COMMENTARY Absolute and Relative Pitch Processing in Tone Learning

DESC_010.fm Page 44 Tuesday, January 28, 2003 11:43 AM

44 Commentary

COMMENTARYBlackwell Publishing Ltd AbsoluteCommentary and relative pitch processing in tone learning tasks

Sandra E. Trehub

Department of Psychology, University of Toronto, Canada

Saffran offers her results as support for ‘the hypothesis term memory (e.g. the first two tones of Here Comes the that absolute pitch is present in infancy’ (Abstract). Bride: Smith, Kemler Nelson, Grohskopf & Appleton, Before evaluating her claim, it is reasonable to review 1994). some facts about absolute pitch (AP) and about AP Tasks that involve musical labels for tones or intervals possessors. Since the phenomenon was first submitted are necessarily limited to musically sophisticated lis- to scientific scrutiny (Stumpf, 1883), AP has been defined teners who have well-established associations between as the ability to identify a specific tone by its musical verbal labels (note or interval names) and pitches (for AP) name (e.g. C) or to produce it (by singing or adjusting or pitch combinations (for RP). Tasks without such a tone generator) without using a reference pitch. The labelling requirements have provided information about assumption is that individuals with AP have a fixed, untrained adults’ long-term memory for specific pitches or stable, internal template of note names (e.g. C, D, E, (Levitin, 1994; Schellenberg & Trehub, in press) or inter- F, G, A, B) attached to specific frequencies or pitches vals (Drayna et al., 2001). In any case, concepts related (Ward, 1999). When AP possessors wake up in the to AP and RP are largely irrelevant to Saffran’s current morning, they can name any isolated tone whose pitch study of statistical learning or to her previous explora- corresponds to a note on the musical scale. Indeed, tions of tone learning (Saffran & Griepentrog, 2001; such unconventional test procedures have been used to Saffran, Johnson, Aslin & Newport, 1999), which did preclude adults’ reliance on reference tones (Petran, not explore long-term memory. With no basis for attrib- 1932). More commonly, the test context is configured to uting AP or RP to the infants or adults in these studies, minimize the availability of reference tones (Costall, how can we interpret the findings? We can ask, instead, 1985). whether infants and adults used absolute or relative cues AP is considered special because of its relative rarity to solve the task. Can we accept Saffran’s contention (1 in 10,000: Takeuchi & Hulse, 1993). In general, that infants capitalized on absolute pitch cues and that trained musicians or musically educated amateurs have adults capitalized on relative pitch cues? relative pitch (RP) rather than AP. Instead of the stable Musicians in Saffran and Griepentrog (2001) should template associated with AP, RP involves a movable have obtained reasonable performance levels on the template corresponding to precise relations among notes basis of their RP abilities. Instead, they achieved 63% of the scale (Ward, 1999). RP possessors recognize the correct on the RP task and 56% on the AP task (non- musical interval (pitch distance) between any pair of musicians obtaining 57% and 52%), which raises con- tones (e.g. perfect fifth, major third), which enables them cerns about Saffran’s task designations. Non-musicians to name the second of two tones when the first is named. in Saffran et al. (1999) achieved roughly 65% correct on When they hear a single (unnamed) tone, they can pro- a comparable AP task with different tone sequences, duce a second tone to yield a specified interval. The less which indicates that the component tones of training developed RP skills of musically untrained individuals and test sequences can affect performance. It is likely can be evaluated in other ways: judging a well-known that small pitch distances between the component tones melody as in or out of tune (Drayna, Manichaikul, de of ‘words’ in Saffran et al. (1999) contributed to the Lange, Snieder & Spector, 2001), or judging whether the coherence and memorability of those sequences, in line interval between two tones matches an interval in long- with perceptual grouping principles (Bregman, 1990;

Address for correspondence: Department of Psychology, University of Toronto at Mississauga, Mississauga, Ontario, Canada, L5L 1C6; e-mail: [email protected]

