This Article Reports Two Experiments. In the first experiment, 13 professional singers performed a vocal exercise consisting of three ascending and descending melodic intervals: minor second, tritone, and perfect fifth. Seconds were sung more narrowly but fifths more widely in both directions, as compared to their equally tempered counterparts. In the second experiment, intonation accuracy in performances recorded from the first experiment was evaluated in a listening test. Tritones and fifths were more frequently classified as out of tune than seconds. Good correspondence was found between interval tuning and the listeners responses. The performers themselves evaluated their performance almost randomly in the immediate post-performance situation but acted comparably to the independent group after listening to their own recording. The data suggest that melodic intervals may be, on an average, 20 to 25 cents out of tune and still be estimated as correctly tuned by expert listeners.

The present article focuses on the question of whether the timbre difference of two sounds with harmonic spectra, produced by natural musical instruments or the singing voice, may influence subjective assessments of the pitch of one sound in relation to the pitch of the other. The authors administered a series of perception tests to a group of professional musicians (n = 13) and a group of non-musicians (n = 13). The tests used the following pre-recorded sounds: the singing voice, the sound of the viola, and the sound of the trumpet. The participants had to compare the pitch of pairwise presented successive tones and decide whether the second tone was either 'flat', 'sharp' or 'in tune'. Tests using stimuli in the pitch range around A3 (220 Hz) at a loudness level of approximately 90 phons revealed pitch shifts of significant magnitude likely to affect intonation quality in a musical performance among both musicians and non-musicians. The conclusion drawn from the study is that timbre-induced pitch shifts may attain magnitudes that are likely to lead to conflicts between subjective and fundamental-frequency-based pitch assessments. Situations are described in which such conflicts may arise in actual musical practice.

Fifteen professional singers sang simple vocal exercises at different pitches. The sung excerpts were recorded, and seven modified versions were created from each recording. The modifications were then played to the participant and (s)he had to assess the similarity of these stimuli to the perception of his/her own voice during the act of singing. Participants rated as most similar those stimuli which were modified by the filter whose frequency response most closely resembled a trapezoid, which was created by taking into account 1) the diffracting air conduction component from the mouth of the singer to his ear channel, 2) the bone conduction component, and also 3) the influence of the stapedius reflex on the sensitivity of his/her auditory system.

The voices of 42 students studying classical opera singing at the Estonian Academy of Music were investigated to find any objectively definable qualities possibly correlating with the length of training. Each student's singing of a four-bar sevenword initial phrase from a well-known Estonian classical solo was recorded. The recordings were digitalized and subjected to acoustic analysis yielding the longterm average spectrum (LTAS) for each voice studied. It turned out that the longer a singing student had been trained professionally, the higher was the level of the so-called singer's formant in her/his LTAS. Subsequently the voice quality in each recording was evaluated by four experts using a five-point scale, five points marking the best quality and one point the poorest. It turned out that the average ratings did not show any positive correlation with the length of training, rather, a slightly negative trend (not statistically significant) could be observed. The results seem to support the critical remarks made by some Estonian specialists about domestic teaching of vocal music being perhaps inadequate in some respects (Pappel, 1990). The teaching process seemsto be focused on the development of those qualities that enable the singer to be audible in large halls and with a symphony orchestra, while the timbral qualities recede into the background.

Singing teachers sometimes characterize voice quality in terms of 'forward' and 'backward placement'. In view of traditional knowledge about voice production, it is hard to explain any possible acoustic or articulatory differences between the voices so 'placed'. We have synthesized a number of three-tone melodic excerpts performed by the singing voice. Formant frequencies, and the level and frequency of the singer's formant were varied across the stimuli. Results of a listening test show that the stimuli which were perceived as 'placed forward', correlated not only with higher frequencies of the first and second formants, but also with the higher frequency and level of the singer's formant.