Janata, P., & Paroo, K. (2006). Acuity of Auditory Images in Pitch and Time. Perception &...

7/27/2019 Janata, P., & Paroo, K. (2006). Acuity of Auditory Images in Pitch and Time. Perception & Psychophysics, 68(5), 82

1/17

eScholarship provides open access, scholarly publishing

services to the University of California and delivers a dynamic

research platform to scholars worldwide.

Previously Published Papers

Peer Reviewed

Title:

Acuity of auditory images in pitch and time

Author:

Janata, PetrParoo, Kaivon

Publication Date:

07-01-2006

Publication Info:

Previously Published Papers, Multi-Campus

Permalink:

http://escholarship.org/uc/item/7rc4x52p

Abstract:

We examined the pitch and temporal acuity of auditory expectations/images formed underattentional-cuing and imagery task conditions, in order to address whether auditory expectationsand auditory images are functionally equivalent. Across three experiments, we observed that pitchacuity was comparable between the two task conditions, whereas temporal acuity deterioratedin the imagery task. A fourth experiment indicated that the observed pitch acuity could not beattributed to implicit influences of the primed context alone. Across the experiments, image acuityin both pitch and time was better in listeners with more musical training. The results support a viewthat auditory images are multifaceted and that their acuity along any given dimension dependspartially on the context in which they are formed.
http://escholarship.org/http://escholarship.org/uc/item/7rc4x52phttp://escholarship.org/uc/search?creator=Paroo%2C%20Kaivonhttp://escholarship.org/uc/search?creator=Janata%2C%20Petrhttp://escholarship.org/uc/search?entity=postprintshttp://escholarship.org/http://escholarship.org/http://escholarship.org/http://escholarship.org/


2/17

Although mental images are likel to pla a central rolein perception (Neisser, 1976), the issue of mental imageris usuall ignored in investigations of selective attention,perceptual acuit, or various forms of memor. However,most experiments on cognitive and perceptual processesrel on tasks in which reference mental images are formedand maintained in working memor over some period oftime before the are compared with sensor input. For ex-

ample, selective attention tasks often require the detectionof infrequent deviants that occur in an attended stream.Mental images ma be formed of either the deviant or thestandard stimulus (or both) so that incoming stimuli canbe categorized accordingl. Similarl, tasks that are usedto examine the abilit to maintain features of a referenceitem in working memor across a dela period or a seriesof distractor items also rel on the maintenance of a mentalimage of the reference item until the time at which a probeitem is presented that either matches or differs from thereference item. Despite parsimonious unifing viewpoints(e.g., Neisser, 1976), the interpla between imager, atten-

tion, perception, and memor remains to be clarified.Perhaps the differences in the dominant paradigms and

tasks have impeded the search for a common neural basisof mental image formation. Studies that directl addressthe issue of mental imager generall seek to answer somecombination of two questions. First, does the process of

forming and maintaining mental images depend on re-sources associated with other cognitive functions, suchas attention or working memor? Second, how well doestheform of mental images match corresponding percepts?Both process and form are addressed in studies that exam-ine whether imager is modalit specific and depends onthe same neural substrates as perception (for a review, seeKossln & Thompson, 2000). Behavioral, neuropscho-

logical, and neuroimaging findings across man visualimager experiments overwhelmingl suggest that theproperties of mentall generated visual images in imag-er tasks parallel the properties of perceived stimuli andrecruit the same temporal and occipital brain areas that arenecessar for visual perception (for reviews, see Farah,2000; Kossln & Thompson, 2000).

The literature on auditor imager is sparse in compari-son with the visual imager literature, although man simi-lar conclusions have been reached. For example, the processof imagining notes in familiar and unfamiliar melodies re-cruits the secondar auditor cortex in the temporal lobe, as

has been shown using PET (Zatorre, Halpern, Perr, Meer,& Evans, 1996), fMRI (Halpern & Zatorre, 1999; Krae-mer, Macrae, Green, & Kelle, 2005), and ERPs (Janata,2001), suggesting shared neural substrates for auditorperception and imager. Working memor experimentshave shown that the maintenance of auditor images fortones in working memor is impaired b distractor items(Deutsch, 1970) and the amount of attention that is paid tothe distractors, with the greatest deficits arising for tonaldistractors, in comparison with verbal or visual distrac-tors (Pechmann & Mohr, 1992). Explicit rehearsal of the

tone across a period of distractors strengthens its memortrace, relative to trials in which attention and workingmemor are also directed toward a distractor task (Keller,

829 Copright 2006 Pschonomic Societ, Inc.

This research was supported b National Institutes of Health Grant R03DC05146 to P.J. We thank Sonja Rakowski for help in collecting data andMari Riess Jones, Sean Fannon, Amrita Puri, and two anonmous review-ers for helpful comments on earlier versions of the manuscript. Correspon-

dence concerning this article should be addressed to P. Janata, Center forMind and Brain, Universit of California, One Shields Avenue, Davis, CA95616 (e-mail: [email protected]).

Acuity of auditory images in pitch and time

PETR JANATAUniversity of California, Davis, California

and Dartmouth College, Hanover, New Hampshire

and

KAIVON PAROODartmouth College, Hanover, New Hampshire

We examined the pitch and temporal acuity of auditory expectations/images formed under attentional-cuing and imagery task conditions, in order to address whether auditory expectations and auditoryimages are functionally equivalent. Across three experiments, we observed that pitch acuity was com-

parable between the two task conditions, whereas temporal acuity deteriorated in the imagery task.A fourth experiment indicated that the observed pitch acuity could not be attributed to implicit influ-ences of the primed context alone. Across the experiments, image acuity in both pitch and time wasbetter in listeners with more musical training. The results support a view that auditory images aremultifaceted and that their acuity along any given dimension depends partially on the context in whichthey are formed.

Perception & Psychophysics2006, 68 (5), 829-844


3/17

830 JANATA AND PAROO

Cowan, & Saults, 1995). Although the working memorexperiments cited are not explicitl auditor imager ex-periments, the suggest that attention, working memor,and imager are intimatel intertwined (Kalakoski, 2001).

Auditor images also appear to have perceptual prop-erties, in that the influence the perceptual processing of

auditor stimuli. The interactions between auditor mentalimages and percepts have been examined behaviorall in avariet of was and to varing degrees of precision. Usinga signal detection task, Farah and Smith (1983) found thatthe intensit threshold for detecting a pure tone in noisewas lower when listeners formed an image of the targettone (e.g., 715 Hz) either before or during the observa-tion interval than when the formed an image of a differ-ent tone (e.g., 1000 Hz). The concluded that forming themental image drew attention to the appropriate frequenc,thus facilitating perceptual processing. Similarl, formingan expectanc that a 20-msec target tone will occur at a

specific time following a cue facilitates the tones detec-tion, in comparison with when the target unexpectedl oc-curs 150 msec before or 100 msec after the expected time(Wright & Fitzgerald, 2004). Auditor images are alsospectrotemporall complex, as is illustrated b imagerof timbre (instrument qualit). Imagining that a pitch cuedwith a sine tone has a specific timbre (e.g., guitar) facili-tates the judgment that a subsequent target has the samepitch when the target is plaed with a matching timbre, incomparison with when it is plaed with a different timbre(Crowder, 1989).

Further evidence that auditor images behave like theirperceptual counterparts has come from studies of audi-tor images in musical contexts. Hubbard and Stoeckig(1988) asked listeners to form a mental image of a chordor tone that was one scale degree above a chord or tonecue and, upon forming the image, to decide whether a tar-get consisted of the same pitch(es) as their image. As inthe studies cited above, the mental image facilitated thedecision to respondsame to a target that was the same asthe image. Moreover, the response times and accurac todifferent chords depended on their harmonic proximitto the imagined chord, with a closel related target chord

showing more interference effects than a distantl relatedchord. Thus, an imagined chord primed other chords inharmonic space in the same wa that an externall pre-sented chord would have.

Coarse temporal aspects of a musical stimulus are alsopreserved in mental images. When subjects make judg-ments about the relative pitch height of two lrics in fa-miliar songs, the time required to make those judgmentsincreases with increasing distance (measured in beats)between those lrics in the melod, as though the subjectsimagine the melod in order to make the judgment (Hal-pern, 1988). The interpla of attention, expectation, timing,

and imager is also evident in studies that manipulate howattention is aimed in pitch and time (e.g., Dowling, Lung,& Herrbold, 1987; Jones & Boltz, 1989). For example, thetemporal evolution of imagined endings can be shaped bmanipulating the rhthmic and tonal accent structure of thepreceding melodic phrase (Jones & Boltz, 1989).

