Content uploaded by E. Glenn Schellenberg
Author content
All content in this area was uploaded by E. Glenn Schellenberg
Content may be subject to copyright.
Liking and Memory for Musical Stimuli as a Function of Exposure
Karl K. Szpunar, E. Glenn Schellenberg, and Patricia Pliner
University of Toronto
Three experiments examined changes in liking and memory for music as a function of number of
previous exposures, the ecological validity of the music, and whether the exposure phase required
focused or incidental listening. After incidental listening, liking ratings were higher for music heard more
often in the exposure phase and this association was stronger as ecological validity increased. After
focused listening, liking ratings followed an inverted U-shaped function of exposure for the most
ecologically valid stimuli (initial increases followed by decreases), but this curvilinear function was
attenuated or nonexistent for less valid stimuli. In general, recognition improved as a function of previous
exposure for focused listeners, but the effect was attenuated or absent for incidental listeners.
Since Zajonc’s (1968) report that “mere exposure is a sufficient
condition for attitude enhancement” (p. 15), there have been nu-
merous replications and extensions of the finding that simple
exposure to a novel, neutral stimulus increases liking for it (for a
review, see Bornstein, 1989). On the basis of a meta-analysis of the
relevant literature, Bornstein concluded that the effect is stronger
(a) for complex than for simple stimuli, (b) for exposure to
multiple stimuli than for a single stimulus presented repeatedly, (c)
for brief stimuli (e.g.,1sorless for visual stimuli) with relatively
few repetitions, and (d) when a delay intervenes between exposure
and ratings of liking.
One puzzling and provocative aspect of the mere exposure effect
is the occurrence of increases in liking even in the absence of
stimulus recognition (e.g., Zajonc, 1980). In fact, the effect is
stronger under subliminal than supraliminal conditions (Murphy,
Monahan, & Zajonc, 1995) and when participants have no explicit
memory for the stimuli (Bornstein, 1989; Bornstein & D’Agostino,
1992). In one study (Kunst-Wilson & Zajonc, 1980), participants
were exposed to five 1-ms exposures of each of several polygons.
They subsequently saw pairs of polygons, one of which had been
presented previously. Their tasks were to select the preferred
polygon in each pair and to select the one that seemed more
familiar. Although previously exposed polygons were preferred
more often than chance (60%), recognition was at chance levels
(48%). The influence of exposure on preference in the absence of
explicit memory has been documented repeatedly in the visual
domain (Barchas & Perlaki, 1986; Bonanno & Stillings, 1986;
Bornstein, Leone, & Galley, 1987; Mandler, Nakamura, & Van
Zandt, 1987; Seamon, Brody, & Kauff, 1983; Seamon, Marsh, &
Brody, 1984) and in the auditory domain (Peretz, Gaudreau, &
Bonnel, 1998; Wilson, 1979). On the basis of these findings,
Zajonc (2001) argued that exposure effects can occur without
previous cognitive appraisal, or that “preferences need no infer-
ences” (Zajonc, 1980, p. 151).
Zajonc (2001) attributed the association between exposure and
preference to conditioning, specifically to the absence of adverse
consequences after exposure to the target stimulus. He and his
associates argued, moreover, that exposure effects are diffuse,
leading to enhanced affect for stimuli other than those exposed
previously (Monahan, Murphy, & Zajonc, 2000; Murphy et al.,
1995) and to mood elevation in those who experience repeated
exposures (Monahan et al., 2000). Nonetheless, their perspective
was limited to stimuli presented subliminally. It is therefore of
limited relevance to the vast majority of stimuli encountered in
daily life or in the laboratory.
Two theories provide accounts of mere exposure effects that
extend to supraliminal stimuli. One, the perceptual fluency/attri-
butional model (Bornstein, 1992; Bornstein & D’Agostino, 1994),
emphasizes different cognitive ramifications of recognized and
unrecognized stimuli. Previous exposure is thought to activate
context-free representations of stimuli, resulting in perceptual
fluency (Jacoby, 1983; Jacoby & Kelly, 1987; Jacoby & White-
house, 1989), which refers to rapid and efficient processing of
previously encountered stimuli. Depending on the available con-
textual cues, participants may misattribute perceptual fluency to
properties of the stimulus instead of to previous exposure (Jacoby
& Whitehouse, 1989; Mandler et al., 1987). For example, if
exposure is followed by evaluative questions about an unrecog-
nized stimulus, participants interpret their apparent ease of pro-
cessing as a positive disposition toward the stimulus (Bornstein,
1992). Similarly, questions about stimulus brightness and darkness
result in enhanced judgments of brightness and darkness, respec-
tively (Mandler et al., 1987). According to Bornstein (1992; Born-
stein & D’Agostino, 1994), repeated supraliminal presentation
generates increased awareness of the stimulus, which leads partic-
ipants to “correct” or reinterpret fluency-based affective responses.
Essentially, participants’ explicit memory enables them to interpret
the perceptual fluency they experience in terms of exposure rather
Karl K. Szpunar, E. Glenn Schellenberg, and Patricia Pliner, Depart-
ment of Psychology, University of Toronto, Mississauga, Ontario, Canada.
Karl K. Szpunar is now at the Department of Psychology, Washington
University in St. Louis.
This research was supported by the Natural Sciences and Engineering
Research Council of Canada. We thank Sandra Trehub and Frank Russo for
their helpful comments on earlier versions of this article.
Correspondence concerning this article should be addressed to E. Glenn
Schellenberg, Department of Psychology, University of Toronto at Mis-
sissauga, Mississauga, Ontario L5L 1C6, Canada. E-mail: g.schellenberg@
utoronto.ca
Journal of Experimental Psychology: Copyright 2004 by the American Psychological Association, Inc.
Learning, Memory, and Cognition
2004, Vol. 30, No. 2, 370–381
0278-7393/04/$12.00 DOI: 10.1037/0278-7393.30.2.370
370
than affect. The presumption is that increases in recognition ac-
curacy will be accompanied by decreases in positive affect toward
a stimulus.
An alternative account of associations between liking and ex-
posure is provided by the two-factor model proposed by Berlyne
(1970) and Stang (1974). Evaluations of supraliminally presented
stimuli are said to form an inverted U-shaped function of arousal
potential. Stimuli with little arousal potential—those that are too
familiar or too simple—receive relatively low evaluations, as do
stimuli with too much arousal potential, those that are too unfa-
miliar or too complex. Rather, stimuli with intermediate levels of
arousal potential are evaluated most favorably. The model includes
two opposing processes. One involves the dissipation of neophobia
or the development of “learned safety” (Kalat & Rozin, 1973),
which generates enhanced affect as a stimulus becomes more
familiar and less threatening through benign contact. This compo-
nent is consistent with Zajonc’s (2001) account. The second pro-
cess is boredom, which increases with increasing exposure. Pre-
sumably, exposure generates increases in positive affect until
boredom outweighs the benefits of learned safety, resulting in
satiation, or decreases in liking.
Although stimulus-enhancement effects after repeated exposure
are highly reliable, satiation effects are elusive. In some cases,
satiation occurs after relatively few exposures (e.g., Bornstein,
Kale, & Cornell, 1990; Kail & Freeman, 1973; Stang &
O’Connell, 1974). In others, no satiation is apparent after 80 or
more exposures (Zajonc, Crandall, Kail, & Swap, 1974; Zajonc,
Swap, Harrison, & Roberts, 1971). In a particularly good example
of the ephemeral nature of satiation effects, Zajonc, Shaver,
Tavris, and van Kreveld (1972) found that liking ratings for
reproductions of artworks increased from zero to five exposures
but decreased steadily with further exposure. When the experiment
was repeated with photographs of men and nonsense syllables,
stimulus evaluation increased as a function of exposure, but there
was no evidence of satiation.
Anecdotal observations are suggestive of striking enhancement
and satiation effects involving music. For example, radio and TV
exposure of popular recordings has well-documented effects on
consumer purchases of such recordings, not to mention widespread
singing of these songs in the shower and elsewhere. For a time,
consumers clamor for repeated airing of these recordings, followed
by increasing dislike of the songs heard most frequently (Jakobo-
vits, 1966). A more dramatic pattern of preferences sometimes
emerges in individuals who normally avoid orchestral music.
