RESEARCH ON COMPUTER-ASSISTED INSTRUCTION IN MUSIC

Transcription

1 RESEARCH ON COMPUTER-ASSISTED INSTRUCTION IN MUSIC by PAUL LORTON,jR. University of San Francisco ROSEMARY KILLAM North Texas State Univerisity Denton, Tex. and WOLFGANG KUHN Department ofmusic Stanford University THE STANFORD system for computer-assisted instruction (CAL) in music has been evolving since This system, consisting of an electronic organ and terminal linked to the computing system of the Institute for Mathematical Studies in the Social Sciences (IMSSS), was designed for both instruction in music and research on instruction in music. The purpose of this discussion is to present both the system and the results of the research accomplished thus far on the system. Music instruction demands a highly individualized approach; computerassisted instruction has developed out of efforts to meet this demand. The development of the CAl-Music project has focused on five specific needs, namely, for sound, for real-time interaction, for individualization, for detailed student records, and for basic research.- With these five needs in mind, work on a software system to conduct research and instruction was begun in the Fall of Almost immediately, a curriculum driver was developed and the first few lessons were implemented. From the outset, this work was dedicated toward making the computer program that controls the curriculum presentation as knowledgeable as possible about the nature of the musical device being used and about the structure of the prospective curriculum. This was done so that as much curriculum as was feasible could be generated by the computer program rather than be written by a curriculum author. A wider ranging and better rationalized body of curriculum material was thus made available more rapidly than would otherwise have been possible. 877

2 878 LORTON, KILLAM, & KUHN During the Winter of 1973, the program and curriculum were tested with small groups of elementary-school (ages 7 to II years) and college students. During the 1973 Spring quarter, the system was put into use as a supplement to the regular curriculum of first-year music courses at Stanford, Student sessions of minutes proved to be the most appropriate. Sufficient student data were generated during this first quarter to undertake a study correlating attitudinal and biographical student information with the students' performance on the system (Herrold, 1973). The system has since been used as an adjunct to undergraduate music courses almost continuously. Considerable data have been recorded. Some students have been on the system for four quarters, thus providing a great deal of longitudinal information. 1. DESCRIPTION OF THE SYSTEM The written portion of the curriculum is presented and student responses are entered through a terminal located in a music practice room in the central campus area. Originally, the students worked at a Teletype Model-33 KSR terminal; currently, the terminal is a cathode-ray tube (CRT), which is faster and quieter than the Teletype. Music is presented by means of a Thomas solid-state organ, Model 145, The organ is linked to the IMSSS PDP-IO computer through an interface that translates 8-bit patterns into notes to be addressed (6 bits) and one of the following commands: SET-ON, PLAY, SET-OFF, or OFF. At present, interaction with the organ is one way; notes are sent to the organ but are not received from the organ. The interface provides access to 64 notes on the organ. These notes are located in two octaves on the upper keyboard and three octaves on the lower keyboard. The timing of the notes is dictated by the use of a 300 baud line for the connection with the computer and the fact that the interface was designed to operate at the 300-baud rate. The main computer program developed for this project is divided into two parts. One part was developed to access the curriculum (generating as much as possible) and to handle the student data and history information. The second part has been developed with as much understanding of the organ and of the relationships among musical constructs as demanded by the curriculum and research projects. The various options available to the curriculum author are described in Killam and Lorton (1974), In general, the curriculum is divided into several homogeneous groupings called strands. At present these include intervals, triads, melody, rhythm, chords, modulation,and others that are constructed for specific research projects. Because of the importance of maximum flexibility for both instruction and research, many options are available on line to the student or experi-

