🎹 Musical Pitch Identification by Absolute Pitch Possessors

📋 Study Overview

🎯 Research Question

How precisely can absolute pitch possessors identify isolated musical tones — and can a task be built that separates them from listeners who arrive at the same answers by other routes?

Miyazaki states three purposes (p. 502): (1) to develop an experimental task “that clearly separates AP possessors from nonpossessors” — blocking the strategies of what Bachem (1937) had already labelled pseudo AP; (2) to evaluate how accurately AP possessors categorise tones when the semitone is finely subdivided; and (3) to estimate the range and position of their pitch categories along the frequency continuum — which is the reason the stimuli are spaced only 20 cents apart, and not merely an anti-cheating measure.

One question is explicitly not asked. On the innate-versus-learned controversy Miyazaki writes that “the present study was not concerned with this problem” (p. 501).

🔬 Methodology

Two Experiments

Experiment 1 (N=10)

• Stimuli: Sawtooth waves (synthetic — not piano or any recognizable instrument, to eliminate timbre-based identification)
• Pitch resolution: 60 tones per octave, spaced at 20-cent intervals (5 microtonal steps per semitone)
• Range: C3 (130.8 Hz) to C6 (1046.5 Hz), standard A4 = 440 Hz
• Task: Identify the note name via a computer keyboard mimicking a piano; press space bar if the tone sounded "off-tune"
• Anti-cheating design: Each consecutive tone was displaced to a different octave (>1 octave apart), preventing use of relative pitch between trials. No feedback was given.

Experiment 2 (N=39)

• Stimuli: Piano-like tones (FM synthesis, Yamaha TX-816) — more ecologically valid than sawtooth waves
• Pitch resolution: 180 tones across 3 octaves (same 20-cent spacing)
• Enhanced response: For each note, subjects could indicate "in tune," "lower," or "higher" — three subcategories per pitch class, allowing finer measurement of precision
• Response time measured: Time from stimulus onset to keypress, providing a window into the cognitive process
• Response via: Silent Yamaha DX-7 keyboard (natural musical interface)

📊 Key Findings

1. Experiment 1: Seven of Ten Had AP — and One Who Didn’t Looked Like He Did

Seven of the ten listeners turned out to be AP possessors: 94.1% of their responses fell within less than one semitone of the correct note (individual range 90.2–99.4%), against 27.2% for the other three (range 13.9–42.7%). Note what the metric counts: the proportion of responses with an error smaller than one semitone, not exact note naming. The paper publishes no chance level for it, and describes the non-possessors’ responses as “almost randomly dispersed” within a one-octave range (p. 511).

One of those three, subject K.M., produced a pattern that superficially resembled AP. He “indeed, had no AP, but used all available means in executing the task successfully” — among them “the strain in the vocal organ” and relative pitch anchored on the extremes of his own vocal range. Miyazaki adds that even so, “his responses were so widely scattered that they could not compare with those of genuine AP possessors” (p. 502). This is the pseudo-AP case the whole task was designed to expose, and the clearest demonstration in the paper of why the design was needed.

2. Three Performance Groups, Including “Imprecise AP”

In Experiment 2 the 39 participants were divided, after the fact and “on the basis of response accuracy” (p. 505), into three subgroups. The cut-off values are not published, and the three groups do not appear in the paper’s own abstract, which emphasises instead that the results “clearly differentiated AP possessors from nonpossessors”.

The accuracy column is the proportion of responses with an error smaller than one semitone — not exact note naming. So “69.1%” for the imprecise AP group does not mean 69% of notes named correctly. The paper publishes no chance level for this metric; the only chance level it gives anywhere is 58.3%, for white-key response bias (p. 510). Of the non-AP group Miyazaki writes that they “could not consistently tell the musical note name of the presented tone” (p. 505).

The differences in response delay between the three groups were significant [F(2,36) = 22.36, p < .0001]. Miyazaki reads the speed as a sign that AP possessors retrieve a pitch category directly, while the others work through slower strategies — but he also warns that these delays are too long to be treated as reaction times in the ordinary sense, because speed was not what he asked of participants (see Limitations).

3. An Anomalously High Prevalence of AP

This is one of the claims in the paper’s own abstract: “Although AP was believed to be very rare, it was proved here that a phenomenally large proportion of the subjects tested had AP” (p. 501). In Experiment 2, 22 of 39 participants (56%) showed some degree of AP; in Experiment 1, 7 of 10. Miyazaki sets this against the historical rates he himself cites — 8.8% among musicians (Wellek, 1963), 3.4% (Revesz, 1913), 0.01% among non-musicians (Bachem, 1955) — and mentions an informal test of a large number of music students, run on another occasion, in which “no less than half” of them were AP possessors.

The explanation he offers for the excess is cultural, without denying a hereditary component: the proportion “cannot be explained solely by inheritance, because it is known that the proportion of inherited AP is extremely low”, and “the acquisition of AP is dependent upon social or cultural circumstances; at least in the Japanese culture, where relatively early music lessons are to some extent widespread, an unexpectedly large proportion of music students have AP” (p. 511). Which also means the prevalence figure travels badly: it is a statement about this population, not about music students in general.