With the present set of four experiments, we extendprevious behavioral studies of auditor imager in twowas. First, we investigate the precision of auditor im-ages b measuring their tuning in pitch and time. Sec-ond, we make within-subjects comparisons of the acuitof the images formed in three different task contexts. The

first task can be thought of in the context of attentionalcuing. In Experiments 1 and 2, attention is guided b theinitial seven notes of an ascending major scale that col-lectivel generate a strong expectation to hear the tonic,the most stable note of the ke after which the scale isnamed (Dowling et al., 1987; Jones, 1981). In our case,the tonic serves as the reference point about which a tun-ing or timing judgment is made. In Experiment 3, thetuning and timing judgments are made for the leadingtone. The leading tone is the seventh note in the scale andis less perceptuall stable and characteristic of the kethan is the tonic.

The second task is explicitl an imager task. Subjectsare instructed to listen to the initial three to five notes ofthe scale and imagine the remaining notes prior to makingthe same tuning/timing judgment about a target that oc-curs at or around the time that the last note in the sequencewould have occurred (the tonic in Experiments 1 and 2 andthe leading tone in Experiment 3). B pairing a task thatexamines how auditor attention is deploed to locationsin pitch and time with an imager task in which a greaterdistance must be traversed along a cognitive map in orderto attain the same location in pitch and time, our aim wasto establish whether the expectations and the images thatexist at the moment that the probe is heard are compa-rable in terms of their content. Although comparable acu-it is not proof that mental images formed in the contextof attentional orienting and expectation are the same asthose formed in the context of imager (or recall of notesequences from memor), it seems that comparable acu-it is a prerequisite for stipulating that the mental imagesarrived at via different task instructions are, indeed, iden-tical at the point at which an internall derived mentalimage is compared with sensor input. It bears pointingout that from the standpoint of Neissers (1976) percep-

tual ccle model, our two tasks are functionall identical,with the onl difference being that the anticipated sensorevent in the imager task is at the end of a longer chain ofimages/anticipations.

Finall, in Experiment 4, we utilized a ke membershipjudgment task to compare the image acuit for the tonicversus the leading tone within subjects. Because the taskwas structured as a discrimination task, as in Experiments2 and 3, performance on the task would benefit from ac-curate image formation. However, no imager instruc-tions were given, and the stimuli were modified slightlto discourage the use of the imager strateg that had been

explicitl encouraged in Experiments 13. Thus, this taskassessed the dependence of image acuit on the sense ofke that was primed b the context notes and allowed usto make two predictions. First, if goal-directed mental im-ager imparts no additional clarit to the mental repre-sentations of the notes that are compared with the probes,


4/17

AUDITORy IMAGE ACUITy 831

we would expect that the acuit of the pitch images inExperiment 4 would match the acuit of the pitch imagesin Experiments 2 and 3. In other words, an acuit wouldarise from the primed context alone. Second, we predictedthat the listeners judgments would be worse for the lead-ing tone than for the tonic, because the relationships of

the tonic and leading tone to the ke that is primed b thecontext are so different (Krumhansl, 1990). Subjectivel,the leading tone is judged to fit less well than the tonicinto a primed ke, and identifing it as a member of theke requires more time (Janata & Reisberg, 1988). Weexpected that the difference between the leading tone andthe tonic would be echoed in our task when the listenersused the primed ke information alone but that this deficitwould be ameliorated when the judgment occurred in thecontext of an explicit imager task.

EXPERIMENT 1

In the first experiment, we compared the attentional-cuing and imager conditions, using a method of constantstimuli, in order to obtain an initial estimate of the imagetuning curves in pitch and time. The dependent variableof interest was the proportion ofcongruousthat is, in-tune andon-timeresponses. The proportion of congru-ous responses was calculated for each level of deviationin listen andimagine trial tpes. In the listen trials, we ex-pected that the proportion of congruous responses woulddecrease as the amount of deviation increased. We chosethe deviation levels so that the proportions would var

across the full range of 0 to 1, but the shapes of the dis-tributions in the listen conditions were not of tremendousinterest. Of primar interest was a comparison, betweenthe listen and the imagine conditions, of the shapes ofthe response distributions across different levels of mis-tuning. Specificall, we expected that the mental imagesmight be less precise in the imagine conditions because offewer exogenous cues to guide their formation. Increasedblurriness of the mental images would manifest itself asa greater proportion of congruous (in-tune) responses atlarger deviations. Similarl, we were interested whether

the number of notes to be imagined would influence theshape of the distributions in the imager conditions. Spe-cificall, we suspected that the blurriness of a mentalimage might increase when more notes of the scale hadto be imagined.

MethodSubjects. Twent-four Dartmouth College undergraduate stu-

dents (16 of them female) from the introductor pscholog courseserved as listeners in the experiment in return for partial coursecredit. Their age was 20.76 2.9 ears (mean6 standard deviation).Three of them reported being left-handed. Two reported possessingabsolute pitch. Twelve listeners had at least 1 ear of formal musi-

cal training for voice or instrument, and among those, the amountof training was 7.16 4.1 ears. All the listeners provided informedconsent in accordance with the guidelines of the Committee for theProtection of Human Subjects at Dartmouth.

Stimuli. Tone sequences were ascending diatonic scales in fivemajor kes beginning on the following tonic notes: C4 (261.63 Hz),

Db4 (277.18 Hz), D4 (293.66 Hz), Eb4 (311.13 Hz), and E4 (329.63 Hz).Each note was 250 msec in duration, and the standard stimulusonset asnchron (SOA) was 600 msec. Each note was snthesizedin MATLAB from its fundamental frequenc and the next sevenhigher harmonics. The amplitude, A, of each harmonic was given

b the equation

A Ne

e

N

( ) ,

/=

1 1

1

whereNis the harmonic number (1 5 fundamental frequenc). Allharmonics had the same starting phase. A 5-msec linear ramp wasapplied to the envelope at the beginning and end of each note.

The pitch of a final probe note (an octave above the starting note)in the scale could assume one of seven values. It was either in tuneor mistuned b260,240,220,120,140, or160 cents. Similarl,the SOA between the penultimate and the final notes in the scale waseither the standard SOA or one of six deviant times:290,260,230,130,160, or190 msec. Deviance levels were chosen to fall arounddiscrimination thresholds commonl encountered in the literature.The final note never deviated from the standard either in pitch or intime. A unique sound file, containing the ascending notes and the

probe note, was snthesized for each degree of deviation in orderto avoid an possibilit of timing variabilit that might have beenintroduced b the stimulus presentation program when switchingfrom prime to target sound files.

Procedure. The listeners were seated in an double-walled sound-attenuating chamber (Industrial Acoustics Corporation) in front ofa computer monitor and made responses on a computer keboard.Stimuli were presented at a comfortable level through Son head-

phones. Presentation software (www.neurobs.com) running on aDell Latitude (Model C840) under Windows 2000 was used for ran-domizing and presenting the stimuli and recording responses.

On each trial, two cue words appeared on the screen to indicate thetask and trial tpes to the listener. The top word was either Time

or Pitch and cued one of the two tasks described below. The bot-tom word cued the trial tpe and was either Listen or Imagine.Note that throughout the article, we will refer to the dimension thatlisteners are making judgments on as the task tpe (time or pitch),and we refer to trial tpes (conditions) in terms of the mode of imageformation (listen or imagine). The cues appeared on the screen 1 sec

prior to the onset of the scale. The words remained on the screenthroughout the trial. In listen trials, the listeners heard ever noteof the scale, whereas in imagine trials, the heard either three notesand imagined five (n5 14 listeners) or the heard f ive and imaginedthree (n5 10 listeners). A final note was presented in the imaginetrials at the time it would have occurred had all the notes been pre-sented as in the listen trials.

The listeners were instructed on the tasks and continued to receiveshort blocks of practice trials until the felt that the understood boththe tasks. The listeners were also instructed to refrain from makingan movements or vocalizations that would help them keep timeduring the imager trials.

Listen and imagine trials alternated during blocks of trials inpitch and time tasks. Each task block consisted of 60 trials. Twopitch blocks were presented in alternation with two time blocks, andthe order was counterbalanced across subjects. In the pitch task,the listeners pressed one of two kes upon hearing the final note inthe scale, to indicate whether that note was congruous (in tune) orincongruous (out of tune). Similarl in the time task, the listeners

judged whether the final note was congruous (on time) or incongru-ous (either earl or late). In the imagine trials, the listeners indicated

whether the final note in the scale was congruous with their mentalimage of either the pitch or the onset time of that note. In each task,the listeners received 60 listen and 60 imagine trials, of which 50%were congruous (deviations of 0 cents or 0 msec). Thus, each levelof deviation awa from zero was experienced once for each ke ineach task. No feedback was given regarding response accurac.
http://www.neurobs.com/http://www.neurobs.com/


5/17


Upon completing all the trials, the listeners f illed out a question-naire about their musical background and about their perception ofthe difficult of the pitch and time tasks and a few general aspectsof their mental images. The questions and rating scales are given inthe Results section.

