Human brain potential correlates of voice priming and voice recognition

movedearAI and Robotics

Nov 17, 2013 (3 years and 7 months ago)

211 views

Neuropsychologia 39 (2001) 921–936
Human brain potential correlates of voice priming and voice
recognition
Stefan R.Schweinberger *
Department of Psychology,Uni￿ersity of Glasgow,58Hillhead Street,Glasgow G128QQ,Scotland,UK
Received 18 October 2000;received in revised form 12 January 2001;accepted 19 January 2001
Abstract
This study investigated repetition priming in the recognition of famous voices,recording reaction times (RTs) and event-related
brain potentials (ERPs).In Experiment 1,a facilitation was found in RTs to famous but not to unfamiliar voices when these had
been primed by a different voice sample of the same speaker earlier in the experiment.However,ERPs to both famous and
unfamiliar voices showed repetition priming in terms of an increased P2 component,which is thought to be generated in the
auditory cortex.When the likelihood of conscious retrieval of primes was reduced in Experiment 2,facilitatory priming in RTs
was again observed for famous voices,but inhibitory priming was now observed for unfamiliar voices.This is consistent with
predictions of a bias model of priming.Moreover,substantial priming was observed even when voice primes were backward
speech samples,which were recognised at chance levels.The results suggests that (a) voice priming is mediated to a large extent
by frequency characteristics of a particular voice,rather than by articulatory and other ‘sequential’ features that are eliminated
in backward speech;(b) priming affects the processing of voices in auditory cortical areas within 200 ms after voice onset;and
(c) explicit recognition of a voice in the priming phase is not a necessary condition for priming to occur.© 2001 Elsevier Science
Ltd.All rights reserved.
Keywords:Priming;Event-related potentials;ERPs;Speaker;Voice recognition
www.elsevier.com/locate/neuropsychologia
1.Introduction
Human voices contain a wealth of information not
only about speech content,but also about a speaker’s
identity and mood.Nevertheless,as is also reflected by
the numerous studies currently published on face recog-
nition,the most important stimulus for conveying a
person’s identity is usually considered to be the face.
Even though there are situations in communication
when the face is invisible,it is only recently that voice
recognition has attracted more interest both in cogni-
tive psychology and neuroscience.This holds even
though some research on voice recognition has been
performed in the 1980s [68–71].It appears that voices
can be recognised from surprisingly short excerpts of
speech [55,57].Evidence is also beginning to emerge
that there may be voice-selective areas in the superior
temporal sulcus bilaterally [2] and/or frontal and tem-
poral regions of the right hemisphere [65].
In a task that involves the recognition of famous
voices among unfamiliar ones,the present paper inves-
tigates both explicit and implicit aspects of voice recog-
nition by using repetition priming.Repetition priming
refers to the observation that the processing of stimuli
can be altered,and often improved,when the same
stimuli have been earlier encountered.Research on
priming has contributed to our understanding of the
visual recognition of words [36,37],pictures of objects
[62],or faces [5].Priming is often observed in terms of
decreased reaction times (RTs) or error rates in re-
sponding to repeated as compared with novel stimuli,
and priming can be present in the absence of explicit
memory for the study items [21].Moreover,the effects
of repetition exhibit a considerable degree of perceptual
specificity,in that they are reduced when the perceptual
format of the stimuli is changed between a study and a
test phase [18,64].
* Corresponding author Tel.:+44-141-3303947;fax:+44-141-
3304606.
E-mail address:s.schweinberger@psy.gla.ac.uk (S.R.Schwein-
berger).
0028-3932/01/$ - see front matter © 2001 Elsevier Science Ltd.All rights reserved.
PII:S0028-3932(01)00023-9
S.R.Schweinberger/Neuropsychologia 39(2001)921–936922
Although most research on repetition priming has
been conducted using visual stimuli,researchers have
now begun to study auditory priming effects (e.g.
[10,14,19,34,58,66]).Notably,some of these studies fo-
cussed on non-￿erbal auditory information.For exam-
ple,Stuart and Jones [66] reported repetition priming in
the identification of environmental sounds such as
those from a bell or a typewriter.Priming depended on
the repetition of perceptual rather than conceptual in-
formation,as was suggested by the absence of any
priming when the spoken label of that sound had been
presented previously.Interestingly,effects caused by a
prime sound that was produced by the same type of
object but was acoustically different from the target
sound (e.g.sounds of a manual and an electric type-
writer) were of similar magnitude as were priming
effects between two identical sounds,suggesting that
these auditory priming effects survive some degree of
perceptual transformation.
In independent recent studies,both Ellis and cowork-
ers [19] and Schweinberger and coworkers [58] were
able to show repetition priming in a voice recognition
task in which participants decided whether a speech
sample of neutral content was spoken by a famous or
an unfamiliar speaker.Schweinberger et al.[58] could
also demonstrate priming when primes and targets were
different speech samples of the same person.This is
similar to the findings in environmental sound recogni-
tion mentioned above [66].
The auditory information used for voice recognition
is provided in the context of speech,which poses the
question about the relationship between voice and
speech recognition.Different aspects of the signal may
be important for voice and speech recognition.For
instance,relative frequency differences between for-
mants are important for vowel identification,whereas
absolute formant frequencies may provide cues for
voice recognition [32,45].Disorders of speech recogni-
tion and voice recognition can be dissociated in neuro-
logical patients [71],so that there can be little doubt
that there is some independence between the brain
systems that mediate the analysis of speech content and
speaker identity recognition.However,there is also
increasing evidence that this independence is not abso-
lute.For example,there may be some overlap in the
acoustic information that subserves speech and voice
recognition [51].Also,spoken words are stored in
long-term memory as traces that contain voice informa-
tion (e.g.[20,38,40,63] see [60,75],for similar findings in
the visual modality).
Priming effects in voice recognition have been at-
tributed to a similar mechanism that has been held
responsible for priming in face recognition (cf.
[9,19,58]).Accordingly,priming was thought to affect
the activation strength of modality-specific ‘face recog-
nition units’ (FRUs — view-independent representa-
tions of familiar faces that are stored in long-term
memory) or the connections of FRUs with post-percep-
tual representations of people (PINs — person identity
nodes,see [9,15]).One argument for this assertion has
been the finding that priming occurred only for faces
that were explicitly recognised in the priming phase of
some experiments [6].In terms of an influential model
by Burton and colleagues [9],a simultaneous activation
of FRU and PIN — a necessary condition for a
strengthening of the FRU–PIN link can take place
only for explicitly recognised faces.If priming in voice
recognition were the result of an analogous mechanism
— a change in the link strength between voice recogni-
tion units (VRUs and PINs) — then only voices that
are explicitly recognised in the priming phase should
give rise to priming.
However,there are already findings from face recog-
nition that are somewhat problematic for such an ex-
planation of repetition priming.Face priming is
gradually reduced by using progressively less similar
portraits of the target person as a prime [18].
1
Although
the most current version of Burton et al.’s model [8] is
able to account for this result,another problem is that
repetition may also change the processing of unfamiliar
faces for which no memory representations should exist
(e.g.[17,25]).In this paper,two experiments will be
reported that aim to identify the point in processing at
which repetition priming affects voice recognition.In
order to determine whether priming effects are specific
for voices with pre-existing memory representations,
priming will be investigated for both familiar and unfa-
miliar voices.
Event-related potentials (ERPs) were recorded in Ex-
periment 1 in order to gain more precise information on
the temporal dynamics and the possible neural pro-
cesses affected by repetition priming in voice recogni-
tion.ERPs have been extremely informative in the
study of priming processes [1,26,35,41,53,56,59].Re-
cently,ERP studies have begun to consider the effect of
voice gender changes in source memory for words [73],
have tried to isolate the auditory ERP to one’s own
voice during vocalisation [22],or have investigated the
effects of stimulus phonation types on ERPs and elec-
1
Research in face priming suggests that certain types of stimulus
transformations (such as changes in viewpoint or details of the
pattern of image intensities) reduce priming,whereas other types of
transformation (such as changes in stimulus size) do not [4].Related
findings from the auditory domain come from a study by Church and
Schacter [11],who demonstrated that priming effects in an auditory-
verbal task (i.e.word-stem completion) were reduced by changes in
emotional intonation or fundamental frequency within a speaker’s
voice,but not by changes in overall sound level of the stimuli.At
present,however,no relevant data on priming in voice recognition
are available,except for the demonstration that voice priming does
not require the repetition of the identical utterance between priming
and test [58].
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 923
tromagnetic brain responses to vowels [67].However,
the present study is the first to report ERPs in a voice
recognition task.
2
After presenting evidence,in Experiment 1,that repe-
tition priming affects the perceptual processing of a
voice within the first 200 ms after voice onset,Experi-
ment 2 will be aimed at further narrowing down the
acoustic stimulus features that mediate voice repetition
priming.To this end,the effects of a rather radical
acoustic manipulation — backward presentation of
speech — on priming in voice recognition was studied.
