This study determined how well the “perceived spectrum,” estimated using a pitch similarity rating method, reflected the spectrum and pitch of seven different tonal sounds. The perceived spectrum well-matched the acoustic spectrum for pure tones ranging from 1 to 12 kHz, it also matched the broad frequency range for two complex tones with periodicity pitches of 1 and 2 kHz, but it did not reflect the pitch of the complex tones. These results suggest that while this method may not measure the pitch of sounds, it may be useful for measuring the general perceived frequency range of sounds.

A variety of psychophysical methods exist which can be used to assess auditory perception, with one subset being rating methods that measure perceptual attributes of sounds (e.g., pitch, loudness, or similarity to other sounds). Unlike classical psychoacoustical procedures that generally measure the ability to detect or discriminate between sounds, rating methods can be used to measure the perception of phantom sounds [e.g., tinnitus; Noreña et al. (2002) and Roberts et al. (2008)] or imagined sounds [e.g., Halpern et al. (2004) and Gygi et al. (2007)]. One example is the novel extension of Noreña et al. of rating methods, which measured the pitch of tinnitus by asking subjects to rate the pitch similarity of various pure tones to their tinnitus [see also Roberts et al. (2008) who expanded on this method]. We suggest that with further evaluation of this method, its use could be more widespread to include applications to the measurement of the perceived pitches within acoustic sounds. With that idea in mind, this study aimed to establish whether a pitch similarity rating approach provided an accurate estimate of the spectral composition or the pitches of objective stimuli. This study focused on pure and complex tonal stimuli. Such stimuli provided a strong basis to evaluate this method because they have known pitches and pitch strengths.

In the most widely used application of this method, the “Tinnitus tester” of Roberts et al., listeners used a slider to rate the similarity of the pitch of various comparison sounds to their tinnitus using the terms “identical, very similar, somewhat similar, not very similar, and not at all similar.” The comparison sounds were pure tones or narrow bands of noise with different center frequencies. The resulting function relating the pitch similarity rating to the frequency of the comparison sound was considered the “tinnitus spectrum.” This tinnitus spectrum presumably illustrated the pitches present in tinnitus: The pitches within a person's tinnitus corresponded to frequencies that received high ratings. This procedure was innovative, was relatively quick to administer, and had robust test-retest reliability, particularly in comparison to the tinnitus pitch matching procedure adopted in clinical tests [cf. Penner (1983) and Burns (1984)].

Yet, because this procedure has never been applied to objective stimuli, its limitations or whether it can be used to estimate the spectral content or pitch of objective stimuli are unknown. If we can establish how well this method represents the characteristics of sounds, we can determine its potential for more general application. As such, we must describe some of the assumptions implicit in this method that were evaluated here. Two primary assumptions of the pitch-similarity rating method were that (a) the tinnitus being measured had a pitch and (b) the resulting tinnitus spectrum reflected the pitches within the tinnitus percept. We evaluated the roles of pitch and pitch strength by applying the pitch rating method to objective sounds with different frequencies: High-frequency pure tones with weak pitches (Ward, 1954; Attneave and Olson, 1971), low-frequency tones with strong pitches (Fastl and Stoll, 1979), and complex tones with periodicity pitches. We were able to establish whether the perceived spectrum, estimated using this method, reflected the frequencies of stimuli and/or the periodicity pitch of complex sounds.

We propose that measuring the “perceived spectrum” of a sound could give a rapid assessment of the pitches that a listener perceives within a sound and therefore could provide valuable information about which aspects of a complex sound are best represented by a particular listener's ear. As a result, this method could have more far-reaching applications, including hearing aid fittings and cochlear implant mappings. Thus, in this study we hope to provide an evaluation of the strengths and weaknesses of this method to inform its ultimate application.

Twelve subjects with normal hearing (three male), aged from 21 to 30 years, participated. All subjects had bilateral air conduction audiometric hearing thresholds of 20 dB hearing level (HL) or better at standard audiometric frequencies and no history of tinnitus or otologic disease.

