Interkey timing in piano performance and typing

Interkey timing in piano performance and typing

Schmuckler, Mark A

Abstract In typing, when the fingers executing two successive movements are on the same hand, the time between keystrokes is longer than when the fingers are on different hands. Biomechanical limitations of the hands are thought to account for this difference. The generality of this finding was explored by investigating skilled pianists’ performance of two successive notes. Experiment 1 failed to find comparable differences in timing as a function of the hands involved. Experiment 2, employing both a piano production and a typing task, replicated the previous piano performance results, and revealed that the timing differences in typing were limited to letter sequences requiring fore-aft and lateral finger movements. Experiment 3 extended this finding to piano performance. Together, these findings clarify the nature of biomechanical constraints on skilled manual performance. The ability of adults to coordinate hand and finger movements during skilled manual tasks is impressive, especially in the two prominent exemplars of such skill: typing and piano performance. Typically, a skilled typist produces between 60-80 words per minute, or approximately 5-7 keystrokes per second, with typing speed reaching up to about 200 words per minute (Genther, 1983; Rumelhart & Norman, 1982). Piano performance can be even more rapid. Lashley (1951) noted that piano performance has the potential for 15 movements per second; others suggest that pianists routinely play up to 30 sequential notes per second over extended passages (Rumelhart & Norman, 1982). Accordingly, the study of piano performance and typing can provide insight into processes of skilled manual motor control. By and large, the majority of such work has examined transcription typing, in which typists type short passages of prose. Analyses of the timing of typing, and of errors made during typing, provide a window onto the cognitive and motoric processes involved in this behaviour. One focal point in typing research involves the production of “digraphs,” which are sequences of two different letters, identified with reference to the hands and fingers involved in their production. Two common classes of digraphs are two-hand (2H) and two-finger (2F) digraphs, which account for roughly 50% and 34% of English digraphs, respectively (computed using data in Solso, Barbuto, & Juel, 1979). A 2H digraph consists of two keystrokes typed by fingers from different hands; examples of 2H digraphs include “su” and “le.” A 2F digraph consists of two keystrokes typed by different fingers from the same hand; examples of 2F digraphs include “li” or “dr.” Research investigating typing has uncovered timing phenomena related to the production of different digraph classes (cf., Rumelhart & Norman, 1982; Salthouse, 1986, for reviews). Using standard QWERTY keyboards, researchers have observed that the interkeystroke interval (IKI), or the time between the occurrence of the first and second keystrokes of the digraph, is shorter for 2H digraphs than for 2F digraphs (Fox & Stansfield, 1964; Gentner, 1982, 1983; Larochelle, 1983; Rumelhart & Norman, 1982; Salthouse, 1986), with approximately a 30-60 ms advantage for 2H digraphs. This greater speed in the execution of 2H digraphs has been interpreted in terms of constraints on the structure of the hands. Because the fingers on a hand are not fully independent, there is limited opportunity to prepare the second movement during the first movement of a 2F digraph. In contrast, the second movement of a 2H digraph can be prepared and initiated during the first movement, with this ability to overlap movements critical in affording faster production of these digraphs. In a direct test of this biomechanical constraint explanation, Larochelle (1984) videotaped finger movements during typing, and found that in 2H digraphs the second movement was initiated about 30 ms prior to the first movement. In contrast, for 2F digraphs, the second movement was begun approximately 40 ms after the end of the first movement. Additional support for this biomechanical constraint idea comes from a study by Gentner (1983) in which beginner typists were taught to touch type. At the beginning of practice, the IKI for 2F digraphs was faster than for 2H digraphs, but this difference reversed by the end of practice, suggesting that typists learned to maximize the extent to which successive keystrokes were executed in an overlapping as opposed to serial manner. In its strongest form, this explanation assumes that the timing of hand and finger movements results primarily (and possibly solely) from the mechanics of finger movements. One implication of this assumption is that the task environment plays little role, if any, in producing these differences. It is possible, however, that the task environment plays more of a role in determining digraph effects, with the demands of typing, which include producing sequential motor movements as quickly and as accurately as possible, providing limits on the timing and production of manual movements. One way of testing these two possibilities (biomechanical constraints vs task environment) is through examining the timing of hand and finger movements in skilled manual domains that place somewhat different constraints on performers. One obvious candidate is the motor control involved in piano performance. If the digraph effects found in typing truly reflect constraints arising from the biomechanical structure of the hands, then similar findings should