Backward speech preserves acoustic information about
fundamental frequency and frequency range of the
voices as well as speaking rate.In contrast,backward
speech distorts phonological cues [48,69],which in
adults are thought to be processed predominantly by
the left hemisphere [43].If priming was mediated by
phonological or articulatory cues that are sensitive to
temporal order,one would,therefore,expect an elimi-
nation of priming when backward primes are used.In
contrast,if priming was mediated by features that are
preserved in backward speech,such as the frequency
content of voices,then one would expect a preservation
of priming from backward speech samples.
2.Experiment 1
2.1.Method
2.1.1.Participants
Twelve participants (four women and eight men)
aged between 21 and 30 years (M=25.9 years) were
paid to contribute data to this study.None of the
participants reported any hearing problems.Three fur-
ther participants were excluded and replaced either
because of technical problems in data acquisition (N=
2) or because response accuracy in the test phase was
close to chance level (i.e.less than 60% correct re-
sponses;N=1).
2.1.2.Stimuli and apparatus
Voice samples from 80 famous people from various
areas (e.g.politics,sports,TV) were used in the present
experiment.These celebrities were selected on the basis
of high ratings for ease of voice recognition in a pilot
study.All voice samples were transferred from
videotape recordings of TV broadcasts to the hard disk
of a microcomputer using a digital audio card.Ana-
logue-to-digital conversion was performed using a sam-
pling rate of 40 kHz and 12-bit resolution.
Two different voice samples were prepared for each
celebrity according to the following criteria:(a) The
verbal content of the samples should not provide any
cue for speaker identity.This was verified by having a
group of independent raters judge the content of the
samples (see Appendix A).(b) The utterances should be
spoken in an emotionally neutral tone and with normal
intensity.(c) Every sample was edited in order to
synchronise its onset with the onset of a phrase or
sentence,and contained exactly 2000 ms of continuous
speech.This sample duration was chosen on the basis
of a previous study which suggested that voice recogni-
tion performance was not likely to improve with longer
samples [57].(d) All samples were scaled to equivalent
maximal signal intensities of approximately 77 dB(A),
and were presented binaurally using Revox RH 31
headphones.Voice samples from 80 unfamiliar people
were prepared in an analogous way.These were taken
from the same videotape sources and resembled the
famous voices with respect to speech content and gen-
der ratio.
2.1.3.Procedure
Prior to the experiments,participants received writ-
ten task instructions.Subsequently,the EEG electrodes
were applied.The experiment consisted of a priming
and a test phase.During the priming phase,participants
heard a series of 80 voice samples (40 famous and 40
unfamiliar) in a randomised order.The interval be-
tween the onsets of successive stimuli was 7 s.For every
voice sample participants decided as fast as possible by
speeded two-choice keypresses whether it was spoken
by a famous or an unfamiliar person.Participants were
told that although every voice sample lasted 2 s,they
should respond as soon as possible.Half the partici-
pants used the right index finger to indicate a famous
speaker and the left index finger to indicate an unfamil-
iar speaker;for the other half,this assignment was
reversed.
During the test phase,participants heard a series of
160 voice samples (80 famous and 80 unfamiliar).
Again they decided as fast as possible whether a famous
or an unfamiliar voice was presented.The 80 famous
voices were divided into two experimental conditions of
40 voices each,as follows:(a) Voices of famous people
that had not been presented during the priming phase
(UNPRIMED),(b) Voices of famous people that had
already been presented during the priming phase
(PRIMED).Repetition priming for unfamiliar voices
was also investigated,using the same experimental ma-
nipulations.That is,half of the 80 unfamiliar voices
during the test phase were unprimed and half were
primed by having presented a voice sample of the same
unfamiliar person during the priming phase.In all
2
One reason for this may be that the type of stimulus that is
typically used for voice recognition experiments,that is,extracts from
continuous speech,is quite non-standard for ERP recordings.The
duration and the considerable acoustic variability of continuous
speech stimuli do not appear to be particularly favourable for obtain-
ing ERPs of high quality.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936924
primed conditions,different voice samples from the
same person were always used for the priming and test
phase,respectively.The approximate time lag between
the end of the priming phase and the beginning of the
test phase was 10 min.
The 80 famous people whose voices were used in the
test phase had been randomly assigned to one of two
sets of 40 people each (see Appendix B),and the same
held for the 80 unfamiliar people.The target voices
used for the respective primed and unprimed conditions
were counterbalanced over participants so that any
differences between the priming conditions could not be
ascribed to stimulus differences.During both priming
and test phase,short breaks were allowed after every 40
trials.Ten practice trials were run before the experi-
ment,using voice samples from different people as
those used in the experiment.
2.1.4.Performance
Responses were scored as correct if the correct key
was pressed within a time window lasting from 200 to
4000 ms after voice onset.Errors of omission (no
keypress) and of commission (wrong key) were
recorded separately.Mean reaction times were calcu-
lated for correct responses only.
2.1.5.E￿ent-related potentials
The electroencephalogram (EEG) was recorded with
tin electrodes mounted in an electrode cap (Electro-Cap
International Inc.) at the scalp positions F
z
,C
z
,P
z
,Fp
1
,
Fp
2
,F
3
,F
4
,C
3
,C
4
,P
3
,P
4
,O
1
,O
2
,F
7
,F
8
,T
7
,T
8
,P
7
,P
8
,
TP
9
and TP
10
.Note that the T
7
,T
8
,P
7
,and P
8
locations
are equivalent to T
3
,T
4
,T
5
,and T
6
in the old nomen-
clature,and TP
9
and TP
10
refer to inferior temporal
locations over the left and right mastoids,respectively
[46].The TP
10
(right upper mastoid) electrode served as
initial common reference.Electrode impedances were
kept below 10 k￿ and were typically below 5 k￿.The
horizontal electrooculogram (EOG) was recorded from
the outer canthi of both eyes,and the vertical EOG was
monitored from an electrode below the right eye
against Fp
2
.The time constant of the amplifier was 10
s;low-pass filters were set to 40 Hz (−3 dB attenua-
tion;12 dB roll-off/octave).Event-related potentials
were recorded for 3540 ms starting 200 ms before target
stimulus onset,and sampled at a rate of 100 Hz.
Offline,all trials were visually inspected for artifacts
of ocular (e.g.blinks,eye movements) and non-ocular
origin (e.g.channel blockings or drifts).Trials with
non-ocular artifacts and trials with incorrect behavioral
responses were discarded.For all other trials,ocular
contributions to the EEG were corrected [13].ERPs
were averaged separately for each channel and for the
four experimental conditions.Each averaged ERP was
low-pass filtered at 10 Hz with a zero phase shift digital
filter,and recalculated to average reference [33].
3.Results
3.1.Performance
The mean recognition accuracy for voice samples in
the priming phase was 68.0%,with average percentages
of errors of commission and omission at 31.4 and 0.6%,
respectively.In the test phase,mean recognition accu-
racy was 72.9%,with average percentages of errors of
commission and omission at 26.4 and 0.7%,respec-
tively.For both the mean correct RTs and the percent-
age of errors of commission from the test phase,
analyses of variance (ANOVAs) were performed with
repeated measures on familiarity (famous,unfamiliar)
and repetition priming (primed,unprimed).Where ap-
propriate,epsilon corrections for heterogeneity of co-
variances were performed throughout [27].Performance
data are shown in Fig.1.
3.2.Reaction times
The ANOVA of the reaction times gave rise to a
significant main effect of familiarity,F(1,11)=338.8,
P￿0.001,reflecting the fact that RTs to famous voices
were faster than RTs to unfamiliar voices.There was
also a significant main effect of priming,F(1,11)=7.4,
P￿0.05,which was qualified by a significant interac-
tion between familiarity and priming,F(1,11)=12.1,
P￿0.01.Repetition priming was significant for famous
Fig.1.Top:Reaction times in Experiment 1 to primed and unprimed
target voices.Bottom:Same for the percentage of errors of commis-
sion.Error bars show standard errors of the means.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 925
Fig.2.ERPs to primed and unprimed famous target voices.At C
z
,note the N1 and P2 potentials that are followed by a sustained negative
potential of high amplitude.Also note that the time scale is expanded for the first 500 ms (the segment left to the vertical line),in order to show
the early ERPs more clearly.
￿oices (difference=161 ms,F(1,11)=13.2,P￿￿0.01),
but not for unfamiliar ￿oices (difference=6 ms,F(1,
11)￿1).
3.3.Error rates
Numerically,error rates were somewhat smaller for
famous as compared with unfamiliar voices,but this
difference was not significant,F(1,11)=2.7,P￿=0.13.