3 MUSIC 879 mental subject. These allow free movement, as appropriate to the particular strand involved, throughout the curriculum. Students returning for subsequent sessions may begin at the point where they signed off, restart any material from its initial lesson, transfer to the point of partial completion in strands used previously, or transfer to lessons preceding or following. In addition, students have the option of constructing musical examples that can be stored in the program and played on request. 2. USES FOR RESEARCH The CAl-Music system makes possible a wide range of research into auditory perception and into the acquisition of auditory skills. The potential, and some of the details, of early studies are discussed at length in Lorton, Killam, and Kuhn (1975). There is general agreement among teachers of music theory on the need for students to learn to identify musical intervals and other basic tonal configurations, like triads, prior to more advanced study. However, little is known about how such learning takes place or about the mostproductive sequence of presentation and study of basic skills. Because the CAl-Music system was designed to be flexible and research-oriented, it has been possible to conduct a sequence of studies that explore both fundamental questions of musical perception and applied questions of performance in an instructional setting. The discussion that follows presents one such series of studies that were carried out to explore the perception of intervals (two-note sequences) and triads (three-note sequences). 2.1 Perception of Intervals One of the first studies undertaken on the CAl-Music system was a replication and extension of an earlier study on interval identification by Plomp, Wagenaar, and Mimpen (1973). The earlier study was limited to the presentation ofsimultaneous intervals. The replication was extended to cover presentation of the notes in ascending and descending order. A full report of this research is available in Killam, Lorton, and Schubert (1975). Plomp and his associates investigated the identification of simple harmonic intervals presented for four different time durations: 15,30,60, and 120 milliseconds. Data were presented in the form of a confusion matrix that basically presents the proportion for which each response was used for each stimulus. The Stanford CAl-Music system was used to replicate and extend this study. The simple intervals were presented simultaneously and sequentially at 100 and 200 msec. for each pitch. To facilitate comparison with Plomp's study, his data for 60 and 120 msec. were extracted into separate confusion matrices and converted to percentages.

4 880 LORTON, KILLAM, & KUHN Data collected by the CAl-Music system included the stimuli presented (pitches, speed of presentation, mode of presentation, etc.) and the student response (including the time of the response). Information on student input other than responses to stimuli (requests for new topics, repetitions of the stimuli, etc.) was also saved. By taking the relatively raw data from the Plomp report, making it machine-readable, and putting it through the same analysis procedures as the data from the CAl-Music system, direct comparisons could be made; Figure 1 shows such a comparison. In general, the percent correct on each 100 ~!?l Iiil Plomp et 01 o Ki 110m Lorton ond Schubert 80 ~ ~ :; u <= "~ ~ ol- L- L L L L- L L '-- MI2 MA2 MI3 MA3 PE4 Y PES MI6 MA6 MI7 MA7 PE8 Interval FIGURE 1. Comparisons of Plomp et al. and Killam et al. data. interval type was fairly close. In Figure 2, the error patterns of two students are charted. These two students represent a considerable range in overall probability correct and, as will be indicated in the discussion comparing the interval and triad studies, show one of the key findings of the work on perception-that there appears to be a central target image around which errors will be distributed. A 'series of analyses of variance were calculated on the data from this study. Table 1 is the an~lysis-of-variancesummary table for the main

5 MUSIC ~ g 20 '" 0 c.8 '" "E > l' c '"~ '" I 0 I Student response in half step units FIGURE 2. Error patterns of two students. TABLE I Analysis of Variance Summary Table-Intervals by Duration Source of variation Sum of squares df Mean square F Main effect (1) ** Main effect (j) ** Main effect (K) Interaction (I x J) ** Interaction (I X K) * Interaction (j x K) Within cells (l x J x K) TOTAL ,439 Note. Analysis over 48 intervals by durations of 200 msec. and 100 msec. (all students), where I = subjects, J = imervals, and K = duration. *p <.05. **P <.01.

