(e-mail: lapidaki@mus.auth.gr)

ABSTRACT This study was designed to determine whether listeners from different age groups and musical backgrounds (musicians and nonmusicians) could set tempi in a consistent manner over an extended period of time. The variables of musical style, familiarity, and preference were also considered. Subjects (n=90) heard the same six compositions on four separate occasions. The order of the presentation and the initial tempo of the examples were varied systematically in each session. Subjects were asked to listen to each composition and indicate whether the experimenter should set the tempo "faster" or "slower" until it sounded right to them; they had to adjust an initially wrong tempo to a personally preferred tempo.

Results indicated that the initial tempo significantly dominated subjects’ "correct" tempo judgements: the slower initial tempo generally evoked slower tempo selections, and so on. However, a relatively small number of adults, mostly musicians, were remarkably consistent in their tempo judgements across all four trials. It appeared that these individuals possess an exceptional ability with respect to acute stability of large-scale timing in music.

There was also evidence that the degree of consistency in correct tempo judgements gradually increased from preadolescence through adulthood. Few statistically significant differences in consistency of tempo judgements were found as a result of musical background. The findings strongly suggested that the style of musical examples influenced the degree of tempo consistency across trials. Moreover, there was statistically significant evidence that an increase of familiarity with and preference for the musical examples and the musical styles resulted in an increase of consistency of correct tempo judgements. The study concludes with recommendations for music education.

The question "Do listeners possess a concept of correct tempo for a piece of music, and if so, is this concept consistent?" is the main theme of this study. But there is also a second question, more difficult to answer, that underlines its intent: "Do listeners possess a time sense―an ability for the proper pacing of time―which enables them to render stable tempo judgements over a period of time?" The literature on stability of tempo perception is far from consistent on these questions.

Most experimental investigations concerning consistency of tempo choices were performed around the first half of this century (Frisheisen-Köhler, 1933; Harrison, 1941; Miles, 1937; Mishima, 1956; Rimoldi, 1951; Wallin, 1911). Most commonly, subjects had to tap their responses on a telegraph key or to listen to metronomic clicks proceeding at different rates, and were asked which tempo was felt to be more appropriate. Generally, findings showed that subjects tended to be relatively consistent in their preferred rates.

More recently, two studies by Clynes and Walker (1982, 1986) on temporal stability in musical performance are worth noting. Repeated musical performance by the same musicians and of the same compositions were timed over a number years. The research findings suggested a high degree of consistency and precision in the execution of musical tempo. The researchers reasoned that music appeared to engage and program a psychobiologic clock or clocks which functioned subconsciously, but gave conscious read-outs, and thereby seemed to guide the performers' realisation of musical tempo in an exact and stable manner. These findings are consistent with the timing of a symphony orchestra in several performances of the same compositions over several years at different music halls of the world measured by Winckel (1962). Similarly, Wagner’s (1974) timing of different performances on the piano of the same piece by Herbert von Karajan showed highly consistent tempi.

With respect to this point about tempo stability in music performance, Epstein (1985) claimed:

In addition to tempo studies that employed listening to stimuli like metronome clicks, tapping tasks or music performance, of particular interest were those investigations (Behne, 1972; Farnsworth, Block, and Waterman, 1934; Halpern, 1988; Lapidaki & Webster, 1991; and Levitin & Cook, 1996) that asked the listener to make judgements about the correct tempo of musical examples with mechanical or electronic devices (e.g., Duo-Art player piano with a tempo lever, the Springer-machine, or the software interface on computers) that allowed for variable tempo control over the musical stimulus.

More specifically, Farnsworth et al. (1934) and Halpern (1988) investigated the consistency between the "perceived" and the "imagined" correct tempo of stimuli. The perceived correct tempo was set by manipulating the tempo lever of a player piano or the computer interface, respectively, while the stimuli were played in real time. The imagined tempo was demonstrated by tapping in the Farnsworth et al. study, and by setting the metronome to coincide with what the subject imagined to be the correct tempo, in the Halpern study. The findings of both experiments reported relative consistency between the perceived and the imagined correct tempo. Both sets of results seem to indicate that there is one tempo consistently associated with familiar listening examples.

Furthermore, Levitin & Cook (1996) conducted a similar study in order to investigate if we remember a song in its original tempo. 46 college students unselected for their musical background were asked to choose a song they knew very well among fifty-eight CDs containing the best known popular songs and to hold it in their hands. They were instructed to close their eyes and imagine that the song was actually playing. Then they were told to try to reproduce the song from memory by singing, humming, or whistling. After the first reproduction subjects were instructed to repeat the procedure with another song of their choice. The tempo of each subjects’ reproduction was then compared with the actual tempo of the excerpt on the CDs. Results showed that long term memory for tempi of well-learned songs is very accurate.

These studies were important because of their use of real musical stimuli with hardware that allowed subjects to have control over tempi. Nevertheless, they were limited in that they merely investigated well-known tunes, such as ballroom music (Farnsworth et al.) or familiar popular songs (Levitin & Cook, Halpern). Moreover, these experiments were conducted solely with nonmusic major students.

Interesting as their results may be, they do not demonstrate whether judgements of correct tempo are consistent across separate trials over an extended period of time, especially when subjects are presented with musical compositions chosen because they represent a wide range of musical styles and familiarity. It also seemed important to investigate how tempo judgements might differ among subjects with different musical backgrounds.