These individuals can become devotees of specific pieces (e.g.,
Bolero by Ravel, Canon in D Major by Pachelbel, or Also Sprach
Zarathustra by Strauss) after the music has been used for dramatic
effect in popular films (e.g., 10, Ordinary People, or 2001: A
Space Odyssey). Although it is impossible to make unequivocal
attributions of such liking to simple exposure rather than to spe-
cific, positive associations, these observations encouraged us to
examine relations among exposure, liking, and memory for music.
In the last 100 years, there have been approximately 20 pub-
lished studies concerned with the affective consequences of expo-
sure to music. In most cases, there were small to moderate num-
bers of presentations, which resulted in the usual increases in
liking (Brickman & D’Amato, 1975; Heingartner & Hall, 1974;
Krugman, 1943; Meyer, 1903; Mull, 1957; Obermiller, 1985;
Peretz, Gaudreau, & Bonnel, 1998; Wilson, 1979). In some cases,
increases in affect varied as a function of the type of music
presented (Bartlett, 1973; Bradley, 1971; Downey & Knapp, 1927;
Gilliland & Moore, 1927; Washburn, Child, & Abel, 1927). Sug-
gestive evidence of satiation was apparent in some studies (Bart-
lett, 1973; Brentar, Neuendorf, & Armstrong, 1994; Hargreaves,
1984; Heyduk, 1975; Verveer, Barry, & Bousfield, 1933), but none
of these studies examined possible links between evaluations and
memory. The absence of satiation in most studies may result from
insufficient exposure. Specifically, the characteristic focus on in-
creases in liking as a function of exposure may have led to a
limited number of exposures because increases often plateau after
about 10 exposures (Bornstein, 1989). Indeed, one of the two
studies that involved the use of more than 20 presentations re-
vealed a small decline in liking after 16 presentations (Brentar et
al., 1994).
The principal goal of the present investigation was to examine
liking for music as a function of previous exposure. A second goal
was to identify how liking and explicit memory covary as a
function of exposure in different contexts. Accordingly, partici-
pants rated how much they liked each stimulus and how well they
remembered it. In each of the three experiments, we attempted to
maximize the likelihood of recognition among some listeners by
requiring them to focus on the music during the exposure phase.
We reduced the likelihood of recognition among other listeners by
having them hear the same stimuli incidentally while they focused
on a distractor task. The vast majority of musical experiences
involve incidental listening, with focused music listening (e.g., at
concerts) being an important but much less common activity
(Sloboda, O’Neill, & Ivaldi, 2001).
In all three experiments, we included a wide range of exposure
frequencies that were conducive to liking and to satiation so that
we could evaluate whether moderate numbers of presentations
would produce the usual increase in liking and whether large
numbers would produce a decrease in liking. The experiments
differed in the kinds of musical stimuli and the tasks used in the
focused-listening conditions. This variation was motivated by
Bornstein’s (1989) conclusion that exposure effects depend on
stimulus complexity. We assumed that, in the musical domain,
stimulus complexity was roughly equivalent to ecological validity
and that the most ecologically valid experiences involved listening
to recordings of real music in a musically relevant manner. Thus,
the three experiments differed in the extent to which the exposure
phase was experienced as involving real music. Experiment 1
provided the least musical experience, Experiment 2 provided the
most, and Experiment 3 fell between the two. Accordingly, if
ecological validity affects liking, the results of Experiment 3
should be intermediate between those of Experiments 1 and 2.
Experiment 1
We evaluated listeners’ liking and memory for tone sequences
presented at different exposure frequencies. The stimuli were
monophonic (one tone at a time) sequences of five to nine piano
tones that did not conform to any Western major or minor scale.
They consisted of tones of equal amplitude and duration, such that
the sequences sounded unfamiliar, mechanical, and relatively non-
musical. These relatively impoverished stimuli could be described
in terms of a few parameters (number of tones, specific pitches,
and pitch contour). The sequences were not ecologically valid, but
371
REPEATED EXPOSURE TO MUSIC
they were similar to those used previously (Wilson, 1979). More-
over, although simplicity (or complexity) is difficult to quantify,
our tone sequences were analogous in some respects to the con-
trolled visual stimuli (i.e., polygons, line drawings, and Chinese
ideographs) used in many studies (Bornstein & D’Agostino, 1992,
1994; Bornstein et al., 1987, 1990; Kunst-Wilson & Zajonc, 1980;
Mandler et al., 1987; Monahan et al., 2000; Seamon et al., 1983,
1984, 1995; Whittlesea & Price, 2001), which can also be de-
scribed with a few parameters. The orienting task that required
listeners in the focused condition to attend closely to the stimuli
was also nonmusical. These listeners were required to count the
total number of tones in each sequence.
We expected that repeated exposure would improve recognition
and that this association would be a more robust consequence of
focused than incidental listening. We also expected exposure-
dependent increases in liking to be greater with incidental listening
because of limited memory for the stimuli (Bornstein, 1989). Because
focused listening accelerates the familiarization process, it was ex-
pected to promote boredom resulting from excessive familiarity, with
the nonmusical orienting task exaggerating this process.
Method
Participants. The listeners were 50 undergraduates who received par-
tial course credit or token remuneration for participating in the study.
Apparatus and stimuli. Testing took place in a sound-attenuating
booth. The stimuli were presented over headphones (Sony MDR-CD370)
at a comfortable listening level. The stimuli were 30 sequences of 5 to 9
piano tones from Thompson, Balkwill, and Vernescu (2000). Their re-
search was not designed to test effects of exposure on preference and
memory, but a subsidiary finding was higher ratings for previously exposed
sequences than for novel sequences, even when the former sequences were
not recognized. We modified several sequences (Thompson et al., 2000,
Appendix) so that we had an equal number (i.e., 6) of each of five lengths
(5, 6, 7, 8, or 9 tones). For example, a sequence of 6 tones—1, 2, 3, 4, 5,
6—was extended to 9 tones by adding Tones 5, 4, and 3 to the end.
Examples of the tone sequences are provided in Figure 1 with musical
notation.
Musical Instrument Digital Interface (MIDI) files were created for each
sequence with sequencing software (Cubase 2.8), ensuring that each tone
had identical duration and amplitude. Interonset intervals between tones
within sequences were 300 ms. The files were routed to a MIDI interface
(Mark of the Unicorn) and a Roland JV-90 synthesizer (piano timbre) and
then rerecorded as digital sound files (CD quality, i.e., 16 bit, sampling rate
of 44.1 kHz) with SoundEdit 16 (Version 2) software. Stimulus presenta-
tion and response recording were controlled by a customized program
created with PsyScope 1.1 software (Cohen, MacWhinney, Flatt, & Pro-
vost, 1993). The distracting stimulus was an excerpt from a narrated story
written by Stephen King called The Green Mile (King, 1996). The text was
rerecorded from an audiocassette and stored as a digital sound file.
Procedure. Thirty participants were assigned to focused listening and
20 to incidental listening. We included additional participants for focused
listening to ensure that null effects on liking ratings did not result from a
lack of power. The procedure consisted of three phases that occurred in
immediate succession: (a) exposure, (b) liking ratings, and (c) recognition
ratings. Only the exposure phase differed across conditions.
Participants were informed that the procedure had three phases. For each
participant in focused listening, 18 target stimuli were selected randomly
from the set of 30. Participants heard 6 of the 18 stimuli in the exposure
phase, 2 of which were presented 4 times, 16 times, or 64 times, for a total
of 168 presentations. Randomization was constrained so that the same
excerpt could not occur on consecutive trials. Participants’ task was to
count the number of tones in each sequence and select one of the response
options of 4 to 10. They were told that some sequences would differ very
subtly from others and that we were interested in the effects of this
manipulation on response accuracy and speed. In fact, the task was de-
signed simply to ensure that they listened closely to each presentation.
Trials were self-paced; participants indicated their readiness for the next
trial by pressing the space bar.
Each participant in incidental listening heard one of five different strings
of 168 sequences ordered randomly. Each string contained six different
sequences, with randomization of repetitions identical to focused listening.