6 882 LORTON, KILLAM, & KUHN results of the interval study, Principally, the effect of different intervals was statistically significant while the effect of different durations was not. 2.2 Perception of Triads In a second area of study, the perception of triads (three-note sequences) was investigated. A full report is found in Killam (1976). Students identified traditional triad structures (major, minor, diminished, and augmented) in all possible inversions. Matrices similar to those reported in Killam et al. (1975) were used to chart areas where students confused one structure with others. The auditory material that was presented consisted of an introductory set of 10 triads, followed by six sets of 48 triads each. Each of the six triad sets contained the same 48 triads: 12 each of diminished, minor, major, and augmented. The subsets of 12 diminished, minor, and major triads were constructed on the pitches A2, C3, E flat 3, and G flat 3. Augmented triads were constructed on each note of the 12-tone scale, so that each of the augmented triads occurred in its three possible pitch arrangements. The first two letters of the triad qualities were used to identify all triads and were also the format for subject responses: D1 ~ diminished, MI ~ minor, MA = major, and AU = augmented. After the instructions were printed out, the first triad was played and the subject identified the triad quality. No time limit was set in which the subject had to make a response; once the subject responded, the next triad was played. Snbjects thus controlled their own response times. The computer program recorded the time of each response. Subjects had the option, if tired, of stopping at the end of any of the six triad sets and continuing the experiment at a later time. Triad presentation in each set was randomly accessed from the data files, so that all subjects received a different ordering of triads within the sets. Table 2 summarizes the order and modes of presentation of the triad sets. The first three sets (following the introductory set) were presented at TABLE 2 Order and Modes of Presentation of Triad Sets Triad seta Set 1 Set 2 Set 3 Set 4 Set 5 Set 6 Description.2 seconds/note simultaneous.2 seconds/note ascending.2 seconds/note descending.1 seconds/note simultaneous.1 seconds/note ascending.1 seconds/note descending a Each of the six triad sets contained the same 48 triads.

7 MUSIC seconds per note; the last three sets were presented at.1 second per note. Within each subset of three, triads in the first set were presented simultaneously; in the second set, in ascending pitch order; and in the third set, in descending pitch order. The ordering of these modes of presentation was fixed and thus the same for each subject. A total of 19 subjects (10 males and 9 females) volunteered their participation in the study. They were recruited from the group of students who had successfully (with more than 80% accuracy) completed the triadidentification portion of the CAl program, available to those enrolled in undergraduate music theory courses at Stanford University. Each of the 19 subjects was presented with a total of 288 triads, so that a group total of 5,472 triad stimuli were presented. The responses of the subjects to five of the total number of stimuli were lost in transmission, so that this study was based on 5,467 subject responses. In Table 3, the total subject response to TABLE 3 Confusion Matrix of Triad Data, Converted to Percentages Response Stimulus Percent of total Diminished Minor Major Augmented response Diminished Minor Majpr II Augmented II 5 I TOTAL a a The differences in the totals from 100% reflect rounding errors and unidentifiable responses such as typographical errol's. all triad stimuli is presented in the form of a confusion matrix. Table 4 presents the rank order of percent correct on the four ~riad qualities, as TABLE 4 Rank Order by Percent Correct for Triad Quality Triad quality Percent correct Point of greatest confusion Major 88 Minor (7%) Minor 71 Major (l2%) Diminished 50 Minor (26%) Augmented 43 Diminished (24%)

8 884 LORTON, KILLAM, & KUHN well as the point of greatest confusion for each. The points of greatest confusion for major and minor triads were reciprocal, but the augmented triad quality was most frequently confused with diminished, and diminished with minor. Accuracy of identification of triad quality varied significantly with length of presentation time (F = 15.99, df = 1/54, P <.01), as summarized in Table 5. Table 6, then, combines the data for the two durations and presents accuracy according to mode of presentation (simultaneous, ascending, descending). Accuracy of identification varied significantly (F = 9.12, df = 2/108, P <.01) with mode of presentation. Lowest accuracy within triad quality occurred in simultaneous presentation, with the exception of augmented triads, where the descending mode of presentation had TABLE 5 Quality Accuracy by Length of Presentation Length of Quality presentation time in sees. Diminished Minor Major Augmented.2 54% 75% 90% 45%.1 46% 67% 85% 42% Loss of accuracy at shorter presentation time 8% 8% 5% 3% TABLE 6 Quality Accuracy by Mode of Presentation Mode of presentation Quality Diminished Minor Major Augmented Simultaneous 45% 67% 84% 41% Ascending 50% 73% 92% 50% Descending 55% 73% 87% 39%., the lowest identification accuracy. Highest accuracy occurred in ascending mode of presentation (with highest identification of minor triads being equal in ascending and descending mode) with the exception of diminished triads, where accuracy according to mode of presentation occurred in augmented triads. Table 7 presents a matrix of subject responses to inversions ofthe major, minor, and diminished triads. Since no inversions of augmented triads can be distinguished out of musical context, the SUbject's responses to augmented triads are presented to the right of inversion data for comparison.