There was no significant effect of priming and no
interaction,F
s
￿1.
3.4.E￿ent-related potentials
The mean number of error- and artifact-free single
trials that contributed to an averaged ERP per subject
was 27.6,25.7,25.5,and 25.3 for the primed famous,
unprimed famous,primed unfamiliar,and unprimed
unfamiliar conditions,respectively.The ERPs to the
voice stimuli in the test phase were characterised by the
well-known N1 and P2 components of the auditory
evoked potential.These waves were followed by a
prominent sustained potential (SP),which at central
sites appeared as a uniform negativity lasting until the
end of the stimulus.This SP was somewhat more
prominent over right as compared with left frontal and
temporal electrodes (see Fig.2 and Fig.3).A distinct
early part of the SP was seen between around 450 and
800 ms at some electrodes,with negativity at frontal
electrodes and positivity at parietal electrodes,respec-
tively (see Fig.2 and Fig.3).
ERPs were quantified with mean amplitude measures
in the time segments 80–120 ms (N1),170–240 ms
(P2),450–800 ms (early SP),and 800–2000 ms (late
SP),relative to a 200 ms prestimulus baseline.Mean
amplitudes were analysed using ANOVAs analogous to
the ones for the performance data,but with the inclu-
sion of an additional variable electrode site with 21
levels.Note that because the average reference sets the
mean activity across all electrodes to zero,any condi-
tion effects in the overall analyses are only meaningful
in interaction with electrode site.For significant effects
in the overall ANOVA,subsequent ANOVAs were
performed separately for electrodes over frontal (F
z
,
Fp
1
,Fp
2
,F
3
,F
4
,F
7
,F
8
),central-parietal (C
z
,P
z
,C
3
,C
4
,
P
3
,P
4
),and temporal (T
7
,T
8
,P
7
,P
8
,TP
9
and TP
10
)
regions (Fig.4).
3.5.Effects of priming
There were no significant effects of priming for the
N1,F(20,220)=1.1;P￿0.20.In contrast,priming
effects were significant for the P2,F(20,220)=3.1;
P￿0.05 (see also Figs.2 and 3 Fig.4).In general,P2
S.R.Schweinberger/Neuropsychologia 39(2001)921–936926
Fig.3.Same as in Fig.2 but for primed and unprimed unfamiliar target voices.
amplitudes were larger for primed as compared with
unprimed voices.At frontal recordings,this resulted in
larger amplitude positivity,F(1,11)=6.3;P￿0.05.At
temporal recordings,at which the electrically negative
aspect of P2 is seen most clearly,amplitudes were more
negative for primed as compared with unprimed voices,
F(1,11)=8.2;P￿0.05.No significant differences were
seen at central-parietal electrodes.
With respect to longer latency effects,there were no
significant effects of priming for the 450–800 ms seg-
ment,F(20,220)=1.5;P=0.20.However,the
ANOVA revealed some influence of priming on ERPs
for the 800–2000 ms segment,F(20,220)=2.3;P￿
0.05.Fig.3 shows an increased negativity,particularly
at central-parietal electrodes,for primed unfamiliar
voices.In contrast,Fig.2 does not suggest much differ-
ence for famous voices in the 800–2000 ms segment.
Separate ANOVAs for the 800–2000 ms segment were,
therefore,performed for famous and unfamiliar voices.
Note that these analyses were performed although
strictly speaking they would be justified by an interac-
tion between priming and familiarity,which did not
reach significance,F(20,220)=1.7;P=0.15.The sepa-
rate analyses suggested significant priming for unfamil-
iar voices,F(20,220)=4.6;P￿0.001,but not for
famous voices,F￿1.For unfamiliar voices,the most
clear differences due to priming in the 800–2000 ms
segment were seen at central-parietal sites,F(1,11)=
15.9;P￿0.01.Some differences were present at frontal
sites,F(1,11)=7.9;P￿0.05,with no differences at
temporal sites,F(1,11)=1.1;P￿0.20.
3.6.Effects of familiarity
Effects of familiarity were only seen in longer latency
ERPs,in the 450–800 ms segment,F(20,220)=3.5;
P￿0.01,and in the 800–2000 ms segment,F(20,
220)=4.3;P￿0.001.This reflected the finding that the
SP was less negative for famous as compared with
unfamiliar voices at central-parietal regions,F(1,11)=
9.8,P￿0.01,and more negative for famous voices at
temporal regions,F(1,11)=5.0,P￿0.05 (see also Fig.
2 and Fig.3).No significant effects of voice familiarity
were seen over frontal regions.
3.7.Hemispheric asymmetry
Finally,in order to determine whether the hemi-
spheric differences to be seen in the 800–2000 ms
segment (cf.Fig.2 and Fig.3) were statistically reliable,
three planned pairwise comparisons between the homo-
logue electrode pairs F
3
/F
4
,C
3
/C
4
,and P
3
/P
4
were
performed,and Bonferroni correction was applied for
this series of pairwise comparisons.The SP was signifi-
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 927
cantly larger,i.e.more negative,over the right hemi-
sphere for the F
4
versus F
3
electrode,F(1,11)=8.5,
P￿0.05 (uncorrected P=0.0140),and tended to be
larger for the C
4
versus C
3
electrode,F(1,11)=7.3,
P=0.06 (uncorrected P=0.0206).No asymmetry was
present for the P
4
versus P
3
electrode,F(1,11)￿1.
Topographical maps of the ERPs in the time seg-
ments that underwent statistical analyses are shown in
Fig.4.In line with the statistical analyses,Fig.4 depicts
the clearest priming effects for the P2 (170–240 ms)
segment.
4.Discussion
Clear N1 and P2 components as well as a prominent
sustained potential (SP) were observed in response to
voice samples.The experimental effects on the ERPs
will be discussed after a few more general comments.
Overall,these ERPs to voice samples are similar to
previously reported ERPs to speech stimuli.In sen-
tences with natural speech,although clear N1 and P2
potentials are observed for sentence-initial words,these
components are completely absent for subsequent
words — N1 and P2 are only obtained if a silent gap is
introduced before words [74].The sustained potential
(SP) in the present data is probably a variant of the
long-lasting baseline shift that has been previously de-
scribed as a response to sustained auditory stimuli of a
longer duration (i.e.in the order of seconds,[42,44]).In
line with earlier findings,the end of this SP was time-
locked,with a fixed delay,to the offset of the auditory
stimulus (Fig.2 and Fig.3).
According to previous studies,the generators of the
N1,P2,and of the SP are all in the vicinity of the
supratemporal auditory cortex areas of both hemi-
spheres [12,30,42,67].This assumption is also consistent
with the scalp topographies of these components in the
present study.However,the generators of the SP were
previously found to be located slightly medially and
anteriorly to the source of the M100 — the magnetic
counterpart of the N1 [42].Similarly,the generators of
the P2 appear to be located anterior to the generators
of the N1 [67],supporting the idea that although both
phenomena reflect cortical activity related to auditory
processing,they have different generators.Recently the
auditory P2 was demonstrated to be larger to vowel
utterances than to tones that were matched in fre-
quency spectra [67],suggesting a speech-related effect.
3
There are two important ERP findings of this experi-
ment.First,priming effects in ERPs were observed
remarkably early,suggesting that repetition priming
influenced the perceptual processing of voices within
the first 200 ms after speech onset.
4
Second,similar
priming effects on the P2 were observed for both fa-
mous and unfamiliar voices,although RT effects of
priming appeared to be specific to famous voices.
Fig.4.Top:Topographical maps of ERPs in Experiment 1 to primed
and unprimed famous target voices.Bottom half:Same for primed
and unprimed unfamiliar target voices.Note:Maps are shown for
average amplitudes in those time segments for which statistics were
computed.Positivity is shaded.Isopotential lines are every 0.5 ￿V for
the N1 (80–120 ms) and P2 (170–240 ms) segments,and every 1.5 ￿V
for the sustained potential segments (450–800 ms and 800–2000 ms,
respectively).Maps were obtained by using spherical spline interpola-
tion (cf.Perrin,Pernier,Bertrand,& Echallier,1989).Top view of the
head extends from C
z
to 110° in all directions.Black dots indicate the
electrode locations.
3
Moreover,the difference was not seen in the P2m — the mag-
netic counterpart of the P2,which led the authors to conclude that
electric recordings might be more sensitive than magnetic ones for
detecting these speech-related effects.Of potential interest are also
differences in brain responses due to differences in glottal excitation
(‘pressed’ versus ‘soft’ vowels,phonation styles which in human
communication may be used to express for example the emotional
content of the speech message).The P2/P2m generators appeared to
change with glottal excitation in the right but not in the left hemi-
sphere (see Figure 5 in [67]).Unfortunately,it was not reported
whether this difference was significant.
