Infants' Perception of Consonance And

INFANTS’ PERCEPTION OF CONSONANCE AND DISSONANCE IN MUSIC

Marcel R. Zentner’ Jerome Kagan Harvard

The origins of the perception of consonance and dissonance in music are a matter of debate. The present study examined the hypothesis of an innate preferential bias favoring consonance over dissonance by exposing 4 month old infants to consonant and dissonant versions of two melodies. Infants looked significantly longer at the source of sound and were less motorically active to consonant compared with dissonant versions of each melody. Further, fretting and turning away from the music source occurred more frequently during the dissonant than the consonant versions. The results suggest that infants are biologi- cally prepared to treat consonance as perceptually more pleasing than dissonance.

INTRODUCTION a combination of two or more frequencies occuring simultaneously. When the combina- The possibility of perceptual universals in tion is experienced as pleasant, the sound is music has been a question of interest for a long classified as consonant. When the combination time. After many centuries of speculative of frequencies is experienced as unpleasant, debate this question is now being empirically the sound is judged as dissonant. This can be addressed. The study of infants represents one illustrated with the simple example of the way to explore musical universals. This interval. An interval is a combination of two research investigates the hypothesis of an tones of different frequencies. The difference innate bias for consonance over dissonance. in frequencies between the two tones of an Consonance and dissonance refer to the interval, also refered to as interval size, is often subjective judgement of a listener, exposed to expressed as either semitones or multiples of

l Marcel R. Zentner, Department of Psychology, University of Geneva, 9 Rte de Drize, CH-1227 Geneva-Carouge; e-mail: [email protected].

INFANT BEHAVIOR & DEVELOPMENT 21 (3), 1998, pp. 483-492 ISSN 0163-6383 Copyright 0 1998 Ablex Publishing Corporation All rights of reproduction in any form reserved. 484 INFANT BEHAVIOR & DEVELOPMENT Vol. 21, No. 3, 19%) semitones. The semitone is the smallest differ- dyadic intervals, put forward by Helmholtz ence in Western musical scales and corre- (,1954), is that tone pairs are judged consonant sponds to a frequency ratio of the two tones of when the maximum number of upper harmon- approximately 1.06.2 ics match. Harmonics are the frequency values Although consonance and dissonance have of individual pure-tone components. These been described as refering to a subjective harmonics are usually integer multiples of the experience, adults asked to rank intervals in fundamental frequency. For example, a com- accord with their pleasantness, produce similar plex tone with a fundamental frequency of 100 results across varied studies. These studies Hz has of upper harmonics of 100, 200, 300, have been reviewed by Schellenberg and Tre- 400,500 Hz, and so on. To maximize the num- hub (1994). Adults judge as most consonant ber of matching harmonics, the frequency either the octave (difference of 12 semitones), ratios between the fundamentals should be the fifth (7 semitones), or the major third (4 expressible as a ratio of small whole numbers semitones). Adults judge as most dissonant the (2:1, 3:2, 4:3, 5:4, 65 etc.). Figure 1 presents minor second (difference of 1 semitone). the example of three intervals with decreasing These robust facts suggest a lawfulness to the degrees of consonance as a function of judgement of consonant or dissonant intervals. decreasing This regularity has motivated a search for the objective acoustic properties that character- Plomp and Levelt ize consonance and dissonance. A popular (1965) who defined a critical band for two explanation of consonance and dissonance of pure sinusoidal waves as the minimum fre-