Test stimuli were seven different sounds: five pure tones (1, 3, 5, 8, and 12 kHz) and two complex tones presented at an overall level of 65 dB sound pressure level (SPL) re: 20 μPa. One complex tone (CS2tone) had two frequencies: 1 and 6 kHz at equal amplitude. The other (CS5tone) had five frequencies 4, 6, 8, 10, and 12 kHz, with a small spectral prominence at 8 kHz. Subjects L1–L6 were tested on the pure-tone test stimuli using 11 comparison stimuli, which had frequencies from 0.5 to 12 kHz. Subjects L7-L12 joined the study later and were tested on the pure-tone test stimuli using 13 comparison stimuli, which had frequencies from 0.5 to 16 kHz, and the complex-tone test stimuli. All stimuli were pulsed “on” for 200 ms with 20-ms rise/fall times and 400 ms “off.” When both test and comparison stimuli were presented together, they were presented in an alternating sequence with 100 ms between test and comparison stimuli.

Subjects began by selecting and listening to a pulsed test stimulus. After this familiarization, subjects listened to three comparison stimuli (alternating with the test stimulus) centered at 2 kHz: a pure tone, a band of noise with a 10% bandwidth, and a band of noise with a 30% bandwidth. Subjects selected the comparison stimulus that matched the quality of the test stimulus. The stimulus quality selected was used for all subsequent comparison stimuli in the next phases of the experiment. Notably, all subjects selected the tone in this phase of the experiment, and so all comparison stimuli were pure tones.

Next, in the level matching phase, the frequencies of the comparison stimuli were selected at random, and the subjects listened to each of the comparison stimuli alternating with the test stimulus. Using a slider, subjects adjusted the levels of the comparison stimuli to match the loudness of the test stimulus. Three replicate loudness matches were obtained for each of the comparison stimuli. The average sound level across the three replicates was used as the presentation level for the perceived spectrum measurements.

Last, in the pitch similarity rating phase, subjects used a slider to indicate the similarity of the pitch of each comparison sound to the test sound. The frequencies of the comparison stimuli were chosen at random, and, when stimuli were presented, alternated with the test stimuli. Subjects were told: “Using the slider, rate the similarity of the pitch of each of the sounds to the pitch of the test sounds.” The following terms were adjacent to and equally spaced along the slider: “identical, very similar, somewhat similar, not very similar, and not at all similar.” Ratings were mapped to the Borg CR100 (Borg and Borg, 2001) scale using the full 100 points with terms separated by 25 points (e.g., identical = 100 and not at all similar = 0). Three replicate ratings were obtained for each comparison stimulus and were averaged together to produce a final similarity rating.

Pilot testing indicated that a practice session facilitated reliable data and ensured that subjects understood the experimental procedures. As a result, prior to experimental testing all subjects were given at least one practice session with a 0.5-kHz test stimulus, selected because of its clear and strong pitch. During the practice, subjects could ask questions, and experimenters reviewed the data to verify that subjects understood the task. To proceed to the experimental phase, subjects must have given the 0.5-kHz comparison tone the highest pitch similarity rating and have used at least 80 points of the rating scale. When subjects did not meet both criteria, they were reinstructed and completed the practice again. Data from two additional subjects were not included because one experimenter failed to give practice to one subject and a second subject was unable to meet these criteria.

Stimuli were generated digitally and played through a 24-bit Digital-Analog Converter [DAC; Tucker Davis Technologies (TDT) RP2.1] at a sampling rate of 48 828 Hz. The output of the DAC was fed into an attenuator (TDT PA5), headphone buffer (TDT HB6), and played through a Sennheiser HD280 Pro headset. Both test and comparison stimuli were presented to the subject's right ear. All test procedures were approved by the Indiana University Human Subjects Office and the Institutional Review Board.

Due to our primary interest in the pitch similarity rating data, the loudness matching data are reported here in brief. Generally speaking, loudness matching data followed equal-loudness contours. The range of dB-matched levels varied between 20 dB SPL (for the low-frequency comparison tones when high-frequency test tones were tested) and 100 dB SPL (the maximum level tested), which typically occurred for the 16-kHz comparison stimulus. For test frequencies of 1, 3, 5, 8, and 12 kHz, the average same-frequency level match was 64.6, 65.0, 64.8, 62.8, and 61.5 dB SPL, respectively. Data across subjects did vary substantially, particularly for the 8- and 12-kHz test frequencies and the two complex tones. For these conditions, standard deviations across subjects ranged between 5 and 18 dB and were generally around 10 dB.

The perceived spectrum was generated by plotting the pitch similarity rating versus the frequency of the comparison stimulus. Perceived spectra from all 12 subjects for the pure-tone stimuli are shown in Fig. 1, with data plotted in ascending order based on the test frequency (i.e., 1 kHz at the top and 12 kHz at the bottom).

Fig. 1.