As a means of minimizing variation among participants in the duration of
the test session, the randomization was constrained so that the sequences
presented 64 times had 6 and 7 tones, those presented 16 times had 5 and
9 tones, and those presented 4 times had 8 tones. Consecutive sequences
were separated by 300 ms of silence (corresponding to the duration of one
tone). The distracting stimulus (the excerpt from a narrated story) matched
the overall duration of the stimulus strings. Participants were instructed that
they would hear two stimuli at once. In their right ear, they heard the
narrated story, its maximum amplitude equaling that of sequences in
focused listening (i.e., the peak amplitude was 80% of maximum before
distortion). In their left ear, they heard the musical excerpts presented at
reduced amplitude (30% of maximum). Participants were told to focus their
attention on the story and to try to ignore the distracting stimulus in their
left ear. Their task was to press one button on a button box when they heard
the narrator say the word the and another button when they heard the word
and. Participants were also instructed to count and remember how often
they heard the word but. They were warned of a forthcoming memory test
on the content of the story, but there was no such test. Rather, the
instructions and distractor task were designed to focus listeners’ attention
away from the music. Participants heard one of the five strings of tone
sequences in their unattended (left) ear. This type of incidental exposure is
similar to when one is engaged in conversation but nonetheless aware of
music playing in the background (e.g., at a party).
In the second phase of the experiment, participants rated how much they
liked 12 of the 18 sequences using a 7-point scale (1 ⫽ dislike extremely,
7 ⫽ like extremely). The set included the 6 sequences heard in the exposure
phase along with 6 additional sequences selected randomly from the 12
Figure 1. Examples of tone sequences used in Experiment 1. There were
six different sequences for each different length (5, 6, 7, 8, or 9 tones).
372
SZPUNAR, SCHELLENBERG, AND PLINER
unheard sequences (to provide a baseline measure of liking). The 12 stimuli
were presented in a different random order for each participant.
In the third phase, participants rated how confident they were that each
of the 18 target sequences had been presented in the exposure phase.
Ratings were again made on a 7-point scale (1 ⫽ extremely sure not
presented, 7 ⫽ extremely sure presented). The excerpts included the 6
sequences from the exposure phase, the 6 sequences that were added to the
second (liking ratings) phase, and an additional 6 sequences that had never
been heard before (as a baseline measure of memory). The order of the 18
sequences was randomized separately for each participant. The entire
procedure lasted approximately 20 min for participants in focused listening
and approximately 10 min for participants in incidental listening. The
difference in duration between the two conditions resulted from the fact
that participants in the focused group were required to make a response
after hearing each sequence and those in the incidental group were not.
Results and Discussion
Liking. For each listener, we calculated average liking ratings
for excerpts heard 64 times, 16 times, and 4 times during the
exposure phase (two excerpts in each case). We also calculated an
average liking rating for the six excerpts heard for the first time in
Phase 2, which served as a baseline measure of liking. Means are
illustrated in Figure 2. We used trend analysis to examine changes
in liking ratings as a function of exposure. Effect sizes are reported
in Table 1. We observed a near-significant interaction between
listening condition and a linear trend in ratings, F(1, 48) ⫽ 3.63,
p ⫽ .063, but no interaction between condition and a quadratic
trend.
In the case of focused listening, both linear and quadratic trends
were nonsignificant. In fact, mean liking ratings varied minimally,
from a low of 3.72 (four exposures) to a high of 3.85 (baseline). No
pair of means differed significantly.
For incidental listening, the linear trend in liking ratings was
reliable, F(1, 19) ⫽ 13.81, p ⫽ .001, but the quadratic trend was
not. Comparisons of adjacent means revealed no reliable differ-
ences. Nonetheless, liking ratings were significantly above base-
line levels for sequences heard 16 and 64 times, t(19) ⫽ 2.56, p ⫽
.019, and t(19) ⫽ 4.20, p ⬍ .001, respectively (see Figure 2). In
addition, liking ratings increased reliably from 4 to 64 exposures,
t(19) ⫽ 2.35, p ⫽ .030. Thus, incidental listening led to small but
significant increases in liking as a function of exposure to these
impoverished stimuli, but focused listening did not.
Memory. For each listener, we calculated average memory
ratings for sequences presented 4 times, 16 times, and 64 times.
We also calculated average ratings for the six sequences heard
once in the second (liking ratings) phase and the six sequences
heard for the first time in the third phase (baseline measure). Mean
ratings are illustrated in Figure 2.
Trend analysis revealed a robust interaction between listening
condition and linear increases in recognition as a function of
exposure, F(1, 48) ⫽ 261.37, p ⬍ .001. There was no interaction
between condition and a quadratic trend. Effect sizes for linear and
quadratic trends are provided in Table 1 separately for focused and
incidental listening.
For focused listening, the linear trend in recognition confidence
was dramatic, F(1, 29) ⫽ 574.47, p ⬍ .001, as shown in Figure 2
and Table 1. On a 7-point scale, memory ratings varied from a low
of 1.72 (baseline) to a high of 6.70 (64 exposures). Each pair of
adjacent means differed significantly. Specifically, recognition
confidence was higher for sequences heard once in the second
phase than for baseline stimuli, t(29) ⫽ 4.38, p ⬍ .001. In other
words, the former sequences were falsely recognized as occurring
in the exposure phase. Recognition confidence was also higher for
sequences heard 4 times than for those heard once, t(29) ⫽ 8.24,
p ⬍ .001; for those with 16 relative to 4 exposures, t(29) ⫽ 4.02,
p ⬍ .001; and for those with 64 relative to 16 exposures, t(29) ⫽
Figure 2. Mean liking and memory ratings in Experiment 1 as a function
of exposure frequency and listening condition. Error bars represent stan-
dard errors.
Table 1
Effect Sizes (
2
) for Linear and Quadratic Trends in Liking and
Memory Ratings as a Function of Exposure
Experiment
Focused listening Incidental listening
Linear Quadratic Linear Quadratic
Liking ratings
1 ⬍.001 .013 .421 .021
2 .002 .467 .632 .001
3 .001 .206 .527 .003
Memory ratings
1 .952 .269 .105 .038
2 .934 .209 .414 .070
3 .932 .433 .345 .048
Note. Significant effects are italicized.
373
REPEATED EXPOSURE TO MUSIC
2.33, p ⫽ .027. The quadratic trend was significant as well, F(1,
29) ⫽ 10.65, p ⫽ .003, because memory ratings reached near-
ceiling levels after 16 exposures (M ⫽ 6.43).
For incidental listening, there was no linear or quadratic trend in
recognition confidence. Memory ratings varied minimally, from a
low of 3.83 (baseline) to a high of 4.38 (16 exposures). No pair of
means differed reliably. In fact, the largest difference between
means approached conventional levels of significance, t(19) ⫽
1.90, p ⫽ .073. Thus, these listeners had virtually no explicit
memory for the sequences presented in the exposure phase.
Summary. Supraliminal presentation of these tone se-
quences yielded findings consistent with Bornstein’s (1989)
conclusion that the mere exposure effect is stronger in the
absence of explicit memory. Our data provided no evidence of
satiation, but they revealed completely different patterns of
exposure effects for liking and explicit memory in the two
listening contexts. When listeners’ attention was focused on our
impoverished stimuli, repeated presentation resulted in im-
proved recognition but did not affect listeners’ evaluation of the
stimuli. By contrast, incidental listening to the same stimuli
produced reliable increases in liking as a function of exposure
but no explicit memory for the stimuli. These data are consis-
tent with the perceptual fluency/attributional account of mere
exposure effects, in which recognized and unrecognized stimuli
have different attributional outcomes with correspondingly dif-
ferent affective responses.
Experiment 2
In Experiment 2, our stimuli were 15-s excerpts from commer-
cial recordings of orchestral music. Thus, the stimuli were more
complex than those of Experiment 1. Each excerpt contained
multiple instruments playing different parts, and some of the
instruments (e.g., keyboard instruments) played multiple tones.