- 1 5 6 6 6

5 5 4 5

4 4 4 3

3 3 3

2 2

‘t2 Pl - p1 I ‘f2 fifth major third minor! second highly consonant consonant highly dissonant Matches at: MatchI at: No match 3:2 5:4 6:4 FIGURE 1 The upper harmonics of representative musical intervals and their superposition. Matches of upper harmonics are bold. Consonance and Dissonance 485 quency separation between them that does not cross-cultural studies is scant and ambiguous. give rise to dissonant effects. The critical band Consonance judgements were invariant among is roughly a little lower than a minor third (3 Americans and Japanese (Butler & Daston, semitones). This means that, for pure sounds, 1968), but dissimilar among Canadians and any interval except a full tone (2 semitones) is Indians. Indians showed greater tolerance not dissonant. This generalization does not toward dissonant intervals (Maher, 1976). hold for more complex sounds because they Hence, the cross-cultural study of consonance may have pairs of upper harmonics that fall in preferences has not yet proved to be informa- the critical band. The theory implies that com- tive. Although additional evidence would be of plex tones related by complex frequency ratios interest, it is difficult, if not impossible, to find are more likely to give rise to dissonant effects cultures that are totally isolated from Western not because of the complexity of frequency music. ratios in itself but because of the greater num- A second strategy is to study the reactions ber of overlapping critical bands among adja- of animals to consonance and dissonance. cent harmonics that characterizes intervals Borchgrevink (1975) examined such conso- with complex frequency ratios. nance preferences in 34 albino rats. He con- Although Helmholtz and later physicists nected a tape recorder to a test chamber believed that consonance judgments were the containing a loudspeaker and two pedals. psychological result of physical-acoustical When the rat pressed one pedal a consonant laws operating on inborn properties of the chord was heard, pressing the other pedal was auditory system (Plomp & Levelt, 1965; Ter- followed by a dissonant chord (unfortunately, hardt, 1984), there is, at present, no proof of the chords were not specified in the article). this biological preparedness. One could argue The association between the consonant and that psychoacoustic laws, as posited by Helm- dissonant chords and the position of the pedals holtz and others, could reflect an acquired, was randomly distributed among animals. rather than an inborn, form of auditory pro- Each rat spent 15 min a day in the test chamber cessing. Indeed, some investigators have in a stipulated sequence for three weeks. A argued that consonance judgments are preference for each animal was defined as the acquired competences based on exposure to difference between the number of presses on the music of a particular culture (Frances, the two pedals. The results show that the rats 1988; Lundin, 1985). This view is popular developed a consonance preference. After an among modern music theorists and composers initial period of small and unreliable differ- who adopt a sceptical stance toward an abso- ences, the number of presses on the conso- lute notion of consonance (e.g. Boulez, 1971). nance producing pedal increased, and was at This “nurture” view can be traced to Schdn- the end almost twice the number of presses on berg’s declaration that the concept of conso- the dissonance producing pedal (Borchgre- nance has no useful meaning. Not only did vink, 1975). Although a generalization from Schiinberg proclaim the “emancipation of dis- rats to humans is speculative-and the prior sonance,” but he also proscribed an elimina- musical experience of the rats is not known- tion of consonances (Schonberg, 1984). this study provided support for a possible bio- Different strategies can be chosen to probe logical basis for consonance preference. the origins of our perception of consonance In a more recent study European Starlings and dissonance. One strategy is to compare (Sturnus Vulgaris) were trained to peck at one judgments of subjects from cultures with dif- key when a consonant chord was presented ferent musical systems. If consonance judg- and at another key when a dissonant chord was ments are primarily a function of exposure we presented. The birds generalized the response should find variability in judgments across cul- to a new pair of consonant and dissonant tures. Unfortunately, evidence from such chords (Hulse, Bernard & Braaten, 1995). This 486 INFANT BEHAVIOR & DEVELOPMENT Vol. 21, No. 3,1998

