrial periods. Neutral sentence imagery facili-
tated latency relative to intertrial startles during Period
one and Period two. It may be that startle facilitation in
these contexts involves task-induced activation; that is,
engagement in a cognitive task or changing from one task to
another increases non-specific arousal. This is consistent
with Putnam's (1975) finding of startle facilitation during
a meaningless but arousing foreground of white noise. Ano-
ther possibility is that all of the sentence contents (even
primarily visual ones) engage the tendency to respond to
auditory stimuli to a greater extent than does the inter-
trial "Count 'one'" task. Further, there may be some task-
specific priming of the auditory channel: The speech-like
task of silent articulation may engage subjects' readiness
to listen, resulting in facilitated response in the acoustic
It may appear contradictory to suggest that sentence
processing has facilitating and attenuating effects; how-
ever, several processes are hypothesized to occur during
this time. First, the context of the entire experiment
requires an internal processing orientation, including the
well-practiced, meditation-like "Count 'one"' task
(Cuthbert, Kristeller, Simons, Hodes & Lang, 1981). This
leads to a general attenuation of the startle response which
is highlighted during sentence imagery rated as particularly
vivid. Second, sentence processing generally facilitated
the startle response relative to the "Count 'one'" task.
Whether this was due to task-induced activation or sensory
engagement could be teased out by designing tasks which,
unlike the sentence processing tasks, minimized sensory
content. Third, the particular content of the sentence
materials modulated the startle response: Fear facilitated
relative to neutral content; auditory facilitated relative
to visual content. Finally, the sentence content diffe-
rences were generally attenuated at the Early startle probe
position, caused either through direct inhibition by the
preceding signal tone (Graham, 1975) or perhaps because
subjects had not yet fully initiated sentence processing.
Thus the influence on startle reflex of most interest in
this experiment (sentence content differences) occurred in
the context of other events known to modulate the startle
reflex. The various modulating variables were apparent in
the influences of different aspects of the experimental
design on the startle response.
Several investigators have proposed that startle magni-
tude is more sensitive to sensory content while latency
responds to generalized activation (Bohlin et al., 1981;
Silverstein et al., 1981). There was some indication of
this in the current study in that latency differentiated
sentence imagery and intertrial startle responses more con-
sistently than did magnitude. For the most part, however,
magnitude and latency results were similar in pattern with
latency being somewhat less sensitive to the experimental
variables. This lack of sensitivity is not surprising.
Resolution of measurement for reflex latency was one milli-
second, and the overall latency difference between startles
elicited during neutral and fear processing was 1.7 milli-
second. In contrast, resolution of startle magnitude was
one A-D unit, a fraction of the 48 A-D unit magnitude dif-
ference between neutral and fear startles.
Summary and Conclusions
Whereas much has been written about human cardiac re-
sponsivity in affective (Cuthbert et al., in press) and
nonaffective contexts (Jennings, 1986; van der Molen, Somsen
& Orlebeke, 1985), the recent human startle literature has
focused on nonaffective weak prestimulation and selective
attention to environmental stimuli (Anthony, 1985). Current
results require several modifications and additions to the
startle reflex modulation literature. First, the startle
reflex is enhanced in negatively-affective contexts regard-
less of task requirements (e.g., stimulus intake, internal
processing). Thus, consistent with earlier human subject
research (Ross, 1961; Spence & Runquist, 1958) and recent
animal research (Berg & Davis, 1984; 1985), startle facili-
tation following engagement of an aversive response disposi-
tion appears to override the modality-specific modulation
which has been the focus of startle reflex research over the
past decade (Anthony, 1985). The generality of this reflex
facilitation suggests the startle as an important, and per-
haps unique, measure of the valence dimension of emotion.