Excerpts were selected from pieces that were harmonically and
rhythmically well defined (i.e., with a clear key and meter) with
expressive timing and phrasing. A major advantage of these stim-
uli was their ecological validity, but a disadvantage was the mul-
tiple dimensions of difference among excerpts (e.g., orchestration,
tempo, and pitch range). Nevertheless, stimulus randomization
eliminated potential threats to internal validity. In addition, the
task for focused listeners was a musical one that required identi-
fication of the lead instrument (i.e., the one playing the melody) in
pieces that had several instruments playing simultaneously. In
short, the stimuli and orienting task were ecologically valid in
Experiment 2 but not in Experiment 1.
In line with Bornstein (1989), we predicted that exposure effects
on liking (both increases and decreases) would be stronger for
these rich, complex stimuli than for the impoverished stimuli in
Experiment 1. In particular, we expected focused listening to
enhance memory for the stimuli and to increase the likelihood of
satiation.
Method
Participants. The participants were 40 undergraduates who were re-
cruited as in Experiment 1.
Apparatus and stimuli. The apparatus was the same as that of Exper-
iment 1. The stimuli were 18 excerpts of 15 s each taken from recordings
available on compact disk (Table 2), 6 each from pieces written in Ba-
roque, Classical, or Romantic/Late-Romantic styles. For the most part, we
selected concerti to permit the isolation of portions featuring a lead instru-
ment. Each excerpt had one of six different lead instruments. There were
equal numbers of each lead instrument (i.e., three), but style and lead
instrument were partly confounded because musical period and instrumen-
tation are not independent. All excerpts were rerecorded from compact
disks onto the computer, saved as digital sound files (CD quality), and
normalized to hold peak amplitude constant across excerpts. The distract-
ing stimulus was the same as in Experiment 1.
Table 2
Pieces From Which Stimuli Were Excerpted in Experiment 2
Style and composer Composition Lead instrument
Baroque
Bach Brandenburg Concerto No. 1 in F Horn
Bach Concerto in A for Oboe Oboe
Bach Flute Concerto in E Minor Flute
Handel Oboe Concerto No. 1 in B Flat Oboe
Vivaldi Cello Concerto in G Major Cello
Vivaldi Oboe Concerto in C Major Oboe
Classical
Beethoven Piano Concerto No. 1 in C Major, Op. 15 Piano
Beethoven Violin Concerto in D, Op. 61 Violin
Haydn Cello Concerto No. 1 in C Cello
Haydn Cello Concerto No. 2 in D Cello
Mozart Horn Concerto No. 1 in D, K. 412 Horn
Mozart Horn Concerto No. 3 in E Flat, K. 447 Horn
Romantic/Late Romantic
Faure´ En Priere for Flute and Orchestra Flute
Faure´ Pavane for Flute and Orchestra Flute
Schubert String Quartet No. 10 in E Flat Major, D87 Violin
Schubert String Quartet No. 13 in A Minor, D804 Violin
Tchaikovsky Piano Concerto No. 2 in D Minor, Op. 23 Piano
Tchaikovsky Piano Concerto No. 3 in E Flat, Op. 75 Piano
374
SZPUNAR, SCHELLENBERG, AND PLINER
Procedure. Half of the participants were assigned to focused listening
and half to incidental listening. The procedure was identical to that of
Experiment 1 with the following exceptions.
Focused listeners heard 84 excerpts of 15 s each of classical music
during the exposure phase. Specifically, they heard 6 different excerpts
selected randomly from the set of 18 (constrained such that each of the six
lead instruments was featured in one excerpt), 2 of which were presented
twice, 8 times, or 32 times. The orienting task was to identify the lead
instrument in each excerpt. Participants were first familiarized with each of
the six target instruments (cello, flute, horn, oboe, piano, and violin). The
purpose of the task was to ensure that participants listened closely on each
trial.
Incidental listeners also heard 84 excerpts in the exposure phase, with 6
excerpts presented multiple times (2, 8, or 32 times), as for focused
listeners. The excerpts were presented consecutively, separated by 500 ms
of silence. There were five different orders of the 84 excerpts, with each
order randomized and constrained as for focused listeners. Participants
heard one of the five strings of musical excerpts at reduced amplitude in
their unattended ear.
In the second phase of the experiment, listeners rated how much they
liked each of 12 excerpts, the 6 heard in Phase 1 along with an additional
6 selected randomly from the 12 unheard excerpts, also constrained so that
each of the novel excerpts had a different lead instrument. In the third
phase, participants rated their recognition confidence for the 6 excerpts
from the exposure phase, the 6 that were added to Phase 2, and the 6
remaining (novel) excerpts. The entire procedure lasted approximately 40
min for participants engaged in focused listening and approximately 30
min for participants engaged in incidental listening.
Results and Discussion
As in Experiment 1, we calculated four average liking ratings
and five average memory ratings for each participant (see
Figure 3).
Liking. As illustrated in Figure 3, liking ratings formed an
inverted U-shaped function of exposure for focused listening but a
linear function for incidental listening. Trend analysis confirmed
that both linear and quadratic trends interacted with listening
condition, F(1, 38) ⫽ 7.51, p ⫽ .004, and F(1, 38) ⫽ 12.88, p ⬍
.001, respectively. Effect sizes are provided in Table 1.
For focused listening, the quadratic trend was significant, F(1,
19) ⫽ 16.64, p ⬍ .001, but the linear trend was not. Ratings
increased from baseline to 2 exposures, t(19) ⫽ 2.52, p ⫽ .021,
and increased again marginally from 2 to 8 exposures, t(19) ⫽
1.91, p ⫽ .072; however, they decreased from 8 to 32 exposures,
t(19) ⫽ 2.93, p ⫽ .009. After 32 exposures, satiation was so
complete that benefits of exposure were no longer apparent. In
fact, ratings for these excerpts did not differ from baseline ratings.
For incidental listening, the linear trend was significant, F(1,
19) ⫽ 32.59, p ⬍ .001, but the quadratic trend was not. Liking
ratings were a monotonic function of exposure, increasing reliably
from baseline to 2 exposures, t(19) ⫽ 2.53, p ⫽ .021; from 2 to 8
exposures, t(19) ⫽ 2.49, p ⫽ .022; and from 8 to 32 exposures,
t(19) ⫽ 2.20, p ⫽ .040.
Memory. Trend analysis revealed a robust interaction between
listening condition and linear improvements in recognition as a
function of exposure, F(1, 38) ⫽ 102.77, p ⬍ .001. The interaction
between condition and quadratic trend was not significant. As
shown in Figure 3, recognition confidence increased with exposure
in both conditions, although the linear trend was much stronger for
focused listening, F(1, 19) ⫽ 268.10, p ⬍ .001, than for incidental
listening, F(1, 19) ⫽ 13.42, p ⫽ .002. Effect sizes for linear and
quadratic trends are reported in Table 1.
For focused listening, average ratings varied from below 2 to
almost 7. Recognition confidence was greater for stimuli presented
in the second phase relative to baseline, t(19) ⫽ 2.90, p ⫽ .009,
reflecting participants’ false belief that the stimuli had appeared in
the exposure phase. Recognition confidence was also greater for
stimuli heard twice than for those heard once, t(19) ⫽ 7.72, p ⬍
.001; greater for stimuli heard 8 times than for those heard twice,
t(19) ⫽ 4.16, p ⬍ .001; and marginally greater for stimuli heard 32
rather than 8 times, t(19) ⫽ 2.02, p ⫽ .058. There was a modest but
significant quadratic trend, F(1, 19) ⫽ 5.03, p ⫽ .037, that arose
from memory ratings approaching ceiling (M ⫽ 6.55) after 8
exposures.
For incidental listening, ratings varied from 4 to 5. The only
significant difference between adjacent means occurred for ex-
cerpts presented once in the second phase relative to baseline,
t(19) ⫽ 3.44, p ⫽ .003 (i.e., “false” recognition). Means for
excerpts heard 2, 8, or 32 times in the exposure phase did not
differ, but recognition confidence exceeded baseline levels in each
case, ts(19) ⫽ 2.35, 4.07, and 3.55, respectively, ps ⬍ .03. Rec-
ognition confidence was also higher for excerpts presented 8 or 32
times relative to those falsely recognized, ts(19) ⫽ 2.54 and 2.33,
Figure 3. Mean liking and memory ratings in Experiment 2 as a function
of exposure frequency and listening condition. Error bars represent stan-
dard errors.
