and intensity (decibels). Applying the Preference for tones as a function of hypothesis to sound frequency is a modest extension of the original, frequency (hertz) and intensity (decibels)* since, in the language of Stevens (1957), the hypothesis has been extended from prothetic continua to a metathetic continuum. In addition, PAUL c. VITZt the present study is a very elementary New York University, New York, New York 10003 experiment in aesthetics. Pure tones are extremely simple aesthetic stimuli, An extension of the Wundt hypothesis concerning stimulus intensity and but nevertheless our reactions to hedonic tone leads to the prediction that Ss prefer tones of intermediate individual tones must be some part of frequency (hertz) and of intermediate intensity (decibels). In Experiment L, our total reaction to music. Of more tones varying from 60 to 5000 Hz were presented in a paired comparison importance is the fact that the procedure, and, as predicted, Ss' preference was an inverted-Usshaped function hypothesis being tested can be :>1 frequency, with the most preferred tones in the range of 400-750 Hz. In interpreted as a special case of a Experiment 2, tones varying in intensity from 40 to 90 dB were also presented general theoretical position already in a paired comparison procedure, and again the predicted inverted-U-shaped shown to have wide support and preference function was found, with 50 dB as the preferred intensity. Related including the evidence that humans research and some unexpected findings are discussed. prefer moderate variation and moderate rates in tone sequences. The It was Wundt (1874, p. 468) who sequences of tones sounding like possibility of deriving affective first postulated that affect, or hedonic simple music that presented moderate response to individual pure tones from state, was related to stimulus intensity amounts of variation. Stimulus the same general theory used to derive in the simple way shown in Fig. 1. variation was presented on the responses to more complex tone This function, roughly shaped like a dimensions of tone frequency, sequences would certainly broaden the skewed and inverted U, has received intensity, and duration. Boesch, significance of the results for a theory support primarily from studies on the Baltes. and Schmidt (1968) presented of aesthetics. taste dimensions of sourness, saltiness, a 1000-Hz tone with an intensity of bitterness, and sweetness. Engle 55-60 dB at rates ranging from .3 to EXPERIMENT 1 (1928) had human Ss rate the 10/sec; they found that Ss had a TONE FREQUENCY pleasantness of solutions having one of preferred rate of tone stimulation. The Method these properties and found curves for many other studies providing evidence Su bj e c t s , Sixteen university sourness, saltiness, and bitterness very in support of the "moderate students, eight male and eight female, consistent with Wundt's hypothesis. stimulation" position are thoroughly served as Ss, All Ss reported that to Pfaffrnann (1960) also found similar integrated and summarized by Berlyne the best of their knowledge they had curves using rats as Ss. The dimension (1967). normal hearing, but no tests were of sweetness was something of an The purpose of the present study is made for possible unknown hearing exception in both studies since neither to test Wundt's hypothesis on the abnormalities. investigator found good evidence for a dimensions of sound frequency (hertz) Stimuli. Eight tones were selected decline in preference for sweetness at high levels of concentration.. Wundt's hypothesis is restricted to simple stimulus intensity; however, if one moves to the general but less t precise dimensions of stimulus .. CD complexity, variation, novelty, etc., c l: then there are many theorists who o I/) have predicted that humans and o animals prefer a moderate amount of CD stimulation: Berlyne (1960, 1967), a:: Dember and Earl (1957), Fiske and Maddi (1961), and especially in the case of lower animals, Schneirla (1959, 1965). This general theoretical Indifference 1---1'------3lt------position represents a major t development in the combined areas of Absolute threshold motivation, reinforcement, and attention. Evidence supporting this position of direct relevance to the present study is reported by several investigators. Vitz (1966) found that human Ss had a definite preference for

*This work was supported bY the New York University Arts and Sciences Fund and bv a New York University-NSF institutional grant. The assistance of Deborah Miller in data collection and analysis was invaluable. tAddress: Department of Psychology, New York University. 707 Broadway, New Fig. 1. The theoretical relation between stimulus intensity and affect as York. New York 10003. postulated by Wundt (1874).

