🧬 Genetic and Nongenetic Components of Absolute Pitch

📋 Study Overview

🎯 Research Question

Can the AP phenotype be pinned down precisely enough to serve as the basis for future genetic mapping — and, along the way, is early musical training on its own enough to produce AP?

The paper states three aims for the survey: “(1) to determine the percentage of musicians in this sample who claim to possess AP; (2) to ascertain whether the age at first formal musical training correlates with the development of AP; and (3) to determine whether AP aggregates in families.”

This is above all an instrument paper — its title says so. Its products are the AP-1 to AP-4 classification and a cohort of families for later mapping. It estimates no heritability and partitions no variance, and no DNA was collected or analysed here. The genome-wide work it was preparing for arrived eleven years later, in Theusch et al. (2009).

🔬 Methodology

Survey (N = 612)

Three main areas assessed:

1. AP status: Self-reported ability to recognize pitch without reference, with specific questions about speed, accuracy, instrument-dependence, and vocal production
2. Musical training: Age at first formal lessons, extent of training, methods used
3. Family history: Presence of AP among first-degree relatives

Objective Testing (99 self-reported AP possessors + 12 controls)

• Where the 99 came from: a first round of 48 self-reported AP possessors — 20 identified through the survey, 28 who “came to our attention through word of mouth among musicians” — tested alongside 12 non-AP musicians of similar training as controls; then 51 more self-reported AP possessors who likewise “came to our attention, via word of mouth, after the initial survey”
• Two stimulus types: 40 pure sine-wave tones (C2–G#8) and 40 piano tones (C1–G#7, from the McGill University Steinway samples). 40 trials per test, in 4 blocks of 10, 3 s apart; the note name written down immediately, with no reference tone
• Anti-cheating: successive tones separated by more than two octaves and a semitone, so relative pitch could not bridge them; no practice, no feedback until the end
• Scoring: 1 point for the exact note, ¾ point for a semitone error, 0 beyond that. The four tones in each test’s extreme octave were dropped from scoring, so the maximum is 36 points per test. Subjects over 45 got full credit for semitone errors (see “AP drifts with age” below)
• Chance level: none is reported. The paper does not say whether the answer ignored the octave, and never computes a chance baseline, and the ¾-point semitone credit means a near miss also scores — so the raw scores below cannot be read as a percentage above guessing

📊 Key Findings

1. Self-Reported AP in This Sample: 15%

Of 612 respondents, 92 (15%) self-reported AP, with no significant difference between men and women (the paper gives this as “data not shown”). This is a rate of self-report among elite conservatory and orchestra musicians — not a population estimate, and the paper offers none. The authors’ own first caveat is that the sampling “may not have been completely random” and that “it is possible that response rates differed” between musicians with and without self-reported AP, without saying in which direction.

2. The Age-of-Training Gradient

The steep decline (40% → 27% → 8% → 4% → 2.7%) is a strong correlation — which is the paper’s own word for it: “These results indicate that there is a correlation between early musical training and the development of AP.” The authors then set out two competing readings and choose neither. One is a developmental critical period, of the kind demonstrated for song learning in songbirds (Doupe 1993) and for language in humans (Neville 1991) — note that the demonstration is in birdsong and language, not in AP. The other is the reverse arrow: that children predisposed to AP are the ones who start lessons early. See “The Chicken-or-Egg Caution” below.

And the column that is easy to skip: 60% of those who began before age 4 did not report AP. The authors themselves anchor the “necessary but not sufficient” conclusion on the wider band — of the 232 who began by age 6, 69% reported no AP. Note also that the table contains 7 self-reported AP possessors who started at 9 or later (4 of 104, and 3 of 112), so even the “necessary” half is the paper’s “nearly all”, not a strict rule.

3. Familial Aggregation: 4x More Reports of an AP Relative

χ² = 38.6, 1 df, P < 10⁻⁵ — self-reported AP possessors were about 4 times more likely to report a first-degree relative with AP (48% vs 14%).