To investigate these issues, Lapidaki & Webster (1991) conducted a study in which subjects were 15 highly experienced musicians (5 composers, 5 performers, and 5 music education specialists) recruited from a pool of professors and graduate students of a School of Music in the Midwestern United States and 5 nonmusicians who were professors and graduate students from other departments of the university and had little formal music education and involvement in musical activities. Three music examples (e.g., J. S. Bach’s "Air in D Major" from the Suite Number 3 in D major; F. Chopin’s Prelude Number 7, Op. 28, and A. Schoenberg’s second piece from "6 kleine Stücke," op. 19) were chosen because they represented a wide range of musical styles and familiarity. All subjects were tested individually at three sessions at three-day intervals. For each of the three testing sessions, subjects were asked to make correct tempo judgements of each of the three compositions. The initial tempo of the presentation of the compositions was systematically in each session.

The findings of Lapidaki & Webster’s study (1991) showed that when tempo is judged by highly skilled musicians in repeated listening tasks of the same compositions, initial tempo has a dominant effect on correct tempo judgements. Simply stated, no single correct tempo emerged as a consistent entity of individual or group performance across the three trials. The sample of adult nonmusicians indicated a basis for a similar conclusion. Nevertheless, this tended to vary according to the composition in question. These results did not support the observations reported by Farnsworth, et. al. (1934), Halpern (1988), and Levitin & Cook (1996) that one tempo is consistently associated with particular listening examples. On the contrary, listeners’ perceptions of correct tempo for a particular composition varied dramatically from trial to trial. Few statistically significant differences in consistency of tempo judgements were found as a result of musical background and compositional style. Many of these tendencies suggested important questions for further study.

It was obvious, however, that additional work was necessary with larger and more varied musical samples and with better measures of individual familiarity with, and preference for judged compositions. Also of interest would be how these judgements may differ among subjects from different age groups and musical background.

The Study

The present study therefore followed this up by investigating the consistency of tempo judgements over a period of time observed in individuals during the listening process. More specifically, the purpose of the study was to examine the consistency of "correct" tempo judgements of compositions of various musical styles among subjects with differing musical background, age, familiarity with, and preference for selected music. Correct tempo is seen here as a subjective unifying construct of music in the listener’s mind whose function is the meaningful synthesis of finite juxtaposed musical elements, such as harmonic as well as rhythmic structure, melody, texture, dissonance, form, and ornamentation, among other things, in real time (Barry, 1990). It should be noted that the study was about the extent to which individuals can set consistent tempi across four separate trials: no attempt was made to establish whether or not these tempi were correct as compared with those set by the composers in the original pieces. Along these lines, it was reasoned that if a correct tempo did exist, subjects ought to be able to arrive at consistent judgements about the tempo of examples despite the examples being presented with differing initial tempi in every session.

The majority of empirical studies on tempo perception have been carried out on adults (Farnsworth et. al., 1934; Halpern, 1988; Hodgson, 1951; Lapidaki & Webster, 1991; Levitin & Cook, 1996; Lund, 1939). However, there is general agreement that the experience of musical time is not separable from the subjects’ age (Bamberger, 1994; Petzold, 1963; Shuter-Dyson & Gabriel, 1981; Zenatti, 1993). To counter this deficiency, it has proved necessary to investigate the following question: Is the capacity for consistent tempo judgements for particular pieces of music affected by the age of listeners (e.g., preadolescents, adolescents, and adults)? Once the age question has been answered, it might be then possible to set varied music educational standards for each age level by considering the often overlooked development of temporal perception in students and, in turn, create a more effective condition for the growth of musical experience.

Furthermore, the capability to perceive different musical parameters, such as tonality, harmony, form, and rhythm, without being able to identify and analyse them, is considered to be the outgrowth of implicit musical knowledge or acculturation (Hargreaves, 1986; Francès, 1958/1988; Bigand, 1993). In other words, in this situation what listeners know is not something they are aware of knowing, but rather it is acquired from knowledge that is implicitly or subconsciously built into their auditory systems through common everyday exposure to music in their cultural environment. There is general agreement among researchers, on the other hand, that this knowledge becomes explicit or conscious only after musical training (Dowling, 1993). In essence, musicians presumably possess a fuller understanding and appreciation of a piece of music, due in part to their ability to possess a sophisticated scheme or set of rules for encoding its musical events in terms of musical meanings and, thus, to assign to it a stable structural description (Sloboda, 1994; Dowling, 1994; Wolpert, 1990; Lerdahl & Jackendoff, 1983). The study was therefore concerned whether the musical background of listeners, that is, the level of formal music education and/or participation in specialised musical activities, affected the consistency in the perception of the correct tempo.

Another issue that this study investigated was whether the stylistic framework in terms of which listeners interpreted the specific musical examples influenced tempo consistency. According to Rosen (1972) the musical elements, among which tempo appears to take an important part, that make up the synthesis of a work of music, " . . . cannot be divorced from its affective and sentimental, as well its intellectual, significance within the work and, consequently, more generally within the stylistic language," (p. 21). With respect to this point about the significance of style on perception, Barry (1990) pointed out:

It was also assumed that familiarity with the musical examples and/or their respective musical styles or genres might affect consistency of judgements concerning the correct tempo of the stimuli. Familiarity with the relevant musical idiom or musical piece gained from prior formal or informal exposure to music contributes to the listener's ability to extract meaningful criteria that help him/her to make sense of incoming musical stimuli (Barry, 1990). By drawing from a known musical context or idiom and/or the same or similar musical experiences from the past the listener can organise the musical information in melodic, motivic-thematic, harmonic, and temporal structures characteristic of a musical idiom. According to Lerdahl & Jackendoff (1983), once a listener "becomes familiar with the idiom, the kind of organisation that he attributes to a given piece will not be arbitrary, but will be highly constrained in specific ways" (p. 3). Furthermore, there is ample evidence that children seem to show culturally favoured responses to music with age, because they spontaneously or subconsciously become familiar with musical sounds heard in the social environment to which they belong (Farnsworth, 1958; Francès, 1988; Hargreaves, 1988; Imberty, 1981; Zenatti, 1993). Geringer & Madsen (1987) also found a marked influence of excerpt familiarity on listeners.