result suggests that consonance and disso- The present experiment was designed spe- nance may be an important cue for auditory cifically to test the hypothesis of an innate communication among songbirds. preferential bias favoring consonance over dis- The problems with conducting cross-cul- sonance. Unlike studies that used isolated pairs tural studies and the interpretation of animal of intervals, infants were presented with actual work motivate a search for other strategies. music. Two different, unfamiliar melodies The study of young infants provides an oppor- were composed with a synthesizer, and a con- tunity for the exploration of auditory predispo- sonant and dissonant version was created for sitions in humans. Although there is a body of each melody. Because our interest was in the literature on infants’ music perception, studies infants’ preference, and not in perceptual capa- typically focused on infants’ processing capa- bility, the dependent measures of visual fixation of the music source, motor activity, bilities rather than preferences and/or affective vocalization, fretting and turning away from responses (Trehub & Trainor, 1993). More- the music source were all coded as relevant over, these studies of infants’ interval percep- indicators. It was expected that the infants tion have examined responses to melodic would be more attentive and show less fre- intervals where tones are presented succes- quent signs of distress to the consonant com- sively, rather than harmonic intervals where pared to the dissonant melodies. tones are presented simultaneously. One of the exceptions is a study by Crow- der, Reznick and Rosenkrantz (1991) origi- METHOD nally designed to test 6 month-old infants’ reactions to major and minor chords. Although there were no differences to major and minor Participants modes, the introduction of a dissonant chord Thirty-two full term, healthy 4 months old seemed to provide reliable differences with infants (16 males, 16 females) participated in prolonged visual fixation on the loudspeaker to the experiment (age range 16.7-21.1 weeks, M the consonant chords. Because the study was 17.9 weeks). Nine additional infants were not designed to test infants’ reactions to conso- tested but excluded because they became fussy nance and dissonance, the number of infants in (7) or because of experimenter error (2). The the consonance/dissonance condition was very parents were middle-class Caucasian residents small (n = 9). of the Boston, Massachusetts metropolitan In a recent experiment by Schellenberg and area and were recruited through the birth Trehub (1996) nine month old infants were records of local city halls. able to detect changes from harmonic intervals with simple frequency ratios to intervals with Stimuli complex frequency ratios, but failed to detect changes from complex to simple frequency Each infant was presented with two different ratios. The particular intervals used were the melodies, each 3.5 set duration, in a consonant fourth and the fifth (= simple frequency ratios) and a dissonant version for a total of 4 trials. The as opposed to the tritone (= complex frequency consonant and dissonant versions of the two ratio). The authors concluded that infants pos- melodies were identical with regard to param- sessed a greater facility to discriminate inter- eters such as pitch, tempo, rhythm, timbre, con- vals with simple compared to intervals with tour. The dissonant version was composed in complex frequency ratios. This result is inter- parallel minor seconds (interval size of 1 semi- esting as it indicates that the distinctiveness tone), the consonant version in parallel major and memorability of certain intervals may and minor thirds (interval size of 4 and 3 semi- underlie a preference. tones respectively), using only two synthetic Consonance and Dissonance 487 voices. The alteration of major and minor thirds not think this creates a serious problem for the followed the structure of the diatonic scale. interpretation of our results. The definition of dissonance and conso- We used two melodies to ensure that the nance was based on prior evidence indicating responsewasnotareactiontotheparticularprop- that the minor second is consistently rated as erties of one melody. Both melodies were Cen- the most dissonant interval, while the third is tralEuropeanfolksongs.Theparentswereasked experienced as consonant (Schellenberg & if they were familiar with the songs. No parent Trehub, 1994). However, to insure that the replied that they had heard the songs before. The melodies too would be perceived as consonant stimuli were created using a computer program and dissonant, 24 undergraduates were asked (midi sequencer, Passport Designs Pro4) that to judge which of the two versions they pre- controlled an attached music synthesizer (midi ferred. Twenty-three showed a preference for synthesizer/Roland D-5). Figure 2 presents the the consonant version and, when asked to com- first of the two melodies in musical notation and ment on this preference, frequently answered in Hertz for each pair of tones. that the dissonant versions were “hurting their ears.” The choice of seconds and thirds was also based on the fact that the interval size between the two versions differed only mini- mally. As a result no major pitch difference would be introduced as a confound.3 Although a small difference could not be completely eliminated, we will explain later why we do