Second, finding modulation of the startle during internally-
generated processing of modality-specific content, rules out
some environmentally-driven accounts of the selective atten-
tion effect (Anthony, 1985), and suggests a top-down effect,
involving associative priming of a disposition to respond to
stimuli in the selected modality. There was also evidence
that the startle response provided an index for degree of
internal orientation involved in cognitive processing (modu-
lation of the response by image vividness), as well as gene-
ralized activation and/or sensory engagement by a cognitive
In summary, negatively-valent stimulus material, acti-
vation, sensory engagement, and internal-external processing
orientation all seem to have well-defined effects on the
startle reflex. Given these well-defined parameters, the
startle reflex presents itself as a unique tool in the study
of emotion. It appears to retain a directionally-specific
sensitivity to the valence of affective stimuli, even in
contexts where heart rate and other autonomic responses are
drastically altered by the cognitive task (Jones & Johnson,
1978; 1980; Vrana et al., 1986). The eyeblink is reflexive,
and therefore not subject to the vagaries of verbally-
mediated measures of emotion. Because the startle can be
measured without preparatory instruction or verbal response,
it is ideal for developmental, cross-cultural, and cross-
species investigations. Indeed, animal studies already
suggest startle as a measure of treatment outcome in anxiety
disorders (Berg & Davis, 1984). The startle response, par-
ticularly in combination with other measures of affect,
promises to be a versatile and fruitful means to study emo-
Anderson, J.R. & Bower, G.H. (1973). Human associative
memory. Washington, D.C.: Winston.
Anthony, B.J. (1985). In the blink of an eye: Implications
of reflex modification for information processing. In
P.K. Ackles, J.R. Jennings, and M.G.H. Coles (Eds.),
Advances in Psychophysiology (Vol. 1). Greenwich,
Conn.: JAI Press.
Anthony, B.J. & Graham, F.K. (1983). Evidence for sensory-
selective set in young infants. Science, 220, 742-744.
Anthony, B.J. & Graham, F.K. (1985). Blink reflex modifica-
tion by selective attention: Evidence for the modula-
tion of 'automatic' processing. Biological Psychology,
Balaban, M., Losito, B., Simons, R.F., & Graham, F.K.
(1986). Off-line latency and amplitude scoring of the
human reflex eyeblink with Fortran IV.
Psychophysiology, 23, 612.
Bauer, R.M. & Craighead, W.E. (1979). Psychophysiological
responses to the imagination of fearful and neutral
situations: The effects of imagery instructions.
Behavior Therapy, 10, 389-403.
Berg, W.K. & Davis, M. (1984). Diazepam blocks fear-
enhanced startle elicited electrically from the brain-
stem. Physiology and Behavior, 32, 333-336.
Berg, W.K. & Davis, M. (1985). Associative learning modi-
fies startle reflexes at the lateral lemniscus.
Behavioral Neuroscience, 90, 191-199.
Betts, G.H. (1909). The distribution and functions of men-
tal imagery. New York: Columbia University Teachers
College, Contribution to Education Series, No. 26.
Bohlin, G. & Graham, F.K. (1977). Cardiac deceleration and
reflex blink facilitation. Psychophysiology, 14, 423-
Bohlin, G., Graham, F.K., Silverstein, L.D., & Hackley,
S.A. (1981). Cardiac orienting and startle blink modi-
fication in novel and signal conditions.
Psychophysiology, 18, 603-611.
Bradley, M.M., Cuthbert, B.N., & Lang, P.J. (1988).
Lateral presentation of acoustic startle stimuli in a
varying affective foreground. Psychophysiology, 25,
Brown, J.S., Kalish, H.I., & Farber, I.E. (1951). Condi-
tioned fear as revealed by magnitude of startle
response to an auditory stimulus. Journal of
Experimental Psychology, 43, 317-328.
Cook, E.W., Spence, E.L., Gray, J.D., & Davis, T.L. (1988).
Acoustic startle is modulated by affective associative
learning. Psychophysiology, 25, 440-441.
Cuthbert, B., Kristeller, J., Simons, R., Hodes, R. & Lang,
P.J. (1981). Strategies of arousal control:
Biofeedback, meditation, and motivation. Journal of
Experimental Psychology: General, 110, 518-546.
Cuthbert, B.N., Vrana, S.R. & Bradley, M.M. (in press).
Imagery: Function and physiology. In P.K. Ackles, J.R.
Jennings, and M.G.H. Coles (Eds.), Advances in
psychophysiology (Vol. 4). Greenwich, Conn.: JAI
Davis, M. & Astrachan, D.I. (1978). Conditioned fear and
startle magnitude: Effects of different footshock or
backshock intensities used in training. Journal of
Experimental Psychology: Animal Behavior Processes, 4,
Graham, F.K. (1975). The more or less startling effects of
weak prestimulation. Psychophysiology, 12, 238-248.
Greenhouse, S.W. & Geisser, S. (1959). On methods in the
analysis of profile data. Psychometrika, 24, 95-112.
Greenwald, M.K., Hamm, A.O., Bradley, M.M. & Lang, P.J.