9 MUSIC 885 TABLE 7 Confusion Matrix of Triad Inversions Stimulus Response Root position First inversion Second inversion Augmented DP Mlb MA' DI MI MA DI MI MA (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) DI 58 8 I MI MA IO AUd <l DI = diminished, b MI = minor, C MA = major, d AU = augmented. Although percentage of accuracy varied with inversion, the points of greatest confusion remained the same for all inversions except the secondinversion minor triad, which was most frequently confused with diminished triad quality. Table 8 presents the rank order of accurate response to triad inversions. Identification of triad quality retained its previous ordering, and the ordering for inversions of triad quality are the same within each quality. TABLE 8 Rank Order of Inversion Identification Triad quality and inversion Percent of accurate response Major, root position 92 Major, 2nd inversion 87 Major, 1st inversion 83 Minor, root position 75 Minor, 2nd inversion 73 Minor, 1st inversion 65 Diminished, root position 58 Diminished, 2nd inversion 47 Diminished, 1st inversion 44 Augmented (no inversion distinguishable) 43 Individuals ranged in accuracy of identification of each triad quality from 2% to 100%. Ranges for individual subjects by triad quality were:

10 886 LORTON, KILLAM, & KUHN Diminished, 9% to 93%; Minor, 35% to 100%; Major, 56% to 98%; and Augmented, 2% to 83%. The accuracy of recognition corresponds with frequency of occurrence of triads in the literature of tonal music, 'where major triads are by far the most common and augmented triads the least common. Thus, the subjects' ability to identify triad quality may be influenced by the frequency with \vhich those triads are encountered in auditory experience. The use of the different triad qualities in the total response did not correspond entirely with frequency of use in the literature; the response "minor" was given one percentage point more often (31%) than the response "major" (30%). Shortened duration of stimuli had a significant effect on accuracy, in contrast to the interval study discussed above. The mode of presentation significantly affected the accuracy of recognition, but did not result in a clear rank-ordering of quality or of mode. Descending mode was responsible for the highest accuracy on diminished triads and the lowest accuracy on augmented triads. The' highest accuracy produced by the combined effects of quality, mode of presentation, and duration was 94% (major,.2 sec., ascending) and the lowest was 38% (augmented,.1 sec., descending). The effect of inversion upon identification accuracy produced identical rank-orderings within quality. The root position of each triad quality was the most accurately recognized inversion; first inversions of all qualities were less accurately identified than second inversions. This does not correspond to use of triad inversions in the literature, where second-inversion triads are far less used than first-inversion or root-position triads. Total interval width outlined by the triads was not a strong ordering pattern. The triads outlining the interval of a major sixth (second-inversion major, first-inversion minor, and first- and second-inversion diminished triads) ranked second, sixth, eighth, and ninth, respectively. The 1'Ootposition major and minor triads outlining a petfect fifth ranked first and fourth, respectively. Individual accuracy varied greatly. There is some evidence that subjects with particularly low scores on some triad qualities had no distinct auditory concept of these qualities, or perhaps even reversed some quality concepts. A reciprocity of confusion of major and minor triad qualities was found, but points of major confusion for diminished and augmented were not reciprocal. Since the confusion of diminished with minor triad quality occurred on all three inversions of the diminished triad, the confusion would seem to be one of triad quality rather than possible overall triad width. Triad quality appears to be a strong central factor in triad perception, stronger than mode of presentation or duration. There was no overlap of identification accuracy between the poorest identified quality (augmented) under the most favorable conditions (ascending,.2 sec., 53%) and the best

