INTRODUCTION

ABSOLUTE pitch (AP) is the rare ability to identify or produce a given pitch without a reference tone. This ability is often associated with a higher level of musicianship (Levitin, 2008). This seems obvious, since only musicians can identify a tone by correct verbal labeling. However, it has been shown that even children are able to recall familiar songs on the precise pitch level that they had used or heard many times before. This type of latent AP also indicates a relevant portion of pitch memory which is involved in the recall of pitch, independent of its octave invariant chroma or physical spectrum (Jakubowski et al., 2017). Therefore, two sub-processes of AP recognition can be differentiated: perceptual structures of processing pitch chroma, and cognitive associations with a verbal label (Elmer et al., 2015; Kim & Knösche, 2017a).

During auditory perception, the human brain analyzes both time and frequency simultaneously. Since information for AP comes from sound spectrum and time processing, time- frequency analysis of musical stimuli in AP might be an essential tool. Moreover, there is psychophysical evidence that a musical tone is characterized by two distinct dimensions: Pitch height, which represents a linear dimension caused by the increase of frequencies (e.g. by transposition by an octave), and pitch chroma, based on a cyclic octave-independent dimension of recurring tone qualities, so-called Tonigkeit (Révész, 1926; Wellek, 1963). Both have distinct representations in the human auditory cortex (Warren et al., 2003) and are crucial for absolute pitch recognition. Whilst chroma determines the harmonic and melodic aspects of music, it is robust in timbre and dynamics. Although it is not yet well understood how both pitch properties interact in AP, it seems clear that chroma is highly relevant to pitch recognition ability (Korpell, 1965), when auditory perception and identification tasks focus on pitch class rather than octave position.

It is important to consider different dimensions of sound perception that account for clearly distinguishable modes of pitch perception. These depend upon which aspect of the sound is of primary focus: either tones with independent recognition of timbre and fundamental pitch, or spectral components. Thus, Schneider and Wengenroth (2009) differentiated between two types of listeners according to their aural orientation: holistic (or fundamental) and spectral listeners. It has been demonstrated that both modes are reflected by structural and functional asymmetries in Heschl's gyrus (Schneider et al., 2005a). It remains to be shown whether AP, as a special hearing ability, is influenced by the pitch perception preference, indicated by the individual pitch perception index measured in the Pitch Perception Preference Test (PPPT, Schneider & Bleeck, 2005; Schneider et al., 2005b).

However, a great deal of discourse concerning AP centers around the question of whether it is acquired through early exposure and practice, or whether it is characterized by neural predispositions (Brown et al., 2002; Levitin & Zatorre, 2003; Ward, 1999). Twin studies reveal a familial aggregation of AP (Baharloo et al., 2000), whereas others query a simple inheritance in place of an effect of environmental, epigenetic and stochastic factors contributing to the genesis of AP (Theusch & Gitschier, 2011). Therefore, the focus here is on both presumably interacting resources. A neural predisposition for acquiring AP introduces a genetic factor, whereas exposure and early musical training during a critical period (Zatorre, 2003) brings environmental aspects of learning and culture into play. Native speakers of tonal languages (e.g. Mandarin, Cantonese, Southern Min, Vietnamese etc.) who have AP support the importance of early learning, since small children can learn to correctly pronounce a pitch-accent language, and through this develop their absolute pitch memory for tone syllables (Deutsch et al., 1999; 2004). However, the nature-nurture debate continues, and suggests an interaction of both resources; even the evidence of neural correlates and detectable substrates cannot rule out the possibility that neural changes are a possible consequence of training and early exposure.

When investigating AP, one inevitably invokes AP, as well as relative pitch (RP), with the memory system (Deutsch, 1999; Levitin, 1999). Previous research has shown that AP possessors remain more stable in their high recognition rate compared to non-possessors in a long-term memory task (Rakowski & Rogowski, 2007). Similar results are reported by Siegel (1974), showing that AP possessors substantially outperform non-possessors after a 15-seconds retention interval, when the difference between two stimuli was a semitone compared to 1/10 semitone. This might be attributed to a different coding strategy in listeners that leads to different mental representations. Therefore, AP may be seen, to some extent, as a form of tonal memory and recall from musical experience.