While there has been a large number of studies on the effect of tempo on listening preferences, inquiries on the nature of tempo perception have failed to take adequate account of the effect of experimental stimulus familiarity. One reason for this paucity seems to be that most auditory stimuli used in tempo research consist of musical material which are not selected so as to vary widely in their likely familiarity to the subjects (Farnsworth, Block, & Waterman, 1924; Halpern, 1988; Levitin & Cook, 199; Sheldon, 1994).

Finally it seemed worth considering whether musical preference might influence consistency of tempo judgements. The musical stimuli therefore were selected so as to vary widely as to their preference to the listeners. The term "preference" is seen here to reflect an individual's relatively transitory liking for one musical piece as compared with another (Hargreaves, 1988; Konecni, 1982; Zenatti, 1993). For instance, it is likely that listeners of the same age group with similar musical training and cultural background may not elicit judgements of a composition's correct tempo in a uniform way or with the same degree of over time consistency. Because of their individual preference of likes and dislikes in music, some listeners will tend to have a more positive attitude to some musical excerpts than others during the listening process with correspondingly higher degree of attention to and concentration on the organisation of musical events. For that reason, they may want to explore the preferred excerpts further or be more eager to arrive at a judgement (LeBlanc, 1987). As Barry (1990) noted:

Moreover, the issue of preference seemed to be particularly crucial for the present study, since it uses large-scale authentic musical pieces as stimuli. According to Barry (1990), "individual preference plays a much larger and more important part in substantial sections or complete musical works than in the very short, stylistically neutral material used in many experiments in musical perception" (p. 21). Finally, it is worth noting that no other study in tempo perception appears to speak directly to the question of preference. As Hargreaves (1988) rightly pointed out: "When preferred styles are employed in experiments alongside the ‘high art’ works that are the most common stimuli in psychological research on the arts, the results may be quite different" (p. 58).

Six specific questions were posed:

(1) Is there a consistent judgement of correct tempo across four separate sessions of the same musical examples using varying initial tempi for each trial?

(2) Is the consistency of tempo judgement affected by the age of the listener?

(3) Is the consistency of tempo judgement affected by the musical background of the listener?

(4) Is the consistency of tempo judgement affected by the style (or genre) of music?

(5) Is the perception of tempo affected by

a) the familiarity or unfamiliarity (novelty) with the individual pieces and

b) the overall style the music?

(6) Is the consistency of tempo judgement affected by the listener’s preference for a particular musical example?

Six music examples were used: C-major and A-minor Two-Part Inventions (around 1720) by J. S. Bach (Baroque), Clair de Lune from the "Suite Bergamasque" (1905) by Claude Debussy (Impressionistic), the Piano Piece (1993) by Michalis Lapidakis (contemporary idiom), Yesterday (1965) by John Lennon and Paul McCartney (Rock ballad), and The Children of Piraeus by Manos Hadjidakis (1960), title song for the film "Never on Sunday" by Jules Dassin, featuring Melina Mercouri. (dance music).

The musical examples used in the present study traverse almost three centuries, from 1720 to the present. It was felt that the listening material should reflect the various levels of flexibility set by composers with regard to the interpretation and perception of tempi. Nevertheless, due to the time constraints for each session, it appeared essential to limit the number of the stylistically diverse listening musical examples. Based on my experience with using listening examples in the classroom, all pieces were approximately from one to two minutes in duration. The length of the musical examples appeared to be long enough to provide a more comprehensive interaction with the music, yet short enough to keep subjects’ attention.

All examples were chosen because they represented a wide range of style, familiarity, and preference. For the sake of uniformity, all musical examples were chosen from the piano repertoire, excluding The Children of Piraeus and Yesterday which were originally composed for voice and instrumental ensemble. Nevertheless, both pieces have been scored, performed or recorded in piano versions. Only complete compositions were used as listening examples, so that listeners could have a sense of actual musical context, in contrast with stylistically neutral, trivial or fragmented auditory material used in many studies in music perception.

In view of the four individual meetings for each subject and the volume of data that would be generated, the maximum number of participants in the study was set at 90 (n=90). Thirty were adults (28-52 years), 30 were adolescents (16-18 years old), and 30 were preadolescents (10-12 years old). Thus, there were three developmental stages or corresponding age groups, with 30 subjects per group. Within each age group, half the subjects were musicians, half were non-musicians.

The adults (n=30) used in the study were not sampled, bur rather recruited on the basis of a set of criteria: age (28-52 years) and musical background. With regard to musical background, the group of adult musicians represented a wide array of significant experience in music. It consisted of 15 highly creative North American and Greek individuals from a pool of musicians who have remarkable career profiles. In addition, all subjects, except four of them, held a graduate degree in music from a European or a North American university.

The 15 adult North American and Greek nonmusicians were professionals in fields other than music; all held a graduate degree in different disciplines. These individuals had no training in music, but did have wide ranging backgrounds in music listening consistent with what might be expected for well-educated adults.