be observed in piano performance. In contrast, although piano performance and typing are related in that they use the same physical system for productions (i.e., hands and fingers), the demands of piano performance diverge from typing. Although typing places a premium on speed, this is of less importance in piano playing, in which temporal and rhythmic precision are critical. Moreover, piano performance, unlike typing, necessitates a mixture of simultaneous and successive keystrokes, occurring both within a hand and between the two hands, providing pianists with more experience in overlapping finger movements than typists. Given such differences, if the digraph effects found in typing rely in any way on aspects of the task environment or goals of performance, then these digraph differences should be negligible in piano performance. Although previous research on piano performance has examined issues of timing (Clarke, 1985; Mackenzie, Van Eerd, Graham, Huron & Wills, 1986; Mackenzie & Van Eerd, 1990; Palmer, 1989a, 1989b; Shaffer, 1981, 1982; Shaffer, Clarke, & Todd, 1985; Sloboda, 1983, 1985; Todd, 1985) and errors during performance (Palmer, 1992; Palmer & van de Sande, 1993), the detailed goals of this work diverge from the current questions in that they reflect concerns more intrinsic to the nature of musical structure and performance. For example, some work (e.g., Mackenzie & Van Eerd, 1990; Shaffer, 1981, 1982) has explored rhythmic consistency and/or precision as a function of repeated performances of a single piece of music, or the rate of performance. Other work (e.g., Clarke, 1985; Mackenzie et al., 1986; Palmer, 1989a, 1989b; Shaffer et al., 1985; Todd, 1985) has examined deviations from mechanical regularity as a function of musical structure or the expressive intentions of a performer. As with typing, both the timing and error data indicate the structure of the mental representations assumed to be operative during musical comprehension and production. Mackenzie and Van Eerd (1990) provide the most relevant data on interkey timing in their research on rhythmic precision in the performance of piano scales. In this work, pianists performed ascending and descending major scales at a variety of tempos, and in different hand combinations (i.e., one hand alone, both hands in parallel motion, and both hands in contrary motion). Their most pertinent results concern analyses of the IKI, with these authors finding that IKI varied with note position, or the fingers performing these notes, with such variations reflecting biomechanical limitations involved in moving fingers and shifting hand positions. Such results imply that basic biomechanical factors constrain piano performance, and that they can be indexed via analysis of timing parameters of piano production. Unfortunately, because these studies examined IKIs only for successive notes played within a hand, this work does not address whether IKIs vary as a function of two finger movements by a single hand, relative to finger movements across the two hands. The primary goal of this project was to explore the importance of biomechanical and task environment constraints on skilled manual activity. Towards this end, these experiments examined interkey timing during piano performance, relating such productions to the interkey timing results typically observed in typing tasks. Again, if differences in digraph productions result primarily from biomechanical constraints, then similar findings should arise in piano performance. In contrast, divergent results from those typically found in typing would indicate a role for more contextual factors. Experiment 1: 2F Versus 2H Dyads During Piano Performance Experiment 1 represents an initial exploration of this issue by having trained pianists perform a musical analogue to a “digraph typing” task (Bosman, 1993, 1994). The digraph typing task involves presenting letter pairs (digraphs) on a computer screen and asking subjects to type these digraphs as quickly and as accurately as possible. Bosman (1993) has found that this task reveals differences in IKI as a function of digraph class. For use with musical stimuli, we presented pairs of sequential musical notes, or dyads, on a computer screen, and asked pianists to perform these dyads as quickly and as accurately as possible; this task will be called the “dyad playing” task. Why would we use this dyad playing task, as opposed to a more natural production task? One reason is that, unlike transcription typing, the ability to play at first sight, or “sight-read,” an extended passage of music is a skill in which musicians vary greatly. In practical terms, the dyad playing task reduces sight-reading demands to a minimum, making it equally applicable to most pianists. Second, presenting pianists with minimalistic musical passages, such as dyads, effectively eliminates expressive and/or artistic concerns. Although not a consideration in typing, musical sequences are often performed expressively; unfortunately for our purposes, expressive intentions have a significant impact on the timing of performances (see Palmer, 1989a). Finally, the dyad playing task provides multiple possible dependent measures. The first measure is the already discussed interkeystroke interval (IKI). A second measure involves the time between the appearance of the to-be-performed stimulus and the first keystroke, or the initial latency. Previous research in typing (Bosman, 1993, 1994; Sternberg, Monsell, Knoll, & Wright, 1978) suggests that these measures are sensitive to different stages of motor control, with initial