375
REPEATED EXPOSURE TO MUSIC
respectively, ps ⬍ .04. There was no quadratic trend. Although the
significant linear trend confirmed greater recognition confidence
with increasing exposure, the changes were small relative to fo-
cused listening.
Summary. For focused listening, memory ratings increased
dramatically and monotonically as a function of exposure. Liking
ratings increased initially but subsequently decreased, revealing
reliable satiation for these ecologically valid stimuli. This pattern
for the liking data is in line with predictions from the two-factor
model, which posits increases in positive affect with increasing
exposure until the effect of boredom supersedes the dissipation of
neophobia. The two-factor model has no prediction about the
memory data, but memory was high when liking was low, which
is consistent with the perceptual fluency/attributional model.
For incidental listening, memory ratings increased modestly
with exposure. Liking ratings also increased with increasing ex-
posure, in line with the mere exposure effect, albeit without any
evidence of satiation (as in Experiment 1). Contrary to predictions
from the perceptual fluency/attributional model, increases in rec-
ognition accompanied increases in liking.
Experiment 3
Recognition ratings were similar in Experiments 1 and 2, with
recognition confidence attenuated greatly for listeners who heard
the stimuli incidentally. Striking differences were evident, how-
ever, for liking ratings. These differences were most likely due to
differences in the ecological validity of the stimuli and methods,
which would have made the listening experience more complex
and interesting in Experiment 2 than in Experiment 1. Observed
differences in liking on the basis of complexity are consistent with
Bornstein’s (1989) review of the literature and with predictions
motivated by the two-factor model (Berlyne, 1970; Stang, 1974).
Nonetheless, the experiments differed in other aspects that could
have been the source of differential responding, at least in princi-
ple. For example, the stimuli varied in duration in Experiment 1
but not in Experiment 2. Other differences included the overall
duration of testing sessions, the duration of the interval between
consecutive stimuli in the incidental conditions, and the number of
presentations in the exposure phase.
The present experiment was designed so that the stimuli and
procedure would be intermediate in ecological validity between
those of Experiments 1 and 2. Accordingly, we predicted that
response patterns for liking ratings would be midway between
those of Experiments 1 and 2. Such a finding would confirm that
differences among the three experiments are best attributed to
differences in ecological validity and stimulus complexity.
Specifically, the stimuli consisted of simple monophonic tone
sequences (as in Experiment 1), but all tone sequences were of
equal duration (as in Experiment 2). We made the stimulus set as
a whole more heterogeneous (and more like Experiment 2) by
presenting the sequences on six different musical instruments. In
addition, the orienting task for focused listeners was musically
relevant, being identical to that of Experiment 2 (i.e., identify the
particular musical instrument).
Because the tone sequences of the present experiment were
shorter in duration than the orchestral excerpts of Experiment 2, it
was impossible to equate the two experiments on total duration of
the testing session, number of presentations in the exposure phase,
and duration of the interval between stimuli in the exposure phase
of the incidental conditions. Moreover, increasing the number of
presentations beyond that used in Experiment 1 (maximum of 64)
could surpass the threshold of what listeners can tolerate in a single
testing session. As a compromise, we made the interval between
stimuli in the incidental condition twice as long as it was in
Experiment 2, which reduced the difference in total duration
between Experiment 2 and the present experiment. If the interval
between stimuli was the source of the differences between liking
ratings in Experiments 1 and 2, differences between the present
experiment and Experiment 1 should be exaggerated (in relation to
Experiment 2) rather than attenuated (as predicted).
Method
Participants. The participants were 60 undergraduates who were re-
cruited as in Experiments 1 and 2.
Apparatus and stimuli. The apparatus was the same as in Experiments
1 and 2. The stimuli were 18 of the 30 tone sequences from Experiment 1.
To produce 18 sequences of equal duration (i.e., seven tones), we left the
seven-tone sequences unchanged and deleted the last tone of the eight-tone
sequences and the last two tones of the nine-tone sequences. In contrast to
Experiment 1 (i.e., all stimuli played on piano), the stimuli were divided
into six sets of 3 sequences, with each set played on a different musical
instrument (cello, flute, horn, oboe, piano, or violin). The set of six
instruments was the same as the set of lead instruments from Experiment
2. The distracting stimulus in the incidental-listening condition was the
same as in Experiments 1 and 2.
Procedure. Half of the participants were assigned to focused listening
and half to incidental listening. The procedure was identical to Experiment
2 except that the tone sequences were substituted for the orchestral excerpts
and the sequences were presented 4, 16, or 64 times in the exposure phase,
separated by1sofsilence in the incidental-listening condition. The entire
procedure lasted approximately 20 min for focused listeners and 15 min for
incidental listeners.
Results and Discussion
As in Experiments 1 and 2, we calculated four average liking
ratings and five average memory ratings for each participant (see
Figure 4).
Liking. Effect sizes derived from a trend analysis are reported
in Table 1. The interaction between listening condition and linear
increases in liking rating was reliable, F(1, 58) ⫽ 11.20, p ⫽ .001,
as was the interaction between condition and a quadratic trend,
F(1, 58) ⫽ 5.53, p ⫽ .022.
For focused listening, the quadratic trend was significant, F(1,
29) ⫽ 7.52, p ⫽ .010, but the linear trend was not. As shown in
Figure 4, liking ratings formed an inverted U-shaped function of
exposure, although the function peaked earlier (at four exposures)
and was less exaggerated than in Experiment 2 (see Table 1). For
example, the difference between the lowest and the highest mean
rating (3.54 and 4.07, respectively) was less than half the magni-
tude of the same comparison in Experiment 2 (3.50 and 4.65; see
Figure 3). Comparisons between adjacent means confirmed that
the effect was weak in relation to that in Experiment 2. In fact, the
difference between the lowest (baseline) and highest (4 exposures)
means was marginal, t(29) ⫽ 1.96, p ⫽ .060. Although increases
in liking from baseline to 16 exposures were reliable, t(29) ⫽ 2.42,
p ⫽ .022 (as a result of decreased variance), no other difference
between means was significant.
376
SZPUNAR, SCHELLENBERG, AND PLINER
For incidental listening, the linear trend in liking ratings was
significant, F(1, 29) ⫽ 32.27, p ⬍ .001, but there was no quadratic
trend. Comparisons between adjacent means revealed that liking
ratings were higher for sequences heard 4 times relative to baseline
levels, t(29) ⫽ 2.31, p ⫽ .028, and higher for sequences heard 64
times than for those heard 16 times, t(29) ⫽ 2.92, p ⫽ .007. Only
the increase in liking from 4 to 16 exposures was not significant.
To confirm that the change in stimuli and methods influenced
response patterns, we conducted a combined analysis of the data
from Experiments 1 and 3, with “experiment” as one of the
independent variables. The main effect of experiment was not
significant, nor was its interaction with listening condition. None-
theless, there was a three-way interaction involving experiment,
listening condition, and quadratic changes in liking as a function of
exposure, F(1, 106) ⫽ 5.22, p ⫽ .024. The two experiments
produced similar response patterns for incidental listening (i.e.,
significant linear increases in liking as a function of exposure) but
not for focused listening. In the present experiment, the focused
listeners exhibited a small but significant quadratic trend in liking
ratings that was not evident for their counterparts in Experiment 1.
Memory. Trend analysis revealed a robust interaction between
listening condition and linear increases in recognition as a function
of exposure, F(1, 58) ⫽ 80.32, p ⬍ .001. The interaction between
condition and a quadratic trend, although weaker, was also signif-
icant, F(1, 58) ⫽ 6.66, p ⫽ .012. Effect sizes are reported in
Table 1.
For focused listening, the linear trend in recognition confidence
was dramatic, F(1, 29) ⫽ 396.54, p ⬍ .001, as it was in Experi-
ments 1 and 2 (see Figure 4). Memory ratings varied from a low
of 2.07 to a high of 6.52. As before, each pair of adjacent means
differed significantly, with the exception of excerpts heard 16 and
64 times. Recognition confidence was higher (a) for stimuli heard
once in Phase 2 than for baseline stimuli, t(29) ⫽ 6.29, p ⬍ .001
(i.e., “false” recognition); (b) for stimuli heard 4 times than for
those heard once, t(29) ⫽ 7.90, p ⬍ .001; and (c) for those heard
16 times than for those heard 4 times, t(29) ⫽ 4.23, p ⬍ .001. The
quadratic trend was also significant, F(1, 29) ⫽ 22.15, p ⬍ .001,
as it was in Experiments 1 and 2, because memory ratings reached
ceiling levels after 16 exposures (M ⫽ 6.48).