4
Although this finding is interpreted in terms of an effect on
perceptual processing,this does of course,not exclude the possibility
that the effect may be sensitive to attention.Attentional modulation
has been frequently observed for auditory ERP or MEG components
with latencies earlier than 200 ms,and this may also hold for
components elicited by speech sounds (e.g.[47]).
S.R.Schweinberger/Neuropsychologia 39(2001)921–936928
With respect to the original question of whether or
not one would find priming for unfamiliar voices,there
is therefore,a discrepancy between behavioural and
electrophysiological data in Experiment 1.One possibil-
ity is that this discrepancy relates to a current contro-
versy about the interpretation of priming (e.g.
[49,50,54]).Ratcliff and McKoon [49] claimed that
priming is not specific to familiar items,but rather that
it causes a bias to respond ‘familiar’ for both familiar
and unfamiliar items.This bias effect may be based on
familiarity or perceptual fluency.For unfamiliar items,
this bias can be opposed by conscious retrie￿al of
previous item exposure [49],which could explain the
present null priming effect in RTs for unfamiliar voices.
A similar two-process account of priming was presented
by Jacoby et al.[29].Ratcliff and McKoon [49,50]
demonstrated that,when conscious recollection of items
from the priming phase is made less likely (e.g.by
having participants respond to test stimuli within a
short deadline),the typical pattern of priming shows
bias,with benefits for familiar responses but costs for
unfamiliar responses.However,when conscious recol-
lection is made more likely,costs for unfamiliar re-
sponses tend to disappear.
For unfamiliar voices,the ERP data in this study
indicate two effects of priming,rather than an absence
of priming.These two effects of priming on the P2 and
on the sustained potential components might tenta-
tively be ascribed to familiarity and conscious recollec-
tion,respectively.The null effect of priming on RTs to
unfamiliar voices,therefore,may have been produced
by conscious recollection counteracting the misleading
familiarity effects for repeated unfamiliar voices.If this
interpretation is correct,then it should be possible to
produce a bias pattern of priming in which priming of
unfamiliar voices becomes inhibitory,simply by reduc-
ing or eliminating conscious recollection of the prime
stimuli.This possibility was explored in Experiment 2.
4.1.Experiment 2
The first aim of Experiment 2 was to reduce or
eliminate conscious recollection of the primes,by in-
creasing the number of stimuli shown in the priming
phase,and by including backward voices to the priming
phase.Voice recognition rate drops when speech is
played backward [69].Under these conditions,two-pro-
cess models of priming [50] would,therefore,predict a
bias pattern of priming,that is largely governed by the
familiarity effects of repetition.Thus,facilitatory prim-
ing would be observed for famous voices but inhibitory
priming would be seen in the behavioural responses to
unfamiliar voices.
The finding of an influence of repetition priming on
the P2 — i.e.within the first 200 ms after voice onset
— also has implications concerning the level of audi-
tory processing influenced by voice repetition.At that
point in time,only relatively low-level auditory infor-
mation — such as information about voice quality or
frequency characteristics — should be available.How-
ever,the P2 results should be considered at best as
indirect evidence for the acoustic features that mediate
priming.It is clear that the role of different acoustic
features can be studied more directly when the stimuli
are manipulated in a way that affects certain types of
information selectively.
During the priming phase of Experiment 2,some
voice samples were played backward.Backward speech
is incomprehensible,and it also eliminates information
about articulatory patterns that are sensitive to tempo-
ral order.In contrast,information about the frequency
characteristics of a voice is preserved in backward
speech (e.g.[69]).It should also be noted that although
phonetic information is clearly distorted,backward
speech may convey a limited amount of phonetic pat-
terns [3].This may be explained by the fact that some
phonemes (such as long-duration vowels) are almost
symmetrical temporally and are thus not sensitive to
temporal order.Nevertheless,comparing voice priming
effects between forward and backward primes will be
useful in order to determine those aspects of auditory
information that mediate priming.
5.Method
5.1.Participants
Eighteen participants (10 women and 8 men) aged
between 20 and 37 years (M=23.5 years) were paid to
contribute data to this study.None of the participants
had served in Experiment 1,and none reported any
hearing problems.Three further participants had to be
excluded and replaced either because of technical prob-
lems in data acquisition (N=2) or because response
accuracy in the test phase was close to chance level (i.e.
less than 60% correct responses in the test phase;N=
1).
5.2.Stimuli and apparatus
Voice samples were prepared and presented in the
same way as in Experiment 1.However,stimuli were
prepared from 38 additional speakers (19 famous and
19 unfamiliar),expanding the stimulus pool to a total
of 99 famous and 99 unfamiliar speakers (see Appendix
C).
5.3.Procedure
Prior to the experiments,participants received writ-
ten task instructions.The experiment consisted of a
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 929
priming and a test phase.During the priming phase,
participants were presented with a total of 132 voice
samples (66 famous and 66 unfamiliar),evenly divided
into four blocks of 33 trials each.Within each block,
the number of famous voice samples was either 16 or
17,keeping the frequency of famous and unfamiliar
voice samples broadly equivalent.Samples were pre-
sented forward during Blocks 1 and 4,and different
samples were played backward during Blocks 2 and 3.
The interval between the onsets of successive stimuli
was always 7 s.For every voice sample,participants
decided as fast as possible by speeded two-choice key-
presses whether it was spoken by a famous or an
unfamiliar person.Participants were told that al-
though every voice sample lasted 2 s,they should
respond as soon as they were reasonably sure as to
whether they heard a famous or an unfamiliar
speaker.Half the participants used the right index
finger to indicate a famous speaker and the left index
finger to indicate an unfamiliar speaker;for the other
half of the participants,this assignment was reversed.
During the test phase,participants were presented
with a series of 198 voice samples (99 famous and 99
unfamiliar).It is important to note that all voices in
the test phase were presented in forward speech.As in
the priming phase,participants were to decide as fast
as possible whether a famous or an unfamiliar voice
was presented.The 99 famous voices were divided
into three experimental conditions of 33 voices each,
as follows:(a) Voices of famous people that had not
been presented during the priming phase (UN-
PRIMED);(b) Voices of famous people that had been
presented during the priming phase,using forward
presentation (PRIMED FORWARD);and (c) Voices
of famous people that had been presented during the
priming phase,using backward presentation
(PRIMED BACKWARD).Repetition priming for un-
familiar voices was investigated using the same experi-
mental manipulations.That is,33 of the 99 unfamiliar
voices during the test phase were unprimed,33 were
primed forward and 33 were primed backward.In all
primed conditions,different voice samples from the
same person were used in the priming and test phase,
respectively.The approximate time lag between the
end of the priming phase and the beginning of the
test phase was 10 min.
The 99 famous people whose voices were used in
the test phase had been assigned to one of three sets
of 33 people each (see Appendix C),and the same
held for the 99 unfamiliar people.The target voices
used for the respective experimental conditions were
counterbalanced over participants so that any differ-
ences between the priming conditions cannot be as-
cribed to stimulus differences.During both priming
and test phase,short breaks were allowed after every
33 trials.Twelve practice trials (six with forward and
six with backward presentation) were conducted be-
fore the experiment.The voice samples used in these
practice trials were from different people to those
used in the experiment.RTs and error rates were
scored in an analogous way to Experiment 1.
5
6.Results
6.1.Performance
The first important observation in this experiment
was that participants appeared unable to recognise
voices from the backward speech samples in the prim-
ing phase above chance levels.The mean accuracy of
51.4% was not significantly different from the chance
level of 50%,t(17)=1.2,P￿0.20 (average percent-
ages of errors of commission and omission were 47.9
and 0.7%,respectively).In contrast,the mean recogni-
tion accuracy for forward speech samples in the prim-
ing phase was 59.9%,which,though not very high,
was significantly above chance level,t(17)=6.5;P￿
0.001 (average percentages of errors of commission
and omission were 39.2 and 0.9%,respectively).
The mean recognition accuracy for the forward
speech samples in the test phase was 70.2%,with av-
erage percentages of errors of commission and omis-
sion at 29.3 and 0.5%,respectively.For both the
mean correct RTs and the percentage of errors of
commission from the test phase,ANOVAs were per-
formed with repeated measures on familiarity (fa-
mous,unfamiliar) and repetition priming (primed
forward,primed backward,unprimed).Where appro-
priate,epsilon corrections for heterogeneity of covari-
ances were performed with the Huynh–Feldt method
throughout.The performance data are shown in Fig.
5.
6.2.Reaction times
The ANOVA of the reaction times gave rise to a
significant main effect of familiarity,F(1,17)=54.4,
P￿0.001,reflecting the fact that RTs to famous
voices were faster than RTs to unfamiliar voices.