(1988). The acoustic startle probe in the assessment
of classical aversive conditioning. Psychophysiology,
Hackley, S.A. & Graham, F.K. (1984). Early selective atten-
tion effects on cutaneous and acoustic blink reflexes.
Physiological Psychology, 11, 235-242.
Jennings, J.R. (1986). Memory, thought, and bodily
response. In M.G.H. Coles, E. Donchin, and S.W. Porges
(Eds.), Psychophysiology: Systems, processes, and
applications. New York: Guilford Press.
Jones, G.E. & Johnson, H.J. (1978). Physiological respon-
ding during self generated imagery of contextually
complete stimuli. Psychophysiology, 15, 439-446.
Jones, G.E. & Johnson, H.J. (1980). Heart rate and somatic
concomitants of mental imagery. Psychophysiology, 17,
Kintsch, W. (1974). The representation of meaning in
memory. Hillsdale, N.J.: Laurence Erlbaum.
Kurtz, K.H. & Seigel, A. (1966). Conditioned fear and
magnitude of startle response: A replication and exten-
sion. Journal of Comparative and Physiological
Psychology, 62, 8-14.
Lacey, J.I., Kagan, J., Lacey, B.C. & Moss, H.A. (1963).
The visceral level: Situational determinants and
behavioral correlates of autonomic response patterns.
In P.H. Knapp (Ed.), Expression of the emotions in man.
New York: International Universities Press.
Lang, P.J. (1979). A bio-informational theory of emotional
imagery. Psychophysiology, 16, 495-512.
Lang, P.J. (1983). Cognition in Emotion: Concept and
Action. In C. Izard, J. Kagan and R. Zajonc (Eds.),
Emotion, Cognition, and Behavior. New York: Cambridge
Lang, P.J. (1985). The cognitive psychophysiology of
emotion: Fear and anxiety. In A.H. Tuma and J.D. Maser
(Eds.), Anxiety and the Anxiety Disorders. Hillsdale,
N.J.: Laurence Erlbaum.
Lang, P.J., Kozak, M.J., Miller, G.A., Levin, D.N., &
McLean, Jr., A. (1980). Emotional Imagery: Conceptual
structure and pattern of somato-visceral response.
Psychophysiology, 17, 179-192.
May, J.R. (1977a). Psychophysiology of self regulated
phobic thoughts. Behavior Therapy, 8, 150-159.
May, J.R. (1977b). A psychophysiological study of self and
externally regulated phobic thoughts. Behavior
Therapy, 8, 849-861.
Miller, G.A., Levin, D.N., Kozak, M.J., Cook, E.W., McLean,
A., & Lang, P.J. (1987). Individual differences in
imagery and the psychophysiology of emotion. Cognition
and Emotion, 1, 367-390.
Osgood, C.E., Suci, G.J., & Tannenbaum, P.H. (1957). The
Measurement of Meaning. Urbana: University of Illinois
Putnam, L.E. (1975). The human startle reaction: Mechanisms
of modification by background acoustic stimulation
(Doctoral Dissertation, University of Wisconsin, 1975).
Dissertation Abstracts International, 36, 6419-B.
Quillian, M.R. (1966). Semantic memory. In M.L. Minsky
(Ed.), Semantic information processing. Cambridge,
Mass.: MIT Press.
Ross, L.E. (1961). Conditioned fear as a function of CS-UCS
and probe stimulus intervals. Journal of Experimental
Psychology, 61, 265-273.
Schwartz, G.E. (1971). Cardiac responses to self-induced
thoughts. Psychophysiology, 8, 462-466.
Segal, S.J. & Fusella, V. (1970). Influence of imaged
pictures and sounds on detection of visual and auditory
signals. Journal of Experimental Psychology, 83, 458-464.
Sheehan, P.W. (1967). A shortened form of Betts' question-
naire upon mental imagery. Journal of Clinical
Psychology, 223, 380-389.
Sheehan, P.W., McConkey, K.M. & Law, H.G. (1978). Imagery
facilitation and performance on the Creative Imagina-
tion Scale. Journal of Mental Imagery, 2, 265-274.
Silverstein, L.D., Graham, F.K., & Bohlin, G. (1981). Selec-
tive attention effects on the reflex blink.
Psychophysiology, 18, 240-247.