For incidental listening, recognition confidence increased lin-
early, F(1, 29) ⫽ 15.27, p ⬍ .001, although the increase was
greatly attenuated relative to focused listening, as it was in Exper-
iments 1 and 2. The quadratic trend was not reliable. There were
no reliable differences between adjacent means, although the dif-
ference in recognition confidence between sequences heard 4
times and those heard once in the liking phase was marginal,
t(29) ⫽ 1.90, p ⫽ .068. Recognition confidence also exceeded
baseline levels for excerpts presented 4, 16, and 64 times, respec-
tively, t(29) ⫽ 2.49, p ⫽ .019; t(29) ⫽ 4.36, p ⬍ .001; and t(29) ⫽
3.61, p ⫽ .001. Overall, the means varied over a relatively narrow
range, from a low of 3.56 (baseline) to a high of 4.75 (16
exposures).
Joint analysis of the present memory data and those from
Experiment 1 confirmed that there was no main effect of experi-
ment and no interactions involving differences between experi-
ments, with one exception. There was a reliable three-way inter-
action involving experiment, listening condition, and linear
increases in memory as a function of exposure, F(1, 106) ⫽ 6.88,
p ⫽ .010. This finding confirms that response patterns differed for
memory ratings across the two experiments. Specifically, the two-
way interaction between listening condition and linear increases in
recognition confidence was stronger in Experiment 1 than in
Experiment 3. In Experiment 1, the linear trend was significant in
the focused condition but not in the incidental condition. In the
present experiment, although the linear trend was relatively weak
in the incidental condition, it was significant in both conditions, as
it was in Experiment 2. Presumably, the homogeneous stimulus set
of Experiment 1 (i.e., all sequences played on piano) made it
difficult for listeners in the incidental condition to discriminate
novel from previously heard sequences, whereas the heteroge-
neous set of the present experiment facilitated discrimination.
Summary. As predicted, response patterns for liking ratings
were midway between those observed in Experiments 1 and 2,
which provides support for the hypothesis that response patterns
are influenced by the ecological validity of the stimuli and method.
Specifically, liking ratings in the focused condition of Experiment
1 did not vary as a function of number of exposures, whereas
ratings in Experiment 2 formed an inverted U-shaped function—as
predicted by the two-factor model—with an initial increase fol-
lowed by a decrease. In the present experiment, ratings from
listeners in the focused condition also formed an inverted U, but
the increase and subsequent decrease were smaller than those
Figure 4. Mean liking and memory ratings in Experiment 3 as a function
of exposure frequency and listening condition. Error bars represent stan-
dard errors.
377
REPEATED EXPOSURE TO MUSIC
witnessed in Experiment 2. In fact, the effect size of the quadratic
trend was less than half the magnitude of the effect in Experiment
2 (see Table 1). The difference between experiments in the time
course of the satiation effect was also consistent with the two-
factor model’s prediction that the entire liking trajectory (i.e.,
increases and decreases as a function of exposure) should be more
rapid for simple than for complex stimuli.
As in Experiments 1 and 2, liking data in the incidental condi-
tions exhibited increases with increasing exposure with no evi-
dence of satiation, in line with the original mere exposure hypoth-
esis. More important, changes in liking ratings were midway
between those of Experiments 1 and 2: The function was weakest
in Experiment 1, strongest in Experiment 2, and intermediate in the
present experiment. For example, no adjacent means differed sig-
nificantly in Experiment 1, two of three comparisons were signif-
icant in the present experiment, and all comparisons were signif-
icant in Experiment 2. Moreover, the effect size of the linear trend
in the present experiment was midway between effect sizes from
Experiments 1 and 2 (see Table 1).
For the most part, recognition ratings were similar to those from
Experiments 1 and 2. For focused listeners, recognition confidence
increased dramatically as a function of exposure. For incidental
listeners, recognition confidence was consistently attenuated. Lin-
ear increases in explicit memory as a function of incidental expo-
sure were reliable in the present experiment and in Experiment 2
but not in Experiment 1. The effect sizes reported in Table 1
suggest that the association between recognition and incidental
exposure becomes stronger as the to-be-remembered stimuli in-
crease in complexity and ecological validity.
General Discussion
We examined liking and memory for musical stimuli as a
function of number of previous exposures. Our findings are nota-
ble in at least four respects. First, the effect of exposure on
preference varied with the ecological validity of the stimuli. Sec-
ond, ecological validity also affected how liking and memory
covaried as a function of exposure. Third, high levels of recogni-
tion confidence were associated with lower evaluations of the
stimuli. Finally, we found evidence of reliable satiation effects in
the laboratory that mirror real-life experience.
In Experiment 1, focused listening led to liking ratings that were
independent of exposure but to memory ratings that increased
monotonically with increasing exposure (see Figure 2). By con-
trast, incidental listening led to liking ratings that increased mono-
tonically with exposure but to memory ratings that were indepen-
dent of exposure. These response patterns for our supraliminal
stimuli mirror those reported with subliminal exposure (Zajonc,
2001). They also provide support for the proposal that listeners’
evaluation of musical stimuli is independent of explicit memory
(Peretz, Gaudreau, & Bonnel, 1998). Indeed, the findings are
consistent with neuropsychological evidence of general dissocia-
tions between cognitive and affective responses to music. For
example, a patient tested by Peretz and her colleagues could
classify happy-sounding and sad-sounding tunes correctly, but she
could not recognize familiar tunes or discriminate gross changes in
pitch contour (Peretz, 2001; Peretz & Gagnon, 1999; Peretz, Gag-
non, & Bouchard, 1998).
Patterns of evaluation and recognition were quite different in
Experiments 2 and 3. After focused listening, liking ratings formed
an inverted U-shaped function of exposure, but memory ratings
increased linearly (see Figures 3 and 4). After incidental listening,
liking and memory ratings increased linearly with exposure, al-
though increases in recognition confidence were much smaller
than they were for focused listening. Of particular interest was the
attenuation of liking in the context of high levels of recognition
confidence. This finding was also evident among focused listeners
in Experiment 1; low but not high levels of recognition confidence
were accompanied by increases in liking. These results are con-
sistent with predictions from the perceptual fluency/attributional
model but inconsistent with proposals of independence between
cognitive and affective responding (Peretz, Gaudreau, & Bonnel,
1998; Zajonc, 1980, 2001).
Other findings contradict the perceptual fluency/attributional
model. The model predicts decreases in stimulus evaluation with
increasing recognition accuracy, but in listening contexts with
ecological validity (Experiments 2 and 3), increased exposure led
to increases in both liking and recognition. In fact, in a context
with intermediate ecological validity (Experiment 3), as few as
four exposures led to an increase in recognition and a subtle
increase in liking. In our most ecologically valid stimulus contexts
(Experiment 2), as few as two exposures led to enhanced memory
and evaluation relative to novel stimuli. Similarly, our incidental
task in Experiments 2 and 3 produced monotonic increases in
liking as a function of exposure frequency as well as small but
reliable increases in recognition confidence. Thus, in ecologically
valid stimulus contexts, repetition led to greater liking as well as
increased recognition. The negative association between liking and
memory predicted by the perceptual fluency/attributional model
became evident only when recognition confidence was extremely
high.
To date, robust satiation effects such as the one discovered in
Experiment 2 have been found infrequently. For focused listeners,
liking ratings increased reliably from baseline to 8 exposures,
falling to baseline levels after 32 exposures. Similar but smaller
increases and decreases in liking were evident in Experiment 3.
These findings are unique in providing unequivocal evidence of
satiation for musical stimuli in a controlled experimental setting.