4. The White-Key Advantage — and Its Confound

Within the precise AP group, white-key notes drew faster responses than black-key notes: 1.524 s against 1.631 s [F(1,11) = 23.17, p < .001]. The fastest categories were G and C. The full rank order was G, C, E, D, A, F, C♯, A♯, B, D♯, F♯, G♯ [F(11,121) = 2.73, p = .0035] — with one exception Miyazaki flags himself: “except for B”. B is a white key and came 9th of 12, behind two black ones.

Accuracy pointed the same way, with two qualifications. The paper reports no significance test for accuracy by key class — the F values above are for response delay — and Miyazaki credits that half of the finding to earlier work: “This has been reported previously by several investigators (Bachem, 1950; Wellek, 1963)” (p. 511). What is new in this paper is the speed advantage.

The same effect appeared in listeners without AP. In the non-AP group, white-key notes drew responses at 2.733 s against 2.957 s for black keys, also significant [F(1,9) = 6.57, p = .03]. Miyazaki argues it “should be interpreted differently” there: those participants put 85.4% of their responses on white keys, against a chance level of 58.3%, and nine of them made almost no black-key responses at all. The white-key advantage is therefore not a signature unique to AP; in the control group it is a keyboard response bias.

What Miyazaki makes of it. He suggests AP possessors “cannot recognize all 12 notes equally well. Instead, they seem to recognize best the white-key notes, or perhaps only 4 or 5 important notes, from which other notes can be inferred” (p. 512), and connects the dominance of C major to early piano lessons that start with the C-major scale and the primary triads in C. But he reports the finding as an awkward one: tonal hierarchy is in principle a relative-pitch phenomenon, so “these results were unexpected”, and he goes no further than saying the dominance of C “may be related to the way in which they acquired AP at an early age”. The paper draws no training recommendation from this.

5. Categorical Perception with Microtonal Precision

In Experiment 2 each note carried three response subcategories — “in tune”, slightly lower, slightly higher. These are categories the listener chose, not properties of the stimulus: an “in-tune” response means the tone sounded acceptable as that note.

• “In-tune” responses: 1.571 sec average
• “Lower” 1.734 sec and “higher” 1.773 sec [F(2,22) = 14.45, p = .0001]; each differs from “in tune” at p < .01, while “lower” versus “higher” is not significant
• In one extremely accurate participant (Figure 5) the “lower”/“higher” responses tracked the actual deviation of the stimulus — a “lower G” response came up mainly when the tone was in fact slightly below G — which suggests categories finer than the standard 12-note tuning system

6. Tone Chroma vs. Tone Height — and Miyazaki’s Actual Argument

The distinction itself is not Miyazaki’s, and the paper says so plainly: tone height and tone chroma come from Revesz (1913) and Bachem (1937), “originally… formalized by Revesz (1913) as the two-component theory of pitch” (p. 511).

• Tone chroma: the musical “character” of a note (its C-ness or D-ness) — cyclic, repeating every octave
• Tone height: simply how high or low a sound is — a linear dimension from bass to treble

Miyazaki’s own contribution is a criticism of how chroma is usually defined. Tying it to absolute pitch, he argues, confines it “to a limited case of AP and cannot be regarded as a general tonal attribute; thus, the two-component theory of pitch loses its generality” (p. 511). His proposal is to redefine chroma in relation to relative pitch — the musical property each tone takes on inside a tonal context — so that listeners without AP perceive chroma too, provided there is context; the AP possessor is peculiar only in perceiving it without one.

That is close to the opposite of treating chroma as the marker of AP. And Miyazaki goes further, in a sentence a site about absolute pitch is tempted to leave out: “This relativity of the musical pitch (relative pitch) is much more important than AP in music” (p. 511).

What the non-possessors did is consistent with this: stripped of musical context, they could sort tones roughly from low to high but could not say which note was which, and their responses within an octave were “almost randomly dispersed”.

💡 Main Conclusions

“It has been held that the uniqueness of AP possessors is that they have a remarkable ability to identify pitch in the absence of any contexts; however, it can be argued that they are actually unique in that they have memorized from early childhood the musical qualities of individual tones (tone chroma) in only one fixed tonal context and, therefore, have had no need to develop relative pitch sense.” — Miyazaki, 1988 (p. 512)

The hedge is his, and it carries the argument: this is a proposal Miyazaki puts against the received view (“it has been held that… however…”), not a finding he reports.

