Sie sind auf Seite 1von 16

Journal of Expert mental Psychology: Human Perception and Performance 1998, Vol. 24, No.

5,1546-1561

Copyright 1998 by the American Psychological Association, Inc. 0096-1523/9WS3.00

The Interplay Between Metric and Figural Rhythmic Organization


Stephen Handel University of Tennessee, Knoxville
Three experiments investigated the relative importance of figural and metric rhythmic organizations. Figural organization is determined by the numbers of tones in successive groups. For figural organization alone, the timings between the onsets of each group are relatively unavailable, so listeners cannot discriminate between 2 rhythms that have the same sequence of groups but different timings between the groups. Thus, traditional views argue that a metric organization is necessary: The timing between adjacent groups is perceived by means of the strong-weak sequence of beats. These experiments, however, suggest a limited role for meter. The metric strength of the individual rhythms affected discrimination of pairs of different rhythms with the same figural organization only when an external meter pulse accompanied the rhythm and only when the rhythm with the stronger meter was the first of the pair.

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The purpose of the present experiments was to investigate the roles of the metric (i.e., periodic) and figural (i.e., grouping or serial) organizations in the perception of rhythmic timing patterns composed of identical tones. Traditionally, rhythmic organization has been thought to depend on identification of the meter or beat of the passage. Passages in which such a meter cannot be easily created are considered to be unstable and therefore difficult to encode and reproduce. The results of the present experiments argue that this is an overly simplified conceptualization and that the metric organization functions only within the more fundamental figural organization created by element grouping. Metric organization is the sense of a regular periodic sequence of subjectively stronger and weaker beats that characterize music. The meter forms a time-based lattice that serves to create the rhythmic organization. The meter occurs at several hierarchical levels at once so that the beats at higher levels occur at integer multiples of the beats at lower levels. The strength of any beat is determined by the number of levels at which the beat appears. For example, consider four-beat meters based on a repeating unit of 16 elements as used here. The strong beats would occur at Elements 1,5, 9, and 13; stronger beats would occur at Elements 1 and 9; and die strongest beat would occur at Element 1 (Palmer & Krumhansl, 1990). Notes that fall at the points of the strong beats become accented, and notes that fall at the points of weak beats are unaccented. Alternative meters could be based on units of three, so that for a repeating unit of 12, the strongest beats would occur on Elements 1,4,7, and 10. To define strong and weak metric rhythms, I made use of the rules suggested by Povel and Essens (1985) for seI thank Piet Vos, Greg Sandell, and Mari Riess Jones for their helpful comments on the manuscript and Hancel Woods for his help in completing Experiment 3. Correspondence concerning this article should be addressed to Stephen Handel, Department of Psychology, University of Tennessee, Knoxville, Tennessee 37996-0900. Electronic mail may be sent to shandel@utk.edu.

quences of identical elements separated by different lengths of silent intervals. Povel and Essens began by observing that in a sequence of identical tones, certain tones appear accented. These include (a) relatively isolated tones, (b) the second tone of a pair of tones, and (c) the first and last tones in a series of tones. Povel and Essens continued by arguing that a strong meter emerges when those accented tones fall at regular intervals (i.e., beat positions). Empirically (Essens & Povel, 1985), the most important factor in determining the ease of reproducing the rhythm and, by inference, in determining the strength of the meter is that a tone should occur at the positions of the strong meter beats, and a silence or rest should not. Theoretically, in determining the bestfitting meter, Povel and Essens (1985) weighted the lack of coincidence of a tone and beat as the most important factor. On this basis, for rhythms 16 elements long I operationally defined the metric strength of a rhythm by the occurrence of tones at the beat positions 1, 5,9, and 13. Thus, the rhythm X.X.X...X...X... (me Xs represent tone elements, the dots represent isochronous silent time intervals, and tones separated by one time unit are heard as forming a single group) would be strongly metric, because tones fall on the stronger beats at Elements 1, 5, 9, and 13, but the rhythm X.X...X..X.X.... would be only weakly metric because tones fall only at Element 1. From this perspective, rhythms are not simply metric or nonmetric. Instead, each rhythm is metric to some degree, depending on the strength of the meter interpretation it evokes. Povel and Essens (1985) argued that listeners attempt to find a meter to fit a rhythmic pattern (i.e., a template) and that highly metric rhythms more easily induce an internal clock that encodes the rhythm in terms of the meter. In contrast, weak metric rhythms do not induce an internal clock, and therefore the rhythm cannot be encoded in terms of a temporal grid in which every element can be located and timed. Figural organization is the sense that a sequence of tones is heard as a series of discrete groups. This organization may be based on shared acoustic properties such as duration, pitch, timing, or timbre or may be based on a trajectory such
1546

METRIC AND FIGURAL RHYTHMS

1547

as an ascending or descending scale. In the present context, the grouping is based on the timing between tones, so that the groups are composed of tones separated by one silent intertone interval (one dot in the representation). Bamberger (1978) and Povel and Essens (1985) termed this figural grouping because the groups are figures perceived against an ordinal ongoing time. The rhythms are organized into bounded groups of elements that follow one another, but the tunings between the onsets of the successive groups are not encoded or compared. For example, the rhythm X.X..X....X.X... would be coded as 2 tones, silence, 1 tone, silence, 2 tones, silence (written as 2-1-2-); the lengths of the silent intervals separating the groups would be coded roughly, if at all. One would expect a similar rhythm with the identical figural organization (e.g., X.X....X..X.X..., which is also coded 2-1-2-) to be easily confused with the former rhythm because the timing differences between the groups are not used. In previous research (Handel, 1992) I demonstrated that for weak metric rhythms composed of identical tones separated by different intertone timings (similar to those discussed above), listeners heard the rhythms in terms of the groups of tones and could not accurately judge the timings between groups. When two different rhythms had the same figural organization, listeners perceived the two rhythms as being identical, and discrimination was below chance. In contrast, when two different rhythms had different figural organizations, discrimination was quite good. My goal in the experiments reported here was to investigate how the figural grouping organization interacts with the metric organization by using rhythms with stronger as well as weaker meters. Although all theories of rhythm postulate that both types of organization jointly determine the emergent rhythm, the precise relationship between the two is unspecified. For example, Lerdahl and Jackendoff (1983) suggested that the metric organization dominates shorter sections of a composition but that the grouping organization dominates longer sections. If a strong meter leads to a percept based on a fixed hierarchical timing structure, then that structure should preserve the timings both within and between groups of tones. In general, it should be easier to distinguish between two different strong metric rhythms or between a strong and a weak metric rhythm than between two weak metric rhythms. More specificallyand this is the focus of these experimentsit should be possible to discriminate between two different rhythms with the same figural organization that differ in meter. However, if a strong meter does not enhance rhythmic organization, then the ability to distinguish between two rhythms should depend on whether their figural organizations are the same, not on the strength of the meter of either rhythm. Moreover, if the figural organization is primary, and the metric organization subsequently elaborates the timing of the figural organization, then the effect of an imposed external meter pulse should depend on the specific figural organizations of the two rhythms and should not improve discrimination among all pairs.

Experiment 1
My specific purpose in Experiment 1 was to investigate discrimination among a wide variety of strong and weak metric rhythms. The goal was to determine if metric strength affected discrimination of pairs of identical rhythms and pairs of different rhythms with the same or different figural organization(s).

Method
Participants. All 57 participants were undergraduates at the University of Tennessee who received course credit for their participation. They were tested in groups of from 1 to 3. Rhythms. All rhythms were based on five tones embedded in a repeating pattern of 16 grid elements. The rules used to construct the rhythms were as follows: (a) A tone always occurred on the beat at the first element; (b) a tone always occurred on the beat at the 13th element but never occurred on the final three grid elements; and (c) there was at least one silent element, but no more than three silent elements, between any pair of adjacent tones. The combination of these three rules made the longest silent interval three elements, and one such interval always occurred at the end of the rhythms. However, there could be other equally long silent intervals, so that organization according to the gap rule, in which the longest silent interval ends the rhythm, might be ambiguous (Garner, 1974; Handel, 1974). However, in practice, this did not prove to be a problem, and no participant claimed that the rhythms were transformed from their starting configuration. I generated all of the rhythms using a 16-element template for several reasons. First, the number of possible rhythms mat satisfy the above rules is relatively small, so it is possible to adequately sample the range of rhythmic complexity. Second, the rhythms include simple and complex ones, but no one rhythm is so difficult that it is impossible to pick up in from two to four repetitions. Third, the length of one repetition played at normal tempo is well within the memory span. Fourth, a 16-element grid is fit perfectly by either a two- or four-beat meter. The strongest beats occur at Elements 1 and 9, weaker beats occur at Elements 5 and 13, and the weakest beats occur at Elements 3,7,11, and 15. A four-beat meter seems most natural to listeners of Western music: Bolton (1894) noticed that listeners spontaneously grouped isochronous tones into units of 4 and 2 as opposed to 3, and Smith and Cuddy (1989) found that four-beat meters produced better performance man three-beat meters. The preceding rules could generate 19 possible rhythms. Three rhythms were perfectly metric, having tones at Elements 1, 5, 9, and 13; four rhythms were strongly metric, having tones at Elements 1, 9, and 13; four rhythms were metric, having tones at Elements 1, 5, and 13; and eight rhythms were weakly metric, having tones only at Elements 1 and 13. There were 19 pairs that contained two identical rhythms, one for each of the possible rhythms. To restrict the number of pairs that contained two different rhythms and to determine if participants could identify which tone differed between the two rhythms, I selected the pairs of different rhythms such that the two rhythms of a pair differed by one tone shifted one grid element. There were 29 such pairs (disregarding the order of the two rhythms). A total of 48 pairs of rhythms were used in the experiment. Nineteen pairs contained two identical rhythms (all of the possibilities). Twenty-nine pairs contained two different rhythms: From the possible set of 29 pairs, 21 pairs were presented in one order, and 4 different pairs were presented in both orders (8 in total). Across the pairs of different rhythms, the five instances in which the two

