Abstract
In two experiments, we assessed recovery of a tonal hierarchy in tone sequences. In Experiment 1, sequence tones were five tones of a Korean pentatonic scale plus seven nonscale tones located between scale tones. Sequences included all 12 tones, randomly ordered. Duration of scale tones in each sequence corresponded to the total duration of each tone in a piece of Korean music, as quantified by U. Nam (1998). Nonscale tones were shorter than scale tones. Listeners were either familiar or unfamiliar with the style of Korean music. Sequences were played 12 times, each time followed by 1 of 12 probe tones that had occurred in the sequence. Participants rated goodness-of-fit of the probe tone to the sequence. Ratings by both groups reflected the Korean tonal hierarchy including the relative salience of scale tones. Experiment 2 followed the same method and tones, but duration was assigned to tones quasi-randomly so that duration did not emphasize intervallic relationships in the Korean scale. Ratings differentiated long and short tones, but showed no other clear organization among long tones. Differences in results between experiments suggest that duration helps listeners organize pitch structure only when duration emphasizes intervallic relationships such as the near-perfect fifth.
The current study addresses perceived pitch structure arising from the durations and tuning of tones in traditional Korean court music. It deals specifically with tonal hierarchies – the abstract system of tonal relations by which tones are ordered according to relative salience and importance. We ask whether listeners can perceive a hierarchy among the set of pitches in Korean music and whether listeners must be familiar with the Korean musical style in order to do so.
The concept of tonal hierarchy applies to music of different cultures and styles. The tonal hierarchies of Western tonal-harmonic music have received the most extensive study. For the Western system, both music-theoretic (e.g., Jackendoff & Lerdahl, 2006; Lerdahl, 1988; Meyer, 1956), behavioral (e.g., Cuddy & Badertscher, 1987; Janata, Birk, Tillman, & Bharucha, 2003; Krumhansl, 1990; Krumhansl & Kessler, 1982; Krumhansl & Shepard, 1979; Marmel & Tillmann, 2009), and neuroscientific (Brattico, Tervaniemi, Naatanen, & Peretz, 2006; Krohn, Brattico, Valimaki, & Tervaniemi, 2007) approaches have identified a multi-layered hierarchy of tonal relations, with the tonic as the most prominent, important and stable tone, followed by the tones of the tonic triad, the rest of the scale tones, and finally the nonscale tones.
Numerous experiments verifying the psychological reality of the Western tonal hierarchy have employed the probe-tone technique (for reviews, see Krumhansl, 1990, 2000; Krumhansl & Cuddy, 2010). The probe-tone technique involves the presentation of a musical sequence followed by a probe tone, one of the 12 tones of the chromatic scale. The listener is asked to rate the goodness of fit of the probe tone to the sequence. The sequence is then repeated followed by a different probe tone. The set of ratings averaged across a variety of key-defining sequences is called the standardized probe-tone profile (Krumhansl & Kessler, 1982) and corresponds to the music-theoretic description of the Western tonal hierarchy.
Krumhansl’s (1987b) account of how listeners acquire a sense of tonality is based on two assumptions. First, listeners are sensitive to the statistical distributions of tones in music, including both the frequency of occurrence and relative duration of the tones. Second, with long-term exposure to the music, the distribution becomes internalized as a mental schema (Krumhansl, 1990).
Experimental studies support both assumptions. The first assumption is supported by probe-tone evidence that listeners are sensitive to pitch distributions that do not conform to the traditional Western tonal-harmonic system, whether novel (Cuddy, 1997; Lantz & Cuddy, 1998, 2006; Loui, Wessel, & Kam, 2010; Oram & Cuddy, 1995) or culturally unfamiliar (Castellano, Bharucha, & Krumhansl, 1984; Kessler, Hansen, & Shepard, 1984; Krumhansl, Louhivuori, Toiviainen, Järvinen, & Eerola, 1999; Krumhansl et al., 2000). Probe-tone profiles showed a hierarchic ordering of tones based on pitch distributional properties at the musical surface of the sequences.
The second assumption is supported by evidence that listeners apply internalized schema in perceiving pitch structure. Tone sequences that do not contain the full tonal hierarchy, such as C E C G (Cuddy & Badertscher, 1987) or three-note triadic chords (Krumhansl & Kessler, 1982), nevertheless result in probe-tone profiles that reveal the full four-tiered tonal hierarchy.
Cross-cultural studies allow examination of familiarity with tonal schemas (Castellano et al., 1984; Kessler et al., 1984; Krumhansl et al., 1999, 2000). For example, Castellano et al. (1984) found that probe-tone responses from both listeners familiar or unfamiliar with Indian music reflected the tonal hierarchy based on tone durations in the contexts. However, only responses from listeners familiar with the style also reflected structural properties of the Indian scale underlying the contexts.
Other studies are suggestive of assimilation in the presence of deviation from the Western tonal hierarchy. Jordan (1987) and Vurma and Ross (2006) further found that mistuned tones within a piece of music could vary by as much as 25 cents without the music seeming to be out of tune. Mistuned intervals up to 25 cents could be assimilated into a Western tonal schema. However, mistuned intervals greater than 25 cents are unlikely to be assimilated into the Western tonal hierarchy.