There was no significant main effect of priming,F(2,
34)￿1,but a significant interaction between priming
and familiarity,F(2,34)=13.5,P￿0.001.Fig.5 sug-
5
For exploratory purposes,ERPs were also recorded in a similar
way as in Experiment 1.As was to be expected from the inclusion of
more experimental conditions,it turned out that the number of error-
and artifact-free single trials that could contribute to an averaged
ERP per subject was considerably lower than in Experiment 1,with
less than 10 trials per average in some subjects and conditions.A
preliminary analysis of ERP data in Experiment 2 suggested effects or
trends that are consistent with the arguments put forward in this
paper,but it was decided not to include these data because of the
somewhat unsatisfactory data quality.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936930
Fig.5.Top:Reaction times in Experiment 2 to target voices that were
unprimed,or had been primed by a forward prime or a backward
prime.Bottom:Same for the percentage of errors of commission.
Error bars show standard errors of the means.
P￿￿0.01.This main effect was due to the observation
that error rates were slightly higher in the primed than
in the unprimed conditions.However,the main effect
was qualified by a significant interaction between famil-
iarity and priming,F(2,34)=4.4,P￿0.05.For famous
voices,the numerical facilitatory effect of priming (Fig.
5) was not significant,F(2,34)=2.2,P=0.12.For
unfamiliar voices,the inhibitory effects of priming was
significant,F(2,34)=5.7,P￿0.01.Relative to the
unprimed condition,inhibitory priming in terms of
increased false familiarity error rates were observed both
for voices that had been primed by forward primes,F(1,
17)=8.7,P￿0.05,and those that had been primed by
backward primes,F(1,17)=9.7,P￿0.05.Error rates
were similar for voices that had been primed by forward
versus backward primes,F(1,17)￿1.
7.Discussion
There are two important findings of Experiment 2.
First,backward voice primes caused very clear and
significant priming effects in RTs to forward voice
targets,even though there was a very considerable lag
both in terms of time and number of stimuli intervening
between primes and targets.Overall,the effects caused
by backward primes were only slightly (and not signifi-
cantly) smaller than the priming effects caused by for-
ward voice primes.This was observed even though the
explicit recognition of the backward voice primes was at
chance levels.These results suggest that priming was
mediated to a large extent by acoustic information that
was preserved in backward speech.Moreover,explicit
recognition of voices in the priming phase is apparently
not a necessary condition for priming to occur.This
supports the idea that priming facilitated an early stage
of auditory processing that may be related to the analysis
of frequency information,which is preserved in back-
ward speech.Because phonetic information is distorted
in backward speech [3],a strong contribution of phonetic
information to priming appears unlikely.Nevertheless,it
must be noted that the present design does not allow to
completely exclude a contribution of those phonetic
aspects that are preserved in backward speech (i.e.
phonemes that are near-symmetric temporally).This
issue may require further clarification.However,the
present results clearly suggest that articulatory and
phonetic information that is sensitive to temporal order
seems to be of relatively minor importance in voice
priming.
The second important finding is that in contrast to
Experiment 1,clear inhibitory priming effects were
observed for unfamiliar voices.The lowrecognition rates
of items fromthe priming phase in Experiment 2 suggest
that relative to Experiment 1,the conscious recollection
of primes was indeed effectively reduced if not elimi-
gests that this interaction reflected the fact that repetition
priming speeded RTs for famous voices but slowed RTs
for unfamiliar voices.
The facilitatory priming effect for famous voices was
significant F(2,34)=3.4,P￿0.05.ANOVAs testing the
contrasts between conditions revealed a significant prim-
ing effect from forward primes,F(1,17)=6.0,P￿0.05,
and a tendency of priming from backward primes when
compared with the unprimed condition,F(1,17)=2.9,
P=0.11,but no difference between the effects caused by
forward and backward primes,F￿1.
The inhibitory priming effect for unfamiliar voices was
also significant,F(2,34)=16.1,P￿0.001.ANOVAs
testing the contrasts between conditions revealed signifi-
cant inhibitory priming fromforward primes,F(1,17)=
29.4,P￿0.01,and significant inhibitory priming from
backward primes F(1,17)=16.9,P￿0.01,but again no
difference between the effects caused by forward and
backward primes,F(1,17)=1.2,P￿0.20.Taken to-
gether,it is clear that priming did not cause a facilitation
for famous voices only,but rather caused a bias effect,
most clearly reflected in the interaction between priming
and familiarity (also cf.Fig.5).
6.3.Error rates
In error rates,there was no main effect of familiarity,
F￿1,but a main effect of priming,F(2,34)=5.6,
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 931
nated.The pattern of inhibitory priming for unfamiliar
voices is consistent with predictions by bias models of
priming [50].
8.General discussion
The present data corroborate and extend earlier find-
ings on voice priming [58].Specifically,the ERP results
suggest that repetition priming in voice recognition
influences processing within the first 200 ms after stimu-
lus onset.By demonstrating that the P2 is sensitive to
voice priming,the present study extends recent findings
of a sensitivity of the P2 amplitude for speech processing
[67].At the same time,the P2 results suggest that priming
modulated processing within auditory cortical areas.
Specifically,other results suggest that regions anterior to
the primary auditory cortex such as area 22 may be
involved [2,67].
Backward voice primes yielded a substantial RT prim-
ing effect,which was only slightly (and not significantly)
smaller than the priming effect caused by forward primes.
The priming effect must,therefore,depend to a large
extent on the repetition of auditory information that is
preserved in backward primes.As indicated above,this
suggests that priming is mediated by the frequency
characteristics of a voice,rather than by articulatory or
phonetic cues that are sensitive to temporal order.Of
particular interest,although forward speech may engage
left-hemispheric areas that are not engaged by backward
speech [43],both forward and backward speech were
recently reported to activate regions in the superior
temporal sulcus [3] when the task does not require
lexical-semantic or syntactic processing.In line with these
findings,the present ERP results could indicate that
auditory areas mediate voice priming by both forward
and backward voices.
With respect to the hemispheric distribution of the
sustained potential,it is noteworthy that the higher
amplitude negativity over right than left frontal and
central areas contrasts with the left-lateralised sustained
negativities that are usually found in verbal tasks [31].
Although a closer link would clearly require further
research,this asymmetry may,therefore,be related to
other studies which suggest a right hemisphere superior-
ity for voice recognition (e.g.[71],see also [39]).
The pattern of results in these experiments is in line
with the predictions by two-process models of priming
[29,49,50].This becomes clear when considering the
priming effects for unfamiliar voices.For unfamiliar
voices in Experiment 1,priming effect were absent in
RTs,although the ERPs revealed both early (P2) and late
(SP) priming effects.In Experiment 2,in which the
likelihood of conscious recollection of primes was re-
duced,a clear inhibitory RTpriming effect for unfamiliar
voices was revealed,as predicted by these models.This
supports the idea that repetition priming of unfamiliar
stimuli can cause two opposite effects that tend to cancel
each other in performance.In Experiment 1,a repeated
unfamiliar voice may have elicited some familiarity,
initially interfering with its classification as unfamiliar.
Subsequently,however,participants may have become
aware that the voice appeared familiar because it had
been heard in the priming phase (rather than because it
is a famous voice),facilitating its classification as unfa-
miliar.In Experiment 2,conscious recollection of primes
— and any compensatory effects thereof — were largely
eliminated,which explains why unfamiliar voices yielded
inhibitory priming under these conditions.While it is
clearly desirable to performfurther experiments that are
specifically designed to explore models of priming,an
important implication of the present study is that ERPs
might be well suited to disentangle the relative contribu-
tions of familiarity and conscious recollection in repeti-
tion priming.
Current psychological accounts of priming can be
broadly classified into structural and episode-based the-
ories of repetition priming (e.g.[16]).Structural theories
(e.g.[9,16]) assume that priming depends on structural
changes within the systems that mediate recognition of
familiar stimuli.Repetition may change either the status
of a representation (e.g.the threshold or baseline activa-
tion of a recognition unit;[37]) or the strength of a link
between representations at different levels (e.g.[9,16,72]).
The present data provide difficulties for these models,
especially because at present they provide no mechanism
that explains priming for unfamiliar stimuli.It is proba-
bly fair to say that,because there is currently some
controversy about possible differences in the processing
levels involved in face and voice recognition,respectively
[23],the present results need not be fatal for those models
which were mainly developed with respect to face recog-
nition.Nevertheless,it may be noted that an occasional
finding in the face recognition literature has been that
repetition priming may slow RTs for unfamiliar stimuli
(e.g.[17,76]).The implication is that structural models
will need to take into account the issues of representation
of unfamiliar stimuli and the familiarisation with novel
stimuli (see also [7]).