Melody A

Upper Voice (Hz): 493.88 440.00 415.30 369.W 329.63 245.94 245.94

consoMnr: 415.30 369.99 329.63 293.66 196.M IS6.00 IY6.M~

Dissmlanl: 466.16 415.30 392.00 349.23 17461 233-M 233.08

UpperVoice: 415.30 415.30 493.88 493.88 415.30 41S.M 4Y3.88

LavderVoice: c

Dissonant: 392.00 392.00 466.16 466.16 392.00392.00 466.16

“ppm voice: 554.31 493.88 440.00 41530 369.99 369.99 329.63

Lower Voice: consmwu: 440.00 41530 369.99 329.63 293.66 293.66 207.65

Di.wnmc 523.25 466.16 JI5.30 392.00 349.23 349.23 174.61 FIGURE2 The first of the two melodies (Melody A) in musical notation and in Hertz. 488 INFANT BEHAVIOR & DEVELOPMENT Vol. 21, No. 3,1998 and was covered with an attractive pattern of variable was total fixation time during the 35 black and white concentric circles. A video set of each trial, the duration of the initial fix- camera recorded the infant’s behavior. There ation following the onset of each trial, and the was a wall in front and to the left of the infant. duration of fixation on speaker during the A research assistant and the parent, who interstimulus intervals. Motor activity was remained in the room, sat behind the infant, defined as the total duration of (a) flexion of outside of the infant’s visual field. Initially, one or both arms and legs of at least 60 there was a 35 set quiet baseline followed by degrees, (b) movement of one or both arms and the 4 melodies with a quiet interval between legs by at least 60 degrees to the right or left each melody of 8 sec. The order of the presen- and more than 2.5 cm in vertical direction. tation of the melodies was counterbalanced Turning away was defined as the frequency of (Latin Square design). The music was deliv- turning the head away from the pattern of con- ered at a SPL of 60dB(A). centric circles by 90 degrees or more. Vocal- izations, non-distress vocal sounds, was Measures defined as the frequency of vocalizations per trial. There had to be a quiet period of at least The videotape record was coded for (1) fix- 3 set between two vocalizations in order to ation time directed at speaker (2) motor activ- code them as separate. Fretting, defined as dis- ity (3) turning away from the speaker (4) tressed vocal sounds, was coded in a similar vocalization and (5) fretting. The fixation time manner to vocalizations.

14 Visual Fixation Motor Activity Cl Consonant m Dissonant I 13

12 M e 11 a n 10 T i m 8 S T e 7 c 6

4 5 - Melody A Mel Melody A AMelody B FIGUREa 3 Left side: Mean total visual fixation time to music source during the consonant and dissonant versions of two melodies. Right side: Mean time of motor movement during the consonant and dissonant versions of the two melodies. Error bars represent standard errors. Consonance and Dissonance