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1548

HANDEL

rhythms had the same figural organization were included, whereas the remaining pairs were made up of two rhythms with different combinations of metric strength. All individual rhythms occurred at least once, and nearly all of the rhythms occurred equally often as the first or second rhythm of the pair. Task. On every trial, two rhythms were presented, and the participant judged whether the two rhythms were the same or different in terms of the relative timing of the tones. If the participants thought that the two rhythms were identical, they circled the word same on the answer sheet. If they thought the two rhythms were different, they circled one of five equally spaced Xs that represented the individual tones in order to indicate which one had changed timing (the participants were instructed beforehand that only one of the three middle tones could change, so that, effectively, they were to circle one of the three middle Xs). The first rhythm was always presented with a higher pitch tone (586 Hz, triangle waveform), and the second rhythm was always presented with a lower pitch tone (440 Hz, triangle waveform). Presentation rate. The tones were presented at a moderate rate: The duration of each grid element was 133 ms, so the length of one repetition was 2.13 s (16 X 133 ms) and there were 2.3 elements/s. The interval between Beats 1,5,9, and 13 was about 500 ms, close to the preferred tempo of major beats (Fraisse, 1982; Parncutt, 1994). Each tone was composed of a 10-ms onset ramp, a 50-ms steady state, and a 10-ms offset ramp (roughly 50% of the duration of a grid element). The rhythms were generated with BRS-Foringer modules and were prerecorded and presented to participants on cassette tapes. The experimental session took place in a small room (3 X 4 m) with acoustical ceiling tile. The participants were seated 2.5 m from two vertically stacked speakers, each of which presented the entire rhythm. The rhythms were presented at a comfortable listening level, approximately 65 dB (SPL); participants were allowed to adjust the loudness if they wished. Alternation conditions. The rhythm-pairs were presented in two ways. In the first, each rhythm was presented once, and then the pair was recycled three times (i.e., notated AXAXAXAX to indicate that the presentation would be AA for pairs of identical rhythms and AB for pairs of different rhythms). In the second, each rhythm was presented two times, and then the pair was recycled (i.e., AAXXAAXX). For both conditions, the following rhythm started at the finish of the three silent grid elements that ended each rhythm (i.e., at 2.13-s periods). There were no differences between the repetition of one rhythm and the alternation between different rhythms. I chose these two conditions to determine if the number of repetitions and alternations between the two rhythms affected discrimination. Experimental design. The design was within subjects: 2 alternation conditions X 48 rhythm-pairs. Each participant was presented with two blocks of trials, one block for each alternation condition. For each alternation condition, three different sequences of the pairs were constructed that roughly counterbalanced order. The order of presentation of the alternation conditions and the sequences within each condition were counterbalanced across participants. Procedure. Two strategies were used to acquaint participants with each condition. First, before the actual presentation of each alternation condition, there were four practice trials that used simpler four-element rhythms. Two consisted of identical pairs, and two consisted of different pairs. If participants were confused, then these rhythms were repeated until participants felt confident. Second, the first two rhythm-pairs were repeated later among the 48 experimental trials (thus, there were actually 50 trials per block). The participants were not told that these were practice trials, and the results were not used. There was an 8-s interval between each

trial, during which participants made their responses. Participants did not receive any feedback. There was a short break between the two blocks, and the experimental session lasted about 50 min. Results For the pairs of identical rhythms, the percentage of "identical" responses was the only possible measure. For the pairs of different rhythms, there were two measures: (a) the percentage of responses correctly discriminating the two rhythms and (b) the percentage of responses that correctly identified which tone had changed timing. Preliminary analyses indicated that there were no differences in discrimination among the three orders for each alternation condition (the average absolute difference was 7%), that there were no differences between the first and second blocks (79% correct for Block 1; 80% correct for Block 2), and that there were no differences between the alternation conditions (the average absolute difference across rhythms was 3%). Thus, the differences between the control variables and between the two alternation conditions were so small that all the results were combined. The percentage correct is based on two responses from each of the 57 participants (114 responses in total), one from each alternation condition. Pairs of identical rhythms. As described in the Method section, the 19 rhythms and resulting pairs of identical rhythms logically could be classified into four levels of metric strength on the basis of whether tones occurred at Element 5, Element 9, both Elements 5 and 9 (5&9), or neither Element 5 nor Element 9 (none). The results are shown in Table 1. Overall, there was a significant difference between the pairs of identical rhythms, F(18, 1008 = 3.9, p < .005, MSB = 0.11. The pairs were then placed into groups of equivalent discrimination performance through the use of 1\ikey's honestly significant difference (HSD) procedures. These procedures indicated that the 19 pairs could be placed into three groups and that each group contained pairs with different metric strength. The first group contained (a) the three perfectly metric rhythm-pairs with tones at both Elements 5 and 9 and (b) Rhythm 16. In this group, discrimination was nearly perfect; 95% of the responses indicated that the two rhythms were identical. The second group contained 10 rhythm-pairs, essentially those for which the percentages correct were in the 80% range. The third group contained the five rhythm-pairs with percentages correct that were below 80%. These latter rhythms tended to have weaker metrics. In sum, there is some evidence that metric strength affected discrimination, but the difficulty of the rhythms varied across a strict metric categorization. Pairs of different rhythms. These parrs can be classified in two ways. The first is according to whether or not the two rhythms had the same figural organization. There were 5 pairs that had the same figural organization and 24 that did not. The second is to create a 4 X 4 table defined by the metric strength of the first and second rhythms (the four levels being 5, 9, 5&9, or none). For example, one possibility could be defined by the first rhythm having a tone at

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

METRIC AND FIGURAL RHYTHMS

1549

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 1 Percentages of Correct Discriminations for Pairs of Identical Rhythms for Experiments 1, 2, and 3 Alternation condition Experiment 1 : AXAXAXAX Experiment 2: Metric Experiment 3: AAXX strength AAXXAAXX AAAAXXXX Rhythm Perfectly metric 96 84 97 5&9 1. X...X.X.X...X... 95 96 2. X...X...X.X.X... 86 5&9 92 97 3. X.X.X...X...X... 5&9 86 94 97 85 M Strongly metric or metric 88 98 4. X.X...X.X...X... 88 9 80 64 89 5 X XX X X... 9 80 92 6. X.X..X..X...X... 60 9 86 7. X..X..X.X...X... 9 82 8. X...X.X..X..X... 5 67 89 75 9 X X X X.X... 76 92 5 74 10. X...X..X.X..X... 5 75 93 11. X...X.X...X.X... 86 5 81 72 92 M Weekly metric 80 83 12. X.X...X..X..X... None 66 13. X.X..X...X..X... None 75 63 77 14. X..X...X.X..X... 74 71 84 None 83 15. X.X...X...X.X... None 78 91 16. X X. X..X..X... None 92 86 17. X..X..X...X.X... None 18. X..X.X...X..X... None 82 76 19. X..X...X..X.X... None M 81 70 84 Note. Empty cells indicate that a rhythm was not used in an experiment. "Indicates whether a tone occurred at Element 5, Element 9, both Elements 5 and 9 (5 & 9), or neither Element 5 nor Element 9 (None). Element 5 and the second rhythm having tones at both Elements 5 and 9 (e.g., X...X..X..X.X... and X...X...X.X.X...). In such a table, 5 of the 16 cells cannot occur because of the restriction that the two rhythms differ only in the position of one element. For example, this restriction eliminates a pair of rhythms in which one rhythm has a tone at Element 5 but not at Element 9 and the second rhythm has just the reverse. The pairs are shown in Table 2, those with the same figural organization appearing first followed by those with different figural organizations. Both types of pairs were placed in tables defined by the metric strength of the first and second rhythms. The rhythm-pairs with different figural organizations were divided into three groups based on the relative metric strength of the first and second rhythms: (a) The metric strength of the two rhythms was equal; (b) the metric strength of the second rhythm was stronger; or (c) the metric strength of the first rhythm was stronger. The percentages of different-rhythm (i.e., correct) judgments and the percentages of correct element identification are shown. All of the rhythm-pairs with identical figural organizations and representative instances of the rhythm-pairs with different figural organizations are shown (the numbers of pairs are indicated in parentheses following the instances, and the ranges of percentages correct are shown following the average values). Overall, there was a significant difference between pairs in terms of the percentage of correct discriminations, F(28,1568) = 20.0,;? < .005, MSB = 0.13, and the percentage of correct identifications, F(28, 1568) = 12.1, p < .005, MSE = 0.20. There was no interaction for either measure. First consider the same-different figural organization distinction. There was no overlap in discrimination accuracy between pairs of different rhythms with the same figural organization and those with different figural organizations (Tukey's HSD test). For the five pairs with identical figural organizations, the mean percentage of judgments that the two rhythms were different was 52%, nearly identical to chance performance, and the mean percentage of correct element identifications was 23%. Thus, even when participants did perceive the two rhythms as being different, they were unable to identify which tone had changed (chance performance being one third of the trials on which listeners detected a difference, or 17% in this instance). In contrast, for the 24 rhythm-pairs with different figural organizations, the mean percentage of judgments that the rhythms differed