latencies indexing preparation of motor performance, and IKI assessing execution processes. Research by Sternberg et al. (1978, p. 142, Figure 5A) has found that typists take longer to initiate the production of letter strings requiring alternating hand movements, relative to letter strings requiring only a single hand, with these differences most pronounced for strings of three letters or more. It is also of interest to examine the timing of the musical dyads as a function of left vs right hand performance, looking for evidence of hand asymmetries in production. Previous research on piano performance (MacKenzie & Van Eerd, 1990) has found significant hand asymmetries in piano productions, with pianists’ emphasizing the right hand more than the left hand in terms of the amount of overlap between notes and note intensity. Accordingly, it is possible that the timing of the production of simple musical dyads will vary systematically as a function of left vs right hand performance. METHOD Subjects Eight trained pianists took part in Experiment 1. They had been playing the piano for an average of 16.6 years, had received formal instruction for an average of 10.4 years, and were currently playing the piano for an average of 6.9 hours per week. All pianists had achieved at least grade 8, Royal Conservatory of Music ranking (mean RCM level = 9.1), and were right-handed. All subjects were either paid, or received course credit in an introductory psychology class, for participating. Apparatus and Stimulus Materials All stimuli were generated by an IBM-PC 286 compatible computer, displayed on a 14″ VGA monitor, and consisted of pairs of musical notes displayed on a musical staff. Ten notes were used, with five of these notes played by the left hand and five played by the right hand. The notes performed by the left hand consisted of the five white keys ranging from E”Symbol not transcribed” (equal tempered tuning = 330 Hz) to B”Symbol not transcribed” (494 Hz); the notes played by the right hand consisted of the five white keys ranging from Cs (523 Hz) to Gs (784 Hz); Figure 1 presents a schematic piano keyboard, with the stimuli labeled. The musical staff consisted of a treble clef and five lines — together this display was approximately 6.0 cm long and 2.2 cm high, and appeared in the center of the computer screen. The musical notes appeared side by side on the staff, separated by approximately 1.9 cm. For convenience in programming, all notes appeared as open ovals with stems (half notes). The notes to be played by the left hand had downward stems; the notes to be played by the right hand had upward stems. Pianists played these notes on a Yamaha DX7 synthesizer, which has an unweighted, velocity-sensitive keyboard. The harmonic structure of the voice used by the DX7 approximated that of a piano, with each tone having about a 15-ms rise to peak amplitude, a gradual decay over the length of the tone until its release, followed by an approximately 80-ms fall to zero amplitude. All tones were amplified by a Peavey KB60 amplifier. The DX7 synthesizer was connected to the same IBM-PC that controlled the presentation of stimulus notes, using a Roland MPU-401 MIDI interface. The internal clock of the IBM-PC calculated all timing measures. This clock had a 1 ms resolution, with the resolution of the MPU-401 at plus or minus 2 ms. The computer recorded the time at which the stimulus originally appeared on the monitor, as well as the onsets and offsets of both keystrokes. Conditions There were two conditions in this study, corresponding to the combination of hands used to play the stimulus notes. In the 21: condition, the notes were played by different fingers on the same hand, either left or right. For each hand there were five stimulus notes, producing 25 possible stimulus dyads per hand. Because our interest was in examining the timing of notes played by different fingers (either within the same hand or between the two hands) the 5 notes involving a repetition of the same finger (e.g., E”Symbol not transcribed” followed by E”Symbol not transcribed”) were eliminated. This left 20 stimulus dyads per hand, for a total of 40 stimulus dyads. In the 2H condition, the stimulus notes were played by one finger from each of the hands (e.g., E”Symbol not transcribed” followed by E”Symbol not transcribed”). Again, five stimulus notes for each hand produced 25 stimulus dyads with the left hand playing the first note and the right hand playing the second note, and 25 stimulus dyads with the reverse order, producing 50 stimulus dyads in all. Together, the 40 2F and 50 2H dyads constituted a single block of trials, with all pianists receiving a different random order for each block of trials. Procedure All pianists were seated at the DX7 keyboard, and were told that they were participating in a study examining the performance of short passages of music. At the beginning of each trial, a musical staff and treble clef would appear on the computer screen, followed by two notes. The pianists’ task was to play these notes as quickly and as accurately as possible, making sure to perform these notes successively, rather than simultaneously. Pianists were cautioned that although the notes would appear on the screen as half-notes (open ovals) this was done for convenience in computer programming, and that they should play these notes as quickly as they could. Pianists were also told that there would be five notes played by the left hand, and five notes played by the right hand, notated by the downward or upward stems of the notes as well as by labels attached to the piano