Proceedings paper

 

ADOLESCENT ENGAGEMENT IN COMPUTER-BASED COMPOSITION: AN ANALYSIS OF THE PROCESS OF COMPOSITION

Frederick A. Seddon and Susan A. O'Neill, Department of Psychology, Keele University

Background. Past research suggests that during computer-based composition, musically trained adolescents experiment less with possibilities offered by the computer and produce compositions with more 'fixed ideas' about creating music than untrained adolescents (Scripp, Meyaard and Davidson, 1988; Folkestad, 1998). Although a recent study with 11 year olds found no difference in teachers' ratings of computer-based compositions by children with and without experience of formal instrumental music tuition (FIMT), the two groups of children appeared to adopt different strategies when engaged in the computer-based composition task (Seddon and O'Neill, 1999). Further research is needed which examines these differences. Additionally, the growing use of computer technology in music education makes it necessary for teachers to be aware of the different ways in which adolescents might engage with this technology based on their prior musical experience. Thus, the present paper outlines the development of a method for investigating and analysing the compositional process used by adolescents engaged in a computer-based composition task.

Investigation of the process of composition may be conducted by observing the composer while composing (Sloboda, 1985) or by the composer giving a verbal report during or after composition (Richardson & Whitaker, 1996). Past studies have used either or both of these approaches. Observation, even through the use of a video camera (Odman, 1992; Daignault, 1996) runs the risk of interfering with the process of composition by producing a 'surveillance effect'(Hickey, 1997). Concurrent 'verbal report' runs the risk of interfering with the composition process, and retrospective 'verbal report' relies extensively upon memory. In addition to these disadvantages 'verbal report' assumes composers are: a) aware of the process they are engaged in and, b) able to articulate full and accurate accounts of the process. These methodological disadvantages may be particularly significant in relation to novice or relatively inexperienced composers such as adolescents.

Computer technology has made it possible to record the development of compositions by saving 'midi files' at different points in time creating 'snapshots' of the composition process (Hickey, 1995; Folkestad, 1998). This data collection method is analogous to methods adopted in a study by Getzels and Csikszentmihalyi in 1976 where photographs were taken every six minutes recording the development of visual artwork. However, recording the composition process by saving 'midi files' may still omit important data (e.g., ideas experimented with but never reaching the recording stage). Our aim was to develop a method of data collection which would reduce 'surveillance effects' whilst at the same time enabling the full compositional process to be examined.

Method. The method involved asking participants to engage with a computer-based composition task after two scripted thirty minute training sessions. The training sessions were scripted to control for variations in training and focused on how to use the composition program but did not provide any instruction in the compositional process itself. No musical examples were given that could have implied 'correct models' to copy. Having been instructed to 'compose a piece that sounds good to you', participants worked with a Yamaha PSR 530 music keyboard connected via MIDI to a computer with a researcher modified version of Cubase Score composition software program installed. Following on from the two training sessions, participants had three individual 30-minute composition sessions on three consecutive days in order to complete their composition. The composition sessions took place in a room designated solely for the use of the participant to ensure privacy. Participants were asked to choose three sounds from ten sounds available. During the composition sessions all 'on screen manipulations' of the program were unobtrusively recorded to videotape through a 'video-card' installed in the computer. In addition to this videotape data, 'midi files' were saved using different name references via the 'save as' method (e.g., David 1, David 2 etc.) for each participant at the end of each composition session. This would make it possible to cross-reference the videotape date with the 'midi files' at three specific points in time.