1550

HANDEL

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 2 Experiment 1: Percentages of Correct Discriminations and of Correct Identifications of Which Tone Changed Position for Pairs of Different Rhythms Metric strength" First Second rhythm rhythm Rhythm-pan* Discrimination' Identification' Identical figural organizations None None X.X..X...X..X... 47 19 X.X...X..X..X... None 5 X..X...X..X.X... 20 50 X...X..X..X.X... None 22 X...X..X.X..X... 5 51 X..X...X.X..X... None X.X..X..X...X... 9 30 56 X.X..X...X..X... X..X.X..X...X... 24 54 X..X.X...X..X... M 23 52 Different figural organizations None None X.X...X...X.X... 51 (37-65) 83 (70-94) X..X..X...X.X... (6) 5 X...X..X.X..X... 5 55 (44-65) 81 (79-85) X...X.X..X..X... (3) 9 50 (38-61) X..X.X..X...X... 82 (76-91) 9 X..X..X.X...X... (6) M 52 83 None
5 9 5 5&9 5&9

X..X..X..X..X... X...X.X..X..X... (2) X...X..X..X.X... X...X...X.X.X... X..X..X.X...X... X...X.X.X...X... (2) M X..X..X.X...X... X..X X X X X...X.X.X...X... X...X.X..X..X... X...X.X.X...X... X..X..X.X...X... (2) M

87 (80-94)
85

36 (32-39)
41

87 (80-93)
86 93 94

42 (32-52)
39 56 60

9 5&9 5&9

None
5 9

88 (84-92)

58 (46-70)

58 92 'Indicates whether a tone occurred at Element 5, Element 9, both Elements 5 and 9 (5 & 9), or neither Element 5 nor Element 9 (None). Rhythm-pairs with two possible combinations of metric strengths are not shown because of sampling among all the rhythm-pairs. The two combinations are (a) 5 followed by None, and (b) None followed by 9. The number of instances for combinations of metric strength are shown in parentheses following a representative pair. The percentages of correct discriminations and identifications are the averages of all the possible pairs. The lowest and highest performances across the rhythm-pairs are shown in parentheses following the averages unless there was only one possible instance.

was 81% and the mean percentage of correct tone identifications was 48% (chance performance is 27%). Second, consider the metric strength analysis. What is clear is that performance did not differ among the combinations of metric strength for the two kinds of different

rhythm-pairs. For the rhythm-pairs with identical figural organizations, there were no differences between pairs in which both rhythms had a weak metric and pairs in which the first rhythm had a strong metric (a tone at Element 9). For the rhythm-pairs with different figural organizations,

METRIC AND FIGURAL RHYTHMS

1551

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

there were no differences that were due to the grouping according to relative metric strength that is shown in Table 2, F(2,21) = 2.8,p > .05. Moreover, there were no differences that were due to the overall metric strength at the following four levels: (a) none-none; (b) 5-none and none-9; (c) 5-5 and 9-9; and (d) 5-S&9, 9-S&9, 5&9-S, and S&9-9, F(3, 20) = 0.8. There was no simple pattern of outcomes among the cells. Consider the two cells that have the greatest number of rhythm-pairs. Both the average percentage and the range of percentages were nearly identical for (a) pairs in which both rhythms had the weakest metric structure (no tones at Elements 5 or 9) and (b) pairs in which both rhythms had a strong metric structure (tones at Element 9). There were differences between pairs of rhythms in the participants' ability to identify which tone shifted position. For some pairs of different rhythms, the identification of which tone had changed timing was relatively poor, although it was easy to discriminate the two rhythms. A typical example was the following pair of rhythms: X..X..X..X..X... followed by X...X.X..X..X... In this case, the two rhythms were judged as being different in nearly all of the trials (93%), but the choice of the second tone (underlined) as the one that had changed was at chance (32%). Instead, participants usually chose the third tone as the one that had moved (48%). What is characteristic is that in the first rhythm, there is an isolated tone (separated by two or more silent intervals on both sides) that shifts one position to form a double group in the second rhythm. Participants tend to perceive the second tone of the double group as having shifted, not the first tone. This result points out the primacy of the initial tone of the group; it is perceived as being stable and as defining the timing of the entire group.

Discussion
In sum, the results give little support to the notion that the metric strength or metric availability influences the discrimination between this class of rhythms. If the two rhythms of a pair had the same figural description, then participants discriminated the two rhythms only at the chance level, independent of the metric structure. If the two rhythms had different figural descriptions, then participants were able to discriminate them easily, but performance again was independent of the metric structure. The only evidence that the metric structure improved discrimination was found for pairs of identical rhythms: Rhythms with a perfect metric organization were more accurately perceived as being identical. It is possible to detail the relationship between the metric and figural organizations from two perspectives. The first perspective involves considering the effect of metric strength on the five pairs with identical figural organizations. The rules used to construct the rhythms impose constraints that limit the possible rhythms. These constraints yield only a small set of rhythm-pairs with identical figural organizations, and within this set, the two rhythms tend to have weaker metrics. In one case, both rhythms have a weak meter, and in four cases, the rhythms have different strengths. Performance was equivalent even though it might be ex-

pected that a change in meter would highlight differences in tuning (much the same as would a transition from tonal to atonal melodies). This difference in metric strength would be maximized in the last two pairs, in which the first rhythm had a strong meter, including a tone at the second strongest beat at Element 9, and the second rhythm did not. Yet discrimination was no better than chance. The second perspective involves considering the effect of figural organization on pairs with identical metric strengths. In all instances, there was no overlap in discrimination between pairs with identical figural organizations and pairs with different figural organizations. Consider the pairs with the weakest metrics: the none-none pairs. Discrimination for the pair with identical figural organizations was below chance, whereas discrimination for the six pairs with different figural organizations ranged from 70% to 94%. Now consider pairs with the stronger metrics: 9-none. In these pairs, the first rhythm had beats at Elements 1, 9, and 13. Nonetheless, discrimination for the two pairs with identical figural organizations was barely above chance (54% and 56%). In contrast, discrimination for the pair with different figural organizations was nearly perfect (93%). To summarize this argument, the difference in discrimination between pairs with identical and different figural organizations occurs equally for pairs with weaker metrics and for pairs with stronger metrics. Thus, the predominance of the figural organization is not limited to weaker metric rhythms. Even a quick perusal of the outcomes suggests that discrimination is affected by many factors. For example, discrimination involving Rhythm 16, X..X..X..X..X..., is quite good even though the four-beat metric is weak. Possibly, this rhythm is organized according to a three-beat metric despite the fact that such a meter would not split the rhythm evenly. In this case, the pairs containing Rhythm 16 should be considered highly metric. A reanalysis of the pairs of rhythms with different figural organizations in which those pairs containing Rhythm 16 were omitted did not change any of the outcomes: There was no difference between the groups defined by relative metric strength (in Table 2) or by overall metric strength as defined above. Thus, the evidence for a three-beat meter is equivocal. The rhythms might simply be organized according to the even spacing of the tones without a meter being induced at all (without a sense of stronger and weaker beats). There are several possible explanations of why these results did not show any effect of the metric structure. 1. The discrimination task may not have been sufficiently sensitive. The percentage correct for pairs with different figural organizations ranged from around 80% to 85%, and this may have represented a ceiling effect that masked any effects of the metric structure. Povel and Essens (1985) used a reproduction task, and Smith and Cuddy (1989) used a reaction time task; both of these may have allowed meter effects to occur. However, even the more sensitive measure used here, the identification of which tone changed timing, did not show any differences, and a ceiling effect is unlikely to have affected that result. 2. The metric structure may not have been apprehended strongly. There are two parts to this argument. First, the

1552

HANDEL

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

single or double alternation may have precluded the building up of the metric structure of either rhythm. These alternation conditions were chosen because they produced the best discrimination in previous work (Handel, 1992), but neither one gives the listener repeated "looks" at the possible organizations to induce the metric organization of each rhythm. Second, these are relatively hard rhythms to pick up. On the whole, none of them follows simple rules, so the meter may have been relatively hidden. The rhythms with the strongest metric structure are the simplest because of the tone locations, so it is difficult to tease out clearly the contribution of the metric structure. Taken together, the rapid alternations and the difficulty of the rhythms might argue that the metric structure did not affect discrimination because listeners could not pick up and make use of the meter. To counter these objections, in Experiment 2 I changed the methodology to try to enhance the perception of the metric structure.