Much research has focused on further behavioral aspects of pitch identification (Deutsch, 2013). Effects of instrumental timbre (West, Marvin & Brinkman, 2000) and key color (Miyazaki, 1988; 1989) have been identified. Differences of reaction times (RT) in pitch labeling tasks appear between black and white key notes, with responses to white-key tones being significantly faster than to black-key tones (Bermudez & Zatorre, 2009; Miyazaki, 1989). Similarly, response time for piano tones is quicker than for string tones (West, Marvin & Brinkman, 2000), which may indicate varied degrees of easiness and fluency according to instrumental timbre. Interestingly, an interaction between AP and RP in reaction times has been found; a more global processing of listening tasks, as in AP tests, may slow down the reaction time (Ziv & Radin, 2014). The obvious prevalence of AP for different pitch classes might differ with the frequency of occurrence in the music to which the listener has been exposed (Huron, 2006).

Recently, the investigation of neural mechanisms has developed a better understanding of the underlying procedures that might explain absolute pitch processing, although such studies may be more intriguing for neuroscientists rather than musical practitioners (Schlemmer, 2008). Just recently, Hou et al. (2017) reviewed the topical literature of neuroscientific research on AP, and recapitulated the scientific knowledge of neural processes involved in AP. Here, we briefly summarize the main results. For processing pitch chroma the ventral pathway plays a critical role (Kim & Knösche, 2017b). Furthermore, left-hemisphere mechanisms facilitate pitch processing (Behroozmand et al., 2014). AP possessors also show an increased functional connectivity during music listening (Loui et al., 2012b), and a hyperconnectivity in bilateral superior temporal lobe structures (Loui et al., 2011). AP musicians exhibit a significant smaller size of the right planum temporale (Keenan et al., 2001). This increased asymmetry in AP musicians may be ascribed to genetic influences (Keenan et al., 2001). Wengenroth et al. (2014) found that gray matter volume of the right Heschl's gyrus is highly correlated with AP, and the primary auditory responses localize more posteriorly in the right hemisphere. However, the accuracy of AP skills may also be related to an early exposure to cultural norms (Hedger et al., 2013).

Finally, the memory system of AP might partly relate to synesthesia because colors and visual clues may function as markers for the characteristics of pitch chroma. Therefore, synesthesia often co-occurs with AP (Glasser, 2018; Gregersen et al., 2013). Research has demonstrated that there are shared as well as distinct neural substrates between both abilities reflected by common activation patterns (Loui et al., 2012a).

In general, the large number of research studies on AP (for a survey, see Deutsch, 2013) has revealed different dimensions involved in AP, such as perceptual and cognitive development, habitual disposition, environmental and genetic factors, and the effects of neuroanatomical conditions. In the present research, other phenomenological aspects beyond neuroscientific approaches to AP will be empirically investigated, such as the stability of the hit rate when physical sound parameters are manipulated, which might influence the accuracy of pitch recognition. The following questions stimulated the current study:

• Do pure sinusoidal tones and complex sounds differ in AP possessors' perception?
• Is the spectral structure of a complex sound relevant to pitch recognition? Does it affect the accuracy of pitch recognition?
• Which psychoacoustic parameter has the strongest influence on pitch recognition?
• Is there any correlation between the beginning of formal instrumental instruction and AP performance accuracy?
• Does the individual pitch perception preference impact pitch labeling accuracy in AP possessors?

METHOD

RESULTS

AP performance with manipulated stimuli

The total scores of TTAP significantly correlated with the Schneider AP test (r = .894, p < .01) which validates the data based on TTAP. There is also a stronger relationship between TTAP and mean accuracy for items with the manipulated sound structure of missing fundamentals (r = -.965, p < .01; see Figure 2).

A comparison of the two types of AP possessors (full and partial) displays a clear differentiation between the mean scores of full AP possessors and those with partial AP (Figure 3). They score significantly lower than full AP possessors (p < .05), who responded with more correct answers. This indicates a stronger impact of psycho-acoustic parameters on the decrease of pitch recognition accuracy for partial AP possessors.

With respect to the absolute hit rate, there is a clear decrease of correct hits from violin (1.65), through to the clarinet (1.74), trumpet (1.83) detuned (1.88), missing fundamental (1.97), and sine wave stimuli (2.06). All means are significantly different (p < .05), except between violin and clarinet stimuli, and between trumpet and detuned stimuli (Table 1).