11 MUSIC 887 identified quality (major) under the least favorable conditions (simultaneous,.1 sec., 62%). Rank ordering of triad inversion occurred within triad quality. The subjects' confusion of triad qualities formed identifiable patterns. Triad recognition presents clear patterns of auditory perception not ordered by total interval width. The stable patterns of triad-quality perception and confusion that prevailed over the effects of duration, mode of presentation, and inversion have possible implications for theories of auditory imagery. 2.3 Comparison ofinterval and Triad Perception In addition to studies involving perception, two studies were conducted concerning student performance on the CAL material. These two performance studies looked at the data from student CAL lessons with the same kind of analytic procedures used for the perception studies. These four studies all considered auditory perception of musical stimuli in a fairly standard setting. We have been able to make some interesting comparisons from the confusion matrices in each study. Because the original detailed stimulus-response data on the Stanford interval and triad studies were available for reanalysis, direct comparison ofinterval and triad perception was accomplished (Lorton & Killam, 1976). Several findings resulted from this analysis. For example, female accuracy was greater than male in all modes of presentation of intervals, while female accuracy was lower than male for simultaneous triads and only slightly higher than male for ascending and descending triads (see Table 9). TABLE 9 Comparison of Triad and Interval Accuracy by Sex of Subjects Sex of subject Mode of presentation Simultaneous Ascending Descending Intervals Female Male 73% 61% 85% 76% 83% 78% Triads Female Male 58% 62% 67% 66% 65% 63% The effect of the vocalrange of the subjects on accurate identification of stimuli was considered. For this reason, both triads and intervals were

12 888 LORTON, KILLAM, & KUHN constructed over a relatively wide pitch-range. Intervals encompassed a two-octave range and triads, one octave plus a major sixth. Stimuli were arbitrarily separated into categories according to the lowest note. Accuracy by sex of the subjects was determined for the categories thus established. The results are presented in Table 10. Comparisons are complicated by the need to construct triads and intervals on varying pitch schemes. However, certain comparisons can be made. No general rank-order of accuracy correlated with lowest pitch emerged for either male or female responses. Similarly, no obvious accuracy pattern of males compared to females occurred at either end of the pitch range. (Accuracy on interval pitch range should be viewed bearing in mind that mean female accuracy exceeded male by 10%. The greater accuracy of females in the interval study was not found in the triad study.) Table 11 summarizes the comparison between triad and interval identification according to the rank order of the accuracy of identification. The clear-cut rank-ordering of triad accuracy was not found in the rank ordering of interval accuracy. Triad qualities clustered throughout their rank ordering, and the inversion ordering within qualities remained the same for major, minor, and diminished. In contrast, several of the intervals received the same response-accuracy. Accuracy rank-order appears to have little to do with the consonance or dissonance of intervals. Both the highest ranked (P8) and lowest ranked (m6) intervals are considered consonant. Any anticipated correlation of traditional dissonances, such as m2 (ranked third) and M7 (ranked tenth), did not occur. This has particular significance in the case of the m2 and M7, since they are inversions of each other. The ranking did not show a clear relationship to overall interval width, although the three lowest ranked intervals (M7, m7, and m6) are larger than the tritone, and the three highest ranked can be considered smaller than the tritone if the P8 is considered as an inversion of the unison. Comparison of major points of confusion of triads and intervals by rank-order correct is presented in Table 12. Since only four response categories (the triad qualities) were available to subjects in the triad study, as opposed to 12 response categories in the interval study, the percentages of points of greatest confusion were higher on the triad study than on the interval study. In general, the point of greatest confusion on triad stimuli received the higher percentages of responses as the accuracy decreased. The point of greatest confusion for augmented triads was below that of all inversions of diminished triads. No such pattern emerged in the interval study. Three of the points of greatest confusion exceeded 10%. As might be expected, the two lowest-ranked intervals were among these three (m6 and m7), but the third (m2) was third-ranked in order of accuracy. In general, any stimulus to which the response shows a high percentage of confusion with one incorrect response suggests confusion of the actual

13 MUSIC 889 TABLE 10 Identification Accuracy According to Lowest Pitch Lowest Pitch Sex of subject Male Female Total C' Xb C X C X Major, minor, diminished triads A C D F Augmented triads A C C D D E F F G G A B Intervals F G G A A B C C D II D E F F G G A A B a C = percent correct. b X = percent incorrect.