Data analysis. The method of analysis used is adapted from qualitative inductive analysis of text where what is important to analyse in the data emerges through a process of inductive reasoning from the data itself rather than being grouped according to predetermined hypothetical categories (Maykut and Morehouse, 1994). This method is based on the notion of grounded theory (Glaser and Strauss, 1967) which is a theory 'inductively derived from the study of the phenomenon it represents' (Strauss and Corbin, 1990, p.23) and differs from traditional scientific method where hypotheses are generated indicating the relevant data to be collected. These hypotheses are then tested by mathematically analysing the data (Goertz and LeCompte, 1981). The substantial quantity of data collected in this study makes the constant comparative method (Glaser and Strauss, 1967; Lincoln and Guba, 1985) the most appropriate method of inductive analysis for our purposes. The constant comparative method of analysing qualitative data involves the identification of 'units of meaning' that emerge from the data and are coded by comparison to similar 'units of meaning' into 'categories'. When applied to text this involves the transfer of raw data such as audio-taped interviews into clearly readable text from which 'units of analysis' are sought and coded with reference to their original source. These units of analysis are then grouped into categories containing units of analysis of similar meaning. If a unit of analysis emerges that can not be included in an existing category then a new category is formed to accommodate it leaving room for continuous refinement in the process of analysis. Repeated reviews of the data create a high degree of familiarity with it facilitating the categorisation process. The next stage is to formulate a 'rule for inclusion' for each category which units of analysis must fulfil for inclusion in that category. The rule for inclusion is a propositional statement drawn from the units of analysis already assigned to the categories. Assigning units of analysis to a category is then achieved by units of analysis fulfilling the rule for inclusion in the propositional statement. Some of these propositions will stand alone and some will require grouping together to become outcome propositions contributing to an understanding of the investigation.

During an initial exploratory review of the video data the videotapes were transcribed into general descriptions of 'activities' engaged in. This first transcription process lead to the formulation of 'analysis codes' to facilitate the identification and classification of observable 'activities'(see appendix A). A second transcription of the videotapes made with the aid of the 'analysis codes' resulted in 'coded transcripts' with sequential numbered 'events' containing coded 'activities'. These 'activities' are analogous to 'units of analysis' in constant comparative method.

'Activities' ('units of analysis')

These observable 'activities' fall into two styles 'on task' (when the participant is clearly engaged with the presented composition task) or 'off task' (when the participant is clearly following a different agenda). The main 'on task' activities were labelled as: 'playing the keyboard', 'recording with the keyboard', 'replay' and 'editing'. Descriptions for each of these activities and 'off task' activities are given below (in Table 1).

Table 1

 

It is possible to engage in these' on task' activities with or without employing the replay facility available in the 'Cubase' composition program. This replay facility enables previously recorded parts to be heard during 'on task' activities. The 'click' (an electronic metronome device to assist performance in strict time) may be used as a substitute for replay of a previously recorded part if preferred. If replay (or 'click') facilities are engaged during 'on task' activities this is labelled as engaging with 'aural reference', if replay (or 'click') facilities are not engaged during 'on task' activities this is labelled as engaging without 'aural reference'. The second transcripts were then coded using the 'analysis codes'. The 'coded transcripts' containing sequentially numbered coded 'events' were also colour co-ordinated for the type of instrumental sound used (see Table 2)

Table 2

 

Coding of transcript

Instruments chosen: Sound one, bass and drums

Session one

Event I: PK/NAR

Playing keyboard on drum sound experimenting with different sounds.

Event 2: PK/NAR

Playing keyboard on bass sound experimenting.

Event 3: PK/NAR

Playing keyboard on sound one experimenting

 

 

Using 'cut and paste' techniques a 'construction of parts' document was produced from the 'coded transcripts' for each participant in order to trace the development of each instrumental sound part in the sequential order of 'events' (see Table 3).

Table 3

 

Construction of parts

Sound one part

Session one

Event 3: PK/NAR

Playing keyboard on sound one experimenting.

Event 5: PK/NAR

Playing keyboard on sound one, playing theme that is eventually used.

Event 8: PK/AR

Playing keyboard on sound one with 'click'

 

 

Using 'midi files' a 'musical score' of each instrumental part (as it was saved at the end of each recording session) was produced to allow for comparison and cross referencing with the 'construction of parts' document (see Example 1).

By cross referencing all three documents (coding of transcript, construction of parts and musical scores) it was possible to trace sequentially the 'activities' involved in the development of each instrumental part to include not only musical material that was recorded and retained but also musical material that had been discarded. This gave a very detailed record of the composition process for each participant. Examination of these detailed records revealed emerging patterns of behaviour enabling the formulation of propositional statements leading to 'rules of inclusion'. Typical examples of two of the emerging patterns are described below along with their propositional statements and 'rules of inclusion'.

Example of 'Pattern A' composition strategy.

Instruments chosen are: piano, strings and cello.