Reliability of the coders was assessed by Vocalizations, Fretting and Avoidance having two persons independently code these variables for 20% of the subjects. Pearson Eight of 32 infants fretted or showed avoid- product moment correlation coefficients were ance during the dissonant versions, but neither .94 for fixation time, .91 for motor activity, .96 fretted nor avoided during the consonant ver- for turning away and .89 for vocalizations and sions. No infant fretted or avoided only during the consonant versions but not during the dis- for fretting. sonant versions. The probability of this result In addition, each parent filled out a ques- occurring by chance is p < 0.01 by the bino- tionnaire about listening habits, musical back- mial theorem. Similarly, 7 infants vocalized ground and education, and their memories of during the consonant, but not the dissonant the type of music the infants’ might have heard versions, while 1 infant vocalized during the at home. dissonant but did not do so during the consonant versions [binomial test: p < 0.051. RESULTS Musical Background Variables Visual Fixation and Motor Activity In order to examine possible effects of the A repeated measures analysis of vari- infants’ and parents’ musical experience we ance, with sex and order as the between-sub- created a new variable by subtracting total fix- ject variables and melody (A or B) and ation time to the dissonant versions from total mode (consonant or dissonant) as the fixation time to the consonant versions: The repeated measure, was implemented for larger this difference score, the longer the each dependent variable. A square root trans- infant fixated the concentric circles during the consonant melodies. Similarly, we subtracted formation was performed because the scores the total motor score to the consonant melodies were not normally distributed. The repeated from the total motor scores to the dissonant measures ANOVA yielded a significant ones: The larger the difference, the more active main effect of musical mode on total tixa- the infant was during the dissonant conditions. tion time [F(1,24) = 11.50 p < 0.005], but Pearson product-moment correlations were also on duration of initial fixation [F( 1,24) computed between infants’ exposure to music = 4.72 p < 0.051 and fixation time during at home (average hours/week) and these differ- the 8 set intervals [F(1,24) = 5.30 ence scores for fixation time and motor activ- p < 0.041. The infants looked significantly ity: All correlations were nonsignificant longer at the speaker for the consonant, com- (fixation: r = -.02; motor activity: r = -.lO). pared with the dissonant, version of each Similarly, there were no significant correla- melody (see Figure 3). tions between the infants’ fixation time or Infants also moved sigificantly more during motor behavior and either the parents’ musical the dissonant compared with the consonant background or practice of a musical instrument versions of the melodies [F( 1,24) = 6.64 (fixation: r = -.22; r = -.08 - motor activity: p c 0.021 (see Figure 3). There was no signifi- r = .19; r = -.06). cant effect for melody (A or B), sex, or order of presentation. There was an unexpected, but significant melody x sex interaction DISCUSSION [F(1,24) = 15.40 p c 0.0011. Girls were more active motorically during melody B, boys were Infants looked significantly longer at the more active during melody A. Melody B had a speaker and were less motorically active when faster tempo than melody A. hearing the consonant as compared to the dis- 490 INFANT BEHAVIOR & DEVELOPMENT Vol. 21, No. 3, 1998