The distinction between sensitivity to distributional information and listeners’ internalization of a tonal hierarchy motivated Smith and Schmuckler (2004) to devise an experiment in which neither the pitch set nor the order of tones conveyed a tonal structure. Sequences contained all 12 chromatic tones presented in random order. Structure was present in the distribution of durations to tones, with each tone consistently paired with a particular duration.
Smith and Schmuckler (2004) chose 12 different durations in proportion to the 12 standardized probe-tone profile ratings from Krumhansl and Kessler (1982). In a hierarchical condition, assignment of duration to tones was in accordance with the tonal hierarchy – tones higher in the tonal hierarchy were assigned longer durations than tones lower in the hierarchy. In the nonhierarchical condition, assignment of durations to tones was random – any tone could be assigned any one of the possible durations. Thus, the durational information did not match the tonal hierarchy. Listeners recovered the tonal hierarchy only in the hierarchical condition. Probe-tone profiles in the nonhierarchical condition did not reflect tonal hierarchy or duration. In fact, nonhierarchical probe-tone profiles were relatively flat, with mean ratings of all probe tones around the midpoint of the rating scale, suggesting that listeners could not organize the tones in the nonhierarchical condition at all. Listeners apparently relied on tonal schema to recover the tonal hierarchy from tone duration.
Our study specifically examines tone duration and familiarity with schema for traditional Korean court music. The structure of Korean court music is generally based on a five-note, pentatonic scale. Nam (1998) examined three distinct traditional Korean pieces and produced findings suggestive of a tonal hierarchy. She measured the tunings from the steady-state part of each note and quantified the total duration of each note in each piece. All three pieces contained pentatonic scales (two with similar patterns of pitch intervals) and shared a similar durational distribution of pitches. The first scale degree was sounded for the greatest proportion of the total duration of each piece, followed by the fourth scale degree. Also, the note occurring most often and for the longest duration in each piece tended to finish phrases in that piece, just as stable and important tones in Western music tend to be played at phrase boundaries. Based on these findings, Nam argued that a tonal hierarchy is available in the surface properties of Korean music.
Experiment 1
The question of whether the tonal structure suggested by Nam’s (1998) study is actually perceived by listeners has not been addressed. The first purpose of the present study was to submit Nam’s (1998) analysis of the hierarchical pitch structure of Korean court music to a perceptual probe-tone investigation. Tone sequences were constructed in which the selection of scale tones and their durations corresponded to the distribution of one of the Korean pieces analyzed by Nam (1998). Seven nonscale tones of shorter duration than the scale tones were also included in the experimental sequences. These were musically artificial with respect to the Korean pentatonic scale but allowed a cognitive load (12 tones) comparable to most experimental work with the Western tonal hierarchy and allowed full examination of possible assimilation of probe-tone results to Western tonality.
The second purpose was to examine the role of tonal schemas and sensitivity to statistical regularities in tone sequences in the perceptual organization of tones. We tested listeners familiar with both Korean music and Western music and listeners familiar with Western music but not Korean music. Korean music is not well known worldwide but existing sources (e.g., Lee, 1981; Pratt, 1987) indicate that Korean court music is familiar within the Korean culture, is taught in the general high school curriculum, and is formally used during ceremonial occasions.
Korean-familiar listeners were expected to be sensitive to both distributional and schematic cues to structure, Western-familiar listeners to be sensitive to distributional cues only. In addition, the interval structure embedded in the Korean sequences might act as a cue for both groups of listeners. The sequences in Experiment 1 included one near-perfect fifth (705 cents) that was emphasized through relatively long durations of the notes. Other such properties of the pentatonic system, such as asymmetry in the pattern of large and small intervals (see Krumhansl, 1987a, 1990), may help perceptual organization. It is also possible that listeners unfamiliar with the tuning of the Korean pentatonic scale may assimilate the tones to the familiar Western diatonic scale.
Method
Participants
Twelve Western-familiar (3 men, 9 women) and 12 Korean-familiar (3 men, 9 women) listeners took part. The age of the Western-familiar listeners ranged from 18 to 27 years (M = 22.1, SD = 3.4), and the age of the Korean-familiar listeners ranged from 18 to 32 years (M =23.3, SD = 3.5). All were recruited from the Queen’s University community and received either $5 or course credit in exchange for one 40-minute experimental session.
Western-familiar listeners were required to have at least 6 years of formal training in Western music and to have lived in North America for their entire lives. None reported familiarity with any Asian music, including Korean. Korean-familiar listeners were also required to have at least 6 years of formal training in Western music. In addition, they were required to have completed their entire elementary school education and at least 4 out of 6 years of high school education in Korea. Korean education includes mandatory music courses throughout elementary and high school covering general knowledge of both Western and Korean music and instruction on certain Korean instruments. Korean-familiar listeners did not report familiarity with Asian music other than Korean.
For Western music training, a points system was used. Total number of points was calculated as the number of years of private training on a primary instrument plus the number of years of group training weighted by 0.5. Points for the Western-familiar listeners ranged from 6 to 14.5 (M = 12.46, SD = 3.41), and points for the Korean-familiar listeners ranged from 10.5 to 21 (M = 12.96, SD = 2.76). For Korean music training, we recorded the number of years of training within the Korean high school system (i.e., beyond elementary school). These years ranged from 4 to 6 (M = 5.3, SD = 1.0). Three Korean-familiar listeners reported 1 or 2 years of further training in Korean music.