No relative was tested for this comparison. Both figures rest on a respondent judging someone else’s AP, which the authors flag as a caveat: “some respondents may have been more critical than others in judging their own or their family members’ AP ability.” Someone who has AP knows what it is and recognises it in a sibling; someone who does not, may not — so this measures a rate of reporting, not a measured family prevalence.

4. Sibling Sub-Analysis — Among Siblings Who Also Started Before Age 6

A sub-analysis restricted to families in which both the respondent and one or more siblings had received musical training before age 6:

• Siblings of self-reported AP possessors: 9 of 15 (60%) were reported by the probands to possess AP
• Siblings of respondents without self-reported AP: 2 of 23 (9%) were reported by the probands to possess AP
• P = .0001 — the paper does not say which test it used for this comparison; the Methods allow either χ² contingency tables or Fisher’s exact test

That is a reported rate roughly 6.7 times higher. Two things it does not establish. First, none of these 38 siblings was tested: the paper attributes both numbers to proband report, and mentions no testing of them, using the same phrasing for each group — they “were reported by the probands to possess AP”. Second, starting lessons before 6 is the only variable equalised here. Growing up beside someone with AP is not the same auditory household as growing up without one, so the environment is matched on one point, not held identical. The authors’ own conclusion keeps the hedge that matters:

"These data support the conclusion that there may be a genetic component to AP." — Baharloo et al., 1998 (p. 228)

5. Four AP Categories — and a Fifth Group That Fits None

This classification is the paper’s durable product, and it is worth being exact about what it is: score bands on one pair of tests, not four kinds of ear. The categories were cut out of a continuous distribution, and the paper says as much — individuals “whose scores were most deviated above the mean were considered to have the most clear-cut AP”. AP-1 exists because genetic mapping needs a strict phenotype, not because a natural boundary was found at that score.

AP-4, stated carefully: these individuals score markedly worse on pure tones than the other three groups — below 22.11 of 36, which still leaves plenty of room for scores far above guessing, so saying they cannot identify pure tones would overstate it — while judging piano tones “with almost complete accuracy”. The authors speculate that this could rest on a different basis, timbre and harmonics rather than the fundamental frequency, and they keep the hedge every time they say it: AP-4 individuals “may have an inherent ability to make use of…” those cues. It was not tested.

And the 17%: 17 of the 99 people who said they had AP cleared none of the four thresholds — 10 of the first 48 tested, 7 of the next 51. The authors read the same data the other way, and their sentence belongs here: “the results of our auditory testing suggest that self-reports of AP were consistent with objectively measured AP ability, in most cases.” Both halves are true, and together they are a first-order result about how far self-report of AP can be trusted, and it is why the four category shares above add up to 83% rather than 100%.

6. Family Testing: Concordance Among Relatives Who Already Claimed AP

In 11 families at least one further relative was available and tested — and the paper is explicit about who got in: “those relatives of AP possessors who were readily available and who were self-described AP possessors”. In 10 of the 11, every tested member landed in the same category (AP-1). In the eleventh (family 10) the proband scored into AP-1 and his brother into AP-2. One family had 5 members tested (family 11), another 4 (family 9). In 5 of these families there were still other relatives with reported AP whom the authors did not test.

The concordance is therefore real but narrower than it sounds: relatives who already described themselves as AP possessors tend to land in the same score band. That is not the same as showing that AP segregates through families like a Mendelian trait.

7. AP Drifts With Age — and This One Was Confirmed by Testing

Of the 20 tested AP possessors over 45 years of age, 16 reported that their pitch perception had shifted, by as much as a semitone, as they had become older. The paper then adds the sentence that makes this different from most of its other family and sibling data: “This claim was confirmed by the auditory tests of these individuals.” It is one of the few things here that was reported and then objectively verified — and it is why the scoring gives full credit for semitone errors in that group.