Experiment 2
The goal in Experiment 2 was to investigate further the role of metric structure in the discrimination of rhythms. The results of Experiment 1 seemed to show that the effect of metric structure was minimal. Participants organized the rhythms according to the figural grouping structure, and the grouping structure, in turn, determined discrimination. In Experiment 2, I tried to bring the metric structure into perceptual prominence by making two changes in the methodology. The first change was to the alternation condition. In Experiment 2, the first rhythm was repeated four times, and then the second rhythm was repeated four times (i.e., AAAAXXXX). The rationale for this was that playing each rhythm four times in a row would allow the listener to pick up the meter more easily. It is possible that the single and double alternations used in Experiment 1 forced listeners to concentrate on the figural structure. The second change was to introduce a short-duration, low-pitch pulse tone to "time" the first rhythm, in much the same way that a percussion accompaniment occurs on the meter beat in a performance. Two approaches were used. In the first approach, the pulse occurred at Elements 1,5,9, and 13 of the first rhythm whether or not a rhythm tone occurred at those elements. Thus, it was possible to have a pulse without a tone. There were no pulse tones for the second rhythm. I did not use pulse tones for the second rhythm because I was afraid that participants would use the derived strategy of simply attending to the coincidence of pulse and tone in the two rhythms to determine if they were the same or different (pulse tones were used for both rhythms in Experiment 3). This condition is termed the meter-rhythm condition because the pulse timed the underlying meter of the first rhythm. In the second approach, the pulse occurred at Elements 1, 5, 9, and 13 only when a tone also fell on those elements. The purpose of this condition was to eliminate the possible confusion when a pulse occurred without a tone and to emphasize the link between the meter pulses and tones. Because tones always occurred on Ele-

ments 1 and 13, the difference between the two conditions could occur only at Elements 5 and 9. As in the meterrhythm condition, the pulse occurred only on the first rhythm. This condition is termed the accent-rhythm condition because the pulse accented the tones at the metric positions. Theoretically, the pulse could improve discrimination equally for strong and weak metric rhythms by providing a timing reference. Furthermore, this reference would exist whether or not the tones and pulses coincided. However, perceptually, it is probably easier to use such a reference when the pulses and tones coincide, as they do for the stronger metric rhythms, than it is when the pulses and tones fall on different grid elements, as they do for the weaker metric rhythms. In these latter cases, one might expect the effect of the pulse to be variable, depending on the exact timings of the pulses and tones. To summarize, the purpose of Experiment 2 was to determine whether repeating each rhythm four times and introducing a pulse would emphasize the metric structure sufficiently for the strength of the metrical organization to affect discrimination.

Method
Participants. All 61 participants were undergraduate volunteers at the University of Tennessee who received course credit for their participation. Different participants were used in each experiment. The participants were tested in small groups of from 1 to 3. Rhythms. The rhythms were the same type as those used in Experiment 1. A total of 27 pairs of rhythms were used. Thirteen pairs contained two identical rhythms. These were selected from the possible set of 19 (see Experiment 1) so that there was a roughly equal distribution of metric strength: Three pairs had tones at Elements 5 and 9, at Element 9, or at Element 5, and four pairs did not have tones at either of these metric elements. Fourteen pairs contained two different rhythms. Five of these pairs contained two different rhythms that had the same figural organization, and the remaining nine pairs had rhythms with different combinations of metric strength. All rhythms were used. Task. The task was the same as that used in Experiment 1. Presentation conditions. The rhythms were presented at the same rate that was used in Experiment 1, roughly 2.3 elements/s. The first rhythm was presented four times and then the second rhythm was presented four times. The timing of the alternation was the same as that used in Experiment 1. In contrast to Experiment 1, the first rhythm was presented with a lower pitch tone (400 Hz, triangle wave), and the second rhythm was presented with a higher pitch tone (600 Hz, triangle wave). The high and low pitch tones had a steady-state duration of 50 ms and 10-ms/lO-ms rise/fall times, as in Experiment 1. The pulse was a 100-Hz sine wave presented for 35 ms: 15 ms steady state and 10-ms/lO-ms rise/fall transients. All of the other conditions were identical to those in Experiment 1. Experimental conditions. There were three conditions. In the first, termed rhythm-rhythm, each rhythm was presented alone without an accompanying pulse. This condition was identical to the situation in Experiment 1 (with the exception of the alternation method) and served as a replication. In the second condition, termed meter-rhythm, the pulse tone occurred on Elements 1, 5, 9, and 13 for the four repetitions of the first rhythm. If there was a rhythm tone, the pulse and tone started synchronously, although the

METRIC AND FIGURAL RHYTHMS tone continued after the pulse. If there was not a rhythm tone, the pulse was heard alone. In the third condition, termed accentrhythm, the pulse occurred on Elements 1,5,9, and 13 of the first rhythm only if a tone occurred. The timing for the pulse and tone was identical to that in the meter-rhythm condition. Experimental design. The design was within subjects: 3 conditions X 27 patterns. The 27 patterns within a condition were presented in a single block, and the blocks were presented to the participants in the identical order: rhythm-rhythm, meter-rhythm, and accent-rhythm. This order was chosen to maximize the effect of the external pulse. The rationale was that the rhythm-rhythm condition would serve as a reference, allowing participants to become familiar with the rhythms. Following this, participants could make use of the pulse to perceive the timing structure of the rhythms without having to become familiar with the rhythms. The accent-rhythm condition was presented last because it seemed somewhat unrepresentative of natural rhythms and I did not want it to create negative transfer to either the rhythm-rhythm or meterrhythm conditions. For each block, three different sequences of rhythms were constructed that roughly counterbalanced the order of the pairs of rhythms. As in Experiment 1, before each block, participants were presented with four simple examples, and the first two trials in each block were replicated later within the block. The timing of the trials was the same as that used in Experiment 1, and there was no feedback. There were short breaks between the conditions, and the experimental session took about 50 min.

1553

Results The outcomes were analyzed according to the same strategy used in Experiment 1. The results for pairs of identical rhythms are considered first, followed by the results for the pairs of different rhythms. Preliminary analyses indicated that performance for the three orders of each condition was equivalent. For the rhythm-rhythm condition, the percentage correct ranged from 74% to 71%; for the meter-rhythm condition, it ranged from 72% to 70%; and for the accentrhythm condition, it ranged from 76% to 70%. On this basis, I combined the results for the three orders for each experimental condition. The percentages correct are based on 61 responses, one per participant. Pairs of identical rhythms. There were no differences among the three experimental conditions, F(2, 120) = 1.15, and there was no interaction between rhythm and condition, F(24, 1440) = 1.85. The percentages of correct judgments for the rhythm-rhythm, meter-rhythm, and accent-rhythm conditions were 76%, 73%, and 75%, respectively. Overall, there was a significant difference among the rhythm-pairs, F(12, 720) = 15.4, p < .001, MSB = 0.19. The best discrimination occurred for the three rhythms with the strongest metric structure (i.e., with tones at Elements 1, 5, 9, and 13) and for Rhythm 4. For these rhythms, the percentage correct averaged 86%. For the remaining pairs (9 in total), there were no differences as a function of experimental condition or metric strength. Here the average percentage correct was 69%. Thus, the significant effect occurred primarily between rhythms with the strongest metric and all the others. The percentages of correct judgments averaged across experimental conditions are shown in Table 1.

Pairs of different rhythms. For all parrs, the percentage of judgments that the two rhythms were different and the percentage of correct note identifications are shown in Table 3 as a function of condition and of the metric strength of the first and second rhythms. As in Experiment 1, pairs with identical and different figural organizations were separated. Overall, for both the percentage of judgments that the two rhythms were different and the percentage of judgments that correctly identified which element had changed timing, there were significant differences between rhythm-pairs, F(13, 780) = 17.5, p < .005, MSB = 0.21, and F(13, 780) = 9.7, p < .005, MSB = 0.22, respectively; no differences between conditions, Fs(2, 120) = 1.4 and 0.5, respectively; and a significant Rhythm-Pair X Condition interaction, F(26, 1560) = 3.1,p < .005, MSB = 0.15, andF(26,1560) = 2.7, p < .005, MSB = 0.15, respectively. The Rhythm X Condition interaction is the crux of the results. Consider first the rhythm-rhythm condition, equivalent to the situation in Experiment 1. Here there was no overlap in discrimination accuracy between rhythm-pairs with the same figural organization and rhythm-pairs with different figural organizations (Tukey's HSD test). There were no differences among the pairs with identical figural organizations and no differences among the parrs with different figural organizations. Neither the relative nor overall metric strength affected discrimination. For pairs with identical and different figural organizations, the percentages of correct discriminations were 50% and 84%, respectively, and the percentages of correct identifications of the tone that changed position were 19% and 41%, respectively. These percentages were nearly identical to those for the same pairs in Experiment 1. Consider next the meter-rhythm and accent-rhythm conditions. These results are quite different from those of the rhythm-rhythm condition, and there was a striking interaction between pairs with the same figural organization and pairs with different figural organizations as a function of the metric strength of the first and second rhythms. For the parrs with the same figural organization, there were two distinct outcomes. If the first rhythm had the weakest metric structure, with tones at Elements 1 and 13 but not at Elements 5 or 9, then discrimination was below chance. The percentages of correct "different" judgments for the meterrhythm and accent-rhythm conditions were 46% and 40%, respectively (this difference was not significant). However, if the first rhythm had a stronger metric structure, with a tone at Elements 5 or 9, then the pulse in the meter-rhythm and accent-rhythm conditions generated significantly better performance. The percentages of correct "different" judgments for the meter-rhythm and accent-rhythm conditions were 72% and 69%, respectively. For the pairs with different figural organizations, with two exceptions, discrimination was excellent. The percentage of correct "different" judgments was 76%, and the percentage of correct element identifications averaged 39%. As was found for the, rhythm-rhythm condition, there were no effects that were due to metric strength. Even excluding the two pairs with the poorest discrimination, the percentage correct was slightly higher for the rhythm-rhythm condition.