To check the assignment of listeners further, two familiarity tests were administered. The first was a Familiar Melodies test (Liégeois-Chauvel, Peretz, Babai, Laguitton, & Chauvel, 1998), supplied on DAT tape by Professor Isabelle Peretz. The test contains 10 well-known Western tonal melodies, such as “Happy Birthday” and “Silent Night,” randomly mixed with 10 novel Western tonal melodies created for the test (retrogrades of the familiar melodies).
The second was a Korean familiarity test constructed by the second author. The test contained six well-known traditional Korean folk songs (Song, 1974) and six novel melodies that were modified versions of little-known Korean songs with the original rhythmic and scale structure retained. The melodies differed in tempo but all contained eight musical bars. The melodies of the Korean familiarity test were constructed using Cakewalk Professional Software (Hendershott, 1990) and recorded in random order on a DAT tape.
Both tests used a piano timbre and were heard through a Tascam DA-P1 DAT recorder on Sennheiser 480 headphones. The Korean familiarity test was administered first, followed by the Familiar Melodies tests. Participants stated whether each melody was familiar or unfamiliar.
For the 20 trials of the Familiar Melodies test, the scores out of 20 for both the Western-familiar (M = 19.25, SD = 1.06) and the Korean-familiar listeners (M = 19.00, SD = 0.47) were very high, and not significantly different. However, for the 12 trials of the Korean familiarity test, the Western-familiar listeners scored near chance (M = 6.25, SD = 0.45) whereas the Korean-familiar listeners scored significantly higher with near perfect scores (M = 11.33, SD = 0.98), t(22) = 21.67, p < .001. The scores confirm that Korean-familiar listeners were likely as exposed as Western-familiar listeners to simple Western melodies, but that only Korean-familiar listeners were indeed familiar with Korean songs.
Materials
Tone sequences
One of the Korean court music pieces analyzed by Nam (1998) was selected – Sangyongsan in Pyongjo Hoesang performed by Chung (1994). This particular piece was played in a leading Korean double reed woodwind instrument called a p’iri, and accompanied by a Korean string instrument called kayagûm. The music is characterized by a slow tempo and relatively simple melodic structure. The structure contains five principal tones called, in ascending pitch order, Hwang, Tae, Chung, Im, and Nam (Nam, 1997, 1998). The mean fundamental frequencies of the five tones are given in Nam (1997). The proportion of the total duration for each tone is given in Nam (1998) and shown in Figure 1. Given that there were only three Korean pieces analyzed, we chose this piece on which to base our sequences because it was similar in the pattern of pitch intervals and durational distributions to one of the other pieces. We propose that, from the information provided by Nam (1998), there is a strong likelihood that the piece we chose was representative of Korean music.
Proportion of total duration of 5 scale tones in Sangyongsan in Pyongjo Hoesang. Tones presented in order of pitch height (after Nam, 1998; recreated with the permission of the publisher).
Tone sequences were created from a tone set that included the five Korean scale tones plus seven other nonscale tones. Nonscale tones were necessary to evaluate assimilation to the Western tonal hierarchy. Further, 12 tones allowed a parallel between our study and Smith and Schmuckler (2004) in which 12 tones and 12 durations were presented to listeners. Specifically, one tone (K1) bisected the interval between Hwang and Tae, two tones (K2 & K3) trisected Tae and Chung, one tone (K4) bisected Chung and Im, one tone (K5) bisected Im and Nam, and two tones (K6 & K7) were added above Nam. The first three columns in Table 1 show, from left to right, the names of the Korean tones, the exact fundamental frequencies for each tone used in the experiment, and the intervals between the tones, measured in cents.
Korean tones, fundamental frequencies, intervals in cents, and assigned durations of tones in Experiment 1.
Table 1 shows the near-perfect fifth (705 cents) between the principal tones Hwang and Im. (In Western tonal music, the equal-tempered fifth is 700 cents; the just-tempered is 702 cents). The table also shows considerable deviation between the tuning of the remaining tones from equal temperament (in which each semitone is 100 cents). No other interval is within 25 cents of the perfect fifth. Three independently selected Western-familiar listeners heard the equal tempered E major ascending scale in alternation with the seven ascending tones from our Korean tuning nearest in frequency to the E major scale. Over six trials, all listeners were able to distinguish the difference in tuning between our scale and the E major scale with perfect accuracy.
Exact values for the duration of scale tones and nonscale tones were modeled on a distribution used by Smith and Schmuckler (2004, Experiment 1, condition 5). 1 The total duration of the five Korean scale tones was set equal to the total duration of Smith and Schmuckler’s (2004) seven Western scale tones. The duration of each Korean scale tone was then assigned in proportion to the measured proportion of duration in Sangyongsan in Pyongjo Hoesang as given in Nam (1998; see Figure 1). The duration of the nonscale tones was assigned so that the duration of the seven Korean nonscale tones equaled the average duration of the five Western nonscale tones. The duration for each scale and nonscale tone is given on the right hand side of Table 1.