💡 Main Conclusions

"Thus, our results are consistent with the hypothesis that early musical training is necessary but not sufficient for the development of AP." — Baharloo et al., 1998 (p. 229)

The Two-Factor Model the Authors Propose:

The paper’s own summary states it as a suggestion, not a demonstration: “These data suggest that both early musical training and genetic predisposition are needed for the development of AP.”

• Genetic predisposition would create the potential for AP
• Early musical training would realise that potential, within the proposed critical period
• Both needed: most of those trained by age 6 did not report AP (69% of 232), which the authors say “suggests a genetic contribution to the development of this phenotype”; those trained late rarely report it. The study did no genotyping and cannot assign a cause case by case — a person in that 60% may lack a predisposition, or may have had early lessons that never involved naming pitches, or may have started at three years and eleven months instead of at two
• Inheritance pattern: the authors write that the pattern is “compatible with” the autosomal-dominant-with-incomplete-penetrance model proposed by Profita & Bidder (1988), with penetrance possibly “influenced by early musical training”. Compatible with a model is not evidence for it over the alternatives

The Chicken-or-Egg Caution:

The authors explicitly acknowledge an alternative interpretation:

"Alternatively, it is possible that individuals who are genetically predisposed to develop AP may be more likely than others to start musical training early in life. Thus, AP may be part of the general phenomenon of musicality, and an early interest in music could result from greater tonal acuity and increased awareness of sounds, in predisposed children." — Baharloo et al., 1998 (p. 229)

In other words: early training might not cause AP — latent AP might cause early musical interest. The direction of causation is not resolved by this study.

What This Paper Was Built For:

• Its stated purpose was to construct a usable phenotype, not to settle nature versus nurture: the study was undertaken “to characterize the AP phenotype, to set the stage for future studies aimed at identification of genes that predispose to AP”
• The AP-1 band was designed as the strict filter for that later work: “We propose to restrict future genetic mapping studies to AP-1 individuals and their families”
• The genome-wide linkage study that later pursued that goal is Theusch et al. (2009), which reported linkage to chromosome 8q24.21

⚠️ Limitations

• Two different samples, not one: of the 99 people who took the auditory test, only 20 came from the 612 survey respondents — the other 79 were recruited by word of mouth among musicians. The AP-1 to AP-4 categories and the “~70% AP-1” describe that self-selected group, not the surveyed population, and the two must not be read as one sample of 612
• Self-report throughout: AP status, relatives’ AP and age at first training all rest on the respondent’s judgement. Objective testing placed 83% of the self-reported possessors into some category, but 17% cleared none of the four thresholds
• No relative was tested in the key family analyses: both the 4x familial figure and the 9-of-15 sibling result are proband report. Only relatives who already described themselves as AP possessors were tested at all
• Ascertainment bias, admitted by the authors: sampling from elite institutions “may not have been completely random”, and “it is possible that response rates differed” between musicians with and without self-reported AP
• Recall bias: age at first training may be imprecisely remembered
• No molecular genetics at all: no DNA, no genotyping, no linkage analysis, no heritability estimate. Every genetic claim in this paper is indirect
• Causation ambiguity: cannot distinguish whether early training enables AP or AP-predisposed children seek out music earlier (the chicken-or-egg problem)

🔗 Related Research

• Commentary: Gregersen (1998) — invited editorial in same journal issue, discussing implications of this study
• Follow-up genetic study: Theusch et al. (2009) — genome-wide linkage study in AP families, reporting linkage to chromosome 8q24.21
• Precursor: Profita & Bidder (1988) — earlier family study suggesting autosomal dominant inheritance
• Brain structure: Schlaug et al. (1995) — PT asymmetry in AP musicians
• Tone language challenge: Deutsch et al. (2006) — reported far higher AP rates among tone-language-speaking conservatory students, which they read as an environmental contribution — and they cite this study on both sides, for the age gradient and as the genetic alternative to their own reading (“see also Baharloo et al., 1998”). Baharloo, eight years earlier, could not answer back. This is the environmental contribution; the two studies are not weighed against each other by either

📖 Access Full Study

📚 Full Citation