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1554

HANDEL

Table 3 Experiment 2: Percentages of Correct Discriminations (Discr.) and of Correct Identifications (Ident.) of Which Element Changed Position for Pairs of Different Rhythms Experiment condition Metric strength* First rhythm None
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Rhythm-rhythm Rhythm-pair X.X..X...X..X... X.X...X..X..X... X..X...X..X.X... X...X..X..X.X... Discr. Ident.

Meter-rhythm Discr. Ident.

Accent-rhythm Discr. Ident.

Second rhythm None

Identical figural organization

44 49 47 43 57 57 52 85 74 82 87 92 84 90 87 88 77 84 81

14 20 17 18 23 19 20 43 33 39 41 48 41 48 23 36 44 54 49

53 38 46 75 75 67 72 85 52 62 84 82 73 82 79 81 74 84 79

16 18 17 39 31 31 34 43 20 31 43 38 35 36 36 36 34 55 45

36 44 40 61 67 80 69 84 53 54 83 82 72 80 84 82 79 92 86

14 20 17 25 25 30 27 55 26 23 46 48 39 26 33 30 35 61 48

None

M
5 9
None None X...X..X.X..X... X..X...X.X..X... X.X..X..X...X... X.X..X...X..X... X..X.X..X...X... X..X.X...X..X...

M
None None X..X..X...X.X... X..X..X..X..X... X.X...X..X..X... X.X...X...X.X...

Different figural organization

5 9

5 9

X...X.X..X..X... X...X.X...X.X... X..X..X.X...X... X..X.X..X...X... X.X...X.X...X... X.X...X..X...X...

M
None

5 5&9

X..X..X..X..X... X...X.X..X..X... X...X..X..X.X... X...X...X.X.X...

M
5 5&9
None X...X.X...X.X... X..X..X...X.X... X...X.X.X...X... X..X..X.X...X...

"Indicates whether a tone occurred at Element 5, Element 9, both Elements 5 and 9 (5 & 9), or neither Element 5 nor Element 9 (None).

Thus, the pulse did not improve discrimination if the two rhythms had different figural organizations. There were no differences among the three conditions in the participants' ability to identify which tone had changed position. More important, the percentages of correct identifi-

<__

_i

*___!_

_^? *t.

strength in Table 3. There are large differences among the pairs. But, in general, participants were unable to identify more accurately that a tone had moved from a strong to a weak metric position than the reverse.

METRIC AND FIGURAL RHYTHMS

1555

Discussion
These results, completely consistent with those from Experiment 1, clarify the role of the metric structure in rhythm perception. The meter did improve discrimination for pairs of identical rhythms, but the effect was more complicated for pairs of different rhythms. To follow the form of the Discussion in Experiment 1, let us first consider the effect of metric strength on the five pairs of rhythms with the same figural organization. For the rhythm-rhythm pairs, performance hovered about chance even if the initial rhythm had tones at the strongest metric beats (at Elements 1 and 9). Thus, these results perfectly replicate those of Experiment 1. For the meter-rhythm and accent-rhythm pairs, the external pulse brought about a fixed temporal grid that enabled listeners to distinguish between two rhythms with the same grouping or figural organization but with different timings between the groups. However, the pulse improved discrimination only if the pulse coincided with the tones of the initial rhythm. The pulse made the metric structure of the initial rhythm more available so that the timings between adjacent groups could be encoded relative to the meter. However, when the pulse did not coincide with any of the three internal tones of the first rhythm (the first two pairs in Table 3), the pulse did not affect discrimination even if the second rhythm had a stronger meter. Now consider the effect of the figural organization on the pairs with the same metric strength. For the rhythm-rhythm pairs, the results replicate those in Experiment 1. For all comparisons, discrimination for pairs with different figural organizations was well above chance, in contrast to the chance performance for all pairs with identical figural organizations. For the meter-rhythm and accent-rhythm pairs, the results differed because there was overlap among pairs with different and identical figural organizations. This asymmetry, in which a transition from stronger to weaker structure produces better performance than does a transition from weaker to stronger structure, has been found in other kinds of auditory processing. For example, Jones and Boltz (1989) found that a hierarchic (i.e., stronger meter) to nonhierarchic (i.e., weaker meter) transition yielded better performance than the opposite sequence. Bharucha and Pryor (1986) found a similar asymmetry in temporal discrimination. Krumhansl, Bharucha, and Castellano (1982) showed that harmonic discrimination was better if the first melody was strongly tonal and the second atonal than the reverse, and Bartlett (1993) demonstrated that listeners could detect changes in melodic contour more easily if the tonal contour preceded the atonal contour. Jones and Boltz (1989) argued that all of these kinds of results can be subsumed under the concept of expectancy. The initial melody, rhythm, or sequence of tones creates a trajectory about subsequent events. Stronger meters (or more tonal melodies) more tightly constrain the range of expectancies so that rhythmic or tonality deviations can be more easily perceived (expectancy is roughly analogous to Garner's, 1974, notion of inferred sets). Nonetheless, it seems reasonable that the structure of the second sequence, by inducing a set of expectancies or possible alternatives,

ought to affect discrimination. But that was not the case here. There are two possible reasons: (a) The second rhythm was being directly compared with the first rhythm without being encoded, or (b) the fact that there was no pulse for the second rhythm made the metric strength less prominent for the listener.

Experiment 3
What is known at this point is that the metric strength influences discrimination in only one context: if the two different rhythms have the same figural organization, if the first rhythm has the stronger inherent meter, and if the metric elements of the first rhythm are marked by a pulse. I designed Experiment 3 to investigate two issues further. The first issue is why the effects of the metric structure are so limited. One possibility is that because the pulse never occurred for the second rhythm, the effect of the pulse in the first rhythm was weakened and any possible effect of the metric strength of the second rhythm was negated. For this reason, in one condition in Experiment 3 I had the pulse occur on elements 1, 5, 9, and 13 for both rhythms (termed the meter-meter condition) in order to continue the metric structure across the second rhythm. These outcomes can be compared with those when the pulse occurred for the first rhythm only (meter-rhythm) or not at all (rhythm-rhythm). (Although participants usually could determine if the two rhythms were different for the meter-meter condition by simply determining if the coincidences of pulses and tones were identical in the two rhythms, no participant explicitly reported using this derived strategy). The second issue concerns the generality of the asymmetry in outcomes that is due to the relative metric strengths of the two rhythms when there is a pulse. I argued for this asymmetry in Experiment 2 by comparing across rhythmpairs. For example, discrimination was at chance for Pair A, in which the meter was weak-strong, but discrimination was very good for Pair B, in which the meter was strong-weak. Because any pair of rhythms have unique timing characteristics, it is impossible to argue unambiguously that a strong-toweak meter transition always yields better discrimination. For this reason, each pair of different rhythms with the same figural organization was presented twice, the second time with the order of presentation of the two rhythms reversed. If there is a general asymmetry created by the interaction of relative metric strength and pulse, then discrimination of each combination of rhythms should be better if the rhythm with the stronger metric is presented first than if the rhythm with the weaker metric is presented first. In pairs in which both rhythms have equal metric strength, there should be little difference between the two orders of the rhythms.