Each sequence contained 24 randomly ordered tones, with all 12 tones occurring twice. Three different sequences were constructed, each with a different random order of tones. The total duration of each sequence was approximately 11 seconds.
Apparatus
Tone sequences for the probe-tone test were constructed using Cakewalk Professional Software (Hendershott, 1990). All sequences and probe tones were sine tones presented on a Zenith Z-200 computer running Midiplay software (Ergincan, 1993) through a TX81Z FM tone generator and Denon PMA-730 amplifier. The sound was delivered in a sound-attenuated chamber in which listeners heard the stimuli through Optimus Nova 71 headphones. Sound pressure level was set to a comfortable listening level for each listener; the range was approximately 55 to 70 dB SPL. Listeners rated probe tones on a remote Zenith computer terminal within the booth.
Procedure
The first two minutes of the 12-minute piece, Sangyongsan in Pyongjo Hoesang, was presented to establish a Korean setting for the experiment and possibly evoke the Korean schema in the Korean listeners. Western-familiar listeners were told that the music was to give them an idea of what Korean music is like, and Korean-familiar listeners were told that the music is an example of traditional Korean music. Given that schemas develop over a long period of time (Krumhansl, 1990), we did not expect that Western-familiar listeners would develop schema from two minutes of exposure to Korean music.
For the probe-tone task, listeners were presented with three blocks of 14 trials, with a different sequence in each block. For each trial, listeners heard the tone sequence, followed by a 1-second pause and a 2-second probe tone. Listeners then rated how well, in a musical sense, the probe tone fit the melody. They used a 7-point Likert scale, “1” indicating a very poor fit and “7” indicating a very good fit. In each block, the first two trials were practice, and the remaining 12 trials were experimental trials. During the experimental trials, each of the 12 tones of the Korean scale, ranging from Hwang to K7, occurred once as a probe tone with the order of the probe tones randomized for each listener.
Results and discussion
Analysis of variance was conducted with two factors, one within-subjects factor, probe tone (12 levels), and one between-subject factor, group (2 levels). Orthogonal contrasts followed, comparing the longer scale tones to the shorter nonscale tones as well as comparisons among the scale tones. Figure 2 shows mean ratings for each tone in the order of pitch height for both groups. Shaded bars indicate ratings for scale tones with darker bars indicating longer tones. There was a close correspondence between the set of mean ratings from each group, r(10) = .87, p < .001. Moreover, mean probe-tone ratings from both groups corresponded significantly with the distribution of durations for the 12 sequence tones (see Table 1) – for Western-familiar listeners, r(10) = .70, p = .011, and for Korean-familiar listeners, r(10) = .72, p < .01.
Mean ratings and standard error bars of each scale tone and nonscale tone in Experiment 1. Tones presented in order of pitch height with darkest bars indicating longest tones.
The ANOVA showed a significant main effect of probe tone, F(11, 242) = 15.85, p < .001. As is clear in Figure 2, ratings for the scale tones were significantly higher than the ratings for the nonscale tones, F(1, 22) = 40.40, p < .001 (Western-familiar: F(1,11) = 24.49, p < .001; Korean-familiar: F(1,11) = 17.14, p = .002). There was no significant Probe Tone X Group interaction on this contrast meaning participants from both groups rated the longer scale tones significantly higher than the shorter nonscale tones. A tonal hierarchy was also found within the scale tones. Overall, the longest scale tone, Hwang, was rated significantly higher than the other four scale tones, F(1, 22) = 5.74, p = .026 (Western-familiar: F(1,11) = 4.46, p = .059; Korean-familiar: F(1,11) = 1.68, p = .221). The next two longest tones, Im and Tae, were rated significantly higher than the two shortest scale tones, Nam and Chung, F(1, 22) = 8.82, p = .0071 (Western-familiar: F(1,11) = 2.41, p = .149; Korean-familiar: F(1,11) = 6.65, p = .026).
There were no significant Probe-tone × Group interactions on these contrasts (p > .30 for all contrasts) suggesting that the groups perceived the same structure within the sequences. Hwang was rated at the top of the hierarchy, followed by Im and Tae, followed by Nam and Chung. Nonscale tones were rated at the bottom of the hierarchy.
Castellano et al. (1984) and Kessler et al. (1984) found only subtle differences in the way listeners of different cultures perceived familiar or unfamiliar music. Therefore, we examined simple effects to see if there were any significant differences in ratings for any of the scale tones between the groups. Using independent samples t tests, we found no significant differences in mean ratings between the groups for any of the scale tones. There is little evidence to indicate that there are differences in the way the Korean-familiar and Western-familiar listeners perceived the sequences.