keyboard itself. To make this task comparable to standard touch-typing, in which the hands are placed in the homerow position (the middle row of the typing keyboard), the pianists’ fingering for each note was constrained, such that each finger was uniquely associated with the production of a single note. Thus, the note E”Symbol not transcribed” Was always to be played by the little finger of the left hand, F”Symbol not transcribed” was to be played by the ring finger of the left hand, B”Symbol not transcribed” to be played by the thumb of the left hand, C”Symbol not transcribed” to be played by the thumb of the right hand, G”Symbol not transcribed” to be played by the little finger of the right hand, and so on. Although pianists found this limitation unusual, none reported difficulty in adopting this constraint. Following these instructions, pianists placed their hands on the piano keyboard, and the experimenter hit a key on the computer keyboard to initiate a block of trials. After a 2s delay the musical staff appeared on the computer monitor, followed 1s later by the notes to be played. Following the release of the second performed note, the computer monitor blanked, and there was a 2s delay before the occurrence of the musical staff signaled the beginning of the next trial. As described earlier, a single block consisted of 90 trials, with subjects completing four such blocks. The entire first block of trials was considered practice, to allow pianists to adjust to the timing of experimental trials, to the feel of the DX7 keyboard, and to adapt to the constraint imposed on their fingering of the different notes. The experimenter was present during the entire first block of trials to caution pianists about the use of inappropriate fingering, as well as to answer questions. Each block took approximately 7-8 minutes, with pianists allowed to take breaks between blocks as they wished. After completing all blocks, the purpose of this study was explained, and pianists completed a questionnaire concerning their musical background. The entire experiment took approximately 45-60 minutes. RESULTS AND DISCUSSION The two primary dependent measures in this study were the initial latency to produce the first keystroke, and the interkeystroke interval (IKI). In calculating these measures, all trials on which an error in production occurred, such as playing two notes by one finger, or playing the wrong note were removed. Averaging across the three blocks, there was a mean of 6.6% errors, with individual subject error rates ranging from 1.5% to 10.4%. Analyses of error rates failed to reveal any significant differences across the three blocks of trials, r(1,14) = 1.66, nor as a function of condition, F(1,7) = 0.18. Mean initial latency and IKI measures were calculated(f.1) and examined using three-way analyses of variance (ANOVAs), with the within subject factors of condition (2F vs. 2H), block (block 1, block 2, or block 3), and hand (left vs right). For this last factor, initial latencies were coded relative to the hand playing the first note of the dyad, while IKIs were coded in terms of the hand performing the second note of the dyad. The analysis of the initial latencies revealed a main effect of condition, with 2H dyads (M = 792 ms, SD = 112) initiated more slowly than 2F dyads (M = 715 ms, SD = 105), F(1,7) = 50.1, MS”Symbol not transcribed” = 2887.2, p 10%) in either the digraph typing or dyad playing task. None of these subjects participated in Experiment 1. Mean errors for the dyad playing task were 6.0% and 9.0% for 2F and 2H conditions, and 10.2% and 9.5% for 2F and 2H conditions of the digraph typing task. In terms of their musical background, the remaining participants had been playing the piano for an average of 14.6 years, had received formal instruction for an average of 12.0 years, and were currently playing the piano an average of 6.4 hours per week. All pianists had passed at least grade 8 RCM (mean RCM level = 9.1). Seven of the subjects were right-handed, and one subject was left-handed. The participants had been typing for an average of 8.1 years, with a self-reported average typing speed of 60.1 net words per minute, and an actual typing speed of 56.8 net words per minute (see below). Four of the eight participants had held jobs requiring them to type, and seven of the eight subjects were currently typing for three hours or less per week (M = 1.1 hrs per week); the remaining subject was currently typing for 40 hours per week. All subjects were either paid or received course credit in an introductory psychology class for participating. Dyad Playing Task Apparatus, stimuli, and conditions. The experimental apparatus employed was identical to that of Experiment 1. The stimuli performed by each hand were the same as in Experiment 1, consisting of multiple blocks of 40 2F stimuli and 50 2H stimuli. Digraph Typing Task Apparatus. All subjects completed the typing tasks using a standard QWERTY keyboard attached to the IBM-type 286 computer. The computer monitored and recorded the timing of all keypresses. All stimuli appeared in the center of the 14″ VGA monitor attached to the computer. Stimuli. Two sets of digraphs were created, home-row digraphs and non-home row digraphs. The home row digraphs consisted entirely of letters located on the home row of the QWERTY keyboard (i.e., a, s, d, f, g, h, j, k, l). The non-home row digraphs could contain letters that were located on the home row, but could not consist of two consecutive letters located on the home row. Figure ! also presents a schematic of the typing keyboard, with the