Similarly,the clear priming effect frombackward voice
primes,which are recognised at chance levels,is not in
conflict with structural models in general.However,it is
clearly inconsistent with the specific claim that explicit
recognition of a stimulus in the priming phase is a
prerequisite for priming to occur [6]
6
This claim is a
6
In this study,RTs for recognized faces in the test phase were
compared depending on whether or not these faces were sponta-
neously recognized in the priming phase.However,one concern is
that the faces in these two groups (that were formed post-hoc
according to the participant’s performance) may have already differed
before the experiment,e.g.in familiarity.In this case,the condition
differences need not necessarily reflect differential priming.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936932
consequence of the idea that a strengthening,or Heb-
bian update,of the link between an FRU and a PIN
only occurs if PIN activation exceeds a certain
threshold,which is the same threshold that determines
whether or not the stimulus will be recognised.In
contrast,the present data suggest that either the effect
of repetition priming in voice recognition acts on a level
earlier than the VRU-PIN links (for some evidence in
favour of this possibility,see [23]),or that the threshold
of PIN activation required for a strengthening of the
link between a VRU and a PIN is lower than the
threshold required for explicit voice recognition.
Episodic accounts of priming assume that a prior
encounter with a stimulus establishes a specific episodic
memory trace.It is held that a contact with that trace
at a second encounter causes more efficient processing,
for example,by perceptual enhancement [28].A related
view,transfer-appropriate processing [52],suggests that
the degree of repetition priming is a function of the
similarity in processing between priming and test.Some
episodic models may easily explain the priming costs
seen in unfamiliar voices (e.g.[29]).However,if re-
trieval of a specific episodic trace were a critical factor,
then one would expect repetition priming to be reduced
with decreasing similarity between prime and target.In
contrast,there was not much of a difference in priming
when primes were similar to the targets (both compris-
ing forward speech) as compared with when they were
dissimilar (primes comprising backward speech but
targets comprising forward speech).
Compared with face recognition,voice recognition is
a much slower and more error-prone process [19,58].
When recognition is slow and characterised by rela-
tively high response uncertainty,then influences of
memory traces for unfamiliar stimuli formed in the
priming phase may be more likely to inhibit perfor-
mance,relative to conditions in which performance is
usually fast and error-free even for unprimed stimuli
(also cf.[72]).Another possibility is that there are
differences between voices and faces in the level at
which familiarity decisions are taken.For example,
Hanley,Smith,and Hadfield ([23],but see [24]) re-
ported that when compared with faces,famous voices
elicited a disproportionately high percentage of ‘famil-
iar only’ states,in which listeners had a definite feeling
of familiarity but were unable to retrieve any further
information about the speaker.Similarly,participants
in another study [57] could retrieve further information
for only about half of the famous voices that they had
found familiar.Hanley et al.[23] argued that this
phenomenon reflects a blockage between the VRU and
PIN level rather than between PINs and semantic infor-
mation.A structural model that assumes that familiar-
ity decisions for voices are taken at a level prior to
PINs might explain why explicit recognition,in the
sense of a strengthening of the VRU–PIN link,is not a
prerequisite for priming to occur.
In sum,the present findings confirm and extend
previous results on voice priming [19,58].The ERP data
from Experiment 1 show effects of repetition priming in
voice recognition on the P2 component.This suggests
that priming modulates activity in auditory areas within
the first 200 ms after stimulus onset.The findings from
Experiment 2 further support a bias model of priming,
and suggest that the explicit recognition of a prime
stimulus is not a necessary condition for priming to
occur.Moreover,voice priming appears to modulate
the processing of frequency characteristics of a voice,
rather than affecting the processing of those articula-
tory or phonetic cues that are sensitive to temporal
order.
Acknowledgements
Portions of this research were presented at the ‘39.
Tagung Experimentell Arbeitender Psychologen’ (TeaP)
in Berlin,March 1997 and as a book abstract related to
this conference [61].This research was supported by a
Heisenberg fellowship of the Deutsche Forschungsge-
meinschaft.The help of Volker Stief,Anja Herholz,
and Michael Barensteiner in data collection is gratefully
acknowledged.I would also like to thank Mike Burton
and Simon Garrod for helpful comments on a previous
draft of this manuscript,Patrick Berg for software
support,and Juan Delius for his support and encour-
agement of this research.
Appendix A.Section and evaluation of stimulus
materials
Eight subjects participated in a pilot study that
served as a basis for the final selection of stimuli used in
these experiments.They were given a list of 708 names
of familiar people including celebrities from various
domains like politics,sports,art,television,etc.For
every name,they had to judge their familiarity with
each person’s name,face,and voice separately on a
three-point scale from 0 to 2.
For Experiment 1,80 people were then selected from
this pool considering only those with relatively high
voice familiarity ratings.These 80 people were ran-
domly assigned to two sets of 40 each (see below).For
a given subject,these stimulus sets could be assigned to
the two different experimental conditions.As was to be
expected from the random assignment,the two sets
were highly comparable with respect to average rated
familiarity of the celebrities’ voices (M [S.D.] =
1.13[0.49] and 1.15[0.44] for set 1 and 2,respectively.
An ANOVA revealed no difference between the two
sets (F(1,78)￿1).For Experiment 2,99 people were
selected from the same pool,which could be divided
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 933
into three sets of 33 people each.For a given subject,
these three stimulus sets could be assigned to the three
different experimental conditions.The three sets were
highly comparable with respect to rated voice familiar-
ity (M [S.D.] =1.09[0.45],1.02[0.52],and 1.08[0.50] for
set 1 through 3,respectively.An ANOVA revealed no
systematic differences between the three sets (F (2,
96)￿1).
To control for the possibility that the verbal content
of the samples enabled participants to recognise the
speakers,eight additional subjects were presented with
printed text versions of the 160 utterances that had
been used in Experiment 1 (80 from famous and 80
from unfamiliar speakers).They were asked to indicate
for every utterance whether they considered it as having
been spoken by a famous or an unfamiliar person.The
average percentages of items that were classified as
famous on the basis of the printed text versions were
50.8 and 49.5 for utterances actually made by famous
and unfamiliar speakers,respectively,(F(2,78)￿1).
One may,therefore,be confident that the verbal con-
tent of the present voice samples alone did not enable
subjects to discriminate between famous and unfamiliar
people.
Appendix B.Celebrities whose voices were presented in
Experiment 1
Set 1 Set 2
1.Gerhard Polt 41.Oskar Lafontaine
42.Gregor Gysi2.Fritz Egner
43.Horst Seehofer3.Wolfgang Scha¨uble
44.Regine Hildebrandt4.Friedrich Nowotny
45.Ernst-Dieter Lueg5.Dagmar Berghoff
46.Hans Clarin6.Claus Seibel
47.Dieter-Thomas Heck7.Konstantin Wecker
48.Richard von8.Volker Ru¨he
Weizsa¨cker
9.Theo Waigel 49.Heinz Sielmann
10.Franz-Josef Strauss 50.Hans -Dieter Hu¨sch
11.Herbert Wehner 51.Jan Hofer
52.Heinz Erhardt12.Heiner Geissler
53.Karl Dall13.Peter Hahne
14.Gerhard 54.Rita Su¨ssmuth