Assimilation to the Western tonal hierarchy
The durational distribution of the tones in the experiment corresponded closely to a mistuned E major scale, r(10) = .82, p = .001. Multiple regression analyses allowed assessment of the possibility that listeners perceptually organized the sequences by assimilating sequence tones to the Western tonal scale. Predictors of the ratings for each group were the durations of each tone and the standardized probe-tone profile for the E Major scale from Krumhansl and Kessler (1982). There were no significant positive correlations between either groups’ ratings and standardized probe-tone profile of any other major key that could be suggestive of assimilation to the Western tonal scale. The regression model significantly predicted the Western-familiar listeners’ ratings with an adjusted R2 = .44, F(1,10) = 9.61, p = .011. Duration, entered first, significantly added to the predictability of the ratings, t(10) = 3.10, p = .011 and a standardized beta coefficient of .70. The standardized probe-tone profile was entered next and did not significantly add to the predictability of the ratings beyond that of duration. Similarly, the regression model significantly predicted the Korean-familiar listeners’ ratings with an adjusted R2 = .46, F(1,10) = 10.44, p = .009. Duration, again entered first, significantly added to the predictability of the Korean-familiar ratings, t(10) = 3.23, p = .009 and a standardized beta coefficient of .72. The standardized probe-tone profile was entered next and again did not significantly add to the predictability of the ratings beyond that of duration. It appears that neither group assimilated the tones of the sequences to the Western tonal system.
The probe-tone task, we employed was difficult for listeners. Unlike most previous probe-tone paradigms, the pitches of the sequence tones provided no information about a Western tonal hierarchy embedded in the tone sequence. Listeners were required to differentiate tone durations and to associate each pitch with its particular duration. Nevertheless, listeners both familiar and unfamiliar with the style of traditional Korean court music were remarkably adept at abstracting durational information from the sequence and recovering evidence of a Korean tonal hierarchy.
Overall, the similarity of the results for the Korean-familiar and the Western-familiar listeners was striking. It was not attributable to assimilation to the Western diatonic scale with which both groups were familiar. Rather, the similarity between both groups appeared largely due to shared sensitivity to distributional information in the tone sequence. It is likely that all listeners had acquired implicit knowledge that the duration of tones in music conveys information about tonal structure; thus it makes sense to suppose that they would use this knowledge to engage a perceptual strategy for the present task.
The data do not support the suggestion, advanced earlier, that only Korean-familiar listeners would be able to extract the durational pattern in the sequence because only Korean-familiar listeners would have internalized schema for Korean music. The data therefore do not seem consistent with those of Smith and Schmuckler (2004). From their results, it would be expected that Western-familiar listeners would not recover a durational distribution. However, there are several design differences between Smith and Schmuckler and the present study. For instance, Smith and Schmuckler’s conditions contained 12 different durations as opposed to six different durations contained in Experiment 1. It is possible that for nonhierarchical conditions, as defined by Smith and Schmuckler, there is a limit to the number of different durations that a listener can track to build a pitch structure. A further difference between Smith and Schmuckler’s (2004) nonhierarchical condition and ours is that in the present study durations were not randomly distributed to tones but applied to an established pentatonic scale. Properties of the Korean scale, such as the intervallic relationships among tones, may have facilitated the organization of pitch structure for listeners unfamiliar with the scale.
Experiment 2
Experiment 1 showed that both Korean-familiar and Western-familiar listeners could abstract the durational distribution from the sequences to recover a Korean tonal hierarchy. The relationships among the five Korean scale tones appear to have allowed the Western-familiar listeners to recover much of the Korean tonal hierarchy based on the durations of the tones. Consonance has long been suggested as an important factor in the development of music scales (Handel, 1989; Krumhansl, 1990). Simple-ratio frequency interval relationships among tones, such as those occurring in the perfect fifth, are the most consonant pairs of tones and facilitate perceptual organization in melodies and chords (Schellenberg, Bigand, Poulin-Charronnat, Garnier, & Stevens, 2005; Schellenberg & Trainor, 1996; Schellenberg & Trehub, 1996), even across nonadjacent tones (Endress, 2010). They may be the structures upon which tonality is built (Schellenberg et al., 2005). Endress, Dehaene-Lambertz, and Mehler (2007) suggested that there may be perceptual primitives, such as consonance, that are analyzed prior to statistical learning. The idea is supported by findings that music scales historically and across most cultures contain simple-ratio intervals (Gill & Purves, 2009) and that infants as young as nine months old are sensitive to changes to simple-ratio intervals. The suggestion is that we are born with or at least very quickly develop sensitivity to simple-ratio intervals (Schellenberg & Trainor, 1996; Schellenberg & Trehub, 1996). The consonance in the near-perfect fifth between Hwang and Im is the most likely relationship helping listeners to organize the sequences in Experiment 1.
Sequences in Experiment 2 were designed to de-emphasize the intervallic structure of the Korean hierarchy by applying longer durations to only one tone of any pair of tones forming a near-perfect fifth. The near-perfect fifth intervals are still available in the sequences but they are no longer highlighted through duration. If listeners are sensitive to the durational distribution of the sequences, then long tones should receive higher ratings in a hierarchical manner as in Experiment 1. If listeners are sensitive to the intervallic properties of the Korean scale, regardless of duration, then the Korean scale tones should receive higher ratings because the perfect fifth intervals are still present. Indeed, a possibility is that the Korean-familiar listeners may be able to use tonal schema to recover the Korean hierarchy. Finally, if duration is necessary to highlight the intervallic structure of the Korean tones, then little or no structure should be found in the ratings.