home-row position of fingers for the left and right hands notated. Both sets of digraphs were drawn from the set of digraphs legally occurring in English. Homerow stimuli. The home-row digraphs consisted of 21 2F and 21 2H digraphs for a total of 42 digraphs. To make the 21: and 2H digraphs as comparable as possible, the two digraph sets were matched on digraph frequency. For the 2F digraphs, mean frequency was 3391.6 (SD = 8016.4); for the 2H digraphs, mean frequency was 3630.8 (SD = 8299.6). Frequency data represent occurrence per million words, and were obtained from Solso, Barbuto, & Juel, 1979. The 2F and 2H digraphs were selected using the following method. From the set of all home row digraphs, the digraphs beginning with a particular character were identified (i.e., all of the digraphs beginning with a, s, and so on) Once the subset of digraphs beginning with a particular character was identified, 2F and 2H digraphs were selected in pairs so that the finger typing the second letter of the digraph was the same across the 2F and 2H digraph. For example, the 2F digraph “af” was paired with the 2H digraph “aj” because in both instances the second letter is typed by the index finger. For each letter appearing on the home row, all such pairs of 2F and 2H digraphs were selected. Non-homerow stimuli. The non-home row digraphs were created using the procedure employed by Bosman (1993, 1994), and consisted of 30 2F and 30 28 digraphs. The two digraph sets were matched on digraph frequency, with a mean frequency for 2F digraphs of 10343.6 (SD = 8409.4), and for 2H digraphs of 11525.0 (SD = 8797.2). In addition, within each digraph class, each letter in the alphabet appeared an equal number of times in the first and second position within a digraph. Within digraph class and position, an equal number of letters were typed by the left and right hand. To achieve this, it was necessary to repeat some of the letters typed by the right hand because the layout of the QWERTY keyboard is such that 15 characters are typed by the left hand, and 11 by the right. The letters U, I, O, and P, were repeated because they are each typed by a different finger on the right hand. One repetition of all letters, plus an additional repetition of U, I, O, and P, in every position, and in every digraph class, resulted in 30 2H and 30 2F digraphs. Assessment of typing skill. Typing skill was assessed using Paragraph 6 from Form F of the Nelson-Denny Reading Test. This paragraph contains 1398 keystrokes. Prior to typing the skill passage, each subject warmed up by typing a passage taken from Reader’s Digest that was approximately 500 keystrokes in length. Procedure All subjects were told that they were participating in a study comparing piano performance and typing. The instructions and procedure for the dyad playing task were identical to those of Experiment 1, except that there was no constraint in their choice of fingering for the intervals, so long as pianists retained the left vs right hand distinction outlined in Experiment 1. All other experimental details (stimulus presentation, timing of trials) were identical to Experiment 1. In the digraph typing task, subjects first completed the warm-up and skill passages, and then began the digraph typing task. To familiarize subjects with the typing task, 10 practice trials were completed, with the practice stimuli randomly chosen to contain an equal number of 2F and 2H digraphs. The experimental trials consisted of three blocks of 102 trials, with each block containing one repetition of the entire stimulus set. Each subject had a different random ordering of both practice and experimental trials. While performing the task, subjects sat with their hands in home row position, and initiated each trial by pressing the space bar with their thumb. A fixation cross appeared in the middle of the screen 1s after the space bar was pressed, and remained on for 500 ms. The stimulus replaced the fixation cross, and remained on until a response was made. Subjects were instructed to type the digraph as quickly as possible without making errors. The entire experimental session, including both dyad performance and digraph typing tasks, took approximately 7590 minutes. RESULTS AND DISCUSSION Experiment 2 was analyzed in a manner comparable to that of Experiment 1. First, all production errors were removed. In the dyad playing task, averaging across the three blocks of trials, there was a mean of 4.1% errors, with individual subject rates ranging from 1.1% to 7.0% errors. For the digraph typing task, there was a mean of 4.6% errors across blocks, with individual subject rates ranging from 0.5% to 9.0% errors. Analyses of errors failed to reveal significant differences as a function of 2F versus 2H conditions for the dyad playing task, F(1,7) = 1.58, or the digraph typing task, F(1,7) = 0.0. For the dyad playing task, initial latency and IKI values were calculated for each of the blocks of trials as a function of condition (2F versus 2H). For the digraph typing task, initial latency and IKI were computed separately for the homerow and non-homerow stimuli. These measures for the piano and typing tasks were then directly compared in a pair of three-way ANOVAS, with the factors of task (piano, homerow typing, or non-homerow typing), condition (2F vs. 2H), and block (blocks 1, 2, or 3). For the analysis of initial latencies, there were main effects of block, F(2,14) = 7.9, MS”Symbol not transcribed” = 1518.6, p

Copyright Canadian Psychological Association Jun 1997

Provided by ProQuest Information and Learning Company. All rights Reserved