Key Implications:

• A middle band exists in this sample: precise AP, imprecise AP and non-AP, separated by both accuracy and speed. The split was made after the fact, “on the basis of response accuracy”, with thresholds the paper does not publish — and it does not appear in the article’s own abstract, which stresses that the results “clearly differentiated AP possessors from nonpossessors”
• On innate versus learned, Miyazaki declines to take a side. The introduction states that “the present study was not concerned with this problem” (p. 501). In the Discussion he calls the origin of AP “still controversial”, notes that the correlation between AP and piano lessons started at ages 3–5 “may be in line with” the hypothesis that AP is acquired through early learning — a hypothesis he attributes to Shuter-Dyson & Gabriel (1981), not to himself — and adds that “in a few cases it may be inherited” (p. 511)
• What early training may supply is content, not only timing: all 12 precise-AP participants had started piano between 3 and 5 (mean 4.1 years), against 8 of 10 in the imprecise group and 5 of 17 in the non-AP group. Miyazaki’s proposed mechanism is about what is practised: “in the early lessons of piano playing and ear training, they usually began with playing the C-major scale and listening to primary triads in C major” (p. 512). This is a cross-sectional correlation, not a critical-period design — there is no late-starting group and no follow-up
• AP may carry costs: because possessors memorised tones “in only one fixed tonal context”, Miyazaki argues they “have had no need to develop relative pitch sense”, which “may give them not only advantages, but also serious disadvantages” (p. 512). A familiar melody transposed to another key can feel wrong — a downside rarely mentioned in popular accounts of “perfect pitch”
• What the study does not say: there is no adult training here, no longitudinal measurement and no advice about how to practise. “Imprecise AP” is a label for a band of accuracy in this sample — 69.1% of responses landing within one semitone — not a claim that partial pitch identification is musically useful. Miyazaki’s only comment on a listener with intermediate performance is that his responses “could not compare with those of genuine AP possessors”

⚠️ Limitations & Context

Study Limitations

• Japanese population only: All participants were from Niigata University. The high AP prevalence may reflect cultural factors specific to Japan's early music education system
• Synthetic stimuli: Sawtooth waves (Exp. 1) and FM-synthesized piano tones (Exp. 2) may not fully represent the richness of real musical contexts
• No longitudinal data: Cannot determine whether AP precision changes over time or can be improved with practice
• Natural AP only: All participants acquired AP in childhood; adult-trained AP was not examined
• Classification criteria: the three-group division (precise / imprecise / non-AP) was made after the data were in, “on the basis of response accuracy” (p. 505). The threshold values are not published, and no formal clustering method was used
• The response delays are not reaction times in the usual sense — his own caveat. “It should be noted that time delay of the responses observed here is so long that it may [not] be regarded as RT in the ordinary sense” — the “not” is missing in the printed text, but the sentence that follows requires it: “ This is partly because rapid responding was not the primary demand in the experiment.” Participants were told to be as accurate as possible and “if necessary, they were allowed to use any available cues helpful to increase response accuracy”. Miyazaki adds that if speed were demanded first, the non-AP subjects would probably judge on pitch height alone, “resulting in faster RTs” (p. 508) — that is, the speed gap could narrow, or reverse, under a different instruction
• The white-key advantage has a confound visible in the control group: it was also significant among listeners without AP, where Miyazaki attributes it to a keyboard response bias (85.4% of responses on white keys, chance 58.3%) rather than to perception (p. 510)
• Small samples, one institution: ten listeners in Experiment 1 and 39 in Experiment 2, all at the same Japanese university. No measure of relative pitch was taken for comparison

Historical Context (1988 vs. 2020s)

🔗 Related Research

• The continuation Miyazaki himself announced: the paper closes by saying that “another series of experiments” was already under way, on “how AP possessors deal with a relative pitch task in which they are required to identify musical intervals variously transposed” (p. 512) — the line of work he became known for, and not a study of memory interference
• Categorical perception: Siegel & Siegel (1977), Absolute identification of notes and intervals by musicians — AP possessors perceive pitch categorically, with performance Miyazaki describes as “comparable to those of relative pitch subjects” (p. 502). The comparison is with musicians using relative pitch, not with speech sounds
• Where a later study disagrees: Bermudez & Zatorre (2009) name this paper among the “occasional claim of a clearly bimodal distribution among musicians” and report the opposite in their own sample of 51 musicians: performance running continuously from perfect to random, with many intermediates. They replicate the white-key advantage, but find A the best-identified class rather than C and G — a difference they attribute to the all-keyboard sample of a later Miyazaki paper (1990), not to this one, whose Experiment 2 was mostly people with no formal lessons
• Training studies Miyazaki cites but does not engage: Cuddy (1970), Cuddy (1968), Brady (1970), Meyer (1899) and Wedell (1934) appear in his introduction only as background to the origin controversy he then declares out of scope
• Early training correlation: Baharloo et al. (1998) — a survey of 612 musicians reporting self-declared AP falling from 40% (training begun before age 4) to 2.7% (after 12), plus a family-aggregation analysis. No molecular genetics was done; the genetic component there is an inference from self-report
• Category malleability: Hedger et al. (2013) — AP categories can shift with listening experience, which cuts against the idea of a template fixed for life
• Adult training: Wong et al. (2025) — after ~21 h of training, adults raised pitch-naming accuracy from 13.9% to 31.7% (chance 8.3%). Miyazaki tested no adult training himself; the connection is ours

📖 Access Full Study

📚 Full Citation