Perceived spectra for the pure tone test stimuli. Each panel represents a different test frequency, and data from each subject are illustrated with different symbols and line properties.

Fig. 1.

Perceived spectra for the pure tone test stimuli. Each panel represents a different test frequency, and data from each subject are illustrated with different symbols and line properties.

Close modal

Results for all of the test frequencies were very similar, evident in a robust peak at the frequency of the test tone. For the 1, 3, 5, and 8-kHz test frequencies, all subjects gave their highest rating to the comparison frequency that matched the frequency of the test stimulus (the matched frequency). Further, most subjects gave a rating of 100 points to the matched frequency, and each subject used a rating of over 90 points. For the 12-kHz test tone, one subject (L3) gave the 6-kHz tone the highest rating, and two subjects (L8 and L9) gave the highest rating to 12 and 14 kHz. Subject L3 did not use the full range of available ratings (max rating = 72 points). With these small exceptions at 12 kHz, all perceived spectra robustly converged on the acoustic spectra of the pure tones.

Perceived spectra from six subjects for the complex tones are shown in Fig. 2. For reference, the frequencies present in the acoustic spectra are shown as the asterisks above the rating data. Data from the CS2tone (top panel; frequencies = 1 and 6 kHz) illustrated a few notable trends. All six subjects provided the highest rating at 6 kHz (L7, L8, L9, L11) or a rating within two points of the highest rating (L10 and L12). Across all subjects, the perceived spectra reflected a perceptual prominence ranging between 6 and 12 kHz, with many of the perceived spectra showing two prominences: one at 6 kHz and one at 10 kHz. No subjects gave a high rating to the 1-kHz comparison tone, despite 1 kHz being a component in this stimulus. Five of the subjects gave a rating of less than 22 points (i.e., “not very similar”). It also noteworthy that the overall pitch similarity ratings were somewhat lower for this stimulus than for the pure tones, as only two subjects (L8 and L9) provided a rating higher than 90 points.

Fig. 2.

Perceived spectra for the complex tone test stimuli. Data from each subject are illustrated with different symbols and line properties. Asterisks indicate the frequencies present in the acoustic stimuli.

Fig. 2.

Perceived spectra for the complex tone test stimuli. Data from each subject are illustrated with different symbols and line properties. Asterisks indicate the frequencies present in the acoustic stimuli.

Close modal

Perceived spectra of the CS5tone (bottom panel; frequencies = 4, 6, 8, 10, and 12 kHz) illustrated considerable variability across subjects, with only some subjects generating perceived spectra similar to the acoustic spectrum. Most perceived spectra illustrated a broad peak ranging from 4 to 8 or 10 kHz, and a fair number of low and high ratings. Many but not all of the high ratings occurred at the stimulus frequencies, with most of the low ratings occurring for the 10- and 12-kHz tones. Roughly speaking then, the perceived spectra broadly represented the range of frequencies present in the CS5tone. A second feature of the perceived spectra was that only two of the six subjects provided a high rating to 2 kHz, which would be the expected (albeit weak) periodicity pitch for this stimulus (Oxenham et al., 2011). We note, however, that the perceived spectra of CStone2 and CStone5 were very similar in their appearance, despite having rather different frequency composition. On the other hand, both perceived spectra contained much broader frequencies ranges from those obtained for the pure tones.

This method revealed very strong convergence between perceived and acoustic spectra of pure tones. In all cases except one (L3 for the 12-kHz test tone), subjects provided the highest rating to the frequency that corresponded to the test stimulus. Yet, perceived spectra of the complex tones did not robustly match with the acoustic spectra, but this method did generally converge on the frequencies present in the complex acoustic spectra.

We are encouraged that this method robustly estimated the frequency of pure tones. Although this approach differs substantially from frequency-discrimination experiments, those studies strongly support the prediction that this method should robustly converge on the perception of pure tones. Relative to frequency difference limens [generally between 1% and 5% of the center frequency; Wier et al. (1977), Sek and Moore (1994), Dai and Micheyl (2011), and Moore and Ernst (2012)], the comparison frequencies used here were very widely separated. As a result, we expected that the perceived spectra should show a prominent peak at the frequency of the test stimulus, and there was clear evidence in support of this hypothesis. Whether this method measured the pitch of these sounds, however, remains unclear. Varying the frequency of a pure tone changes its pitch, pitch strength, and timbre, and results were robust across all frequencies tested. We had originally hypothesized that this method might yield more variable pitch similarity ratings across subjects or poorer correspondence between acoustic and perceptual spectral for the high-frequency pure tones which had weak pitches, but the data did not support this prediction. One likely possibility is that the procedure used here was not sensitive to pitch strength. A second possibility is that this method estimated quality or the general perceptual characteristics of the sounds and not the pitch, per se.