DESC_010.fm Page 45 Tuesday, January 28, 2003 11:43 AM

Commentary 45

Deutsch, 1999). Regardless, Saffran’s conclusion that References adults used a relative pitch strategy is at odds with their failure to perform better on RP than on AP tasks in the Baharloo, S., Johnston, P.A., Service, S.K., Gitschier, J., & present study. Her use of multiple test trials may have Freimer, N.B. (1998). Absolute pitch: an approach for iden- impaired adults’ performance, perhaps differentially, on tification of genetic and nongenetic components. American AP and RP tasks. Indeed, the presentation of interven- Journal of Human Genetics, 62, 224–231. ing tones is often used to ‘erase’ pitch memory in AP Baharloo, S., Service, S.K., Risch, N., Gitschier, J., & Freimer, N.B. (2000). Familial aggregration of absolute pitch. Amer- tasks (Ward, 1999). ican Journal of Human Genetics, 67, 755–758. If infants detected differences between words and Bregman, A.S. (1990). Auditory scene analysis: The perceptual part-words on the basis of absolute pitch cues, they organization of sound. Cambridge, MA: MIT Press. should have succeeded on RP as well as AP tasks. The Costall, A. (1985). The relativity of absolute pitch. In P. How- test items in both tasks included sequences that were ell, I. Cross & R. West (Eds.), Musical structure and cognition presented frequently during the familiarization period. (pp. 189–208). London: Academic Press. The principal difference between AP and RP tasks was Deutsch, D. (1999). Grouping mechanisms in music. In D. the non-distinctiveness of relative pitch cues among test Deutsch (Ed.), Psychology of music, 2nd edn (pp. 299–348). items in the AP task. Specifically, the AP task had dis- San Diego: Academic Press. tinctive absolute pitch cues, and the RP task had dis- Dowling, W.J. (1999). The development of music perception tinctive absolute and relative pitch cues, along with and cognition. In D. Deutsch (Ed.), Psychology of music, 2nd edn (pp. 603–625). San Diego: Academic Press. distinctive contour cues. These factors make it imposs- Drayna, D., Manichaikul, A., de Lange, M., Snieder, H., & ible to pinpoint the basis for infants’ or adults’ perform- Spector, T. (2001). Genetic correlates of musical pitch recog- ance on the AP and RP tasks. nition in humans. Science, 291, 1969–1972. Saffran’s proposed developmental shift from absolute Flege, J.E., & Fletcher, K.L. (1992). Talker and listener effects to relative pitch processing has parallels in infants’ shift on degree of perceived foreign accent. Journal of the Acous- from language-general to language-specific processing of tical Society of America, 91, 370–389. speech sounds (Werker & Tees, 1999). Her proposal is Levitin, D.J. (1994). Absolute memory for musical pitch: evid- also consistent with the hypothesized critical period for ence from the production of learned melodies. Perception the acquisition of AP (Takeuchi & Hulse, 1993). How- and Psychophysics, 56, 414–423. ever, unlike very early exposure to a second language, Petran, L.A. (1932). An experimental study of pitch recogni- which guarantees accent-free speech (Flege & Fletcher, tion. Psychological Monographs, 42, 1–120. Saffran, J.R., & Griepentrog, G.J. (2001). Absolute pitch in 1992), early musical training (before 6 years) is necessary infant auditory learning: evidence of developmental reorgan- but not sufficient for AP (Baharloo et al., 1998). Genetic ization. Developmental Psychology, 37, 74–85. factors also make important contributions (Baharloo, Saffran, J.R., Johnson, E.K., Aslin, R.N., & Newport, E.L. Johnston, Service, Gitschier & Freimer, 1998; Baharloo, (1999). Statistical learning of tone sequences by human Service, Risch, Gitschier & Freimer, 2000), as they do infants and adults. Cognition, 70, 27–52. for RP (Drayna et al., 2001). Schellenberg, E.G., & Trehub, S.E. (in press). Good pitch In short, Saffran’s proposed developmental shift memory is widespread. Psychological Science. requires much more evidence than that presented. For Smith, J.D., Kemler Nelson, D.G., Grohskopf, L.A., & Apple- example, it remains to be determined whether there are ton, T. (1994). What child is this? What interval was that? developmental changes in pitch memory and in the rel- Familiar tunes and music perception in novice listeners. ative weighting of absolute and relative pitch cues. Cognition, 52, 23–54. Stumpf, C. (1883). Tonpsychologie. Leipzig: Herzel. Despite the prevailing view that adults typically focus on Takeuchi, A.H., & Hulse, S.H. (1993). Absolute pitch. Psycho- relative pitch cues, there is increasing evidence that they logical Bulletin, 113, 345–361. encode novel melodies in absolute rather than relative Ward, W.D. (1999). Absolute pitch. In D. Deutsch (Ed.), terms (Dowling, 1999). Finally, it is possible that the Psychology of music, 2nd edn (pp. 265–298). San Diego: critical period for AP has less to do with age-related Academic Press. differences in pitch memory than with developmental Werker, J.F., & Tees, R.C. (1999). Influences on infant speech changes in the ease of acquiring arbitrary associations processing: toward a new synthesis. Annual Review of Psycho- between verbal labels and pitches. logy, 50, 509–535. February61000 2003

DESC_010.fm Page 46 Tuesday, January 28, 2003 11:43 AM

46 Response

RESPONSE BirdsResponse do it – why not babies?