Method
Participants
Twelve Western-familiar (6 men, 6 women) and 12 Korean-familiar (1 man, 11 women) listeners took part. The age of the Western-familiar listeners ranged from 18 to 28 years (M = 21.8, SD = 2.9), and the age of the Korean-familiar listeners ranged from 19 to 28 years (M = 22.2, SD = 2.6). All listeners received either $5 or course credit for the 40-minute session.
As in Experiment 1, Western-familiar listeners were required to have at least 6 years of formal training in Western music and to have lived in North America for their entire lives. None reported familiarity with any Asian music including Korean. The Korean-familiar listeners were also required to have at least 6 years of formal training in Western music. In addition, they were required to have completed their entire elementary school education and at least 4 out of 6 years of high school education in Korea to ensure adequate exposure to traditional Korean court music.
For Western music training, the same point system described in Experiment 1 was used. Points for the Western-familiar listeners ranged from 6 to 13.5 (M = 10.54, SD = 2.80). Points for the Korean-familiar listeners ranged from 6 to 14 (M = 9.17, SD = 3.15). For Korean music training, we again recorded the number of years of training within the Korean high school system. For the Korean-familiar listeners, training ranged from 4 to 6 years.
The Familiar Melodies test (Liégeois-Chauvel et al., 1998) and the Korean familiarity test were again administered to check familiarity with Western and Korean music, respectively. Both tests were administered in an identical way as in Experiment 1.
For the 20 trials of the Familiar Melodies test, the scores for the Western-familiar (M = 18.08, SD = 1.38) and the Korean-familiar listeners (M = 16.67, SD = 3.42) were both high, and not significantly different. However, for the 12 trials of the Korean familiarity test, the Korean listeners scored near perfect (M = 11.00, SD = 1.28) whereas the Western-familiar listeners scored significantly lower and near chance (M = 6.92, SD = 1.08), t(22) = 8.44, p < .001. The scores confirm that both groups were familiar with simple Western melodies, but that only Korean-familiar listeners were familiar with the Korean folk songs.
Materials
Tone sequences
Tones and durations were identical to those in Experiment 1, as shown in Table 1. However, durations were not assigned to tones according to the Korean tonal hierarchy but were assigned in a quasi-random order. Pairs of tones forming an interval of a near-perfect fifth were identified and durations were assigned such that no more than one tone in each pair was assigned one of the five longer durations. As seen in Table 2, two different assignments of duration to tones were created by assigning the longer duration to a different tone of each pair in each assignment, designated Assignment 1 and Assignment 2. Assigning the longer durations to different tones in each assignment ensured that we did not inadvertently make salient some property of the tones in the sequences that would affect ratings beyond duration. For example, Hwang and Im form a near-perfect fifth so in one assignment of duration to tones, Hwang was assigned a long duration whereas Im was assigned the shortest duration. In the other assignment, Im was assigned a long duration whereas Hwang was assigned the shortest duration. Although not a known musical scale, the long tones in each assignment will be referred to as scale tones and the short tones will be referred to as nonscale tones. Three different sequences were constructed from each assignment of duration to tones, each with a different random order of tones.
Intervals in cents of tone pairs forming a near perfect fifth, and assigned durations of tones in Experiment 2.
Apparatus
The apparatus was identical to that in Experiment 1 with the exception that listeners heard the stimuli through Roland RH-80 headphones.
Procedure
Presentation of the first two minutes of Sangyonsan in Pyongjo Hoesang as well as the Familiar Melodies test and Korean familiarity test were the same as in Experiment 1. The probe-tone task was also identical to Experiment 1 except that there were six sequences of randomly ordered tones resulting in six blocks of 14 trials.
Results and discussion
Analysis of variance, with one between-subject variable (group) and one within-subject variable (probe tone), was applied to the data. Orthogonal contrasts followed, comparing the longer scale tones to the shorter nonscale tones as well as comparisons among the scale tones.
Figure 3 shows the mean ratings in order of pitch height for both groups in both Assignments. Shaded bars indicate the scale tones with darker bars indicating longer tones. Overall, there was a significant main effect of probe tone, F(11, 242) = 4.45, p < .0001. The five longer scale tones were rated significantly higher than the seven shorter nonscale tones, F(1, 22) = 9.66, p = .005 (Western-familiar: F(1,11) = 6.12, p = .031; Korean-familiar: F(1,11) = 3.54, p = .087), showing that duration affected the perception of pitch structure in the sequences. There was no significant Probe Tone × Group interaction on this contrast suggesting both groups rated the tones in a similar pattern across the scale and nonscale tones.
Mean ratings and standard error bars of scale tones and nonscale tones in Experiment 2 for each group. Top panel shows ratings for Assignment 1; bottom panel shows ratings for Assignment 2. Tones presented in order of pitch height with darkest bars indicating longest tones.
None of the orthogonal contrasts among the scale tones were significant meaning that the apparent differences between the scale tones are due to chance. The particular order of mean ratings for scale tones is therefore arbitrary rendering correlations with potential predictors, such as assimilation to Western or Korean tonal hierarchies, meaningless and potentially misleading.