14 890 LORTON, KILLAM, & KUHN T ABLE II Rank Order of Triad Inversion and Interval Identification Triad quality and inversion Percent of accurate response Interval quality Percent of accurate response Major, root position Major, 2nd inversion Major, 1st inversion Minor, root position Minor, 2nd inversion Minor, 1st inversion Diminished, root position Diminished, 2nd inversion Diminished, 1st inversion Augmented (no inversion distinguishable) P8 M3 m2 P4, M6, and P5 M2 and m3 T M7 m7 m TABLE 12 Comparison of Points of Major. Confusion for Triads and Intervals, Triad quality and inversion Accurate response Point of greatest confusion Interval quality Accurate response Point of greatest confusion Major, root position 92% Minor 4% P8 88% M7, P5 2% Major, 2nd inversion 87% Minor 7% M3 84% m3 6% Major, 1st inversion 83% Minor 10% m2 83% M2 12% P4 82% P5 6% Minor, root position 75% Major 10% M6 82% m6 3% Minor, 2nd inversion 73% Dim. 12% P5 82% P4 8% Minor, 1st inversion 65% Major 17% M2 80% m2 8% m3 80% M3 7% Diminished, root position 58% Minor 25% TT 72% P4 6% Diminished, 2nd inversion 47% Minor 26% M7 70% m7 8% Diminished, 1st inversion 44% Minor 26% m7 58% M7 11% m6 55% P4 1I% Augmented (no inversion distinguishable) 43% Dim. 24% auditory image. Conversely, a stimulus with a low percentage of accurate response, but no centered point of confusion, more probably represents merely a poorly recognized stimulus. Comparison of the overall width of triads and intervals must take into account that the intervals produce 12 separate widths, one for each interval, and that all inversions of triads are encompassed within four intervals: diminished and perfect fifths, minor sixths (augmented fifths), and major sixths. Table 13 compares the percentages of accuracy on those intervals.

15 MUSIC 891 TABLE 13 Overall Interval-width Comparison Percent correct Interval width Triad Interval study study D5(Aug 4) P m6(aug 5) M Accuracy was nearly equal on triads and intervals on the P5 and m6. The interval width of root-position major and minor triads is a perfect fifth, while augmented triads, first-inversion major triads, and second-inversion minor triads are included within the overall width of the minor sixth. Comparison of the length and modes of presentation of the triad and interval study shows that halving the length of presentation produced a significant drop in accuracy of recognition of triads, but not of intervals. The recognition of ascending and descending intervals and triads fell 15% and 17%, respectively, while recognition of simultaneous triads fell only 9 percentage points. The possible role of auditory images in triad and interval recognition needs further research. The comparative data presented in this study imply that subjects image sounds. Interval recognition may involve tri~d images also in some sort of "subtractive reference" process. The comparison of percent correct in terms of overall width of triads and intervals may shed some light on the low recognition accuracy of subjects on the minor sixth. This interval width had the lowest percentage of recognition in triads also,.accounted for by the low accuracy on augmented triads. If, indeed, triad images were used to recognize. intervals, then the minor sixth, out of context, presented a problem, since it is also the augmented fifth, the interval encompassing the augmented triad. If the subjects related the m6-as inter-second-inversion minor, this might ac~ count for some of the inaccuracy on the m6, since the augmented triad had no strongly consistent recognition or confusion pattern. Admittedly, this is unlikely, since overall triad recognition was below that of interval recognition. In the absence of further data, interval images seem more likely to have been consulted in the triad-recognition process than vice versa. However, recognition patterns found by rank order were much clearer for triads than for intervals. A reciprocal additive-subtractive recognition process involving interval and triad images is possible. This might also explain the atypical point of