The adolescents (n=30) who participated in the study were junior and senior students (16-18 years) enrolled in a public high school in the Midwestern United States. In selecting the public high school for the study, a school was sought that met the following characteristics: (a) the music program must offer sufficient opportunities to experience music through performance; (b) the music program must incorporate a variety of musical media, such as choral groups, orchestra, bands, electronic and acoustic instruments, and ensembles with mixed instrumentation; and (c) the music program must motivate and prepare students for successful future professional involvement with music. The high school that met the characteristics cited above more than adequately is located in a northern suburb of Chicago and serves a generally affluent community. The makeup of the student body is approximately 82 percent Caucasian and 18 percent minority groups.

After consultation with the director of the high school’s music program and orchestra conductor, fifteen junior and senior students were randomly chosen to participate in the study from a pool of 25 orchestra players who had instrumental training and/or participation in a musical ensemble as members or soloists for more than six years.

With regard to the adolescent nonmusicians, a chemistry teacher and an English teacher at the same high school were asked to select junior and senior students from their classes who had never participated in the school’s music program or did not benefit from previous private music lessons and/or involvement in any kind of musical activity. A total of forty-six students met the above criteria to comprise a pool from which 15 students were randomly selected.

The preadolescents (n=30) who participated in the study were fifth and sixth grade students (10-12 years) recruited from a pool of students at an independent Greek school in the Chicago metropolitan area.

Like other ethnic groups in the United States, Greek-Americans demonstrate a strong connection with their language, heritage, and culture which is being transmitted to the younger generations through the Greek schools. These are private educational establishments that children attend twice a week for approximately 3-4 hours each session, in addition to their attendance at public or private American elementary schools. In selecting this particular school for the study, an educational environment on the elementary level was sought that met the following conditions: (1) all students were comparable in ethnic background, socio-economic milieu, and community traditions; (2) the school director, the teachers, the parents, and the students were willing to support the study on the basis of scheduling and accessibility of private testing space.

In accordance with the district’s request and Northwestern University’s guidelines for research with human subjects, a consent form was sent home to parents of all students in the two fifth-grade classes and the sixth-grade class. The letter informed parents about the study and asked them to indicate on the accompanying permission slip whether they would be interested in having their child participate. Positive responses for each class ranged from 20 to 27, for a total of 71 responses. The three schoolteachers reviewed these responses for students who were systematically taking music lessons and/or were regularly involved in a musical activity for at least three years. Fifteen such students from both the fifth- and sixth-grades were identified by the schoolteachers, building a total pool of 15 preadolescent musicians. To arrive at a total of 15 preadolescent nonmusicians (e.g., students with no musical training), 5 such students were randomly selected from each of the 3 classes.

Musical examples were performed on a YAMAHA P-150 digital piano connected to a Macintosh computer that registered performance data in MIDI files format by means of a sequencing program.

For the playback of the performance data―stored in MIDI files in the microcomputer―the MIDI sound generator Roland Sound Canvas SC-55 was used which was connected via an Opcode MIDI Translator to a Macintosh computer and a Peavey KB15 amplification/speaker system. "Piano 1" was the timbre of sound output and "Room 1" the sharply defined reverberation effect that simulates the natural echo of a room with a broad spread. Besides its realistic quality of piano sound, the Sound Canvas was used because of its compact size which allowed for its easy transportation to the various testing sites.

The software program employed for both recording and playback of performance data was the professional MIDI sequencing program Performer from Mark of the Unicorn. This program was chosen in large part because of its ability to alter the graphic window display on the computer screen so that the metronome controls could be easily manipulated. In addition, the program had the capacity to vary the tempo precisely, without altering any other musical attributes (e.g., pitch, timbre, articulation, etc.).

The tempo of each musical example―that is, the initial tempo―could be easily set by the experimenter prior to each session of each musical example. The mouse was used by the experimenter to manipulate the tempo, following the explicit directions of each subject. Set in manual tempo mode, the tempo slider of the graphic window display on the Macintosh was used to display and change the tempo in real time in the metronome window. To change tempo, the experimenter dragged the triangular indicator along the slider: to the left decreased the tempo, to the right increased it. The experimenter could also use the arrows at either end of the slider: the + (plus) arrow increases the tempo and the – (minus) arrow decreases it. Subjects were not asked to use the mouse themselves, since to do so would have required training for a number of subjects.

Each subject was tested individually by the experimenter (one of the authors) in a private room, isolated from outside noise. All adults were tested at the experimenter’s home music studio. The high school and elementary school students were tested in rooms made available to the experimenter in their respective schools.

In all sites the testing set-up was identically arranged to ensure continuity. The computer was placed on a table facing the experimenter and shielded from the subject’s view, with the sound generator and the amplification/speaker system mounted a comfortable distance away. It was felt that the subject ought not to view the tempo displayed on the computer screen while listening to the musical examples in order to minimise distractions in the decision making process.

For the four testing sessions, subjects were asked to listen to each composition and indicate whether the experimenter should set the tempo "faster" or "slower" until it sounded right to them. One might put it this way: the listeners had to reconstruct the musical examples examining the influence of tempo on the music until they came to the decision of the correct tempo. Each subject was encouraged to take as much time as was needed and to listen to the composition as often as necessary.

The following instructions were given to each of the subjects by the experimenter at the beginning of each piece in all four testing sessions: "I want you to help me find the tempo which you feel most comfortable with for this particular piece. There is no trick involved or right or wrong answer. If you want the tempo to increase, say ‘faster’ and, if you want it to decrease, say ‘slower.’ Then, when you have found the "right" tempo, that is the most appropriate tempo for the piece in your opinion, tell me. There is no time restriction. Are you ready to begin?"