sonant versions of each melody. Although seconds, the two types of intervals also differ only a small number of infants vocalized, fret- in terms of interval size. Thus, one could argue ted or turned away, the occurrence of these that the differential behaviors to the two melo- behaviors was differentially distributed over dies has nothing to do with consonanceldisso- the consonant and dissonant stimuli. These nance, but are due to differences in size of data support the hypothesis of an innate bias interval. favoring consonance over dissonance. However, size of complex-tone intervals Two major factors control infants’ duration (as used in our study) is not associated with of fixation: preference and discrepancy (Fantz, preference in adults. Sometimes large intervals Fagan, & Miranda, 1975). The dissonant ver- are judged as unpleasant (seventh = 10-l 1 sions of the melodies deviated from what most semitones); sometimes they are rated as conso- infants might have heard at home. Therefore nant (sixth = 8-9). The same is true for small they can be treated as discrepant events. If dis- (third = 4-5 vs. second l-2) and medium-sized crepancy was controlling the infants’ fixation intervals (fourth = 5 and fifth = 7 vs. tritone = time we would expect the infants to show 6). Clearly, then, in adults preference for inter- increased fixation time during the dissonant vals is not a function of size, but of frequency versions of the melodies. However, because ratio. Should this be different for infants? In the infants did not show longer fixations to the the study by Schellenberg and Trehub cited dissonant stimuli it is fair to assume that it was earlier, the intervals were the fourth and the a preference, and not discrepancy, that con- fifth (= simple frequency ratios), as opposed to trolled the infants’ fixation time. the tritone (= complex frequency ratio). When Differences in motor activity are more diffi- infants were presented with standard interval cult to interpret. Increased motor activity patterns they were better at discriminating a reflects arousal, but arousal can be pleasant or change from a pattern composed of the fifth or distressed. However, because the dissonant the fourth to the tritone than vice versa. versions elicited a combination of increased Because the tritone has an interval between the motor activity but shorter fixation times and fourth (smaller) and the fifth (larger), fre- increased fretting/turning away, it is reason- quency ratio, and not interval size, has to be able to assume that the increased motor activ- responsible for the results. Further, infants as ity reflected a distressed rather than a pleasant well as adults, were better able to discriminate state. One can also interpret the data by regard- changes from a standard consonant harmonic ing a decrease in motor activity as a measure of interest. In this case the reduced motor activity interval to a dissonant harmonic interval than during the consonant versions leads to the to another consonant interval, even though the same conclusion as the data on fixation time. change in interval size to the consonant inter- Fretting is an obvious manifestation of val was twice as large as the change to the dis- unpleasantness. In sum, these results reveal a sonant interval. In this study, dissonance was pattern suggesting that the human infant may more important than interval size in mediating be born with a biological preparedness that interval discrimination (Schellenberg & makes consonance perceptually more attrac- Trainor, 1996). In sum, although we cannot tive than dissonance. conclusively rule out the alternative hypothe- Some limitations of our study are acknowl- sis, it is less likely than the one proposed. edged. The primary limitation is a function of Another limitation is that seconds and the way consonance and dissonance were thirds represent only one aspect of consonance operationalized. Although the melodies we and dissonance. Therefore, we must be cau- created were based on massive evidence from tious in generalizing our findings to other adult listeners who judge thirds (major and forms of consonance and dissonance. More minor) to be much more consonant than minor ambiguous forms of consonance and disso- Consonance and Dissonance 491 nance might be subject to a much stronger cul- same interval of time, shall be commensu- tural impact. rable in number, so as not to keep the ear *drum in perpetual torment, bending in two Our primary objective in the present inves- different directions in order to yield the tigation has been to examine the hypothesis of ever discordant impulses. an innate bias favoring consonance over dissonance. We can offer no explanations, only Acknowledgment: A portion of this work speculation, on the nature of the innate bias previously appeared in Nature (383, p. 29, that renders consonance perceptually more 1996). Preparation of this paper was supported attractive than dissonance. A number of inves- by a grant from the John D. and Catherine T. tigators contend that, intervals with matching MacArthur Foundation Network on Psychopa- upper harmonics (= consonances) and inter- thology and Development and a fellowship vals with non-matching upper harmonics (= from the Swiss National Science Foundation dissonances) create distinct firing patterns in awarded to the first author. We are grateful to the auditory neural network. This idea, put for- Curt Stallman for his help with the creation of ward by Boomsliter and Creel (1961) and the stimuli, to Emily Hoffman, Steve Most and Moore (1989) (see also Schellenberg & Tre- Meredith Rowe for helping us with testing and hub, 1994), is that complex tones that are coding, and to Carol Krumhansl and Sandra related by simple frequency ratios produce Trehub for their helpful comments at different similar firing patterns in the auditory network stages of this research. because of the relatively frequent match of upper partials (see Figure 1). The absence of shared neural circuits presumed when there is NOTES processing of dissonances might be one reason for experiencing intervals with complex fre- Marcel R. Zentner is now at the Department of quency ratios as dissonant. Psychology at the University of Geneva. The pairs of tones can occur in succession or Another explanation centers on the concept simultaneously. The former type of interval is of critical bandwith. In this view, the auditory called sequential or melodic, the latter is neural network has a limited ability to resolve referred to as harmonic. Although consonance different tones (and its upper haromics) that and dissonance can refer to a sequence of single are too proximate in pitch. Tones that are very tones, they usually refer to harmonic intervals. proximate, but not identical, in frequency can It is only in this latter sense that we are con- not be effectively processed by the basilar cerned with consonance and dissonance here. membrane: hence the excitation patterns over- The exclusive use of thirds would result in the lap. These fluctuations result in a perception of violation of tonality, introducing another type roughness or dissonance. Because the minor of dissonance. Thus, in the few instances where it was unavoidable the third was replaced by its seconds used in the dissonant versions of our natural complement, the sixth, which has a melodies are characterized by more complex very similar consonance quality. Where the frequency ratios and more overlapping critical third was replaced by the sixth in the consonant bands than the thirds used for the consonant version, in the dissonant version the natural versions of the melodies, the two hypotheses complement of the second, the seventh, was are equally plausible. used in order to match the interval size of the These explanations are reminiscent of Gali- consonant version (see Figure 2). leo’s views for he wrote in 1638 (1963, p. 100): REFERENCES