Session one

Event no 1 (PK/NAR) the participant begins by playing the keyboard on piano sound. It is noted that during this event a melody theme emerges that is recognisable as being 'related' to the final piano part. Two attempts are made to record this same theme and are deleted. The third recording of the same theme (Event 9) survives, is volume edited at Event no 10. and is then as it appears in the final composition. (see Ex. 1)

 Ex. 1.

 

Neither of the other two sounds (strings and cello) are engaged with until the piano part is completed. At Event 11 (RK/AR), the participant begins recording the cello 'accompaniment' with aural reference to the piano part. This first recording of the cello part is made without prior experimentation and is deleted. The keyboard is played on cello sound developing the same part. The participant chooses not to aurally reference the piano part while developing the cello part. Event 15 RK/AR results in the 14 bar cello part recorded with aural reference to the piano part. This part is subsequently reduced to 9 bars through a 'cut/delete' note edit of bars 9-14 (Event 17) and remains as in (Ex. 2) until it is deleted at the beginning of session two (Event 43).

Ex. 2.

The string part is started at Event 18: PK/NAR (i.e., playing keyboard on string sound), as with the cello part the participant chooses not to aurally reference the previously recorded parts. Three recordings of the same string part are made with aural reference to the piano and cello parts and are subsequently deleted before the fourth recording of the same part is made. This recording is also made with aural reference to the piano and cello parts. The fourth recording remains but the last three bars 13-15 are 'cut/deleted' in a note edit (Event 32). The string part remains as in (Ex 3) until it is deleted in session two, (Event 54).

Ex. 3.

 

Session two

After replay the cello part is deleted (Event 43) re-recorded this time four bars longer (bars 10-13). This extension to the part is made without any prior keyboard playing activity. Two recordings are deleted before the third is accepted (Event 49) all recordings are made with aural reference to the piano part but with the strings muted. The cello part is now as in (Ex. 4).

Ex. 4.

Comparison of (Ex. 2) with (Ex. 4) reveals that although in bars 1-9 the notation of the part appears different the sound remains unchanged. The cello part remains unchanged from (Ex. 4) in the final composition.

Having completed the cello part the participant resumes work on the string part. The string part is deleted (Event 54) and re-recorded with a note change and slight extension to the part. Two recordings are rejected before the third one is accepted (Event 58) see (Ex. 5). All recordings are made with aural reference to the piano and cello parts.

 

Ex. 5.

Comparison of (Ex. 3) with (Ex. 5) reveals that although in bars 1-9 the notation of the part appears different the sound remains unchanged. The string part remains unchanged from (Ex. 5) in the final composition. There are no changes made to the piano part in this session.

Session three

There are no events during this session that make changes to any of the parts. The session is spent almost exclusively in 'off task' activity in particular playing recogniseable tunes e.g., 'We Three Kings', 'Little Drummer Boy', and 'Super Trouper'.

 

This typical example of 'Pattern A' is charaterised by its lack of experimentation. The melody appears during the first event and although there are three recordings of the melody made these were made to correct performance errors. The recording of this melody which is completed from beginning to end in one recording shapes the whole composition from this point, all that remains is for the participant to harmonise it. The participant accomplishes this harmonisation during the second session displaying little experimentation with possible alternatives. The composition being completed by the end of the second session leaves the third session available for any chosen activity. The participant chooses to engage with performance skills rather than engage with the creative possibilities made available by the computer. It is interesting to note that aural reference to previously recorded parts is not made while playing the keyboard (which could have offered an opportunity to experiment with alternatives) but aural reference is made during recording with the keyboard to guide performance. This is exemplified by the sequence of coded events: PK/NAR- RK/AR- RP/AR followed by a considered response to the outcome. The participant's ability to re-record the same parts on different occasions indicates that the parts remain in aural memory and could indicate that aural reference was being made covertly rather than overtly during playing the keyboard events.

Propositional statement for 'rules of inclusion' for 'Pattern A'

'Pattern A' is characterised by the predominant use of 'practicing' techniques rather than 'improvising' techniques during 'playing the keyboard' activities. Recognisable elements of the final composition appear early in the composition process after few ideas have been introduced. Recording with the keyboard will mainly consist of recording similar ideas correcting performance errors rather than experimenting with alternative ideas. Individual parts will be composed from beginning to end before engaging with other parts usually completing the melody first. Aural reference is employed during 'recording' but not during 'playing the keyboard' activities. The participant will be more likely to engage with 'off task' keyboard playing.