Data analysis. The data were analzed using MATLAB and SAS.Mixed model analses were implemented using PROC MIXED in

SAS for the pitch and time tasks separatel (Littell, Milliken, Stroup,& Wolfinger, 1996). The level of deviation, trial tpe (listen or imagine),and number of imagined notes (three or five) were treated as fixed ef-fects. Level of deviation and trial tpe were treated as within-subjectsrepeated measures. The number of imagined notes was a between-subjects variable. The model was estimated using the restrictedmaximum likelihood method and utilized an unstructured covariancestructure, since this covariance structure ielded better fit statistics(e.g., Akaikes information criterion) than did a variance componentscovariance structure. The Satterwaithe approximation was used to cal-culate degrees of freedom associated with fixed effects.

Results

Image tuning curves. Figures 1A and 1B show thatthe likelihood of judging a probe note to be in tune atan given degree of mistuning was comparable when thelisteners heard all of the notes leading up to the probenote and when the listeners had to imagine either three orfive notes (including the imagined note that was compared

against the probe). The statistical analsis indicated thatthe onl significant effect on the proportion of congru-ous responses was the main effect of level of deviation[F(6,22) 5 44.17, p, .0001]. Not surprisingl, probenotes that were in tune were judged as such 88% and 78%of the time in the heard and imagine conditions, respec-

tivel. A probe note that was tuned flat b 20 cents wasambiguous, with ~50% in-tune responses. When the probewas tuned sharp b 20 cents, the proportion of in-tune re-sponses approached 70%, suggesting that the listenersexpectations and images tended to be mistuned upward.The apparent presence of an upward asmmetr in boththe Imagine 3 and the Imagine 5 groups of subjects wastested with a contrast of the positive mistunings with thenegative mistunings. As a group, positive mistunings werejudged congruous 16% more often than were the corre-sponding group of negative mistunings [t(1,22) 5 5.41,p, .0001]. Of greatest importance was the absence of

significant interactions of trial tpe with level of devia-tion and trial tpe with number of imagined notes and ofa significant three-wa interaction of deviation, trial tpe,and number of imagined notes, indicating that the images/expectations that the listeners formed for the probe noteswere comparable in all the conditions. The onl other

Figure 1. Tuning functions for mental images in pitch and time. The height of each bar in a panel

gives the probability that a probe note at a particular deviation will be judged as in tune in the pitchtask (panels A and B), or on time in the time task (panels C and D). Gray bars show the judgments of

probe notes in the listen conditions when the listeners heard all of the notes leading up to the probenote. Black bars give probabilities for probe notes in the imagine conditions when the listeners either

heard five notes and imagined three (panels A and C), or heard three notes and imagined five (panels

B and D). Error bars indicate 1SEM.

60 40 20 0 20 40 60

60 40 20 0 20 40 60

0

.2

.4

.6

.8

1Imagine 3 Pitch

Imagine 5 Pitch

Imagine 3 Time

Imagine 5 Time

n = 10

90 60 30 0 30 60 90

90 60 30 0 30 60 90

n = 14

Deviation (cents)

ProportionCongru

ous

0

.2

.4

.6

.8

1

ProportionCongruous

0

.2

.4

.6

.8

1

ProportionCongru

ous

0

.2

.4

.6

.8

1

ProportionCongruous

Deviation (msec)

ListenImagine

A C

DB


6/17


outcome of the statistical analsis was an interaction be-tween level of deviation and number of notes imagined[F(6,22)5 3.46,p5 .015], which was probabl due toa greater proportion of congruous judgments in both thelisten and the imagine conditions at the extremes of thedeviation range (260,240, and160 cents) for the listen-

ers in the Imagine 5 group (Figure 1B). In other words, thelisteners in the Imagine 5 condition tended to form lessaccurate images/expectations overall.

The results from the time task differed markedl fromthose in the pitch task, as can be seen in Figures 1C and1D. The requirement to imagine either three or five notesconsiderabl impaired a listeners abilit to determinewhether a probe note occurred earl or late, when com-pared with having to determine whether the next note in asequence of notes arrived on time. Whereas the listenerswere reliabl judging notes with 60-msec deviation as ar-riving too earl or too late in the listen trials, the were

uncertain about whether notes were on time throughoutthe range of290 to 190 msec in the imagine trials, ir-respective of the number of notes to be imagined. Theseeffects manifest themselves as a significant main effectof trial tpe [F(1,22)5 42.35,p, .0001] and a signifi-cant trial tpe3 level of deviation interaction [F(6,22)524.05, p, .0001]. The main effect of level of deviationwas also significant [F(6,22)5 32.54,p, .0001], as wasthe interaction of trial tpe with number of imagined notes[F(1,22)5 5.22,p5 .0324].

Assessments of task difficulty and image clarity.Mixed model analses of the task difficult ratings (seeTable 1) conf irmed that the listeners found trials in whichthe had to imagine notes in the scale more difficult thanthose in which the heard all of the notes [pitch,F(1,21)525.46, p, .0001; time, F(1,21)5 23.16, p, .0001].Neither the main effect of number of imagined notes northe interaction of number of imagined notes with trial tpewas significant for either the pitch or the time task.

DiscussionAs was expected, the listeners readil detected mis-

tuned or mistimed targets when the target was the next

expected note in the scale. Interestingl, when having toimage a sequence of notes, their acuit differed dramati-call with the tpe of judgment the were asked to make.Intonation judgments were rendered with an accuraccomparable to that in the attentional-cuing task and wereunaffected b the number of notes that had to be imag-ined. However, timing judgments suffered greatl. In fact,the range of f ixed temporal deviations was insufficient tofind a tthat resulted in reliable incongruous ratings inthe imagine condition. This result, as well as an observa-tion that individual listeners differed considerabl in theiracuit, prompted us to modif the procedure and estimate

individual thresholds.

EXPERIMENT 2

In this experiment, we used a standard pschophsicaltwo-alternative forced choice (2AFC) staircase procedure

Ta

ble1

Subj

ectiveAssessmentsofTaskDifficulty

andImagePropertiesinExperiments1

3 Experiment1

Experiment2

Experiment3

Question

Imagine5

Imagine3

Imagine3

Imagine3

HowdifficultdidyoufindthePITCHtaskwhenyouHEARDtheentirescale(15v

eryeasy;7

5v

erydifficult)?

3.061.2

2.361.2

3.961.3

3.361.6

HowdifficultdidyoufindthePITCHtaskwhenyouIM

AGINEDthescale(15v

eryeasy;75

verydifficult)?

4.861.4

4.661.5

5.061.4

4.661.8

HowdifficultdidyoufindtheTIMEtaskwhenyouHEARDtheentirescale(15v

eryeasy;75v

erydifficult)?

3.361.5

2.861.2

2.961.8

3.261.4

HowdifficultdidyoufindtheTIMEtaskwhenyouIMAGINEDthescale(15v

eryeasy;75verydifficult)?

5.161.5

5.261.5

4.261.7

5.161.9

Howoftendidy

oufindyourselfsingingalonginyourm

indwiththeheardscales(15o

ften;75n

otoften)?

1.761.0

2.262.0

2.862.0

2.461.2

Howvividwere

yourimagesofthenotesduringtheimaginedscales(15v

eryvivid;75h

adnoimage)?

3.661.5

3.862.0

4.661.5

4.262.5

Ratetheeasewithwhichyoucouldmaintainthetempo/rhythmwhenyouwereimaginingthenotes(15v

eryeasy;75v

erydifficult).

3.561.5

3.861.3

3.061.2

4.261.5

NoteAllvaluesaregivenasmean6s

tandarddeviation.


7/17


for estimating image acuit. This procedure allowed us toestimate each listeners point of subjective equivalence interms of the offset of the image from perfectl in tune oron time, as well as the sharpness of the image in termsof the difference between the two threshold estimates foreach dimension in each condition. Thresholds were es-

timated simultaneousl for both directions of deviationalong each dimensionthat is, flat and sharp for pitch andearl and late for time in both the listen and the imagineconditions.

MethodSubjects. Fourteen Dartmouth College undergraduate students (8

of them female) from the introductor pscholog course served aslisteners in the experiment in return for partial course credit. Their age

was 20.1 6 1.2 ears (mean 6 standard deviation). All were right-handed. One reported possessing absolute pitch. Eleven listeners hadat least 1 ear of formal musical training for voice or instrument, andamong those, the amount of training was 7.2 6 5.1 ears. All the lis-teners provided informed consent in accordance with the guidelinesof the Committee for the Protection of Human Subjects at Dartmouth.

None of the listeners had participated in the previous experiment.