The experimenter changed the tempo as directed until the subject was satisfied. When the subject indicated that the tempo sounded right, the experimenter recorded the metronome value, and then reset the computer for the next composition in the session. Once the six compositions were judged, the subject was asked to return within at least four days time for the next session. This slow pacing of trials was observed in order to prohibit memory carryover from one trial to another.

Each session for each subject systematically varied the order of the compositions and the initial tempo of the listening examples in order to eliminate the possibility of contextual cues. Two initial tempi were used: M.M.= 20 and M.M.= 200; all tempo judgements in the Lapidaki & Webster study (1991) had lain within this range. Each initial tempo was repeated twice: either in the first and third or in the second and fourth trials. Table 1 illustrates the order and the initial tempi in which the six examples were presented to subjects in the four testing sessions.

All judgements were recorded by noting the beats per minute, usually referenced as the metronome marking (M.M.). These values were used as continuous scales for the statistical analyses.

In order to examine subjects’ familiarity with the listening examples a questionnaire form was handed to them at the beginning of the first testing session. Subjects had to answer questions concerning their familiarity with the particular example and its relevant musical style, after they judged the correct tempo of the example. Subjects had to rank their familiarity with each piece on a scale of five levels which were based on the times they had heard the musical examples prior to the first trial (e.g., familiarity level 1 = Never, 2 = Once, 3 = 2 to 5 times, 4 = 6 to 10 times, and 5 = more than 10 times). Subjects had to rank their familiarity with each style on a scale of three levels (e.g., familiarity level 1 = I just do not know this style at all; 2 = I have heard music of this style, but not very often; and 3 = I really know this style of music).

Finally, with regard to the question of their individual preference for a particular musical example, subjects were asked to rate it on a scale ranging from 1 (least-liked or poor) to 4 (most-liked or excellent), after they judged the correct tempo of the example at the fourth testing session. This information was recorded and used in later analyses.

To test the hypothesis that listeners would make consistent judgements of tempo, independently from the initial tempi, a one-way repeated measures ANOVA for each musical example was performed using tempo judgements across all subject groups at each of the four trials as the independent variable. The results for these analyses show that listeners did not exhibit significant consistency in their judgements of the most appropriate tempo of the musical examples across the four trials (Bach I, F=84.43, p < .0001; Bach II, F=86.27, p < .0001; Debussy, F=80.37, p < .0001; Lapidakis, F=139.07, p < .0001; Beatles, F=59.02, p < .0001; Greek dance, F=78.856, p < .0001).

Further examination of the average means across all subject groups for all trials of each musical example revealed that both means of tempo judgements for the trials with the fast initial tempi were higher than the means for the trials with the slow initial tempi with respect to all musical examples (see Tables 2a & 2b). In other words, there was a clear relationship between the initial tempi and listeners’ final choices of most-appropriate tempo: the slower initial tempo generally evoked slower preferences, and so on.

Given this finding, it seemed appropriate to investigate differences between listeners’ tempo judgements observed at trials with the same initial tempo (e.g., differences between trials 1 & 3 and trials 2 & 4). An independent samples paired t-test was performed for each musical example. The results did not show evidence that repetition of the fast initial tempo evoked more consistent judgements than the repetition of the slow initial tempo or the contrary.

To test the effect of age on the consistency of tempo judgements, the most appropriate way would be to perform a one-way repeated measures ANOVA for each listening example using tempo judgements of trials and age as variables. However, the sphericity assumptions for the univariate test for an age effect were clearly violated (p value=0.0000) .. Therefore, either a repeated measures MANOVA procedure or the adjusted univariate test had to be utilised for each musical example. Both analyses led to the same conclusion, namely that consistency of tempo judgements across the four trials for all examples were significantly influenced by the age group of the listener (p < .02).

Furthermore, in order to ascertain which age group exhibited the highest degree of consistency, the individual deviation scores (IDS) averaged over the four trials of each piece were used as an additional measurement of tempo judgement consistency for each musical example. IDS reflects the standard deviation of the four different tempo judgements (Y1, Y2, Y3, and Y4) at the four trials for an individual. .IDS gives a more global sense for the deviations of each group. IDS was used as a primary response variable to answer questions about consistency associated with other factors of interest such as age, musical background, familiarity, and preference.

As shown in Table 3, results clearly indicated that adults were the most consistent and preadolescents the most inconsistent with regard to all musical examples (p < .001). In other words, the following consistency scale for all musical examples was observed with respect to the three age groups: preadolescents < adolescent < adults.

To examine the effect of musical background, a repeated measures MANOVA for each musical example was employed using tempo judgements of each trial and musical background as variables. The reason that the regular repeated measures univariate analysis of variance (ANOVA) was not performed was that data did not exhibit sphericity. The results showed that the musical background of the listener did not significantly affect consistency of tempo judgements for all six pieces (Bach I, F=0.79, p < .01; Bach II, F=0.73, p < .01; Lapidakis, F=0.73, p < .01; Beatles, F=0.47, p < .01; Greek dance, F=0.97, p < .0001), with the exception of the Debussy composition (F=4.00, p < .01).

Furthermore, to investigate whether musicians were more consistent than non-musicians, an independent samples t-test was performed which used IDSs among the trials of each piece as an additional measurement of tempo judgement consistency (see Table 4). The results clearly indicated that the only time musicians and non-musicians differed in consistency of tempo judgements was in Debussy (p < .01).