The robust satiation effect in Experiment 2 cannot be attributed
simply to amount of exposure. Participants in Experiments 1 and
3 had twice as many exposures as did those in Experiment 2, but
they showed either no evidence of satiation (Experiment 1) or a
smaller effect (Experiment 3). Our data indicate that satiation
effects are more likely to result from repeated exposures to com-
plex, ecologically valid stimuli than from repetition of simple
stimuli. Why should this be the case? One possibility is that the
degree to which liking initially increases is the best predictor of
subsequent decreases. In the present study, the largest amount of
satiation was evident after the largest initial increases in liking
(i.e., from 0 to 8 exposures, Experiment 2, focused listening). In
other words, complex stimuli may engender relatively large in-
creases in liking as a function of exposure, and substantial in-
creases in liking may be a prerequisite for substantial decreases.
Recent data from the visual domain provide evidence consistent
with this perspective. Event-related brain potentials reveal that
perceptual encoding of “affect-laden” stimuli (i.e., those likely to
induce affective responding) is facilitated relative to encoding of
378
SZPUNAR, SCHELLENBERG, AND PLINER
neutral stimuli (Schupp, Jungho¨fer, Weike, & Hamm, 2003). On
the one hand, then, perception of stimuli with affective potential
may be inherently fluent relative to neutral stimuli. On the other
hand, perceptual fluency increases affective responding. In other
words, the association between affect and exposure may be some-
what circular. In the present study, it seems safe to assume that
stimuli with greater ecological validity were also the most affect
laden, which would have exaggerated fluency effects of exposure
and increased the possibility of greater increases and decreases in
liking. Presumably, our “real music” stimuli of Experiment 2 were
intrinsically more likeable than the somewhat impoverished stim-
uli of Experiment 3, which were more likeable than the very
impoverished stimuli of Experiment 1.
1
This also means that liking
as a function of exposure in the experiments would have proceeded
from different baseline levels. In most of the relevant research
(Bornstein, 1989; Zajonc, 2001), stimuli are initially neutral with
respect to affect before the exposure phase. Interestingly, in one
instance in which aesthetically pleasing stimuli (artworks) were
presented (Zajonc et al., 1972), satiation effects were similarly
robust to the one observed in Experiment 2.
Conclusions about differential responding based on stimulus
complexity and ecological validity remain somewhat speculative,
however, until further data are collected. Stimuli for our three
experiments differed in multiple ways (e.g., timbre, texture, typi-
cality, and aesthetic value) that may be separate from, or interact
with, overall complexity. Any combination of these variables
could have contributed to differences in responding among
experiments.
Regardless, our results highlight the strengths and weaknesses
of the various theoretical perspectives that purport to explain
associations between exposure and liking. Zajonc’s (2001) account
is not intended to account for supraliminally presented stimuli that
may be remembered as well as liked or disliked. Both the percep-
tual fluency/attributional model and the two-factor model account
for some of the data reported here, yet neither explains all of the
results adequately. To summarize, the perceptual fluency/attribu-
tional model accounts for dissociations between liking and explicit
memory as well as for all cases of reduced liking that were
accompanied by increases in recognition confidence. It fails, how-
ever, to account for increases in liking that accompanied increases
in recognition confidence. The two-factor model accounts for the
satiation effects we observed, as well as for the increases in
recognition confidence and liking that occurred simultaneously. It
fails, however, to account for contexts in which liking and memory
were independent. Thus, a hybrid of the two approaches (see also
Bornstein, 1989) may provide the best account of associations
among exposure, liking, and memory.
Our proposed hybrid model makes a distinction between im-
poverished, simple, and affectively neutral stimuli, on the one
hand, and ecologically valid, rich, and affect-laden stimuli, on the
other hand. When the former are presented subliminally or supra-
liminally, previous exposure to such stimuli increases their per-
ceptual fluency. In the absence of explicit memory for the stimuli,
participants are likely to misattribute this fluency to any dimension
introduced by the experimenter (e.g., liking or brightness). When
they remember the stimuli explicitly, however, they have an ade-
quate explanation available for fluency effects, which they at-
tribute correctly to previous exposure. For ecologically valid stim-
uli that may have preexisting affective status, increased exposure
initially corresponds with relatively large increases in liking as
well as improved memory. Presumably, part of this exposure
process includes increasing familiarity with the higher order struc-
ture of the stimuli, which allows the perceiver to appreciate their
structural complexity. As exposure increases to the point at which
the perceiver has nothing left to learn, boredom sets in and satia-
tion begins.
A central goal of the present study was to improve our under-
standing of how repeated exposure to rich and meaningful stimuli
influences how they are liked and remembered. Our results reveal
that controlled laboratory experiments can produce findings that
are consistent with the increases and decreases in liking for music
that characterize phenomenological experience. Our findings also
point to ways in which theories of exposure and liking might be
modified and improved. A good theory should be able to explain
responses to impoverished, highly controlled, affectively neutral
stimuli, as well as responses to complex, ecologically valid, and
affect-laden stimuli that people encounter regularly in their daily
lives.
Our interest in effects of repeated exposure to complex stimuli
in general, and to artworks in particular, parallels the approach
taken by Cutting (2003). Cutting found that mere exposure effects
can explain how a canon of classic French Impressionist paintings
is promoted and maintained. Several years earlier, Berlyne noted
that the goal of his new experimental aesthetics was “not only to
throw light on aesthetic phenomena but, through the elucidation of
aesthetic problems, to throw light on human psychology in gen-
eral” (1974, p. 5). Thirty years later, we believe that this goal
remains laudable. Future studies of exposure and liking for com-
plex, real-world stimuli are bound to improve our understanding of
associations among exposure, liking, and memory that are central
to human experience.
1
As illustrated in Figures 2, 3, and 4, differences across experiments in
the absolute magnitude of liking ratings were relatively small. Presumably,
ratings were context dependent and made in relation to the particular
stimulus set as a whole. We would expect larger differences if the same
participants heard tone sequences (Experiments 1 and 3) and orchestral
excerpts (Experiment 2).
References
Barchas, P. R., & Perlaki, K. M. (1986). Processing of preconsciously
acquired information measured by hemispheric asymmetry and selection
accuracy. Behavioral Neuroscience, 100, 343–349.
Bartlett, D. L. (1973). Effect of repeated listenings on structural discrim-
ination and affective response. Journal of Research in Music Education,
21, 302–317.
Berlyne, D. E. (1970). Novelty, complexity, and hedonic value. Perception
& Psychophysics, 8, 279–286.
Berlyne, D. E. (1974). The new experimental aesthetics. In D. E. Berlyne
(Ed.), Studies in the new experimental aesthetics (pp. 1–25). Washing-
ton, DC: Hemisphere.
Bonanno, G. A., & Stillings, N. A. (1986). Preference, familiarity, and
recognition after repeated brief exposures to random geometric shapes.
American Journal of Psychology, 99, 403–415.
Bornstein, R. F. (1989). Exposure and affect: Overview and meta-analysis
of research, 1968–1987. Psychological Bulletin, 106, 265–289.
Bornstein, R. F. (1992). Inhibitory effects of awareness on affective re-
379
REPEATED EXPOSURE TO MUSIC
sponding. In M. S. Clark (Ed.), Emotion: Review of personality and
social psychology (No. 13, pp. 235–255). Thousand Oaks, CA: Sage.
Bornstein, R. F., & D’Agostino, P. R. (1992). Stimulus recognition and the
mere exposure effect. Journal of Personality and Social Psychology, 63,
545–552.
Bornstein, R. F., & D’Agostino, P. R. (1994). The attribution and discount-
ing of perceptual fluency: Preliminary tests of a perceptual fluency/
attributional model of the mere exposure effect. Social Cognition, 12,
103–128.
Bornstein, R. F., Kale, A. R., & Cornell, K. R. (1990). Boredom as a
limiting condition on the mere exposure effect. Journal of Personality
and Social Psychology, 58, 791–800.
Bornstein, R. F., Leone, D. R., & Galley, D. J. (1987). The generalizability
of subliminal mere exposure effects: Influence of stimuli perceived
without awareness on social behavior. Journal of Personality and Social
Psychology, 53, 1070–1079.
Bradley, I. L. (1971). Repetition as a factor in the development of musical
preferences. Journal of Research in Music Education, 19, 295–298.