Stimuli. The stimuli were snthesized with the same parametersas those in Experiment 1. However, onl the ascending D-majorscale (starting at D4, 293.67 Hz) was used, and in the imagine con-ditions, the listeners heard five notes and imagined three. As in Ex-

periment 1, a set of sound f iles was created, with each sound filecorresponding to a level of deviation in either pitch or time. Between2300 and250 msec, temporal deviation steps were 10 msec. Be-tween 250 and 0 msec, the step size was 5 msec. The same stepsizes were used for deviations greater than 0. Pitch deviations ranfrom250 to150 cents in 2.5-cent intervals.

Figure 2. Tuning and timing thresholds for individual listeners judging

(A) the intonation of the tonic, which is the final note of an ascending majorscale and the most representative note of the key, or (B) the timing of the tonic.Gray bars are data from the listen condition, in which the listeners heard the

seven notes of the scale preceding the final note. Black bars are data from theimagine condition, in which listeners heard five notes of the scale and imagined

the remaining three. Error bars represent 1SEM, based on the six turnaroundpoints used to calculate each threshold.

400

300

200

100

0

100

200

300

400

Time(ms

ec)

Temporal Thresholds for the Tonic

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L1410080

60

40

20

0

20

40

60

80100

Listener

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14

Listener

Pitch

(cents)

Pitch Thresholds for the TonicA

B

ListenImagine

ListenImagine


8/17


Procedure. Thresholds were determined using a 2AFC task.Each trial had the following structure. The listeners saw the cuesspecifing the tpe of task (pitch or time judgment) and the modeof imager (listen or imagine) and heard the f irst of two sequences.One sequence was the standard sequence in which the eighth note ofthe ascending scale was in tune and on time. The other was a probesequence in which the eighth note was either mistuned or mistimed,

depending on the task modalit of the current trial. As in Experi-ment 1, the listeners heard all eight notes of the ascending scale inthe listen conditions, whereas the heard five and imagined three inthe imagine conditions, with a probe note occurring at or around thetime at which the eighth note should have occurred (4,200 msec).Immediatel upon termination of the first sequence, 1 of 10 dif-ferent tonal masks was plaed. The tonal masks were random se-quences of four tones within one semitone of the standard probe

pitch (587 Hz). Each tone had the same timbre as the notes in thetask sequences, was 250 msec in duration, and was plaed once. Thesecond sequence began 1 sec following the termination of the mask.After hearing the second sequence, the listener had to specif, b

pressing one of two buttons, which of the sequences contained themistuned/mistimed probe note. The order of the standard (in-tune or

on-time) sequence and the probe sequence was randomized.The degree of deviation of the probe note was adjusted using a

slightl modified two-down/one-up staircase procedure according tothe following rules. The initial deviation was set to the maximum ofthe rangefor example,6300 msec or650 cents. In order to bringthe listeners to a range around their thresholds more quickl, theamount of deviation was reduced on each trial b one step until twoerrors were made. At this turnaround point, the amount of deviationwas increased b one step on each trial until a correct response wasmade. Two successive correct responses resulted in a one-step reduc-tion of the deviation. The procedure continued until 10 turnaroundshad been accomplished in each condition.

Conditions were randoml interleaved across trials so that thresh-

olds could be estimated for both listen and imagine in both the timeand the pitch tasks in parallel. As in Experiment 1, the listeners wereinstructed on the tasks and received short blocks of practice trialsuntil the felt that the understood the tasks. The listeners were in-structed to refrain from making an movements or vocalizations thatwould help them keep time during the imager trials. The listenerswere given the opportunit to take a break ever 55 trials before con-tinuing. The total number of trials varied across listeners (264354)

but was ~300 on average. The procedure lasted approximatel 1.5 h.The listeners then completed a questionnaire, as in Experiment 1.

Data analysis. Acuit thresholds were defined for each subjectas the average of the deviation values at turnaround points 510.Two properties of the mental images were of particular interest.The first was the width of the mental image. The width was definedas the difference between the positive and the negative thresholds.

The second was the offset of the center of the mental image tuningcurve, defined as the average of the positive and negative thresholds.The question of whether image acuit, as defined b the width, re-mained constant between the listen and the imagine conditions wasaddressed using a pairedttest. Similarl, the question of whether thecenter of the mental image differed between the two conditions wasassessed with a pairedttest comparing the listen and the imagineconditions. Finall, one-sample ttests were performed separatelfor the listen and the imagine conditions, testing whether the averagecenter of the image was equal to a deviation of zero or whether thelisteners as a population tended to sstematicall shift their mentalimages in one direction or another. All ttests were two-tailed.

Results

The width of the pitch component of the images did notdiffer significantl between the listen and the imagine con-ditions [t(1,13)5 1.2, n.s.; see Figure 2A and Table 2].The average widths of the pitch images were 30 cents inthe listen condition and 33 cents in the imagine condition.Given a fundamental frequenc of 587 Hz for the target,these widths correspond to 10.3 Hz in the listen conditionand 11.2 Hz in the imagine condition. Thus, the amount ofmistuning at threshold in both conditions was about 1%of an in-tune target. The average offset of the image in theimagine condition was 4.3 cents more positive (sharp) than

zero [t(1,13) 5 2.5,p, .03; see Table 3]. The offset inthe imagine condition was significantl more positive (b3.4 cents) than was the center of the image in the listencondition [t(1,13)5 3.6,p, .004]. Figure 2A illustratesthat the individual listeners varied in their tendenc toimagine sharp or flat. For example, L2s images were vernarrow in both the listen and the imagine conditions but

Table 2Image Widths in Experiments 2, 3, and 4

Pitch Time

Experiment Listen Imagine Listen Imagine

Experiment 2: Eighth note (tonic) 306 15 336 20 516 18 2866 71Experiment 3: Seventh note (leading tone) 306 14 476 24 706 12 2646 74Experiment 4: Tonic 786 40

Experiment 4: Leading tone 1076 32

NoteAll values are given as mean6 standard deviation. Pitch values are given in cents. Timevalues are given in milliseconds.

Table 3

Image Offsets in Experiments 2, 3, and 4

Pitch Time

Experiment Listen Imagine Listen ImagineExperiment 2: Eighth note 0.96 4.4 4.36 6.2 46 10 96 40Experiment 3: Seventh note 2.66 9.0 4.46 6.8 96 6 206 34Experiment 4: Tonic 20.96 15.0Experiment 4: Leading tone 4.06 13.0

NoteAll values are given as mean6 standard deviation. Pitch values are givenin cents. Time values are given in milliseconds.


9/17


consistentl flat, whereas L4s images were much broaderand consistentl sharp. Remarkabl, the images of one lis-tener (L12) were more precise in the imagine condition.

Unlike pitch images, temporal images were much widerin all the listeners when the notes leading up to the probenote had to be imagined than when the were heard (see

Figure 2B and Table 2). The average image width increasedb 235 6 22 msec (mean 6 standard error) from 51 to286 msec. Despite the increase in image width, the offsetof the image in the imagine condition was neither earliernor later than that in the listen condition [t(1,13)5 0.4,n.s.]. When compared with the canonical onset of the probenote, neither of the measured offsets differed significantl[listen, t(1,13)5 1.7, n.s.; imagine, t(1,13)5 0.9, n.s.;Table 3]. Thus, although the listeners were much less ac-curate in judging when the probe note was to occur whenasked to imagine the preceding notes, as a group, the didnot sstematicall move their image or expectation earlier

or later in time. Nevertheless, inspection of the individuallistener data in Figure 2B suggests that the temporal imagesof some listeners were biased toward earlier times (e.g., L1and L3), whereas those of others were delaed (e.g., L9 andL14). In addition, there was intersubject variabilit in thewidth of the temporal component of the images, with somelisteners (e.g., L7 and L12) showing higher acuit than didothers (e.g., L5 and L6).

DiscussionThe results of Experiment 2 confirmed the observa-

tion in Experiment 1 that the temporal component of theimages suffered greatl when the listeners were asked toimagine several notes preceding the target event, whereasthe images along the pitch dimension remained tightlfocused. The temporal thresholds in the listen conditionare in direct agreement with thresholds for detecting de-viations from isochron at similar interstimulus intervals(Ehrle & Samson, 2005).