In order to answer the research question which deals with the effect of the musical style of the listening examples on the consistency of tempo judgements across the four trials, a repeated measures ANOVA procedure using style as the experimental factor (five levels corresponding to the five different musical styles of the pieces) and the mean number of IDS averaged over the four trials of each style as the response variable was performed. The results revealed that the style of rock ballad exhibited the highest degree of consistency (M=23.27, SD=22.54) followed by the styles of Greek dance music (M=30.90, SD=25.02), Impressionism (M=35.510, SD=26.29), and Baroque (M=36.51, SD=29.53 (Bach I, M=36.53 & Bach II, M=36.49)), respectively (F=13.68, p < .0001). The tempo judgements for the contemporary idiom were the less consistent among all styles (M=52.55, SD=31.56). In other words, the following consistency scale with respect to the musical styles was observed in subjects’ tempo judgements: Rock ballad < Greek dance music < Impressionism < Baroque < Contemporary idiom.

In order to investigate the effect of familiarity level on tempo judgements for each musical example, a repeated measures MANOVA was performed using tempo judgements for each example averaged over the four trials and the 5 familiarity levels based on the times they had heard the musical examples prior to the first trial (e.g., familiarity level 1 = Never, 2 = Once, 3 = 2 to 5 times, 4 = 6 to 10 times, and 5 = more than 10 times) as variables. The results indicated that familiarity with musical examples significantly influenced tempo judgements (p < .001).

In order to investigate the effect of familiarity level on tempo judgements for each musical example, a subsequent repeated measures ANOVA was performed using familiarity levels as the experimental factor and the mean number of IDSs averaged over the four trials of each musical example as the response variable (see Table 5). The results showed that an increase of familiarity had a significant effect on consistency of tempo judgements for all musical examples (Familiarity level 1, M=46.51, SD=28.71; Familiarity level 2, M=42.802, SD=27.80; Familiarity level 3, M=34.13, SD=28.04, Familiarity level 4, M=24.13, SD=18.56; Familiarity level 5, M=13.13, SD=17.64, F=25.17, p < .05). Concerning the Lapidakis composition, the familiarity effect could not be estimated, since there was only one level of familiarity among all subjects (e.g., familiarity level = 1).

In order to answer the question whether familiarity with the musical styles that the listening examples represented had an effect on subjects’ consistency of tempo judgements, all subjects had to rank their familiarity with each style on a scale of three levels (e.g., familiarity level 1 = I just do not know this style at all; 2 = I have heard music of this style, but not very often; and 3 = I really know this style of music). A repeated measures ANOVA was employed using IDSs after averaging over trials and familiarity levels as variables. The results showed that an increase of familiarity with the style of the musical examples resulted in an increase of tempo judgement consistency (overall F=43.38, p < .0001).

With respect to each individual style, the ANOVA tests of significance demonstrated that increased familiarity with the styles of Baroque (F=17.23, p< .0001), Impressionism (F=11.57, p< .0001), Rock ballad (F=40.22, p< .0001), and Greek dance music (F=7.92, p< .0001) significantly influenced the degree of consistency of tempo judgements for the particular examples that represented these styles. However, greater familiarity with the contemporary idiom did not lead into more consistent choices of most appropriate tempo for the Piano Piece by Lapidakis (F=0.99, p< .0001).

Preference

To investigate the research question concerning the effect of preference for the individual listening examples on consistency of tempo judgements, all subjects had to rank their preference for each example on a scale of four levels (e.g., preference level 1 = Poor, 2 = Good, 3 = Very good, and 4 = Excellent). A repeated measures ANOVA procedure was performed using preference levels as the experimental factor and the mean number of IDSs averaged over the four trials of each example as the response variable. The results revealed that there was a significant influence of the degree of preference on consistency, with the exception of the Greek dance (F=0.01, p < .001), Lapidakis (F=2.29, p < .001), and the Beatles (F=0.28, p < .001). In other word, with regard to Bach I (F=6.07, p < .001), Bach II (F=9.20, p < .001), and Debussy (F=15.81, p < .001), consistency clearly increased with preference.

The results show little evidence to support the claim that correct tempo judgements are consistent in the face of initial tempo changes. It is quite clear that when tempo is judged in repeated listening tasks of the same compositions, initial tempo of stimuli has a dominant effect on judgements of correct tempo: the slower initial tempo generally evokes slower tempo selections, and so on.

However, it should be stressed, in this respect, that a relatively small number of adults, mostly musicians, were remarkably consistent in their tempo judgements across all four trials. It appears that these individuals possess an exceptional ability with respect to acute stability of large-scale timing in music. This ability to give over time consistent tempo judgements to a piece of music in conditions seemingly devoid of an external tempo reference (a score or the body interaction involved in performance) may be referred to as absolute tempo, analogous to absolute pitch.

Interestingly enough, these subjects reported that they were surprised when they heard that their right tempo choices were virtually precise across trials. In addition, it seems that physical, psychological, and environmental factors, such as, fatigue, mood or time of day, do not have an effect on the ability to make consistent tempo judgements. Another proposal is that absolute tempo is a form of "implicit cognition" that has been the subject of increasing interest and debate in the most recent psychological research (Dorfman, Shames & Kihlstrom, 1996, p. 259; Underwood & Bright, 1996). The term implicit cognition has been used to characterise situations in which mental processes can influence perception outside of phenomenal awareness and voluntary control. For example, implicit cognition can refer to a subject’s behaviour that is shown to be influenced by stimulus events such as tempo changes that are too subtle to be consciously perceived (Lapidaki, 1990). With regards to consistency of correct tempo judgements, Epstein (1995) summarised this position as follows: "Ultimately decisions about rightful tempo rest upon intuition. Intuition is not absolute, however: it, too, can modify, our perception of music enriched by concepts structural, affective, historical, neural" (p. 107).