Regarding the complex tones, the method broadly characterized the acoustic spectra but missed in the fine details. Although the acoustic spectrum of the two-tone complex sound (CS2tone) included components at 1 and 6 kHz, the perceived spectrum contained a broad spectral peak ranging from 6 to about 12 kHz and no peak at 1 kHz. Thus, this method did not robustly estimate the acoustic spectrum of this stimulus, nor did it estimate the pitch [which would have been 1 kHz; Oxenham et al. (2011)]. The low pitch similarity ratings for the 1-kHz component were surprising, as we expected that listeners could “hear out” each of these perceptually distant tones (>2 octaves apart) tone from within the complex (Hartmann et al., 1990; Moore and Ogushi, 1993). Interestingly, for 5 of the 6 subjects, the perceptual spectrum of CS2tone was broad in frequency and did not have a single spectral peak like those of the pure tones, suggesting that this stimulus did not sound like a pure tone to the six subjects.

Results from the CS5tone were more similar to the acoustic spectrum than was observed for the CS2tone. Many of the subjects provided high ratings to the tones present in the stimulus (4, 6, and 8 kHz in particular), although there were substantial individual differences in the perceptual spectra. Broadly speaking, the method converged on the general frequency range of this stimulus, suggesting that the methodology did provide a global estimate of the frequencies present in the sound. On the other hand, it is interesting that only two subjects gave a high rating to 2 kHz, which would be the expected periodicity pitch of this stimulus (Oxenham et al., 2011). Yet, as with CS2tone, the periodicity pitch would also be relatively weak.

Thus, for the complex tone stimuli, it appeared that this method indeed did not measure the pitch. Rather, the perceived spectra reflected the general frequencies present in the stimuli. Therefore, perhaps the method characterized a global perception of the frequencies within the sound, rather than the pitch per se. This interpretation is bolstered by the observation that the frequency ranges of the perceived spectra for CS2tone and CS5tone were very similar.

Given the evidence presented here, we must consider that the pitch similarity ratings were based on general perceptual similarity rather than the pitch, a possibility noted by Noreña et al. (2002) in their measurements of tinnitus. The evidence here is as follows: First, with regards to the pure tones, a large proportion of the ratings to unmatched frequencies (i.e., comparison frequencies different from the test frequency) were on the order of “somewhat similar (rating = 50)” despite very large frequency differences between test and comparison stimuli. Second, octave frequencies of the test frequency were not associated with high ratings, which would have been expected if this method measured the pitch, further bolstering the hypothesis that these ratings were not based on the pitch. Finally, the perceived spectra for the complex sounds did not reflect a component at the periodicity pitch, which would have been 1 kHz for CS2tone and 2 kHz for CS5tone. Indeed, more work testing complex sounds in which pitch and pitch strength is manipulated (e.g., harmonic sounds with low fundamental frequencies or rippled noises in which pitch strength can be varied systematically) is needed. However, we tentatively conclude that this method did not measure the pitch perception of complex sounds but that rather the method provided a general characterization of the frequencies present within the test sounds. Such as conclusion will remain tentative, however, until more extensive work on this method is completed.

Despite this observation, we remain optimistic regarding in the application of this method to measuring the characteristics of tinnitus. Generally, the ultimate goals of applying this method to tinnitus are to establish which neural populations underlie the tinnitus percept or to establish parameters of stimuli to be used in tinnitus treatments (Roberts et al., 2008). We argue that it may be of little consequence that subjects were not rating pitch similarity. First, the method generally resulted in a perceptual spectrum that reflected the frequencies present in sounds. Second, perceived spectra measured for the complex tones were much broader than those for the pure tones. To a first approximation, this method differentiated between a pure tone and a complex one and reflected the general frequencies within all stimuli. As such, this method might be able to provide a description of the frequencies of the neural components involved in tinnitus perception, or at least the frequency range.

Perceptual spectra obtained using the pitch similarity rating method had excellent correspondence to the acoustic spectra of pure tones. In contrast, the relationship between the perceptual and acoustic spectra of complex tones was not as robust. Although subjects were probably not rating the pitch similarity of comparison sounds to the test stimuli, the method globally converged on the frequencies present in acoustic stimuli. More work testing sounds with very clear pitches is needed to establish whether a pitch similarity rating method is a viable approach to measuring the frequency content of acoustic sounds.