EXPERIMENT 3

In order to facilitate image formation in the present set

of experiments, we chose to use ascending scales, becausethe are relativel simple and familiar, et the force lis-teners to move their mental images in pitch space. Whenascending scales are used, a convenient location aroundwhich to place probes is the tonic, which serves as boththe starting note and the final note (an octave above thestarting note). The tonic has the propert that it is the mostrepresentative member of major and minor kes (Krum-hansl, 1990) and, when primed with a scale, tends to becategorized most quickl and accuratel as belonging tothe ke (Janata & Reisberg, 1988). Thus, in Experiments1 and 2, the listeners ma have been evaluating the tuning

of the probe note relative to a mental image of the tonicthat was primed b the heard notes in the scale, rather thanarriving at the mental image of the tonic b imagining allof the notes leading up to it. In Experiment 3, we soughtto eliminate the potential use of this alternative strateg b

terminating the sequence with a probe note on the seventhdegree (the leading tone) of the scale. The leading tone isregarded as one of the least characteristic members of amajor ke, and its representation is, therefore, not stronglprimed b the first five notes of the ascending scale. Weexpected that if the listeners were not forming accurate

mental images of each note remaining in the scale but, in-stead, reling on a primed tonal context, the tuning curveassociated with probes around the seventh degree of thescale would be broader and shifted upward in pitch, incomparison with images formed for the eighth and morestable degree of the scale. An upward shift toward thetonic might arise via biasing of the leading tones repre-sentation b the strongl primed tonic to which the lead-ing tone tends to resolve. In other words, if across mantrials numerous conflicts arise between the pitch that isto be imagined and its upper neighbor that has a strongerrepresentation based on the context, the outcome might be

an image that is both broader and shifted upward.

MethodSubjects. Fourteen members of the Dartmouth communit (7

of them female) served as listeners in the experiment and received$10/h for their participation. Their age was 23.16 6.5 ears (mean6 standard deviation). All were right-handed. Ten listeners had atleast 1 ear of formal musical training for voice or instrument, andamong those, the amount of training was 8.1 6 4.7 ears. All thelisteners provided informed consent in accordance with the guide-lines of the Committee for the Protection of Human Subjects atDartmouth. None of the listeners had participated in either of the

previous experiments.

Stimuli. The stimuli were identical to those used in Experiment 2,with the exception that the note sequences ended on the note C #5,the seventh degree of the D-major scale. Thus, probe notes varied in

pitch around 554.37 Hz.Procedure. The procedures used for obtaining pitch and temporal

acuit thresholds were the same as those in Experiment 2, and thelisteners completed a questionnaire as in Experiment 1.

Data analysis. The data were analzed as in Experiment 2. Inaddition, a mixed model analsis was implemented in SAS in orderto assess an differences between Experiments 2 and 3 in terms ofimage widths and offsets. Narrower pitch images in Experiment 2might indicate that the listeners benefited from a coincidence ofthe probe pitch with a primed representation of the tonic. Further

evidence that the listeners relied on a primed representation of thetonic, instead of generating the correct sequence of mental pitchimages, when making probe note judgments would be a dramatic

positive offset (on the order of one semitone) of pitch images inExperiment 3. The parameters of the mixed model analsis werethe same as those in Experiment 2. One listeners (L13) thresholdswere considerabl worse than those of all the other listeners in all theconditions (see Figure 3), so this persons data were excluded fromthe statistical analses.

ResultsIn contrast to Experiment 2, pitch images were signifi-

cantl broader in the imagine condition than in the listen

condition [16.56 4.9 cents; t(12)5 3.35,p5 .0058; seeFigure 3A and Table 2]. The offset of the pitch image didnot differ significantl across the two conditions [1.8 61.9 cents; t(12)5 0.92, n.s.], although the image was sig-nificantl sharper than zero in both the listen [2.6 6 0.8


10/17


cents; t(1,12)5 3.18, p5 .0095] and the imagine [4.46 2.0 cents; t(1,12)5 2.2,p5 .0480] conditions. As inExperiment 2, there was considerable variabilit in imageacuit across listeners.

As in Experiment 2, temporal images were significantlbroader (mean difference5 1956 21 msec) in the imag-ine than in the listen condition for all the listeners (seeFigure 3B and Table 3). Although the center of the tempo-ral images did not differ significantl between conditions[11.46 msec; t(1,12) 5 1.01, n.s.], the expectation for a

slight, 9.5-msec dela in the listen condition was signifi-cant [t(1,12)5 5.02,p5 .0003], and there was a tendenc,albeit more variable, toward the same in the imagine condi-tions [19.9 msec; t(1,12)5 2.05,p5 .0625; see Table 3].

The combined analsis of pitch image width estimatesfrom Experiments 2 and 3 identified a significant main ef-

fect of imager condition [F(1,25)5 13.19,p5 .0013],in which images were broader, on average, in the imaginecondition b 9.6 cents. This increased width was attribut-able to the imagine condition in Experiment 3, as suggestedb a significant interaction between imager condition andterminal note [F(1,25)5 6.87,p5 .0147] and the pairwisecomparison between imager conditions in Experiment 3noted above. Neither a pairwise comparison of imagewidths between Experiments 2 and 3 for the imagine con-dition nor one for the listen condition showed a significant

difference between the two experiments. Thus, the broaden-ing of the image width when the leading tone is imagined israther modest. Pitch images in the imagine condition weremore sharp b 2.5 cents [F(1,25)5 6.25,p5 .0193], butthere was neither a main effect of target note nor an interac-tion of target note and trial tpe.

Figure 3. Tuning and timing thresholds for individual listeners judging

(A) the intonation of the leading tone, the penultimate note of an ascendingmajor scale, or (B) the timing of the leading tone. In the listen condition (gray

bars), six notes were presented prior to the target note, whereas in the imaginecondition (black bars), four notes were presented and three were imagined.

Error bars represent 1SEM, based on the six turnaround points used to calcu-

late each threshold.

400

300

200

100

0

100

200

300

400

Time(msec)

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14100

80

60

40

20

0

20

40

60

80

100

Listener

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14

Listener

Pitch

(cen

ts)

Temporal Thresholds for the Leading Tone

Pitch Thresholds for the Leading ToneA

B

ListenImagine

ListenImagine


11/17


As was expected, the combined analsis of offsets inthe temporal component showed no significant effects,and the analsis of temporal image width showed onla highl significant main effect of trial tpe [F(1,25) 5208.21,p, .0001], but neither a main effect of target notenor an interaction of target note and trial tpe.

Finall, we calculated the correlations between the dura-tion of musical training and each of our dependent measuresof auditor image acuit. Musical training significantlinfluenced the acuit of auditor images, primaril alongthe pitch dimension (Figure 4A). Pooled across the 22 lis-teners in Experiments 2 and 3 for whom musical training

data were available, the width of the image was negativelcorrelated with ears of musical training in both the lis-ten conditions [r(20)52.74,p, .0001] and the imagineconditions [r(20)52.64,p5 .0004]. Although the widthof temporal images was uniforml good when the listenersheard all of the notes preceding the target, more musicalltrained listeners again showed significantl tighter thresh-olds when asked to imagine three notes [r(20)52.42,p5.0320; see Figure 4B].

DiscussionExperiment 3 replicated the results of Experiments 1

and 2, showing that the temporal component of auditorimages is considerabl less precise when multiple notesin a scale must be imagined than when the timing of thenext event in an isochronous note sequence is to be evalu-ated. The small but significant broadening of images inthe pitch dimension in the imager task suggested thatthe listeners had a more difficult time forming an accu-rate image of the leading tone than of the tonic. Becausetheir images of the leading tone were not biased upwardin pitch toward the tonic, it is unlikel that the broaderthresholds arose from a blending or confusion of an imagefor the leading tone with a primed image for the tonic. Itis possible that the process of imagining a sequence ofpitches is inherentl associated with increased variancethat results in a broader mental image for the final pitchin the sequence than with the expectanc associated withthe next heard pitch. Such an interpretation would sug-gest that the slightl narrower width of the image for thetonic reflects a facilitation of image acuit b the primedtonal context. Such an account needs further support bshowing that image acuit broadens similarl for othernotes surrounding the tonic, such as a continuation of thescale to the supertonic. However, previous work shows a

context-dependent facilitation of tuning judgments whenthe target pitch is at the end of a familiar melod and, to alesser extent, at the end of a repeating or alternating tonesequence, in comparison with judgments made on isolatedpairs of tones (Warrier & Zatorre, 2002).

EXPERIMENT 4

We performed a final experiment to test whether thecomparable acuit of pitch images in the imager condi-tions of Experiments 2 and 3 might be attributable solelto priming of the ke to which the probes belonged. Given

that the leading tone is judged to be much less stronglassociated with a primed ke than is the tonic (Janata& Reisberg, 1988; Krumhansl, 1990), it is unlikel thatthe comparable image acuit for tonic and leading toneprobes in the imager conditions in Experiments 2 and 3is attributable to priming effects alone. However, to assess

Figure 4. Pitch and temporal image acuity improve with music

training. For the 22 listeners in Experiments 2 and 3 for whom thenumber of years of musical training was known, the duration of

musical training was positively correlated with smaller thresholdsin both the listen and the imagine conditions for pitch (panel A).

Temporal images showed an improvement with musical training

only in the imagine condition (panel B). The solid line is the slopeof the regression line from the listen conditions, and the dashed

line is the slope for the imagine conditions.

