📖 Absolute Pitch

Perception of Pitch Relations

• AP possessors do perceive pitch relations (intervals, chords), just as non-possessors do
• But absolute perception seems to occur automatically and involuntarily
• Evidence: Stroop-like interference when incongruent pitch names are presented (Zakay et al. 1984)
• Corliss (1973) reported identifying intervals indirectly — first the component pitches, then the relation — but that is one author’s introspection about herself, and T&H report it in order to test it
• The test went against it: Benguerel & Westdal (1991) used mistuned notes and found that 9 of the 10 AP possessors showed no evidence of identifying intervals indirectly. T&H offer this as one interpretation: that AP possessors “identify the pitch intervals directly, just as AP nonpossessors do” (p. 354)
• AP possessors are reported to find transposition harder — but T&H suspect the difficulty is in producing, not perceiving, and say it “appears unlikely that AP possessors represent pitch structures purely in absolute terms” — though T&H note that “no formal study of the comparative abilities of AP possessors and nonpossessors to produce pitch transpositions has been undertaken”

What varies instead: degree of AP

Where T&H do sketch a structure, it is a hierarchy of degree, not a quasi/genuine split. Sensitivity to timbre, register and pitch class marks the lower levels of AP; the ability to produce pitches absolutely marks the higher ones. Their evidence for the ordering is an absence: “there are no documented cases of AP possessors who can produce pitches absolutely but are unable to identify pitches” (p. 353).

Absolute Tonality — and why it is not simply “weak AP”

Some people can name the key of a piece without being able to name individual pitches, an ability the paper calls absolute tonality. It looks like a natural intermediate step, and Sergeant (1969) found that people who have it began musical training later than AP possessors did — independent support for the early-learning theory.

But T&H close the simple reading: “if absolute tonality is a weak form of AP, then AP possessors should also have absolute tonality. This is not quite the case.” (p. 352) AP possessors tend to arrive at the key indirectly, by naming notes first. They therefore contest Vernon’s (1977) framing of absolute tonality as a weak form of AP.

Residual AP in Adults

Several studies suggest that even non-possessors retain some absolute pitch ability:

• Halpern 1989: Both musicians and non-musicians consistently start familiar songs at the same absolute pitch
• Levitin 1992: college students sang songs from familiar pop recordings. Ignoring octave errors, 24% of subjects reproduced the absolute pitches of the songs and 51% came within one semitone (published as Levitin 1994, also in this collection)
• Terhardt & Seewann 1983; Terhardt & Ward 1982: Non-possessors detect transpositions above chance
• Deutsch et al. 1988, 1990: Perception of paradoxical pitch patterns consistent within individuals, suggesting absolute internal standards

Takeuchi & Hulse read these as consistent with their theory — that everyone initially has the potential for AP and most people lose it as relational perception takes over. They are suggestive findings about residual absolute encoding, not a demonstration of that developmental claim.

The Early-Learning Theory (Central Thesis)

Takeuchi & Hulse propose a unified theory for AP development. They preface it with a hedge that the paper’s reputation has since worn away: “Although the data are insufficient to unequivocally support any theory of the etiology of AP, we believe the existing evidence points toward the early-learning theory” (p. 355).

1. AP can be acquired by anyone during a limited period of development
2. This period corresponds to ages ~3 to 5–6 years
3. After this, a general developmental shift from perceiving absolute features to relational features occurs
4. This shift makes AP acquisition difficult or impossible in older children and adults — in the paper’s own closing phrasing, the potential is “lost or at least greatly attenuated” (p. 358)

Two qualifications belong with the theory:

• The window alone is not enough. “AP may only develop if early musical training includes the association of pitch names with particular absolute pitches. If musical training focuses solely on the relational aspects of pitch […] then the child may not develop AP” (p. 355).
• The child-vs-adult comparison had never been run. T&H say so themselves: “a rigorous demonstration that the same techniques for training AP are more successful with children than with adults is still lacking” (p. 356).

The proposed mechanism is also softer than it is usually reported. It is not neural maturation or a closing window of cortical plasticity but a shift in attentional preference: young children “prefer to process the absolute rather than the relative pitches of musical stimuli. As a child matures, this preference shifts toward relative pitch, and we speculate that as a consequence, the ability to process pitches absolutely declines” (p. 356). This matters when comparing the paper with Gervain 2013, which is about neurochemical plasticity — a claim T&H never make.