Stoltenberg
15.Inge Meysel 55.Willy Millowitsch
16.Helmut Fischer 56.Hans-Jochen Vogel
17.Helmut Schmidt 57.Marianne Sa¨gebrecht
18.Bernhard Grzimek 58.Manfred Krug
19.Martin 59.Ulrich Wickert
Semmelrogge
60.Frank Elstner20.Uschi Glas
61.Joachim Fuchsberger21.Roberto Blanco
22.Klaus Kinkel 62.Berti Vogts
63.Horst Tappert23.Sabine Christiansen
64.Roman Herzog24.Didi Hallervorden
25.Loriot 65.Franz Beckenbauer
66.Norbert Blu¨m26.Boris Becker
67.Heinz Riesenhuber27.Lothar Mattha¨us
28.Hans-Dietrich 68.Gerd Fro¨be
Genscher
69.Heinz Ru¨hmann29.Rudolf Scharping
70.Bjo¨rn Engholm30.Joschka Fischer
71.Margarethe31.Brigitte Bastgen
Schreinemaker
32.Gu¨nther Verheugen 72.Hildegard Knef
73.Helmut Kohl33.Rainhard Fendrich
74.Joachim Brauner34.Kari Valentin
35.Peter Frankenfeld 75.Willy Brandt
76.Angela Merkel36.Dieter Hildebrandt
77.Ju¨rgen von der Lippe37.Alfred Biolek
38.Fritz Wepper 78.Gustl Bayrhammer
79.Thomas Gottschalk39.Gu¨nther Rexrodt
80.Marcel Reich-Ranicki40.Harald Juhnke
Appendix C.Celebrities whose voices were used in
Experiment 2
Set 2Set 1 Set 3
1.Ju¨rgen von 34.Steffi Graf 67.Antje-Katrin
Ku¨hnemannder Lippe
2.Margarethe 35.Boris Becker 68 Hans-Dieter
Hu¨schSchreinemaker
3.Peter 36.Helmut 69.Willy
MillowitschSchmidtFrankenfeld
4.Gerhard Polt 37.Joschka 70.Hans-Jochen
VogelFischer
71.Marianne38.Brigitte5.Fritz Egner
Bastgen Sa¨gebrecht
39.Gu¨nther6.Wolfgang 72.Manfred
Verheugen KrugScha¨uble
40.Eva Herman7.Friedrich 73.Ulrich
WickertNowotny
41.Max8.Dagmar 74.Frank
SchautzerBerghoff Elstner
75.Joachim42.Sascha Hehn9.Claus Seibel
Fuchsberger
43.Gerhard 76.Berti Vogts10.Konstantin
Wecker Schro¨der
44.Rainhard11.Volker Ru¨he 77.Horst
TappertFendrich
12.Theo Waigel 45.Karl 78.Roman
HerzogValentin
46.Dieter13.Franz-Josef 79.Ju¨rgen
Mo¨llemannHildebrandtStrauss
S.R.Schweinberger/Neuropsychologia 39(2001)921–936934
47.Alfred14.Herbert 80.Eduard
BiolekWehner Zimmerman
48.Fritz15.Heiner 81.Klaus
WepperGeissler To¨pfer
49.Gu¨nther16.Peter Hahne 82.Katrin
Mu¨llerRexrodt
17.Gerhard 50.Harald 83.Franz
BeckenbauerJuhnkeStoltenberg
51.Oskar 84.Norbert18.Inge Meysel
Lafontaine Blu¨m
52.Gregor Gysi 85.Heinz19.Helmut
Fischer Riesenhuber
53.Horst 86.Gerd Fro¨be20.Hans Moser
Seehofer
21.Kurt 54.Regine 87.Heinz
Biedenkopf Ru¨hmannHildebrandt
88.Bjo¨rn55.Ernst-Dieter22.Oliver
Lueg EngholmHardy (dubbed
voice)
56.Hans Clarin23.Bernhard 89.Hildegard
KnefGrzimek
57.Dieter-24.Martin 90.Helmut
Thomas Heck KohlSemmelrogge
58.Richard von25.Uschi Glas 91.Joachim
BraunerWeizsa¨cker
26.Roberto 59.Theo Lingen 92.Willy Brandt
Blanco
60.Stan Lau- 93.Angela27.Klaus
Kinkel rel(dubbed Merkel
voice)
28.Sabine 61.Marie-Luise 94.Gustl
BayrhammerMarjanChristiansen
62.Heinz29.Didi 95.Thomas
Hallervorden GottschalkSielmann
30.Loriot 96.Marcel63.Jan Hofer
Reich-Ranicki
31.Lothar 64.Heinz 97.Edmund
StoiberErhardtMattha¨us
32.Hans-Di- 65.Karl Dall 98.Michael
Stichetrich Genscher
66.Rita33.Rudolf 99.Konrad
Su¨ssmuth AdenauerScharping
References
[1] Anderson JE,Holcomb PJ.Auditory and visual semantic
priming using different stimulus onset asynchronies:an event-
related brain potential study.Psychophysiology 1995;32:177–
90.
[2] Belin P,Zatorre RJ,Lafaille P,Ahad P,Pike B.Voice-selec-
tive areas in human auditory cortex.Nature 2000;403:309–12.
[3] Binder JR,Frost JA,Hammeke TA,Bellgowan PSF,Springer
JA,Kaufman JN,Possing ET.Human temporal lobe activa-
tion by speech and nonspeech sounds.Cerebral Cortex
2000;10:512–28.
[4] Bruce V,Burton M,Carson D,Hanna E,Mason O.Repeti-
tion priming of face recognition.In:Umilta`C,Moscovitch M,
editors.Attention and Performance XV.Cambridge,London:
MIT Press,1994:179–201.
[5] Bruce V,Valentine T.Identity priming in the recognition of
familiar faces.British Journal of Psychology 1985;76:373–83.
[6] Brunas-Wagstaff J,Young AW,Ellis AW.Repetition priming
follows spontaneous but not prompted recognition of familiar
faces.The Quarterly Journal of Experimental Psychology
1992;44A:423–54.
[7] Burton AM.Learning new faces in an interactive activation
and competition model.Visual Cognition 1994;1:313–48.
[8] Burton AM,Bruce V,Hancock PJ B.From pixels to people:a
model of familiar face recognition.Cognitive Science
1999;23:1–31.
[9] Burton AM,Bruce V,Johnston RA.Understanding face
recognition with an interactive activation model.British Jour-
nal of Psychology 1990;81:361–80.
[10] Chiu CY,Schacter DL.Auditory priming for nonverbal infor-
mation:implicit and explicit memory for environmental
sounds.Consciousness and Cognition 1995;4:440–58.
[11] Church BA,Schacter DL.Perceptual specificity of auditory
priming:implicit memory for voice intonation and fundamental
frequency.Journal of Experimental Psychology:Learning,
Memory and Cognition 1994;20:521–33.
[12] Crottaz-Herbette S,Ragot R.Perception of complex sounds:
N1 latency codes pitch and topography codes spectra.Clinical
Neurophysiology 2000;111:1759–66.
[13] Elbert T,Lutzenberger W,Rockstroh B,Birbaumer N.Re-
moval of ocular artifacts from the EEG — a biophysical ap-
proach to the EOG.Electroencephalography and Clinical
Neurophysiology 1985;60:455–63.
[14] Ellis AW.Modality-specific repetition priming of auditory
word recognition.Current Psychological Research 1982;2:123–
8.
[15] Ellis AW.Cognitive mechanisms of face processing.Philosoph-
ical Transactions of the Royal Society of London B
1992;335:113–9.
[16] Ellis AW,Flude BM,Young A,Burton AM.Two loci of
repetition priming in the recognition of familiar faces.Journal
of Experimental Psychology:Learning,Memory and Cognition
1996;22:295–308.
[17] Ellis AW,Young AW,Flude B.Repetition priming and face
processing:priming occurs within the system that responds to
the identity of a face.The Quarterly Journal of Experimental
Psychology 1990;42A:495–512.
[18] Ellis AW,Young AW,Flude BM,Hay DC.Repetition prim-
ing of face recognition.The Quarterly Journal of Experimental
Psychology 1987;39A:193–210.
[19] Ellis HD,Jones DM,Mosdell N.Intra- and inter-modal repe-
tition priming of familiar faces and voices.British Journal of
Psychology 1997;88:143–56.
[20] Goldinger SD,Kleider HM,Shelley E.The marriage of per-
ception and memory:creating two-way illusions with words
and voices.Memory & Cognition 1999;27:328–38.
[21] Graf P,Squire LR,Mandler G.The information that amnesic
patients do not forget.Journal of Experimental Psychology:
Learning,Memory and Cognition 1984;10:164–78.
[22] Gunji A,Hoshiyama M,Kakigi R.Identification of auditory
evoked potentials of one’s own voice.Clinical Neurophysiology
2000;111:214–9.
[23] Hanley JR,Smith ST,Hadfield JI.Recognise you but I can’t
place you:an investigation of familiar-only experiences during
tests of voice and face recognition.The Quarterly Journal of
Experimental Psychology 1998;51A:179–95.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936 935
[24] Hanley JR,Turner JM.Why are familiar-only experiences more
frequent for voices than for faces?The Quarterly Journal of
Experimental Psychology 2000;53A:1105–16.
[25] Hay DC.Testing instance models of face repetition priming.
Memory & Cognition 2000;28:192–203.
[26] Holcomb PJ.Semantic priming and stimulus degradation:impli-
cations for the role of the N400 in language processing.Psycho-
physiology 1993;30:47–61.
[27] Huynh H,Feldt LS.Estimation of the box correction for degrees
of freedom from sample data in randomized block and split plot
designs.Journal of Educational Statistics 1976;1:69–82.
[28] Jacoby LL.Perceptual enhancement:persistent effects of an
experience.Journal of Experimental Psychology:Learning,
Memory and Cognition 1983;9:21–38.
[29] Jacoby LL,Woloshyn V,Kelley C.Becoming famous without
being recognized:unconscious influences of memory produced
by dividing attention.Journal of Experimental Psychology:Gen-
eral 1989;118:115–25.
[30] Jones SJ,Vaz Pato M,Sprague L,Stokes M,Munday R,Haque
N.Auditory evoked potentials to spectro-temporal modulation
of complex tones in normal subjects and patients with severe
brain injury.Brain 2000;123:1007–16.
[31] Jung R,Altenmu¨ller E,Natsch B.Zur Hemispha¨rendominanz
fu¨r Sprache und Rechnen:Elektrophysiologische Korrelate einer
Linksdominanz bei Linksha¨ndern.Neuropsychologia
1984;22:755–75.