0 2 4 6 8 10 12 14 16 18

10

20

30

40

50

60

70

80

90

Training (years)

Training (years)

ImageWidth(cents)

ImageWidth(msec)

y= 43.3 2.1x; r= .74

y= 57.7 2.9x; r= .74

ListenImagine

ListenImagine

0 2 4 6 8 10 12 14 16 18

50

100

150

200

250

300

350

400

y= 63.9 0.77x; r= .24

y= 314.9 5.8x; r= .42

Pitch

Time

A

B


12/17


more thoroughl the relative influence of tonal primingversus imager instructions on the listeners intonationjudgments, we performed a within-subjects comparisonof tonic and leading tone probe judgments, using slightlmodified versions of the stimulus materials and task.

Listeners performed a 2AFC task in which the heard

the initial four notes of an ascending major scale anddiscriminated among the same set of pitch probes, as inExperiments 2 and 3. However, both the stimuli and thetask differed in one regard. First, the probe alwas oc-curred earlier, with the intent of rendering a strateg ofisochronous image formation of the remaining notes use-less. Second, because we wanted to discourage an ex-plicit strateg of imager, we gave the listeners a contextmembership task, in which the had to determine whichof the two probe items was a note from the ke (context)that was suggested/primed b the first four notes of thescale. Since the in-tune probe is technicall a member of

the primed context, whereas a mistuned probe is not, thetask serves to estimate the degree to which the intonationjudgments in Experiments 13 were based on priming ofthe ke alone.

We were interested in examining two questions. First,would the image widths of the leading tone and the tonicdiffer from one another? The hpothesis that each probeis primed equall b the context would have to be rejectedif the two differed. Second, in order to assess an rela-tive benefit of an explicit instruction to imagine the re-maining notes in the scale over no imager instructionwhatsoever, we compared the image widths obtained inExperiments 24. If our results in Experiments 2 and 3were driven primaril b priming of the ke, image widthsshould not change in this experiment. However, if the useof an imager strateg is explicitl supported b both taskinstructions and the stimulus structure (as in Experiments2 and 3), we would expect image widths to broaden whenimager instructions are absent and the assumptions aboutwhen the probes should occur in an isochronousl pre-sented scale are disrupted.

MethodSubjects. Twent-two members of the Universit of California,

Davis communit (13 of them female) served as listeners in the ex-periment and received either course credit or $10/h for their par-ticipation. One of the listeners did not complete the experiment, onaccount of fatigue, and another was excluded on account of hearing

problems. The remaining listeners ranged in age from 18 to 26 ears[20.6 6 2.0 ears (mean 6 standard deviation)]. Nineteen wereright-handed. Fifteen listeners reported having undergone at least1 ear of formal musical training for voice or instrument (range,215 ears), and among them, the amount of training was 6.3 64.4 ears. All the listeners provided informed consent in accordancewith the guidelines of the Institutional Review Board at the Univer-sit of California, Davis. None of the listeners had participated inan of the previous experiments.

Stimuli. The stimuli used were identical to those used in thepitch conditions in Experiments 2 and 3, with the exception thatthe probes were heard 1,600 msec following the onset of the final(fourth) note in the prime. With respect to an isochronous continu-ation of the scale, the probe occurred both considerabl earlier thanwould be expected and 200 msec earlier than the sixth note of thescale would have been heard. Both of these features of probe timing

were intended to disrupt the imager strateg that was encouragedin Experiments 13.

Procedure. The listeners completed two warm-up exercises be-fore we obtained thresholds. First, in order to familiarize them witha 2AFC task structure, we had the listeners perform eight trials inwhich each interval had two pitches about which the madesame/differentjudgments in order to indicate which interval contained

two different pitches. The frequenc difference varied between 35and 50 cents, in order to make it readil detectable. Once the listen-ers were able to make the discriminations reliabl, the performedeight practice trials of the actual task. The listeners were instructedthat each trial was split into two parts and that, in each part, thewould hear the first four notes of a major scale and then a probe thateither did or did not belong to the same ke/context that the f irst fournotes belonged to. Their task was to indicate which part of the trialhad the probe note that did not fit into the context. The pressed a

button with their left or right index finger to indicate whether thethought that the out-of-context probe was in the first or the second

part, respectivel. Once the listener felt comfortable with the taskstructure, we proceeded to obtain the upper and lower thresholdsfor both the leading tone and the tonic simultaneousl. Following

the threshold estimation procedure, the listeners completed a briefquestionnaire to probe their strategies and awareness of differentaspects of the task.

In earlier tests of the procedure, we noticed that some trajectoriesof pilot subjects did not converge toward the probe notes but trav-eled, instead, to the boundar of one semitone that we had set forour stimuli. In such cases, the threshold estimation procedure forthat particular threshold was terminated if the subject was stuck atthe boundar for three trials.

Data analysis. The data were analzed using mixed model anal-ses of image width and offset, as in the previous experiments.

ResultsIn accordance with our predictions, Figure 5 and Table 2

illustrate that the change in stimuli and task resulted in a sig-nificant decrement in image acuit for the leading tone, rel-ative to the tonic [306 6.8 cents; t(18)5 4.33,p5 .0004].A comparison of tonic and leading tone image widths inExperiment 4 with their counterparts in Experiments 2and 3, respectivel, indicated that image widths were sig-nificantl larger in Experiment 4 for both probes [tonic, 446 11 cents; t(44)5 4.03,p5 .0002; leading tone, 60 611 cents; t(57)5 5.49,p, .0001]. In 14 of 20 listeners inExperiment 4, image widths were considerabl larger for

the leading tone than for the tonic. The effect of tpe of taskon the relative image widths for the leading tone and tonicwas captured in a highl significant experiment 3 probetpe interaction [F(4,33.9)5 78.06,p, .0001]. An anal-sis of the image offsets indicated no significant differencesfrom zero or between probe tpes (Table 3).

Most listeners (17/20) were aware that there were twodifferent probe tones. When asked which probe tone(lower or higher) fit better with the preceding context, 4of the 17 indicated that the leading tone fit better, and13 indicated that the tonic did. This difference in ratingsfits with standard results in the tonal hierarch literature

(e.g., Krumhansl, 1990). Ratings of differential difficultof performing the task for each of the probe notes on a7-point scale (1 5more difficult for lower probe, 4 5no difference, 7 5more difficult for the higher probe),showed no sstematic difference between the two probes[3.9 6 0.6; t(16)52.48, n.s.], although the distribution


13/17


was bimodal, with 9/17 listeners giving a rating of 1 or 2

and 6/17 listeners giving a rating of 6 or 7. When askedto rate how confusing the task was (1 5not at all con-fusing, 4 5somewhat confusing, 7 5very confusing),the listeners gave it a rating of 2.6 6 0.4. Additional evi-dence that the tonic served as a stronger perceptual mag-net, even when mistuned, comes from the observationthat the upper thresholds for the leading tone migrated tothe upper boundar of mistuning in 50% of the subjects,reaching one semitone (the tonic) in all of the cases (8/10)in which the maximal mistuning was one semitone. Thus,in the trials in which the level of upward mistuning of the

leading tone had almost reached the tonic, the listenersdirectl compared the relative stabilit of the leading toneand tonic with respect to the priming context and judgedthe tonic to be more stable. In contrast, onl 2 listenersmigrated to the upper boundar when evaluating the tonic.With regard to flat mistunings, 3 listeners migrated to thelower boundar for both the leading tone and the tonic.

We also queried the listeners about their use of imagerin performing the task, asking them, As the trials went b,did ou find ourself imagining the probe note in our mindbefore ou actuall heard it? (15yes, always, 45some-times, 75no, never). The mean response was 2.6 6 0.4,

indicating that man listeners found themselves tring toform images of the probe notes. When asked to rate howvivid their mental images were (1 5very vivid, 7 5hadno image), the mean response was 3.56 0.4. When askedwhether the had imagined all of the missing notes of thescale or just the single probe note, 41.2% answered that

the had imagined all of the missing notes, whereas 58.5%

responded that the had imagined onl the probe note.As in Experiments 2 and 3, the amount of experience

with plaing a musical instrument had a significant ef-fect on image widths (Figure 6), although there was a sig-nificant interaction of probe and amount of experience[F(1,18) 5 4.82,p, .05]. The interaction was due to asignificant negative correlation between image width andamount of musical experience for the tonic (R25 .20,p,.05), but not for the leading tone (R25 .01, n.s.). A mixedmodel evaluating image widths across Experiments 24and using experience as a covariate showed an overall sig-

nificant main effect of experience [F(1,44)5

11.69,p5

.0014] and an experience 3 probe interaction [F(1,21)55.52, p5 .0285] but no experience 3 experiment in-teraction [F(2,47) 5 1.62, n.s.]. The fact that the latterinteraction was not significant is important insofar as itprovides some reassurance that the differences in imagewidths between Experiments 2 and 3 and Experiment 4are not attributable to differences between Dartmouth andUniversit of California, Davis students. Accounting forthe variance associated with experience did not change thesignificance of the outcome of an pairwise comparisonbetween tonic and leading tone probes.