Finally, absolute tempo has been noticed with regard to musical examples with which listeners were familiar. Nevertheless, taken in isolation, this result should be interpreted with caution for a number of reasons. The major one is that these subjects did not exhibit absolute tempo with respect to all pieces with which they had the same level of familiarity. Contrary to absolute pitch, one might suppose with respect to absolute tempo that the same person seems to follow different cognitive strategies of timing for each individual piece of music which leaves one wondering whether the stability in viewpoint is to some extent discrete more than continuous.

Age

There is evidence that the degree of consistency in correct tempo judgements gradually increases from preadolescence through adulthood. This should lead to the assumption that fluctuations among tempo judgements of the same pieces of music become smaller, as awareness of tempo is attained with age. The interesting aspect of such an assumption is that tempo awareness does not appear to "plateau" after the age of 10, as is the case with the perception of other musical parameters (Imberty, 1981).

More specifically, when listeners between the ages of 10-12 are asked to choose the right tempo of a piece of music, they tend to demonstrate an awareness of tempo which does not reflect simultaneous perceptual organisation and integration of other relevant musical features, such as harmonic, rhythmic, and metric relationships. Therefore, preadolescents are experimenting with their correct tempo selections with greater ease, which leads to greater tempo fluctuations across trials. In adolescence, tempo fluctuations become smaller. Finally, by adulthood, tempo perception is rather accurate, and thus tempo consistency greater. It is therefore likely, though not essentially certain, that only when the representation of music has stabilised in the mind through maturation, is the corresponding temporal consistency noticeable.

The musical background of the subjects may have some bearing on consistency of tempo judgements, but this study reveals no overall statistically significant results to support this contention. In fact, it seems that the musician’s grasp of tempo behaviour is initially very much like the non-musician’s in the listening task in question. A possible explanation―not a parsimonious one, but compatible with neuropsychological data as well as with Davidson’s conclusions (1994) about song singing―is that ability to make consistent tempo judgements in listening tasks is probably a maturational one that is independent of music education. Nonetheless, this explanation is not necessarily true at high levels. Actually, results clearly indicated that adult musicians showed the smallest fluctuations in their tempo judgements and, thus, the highest degree of consistency.

Indeed, it must be acknowledged that musicians and nonmusicians of all age groups significantly differed in consistency of correct tempo judgements with regard to Clair de Lune by Claude Debussy. It should be noted that this particular piece is the only case among the musical examples in which one notices a slow overall tempo and rapid movements within this slow tempo at the same time. Thus, tempo becomes a function not only of the beat rate but also of the number of unfolding events per unit of time.

This leads to the speculation that musically experienced listeners may use an integrative cognitive strategy that enables them to control and synthesise local-level temporal relationships into the large-scale, complex (yet hierarchic), and coherent musical structure when they chose the right tempo of a piece of music. It is therefore crucial to entertain the possibility that musical training increases performance in listening tasks by alerting listeners’ perceptual organisation of stimuli, especially when the regularity of stable musical patterns breaks down.

The findings strongly suggest that the style of the musical examples influenced the degree of tempo consistency across trials. Indeed, it becomes obvious that individual structural characteristics of the styles facilitate listener’s extraction criteria of determining the correct tempo of the particular musical examples. As far as the specific styles of the listening examples are concerned, the style of rock ballad exhibited the highest degree of consistency. Tempo judgements for the contemporary idiom were the less consistent among all styles. Specifically, the following consistency scale with respect to the musical styles was observed in subjects’ tempo judgements: Rock ballad < Greek dance music < Impressionism < Baroque < Contemporary idiom.

To sum up, results support the belief that tempo is not perceived entirely independently of the temporal, melodic, and harmonic elements of a piece, but rather is part of its integrated sonic structure. On the one hand, regularity or consonance in both rhythmic and melodic stylistic implications facilitates tempo perception, as is the case of the rock ballad. On the other hand, if the stylistic constraints of the piece are perceptually too demanding (complex) or unconventional in terms of typical rules of the tonal Western tradition, as is the case of the contemporary idiom, listeners may not be able to develop a stable mental representation of the musical structure, and this results in a corresponding temporal instability.

The results show that an increase of familiarity with the musical examples resulted in an increase of consistency of correct tempo judgements for all examples The straightforward implication of this evidence is that familiarity gained from formal or informal exposure to music, on the one hand, highlights the ability to form a stable mental representation of music that enables the listener to render fairly stable tempo judgements across trials. On the other hand, novelty yields the opposite results in terms of temporal stability of judgements. One may say that only when the representation of music has stabilised in the mind through familiarisation is the corresponding temporal consistency noticeable.

Moreover, there is statistically significant evidence that an increase of familiarity with the styles in question, except for the contemporary idiom, significantly influenced the degree of consistency of tempo judgements. It is interesting that greater familiarisation with the contemporary idiom did not lead into more consistent choices of most appropriate tempo for the Piano Piece by Michalis Lapidakis. The reader should be reminded here that this composition was selected because it had never been publicly performed before the experiment; as a result, all subjects were unfamiliar with it.

This finding appears to exemplify that exposure to or familiarity with the contemporary musical idiom which is not typical of the conventional tonal grammar does not appear to enhance our comprehension of right tempo and the consistency of tempo judgements for contemporary compositions. In this case, tempo consistency seems to depend on familiarity with the particular stimulus pattern that is heard rather than on familiarity with its style. Thus, one could argue that the contemporary idiom does not set up a strong context that facilitates the formation of the concept of right tempo for individual compositions. It is noteworthy, however, that both listeners who rendered absolutely consistent judgements across trials for the Lapidakis’ composition were composers.