84 Copyright 1972, Psychonomic Society, Inc., Austin, Texas Perception & Psychophysics, 1972, Vol. 11 (lB) Table 1 Results Percent Preference for Each Tone Frequency Averaged Across All Intensities The major results are shown in Tone Frequency (Hz) Table 1 and Fig. 2. From Fig. 2 it is clear that the average preferred S 60 110 210 400 750 1410 2660 5000 frequency is around 750 Hz and that 1 71.4 71.4 50.0 66.7 47.6 42.9 33.3 16.7 preference declines regularly on both 2 11.9 45.2 83.3 92.9 88.1 47.6 23.8 7.1 sides of the peak point. 3 19.0 45.2 73.8 85.7 78.6 40.4 40.4 16.7 Table 1 presents the proportion of 4 0.0 16.7 40.4 90.4 97.6 64.3 54.8 35.7 times each frequency was selected for 5 52.4 66.7 66.7 80.9 57.1 38.1 28.6 9.5 each S. These results show that the 6 4.8 33.3 64.3 85.7 80.9 71.4 45.2 14.3 great majority of Ss have preference 7 26.2 66.7 83.3 88.1 66.7 30.9 33.3 4.8 curves very similar to the average 8 0.0 19.0 69.0 90.4 85.7 64.3 45.2 26.2 9 73.8 38.1 42.9 35.7 71.4 54.8 50.0 33.3 curve. One S, No.9, shows two 10 52.4 73.8 80.9 78.6 66.7 26.2 19.0 2.4 "peak" points. 11 16.7 42.9 54.8 88.1 85.7 54.8 42.9 14.3 Further evidence supporting the 12 2.4 19.0 52.4 85.7 90.4 73.8 47.6 28.6 hypothesis is presented in Fig. 3 by 13 2.4 11.9 42.9 76.1 85.7 66.7 64.3 50.0 breaking the results into three sets of 14 4.8 9.5 28.6 47.6 85.7 73.8 76.1 73.8 preferences, one each at 65, 75, and 15 9.5 14.3 54.8 85.7 88.1 59.5 52.4 35.7 85 dB. Again, the Ss show curves with 16 7.1 26.2 40.4 69.0 85.7 52.4 66.7 52.4 peak preferences in the middle range. Mean 22.2 37.5 58.0 77.9 78.8 53.9 45.2 26.3 Looking at the results of individual Ss, Percent we found that at 65 dB 12 Ss had individual curves regularly rising to a peak and then declining regularly; at for the paired comparison task: 60. At the end of the experiment, the S 75 dB 7 Ss had such regular curves and 110, 210, 400, 750, 1410, 2660, and filled out a short questionnaire asking at 85 dB, 10 Ss. Obviously, reducing 5000 Hz. These tones were selected to (1) How would you rate your interest the data by two-thirds means that cover the range of 60-5000 Hz, the in music? (a) indifferent to music; chance effects could easily have caused most important range for human (b) little interest; (c) moderate the small reversals in the predicted hearing. The intervals between tones interest; (d) strong interest; (e) very curves. Evidence for this interpretation were within a few cycles of equal log strong interest. (2) How much time do is that of the 19 curves which units. The tones were pure tones you spend listening attentively to contained a reversal 12 would have generated on two Hewlett-Packard music? (a) none at all; (b) a little; (c) a matched the predicted curve without a audio oscillators, Model 200AB. moderate amount; (d) a great deal of reversal if a single choice had been Intensities were controlled by two time. (3) How much formal music changed. Thus, the predicted curve Hewlett-Packard attenuators, training have you had other than in characterizes both the average curve Model 350D, and were measured and required, general music courses? and the individual S curves at each of equated within ± 1 dB. Intensities were (a) none at all; (b) 1 year or less; (c) 2 the three different intensities. measured and standardized before to 3 years; (d) more than 3 years, An unexpected finding can be seen each experimental session. A General please list the instrument(s) in which in Fig. 3, which shows that as the Radio Company precision sound-level you had this formal training (e.g., intensity of the paired tones increased meter (Type 1561-A) was used in voice, piano, etc.). the preferred frequency also increased. conjunction with a General Radio Company precision microphone C 100 (Type 1560-P7) for measuring and IJJ equating intensities at the earphone. 0:: Headsets were U.S. Government 0:: general issue. IJJ 80 LL Each of the eight tones was IJJ presented at one of three inten­ 0:: sities: 65, 75, and 85 dB (re a. .0002 dynes/em- ). GO Procedure. Each S was presented .... with 168 pairs of tones and asked to Z choose the tone he found most IJJ pleasant; a tone lasted 1.0 sec. These 0 40 0:: 168 comparisons consisted of three IJJ groups of 56 tones. Each group of 56 a. tones consisted of all the possible paired comparisons among eight tones, Z 20 60 all at the same intensity. Each pair of