Table 1. Descriptive statistics for all items in the TTAP.

Stimuli	Violin	Clarinet	Trumpet	Detuned	Missing f.	Sinus
N	48	48	48	48	48	48
Mean	1.65	1.74	1.83	1.88	1.97	2.06
Median	1.50	1.57	1.60	1.75	1.89	1.97
SD	.59395	.59535	.69825	.67321	.69063	.63947

DISCUSSION

The study relates the obvious phenomenological aspects of AP to the psychophysical scaffolding of pitch recognition in AP possessors. In view of this relation, AP constitutes a musical skill that results from genetic factors, and from environmental impact that originates from learning and practice. AP, then, represents a special form of a musical trait that seems to come naturally (one has it or not), but it can also, at least partly, be acquired. The first (one has it) is due to genetics, the latter (partly acquired) to memory and training. Both may occur independently, but they can also interact. The current, significant main effect of an early start in instrumental instruction underscores the importance of an environmentally supported training effect and confirms earlier findings (Sergeant, 1969).

One can assume that the portion of AP that is acquired and memorized is more error-prone than the genetically determined ability, because the former is founded in clear neuronal correlates beyond training, acquaintance or usage and, therefore, might stabilize pitch recognition. This study aimed to show how stable pitch recognition accuracy would be in a broader sample of AP possessors.

Musicians are accustomed to the sound of natural musical instruments. However, the current study also dealt with psychoacoustic manipulations of the physical structure of a perceived pitch. Why does it make a difference to listen to a natural instrumental sound, or to sounds that are slightly detuned or characterized by a harmonic spectrum with a missing fundamental (that is, when the perceived pitch frequency is not physically present, but must be generated in the musician's mind due to the overtone spectrum of that sound)? Why is a pure tone not similarly perceived as a complex tone with its rich overtone spectrum? Musicians exhibit an expert knowledge of sounds by the experience of daily practice. They "feel" that both types of sound structures are acoustically different. Therefore, it was expected that musicians respond to varying types of stimuli in our test differently, and that varied instrumentalists process pitches differently according to the sound spectra of their familiar instruments. Here, the question was how stable pitch labeling accuracy arises under those various conditions.

The data of the test confirm that there is a clear hierarchy of accurate pitch recognition from pitches with a natural spectrum of overtones down to a pure sinusoidal sound. The individual percept of complex instrumental sounds is obviously different from other sound sources, so that manipulated sounds are less accurate in an AP recognition test than natural instrumental sounds, although all items were synthetically produced. It seems reasonable that a sound can best deploy its unique characteristics through an instrument. Moreover, the sound spectrum that is responsible for the instrumental timbre might influence pitch detection in the case of spectral listeners, but less so for fundamental pitch listeners. For fundamental pitch listeners, the pitch percept of a complex sound should be independent of the shape and intensity of the sound spectrum due to its specific focus on the fundamental pitch. In contrast, the pitch percept of spectral listeners may be influenced by the sound spectrum characteristics, due to their sensitivity to spectral pitches and formants. This is clearly reflected in the difficulty of spectral listeners to recognize the correct (fundamental) pitch of the physically manipulated tones. However, pitch chroma is strongly connected with the overtone spectrum, but independent from the formant structure that determines instrumental timbre. Consequently, any intervention of the sound structure by physical manipulation (e.g. by eliminating the fundamentals or detuning elements of the spectrum) will necessarily influence the recognition rate.

On the other hand, pure tones exhibit an astonishing failure in recognition, which confirms the findings of Lee and Lee (2010) and Glasser (2018). There are several reasons for this: first, the weaker recognition ability of sinusoidal tones might be due to the fact that the pitch height of pure tones are less pronounced in higher and lower pitch ranges (Terhardt, 1998). Furthermore, the detailed analyses of the PPP test (Schneider et al., 2005b) demonstrate that the strength of the fundamental pitch percept correlates with the average number of harmonics that are physically present in a missing fundamental sound. In the case of an incomplete complex sound with only two present harmonics, spectral listeners perceive just the spectral pitches corresponding to the frequencies of the two harmonics. In the case of three or more harmonics, the spectral listeners show a slow, systematic increase in strength of fundamental pitch perception, even though the fundamental pitch is still missing. For fundamental listeners, the fundamental pitch percept is always present independent of changes in the sound spectrum, but also strengthens with an increasing number of harmonics (Schneider, Sluming et al., 2005b). Considering these observations, it is not surprising that sinusoidal tones are correctly recognized less often than natural complex tones, although they clearly indicate pitch height in terms of frequency. Therefore, spectral chroma seems to be a prerequisite for pitch detection, independent of individual differences in pitch perception preference.