DiscussionA comparison of the results from Experiment 4 with

those from Experiments 2 and 3 served to assess the rela-tive influence of contextual priming versus imager in-structions on the width of tuning of mental pitch images

Figure 5. Tuning thresholds for individual listeners who made key membership judgments sepa-rately for the leading tone and the tonic, relative to their mistuned counterparts. In all cases, the

listeners heard the initial four notes of an ascending major scale in order to prime the key. No im-agery instructions were given. Gray bars are data from the condition in which the listeners made

judgments about the tonic (eighth) scale degree, and black bars are data from judgments about the

leading tone (seventh) scale degree. Asterisks indicate listeners who reported using imagery some-times to never. All of the other listeners reported forming images at least sometimes.

L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14 L15 L16 L17 L18 L19 L20100

80

60

40

20

0

20

40

60

80

100

Listener

Pitch

(cents)

Pitch Thresholds

Tonic

Leading tone

* * * * * *


14/17


in the context of tuning judgments about probe tones. Thelarge increase in image widths for both the tonic and theleading tone in Experiment 4 indicated that priming of atonal (ke) context could not explain the image acuit inthe earlier experiments. Because we manipulated aspectsof both the task and the stimulus, it is interesting to con-sider the possible causes of the pattern of image acuitdecrements we observed.

One possible reason is the absence of explicit imagerinstructions. Interestingl, most listeners reported fre-quentl attempting some imager strateg, although thedata in Figure 5 suggest that relativel few listenersforexample, L8, L1012, and L15benefited from it. Whatthe dissociation between listener reports and the measuredthresholds nicel illustrates is that the qualit of imager

can be highl variable. In other words, not all attempts atmental imager are equall effective. Even if an imag-er strateg is adopted, thus effectivel turning the taskinto a tpe of imager task regardless of the experimenterinstructions, certain conditions must be met in order toachieve good image qualit.

One stimulus factor that ma have hampered the listen-ers in their attempt to imagine the probe was the temporalincongruit of the probe. The probe arrived earlier thaneither the leading tone or the tonic would have in an iso-chronous continuation of the scale. Even though Experi-ments 13 showed that temporal acuit was considerabl

more impaired than pitch acuit under imager conditions,suggesting that temporal precision in an image ma not bea prerequisite for forming an accurate pitch image, a grosstemporal violation such as that in Experiment 4 might,nonetheless, interact with the abilit to form images ofpitches that normall would have occurred at specific

later times. Such structural interactions of tonal contextand time were documented b Jones and Boltz (1989) inan experiment in which listeners extrapolated the endingtime of melodies missing the final 15 notes. Extrapo-lated ending times could be induced to deviate from theexpected time (1,200 msec) as a function of the preceding

temporal and harmonic accent structure, indicating thatthe perceived accent structure could shape expectanciesreaching several events into the future. In our situation, theprobe violated both the isochronous context and the pitchthat would normall have been expected around that timeif the ascending scale had been continued normall.

Image acuit ma have been impaired also b theunpredictabilit of which scale degree (leading tone ortonic) would be the focus of an given trial. If the listen-ers formed an image of a probe, it ma not have been animage of the correct probe for that trial. A mismatch be-tween imagined probe and actual probe on the first part of

the 2AFC trial might be expected to introduce uncertaintinto the judgment of whether or not the probe that actuallsounded was part of the primed ke, which, in turn, couldmanifest itself as a broader image.

A different result from Experiment 4 is consistent withresults that would be predicted b one of two priming ac-counts. Our observation that mental images associatedwith the leading tone were significantl less focused thanthose associated with the tonic is in accordance with probetone studies of tonal hierarchies that have shown the lead-ing tone to be judged as less strongl associated with a kepriming context than is the tonic (e.g., Janata & Reisberg,1988; Krumhansl, 1990). Specificall, we find that whenconditions for accurate image formation are not optimaland contextual priming effects can presumabl dominate,judgments about the ke membership of mistuned scaledegrees are more precise for the tonic, the more stablenote in the hierarch. One must point out, however, thatthe results could also be explained b repetition priming,arising from the fact that the pitch class of the tonic washeard in the priming context, whereas the pitch class ofthe leading tone was not heard (cf. Oram & Cudd, 1995).Because we were interested in keeping the stimuli in Ex-

periment 4 as similar as possible to those in Experiments2 and 3, we did not compare the acuit of two pitch classesthat were not part of the priming sequence but that dif-fered in their positions in the tonal hierarch.

GENERAL DISCUSSION

We investigated the acuit of mental auditor images,using pschophsical threshold estimation procedures ina series of four experiments. Experiments 13 directlcompared the acuit ofexpectancies for immediatel fol-lowing auditor events with the acuit of the last image

in a sequence of auditor images retrieved from memorfollowing an acoustic cue. In these experiments, we foundthat the acuit of auditor images depended on the di-mension of them that the listeners were asked to judge.Whereas the listeners formed images for pitch that were asaccurate when the imagined two notes preceding a target

Figure 6. Correlation of musical training and tuning curves forjudgments about the tonic (solid regression line) and leading tone

(dashed regression line). Increased musical training results in sig-

nificantly better performance when the tonic is judged.

0 5 10 150

20

40

60

80

100

120

140

160

180

200

Training (years)

ImageWidth(cents)

y= 95.9 3.9x

y= 110.0 0.61x

TonicLeading tone


15/17


as when the heard all of the notes preceding a target, theirimages in time were impaired when imagining of multiplenotes preceding the target was required. B having thelisteners perform both an attentional-cuing task (the listenconditions in Experiments 13) and an imager task (theimagine conditions in Experiments 13), we were able to

provide evidence that the mental representation of pitch atthe point of comparison with an incoming acoustic stimu-lus is practicall equivalent in the two tasks. Absent directneurophsiological data, we cannot prove that the twomental representations are the same in terms of their ac-tivation patterns in the neural substrates that are requiredfor the discrimination judgment. However, we have re-duced the need to posit that pitch expectations and pitchimages differ. Indeed, the idea ofimages as anticipationsput forward b Neisser (1976) provides a unifing frame-work for our data and for thinking about the relationshipof attention and imager tasks. Experiment 4 served to

illustrate that the listeners in Experiments 13 were notbasing their tuning decisions purel on the relationshipof the probe note to the sense of ke that was establishedb the priming sequence. Rather, the results indicate thatsome combination of explicit knowledge about what theprobe would be, its approximate timing, and instructionsto imagine all of the notes leading up to the probe werenecessar for increased image f idelit.

Although our tasks and pitch acuit data fit well withNeissers (1976) mental image as anticipation framework,the divergence of pitch and temporal acuit across theattentional-cuing and imager contexts presents more ofa puzzle. In other words, wh do we accuratel anticipatethe pitch but not necessaril the exact moment at whichit will occur? Neuropschological data suggest that thesedimensions are not alwas tightl bound. Amusicsstroke patients or neurologicall intact individuals withpronounced impairments on tasks requiring fine-grainedpitch judgmentsshow better performance, sometimeseven on par with control listeners, on temporal judgments,relative to pitch judgments, thereb indicating that at leastin such individuals, the neural substrates supporting tem-poral and pitch judgments differ (Aotte, Peretz, & Hde,

2002; Hde & Peretz, 2004; Peretz et al., 2002; Peretz &Kolinsk, 1993). It is, therefore, not surprising that mentalimages formed on each dimension can show different tun-ing characteristics, even if listeners are instructed to imag-ine objects in which these dimensions are integratedforexample, the notes of a melod.

Our data lead us to conclude that images in time aremore susceptible to distortion in the absence of exter-nal stimuli than are pitch images. The broader underl-ing issue is the manner in which timing mechanisms inthe brain interact with control processes, such as atten-tion or working memor, that are implicated in voluntar

image formation, as well as the behavioral measures usedto examine the timing process. Timing processes in thebrain have been examined in a variet of different con-texts (Block & Zaka, 1997; Ehrle & Samson, 2005; Ivr,1996; Jones, 1993; McAule & Jones, 2003; Schner,2002; Wearden, 2003; Wright & Fitzgerald, 2004), but a

detailed discussion of how each of those paradigms mightmap onto the tasks we used is beond the scope of thisdiscussion. We will focus, therefore, on three possibilitiesfor the divergence of temporal acuit that we observedacross our listen and imagine conditions.