[32] Kuwabara H,Ohgushi K.Contributions of vocal tract resonant
frequencies and bandwidths to the personal perception of speech.
Acustica 1987;63:120–8.
[33] Lehmann D.Principles of spatial analysis.In:Gevins AS,Re-
mond A,editors.Handbook of Electroencephalography and
Clinical Neurophysiology.1.Methods of Analysis of Brain
Electrical and Magnetic Signals.Amsterdam:Elsevier,
1987:309–54.
[34] McClelland AGR,Pring L.An investigation of cross-modality
effects in implicit and explicit memory.The Quarterly Journal of
Experimental Psychology 1991;43A:19–33.
[35] Mitchell PF,Andrews S,Ward PB.An event-related potential
study of semantic congruity and repetition in a sentence-reading
task:effects of context change.Psychophysiology 1993;30:496–
509.
[36] Morton J.Interaction of information in word recognition.Psy-
chological Review 1969;76:165–78.
[37] Morton J.Facilitation in word recognition:experiments causing
change in the logogen model.In:Kolers PA,Wrolstal M,
Bouma H,editors.Processing of Visible Language.New York:
Plenum Press,1979:259–68.
[38] Mullenix JW,Pisoni DB.Stimulus variability and processing
dependencies in speech perception.Perception and Psycho-
physics 1990;47:379–90.
[39] Neuner F,Schweinberger SR.Neuropsychological impairments
in the recognition of faces,voices,and personal names.Brain
and Cognition 2000;44:342–66.
[40] Nygaard LC,Sommers MS,Pisoni DB.Speech perception as a
talker-contingent process.Psychological Science 1994;5:42–6.
[41] Paller KA,Gross M.Brain potentials associated with perceptual
priming vs.explicit remembering during the repetition of visual
word-form.Neuropsychologia 1998;36:559–71.
[42] Pantev C,Eulitz C,Elbert T,Hoke M.The auditory evoked
sustained field:origin and frequency dependence.Electroen-
cephalography and Clinical Neurophysiology 1994;90:82–90.
[43] Perani D,Dehaene S,Grassi F,Cohen L,Cappa SF,Dupoux E,
Fazio F,Mehler J.Brain processing of native and foreign
languages.NeuroReport 1996;7:2439–44.
[44] Picton TW,Woods DL,Proulx GB.Human auditory sustained
potentials.I.The nature of the response.Electroencephalogra-
phy and Clinical Neurophysiology 1978;45:186–97.
[45] Pisoni DB.Some thoughts on ‘normalization’ in speech percep-
tion.In:Johnson K,Mullenix JW,editors.Talker Variability in
Speech Processing.San Diego:Academic Press,1997:9–32.
[46] Pivik RT,Broughton RJ,Coppola R,Davidson RJ,Fox N,
Nuwer MR.Guidelines for the recording and quantitative analy-
sis of electroencephalographic activity in research contexts.Psy-
chophysiology 1993;30:547–58.
[47] Poeppel D,Yellin E,Phillips C,Roberts TPL,Rowley HA,
Wexler K,Marantz A.Task-induced asymmetry of the auditory
evoked M100 neuromagnetic field elicited by speech sounds.
Cognitive Brain Research 1996;4:231–42.
[48] Ramus F,Hauser MD,Miller C,Morris D,Mehler J.Language
discrimination by human newborns and by cotton-top tamarin
monkeys.Science 2000;288:349–51.
[49] Ratcliff R,McKoon G.Bias in the priming of object decisions.
Journal of Experimental Psychology:Learning,Memory and
Cognition 1995;22:1383–407.
[50] Ratcliff R,McKoon G.Bias effects in implicit memory tasks.
Journal of Experimental Psychology:General 1996;125:403–21.
[51] Remez RE,Fellowes JM,Rubin PE.Talker identification based
on phonetic information.Journal of Experimental Psychology:
Human Perception and Performance 1997;23:651–66.
[52] RoedigerIII HL.Implicit memory.Retention without remember-
ing.American Psychologist 1990;45:1043–56.
[53] Rugg MD.Dissociation of semantic priming,word and non-
word repetition effects by event-related potentials.The Quarterly
Journal of Experimental Psychology 1987;39A:123–48.
[54] Schacter DL,Cooper LA.Bias in the priming of object deci-
sions:logic,assumptions,and data.Journal of Experimental
Psychology:Learning,Memory and Cognition 1995;21:768–76.
[55] Schellenberg EG,Iverson P,McKinnon MC.Name that tune:
identifying popular recordings from brief excerpts.Psychonomic
Bulletin & Review 1999;6:641–6.
[56] Schweinberger SR.How Gorbachev primed Yeltsin:analyses of
associative priming in person recognition by means of reaction
times and event-related brain potentials.Journal of Experimental
Psychology:Learning,Memory and Cognition 1996;22:1383–
407.
[57] Schweinberger SR,Herholz A,Sommer W.Recognizing famous
voices:influence of stimulus duration and different types of
retrieval cues.Journal of Speech,Language and Hearing Re-
search 1997;40:453–63.
[58] Schweinberger SR,Herholz A,Stief V.Auditory long-term
memory:repetition priming of voice recognition.The Quarterly
Journal of Experimental Psychology 1997;50A:498–517.
[59] Schweinberger SR,Pfu¨tze E-M,Sommer W.Repetition priming
and associative priming of face recognition:evidence from event-
related potentials.Journal of Experimental Psychology:Learn-
ing,Memory and Cognition 1995;21:722–36.
[60] Schweinberger SR,Soukup GR.Asymmetric relationships
among perceptions of facial identity,emotion,and facial speech.
Journal of Experimental Psychology:Human Perception and
Performance 1998;24:1748–65.
[61] Schweinberger SR,Stief V,Barensteiner M.Auditorisches
Langzeitgeda¨chtnis:wiederholungspriming beim Stimmenerken-
nen durch ru¨ckwa¨rts abgespielte Sprache.In:van der Meer E,
Bachmann T,Beyer R,Goertz C,Hagendorf H,Krause B,
Sommer W,Wandke H,Ziessler M,editors.Experimentelle
Psychologie.Lengerich:Pabst Science Publisher,1997 235
(Abstract).
[62] Seymour PHK.Human Visual Cognition.London:Macmillan,
1979.
[63] Sheffert SM.Contributions of surface and conceptual informa-
tion to recognition memory.Perception and Psychophysics
1998;60:1141–52.
[64] Srinivas K.Perceptual specificity in nonverbal priming.Journal
of Experimental Psychology:Learning,Memory and Cognition
1993;19:582–602.
S.R.Schweinberger/Neuropsychologia 39(2001)921–936936
[65] Stevens AA,Schwartzreich E.Dissociating message and messen-
ger:fMRI of word and voice recognition.Journal of Cognitive
Neuroscience 2000;12:S49 Abstract.
[66] Stuart GP,Jones DM.Priming the identification of environmen-
tal sounds.The Quarterly Journal of Experimental Psychology
1995;48A:741–61.
[67] Tiitinen H,Sivonen P,Alku P,Virtanen J,Na¨a¨ta¨nen R.
Electromagnetic recordings reveal latency differences in speech
and tone processing in humans.Cognitive Brain Research
1999;8:355–63.
[68] VanLancker D,Kreiman J.Voice discrimination and recognition
are separate abilities.Neuropsychologia 1987;25:829–34.
[69] VanLancker D,Kreiman J,Emmorey K.Familiar voice recogni-
tion:patterns and parameters.Part I.Recognition of backward
voices.Journal of Phonetics 1985;13:19–38.
[70] VanLancker DR,Canter GJ.Impairment of voice and face
recognition in patients with hemispheric damage.Brain and
Cognition 1982;1:185–95.
[71] VanLancker DR,Kreiman J,Cummings J.Voice perception
deficits:neuroanatomical correlates of phonagnosia.Journal of
Clinical and Experimental Neuropsychology 1989;11:665–74.
[72] Wheeldon LR,Monsell S.The locus of repetition priming of
spoken word production.The Quarterly Journal of Experimental
Psychology 1992;44A:723–61.
[73] Wilding EL.Separating retrieval strategies from retrieval suc-
cess:an event-related potential study of source memory.Neu-
ropsychologia 1999;37:441–57.
[74] Woodward SH,Ford JM,Hammett SC.N4 to spoken sentences
in young and older subjects.Electroencephalography and Clini-
cal Neurophysiology 1993;87:306–20.
[75] Yakel DA,Rosenblum LD,Fortier MA.Effects of talker vari-
ability on speechreading.Perception & Psychophysics
2000;62:1405–12.
[76] Young AW,McWeeny KH,Hay DC,Ellis AW.Access to
identity-specific semantic codes from familiar faces.The Quar-
terly Journal of Experimental Psychology 1986;38A:271–95.