Jenny R. Saffran

Waisman Center and Department of Psychology, University of Wisconsin-Madison, USA

Like many species of songbirds and non-human pri- 1994; Halpern, 1989). Moreover, infants can apparently mates, we suggest in the current paper and elsewhere be induced to track RP cues in our task if AP cues are (Saffran & Griepentrog, 2001) that human infants have rendered non-predictive; preliminary data from our lab the capacity to represent the absolute pitches of sounds, indicate that when RP cues are consistent but AP cues which we defined as ‘the encoding of a pitch independ- are highly variable, 8-month-old infants can track relat- ent of its relation to other sounds’. Whether or not this ive pitch sequences in our stimuli (Saffran, Reeck & mode of perceiving the auditory world is related to the Niebuhr, in progress). This pattern of results suggests a rare ability to label pitches remains unknown, and study- complicated interplay between the perceptual capacities ing the ontogeny of pitch labeling is not the intent of of learners and the cues afforded by auditory stimuli: this line of research. Instead, our interest lies in charac- adult listeners are more successful at tracking absolute terizing the cues utilized by learners at different points pitches as the materials become easier to represent, while in development. That is, to what extent do infant lis- infant listeners are more successful at tracking relative teners share the perceptual primitives – absolute pitches pitches when absolute pitch cues are rendered less – observed frequently in non-human species, and how informative. does this capacity change over age and experience? We agree with Sandra Trehub that the component We suggest that our results support the claim that tones of the training and test sequences can affect per- infants preferentially represent absolute pitches whereas formance; this is clearly evident in the difference in per- adults preferentially represent relative pitches, given the formance between the current study and Saffran and task of segmenting a continuous stream of tones. Griepentrog (2001). However, it is not the case that non- Directly pitting these cues against one another is diffi- musicians in another of our studies (Saffran, Johnson, cult, given that any alteration in absolute pitches neces- Aslin & Newport, 1999) achieved roughly 65% on a sarily alters the relative pitches. It is for this reason that comparable AP task, as suggested by Trehub. That task we chose the tact of paired experiments. Each experi- was not designed to isolate AP cues; RP cues were also ment was designed such that success on the test would available for some of the test discriminations used by require the use of either one type of pitch cue or the Saffran et al. (1999). Comparing performance levels other. As in any experiment, we can only interpret suc- from Saffran et al. (1999) with the results of our more cessful discrimination; in the case of the Saffran and recent studies, in which no RP cues were available for Griepentrog (2001) results, infants successfully discrim- test discrimination, does not provide a valid indicator of inated when AP cues were required, and adults success- the use of different types of pitch cues. It is also import- fully discriminated when RP cues were required. Because ant to note that infants’ failure in the RP tasks is this is a different pattern of successes as a function of unlikely to be due to inherent difficulties with the coher- development, we interpreted these data as indicating a ence and memorability of the ‘words’, given that adults shift in perceptual prioritization. can successfully learn these materials (while failing to Importantly, as noted in our papers, we do not claim learn others on which the infants succeed). that these cues are used to the exclusion of other types Is it the case that infants should have succeeded on the of information. Indeed, adults in the current study do RP task given access to AP cues? We suggest not. While show some success at tracking sequences of absolute words and part-words did necessarily differ in their AP pitches, consistent with other literature demonstrating cues, the design ensured that the transitional probabilit- the use of AP cues by adult non-musicians (Levitin, ies between absolute pitches in the part-words and the

Address for correspondence: Department of Psychology, 1202 West Johnson Street, University of Wisconsin-Madison, Madison, WI 53706, USA; e-mail: [email protected]

DESC_010.fm Page 47 Tuesday, January 28, 2003 11:43 AM

Response 47

words are identical, limiting the usefulness of AP as a learning materials (MacDougall-Shackleton & Hulse, cue for discrimination. Infants’ failure to discriminate 1996); we suspect that humans share this capacity to use suggests that the equating of absolute pitch familiarity both types of pitch cues, as well as the ﬂexibility to across test items provided a sufﬁcient control. Relative switch perceptual primitives as the situation demands. pitch cues – which provided a strong cue for discriminat- ing words from part-words – also failed to lead to discrimination, which we interpret as a lack of evidence for References the use of RP cues given these materials as input. We thus conclude that infants do have access to abso- Halpern, A.R. (1989). Memory for the absolute pitch of familiar lute pitch information. At the same time, we absolutely songs. Memory and Cognition, 17, 572–581. agree with Trehub that much more remains to be learned Levitin, D. (1994). Absolute memory for musical pitch: evidence about the use of pitch cues across development. We sug- from the production of learned melodies. Perception and gest that all of our learners can use both types of pitch Psychophysics, 56, 414–423. cues, but that the circumstances under which each type MacDougall-Shackleton, S.A., & Hulse, S.H. (1996). Concurrent of cue is used are driven by different factors at different absolute and relative pitch processing by European starlings. Journal of Comparative Psychology, 110, 139–146. ages. On this view, the interesting question is what drives Saffran, J.R., & Griepentrog, G.J. (2001). Absolute pitch in infant the relative strengths of the many cues which co-exist in auditory learning: evidence for developmental reorganization. the perceptual world, and why might this weighting Developmental Psychology, 37, 74–85. change over development? Some avian species can be Saffran, J.R., Johnson, E.K., Aslin, R.N., & Newport, E.L. induced to switch from prioritizing absolute pitch to (1999). Statistical learning of tone sequences by human using relative pitch as a function of the structure of the infants and adults. Cognition, 70, 27–52.