The Probe Tone × Assignment interaction approached significance, F(1,22) = 3.63, p = .070. Analysis on each assignment separately showed that, although mean ratings for the scale tones were higher than nonscale tones in both assignments, ratings for scale tones were significantly higher only in Assignment 2, F(1, 22) = 11.81, p = .002. The contrast was in the same direction but not significant in Assignment 1, F(1, 22) = 2.82, p = .107.
None of the interactions concerning the groups were close to reaching significance, suggesting that both groups perceptually organized the sequences in a similar way. There was a significant correlation between the two groups’ mean ratings across both assignments, r(22) = .41, p = .046, although the correlation was not as high as in Experiment 1 (r(10) = .87, p < .001).
The main effect of scale tones versus nonscale tones and the near significant Probe-tone × Assignment interaction suggested that there may have been differences in mean ratings within the two assignments. An exploratory ANOVA was therefore applied to examine differences in mean ratings among scale tones and among nonscale tones separately in each assignment for each group.
Significant differences were found in only two conditions. In Assignment 1, there was a significant difference among Korean-familiar scale tone ratings, F(4,44) = 10.66, p < .001. Post hoc t tests showed ratings for the two highest pitched scale tones, K5 and K6, were rated significantly higher than the three lowest pitched scale tones (p’s ranged from .002 to .015). Further exploratory correlations with factors such as duration, standardized profiles for all major keys and pitch height were conducted to see if there was any evidence that Korean-familiar listeners were following a consistent strategy to find pitch structure. The only significant correlation with mean ratings was pitch height, r(10) = .85, p < .001 with higher pitches receiving higher ratings. It is possible that, in the absence of salient intervallic relationships, Korean-familiar listeners attempted to base perceived pitch structure on pitch height, at least in Assignment 1.
In Assignment 2, there was a significant difference among Western-familiar nonscale tone ratings, F(6,66) = 3.20, p = .008. Post hoc t tests showed K5 was rated significantly higher than every other nonscale tone (p’s ranged from .001 to .021). It is uncertain why one nonscale tone received salience in Assignment 2. It was not the final tone in any sequence, so was not more salient because it was the last tone heard before the probe tone. Further exploratory correlations were conducted to see if there was any evidence that Western-familiar listeners were following a consistent strategy to find pitch structure. The only significant correlation was found between the ratings and the standardized profile for C major, r(10) = .59, p = .043, suggesting that Western-familiar listeners may have assimilated Assignment 2 sequences to the key of C major. However, two points argue against assimilation to C major. First, the correlation is significant but only moderately high. If Western-familiar listeners assimilated the sequences to C major, one would expect closer to a perfect correlation. Second, the lack of differences in mean ratings among scale tones and among the rest of the nonscale tones suggests that the order of mean ratings are due to error variance, making any correlation spurious. The mean ratings in Experiment 2 provide another clue to the difficulty listeners had in organizing the tones. On a scale of 1 to 7, only one tone had a mean rating above 5 or below 3 for either group in either assignment. Apparently, listeners heard most of the tones as more or less neutral, essentially declaring they heard little consistent pitch structure in the sequences, other than the overall scale tone versus nonscale tone division.
Experiment 1 versus Experiment 2
Comparisons were made between Experiments 1 and 2. Listeners in both experiments were able to distinguish between long tones and short tones but only listeners in Experiment 1 were able to recover the full durational distribution. Were listeners in Experiment 1 actually able to organize the pitch structure better than listeners in Experiment 2? In order to show that listeners perceived structure better in Experiment 1 compared to Experiment 2, we combined the ratings from both experiments for analysis on the scale tone versus nonscale tone contrast. Assignments 1 and 2 were treated as separate conditions for this analysis. Analysis of variance was conducted with probe tone (12 levels) and experiment (3 levels: Experiment 1, Experiment 2 – Assignment 1, and Experiment 2 – Assignment 2) as independent variables. The scale tone versus nonscale tone comparison was the only comparison of interest. Data were collapsed across groups and the three tone orders for the analysis.
Figure 4 shows the mean ratings for scale tones and nonscale tones for Experiment 1 as well as both assignments of Experiment 2. The Probe Tone × Experiment interaction was indeed significant, F(1, 66) = 17.80, p < .001. The difference in ratings between scale tones and nonscale tones was significantly greater in Experiment 1 than either Experiment 2 Assignment 1, F(1, 44) = 17.73, p < .001, or Assignment 2, F(1, 44) = 7.82, p = .008. Listeners in Experiment 1 not only were able to differentiate among the scale tones better than listeners in Experiment 2, they were also able to distinguish between the scale tones and nonscale tones to a significantly greater degree than listeners in Experiment 2.
Comparison of mean ratings and standard error bars of scale tones and nonscale tones in Experiment 1 and both Assignments of Experiment 2.
General discussion
Employing Korean scale tuning and durational distributions based on Nam (1997, 1998), Experiment 1 showed that the Korean tonal hierarchy is perceivable by listeners familiar with Korean court music as expected, but that listeners unfamiliar with Korean court music could also abstract the Korean tonal hierarchy from the sequences. Neither schema nor sensitivity to duration alone can explain the data from Experiment 1. If schema alone influenced the ratings, there should have been large differences between the groups as Western-familiar listeners had no schema on which to organize pitch structure. Duration alone could not have influenced the ratings in Experiment 1 because listeners in Experiment 2 were not able to differentiate among the long tones as did listeners in Experiment 1. Results of the two experiments together are indicative of the idea that duration emphasizes the intervallic structure found in the Korean scale, allowing the organization of pitch structure in Experiment 1 but not Experiment 2.