The authors thank Matthew Walker, Grace Haines-Gallagher, Nardine Taleb, and Raquel Mendoza for assistance during data collection. This work was funded by National Institutes of Health Grant No. R21 DC013171 awarded to J.J.L.

1.
Attneave
,
F.
, and
Olson
,
R. K.
(
1971
). “
Pitch as a medium: A new approach to psychophysical scaling
,”
Am. J. Psychol.
84
,
147
166
.
2.
Borg
,
G.
, and
Borg
,
E.
(
2001
). “
A new generation of scaling methods: Level anchored ratio scaling
,”
Psychologia
28
,
15
45
.
3.
Burns
,
E. M.
(
1984
). “
A comparison of variability among measurements of subjective tinnitus and objective stimuli
,”
Audiology
23
,
426
440
.
4.
Dai
,
H.
, and
Micheyl
,
C.
(
2011
). “
Psychometric functions for pure tone frequency discrimination
,”
J. Acoust. Soc. Am.
130
,
263
272
.
5.
Fastl
,
H.
, and
Stoll
,
G.
(
1979
). “
Scaling of pitch strength
,”
Hear. Res.
1
,
293
301
.
6.
Gygi
,
B.
,
Kidd
,
G. R.
, and
Watson
,
C. S.
(
2007
). “
Similarity and categorization of environmental sounds
,”
Percept. Psychophys.
69
(
6
),
839
855
.
7.
Halpern
,
A. R.
,
Zatorre
,
R. J.
,
Bouffard
,
M.
, and
Johnson
,
J. A.
(
2004
). “
Behavioral and neural correlates of perceived and imagined musical timbre
,”
Neuropsychologia
42
(
9
),
1281
1292
.
8.
Hartmann
,
W. M.
,
McAdams
,
S.
, and
Smith
,
B. K.
(
1990
). “
Hearing a mistuned harmonic in an otherwise periodic complex tone
,”
J. Acoust. Soc. Am.
88
(
4
),
1712
1724
.
9.
Moore
,
B. C.
, and
Ohgushi
,
K.
(
1993
). “
Audibility of partials in inharmonic complex tones
,”
J. Acoust. Soc. Am.
93
(
1
),
452
461
.
10.
Moore
,
B. C. J.
, and
Ernst
,
S. M. A.
(
2012
). “
Frequency difference limens at high frequencies: Evidence for a transition from a temporal to a place code
,”
J. Acoust. Soc. Am.
132
,
1542
1547
.
11.
Noreña
,
A.
,
Micheyl
,
C.
,
Chery-Croze
,
S.
, and
Collet
,
L.
(
2002
). “
Psychoacoustic characterization of the tinnitus spectrum: Implications for the underlying mechanisms of tinnitus
,”
Audiol Neurootol.
1
,
358
369
.
12.
Oxenham
,
A. J.
,
Micheyl
,
C.
,
Keebler
,
M. V.
,
Loper
,
A.
, and
Santurette
,
S.
(
2011
). “
Pitch perception beyond the traditional existence region of pitch
,”
Proc. Natl. Acad. Sci. U.S.A.
108
(
18
),
7629
7634
.
13.
Penner
,
M. J.
(
1983
). “
Variability in matches to subjective tinnitus
,”
J. Speech Hear. Res.
26
(
2
),
263
267
.
14.
Roberts
,
L. E.
,
Moffat
,
G.
,
Baumann
,
M.
,
Ward
,
L. M.
, and
Bosnyak
,
D. J.
(
2008
). “
Residual inhibition functions overlap tinnitus spectra and the region of auditory threshold shift
,”
J. Assoc. Res. Otolaryngol.
9
,
417
435
.
15.
Sek
,
A.
, and
Moore
,
B. C. J.
(
1994
). “
Frequency discrimination as a function of frequency, measured in several ways
,”
J. Acoust. Soc. Am.
97
,
2479
2486
.
16.
Ward
,
W. D.
(
1954
). “
Subjective musical pitch
,”
J. Acoust. Soc. Am.
26
,
369
380
.
17.
Wier
,
C. C.
,
Jesteadt
,
W.
, and
Green
,
D. M.
(
1977
). “
Frequency discrimination as a function of frequency and sensation level
,”
J. Acoust. Soc. Am.
61
,
178
184
.