16 892 LORTON, KILLAM, & KUHN greatest confusion for minor sixths-the perfect fourth. Both firstinversion major and second-inversion minor triads contain the minor third and perfect fourth intervals. If the subjects consulted these triad images to assist in identification of the minor sixth, confusion of the minor sixth and perfect fourth may have resulted. The study sheds no light on why confusion would occur with the perfect fourth rather than with the minor third. Comparison of the interval and triad recognition studies provides some preliminary data for further study. Triads appear to be more difficult to identify accurately. The subjects produced much more sharply defined response patterns to triad qualities (no matter what the inversion) than to interval qualities. The effect of duration of presentation for intervals is not significant for intervals but is significant for triads. The effect of the mode of presentation remains relatively the same for both intervals and triads, with simultaneous mode of presentation in both studies appearing to be of a different order of difficulty and perhaps requiring different recognition skills. Preliminary evidence indicating preexistent auditory imagery emerged in both studies. The effect of this on general auditory theory will need to be assessed through further examination and documentation of these findings. In the meantime, certain commonsense suggestions for the teaching of ear-training can be given. An effort should be made to assure that students have clearly defined auditory images of basic musical concepts. For instance, students with no real auditory concept of an augmented triad will probably have difficulty differentiating the augmented from other triad qualities. The ability to discern that there is a difference in quality does not appear sufficient to identify that quality accurately. The idea that descending musical constructs are more difficult to identify than ascending ones is in no way confirmed by this study. Rather, this study indicates that ascending and descending interval and triad patterns represent similar auditory identification tasks. The points of greatest confusion remained the same throughout the mode of presentation of triads, with the exception of descending augmented triads, which were more frequently confused (by 2 percentage points) with minor triads. Differing points of greatest confusion occurred on four intervals from ascending to descending. The identification of simultaneous (as opposed to ascending and descending) intervals and triads presents a more difficult task, perhaps one requiring different auditory skills. The loss of accuracy was more marked in intervals than in triads. The patterns ofconfusion remained the same for harmonic and melodic triads, but differed for four intervals. Individual subject differences were significant on both studies, despite all subjects' prior successful development of skill levels on relevant curriculum areas in the CAl program. The development ofgreater accuracy at

17 MUSIC 893 the level of stimulus difficulty presented in the two studies will require further research in curriculum design and implementation. Systematic studies of an area where fundamental questions of perception overlap applied questions of instructional strategy contribute to the advancement of knowledge in each with an efficiency not otherwise possible. Through studies of interval and triad perception and the subsequent confirmation of the findings with the analysis of performance on similar material in an instructional setting, more precise studies of perception can be developed and more responsive computer-delivered curriculum can be designed. REFERENCES Herrold, R. Computer-assisted instruction: A stud)' of student performance in a CAl eartrainingprogram. Unpublished DMA Final Project, Stanford University, Killam, R. N. A study of triad recognition and confusion by Stanford undergraduate music students. Unpublished DMA Final Project, Stanford University, Killam, R. N., & Lorton, P. Computer-assisted instruction in music: Eartraining drill and practice. Proceedings of the Fifth Conference on Computers in the Undergraduate Curriculum. Pullman: Washington State University, Killam, R. N., Lorton, P., & Schubert, E. D. Interval recognition: Identification of harmonic and melodic intervals. Journal of Musical Theory, 1975,19, Lorton, P., & Killam, R. N. Interval and triad recognition: A comparison oftwo studies of musical perception. Paper presented at the Second Annual Meeting of the National Consortium for Computer-Based Musical Instruction, Minneapolis, August (a) Lorton, P., Killam, R. N., & Kuhn, W. A computerized system for research and instruction in music. In O. Lecarme & R. Lewis (Eds.). Computers in Education: Proceedings-IFIP 2nd World Conference. Amsterdam: North Holland, Plomp, R., Wagenaar, W. A., & Mimpen, A. M. Musical interval recognition with simultaneous tones. Acustica, 1973,29,

18