 

Example of 'Pattern B'composition strategy.

Instruments chosen are: sound one, bass and drums.

Session one

The first four events are playing keyboard events (PK/NAR) all three sounds are experimented with separately. The first surviving idea appears on 'sound one' at Event 5 (PK/NAR). The tempo is changed at Event 7 (TC). The first 14 bars of the eventual 'introduction' on 'sound one' are recorded at Event 9 (RK/AR). This 'introduction' section is recorded with aural reference to the 'click' after some prior practice. Recording the drum part begins at Event 11 (RK/AR) with aural reference to the 'click' and 'sound one'. The first recording is deleted and the second (a 5 bar drum part) is recorded with the 'click' at Event 15. This short part proves to be an experiment which is later deleted (Event 22). At this point the participant leaves the drum part and resumes work on the 'sound one' part extending it by three bars at Event 26 (RK/AR). The part is extended by 'overdubbing' (recording a new section to a part by recording over it without erasing the original part). The sound one part is now as in (Ex. 6).

Ex 6

 

After deleting the 5 bar drum part (Event 22), a period of experimenting on drum sound with and without aural reference to 'sound one' and 'click' results in a 25 bar improvised drum part (see Ex. 7). This part is 8 bars longer than the existing 'sound one' part and is recorded at Event 33 (RK/AR) making aural reference to the 'sound one' and 'click during recording.

Ex. 7.

No bass part is recorded during session one but two events Event 2 (PK/NAR) and Event 28 (PK/AR) reveal experimentation with and without aural reference.

Session two

Bars 18-25 of the session one drum part (Ex. 7) are deleted after replay at Event 37. The deleted section (bars 18-25) of the original drum part is replaced and extended to 39 bars by 'overdubbing' at Event 41 (RK/AR). Comparison of (Ex. 7) and (Ex. 8) reveals this change to the drum part.

 

Ex. 8.

After a brief period of experimentation with possible bass 'riffs', Event 43 (PK/NAR), a bass part is recorded with aural reference to the 'sound one' and drum parts, Event 46 (RK/AR). This recording of the bass part is from bar 21-30 but bars 28-30 are subsequently deleted in a note edit, Event 48 (ED(N)/AR). A period of practice, Events 50 and 52 (PK/AR) where the participant is playing keyboard on bass sound while the drum part is replaying, results in an extension by 'overdubbing' to the bass part from bar 28-43, Event 54 (RK/AR). Bars 36-43 of this recording had not been practiced prior to recording and were an improvised section which was subsequently deleted (Event 57). The bass part at the end of session two is bars 21-36, see (Ex. 9).

 

Ex. 9.

No change to 'sound one' part during this session.

Session three

A note edit is made to the drum part (compare bars 29-30, Ex. 8 to bars 29-30, Ex. 11) then the participant moves to work on the bass part. This begins with playing the keyboard on bass sound, Event 63 (PK/AR), first experimenting, then practicing new ideas to record. The new section of the bass part, bars 37-52 (see Ex. 10) is then 'overdubbed' making aural reference to the drum part (bars 37- 39) and extended beyond the recorded drum part to bar 52.

Ex. 10.

Participant then moves to work on the drum part by playing the keyboard on drum sound, Events 68 and 70 (PK/AR), with aural reference the latest bass recording (see Ex. 10). Eventually the drum part is extended by 'overdubbing' to bar 51, with aural reference to the bass part Event 71(RK/AR) see (Ex. 11).

 

Ex. 11.

 

The final event (Event 73) is the recording of a solo 'coda' on sound one bars 52-65 (see Ex. 12).

Ex. 12.

This typical example of 'Pattern B' is charaterised by the way the composition develops over all three sessions. The participant experiments with alternative musical material for the instrumental parts employing aural referencing techniques. This is exemplified by the sequence of coded events : PK/AR-RK/AR-RP/AR followed by considered response to the outcome. Musical material is reviewed and sections are deleted to be replaced by different material indicating that the closure of the creative process is not reached (if at all) until late in the composition process. On occasion the composition is extended by employing improvisation techniques with the previously recorded parts providing the stimulus for the current improvisation. It is doubtful that these improvised parts remain in aural memory. The absence of any 'off task' behaviours and the final event being a 'recording with the keyboard' event is further indication that the creative process is ongoing rather than completed.