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Method
Participants. All 67 participants were undergraduates at the University of Tennessee who received course credit for their participation. They were tested in groups of from 1 to 3. Rhythms. The rhythms were the same type as those used in Experiments 1 and 2. There were 36 pairs of rhythms in all. Thirteen pairs had two identical rhythms; these were the same pairs

1556

HANDEL

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

used in Experiment 2. Twenty-three pairs had two different rhythms. These can be broken into two sets. The first set consisted of the five pairs of different rhythms used previously that had the same figural organization. Each pair was presented in the two possible orders to generate a total of 10 pairs. The second set consisted of 13 pairs that had different figural organizations. These pairs were chosen to match the metric strengths of the pairs with the identical figural organizations. Task. The task was simpler than the one used in Experiments 1 and 2. In Experiment 3 participants judged whether the two rhythms were identical or different on a 4-point scale (1 = very sure identical, 2 = fairly sure identical, 3 = fairly sure different, and 4 = very sure different). In deriving the percentage correct, I considered judgments of 1 and 2 to be "identical" judgments and 3 and 4 to be "different" judgments. I switched the response in order to maximize performance so that any possible improvement that was due to the meter could be found, and I felt that trying to identify the tone that changed position could be distracting. Presentation conditions. The rhythms were presented at the same rate used previously, roughly 2.3 elements/s. The first rhythm was presented two times, and then, without a break, the second rhythm was presented two times. This alternation condition was chosen to make the task more difficult so that any improvement that was due to the meter would not be obscured by a ceiling effect. The rhythms were generated with the MIDI software package MIDILAB for the IBM PC. The sounds were generated with a Seiko DS-250 keyboard, and the piano timbre was used. The first rhythm was presented with a lower pitch tone (440 Hz), and the second rhythm was presented with a higher pitch tone (660 Hz). Each tone was 75 ms in duration. The pulse was a 99-Hz tone presented for 40ms. Experimental design. The design was within subjects: 3 conditions X 36 pairs of rhythms. The rhythms were presented in three blocks according to condition and across subjects; the order of presentation of the conditions was counterbalanced. For each condition, the rhythms were placed into four sequences that counterbalanced order. The procedures were the same ones used in Experiments 1 and 2. There were short breaks between the three conditions, and the experimental session took about 75 min.

Results
Preliminary results indicated that there were no differences that were due to the sequence order within each condition. The maximum difference in the percentage correct between the sequences for any condition was 11%. Moreover, there were no differences in percentage correct that were due to the order of the blocks. The percentages correct for Blocks 1, 2, and 3 were 77%, 82%, and 82%, respectively. On this basis, the results were averaged over sequences and order. The percentages correct are based on 67 responses, one per participant. Pairs of identical rhythms. There was a small significant difference among conditions, F(2, 132) = 3.9, p < .025, MSB = 0.09, a significant Rhythm X Condition interaction, F(24, 1584) = 2.6, p < .005, MSB = 0.07, and a significant difference among rhythm pairs, F(12, 792) = 9.8, p < .005, MSB = 0.08. For the three rhythms in which tones occurred at all four meter elements, discrimination was nearly perfect. The percentages correct for the rhythm-rhythm, meterrhythm, and meter-meter conditions were 99%, 95%, and 96%, respectively. Discrimination was equivalent for the six rhythms in which tones occurred at Elements 1,5, and 13 or

at Elements 1,9, and 13. (The only exception was Rhythm 4: The percentage correct equaled that for rhythms with the strongest meter). The percentages correct for the rhythmrhythm, meter-rhythm, and meter-meter conditions were 89%, 91%, and 96%, respectively. Discrimination was poorer for the four rhythms that had tones only at Elements 1 and 13. The percentages correct for the rhythm-rhythm, meter-rhythm, and meter-meter conditions were 88%, 78%, and 84%, respectively. Over all pairs, the differences among conditions were small, averaging 2%. For this reason, the percentage correct for each rhythm was combined across conditions and is shown in Table 1. On the whole, discrimination for the pairs of identical rhythms mirrors that in the first two experiments, although the percentage correct is higher in Experiment 3. What is common across the three experiments is that it was easier to identify two strongly metrical rhythms as being identical. What differs in Experiment 3 from Experiments 1 and 2 is the hint of an interaction between metric strength and pulse condition. Discrimination was identical across the various pairs of identical rhythms for the rhythm-rhythm condition without any external pulse, but discrimination was more difficult (i.e., participants were more likely to hear the two rhythms as being different) for the less metrical rhythms for conditions with an external pulse. Pairs of different rhythms. For these pairs, there were a significant rhythm effect, F(22, 1452) = 33.0, p < .005, MSB = 0.22, a condition effect, F(2, 132) = 4.1, p < .025, MSB = 0.21, and a Rhythm X Condition interaction, F(44, 2904) = 4.3, p < .005, MSB = 0.15. The Rhythm X Condition interaction is the main focus of the results (as in Experiment 2). This interaction can be understood from three perspectives. First, consider the overall difference between the 10 pairs with identical figural organizations and the 13 pairs with different figural organizations. As found in Experiments 1 and 2, there were large differences in the percentage of correct discriminations for the rhythm-rhythm and meter-rhythm conditions, F(l, 21) = 58.0, p < .001, and F(l, 21) = 6.5, p < .02, respectively. Discrimination was at the chance level for pairs with identical figural organizations (52%) and well above chance for pairs with different figural organizations (77%). These percentages are nearly identical to those found in Experiment 2. In contrast, there was no difference in discrimination for the meter-meter condition because discrimination for the parrs with identical figural organizations improved sharply (63%) compared with discrimination for pairs with different figural organizations (77%),F(1,21) = 2.5,/.05. Second, consider the 10 pairs of different rhythms with the same figural organizations. For the two pairs in which the two rhythms had equal metric strengths (la and Ib in Table 4) and the four pairs in which the rhythm with the weaker metric preceded the rhythm with the stronger metric (2a5a), discrimination was at the chance level or lower and was essentially equal for the three conditions. The only exceptions occurred for Pairs Ib and 5a for the meter-meter condition. In contrast, if the stronger metric rhythm preceded the weaker one (2b-5b), discrimination was much

METRIC AND FIGURAL RHYTHMS

1557

Table 4 Experiment 3: Percentages of Correct Discriminations for Pairs of Different Rhythms


Metric strength" First rhythm None Second rhythm None Pair no. (la) Rhythm-pair X.X..X...X..X... X.X...X..X..X... X.X...X..X..X... X.X..X...X..X... Experimental condition Rhythm-rhythm Meter-rhythm Meter-meter

Identical figural organization

41 50 46 45 33 38 36 38
56 57

32 63 48 19 29 41 40 33
70 75

49 72 61 32 26 62 65 46
80 69

(Ib)

M
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

None

(2a) (3a)

X..X...X.X..X... X...X..X.X..X... X..X...X..X.X... X...X..X..X.X... X.X..X...X..X... X.X..X..X...X... X..X.X...X..X... X..X.X..X...X...

None

(4a) (5a)

None

(2b)

X...X..X.X..X... X..X...X.X..X... X...X..X..X.X... X..X...X..X.X... X.X..X..X...X... X.X..X...X..X... X..X.X..X...X... X..X.X...X..X...

(3b)

None

(4b)

53 65 58

77 75 74 47
85 72

92 77 80 45
69 91

(5b)

M
None None (6)
(7)

Different figural organization X.X...X..X..X... X.X...X...X.X... X..X..X...X.X... X..X..X..X..X...


X. X X X X . X...X.X..X..X...

73
93 91

None

(11) (12a)

X..X..X...X.X... X...X.X...X.X... X.X...X..X..X... X.X...X.X...X...


X X X X X X..X..X.X...X...

85
86 91

77
77 88

84
83 82

None

(13a) (14a)

88 74 91
96 87

80 73 93
95 87

85 87 96
97 93
(table continues)

5 9

None None

(12b) (13b)

X...X.X...X.X... X..X..X...X.X... X.X...X.X...X... X.X...X..X..X... X..X..X.X...X... X X X X X...

(14b)

1558
Table 4
First rhythm

HANDEL

(continued)
Second rhythm Pair no. Experimental condition Rhythm-pair Rhythm-rhythm Meter-rhythm Meter-meter Different figural organization

Metric strength3

9
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

. . . X . X . . X . . X . . . ( 8 a ) X X . . . X . X . . . X . X . . . (8b) X . . . X . X . . . X . X . . . X . . . X . X . . X . . X . . . (9) X . X . . . X . X . . . X . . . X . X . . X . . X . . . X . . . . . X . . X . X . . . X . . . (10) X X . . X . X . . X . . . X . . . M

58 71 81 77 76

42 61 62 70 61

41 66
71

65 60

"Indicates whether a tone occurred at Element 5, Element 9, both Elements 5 and 9 (5 & 9), or neither Element 5 nor Element 9 (None).

better and there were large differences between the conditions. For the rhythm-rhythm condition, discrimination improved from 38% to 58% although performance was above chance only for Pair 5b. For the meter-rhythm and meter-meter conditions, the improvement was much more dramatic (from 33% to 74% and from 46% to 80%, respectively) and discrimination of all four pairs was above chance. These results are shown in Table 4. The discrimination performance for the four pairs going from a weaker to a stronger metric breaks into two classes. Performance was better for the meter-rhythm condition and considerably better for the meter-meter condition for Pairs 4a and 5a than for Pairs 2a and 3a. Probably, in the two easier pairs, the tone that changed moved to Element 9, which created a stronger metric than in the two other cases in which the tone moved to FJement 5, which created a weaker metric. There was no such difference for the rhythm-rhythm condition, which demonstrates the potential usefulness of the meter if the figural organization does not discriminate two different rhythms. The difference in discrimination between the meter-rhythm and meter-meter conditions for Pairs 4a and 5a illustrates that a pulse that accompanies the stronger meter of the second rhythm can improve discrimination. Third, consider the 13 pairs in which the two rhythms had different figural organizations. For the rhythm-rhythm condition, discrimination for all pairs (with the exception of Pair 8a) was equivalent. There was no effect of metric strength. Moreover, as described above, discrimination for all of these pairs was better than that for pairs with the identical figural organization. For the meter-rhythm and meter-meter conditions, discrimination was far more variable, and the effect of metric strength was inconsistent. Discrimination was poorest when the two rhythms in a pair had the identical metric strength even if both rhythms had strong meters (Pairs 9 and 10). Discrimination improved when the two rhythms had different metric strengths and was maximum when the stronger metric was the first rhythm of the pair. Two rhythm-pairs were significantly more difficult than the others (Pairs 6 and 8a), and performance on them equaled that for pairs with identical figural organizations. These two

pairs were the same ones that produced poorer discrimination in Experiment 2. The identical pattern of results was found even for the best 25% of the participants. These participants, although they performed more accurately across all the pairs, also were unable to discriminate Pairs la-5a for the rhythm-rhythm and meter-rhythm conditions and Pairs la-3a for the meter-meter condition.