Along these lines, it is worth noting that the finding of this study regarding the effect of familiarity on tempo consistency in a listening task may be compared with Clynes’ & Walker’s observation (1986) that the only time performances by the Sidney String Quartet showed low consistency was with a composition with which the string quartet was not as familiar with the music as with the other works reported in the study. According to the researchers: "Concepts need time to settle, to be established. This need of time for musical concepts to stabilise touches the essence of a musical concept" (p. 113).

The results show that tempo judgements across trials were significantly affected by subjects’ preference for the listening examples. This evidence implies that listeners tend to make more stable tempo judgements for the pieces they like than for the ones they dislike. A possible explanation is that listeners seemingly pay greater attention to the music which they prefer, noticing relationships within the large-scale musical organisation more thoroughly during the task of selecting the correct tempo.

This explanation, however, does not imply that we cannot exhibit higher degrees of tempo consistency or that we cannot listen attentively to something we do not like. For instance, it is quite likely that it was not necessary for listeners to like Yesterday by the Beatles, or The Children of Piraeus (the Greek dance) by Manos Hadjidakis (e. g., examples which received the highest ranking in familiarity), in order to exhibit great stability in their judgements for these pieces. Moreover, with respect to the Piano Piece by Michalis Lapidakis which presented the greatest novelty among the listening examples, results also showed that preference did not significantly influence consistency of tempo judgements. Thereby, one might assume that familiarity or novelty has clearly a stronger effect than preference on tempo judgements here. However, it is obvious that many questions remain open at this point. It would therefore be premature to draw any firm conclusions.

This study is based on the premise that tempo is a general cognitive constraint influencing the way we organise long-scale musical events in real time and, thus, make sense of them. Tempo constitutes an important element of music by enabling various sound events to be woven together at a fixed pace in time. In turn, this temporal pacing is what gives music its unique motional, emotional, and sensuous character, as has been acknowledged in most discussions concerned with music aesthetics, theory, and compositional or performance practice (Lapidaki, 1990).

If this is the case, then the development of a more refined or discerning concept of tempo in students should be considered crucial for a thorough understanding of the expressive qualities of music. Instead of focusing solely on reading of notated structural parameters of music (e.g., melody, rhythm, meter, and the like), music instruction should promote the development of listening skills, especially with reference to the imprecisely represented in conventional notation, motional, and emotional aspects of music, as is the case with tempo (Lapidaki, 1992).

It is somewhat surprising to find, however, that tempo is commonly being treated in music instruction of all levels as if it were solely applied to the metronome or to verbal designations. This leads us to believe that students are not supposed to have an awareness that tempo judgements lie deeply within the human mind which intuitively attempts to supply its own tempo to music in order to ensure the meaningful co-ordination and motion of sound relationships through time. In other words, it is the relation of melody, phrasing, harmony, rhythm, timbre, dynamics, style, and other musical features, to tempo that imbues them with a new and exciting perceptual dimension. Music educators can help students to gain a deeper sense of recognition and mastery of all kinds of relations in a piece of music by showing them the power that tempo exerts on their synthesis.

To help students of all ages find a use for the concept of tempo in music, music educators may consider the task utilised in this research which proposes a fascinating, creative, and― most importantly―an intrinsically musical activity reflecting our need to organise and control the passage of time in music. More specifically, listeners were given the task of manipulating the tempo of a piece of music in real time while examining its influence on the way the music sounds. The ultimate objective was to come up with a tempo in which all elements of the piece would fit together naturally, in a right pace.

On the one hand, listeners were provided with the opportunity to choose the most appropriate tempo in their opinion among a vast number of tempi available to them by means of the computer, something that cannot happen when you perform music due to motor-sensory limitations of the performer. On the other hand, listeners were faced with the technical problem of saying "this tempo is not correct" or "that tempo is correct" which is an important aspect of the creative process in music (Lapidaki, 1990).

However, in a real educational setting, students’ musical decisions about correct tempo have to also rest on the teacher’s guidance. The skilful music teacher will use musical materials which are appropriate for each particular age level but will attempt to demonstrate how musical parameters are integrated and flow with a rightful pace in all music. The thoughtful use of examples― particularly, from contemporary music and from music of various genres and cultures―may also help students to understand the different roles tempo fulfils within the musical structure, from aesthetic fulfilment, to inspiration of dance and song, to arousal of a certain emotional ambience. In addition, the teacher’s recommendation that tempo is the parameter with the greatest degree of variability and possibilities in music will give students the freedom to experiment with music with greater ease and curiosity while focusing on the pacing of musical events.

Finally, by using tempo as a reference point in order to teach other musical elements in their proper motion, we can open a new and intriguing dimension for listening. As Stockhausen (Cott, 1973) said referring to the importance of manipulating tempo in compositional practice:

In this context, the finding that most listeners did not prove to be precisely consistent in their correct tempo judgements over a period of time becomes a secondary issue. Indeed we all vary in the abilities with which our aesthetic perceptions operate. After all, we are not metronomes.

Note. aBach I refers to the C-major Two-Part Invention. bGreek Dance refers to Hadjidakis’ The Children of Piraeus. cBach II refers to the A-minor Two-Part Invention.

Table 2a. Average Cell Means for Tempo Judgments Across All
Subject Groups from Individual Trials of Bach I, Bach II, and Debussy

Note. N=90.

*p < .0001.

Note. N=90.

*p < .0001.

Note. N=90. ^an =30.

P < .001.

Cell Means for Individual Deviation Scores (IDS) Averaged over the Four Trials Arranged by Musical Example and Subjects’ Musical Background (e.g., musicians and non-musicians) and F ratio from an Independent Samples T-test Procedure