Simons, R.F. & Zelson, M.F. (1985). Engaging visual stimuli
and reflex blink modification. Psychophysiology, 22,
Spence, K.W. & Runquist, W.N. (1958). Temporal effects of
conditioned fear on the eyelid reflex. Journal of
Experimental Psychology, 55, 613-616.
van der Molen, M.W., Somsen, R.J.M. & Orlebeke, J.F. (1985).
The rhythm of the heart beat in information processing.
In P.H. Ackles, J.R. Jennings, and M.G.H. Coles (Eds.),
Advances in psychophysiology (Vol. 1). Greenwich,
Conn.: JAI Press.
Vrana, S.R., Cuthbert, B.N., & Lang, P.J. (1986). Fear
imagery and text processing. Psychophysiology, 23,
Vrana, S.R., Cuthbert, B.N., & Lang, P.J. (in press). Proc-
essing fearful and neutral sentences: Memory and heart
rate change. Cognition and Emotion.
Vrana, S.R., Spence, E.L., & Lang, P.J. (1988). The
startle probe response: M new measure of emotion?
Journal of Abnormal Psychology, 97, 487-491.
Wegner, D.M., Schneider, D.J., Carter, S.R. & White, T.L.
(1987). Paradoxical effects of thought suppression.
Journal of Personality and Social Psychology, 53, 5-13.
White, K.D. (1978). Salivation: The significance of
imagery in voluntary control. Psychophysiology, 15,
Zelson, M.F. & Simons, R.F. (1986). Sustained attention in
Type A and Type B subjects: A blink reflex analysis.
Psychophysiology, 23, 385-392.
STIMULI AND SUBJECT INSTRUCTIONS
Subject Instructions: Null Task-Image Group,
High Tone=Fear Sentences
I am now going to read you the instructions for this
experiment. In a little while you will memorize two sen-
tences, one fearful and one neutral in content. You will
use these sentences to create images in your mind using the
following procedure. After you memorize the two sentences
and I leave the room, you will hear a series of short tones,
one every six seconds. Each tone will be at one of three
different frequencies. The tone that you will hear most
often is the middle frequency tone. I'll call this the
"normal" tone. Whenever you hear this tone, just relax and
think the word "one" to yourself each time you breathe out.
This is to help clear your mind and help you remain relaxed.
Sometimes you will hear a higher-pitched tone, which
will always be presented twice in a row at the usual six-
second interval. When you hear this tone the first time,
just continue to think "one" to yourself and clear your
mind. At the second high tone, begin to imagine the fear
scene as a vivid, personal experience. When you do this,
try to imagine you are actually in the situation and parti-
cipating in the events described, and not just "watching
yourself" in the scene. To review, you will hear the normal
tone, and think "one" to yourself until the next tone. At
the first higher-pitched tone, continue to think "one" to
yourself. At the second high tone, imagine the fear scene.
Continue with your image until the next normal tone, then
begin to think "one" to these tones again.
Sometimes you will hear a tone that is at a lower
frequency than the normal tone. As for the high tones, the
low tones will always be presented twice in a row at the
normal six-second interval. When you hear the first low
tone, continue to clear your mind and think "one" to your-
self. At the second low tone, imagine the neutral scene as
vividly as you can. Once again, try to imagine you are
actually participating in the situation described. To
review, when you hear the lower-pitched tone for the first
time, continue thinking "one" to yourself; at the second low
tone, imagine the neutral scene until the next normal tone,
then begin to think "one" to these tones again.
Let me summarize. You will hear short tones occurring
every six seconds. Most of the tones will be normal tones,
and when you hear these clear your mind by thinking "one" to
yourself. Every once in a while you will hear a pair of
tones that are either higher or lower in frequency than the
normal tone. When you hear one of these tones, continue to
think one to yourself, then at the next such tone, create an
image to the appropriate sentence. If the different tones
are higher-pitched, create an image to the fear sentence.
If the different tones are lower-pitched, create an image to
the neutral sentence. At times you will hear loud clicks,
like a finger snapping. They are meant to elicit a response
we wish to measure, but you just need to ignore them and
continue with the task. In just a minute I will play
examples of the tones and noise clicks and we can go through
this sequence of events. Do you have any questions about
After you have heard a sequence of tones lasting about
ten minutes, you will hear two tones in quick succession.
At this point open your eyes and rate your images for their
pleasantness, arousal, and vividness. I will show you how
to do this in a little while. After you rate your images,
you will memorize another two sentences, and then go through
the tone series again. There will be a number of sentence
pairs and tone sequences. Now I will start the tone sequence.