Perception & Psychophysics, 1972, Vol. 11 (18) 85 0 100 l&l order a-b and b-a. The four groups of declined thereafter. Thus, they failed II: II: 30 comparisons, each at a fixed to find that low tones were less l&l 80 u, frequency were intermixed so as to preferred than moderate frequencies, l&l II: distribute each frequency in about e.g., 400-750 Hz. [These data were /1. 60 .... equal proportions over the 120 trials. used by Guilford (1954) to derive I-:z theoretical affective curves as a l&l ~,. After 60 trials, there was a 5· to (.J 40 II: V'. 10-min break. At the end of the function of tone frequency. ] There l&l \ "b /1. experiment, the S filled out the same are, however, serious reasons to question the scaling procedure of ~ 20 questionnaire as used in Experiment 1. l&l Singer and Young. It is well known :E 0 1 Results that ratings, in this case, rating tones '00 ,000 10,000 The main findings are presented in on a scale from 0 to 110, can often be TONE FREQUENCY (Log Scale) Fig. 4 and Table 2. The results closely misleading. In addition, the ratings parallel those of Experiment 1. The were reduced to three gross categories: average preference for intensity pleasant, indifferent, and unpleasant. Fig. 3. Preference for tones as a "peaks" at 50 dB and declines on The proportion of ratings in these function of frequency at three either side. The curves at each three categories across all Ss were different levels of intensity (decibels). frequency are shown in Fig. 5. The combined to create the published curve at 400 Hz is highest at 40 dB scale. For these reasons, it seems The mean preferred frequencies were and declines regularly out to 90 dB. probable that the failure to detect a 454 at 65 dB, 616 at 75 dB, and 731 The curves for the other three peak preference around 400 to 750 Hz at 85 dB. The difference of 277 frequencies all peak at some is due to measurement weaknesses. between the preference at 65 and intermediate value. The results in Evidence to substantiate these 85 dB is significant (t =1.78, P < .05, Table 2 show that the preferences of criticisms comes from a study one-tailed, df = 15). This finding is nine Ss peaked at some intermediate discovered after the present study was considerably more significant if an intensity and then declined on either completed by Laird and Coye (1929). analysis is done of the individual side. Six Ss preferred the lowest They presented freauencies rangina choices. When the pair of tones was intensity of 40 dB and had regularly from 64 to 8192 Hz in a 65 dB, the higher frequency was declining preferences out to 90 dB. paired-comparison procedure and selected 45% of the time, at 75 dB, One S, No. 14, again showed two asked ISs which one of the two was 48% of the time, and at 85 dB the "peaks." more annoying. Their results agree higher frequency was selected 54% of The results were divided into four closely with those presented here. The the time. The difference between the sets, one for the preferences at each least annoying tones were 256, 512, percent at 65 and 85 dB is quite frequency, and the same pattern of and 1024 Hz, while 64 and 128 Hz significant (t = 2.75, P < .02, individual preferences was found as in were very significantly more annoying, two-tailed, df = 15). Experiment 1 when the results were and were approximately equal in On the basis of their answers to the divided on the basis of stimulus annoyance value to 2048 and 4098, three questions, the Ss were divided intensity. That is, evidence for respectively; 8192 Hz was the most into a high and a low music group. The individual Ss having a preferred annoying or all. Thus, the case tor a high music group consisted of the intermediate intensity was still preferred level of tone frequency eight Ss who scored highest on the present, although the curves were less seems clear. three questions. No differences were consistent than when averaged across Another issue bearing on the results found between the high and low music all four frequencies. of Experiment 1 is the question of Ss or between the first and last parts In Fig. 5 there is more evidence that how the results of this experiment of the experiment; nor were any sex at higher frequencies Ss preferred might have been affected by differences found. more intense tones. The curves for the differences in loudness. From the different frequencies separate in a Fletcher-Munson (1933) EXPERIMENT 2 regular manner as frequency increases. equal-loudness curves, it is apparent TONE INTENSITY The mean preferred intensity was that the differences between loudness Method 49.1 dB at 100 Hz and 59.1 dB at and intensity are not especially great Subjects. Sixteen new Ss with the 5000 Hz. This difference was quite for 65 and 75 dB, and are negligible same characteristics as those in significant (t = 2.48, P < .02, Experiment 1 were used. two-tailed, df = 15. Again, no Stimuli. Six different intensities: significant differences were found 100 C 40, 50, 60, 70, 80, and 90 dB were between high and low music Ss, l&l It: selected. The same stimulus-generating between first and last trials, or It: l&l 80 equipment was used as in between the sexes. u, l&l Experiment I, except that new and It: /1. 