Discussion
These results support the contention that for these rhythms and tasks, the metric structure operates within the figural or grouping structure. If the figural organizations of two rhythms are different, then adding an external pulse will not improve and may even degrade discrimination. However, the metric structure can help listeners discriminate two different rhythms with identical figural organizations. But the metric structure cannot operate "backward" in time. Unless the timing of the tones of the initial rhythm fits the meter (i.e., tones occur synchronously with the pulse at beat elements), discrimination will not improve. In fact, if the tones of the initial rhythm do not fit the meter, the pulse will not act as a temporal grid and will make discrimination more difficult. Thus the effect of an external meter pulse is context dependent.

General Discussion
The results of these three experiments help to delimit the role of the metric structure. The notion of a meter underlying the perception of every rhythm, or that metric rhythms are prototypical, is seductive, but that notion may overstate the importance of a meter in discrimination. For the rhythms used in these three experiments, listeners appear to initially place the tones into groups based on the shortest interval, and those groups form the basis of the figural organization. If two rhythms have different figural organizations, either in terms of the number of tones in each group or in terms of the order among the groups (e.g., 1-3-2 as opposed to 3-1-2), then listeners easily perceive that the

METRIC AND FIGURAL RHYTHMS

1559

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

two rhythms are different. If the figural organizations are the same, then listeners must use the timing between the onsets of the initial elements of each group to discriminate the rhythms. It is here that the meter can provide a grid to time the intervals between successive groups. What is surprising is the relative difficulty listeners have in making use of the possible grids. When these rhythms were presented without a pulse, there was little evidence that the occurrence of tones at the metric elements affected discrimination. The strength of the inherent meter produced by the timing of the tones did not create a grid that allowed participants to compare timings across groups. When these rhythms were presented with a pulse, the pulse improved discrimination only if the initial rhythm was strongly metric so that the pulse occurred synchronously with tones. Otherwise, the pulse did not improve discrimination and actually impaired it, even though it should logically have instituted the same underlying metric and timing relationships. In sum, in certain instances the meter can be used to distinguish among rhythms with identical figural organizations by creating a temporal grid, in the same way that melodic tonality can be used to distinguish among equivalent melodic contours by creating a harmonic template or that facial features can be used to distinguish among equivalent profiles by creating expressive prototypes.

Generality of Results
Comparisons across experiments. One relevant issue concerns whether the results reported here can be attributed to general experimental design effects, response biases, or both. For example, Yee, Holleran, and Jones (1994) reported that performance for highly skilled participants changed as a function of the types of rhythms included within an experimental session. It is unlikely that such design effects could explain these results even though the three experiments used a variety of presentation conditions, including different response modes, pulse conditions, and alternation conditions, and used a different set of participants in each experiment. I argue this on two grounds. First, performance was remarkably consistent across experiments, particularly for the critical different pairs with identical figural organizations. Consider the five pairs that occurred in all three experiments, which are the first five pairs shown in Table 2. The range in the percentage of correct discriminations for these pairs for the rhythm-rhythm condition across the three experiments is 6%, 5%, 13%, 4%, and 11%. The range in the percentage of correct discriminations for these pairs for the meter-rhythm condition between Experiments 2 and 3 is 21%, 19%, 5%, 2%, and 8%. Second, the use of alternative pulse conditions (meter-rhythm and meter-meter) led to new response patterns for those conditions without changing the responses to the other conditions within Experiments 2 and 3. The meter-rhythm condition (in Experiment 2) yielded better discrimination for strong-to-weak meter pairs but did not change the discrimination for the rhythm-rhythm condition. The meter-meter condition (in Experiment 3) yielded better discrimination for pairs in which the second rhythm had a tone at Element 9 (Pairs 4a and 5a in Table 3)

but did not affect discrimination for the rhythm-rhythm and meter-rhythm conditions. It is also unlikely that response biases affected the outcomes. In each of the three experiments there was a higher percentage of different pairs (60%, 52%, and 64% for Experiments 1,2, and 3, respectively) than of identical pairs. Thus, purely on the basis of frequency, participants should have responded "different" if they were uncertain. But participants judged different pairs with identical figural organizations as identical, going against the probability expectancy. I believe that participants' initial hypothesis in this task is that the two rhythms are identical and that they judge them as different only when they perceive a timing change. Thus, when the participants do not pick up the variation in the between-groups timing for two different rhythms with identical figural organizations, they judge the rhythms as being the same. Alternative meters. Participants may be using alternative meters in perceiving these rhythms. In that case, the analyses based on four-beat meters would not be relevant for many pairs. Rhythm 16 (in Table 1) could be organized according to a three-beat meter because the tones fall on Elements 1, 4, 7, 10, and 13. Similarly, Rhythms 7 and 17 begin with tones on Elements 1,4, and 7. In all three of these rhythms, however, the interval between the onset of the final tone at Element 13 and the onset of the next repetition at Element 1 is four elements, so a strict three-beat meter would progressively fall out of phase with the rhythmic tones across repetitions. This difference in timing is quite noticeable (399 ms vs. 532 ms). Nonetheless, to avoid the possibility that the classification of Rhythm 16 as weakly metric biased the results, I redid all of the analyses of pairs of different rhythms in Experiments 2 and 3 and omitted pairs involving Rhythm 16. The results did not change for any analysis. However, there is some evidence, particularly in Experiment 3, that pairs of different rhythms involving Rhythm 16 were easier to discriminate. Overall, the percentage of correct discriminations for pairs involving Rhythm 16 was 88%, whereas the percentage of correct discriminations for the other different pairs with different figural organizations averaged 72%. What makes the conclusion that participants were using a three-beat meter comparison problematic is that Rhythm 16 has a unique figural organization of five individual elements and differs from all of the other rhythms, which have at least one group of two or more elements. For this reason, the discrimination could be made without any perception of meter at all. Does feedback affect discrimination? It is possible that discrimination would have been better, particularly for pairs of different rhythms with the same figural organization, if participants had been given feedback. This feedback would have allowed the participants to learn to attend to the relevant timing information between groups. Although I could not provide feedback for each trial because of the available equipment, I attempted to test this possibility with a fourth experiment in which I provided intensive pretraining on the rhythms. Before participants began a block of trials for one condition, four different kinds of pairs of rhythms were demonstrated and explained: (a) pairs of identical rhythms that participants misjudged rarely; (b)

1560

HANDEL

pairs of identical rhythms that participants misjudged sometimes (discrimination usually was good for identical rhythms); (c) pairs of different rhythms that participants misjudged rarely; and (d) pairs of different rhythms with the same figural organization that participants judged correctly at less than chance. For each kind of rhythm, several examples were presented auditorily while participants looked at a visual representation of the rhythms indicating which element had shifted (similar to the representations shown in the four tables). After the participants stated that they could hear that the rhythms were identical or different and could explain the visual representations, the experimental trials were started. To keep the experiment simple, I used only the rhythm-rhythm and meter-meter conditions. The orders of the rhythm-pairs and conditions were counterbalanced, and the presentation conditions were identical to those in Experiment 3. A total of 40 undergraduate students participated. In spite of the extensive pretraining, discrimination was identical to that in Experiment 3. The most important results concern the pairs of different rhythms with the same figural organizations. For the meter-meter condition, if the rhythm with the stronger meter preceded the one with the weaker meter, the percentage of correct discriminations was 87%, but the percentage of correct discriminations for the reversed pair was only 40%. (The analogous percentages from Experiment 3 were 80% and 46%). For the rhythm-rhythm condition, if the rhythm with the stronger meter preceded the one with the weaker meter, the percentage of correct discriminations was 64%, but the percentage of correct discriminations for the reversed pair of rhythms was 40%. (The analogous percentages from Experiment 3 were 58% and 38%). Thus, although the pretraining may have improved discrimination slightly for the stronger-to-weaker metric pairs, it did not improve performance for the more difficult weaker-to-stronger metric pairs. Overall, the pattern of results was identical to that in Experiment 3. Specifically, the pulse improved discrimination only if the stronger metric rhythm was the first (i.e., presumably the reference) rhythm. Otherwise, discrimination for the meter-meter and rhythmrhythm conditions was equivalent and remained below chance for rhythms with the identical figural organization. Is the figural organization always predominant? In this context, the figural organization determined whether two rhythms were perceived as identical or different. These results are contrary to the belief that the meter underlies rhythmic perception and that the figural organization represents a weaker or less organized percept. There are several possible reasons for the present outcome. First, the participants were unselected and probably relatively untrained, and metric organization improves with musical training (e.g., Yee et al., 1994). Second, the discrimination task itself might have minimized the usefulness of the meter, and the alternation between rhythms might have made the meter difficult to perceive. Third, the pairs of different rhythms with identical figural organizations tend to have weaker meters. Rhythms with stronger meters might have allowed participants to discriminate between two different rhythms with identical figural organizations. Fourth, the intervals between the onsets of adjacent tones are based on only one unit of 133 ms, and the ratios between different onset

intervals were 1:2:3, which do not readily fit any simple meter. Any one or any combination of these possibilities could account for the weak effect of metric strength. Nonetheless, I do not believe that they undercut the conclusion that the figural organization is primary. A meter can resolve the temporal relationships among the figural groups, but the meter could not exist without the figural groups. The meter is an emergent property of the auditory grouping in the same way that symmetry is an emergent property of visual grouping.