Teaching: a recommendation inside a dilemma

Takeuchi & Hulse are careful to frame this as a problem they cannot resolve. The early-learning theory, “if it turns out to be valid as further data accrue, raises a dilemma regarding the proper content of early musical training” — because music itself is built on relations: “children must learn to perceive relative features of music to understand music” (p. 359). Within that dilemma, their suggestion is one of sequence:

“We suggest that teaching should follow the natural developmental sequence and begin by emphasizing the absolute qualities of tones and reinforcing absolute-pitch-naming behavior. Although the data are not yet conclusive, we believe the child may then develop AP. Subsequent emphasis on relations among tones should direct the child toward perceiving pitch relations directly, without first identifying the absolute pitches.” — Takeuchi & Hulse, 1993 (p. 359)

Two things in that passage are easy to lose. The child may develop AP — the sentence opens with “Although the data are not yet conclusive”, and the hedge is the point. And the second half is the argument from the interval section above: the goal of the later, relational stage is perceiving pitch relations directly, not routing through note names.

The recommendation itself is explicit — “Thus, we encourage early musical training of absolute pitch perception” — and so is the guard against the obvious misreading: such training “should not supplant training of relative pitch perception, which is an indispensable part of music perception and performance” (p. 359).

• Takeuchi & Hulse take no position. They offer two readings of why the family studies disagree with each other — differing criteria or respondent bias, or, “alternatively, the inconsistency may arise because there is no genetic basis for AP, whereas early environment does affect the development of AP, and heredity and early environment are generally confounded in these studies” (p. 357)
• The family evidence itself points in every direction: Bachem 1955 — 40 of 96 possessors claimed relatives with AP; van Krevelen 1951 — only 2 of 17 did, though 14 had musician parents; Spender 1980 — 87% of a group of specially gifted concert performers claimed AP and none reported a sibling or parent with it
• And family studies “cannot necessarily be taken as evidence supporting this view because of the confound that generally exists in these studies between heredity and environment” — they also rely on unconfirmed reports about relatives, with no standard criterion for who counts as a possessor
• Higher incidence among congenitally blind (Bachem 1940; Oakes 1955) — T&H use this against a hereditary account, not for one: it “raises a difficulty for the idea that AP is hereditary unless one is willing to believe that congenital blindness and AP are genetically linked” (p. 358). The natural reading is environmental — a blind child attends more to sound
• Profita & Bidder 1988: found familial aggregation common — it was “common for children of AP possessors to have AP” — but could locate no twin pairs with AP and no couples where both parents had AP: precisely the designs that would separate genes from shared environment
• Resolution attempted: “What is inherited is not AP itself but the predisposition to develop AP” (Bachem 1955; Weinert 1929)
• T&H reject the move flatly: “However, there is no evidence to support this claim” (p. 358)
• (Later work went further, though less far than is often claimed: Baharloo et al. 1998 documented familial aggregation from self-report and collected no DNA, while Theusch et al. 2009 ran the first genome-wide linkage scan — where the significant result held only in the subgroup of 45 European families.)

1993 verdict: absence of evidence, not demonstrated impossibility

This is the claim the paper is best known for, and it is worth quoting exactly, because the reputation is stronger than the text. Reviewing the adult training literature, Takeuchi & Hulse conclude that subjects “generally did not reach an unqualified level of accuracy of absolute pitch identification comparable to that of AP possessors. These studies are therefore not strong evidence that AP can be learned by adults” (p. 358).

That is a statement about the quality of the existing studies. The paper does contain stronger phrasings — “Adults are generally unable to learn AP because they can no longer perceive the pitch of a single tone outside of its context within other tones” (p. 356) — but it never claims a demonstration, and it says outright that the decisive comparison was never made (see the theory section above).

• Several studies did show improvement after training (Gough 1922; Hartman 1954; Lundin 1963; Lundin & Allen 1962; Meyer 1899; Mull 1925; Terman 1965; Vianello & Evans 1968; Wedell 1934; Brady 1970; Cuddy 1968, 1970)
• But in the vast majority of cases, adults did not reach performance comparable to AP possessors
• Two general methods: (a) play notes, ask for identification, supply the answer; (b) discrimination training from a single standard note
• Brady 1970 (N=1, self-taught): reported 96.5% correct within one semitone, without feedback — a lenient criterion. And the limits were unlike those of AP possessors: he could not identify simultaneous tones, nor pitches in any musical context other than unaccompanied melody
• Carroll 1975 appeared to replicate Brady, but subjects answered on a second piano, which fed back the correct pitch
• Lundin & Allen 1962: the two subjects who exceeded 95% at post-test already described themselves as AP possessors beforehand (pre-test 54% and 42%), so the gain is attributable to familiarity with the procedure
• Cuddy 1968, 1970: reported the single-standard method to be superior, but Heller & Auerbach (1972) found no difference between methods and suggested Cuddy’s result was due to factors other than the training method
• Rush 1989: commercial AP training program — no test of retention, and none of generalization to notes outside the training set