60 more comfortable headsets were used. DISCUSSION I- :z Each intensity was presented at one of The general results constitute strong w (.J 40 the following frequencies: 100, 400, support for the initial hypotheses. It: W 1400, and 5000 Hz. However, there are several issues which /1. Procedure. Each S was presented require comment. An experiment by Z 20

86 Perception & Psychophysics, 1972, Vol. 11 (lB) Table 2 shifts the preference on the other Percent Preference for Each Tone Intensity Averaged Across All Frequencies dimension in the same direction. This Intensity (dB) result is difficult to interpret within the present theoretical framework. S 40 50 60 70 80 90 One possibly relevant observation is 1 35.0 55.0 55.0 67.5 55.0 32.5 that when these results were 2 37.5 62.5 62.5 72.5 40.0 25.0 mentioned to several musicians they 3 55.0 70.0 67.5 52.5 45.0 10.0 were not surprised, since positively 4 70.0 80.0 80.0 47.5 22.5 0.0 correlated changes in music are 5 55.0 85.0 80.0 52.5 75.0 2.5 common, e.g., moving to a climax by 6 75.0 75.0 75.0 50.0 21.0 0.0 increasing the loudness, pitch, and 7 50.0 55.0 55.0 75.0 42.5 22.5 8 92.5 72.5 62.5 40.0 27.5 5.0 tempo. 9 80.0 70.0 67.5 50.0 32.5 0.0 An important issue raised by these 10 27.5 67.5 67.5 75.0 45.0 17.5 results is the question of the 11 52.5 62.5 57.5 62.5 42.5 22.5 physiological basis of such sensory 12 95.0 77.5 60.0 45.0 20.0 2.5 pleasure. Although this issue has been 13 82.5 70.0 55.0 52.5 30.0 10.0 largely ignored, there is a study by 14 55.0 52.5 47.5 40.0 42.5 62.5 Berlyne, McDonnell, Nicki, and 15 65.0 65.0 72.5 50.0 37.5 10.0 16 47.5 5.0 Parham (1967) that provides an 80.0 75.0 70.0 22.5 interesting clue. Berlyne et al Mean 62.9 68.4 64.7 55.0 34.7 14.2 presented 1-sec tones of 200,400, Percent 800, and 1600 Hz to 8s and recorded the duration of their EEG desynehronization, They found a for 85 dB. Therefore, it is likely that the present experiment, an AL might U -s h aped function in which the affect of loudness is not apt to develop, as the experiment progressed, desynchronization was least at 400 change the basic shape of the curve. around the mean of the comparison and 800 Hz and areatest at 200 and Moreover, the middle values are those stimuli and this could lead to double 1600 Hz. These results are similar to which are subjectively louder, and this peak curves. those of Experiment 1 and raise the presumably would have tended to The finding that the preferred simple possibility that stimuli that work against the hypothesis. frequency went up when tone cause desynchronous EEG are less The results of Experiment 2, a intensity increased (Experiment 1) and pleasant than those that are associated preferred intensity of approximately that the preferred intensity went up with synchronous EEG. 8uch a simple 50 dB (ranging from 40 to 60 dB, when frequency increased relationship between judged depending on frequency), are (Experiment 2) raises some problems. preference or hedonic state and the somewhat lower than might be Although no prediction was made, the EEG response is probably too simple expected. Considerable previous work author expected that the opposite in view of the known complexity of has been done on most comfortable would be the case. The reasoning for both variables-and yet it has enough listening levels (Pollack, 1952) and this apparently false prediction is: if, plausibility to be given attention. annoyance levels (Spieth, 1956). These for example, a 8's preferred tone at studies typically have the 8 listen to a 75 dB is 700 Hz, then this can be REFERENCES single tone and adjust it, using the BERLYNE. D. E. Conflict, arousal and interpreted as the 8's optimal level of curiosity. New York: McGraw-Hill. 1960. method of adjustment, to the most stimulation on both dimensions, i.e., BERLYNE, D. E. Arousal and comfortable level or to a level which is total optimal stimulation 0 If the reinforcement. In D. Levine (Ed.), almost annoying. In general, most intensity is increased to 85 dB, then Nebraska symposium on motivation 1967. Lincoln: University of Nebraska comfortable listening levels and almost presumably the 8 would want to Press. 