Brentar, J. E., Neuendorf, K. A., & Armstrong, G. B. (1994). Exposure
effects and affective responses to music. Communication Monographs,
61, 161–181.
Brickman, P., & D’Amato, B. (1975). Exposure effects in a free choice
situation. Journal of Personality and Social Psychology, 32, 415–420.
Cohen, J. D., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope:
A new graphic interactive environment for designing psychology exper-
iments. Behavior Research Methods, Instruments, & Computers, 25,
257–271.
Cutting, J. E. (2003). Gustave Caillebotte, French Impressionism, and mere
exposure. Psychonomic Bulletin & Review, 10, 319–343.
Downey, J. E., & Knapp, G. E. (1927). The effect on a musical programme
of familiarity and of sequence of selections. In M. Schoen (Ed.), The
effects of music (pp. 223–242). New York: Harcourt, Brace.
Gilliland, A. R., & Moore, H. T. (1927). The immediate and long-time
effects of classical and popular phonograph selections. In M. Schoen
(Ed.), The effects of music (pp. 211–222). New York: Harcourt, Brace.
Hargreaves, D. J. (1984). The effects of repetition on liking for music.
Journal of Research in Music Education, 32, 35–47.
Heingartner, A., & Hall, J. V. (1974). Affective consequences in adults and
children of repeated exposure to auditory stimuli. Journal of Personality
and Social Psychology, 29, 719–723.
Heyduk, R. G. (1975). Rated preference for musical compositions as it
relates to complexity and exposure frequency. Perception & Psycho-
physics, 17, 84–91.
Jacoby, L. L. (1983). Perceptual enhancement: Persistent effects of an
experience. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 9, 21–38.
Jacoby, L. L., & Kelly, C. M. (1987). Unconscious influences of memory
for a prior event. Personality and Social Psychology Bulletin, 13, 314–
336.
Jacoby, L. L., & Whitehouse, K. (1989). An illusion of memory: False
recognition influenced by unconscious perception. Journal of Experi-
mental Psychology: General, 118, 126–135.
Jakobovits, I. A. (1966). Studies of fads: I. The “Hit Parade.” Psycholog-
ical Reports, 18, 443–450.
Kail, R. V., Jr., & Freeman, H. R. (1973). Sequence redundancy, rating
dimensions, and the exposure effect. Memory & Cognition, 1, 454–458.
Kalat, J. W., & Rozin, P. (1973). “Learned safety” as a mechanism in
long-delay taste-aversion learning in rats. Journal of Comparative and
Physiological Psychology, 83, 198–207.
King, S. (1996). The green mile: Night journey. New York: Penguin
Audiobooks.
Krugman, H. F. (1943). Affective responses to music as a function of
familiarity. Journal of Abnormal and Social Psychology, 38, 388–392.
Kunst-Wilson, W. R., & Zajonc, R. B. (1980). Affective discrimination of
stimuli that cannot be recognized. Science, 207, 557–558.
Mandler, G., Nakamura, Y., & Van Zandt, B. J. (1987). Nonspecific effects
of exposure on stimuli that cannot be recognized. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition, 13, 646–648.
Meyer, M. (1903). Experimental studies in the psychology of music.
American Journal of Psychology, 14, 456–478.
Monahan, J. L., Murphy, S. T., & Zajonc, R. B. (2000). Subliminal mere
exposure: Specific, general, and diffuse effects. Psychological Science,
11, 462–466.
Mull, H. K. (1957). The effect of repetition upon the enjoyment of modern
music. Journal of Psychology, 43, 155–162.
Murphy, S. T., Monahan, J. L., & Zajonc, R. B. (1995). Additivity of
nonconscious affect: Combined effects of priming and exposure. Journal
of Personality and Social Psychology, 69, 589–602.
Obermiller, C. (1985). Varieties of mere exposure: The effects of process-
ing style and repetition on affective responses. Journal of Consumer
Research, 12, 17–30.
Peretz, I. (2001). Listen to the brain: A biological perspective on musical
emotions. In P. N. Juslin & J. A. Sloboda (Eds.), Music and emotion:
Theory and research (pp. 105–134). Oxford, England: Oxford Univer-
sity Press.
Peretz, I., & Gagnon, L. (1999). Dissociation between recognition and
emotional judgments for melodies. Neurocase, 5, 21–30.
Peretz, I., Gagnon, L., & Bouchard, B. (1998). Music and emotion:
Perceptual determinants, immediacy, and isolation after brain damage.
Cognition, 68, 111–141.
Peretz, I., Gaudreau, D., & Bonnel, A. M. (1998). Exposure effects on
music preferences and recognition. Memory & Cognition, 15, 379–388.
Schupp, H. T., Jungho¨fer, M., Weike, A. I., & Hamm, A. O. (2003).
Emotional facilitation of sensory processing in the visual cortex. Psy-
chological Science, 14, 7–13.
Seamon, J. G., Brody, N., & Kauff, D. M. (1983). Affective discrimination
of stimuli that are not recognized: Effects of shadowing, masking, and
cerebral laterality. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 9, 544–555.
Seamon, J. G., Marsh, R. L., & Brody, N. (1984). Critical importance of
exposure duration for affective discrimination of stimuli that are not
recognized. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 10, 465–469.
Seamon, J. G., Williams, P. C., Crowley, M. J., Kim, I. J., Langer, S. A.,
Orne, P. J., & Wishengrad, D. L. (1995). The mere exposure effect is
based on implicit memory: Effects of stimulus type, encoding condi-
tions, and number of exposures on recognition and affect judgments.
Journal of Experimental Psychology: Learning, Memory, and Cogni-
tion, 21, 711–721.
Sloboda, J. A., O’Neill, S. A., & Ivaldi, A. (2001). Functions of music in
everyday life: An exploratory study using the experience sampling
method. Musicae Scientae, 5, 9–32.
Stang, D. J. (1974). Methodological factors in mere exposure research.
Psychological Bulletin, 81, 1014–1025.
Stang, D. J., & O’Connell, E. J. (1974). Computer as experimenter in social
psychological research. Behavior Research Methods and Instrumenta-
tion, 6, 223–231.
Thompson, W. F., Balkwill, L.-L., & Vernescu, R. (2000). Expectancies
generated by recent exposure to melodic sequences. Memory & Cogni-
tion, 28, 547–555.
Verveer, E. M., Barry, H., & Bousfield, W. A. (1933). Change in affec-
tivity with repetition. American Journal of Psychology, 45, 130–134.
Washburn, M. F., Child, M. S., & Abel, T. M. (1927). The effects of
immediate repetition on the pleasantness or unpleasantness of music. In
M. Schoen (Ed.), The effects of music (pp. 199–210). New York:
Harcourt, Brace.
Whittlesea, B. W., & Price, J. R. (2001). Implicit/explicit memory versus
380
SZPUNAR, SCHELLENBERG, AND PLINER
analytic/nonanalytic processing: Rethinking the mere exposure effect.
Memory & Cognition, 29, 234–246.
Wilson, W. R. (1979). Feeling more than we can know: Exposure effects
without learning. Journal of Personality and Social Psychology, 37,
811–821.
Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of
Personality and Social Psychology Monographs, 9(2, Pt. 2), 1–27.
Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences.
American Psychologist, 35, 151–175.
Zajonc, R. B. (2001). Mere exposure: A gateway to the subliminal. Current
Directions in Psychological Science, 10, 224–228.
Zajonc, R. B., Crandall, R., Kail, R. V., & Swap, W. (1974). Effect of
extreme exposure frequencies on different affective ratings of stimuli.
Perceptual and Motor Skills, 38, 667–678.
Zajonc, R. C., Shaver, P., Tavris, C., & van Kreveld, D. (1972). Exposure,
satiation and stimulus discriminability. Journal of Personality and So-
cial Psychology, 21, 270–280.
Zajonc, R. B., Swap, W. C., Harrison, A. A., & Roberts, P. (1971).
Limiting conditions of the exposure effect: Satiation and relativity.
Journal of Personality and Social Psychology, 18, 384–391.
Received December 24, 2002
Revision received August 18, 2003
Accepted August 22, 2003 䡲
381
REPEATED EXPOSURE TO MUSIC