Propositional statement for 'rules of inclusion' for 'Pattern B'

'Pattern B' is characterised by the predominant use of 'improvising' techniques rather than 'practicing' techniques during 'playing the keyboard' activities. New ideas are experimented with throughout the process of composition. Recording will be used to 'capture' improvisations that may or may not survive in the final composition. The composition will evolve in sections rather than each part being through composed. Aural reference will be made during 'playing the keyboard' activities in addition to 'recording' activities.

A further three 'patterns' of composition have emerged from the data: Patterns 'C', 'D' and 'E'. Broad descriptions of these patterns appear below but space restrictions do not allow for detailed examples of these 'patterns' like those provided for 'A' and 'B'.

'Pattern C'

As with 'Pattern A' individual parts are completed from beginning to end before engaging with the next part. The main difference being the harmonic structure of the accompaniment is completed first with the melody composed to fall within the harmonic boundaries created by the accompaniment.

'Pattern D'

The main focus of attention in this 'pattern' is to achieve synchronisation between all three parts in strict time. A melody is recorded from beginning to end using performance skill. On discovering the high levels of performance skill required to exactly synchronise the remaining parts to the original the remaining time is spent manipulating various editing techniques ('copy/paste') to ensure the timing of the parts is exactly the same. This can lead to parts being identical except for timbre.

'Pattern E'

This 'pattern' is predominantly random in nature lacking any observable structure in either the use of the program or composition process. Both the Cubase program and the available sounds are extensively experimented with. The 'composition' is formed by the expiration of time allowed rather than a decision being made to 'save' what has been produced.

 

Implications The data collection methods described above address the data collection shortcomings of past studies. Detailed and complete data can be collected with reduced 'surveillance effect' and without relying upon participants' memory, levels of awareness or articulation skills. The proposed procedures for analysis provide a more appropriate method for this type of data than can be achieved solely through a statistical analysis of time spent in each activity.

We are currently engaged in a study for which data has been collected. The data has been through the initial process of analysis described in this paper. This study involves 48 adolescents (aged 13-14 years). Twenty five (Female=12, Male=13) had between 2-4 years prior experience of FIMT and twenty three (Female=12, Male=11) had no prior experience of FIMT. A review of the initial non-coded transcripts suggests some of the identified 'patterns' of composition were adopted by groups of participants identified by their prior experience and/or gender (e.g., 8 of the 12 females with prior experience of FIMT adopted 'Pattern A' and 6 of the 8 participants who adopted 'Pattern B' had no prior experience of FIMT). These initial interpretations support the results of past studies identifying different ways of composing using music technology. For example 'Pattern A' corresponds closely to 'supplementary use' (where equipment is used as a tool for arranging music) and 'Pattern B' corresponds closely to 'integral use' (where equipment is used as an interactive medium and plays an integral part in composition) (Folkestad, 1991). 'Pattern A' could be described as 'Horizontal' composition and 'Pattern B' could be described as 'Vertical' composition (Folkestad, Hargreaves, and Lindstrom,1998).

 

The detailed coded analysis of data from all 48 participant is in progress and it is predicted that the results will further our understanding of the extent to which: a) adolescents employ different 'patterns of composition' similar to those described above when engaged in computer-based composition and, b) these 'patterns of composition' may be linked to prior experience of FIMT and gender.

  

References

 

Daignault, L. (1996). A study of children's creative musical thinking within the context of a computer-supported improvisational approach to composition. Unpublished doctoral dissertation. Chicago, U.S.A.: Northwestern University.

Folkestad, G. (1991). Music composition in the upper primary school with the help of synthesisers-sequencers. (Report No. 1991:19), Stockholm: Center for Research in Music Education.

Folkestad, G., Hargreaves, D. J., Lindstrom, B. (1998). Compositional strategies in computer-based music making. British Journal of Music Education (1998) 15: (1), (pp 83-97).

Getzels, J., & Csikszentmihalyi, M. (1976). The creative vision: a longitudinal study of problem finding in art. New York: John Wiley.