1967. annoying tones are in the range from reduce the frequency, say to 600 Hz, BERLYNE. D. Eo, McDONNELL, P., 60 to 95 dB. The methodological so as to remain close to the NICKI. R. Mo, & PARHAM. L. C. C. differences between these studies and hypothesized total optimal Effects of auditory pitch and complexity on EEG desynchronization and on the present are large and make the s tim u lation. Instead of this verbally expressed judgments. Canadian results difficult to compare. For compensatory shift, the opposite Journal of Psvchologv, 1967, 21. example, considerable adaptation is appears to be true-namely, that an 346-367. likely in using the method of BOESCH, E. E., BALTES, P. B., & increase or decrease on one dimension SCHMIDT, L, R. Preference for different adjustment, especially in comparison auditory stimulus sequences in various to the present procedure in which each age groups. Psvchonomie Science, 1968. tone lasted 1 sec. The questions asked 100 10, 205-206. 0 DEMBER, w. N., & EARL, R. W, Analysis the 8 are different in important ways, Wa: a: of exploratory, manipulatory and and further, these experiments do not w so curiosity behaviors. Psychological ask for comparisons between two lLw Review, 1957, 64, 329-339. a: ENGLE, R. Experimentelle Untersuchungen frequencies and hence the important Q. 60 I- tiber die Abhangigkeit der Lust und rei a t ive preferences are not Z Unlust von der Reizstarke beim W ~",. determined. o 40 ',,'~~>.. Geschmacksinn. Archiv fUr die gesamte a: Psychologie, 1928, 64, 1-36. Cited by Two Ss, 89 in Table 1 and 814 in ~ ~ '.~~~.~.. Table 2, showed two peaks in their , ... ~ Woodworth, R. S., & Schlosberg, H. Z 20 ','" Experimental Psychology. New York: preference curves. This may be due to wct ." Holt, 1954• :l! an artifact such as an unknown hearing OJ FISKE, D. W., & MADDI, S. R. Functions weakness or gap. Nevertheless, such 40 50 6P 70 so 90 of varied experience. Homewood, m: Dorsey, 1961. curves are very like the double peak or TONE INTENSITY (dB) FLETCHER, H., & MUNSON, W. A. "butterfly" curves reported by Haber Loudness, its definition, measurement (1958), who found such curves of and calculation. Journal of the Acoustical Fig. 5. Preference for tones as a Society of America, 1933, 5, 82:108. preference for discrepancies from a function of tone intensity at four dif­ GUILFORD, J. P. System in the water-temperature adaptation level. In ferent levels of frequency (hertz). relationship of affective value to

Perception & Psychophysics, 1972, Vol. 11 (lB) 87 frequency and intensity of auditory ~~5~e2l~~~s_~~~. affective reaction: 110 Dependence of stimuli. American Journal of Psychology, Society of America. affective ratings upon the 1954. 67.691-695. SCHNEIRLA, T. C. An evolutionary and stimulus-situation. Journal of General HABER, R. N. Discrepancy from adaptation developmental theory of biphasic Psychology, 1941, 24, 303-325. level as a source of affect. Journal of processes underlying approach and SPIETH. W. Annoyance threshold Experimental Psychology. 1958, 56, withdrawal. In M. R. Tones (Ed<>. judgements of bands of noise. Journal of 370-375. Nebraska symposium on motivation the Acoustical Society of America. 1956, LAIRD. D. A •• & COYE, K. Psychological 1959. Lincoln: University of Nebraska 28,872-877. measurements of annoyance as related to Press, 1959. STEVENS. S. S. On the psychophysical law. pitch and loudness. Journal of the SCHNEIRLA, T, C. Aspects of stimulation Psychological Review, 1957, 64. 153·181. Acoustical Society of America, 1929, 1, and organization in approach withdrawal VITZ, P. C. Affect as a function of stimulus 158-163. processes underlying vertibrate behavioral variation. Journal of Experimental PFAFFMANN, C. The pleasures of development. In D. L. Lehrman. R. Psychology. 1966, 71, 74-79. sensation. Psychological Review, 1960, Hinde, and E. Shaw (Eds.), Advances in WUNDT, W. Gruruiz uge der phy siologischen 67.253·268. the study of behavior. New York: Psvchotogie, Leipzig: Engelmann. 1874. POLLACK, I. Comfortable listening levels Academic Press, 1965. for pure tones in quiet and noise. Journal SINGER, W. B., & YOUNG, P. T. Studies in (Accepted for publication June 10, 1971.)

88 Perception &; Psychophysics, 1972, Vol. 11 (lB)