Agreeable consonances are pairs of tones which strike the ear with a certain regular- Boomsliter, P. & Creel, W. (1961). The long pattern ity; this regularity consists in the fact that hypothesis in harmony and hearing. Journal of the pulses delivered by the two tones, in the Music Theory, 5,2-3 1. 492 INFANT BEHAVIOR & DEVELOPMENT Vol. 21, No. 3, 1998

Borchgrevink, H. M. (1975). Musikalske akkord- Maher, T. F. (1976). “Need for resolution” ratings preferanser hos mennesket belyst ved dyre- for harmonic musical intervals: A comparison forsok. [Musical chord preferences in humans as between Indians and Canadians. Journal of demonstrated through animal experiments]. Tid- Cross Cultural Psychology, 7. 259-276. skrift for den Norske Laegeforening, 95, 3.56 Moore, B. C. J. (1989). An introduction into thepsy- 358. chology ofhearing (3rd ed.). London: Academic Boulez, P. (197 1). Boulez on music today. London: Press. Faber and Faber. Plomp, R. & Levelt, W. (1965). Tonal consonance and critical bandwith. Journal of the Acoustical Butler, J. W. & Daston, P. G. (1968). Musical con- Society ofAmerica, 38, 548-560. sonance as musical preference: A cross-cultural Schellenberg, E. G. & Trainor, L. T. ( 1996, June). study. Journal of General Psychology, 79, 129- Sensory cortsonance and the perceptual similur- 142. ity of complex-tone harmonic intervals: Tests of Crowder, R. G., Reznick, J.S. & Rosenkrantz, S. adult and infant listeners. Paper presented at the (1991). Perception of the major/minor distinc- 10th Biennial International Conference on Infant tion. Preferences among infants. Bulletin qf the Studies, Providence, Rhode Island. Psychonomic So&q, 29, 187- 188. Schellenberg, E. G. & Trehub, S. E. (1994). Fre- Fantz, R. L., Fagan. J. F., Miranda, S. B. (1975). quency ratios and the perception of tone pat- Early visual selectivity. In L. B. Cohen & P. Sal- terns. Psychonomic Bulletin & Review, I, 191- apatek (Eds.), Infant perception: From sensation 201. to cognition: Vol 1. Basic visual processes (pp. Schellenberg, E. G. & Trehub, S. E. (1996). Natural 249-341). New York: Academic Press. musical intervals. Evidence from infant listen- Frances, R. (1988). The Perception ofMusic (W. J. ers. Psychological Science, 7, 272-277. Dowling, Trans.). Erlbaum: Hillsdale. Schonberg, A. (1984). Style and idea. Berkeley: Galileo Galilei (1963 /1638). Dialogues concerning University of California Press. two new sciences (H. Crew & A. de Salvio Terhardt, E. (1984). The concept of musical conso- Trans.). New York: McGraw-Hill. nance: A link between music and psychoacous- Helmholtz, H. v. (1954 /1863). On the sensations of tics. Music Perception, 1, 276-295. Trehub, S. E. & Trainor, L. J. (1993). Listening tone as a physiological basis for the theor?: of strategies in infancy: The roots of language and music (A. J. Ellism, Trans.) New York: Dover. musical development. In S. McAdams & E. Big- Hulse, H. S., Bernard, D. J. & Braaten, R. F. (1995). and (Eds.), Thinking in sound: The cognitive Auditory discrimination of chord-based spectral psychology qf human audition (pp. 278-327). structures by European Starlings (Stumus vul- London: Oxford University Press. garis). Journal of Experimental Psychology: General, 124,409~I23. Lundin, R. W. (1985). An objective psychology of music. (3rd ed.). Malabar, FL: Krieger. 12 December 1996; Revised 28 October 1997 n