Start Practice Trials
These are the normal tones. They will occur every six
seconds. When you hear these tones, just clear your mind
and think "one" to yourself each time you exhale. That was
a high tone. Continue thinking one to yourself. When you
hear this high tone imagine the fear sentence. Continue
your image until was one of the noise clicks. Just ignore
it and continue counting "one" to yourself. That was the
lower tone. When you hear that tone continue to think one
until this low tone. Continue your image until the next
normal tone. Now go back to thinking "one" to yourself.
Continue with this until you hear a double tone. Like this
one; then open your eyes to rate your images to the
sentences. I want you to make each rating based on the
average of all your images to that sentence.
Here are the first two sentences to memorize. Tell me
when you have them memorized, and then I will have you
repeat the parts in capital letters back to me. Please read
the whole sentence and use all the information in it for
your image, but you only have to memorize and repeat the
phrases in capital letters.
Good. I would like to remind you how the tone sequence
will go. When I leave the room, close your eyes and get as
comfortable as you can in the chair. When you hear each
"normal" tone think "one" to yourself. Try to clear your
mind and not think of anything but the number "one" at this
point. When you hear a high-pitched tone, continue to clear
your mind and think one and then at the next, high-pitched,
tone create an image to the fear sentence you just memo-
rized. When you hear a low-pitched tone, continue thinking
one and then at the next, low-pitched, tone create an image
to the neutral sentence you just memorized. Stop your image
at the next normal tone and go back to thinking "one" at
each tone until you hear the next high- or low- pitched
tone. Continue this until you hear two quick tones, then
make your ratings. Do you have any questions?
Remember, your main job in this experiment is to create
vivid images, and to imagine you are really participating in
the scene. Just ignore the noise clicks--the first few may
make you jump a bit but after that you'll get used to them.
I will tell in just a minute when the tones will start. Try
to move as little as possible through the whole tone
sequence, as this will effect the physiological recording.
Put the headphones on and get yourself seated as comfortable
as possible now and we will get started in just a minute.
Other Subject Group Instructions
Instructions for this group were the same, except they
were told that at the first high- or low-pitched tone they
were to "'think' the sentence. This means repeat the words
of the sentence over in your head. At the second high tone,
begin to imagine the scene as a vivid, personal experience."
Instructions for this group were the same, except they
were told that at the first high- or low-pitched tone they
were to "create an image to the sentence as a vivid,
personal experience. When you do this, try to imagine you
are actually in the situation and participating in the
events described, and not just watching yourself in the
scene. At the second high tone, again create an image to the
scene as a vivid, personal experience."
The bell sounds, the students wait impatiently, MY
HEART POUNDS AS I BEGIN MY SPEECH TO THE CLASS.
I grip the chair, heart racing, as THE DENTIST HOOKS MY
GUMLINE AND COLD STEEL SCRAPES ACROSS MY TEETH.
I tense as THE NURSE SLOWLY INJECTS THE SHARP NEEDLE
INTO MY UPPER ARM, and beads of sweat cover my forehead.
I flinch at the screech of brakes; MY COMPANION IS
STRUCK BY A SPEEDING CAR; HER LEG IS CRUSHED, bone
protruding, AND BLOOD PUMPS ONTO THE ROAD.
Taking a shower, ALONE IN THE HOUSE, I HEAR THE SOUND
OF SOMEONE FORCING THE DOOR, and I panic.
ALONE IN BED, I FEEL a scuttling along my bare leg; I
switch on the light, and trembling, see A LARGE, BLACK
SPIDER MOVING UP MY THIGH.
I AM RELAXING on my living room couch LOOKING OUT THE
WINDOW ON A SUNNY AUTUMN DAY.
I AM sitting in a lawn chair on the front porch
WATCHING THE SOFT SUMMER BREEZE SWAY THE LEAVES ON THE
A wood fire dances in the hearth, I FEEL SNUG AND WARM
IN THE CABIN, READING THE BOOK ON MY LAP, enjoying a well-
SOFT MUSIC IS PLAYING ON THE STEREO, AS I SNOOZE LAZILY
on my favorite chair.
I AM LYING ON THE SAND on a warm day, LISTENING TO
CHILDREN PLAYING DOWN THE BEACH, their soft voices mingling
with the sound of the waves.