Similar to Castellano et al. (1984) and Kessler et al. (1984), there was no evidence that listeners assimilated the sequence into the Western-tonal schema to help them organize the pitch structure. In Experiment 1, mean ratings did not correlate with any of Krumhansl and Kessler’s (1982) Western major probe-tone profiles. In Experiment 2, there was no significant differentiation among ratings for the scale tones or among ratings for the nonscale tones. The order of the mean ratings was therefore due to random error and any correlational analyses would be meaningless. In the exploratory post hoc tests, a moderately high correlation with the standardized profile of C major was found in one condition and only by Western-familiar participants, suggesting the possibility of assimilation of the sequences to the Western C major hierarchy. However, the correlation, though significant, is not really strong enough to support the idea of assimilation.
There were similarities between our experiments and Smith and Schmuckler’s (2004) experiment. In Experiment 1, and Smith and Schmuckler’s hierarchic condition, duration emphasized the intervallic relationships of a scale and a tonal hierarchy was recovered. In Experiment 2, and Smith and Schmuckler’s nonhierarchic condition, the durational distributions did not emphasize any intervallic relationships. However, in Experiment 2, listeners were still sensitive to duration although only at a coarse level, differentiating long tones from short tones whereas Smith and Schmuckler’s listeners were not at all sensitive to duration as a cue to pitch structure. Smith and Schmuckler’s (2004) nonhierarchic sequences included 12 different durations in which the shortest tones were 24 ms. Our Experiment 2 sequences included only six different durations in which the shortest tones were 168 ms. The differences in the number and length of the durations appear to have made it more difficult for Smith and Schmuckler’s listeners to perceive any pitch structure in the nonhierarchic sequences.
Other cross-cultural studies found only subtle differences between those familiar and unfamiliar with a music style (Castellano et al., 1984; Kessler et al., 1984; Krumhansl, 1999, 2000). There was no evidence to support cross-cultural differences in Experiment 1, in which the durational distributions of tones matched a Korean tonal hierarchy. Western- and Korean-familiar ratings were significantly correlated to a high degree and there was no significant contrast by group interaction of scale tones versus nonscale tones in the analysis of variance nor even simple effects comparing between-group ratings. In Experiment 2, there were neither main effects of group nor any interactions involving group in the analysis of variance.
Only one of the Korean pieces analyzed by Nam (1997, 1998) was used to generate the distributional properties of the tone sequences in Experiment 1. Although it is likely that this piece is representative of Korean court music, other pieces in the repertoire could be used to test the generality of our findings. The stimulus materials were artificially constructed rather than excerpts from real Korean pieces. They may, however, provide a standardized probe-tone profile that could be compared to profiles obtained with real Korean music in future research. Although Korean court music is familiar within Korea (Lee, 1981; Pratt, 1987), our Korean-familiar listeners received training in Korean music only at the high school level. A future study might recruit more highly trained Korean musicians. Use of real Korean music and more highly trained musicians could possibly reveal closer engagement of a Korean tonal schema. Cross-cultural and novel scale systems have been categorized according to natural properties such as number and tuning of tones, consistency of tuning, distribution of interval sizes, number of instances of each interval size, and so forth (Krumhansl, 1987a, 1990). As Krumhansl (1987a) noted, the degree to which these properties constrain the emergence of a tonal hierarchy is a matter of question. Our results provide evidence that the perfect fifth may have a large influence on perceiving structure within a scale system, but that the perfect-fifth relationship may need to be emphasized in some way, such as with longer duration. De-emphasizing the perfect-fifth intervals in Experiment 2 led to the decreased ability of either group to recover a hierarchical durational distribution. A systematic examination of the effects of the perfect fifth and other natural scale properties on perceived structure would be an important topic for future research.
Summary
The purpose of our study was to examine the perceived pitch structures of tone sequences in which the durational distribution of tones matched (Experiment 1) or did not match (Experiment 2) a Korean tonal hierarchy found by Nam (1998). We found that listeners both familiar and unfamiliar with Korean music were able to perceive pitch structure when the durational distribution of tones matched the Korean tonal hierarchy. There was little evidence to suggest the Korean-familiar listeners had an advantage over the Western-familiar listeners in recovering a Korean tonal hierarchy. When the durational distribution did not match the Korean tonal hierarchy, neither group appeared to perceive a clear pitch structure beyond a long tone versus short tone distinction. Our results suggest that listeners can perceive pitch structure when the durational distribution of the tones emphasizes the intervallic relationships within the intervals of the Korean pentatonic scale.
Footnotes
Funding
The research was supported by a Summer Undergraduate Research Award to the second author and a Discovery Grant to the third author from the Natural Sciences and Engineering Research Council of Canada [grant number NSERC RGPIN 333-10]. We thank C. L. Krumhansl, U. Nam and M. R. Jones for supportive discussions, comments from three anonymous reviewers, and C. K. Koh, N. A. Smith, M. Cassels, and A. K. Kelly for research assistance.