In the first, the abilit of isochronous tone sequences

to automaticall induce expectations for events to occurat the established interonset interval is examined. Pitch-matching judgments in a working memor task are moreaccurate if the target pitch occurs at an interonset intervalthat is established b a sequence of task-irrelevant distrac-tor tones presented with an isochronous rhthm (Jones,Monihan, MacKenzie, & Puente, 2002). Even if the lastpacing note is omitted, deviations as small as 15 msecfrom the established period of 600 msec result in impairedpitch-matching performance, indicating that the tempo-ral expectation persists for at least one extra period. Thisresult supports a model in which isochronous stimulus

sequences serve to orient attention in time b inducingan oscillator process that entrains to the regularit ofthe note sequence preceding the target (Barnes & Jones,2000; Large & Jones, 1999; McAule & Jones, 2003).Our data in the listen condition are entirel consistent withthis account and with the temporal precision observed bJones and colleagues.

In order to juxtapose our imager results against an at-tentional oscillator context, it is important to point to adistinction between bottom-up and top-down temporalexpectations suggested b Barnes and Jones (2000). Thecrux of the distinction is that entrained attentional processesthat give rise to temporal expectations depend on externalinput. Thus, these temporal expectations are the result of apassive, stimulus-driven, bottom-up process, rather than atop-down process guided b voluntar orienting of atten-tion. Indeed, in their tasks, there was no requirement forsubjects to generate either covert or overt events during thetemporal expectation-inducing isochronous sequence, sothere is presumabl relativel little top-down influence onthe expectations. Our imager task, on the other hand, canbe thought of as a top-down process, because it involves theexplicit formation of images for each event that remains in

the sequence. Further experiments will be needed to deter-mine how internall driven temporal expectations fit into adnamic attending framework.

A second explanation for the deterioration of the acuitof temporal images in our experiments takes into consid-eration the sensorimotor constraints we placed on the lis-teners. We explicitl instructed them to refrain from mak-ing an movements, including vocalizations, that wouldhelp them keep time. Although we did not videotape thelisteners or obtain mographic recordings to objectivelverif their compliance with our instructions, there is noevidence that an of them benefited from covert time-

keeping movements the ma have made. It would be ofinterest to know to what extent the sensorimotor feed-back that would be provided b allowing listeners to taptheir fingers or feet would assist in maintaining temporalacuit. Studies of snchronization and continuation tap-ping provide a partial answer to this question. During the


16/17


production of 580-msec intervals, the standard deviationof interresponse intervals is between 15 and 20 msec inboth snchronization and continuation conditions, andthe mean interresponse interval in the continuation condi-tion is a modest 10 msec shorter than the target interval(Semjen, Schulze, & Vorberg, 2000). The relative similar-

it in timing variabilit between externall and internallguided tapping behavior, when contrasted with the largedifferences in time estimation observed in our experi-ments, suggests that the sensorimotor component of thetapping tasks is essential to maintaining the accurac ofthe internal timekeeper across multiple mentall gener-ated events when an external pacing signal is absent. Thisaccount is also consistent with the notion that attentionaloscillators ma require some form of external input tomaintain a precisel timed attentional focus.

Finall, the relative difficult in generating mental im-ages in time with the restrictions we placed on our listen-

ers might have also biased them toward using a differentimager strateg. Specificall, given that the in-tune probenotes in Experiments 13 were unique and, therefore, fullpredictable, the listeners ma have formed a memor ofthe prime and probe that the could use to accuratel placetheir image in pitch without regard for the exact timing. Inother words, since the listeners knew which pitch wouldbe probed, there was less incentive to arrive at the properpitch location incrementall b imagining each successivenote than might have been the case if the pitch location tobe probed had not been known ahead of time.

Similarl, when performing the time judgments in theimager trials, the listeners ma have formed a memortrace of the interval separating the last context note and theprobe, instead of imagining each of the notes leading up tothe probe at the right time. Thus, the increased variance oftemporal estimates we observed ma have been a productof the increased temporal separation from the last refer-ence interval, as would be expected given the extensiveinterval timing literature. Because our experiments werenot structured to verif that the listeners were imaginingever note leading up to the probe note, we cannot rule outthe possibilit that this alternate strateg was used.

Finall, we found significant correlations between theamount of musical training and image thresholds on boththe intonation and the context membership tasks. Moremusical training was accompanied b more narrowltuned images, primaril along the pitch dimension. Al-though we did not find an effect of musical training on theabilit to discriminate a deviation of the probe note fromisochron (cf. Ehrle & Samson, 2005), we did find thatincreased musical training was correlated with a betterabilit to judge whether a probe note occurred on time inthe imager conditions. With regard to pitch images, weobserved a more interesting dissociation across Experi-

ments 24, in that the benefit of musical training in thecontext membership task was apparent for the tonic, butnot for the leading tone. It might be expected that musi-cians would have benefited from their training when mak-ing judgments about the leading tone also. One possiblereason for the discrepanc between the tonic and the lead-

ing tone image widths as a function of musical training isthe observation that ratings of how well a tonic probe fitswith a preceding tonal context are higher among musi-cians than among nonmusicians, whereas ratings for dia-tonic tones outside the major triad are comparable for thetwo groups, if not lower for musicians (Halpern, Kwak,

Bartlett, & Dowling, 1996). Thus, if the listeners in ourstud were basing their membership judgments on a senseof how well the tones fit, musicians would be expected tobenefit when judging trials with the tonic.

What remains to be determined is whether musicalltrained listeners showed the benefits the did because atask in which the are asked to listen carefull and makesubtle acoustic discriminations comes more naturall tothem and the find it easier to focus their auditor atten-tion as directed, or whether the sharper tuning curves re-flect a finer tuning of the neural circuitr underling audi-tor attention and image formation. Recent recordings of

brain electrical activit in musicall trained and untrainedlisteners performing a pitch discrimination judgment atthreshold showed that musical training was unrelated tothe strength of a brain potential index of automatic pro-cessing of acoustic deviance, the mismatch negativit,whereas musical training did increase the magnitude ofevent-related potentials associated with attentional pro-cessing when finer discriminations were made (Tervani-emi, Just, Koelsch, Widmann, & Schroger, 2005).

Overall, our results support a view that auditor imagesare multifaceted and that their acuit along an given di-mension depends, in part, on the context in which the areformed, the manner in which the are probed, and the mu-sical expertise of the listener. The data suggest that imageson the dimension of pitch are isomorphic when generatedb attention-cuing and imager tasks. Electrophsiologicalstudies showing similar responses of the auditor cortex toexpected/imagined auditor events and their heard counter-parts provide more direct evidence for this claim (Hugheset al., 2001; Janata, 2001) and can, perhaps, shed light onhow the temporal acuit of auditor images is shaped.

REFERENCES

Ayotte, J., Peretz, I., & Hyde, K. (2002). Congenital amusia: A groupstud of adults afflicted with a music-specific disorder. Brain,125,238-251.

Barnes, R., & Jones, M. R. (2000). Expectanc, attention, and time.Cognitive Psychology,41, 254-311.

Block, R. A., & Zakay, D. (1997). Prospective and retrospective du-ration judgments: A meta-analtic review.Psychonomic Bulletin &

Review,4, 184-197.Crowder, R. G. (1989). Imager for musical timbre.Journal of Experi-

mental Psychology: Human Perception & Performance, 15, 472-478.Deutsch, D. (1970). Tones and numbers: Specificit of interference in

immediate memor. Science,168, 1604-1605.Dowling, W. J., Lung, K. M.-T., & Herrbold, S. (1987). Aiming at-

tention in pitch and time in the perception of interleaved melodies.Perception & Psychophysics,41, 642-656.

Ehrle, N., & Samson, S. (2005). Auditor discrimination of aniso-chron: Influence of the tempo and musical backgrounds of listeners.

Brain & Cognition,58, 133-147.Farah, M. J. (2000). The neural bases of mental imager. In M. S. Gaz-

zaniga (Ed.), The new cognitive neurosciences (pp. 965-974). Cam-bridge, MA: MIT Press.
http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0006-8950()125L.238[aid=7537782]http://www.ingentaconnect.com/content/external-references?article=0010-0285()41L.254[aid=1534186]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=1069-9384()4L.184[aid=1321928]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0096-1523()15L.472[aid=289813]http://www.ingentaconnect.com/content/external-references?article=0036-8075()168L.1604[aid=317326]http://www.ingentaconnect.com/content/external-references?article=0031-5117()41L.642[aid=2989675]http://www.ingentaconnect.com/content/external-references?article=0278-2626()58L.133[aid=7537781]


17/17


Farah, M. J., & Smith, A. F. (1983). Perceptual interference and fa-cilit

Janata, P., & Paroo, K. (2006). Acuity of Auditory Images in Pitch and Time. Perception &...

Documents

Transcript of Janata, P., & Paroo, K. (2006). Acuity of Auditory Images in Pitch and Time. Perception &...