I AM LYING IN BED on a Sunday morning, half asleep and
LISTENING TO THE DISTANT SOUND OF BELLS, relaxing on my day
DIFFERENCES BETWEEN STARTLE AND NON-STARTLE EXPERIMENT
The startle experiment reported in the Results section
and the non-startle study reported in the Introduction are
nearly identical in design, procedure, and materials. The
differences between the two studies are stressed here in
order to assist in clear interpretation of the results. The
greatest difference between the two studies is the inclusion
of the startle reflex measure in the second study; this
required two additional electrodes placed near the subject's
eye as well as presenting the white noise bursts during and
between trials, as described in Methods. All auditory stim-
uli in the startle experiment were presented using head-
phones, rather than the speaker used in the non-startle
study, in order to present the 95 dB noise burst without
disrupting other activities in the laboratory. The fre-
quency of the low, medium, and high tones in the startle
study were 800, 1100, and 1500 Hz, respectively, rather than
the 500, 800, and 1100 Hz tones presented in the non-startle
experiment. This was to eliminate the sound pressure level
differences found in the three tone frequencies in the
initial, non-startle study.
The materials were somewhat different in the two
studies. Four neutral sentences were slightly re-written to
explicitly refer to the auditory or visual modality, and
word-for-word sentence memorization in the non-startle study
was modified to word-for-word memorization of only key
phrases of each sentence for the startle study. The dif-
ferences in sentences can be examined by comparing the non-
startle sentence materials at the end of this Appendix with
the startle study sentence materials in Appendix A. Imagery
ratings (pleasure, arousal, dominance and vividness) were
performed by making a numerical rating in the non-startle
study; in the startle study these ratings were performed by
marking a horizontal line. These ratings were quantified so
that each dimension had the same range (0-29) and the same
meaning in each experiment (for example, a rating of twenty-
nine on the valence dimension represents maximum pleasure in
Finally, obtaining the desired number of data points
for the startle reflex required increasing the number of
trials in each block from eight (four neutral and four fear
trials) to twelve (six neutral and six fear). This
increased the length of each block of trials from about five
and a half minutes in the non-startle experiment to over
eight minutes in the startle experiment.
Fear Sentences: Non-Startle Study
The bell sounds, the students wait impatiently, my
heart pounds as I begin my speech to the class.
I grip the chair, heart racing, as the dentist hooks my
gumline and cold steel scrapes across my teeth.
I tense as the nurse slowly injects the sharp needle
into my upper arm, and beads of sweat cover my forehead.
I flinch at the screech of brakes; my companion is
struck by a speeding car; her leg is crushed, bone
protruding, and blood pumps onto the road.
Taking a shower, alone in the house, I hear the sound
of someone forcing the door, and I panic.
Alone in bed, I feel a scuttling along my bare leg; I
switch on the light, and trembling, see a large, black
spider moving up my thigh.
Neutral Sentences: Non-Startle Study
I am relaxing on my living room couch looking out the
window on a sunny autumn day.
I am sitting in a lawn chair on the front porch
enjoying the soft summer breeze.
A wood fire dances in the hearth, I feel snug and warm
in the cabin, a good book in my lap, enjoying a well-
Soft music is playing on the stereo, as I snooze lazily
on my favorite chair.
I am lying on the sand on a warm day, children are
playing down the beach, and their soft voices mingle with
the sound of the waves.
I am lying in bed on a Sunday morning, half asleep and
listening to the distant sound of bells, relaxing on my day
HEART RATE FIGURES
Figure 9. No-startle study data: Continuous heart rate
waveform in half-second intervals for each group for
the "Count 'one'" period and the first sentence proc-
essing period. The tone cueing retrieval of the neut-
ral or fearful sentence is signified by a vertical line
at the six second mark in each graph.
76 "ONE" NULL
68 h/ -
0 6 12
S 72 "ONE" ARTICULATE
w -0-- NEUTRAL
S 68 -*- FEAR
0 6 12
0 6 12
Figure 10. Startle study data: Continuous heart rate
waveform in half-second intervals for each group for
the "Count 'one'" period and the first sentence proc-
essing period. The tone cueing retrieval of the neut-
ral or fearful sentence is signified by a vertical line
at the six second mark in each graph.
TABLES OF STARTLE REFLEX DATA
Null-Image Group: Startle Reflex Magnitude
and Startle Probe Time
by Content, Period,
Null-Image Group: Startle
and Startle Probe Time
Reflex Latency by Content, Period,
Articulate-Image Group: Startle Reflex Magnitude by Content,
Period, and Startle Probe Time