Glaser, B.G., & Strauss, A.L. (1967). The discovery of grounded theory. Chicago, Il: Aldine.

Goertz, J.P. & LeCompte, M.D. (1981). Ethnographic research and the problem of data reduction. Anthropology and Education Quarterly, 12, pp.51-70.

Hickey, M. (1995). Qualitative and quantitative relationships between children's creative musical thinking processes and products. Unpublished doctoral dissertation. Chicago, U.S.A.: Northwestern University.

Hickey, M. (1997). The Computer as a tool in creative music making. Research Studies in Music Education No.8 July 1997.

Lincoln, Y. & Guba, E. (1985). Naturalistic enquiry. Beverly Hills, CA: Sage.

Maykut, P., & Morehouse, R. (1994). Beginning qualitative research: A philosophic and practical guide. London: The Falmer Press

Richardson, C.P., and Whitaker, N.L. (1996). Thinking about think alouds in music education research. Research Studies in Music Education, No. 6. June 1996, pp. 38- 49.

Odman, P. J. (1992). Didactical/phenomenological aspects of creative music making with the help of computers. In Datorer I musikundervisningen (11-21), Stockholm: Center for Research Music Education.

Scripp, L., Meyaard, J., and Davidson, L. (1988). Discerning musical development: Using computers to discover what we know. Journal of Aesthetic Education, 22 (1), 75-88.

Seddon, F.A., & O'Neill, S.A. (1999). An evaluation study of computer-based compositions by children with and without prior experience of formal instrumental music tuition. Accepted for publication Psychology of Music January 1999.

Sloboda, J. A. (1985). The Musical Mind: The Cognitive Psychology of Music. Oxford University Press.

Strauss, A. & Corbin, J. (1990). Basics of qualitative research: Grounded theory procedures and techniques. Newbury Park, CA: Sage.

 

Appendix A

Analysis codes:

'On task activities'

 

'Playing the keyboard' codes

 

PK/AR Playing the musical keyboard with aural reference to previously recorded part(s) or with 'click' (similar to playing with a metronome) during replay, experimenting with and developing ideas or practising prior to recording.

 

PK/NAR Playing the musical keyboard without aural reference to previously recorded part(s) or 'click' (either because this is the first sound engaged with and 'click' has not been activated or previously recorded parts have been 'muted', or replay is not activated), experimenting with and developing ideas or practising prior to recording.

'Recording with the keyboard' codes

 

RK/AR Recording using the musical keyboard with aural reference to either 'click' or a previously recorded part(s).

RK/NAR Recording using the musical keyboard without aural reference to either 'click' or a previously recorded part(s).

'Replay' Codes

('Specific' replays that take place immediately following editing procedures are included in 'edit' coding with and without AR respectively.)

RP/AR 'Global' replay with aural reference to previously recorded parts or 'click'.

RP/NAR 'Global' replay without aural reference to previously recorded parts or 'click'.

 

'Editing' Codes.

ED(N)/AR Edits performed to change notes (or a group of notes) in time and/or pitch, erase, insert, extend, or identify, with aural reference to previously recorded part(s).

ED(N)/NAR Edits performed to change notes (or a group of notes) in time and/or pitch, erase, insert, extend, or identify, without aural reference to previously recorded part(s).

 

ED(V)/AR Edits performed to change the volume of notes (or a group of notes) with aural reference to previously recorded part(s).

ED(V)/NAR Edits performed change the volume of notes (or a group of notes) without aural reference to previously recorded part(s).

DP Deletes part

C/P Copy/paste of part.

TC Tempo change

'Off Task activities'

'Off task' behaviours codes

PI Period of inactivity (Mouse remains motionless for more than 5 seconds).

PPI Prolonged period of inactivity (Mouse remains motionless for longer than one minute).

PK/OT Playing the keyboard in 'off task' way, random, displaying 'frustration' or 'recognisable' tunes.

'Error' Codes

EUP Events resulting from 'errors' using the program, 'accidents' or actions exploring the program that have no obvious intent.

PM Program malfunctions resulting from unknown sources or misuse of the program.

PD Program defaults preventing participants actions (e.g., 'cut' mid bar or when program defaults to the start when record button is pressed.)

 

 

 Back to index