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Is a Concept of Meter Necessary?


The concept of a meter is deeply ingrained in music theory and empirical research. Nearly every theorist argues for the multilevel nature of a regular beat and the interplay between the metric and grouping structure (Yeston, 1976). Moreover, the most influential recent treatment of the subject, by Lerdahl and Jackendoff (1983), argued that the metric and grouping structures are independent. Similarly, there are extensive findings in both motor and music research that there is a basic temporal unit that is used to create all of the intervals. Fraisse (1982) summarized research demonstrating that when participants reproduce intervals between three or four tones, the intervals are systematically distorted so that they approach one of two values: a short interval that is roughly one half the length of a long interval. Povel (1981) and Povel and Essens (1985) argued that listeners attempt to create an internal clock (i.e., a meter) when listening to a rhythm. Rhythms that are easily encoded are those in which accented tones fall at equal intervals, and listeners are assumed to search for a clock that "hits" the accented tones. In addition, Jones and coworkers (Jones, 1993) have argued for a model of rhythmic attending in which the presence of a regular pattern of strong and weak beats allows the listener to extrapolate the passage and identify the important structural elements. Finally, Collier and Wright (1995) suggested that there are innate preferences for specific rhythmic ratios and that musicians can scale these ratios faster or slower. But even experienced musicians find it difficult to learn complex ratios, and none could scale these complex ratios. What this all means is mat there is a consensus that rhythmic production is based on a temporal unit (or possibly two independent units) and that this unit is subdivided or multiplied to produce the various intervals in a passage. Using this unit as a reference, performers can stretch or contract durations to produce the desired artistic effect (Clarke, 1985). The problem, then, is to reconcile the overwhelming evidence for the importance of a meter in reproduction tasks to the relatively small effects found for a meter in the present discrimination tasks. One way is to make use of ideas from information theory that Garner (1974) used to explain differences in recall and discrimination tasks. All theorists have explicitly or implicitly argued that the meter simplifies rhythms by making them temporally predictable (and thus, in Garner's terms, making the inferred set smaller). Thus, there is a simple contingency between the beats and the tones that fall on the beats. The interval between strong beats

METRIC AND FIGURAL RHYTHMS

1561

is always the same, and this allows the listener to extrapolate the beat and tonal rhythm into ttie future (cf. Jones & Yee, 1993). This regularity allows the person producing a rhythm to preplan a repetitive motor response for the strong beats and to attend to the more irregular intermediate weaker beats. The memory load is reduced, and therefore we would expect that reproduction of metric rhythms would be more accurate than reproduction of nonmetric rhythms. However, the simple contingency among the stronger beats of metric rhythms makes them less differentiated, and therefore we would expect discrimination among metric and nonmetric rhythms to be equivalent because of two competing factors. On the one hand, it is easier to encode metric rhythms because of the simple contingencies, and that should improve discrimination. On the other hand, the simple contingencies make the metric rhythms more similar, and that should make discrimination more difficult. In these experiments, the pulse improved discrimination, presumably by allowing participants to more efficiently encode the metric rhythm so that it could be better differentiated from the following nonmetric rhythm. This argument points out once again the contextual nature of rhythmic organization: The usefulness of any organization depends on the specifics of the task, and different concepts may be necessary as the task varies. From this perspective, the better reproduction of metric rhythms occurs not because they are metric, but because they are simpler, and "simpler" would be the preferred explanation because it is applicable in a broader domain. There is an inevitable confound here in that metric rhythms are necessarily simpler because the strong beats occur at equal temporal intervals and are therefore more easily predicted. I do not believe that this is merely a semantic issue because it goes to the heart of whether rhythms are simply one type of pattern. If this is the case, then concepts applicable to a broad cross-section of visual and auditory perceptual phenomena, such as regularity, symmetry, good continuation, common fate, and similarity, might suffice to explain rhythm perception without the need for rhythmspecific concepts. Although I have argued previously that both visual and auditory events are perceived within a hierarchical spacetime framework and that it is possible to form many equivalences for these sorts of concepts (Handel, 1988), the ubiquitous theme-plus-variation structure of music tends to make the concept of periodicity (i.e., a metric beat) synonymous with repeatability, and that need not be true for the visual world. We do need a concept of meter (if only to emphasize the motoric and affective components of rhythm), but the effect of the metric regularity can be understood only within the context of the listening task, and different tasks will require different concepts.

References
Bamberger, J. (1978). Intuitive and formal musical knowing: Parables of cognitive dissonance. In S. S. Madeja (Ed.), The arts, cognition and basic skills (pp. 173-209). St Louis, MO: CEMREL. Bartlett, J. C. (1993). Tonal structure of melodies. In T. J. Tighe &

W. J. Dowling (Eds.), Psychology and music (pp. 39-65). Hillsdale, NJ: Erlbaum. Bharucha, J. J., & Pryor, J. H. (1986). Disrupting the isochrony underlying rhythm: An asymmetry in discrimination. Perception & Psychophysics, 40, 137-141. Bolton, T. L. (1894). Rhythm. American Journal of Psychology, 6, 145-238. Clarke, E. F. (1985). Some aspects of rhythm and expression in Erik Satie's "Gnossienne No. 5." Music Perception, 2, 299-328. ColUer, G. L., & Wright, C. E. (1995). Temporal rescaling of simple and complex ratios in rhythmic tapping. Journal of Experimental Psychology: Human Perception and Performance, 21, 607-627. Essens, P. J., & Povel, D. J. (1985). Metrical and nonmetrical representations of temporal patterns. Perception & Psychophysics, 37, 1-7. Fraisse, P. (1982). Rhythm and tempo. In D. Deutsch (Ed.), The psychology of music (pp. 149-180). New York: Academic Press. Garner, W. R. (1974). The processing of information and structure. Potomac, MD: Erlbaum. Handel, S. (1974). Perceiving melodic and rhythmic patterns. Journal of Experimental Psychology, 103, 922-933. Handel, S. (1988). Space is to time as vision is to audition: Seductive but misleading. Journal of Experimental Psychology: Human Perception and Performance, 14, 315-317. Handel, S. (1992). The differentiation of rhythmic structure. Perception & Psychophysics, 52, 497-507. Jones, M. R. (1993). Dynamics of musical patterns: How do melody and rhythm fit together? In T. J. Tighe & W. J. Dowling (Eds.), Psychology and music (pp. 67-92). Hillsdale, NJ: Erlbaum. Jones, M. R., & Boltz, M. (1989). Dynamic attending and responses to time. Psychological Review, 96, 459-491. Jones, M. R., & Yee, W. (1993). Attending to auditory events: The role of temporal organization. In S. MeAdams & E. Bigand (Eds.), Thinking in sound (pp. 69-112). Oxford, England: Oxford University Press. Krumhansl, C. L., Bharucha, J. J., & Castellano, M. (1982). Key distance effects on perceived harmonic structure in music. Perception & Psychophysics, 31, 75-85. Lerdahl, P., & Jackendoff, R. (1983). A generative theory of tonal music. Cambridge, MA: MIT Press. Palmer, C., & Krumhansl, C. L. (1990). Mental representations for musical meter. Journal of Experimental Psychology: Human Perception and Performance, 16, 728-741. Parncutt, R. (1994). A perceptual model of pulse salience and metrical accent in musical rhythms. Music Perception, 11, 409-464. Povel, D.-J. (1981). Internal representation of simple temporal patterns. Journal of Experimental Psychology: Human Perception and Performance, 7, 3-18. Povel, D.-J., & Essens, P. (1985). Perception of temporal patterns. Music Perception, 3, 411-440. Smith, K. C., & Cuddy, L. L. (1989). Effects of metric and harmonic rhythm on the detection of pitch alterations in melodic sequences. Journal of Experimental Psychology: Human Perception and Performance, 15,457-471. Yee, W., Holleran, S., & Jones, M. R. (1994). Sensitivity to event timing in regular and irregular sequences: Influences of musical skill. Perception & Psychophysics, 56, 461-471. Yeston, M. (1976). The stratification of musical rhythm. New Haven, CT: Yale University Press. Received August 20,1996 Revision received August 8, 1997 Accepted September 15, 1997

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Das könnte Ihnen auch gefallen