Furthermore, in our test it seems remarkable that melodic instrument players always performed at a higher level compared with pianists (see Figure 6). However, pianists are more stable in their correct pitch detection rate, despite a striking drop of correct detection in pure tones. These inverse observations for pianists and melodic instrument players may be explained by the fact that pianists on average show a considerably stronger fundamental pitch percept compared to melodic instrument players and singers (Schneider et al., 2005a). Nevertheless, their higher spectral pitch awareness may have an influence on the performance accuracy of the octave-independent pitch chroma perception. Overall, the strength of AP perception might be considerably enhanced if pitch chroma is not only derived from fundamental pitch, but also from an octave-shifted fundamental pitch with the same chroma. Indeed, the recognition of octave-shifted pitches occurs more frequently in spectral listeners (Schneider et al., 2005b; Schneider & Wengenroth, 2009). As melodic instrument players show a higher rate of spectral pitch perception compared to pianists, they should consequently have a higher AP accuracy compared to fundamental pitch listeners, who derive the AP information predominantly from the non-octavated fundamental pitch. The influence of octave ambiguity on pitch labeling accuracy may be underestimated so far.

It seems plausible that error-proneness is higher when more memory is involved. If recognition is derived from a neurally determined ability, it can be expected that the accuracy of pitch recognition is higher. Therefore, further research might focus on the relation of neural correlates with empirical accuracy of pitch recognition in AP tests, and investigate to what extent neural correlates predict the error-rate or pitch detection accuracy.

The missing effects of the perception indices on AP scores was unexpected, but might be due to the uneven distribution (see Figure 7) of listening modes and instrumental orientation of the participants. The majority consisted of designated holistic listeners (81%) with a predominance of piano players that is purported to be a preferred instrument of this type of sound perception (Schneider et al., 2005a). Both listening modes (holistic vs spectral) are neurally determined and do not exhibit training effects. The measurable perception index refers to a mental disposition of sound processing that is based on neural traits that obviously develop independently from the neural conditions for AP abilities. In contrast, AP has, beyond the neural correlates, a strong association with memory that is irrelevant for the perception index. Therefore, it is reasonable to suggest that perception indices do not qualify as predictors of AP competence.

The apparent differences between black and white keys in pitch recognition shed light on the importance of exposure to music, and the practical acquisition of percepts developed by playing a musical instrument. The superiority of white keys compared to black keys mirrors the statistical dominance of tones without alterations in traditional classical music. This underpins once more the importance of cultural imprints beyond genetics. The Western tonal system forms European instruments as well as the mental representation of their sounds. Consequently, pitches of a Western tonal system are more stably represented than other detuned pitches. To mark the traits of those pitches, the ear needs information from the sound structure; the more this is modified, or even violated, the less accurate pitch detection may be.

An interesting result was found regarding visual clues during pitch recognition in the AP test. Although we cannot assert strong evidence for synesthesia, the small number of participants who refer to visual clues may nevertheless be an indication of a possible involvement of synesthetic effects in pitch recognition, supporting the idea that synesthesia and AP represent "two sides of the same coin" (Loui et al., 2012a). This is confirmed by a study that demonstrates that most AP cases in the sample met the diagnostic criteria for synesthesia (Glasser, 2018). In fact, with regard to a multisensory integration, one can assume that tonal and visual modes of processing are connected in AP (Loui et al., 2012a; Spence & Soto-Faraco, 2010), but this requires further investigation.

From this study, it can be concluded that pitch recognition in AP possessors is not just a distinct ability based on simple mechanisms. Instead, it is based on a mutual interaction of genetic and environmental, mental and memory dependent factors that influence the reliability of pitch labeling ability even in AP possessors.