Page  00000674 Musical Pattern Design Using Contour Icons Charlie Cullen*, Eugene Coylet *Digital Media Centre, Dublin Institute of Technology charlie.cullen@dit.ie tSchool of Control Systems & Electrical Engineering, Dublin Institute of Technology eugene.coyle@dit.ie Abstract This paper considers the use of basic melodic shapes known as contour icons in the design and implementation of musical patterns, for the purposes of detection and recognition in a data Sonification. Existing work in the field (such as that concerning earcon design) has considered the mechanisms by which patterns may be made distinctive, but it is argued that separate consideration must be given to the method of making such patterns memorable. This work suggests that while segregation and detection can best be facilitated by the individuality of a patterns rhythm and timbre, the retention (and hence future recognition) of a musical pattern is concerned more with features such as melodic range and contour. The detection and comprehension of contour icons was tested in comparison to a set of low level reference patterns based on earcon design guidelines (although not earcons themselves). Results show that significant improvement was made due to the use of contour icons, with future work focusing on the many possibilities that such a design framework suggests. 1 Introduction Sonification concerns the delivery of data and information using non-speech audio (G Kramer 1997), and indeed many differing methods of such delivery have been considered. The speed at which stimuli can be detected by the auditory cortex- 2ms for audio (Kail and Salthouse 1994) compared to 100ms for vision- suggests that the delivery of information using Sonification could be far more efficient than with existing visual methods. Allied to this is the sensory independence of the hearing mechanism (Kramer 1994), allowing other (perhaps unassociated) tasks to be performed in tandem in a manner that is not possible using purely visual mechanisms. In this paper, musical patterns based on visual templates- known as contour iconsare presented. As Contour icon patterns mimic basic shapes and structures, they provide listeners with a means of categorising them in a high level manner. 1.1 Auditory Display McGookin (2004) defines an auditory display as "the use of sound to communicate information about the state of an application or computing device to a user". The roots of auditory displays can arguably be traced back to alarm and alert mechanisms, seeking to inform the user of an important or urgent condition. Such displays utilise many of the advantages of audio information delivery, notably the operation of an 'eyes-free (Brewster et al 2003)' interface. Focus independent systems of information delivery have been implemented in situations such as flying a plane (Newell 1995) or driving a car (Vargas and Anderson 2003) as an essential means of processing information, and in so doing highlight on of the main advantages of auditory display. 2 Existing Pattern Design Methods Work in auditory displays with earcons (Blattner et al 1989) has led to the definition of specific guidelines for the design of musical patterns in Sonification. Although earcons are not directly used in this research, the means by which memorable musical patterns can be produced must first consider existing methods. It is the contention of this work that although a pattern can be made detectable using low level features, it does not guarantee that pattern may be sufficiently memorable. Contour icon pattern design methods include higher level cognitive features (melodic contour) to make patterns more memorable than those designed using low level methods. To this end, the earcon design guidelines are used to define a set of low level patterns for use in comparison during testing. Earcon design guidelines (Brewster et al 1995) are primarily intended for use in multiple stream concurrent auditory display (McGookin 2004), and so several of the guidelines do not directly relate to investigations into the effectiveness of single pattern Sonification. This again serves to highlight the limitations of low level pattern design guidelines, as many of the features used in contrast relate to different auditory streams (Bregman 1999) rather than purely musical attributes. The guidelines consider 674

Page  00000675 different musical attributes in turn to define an overall system for earcon creation: Timbre. Musical timbres with multiple harmonics should ideally be used to allow straightforward judgements to be made between them. Discrete sounds are ideal for short, rhythmic patterns but may not be sufficient for longer events such as chord intervals or drones. Register. Register denotes the octave of the patterns used. It is recommended that register is not a singularly reliable means of differentiating between patterns. If used, it is suggested that gaps of 2 or 3 octaves will lead to better recognition. Pitch. The pitch range used should ideally fall in the range 125-150Hz up to 5 kHz. As with register, pitch is not a reliable means of recognition in isolation, but can be utilised effectively along with rhythmic patterns. Rhythm and Duration. Rhythmic patterns used should ideally be as different as possible. The use of different numbers of notes in patterns has been found to be very effective. A recommended guideline suggests a note length of no less than 0.0825 seconds, although in short patterns (of no more than 2 or 3 notes) notes as short as 0.03 seconds can be considered. It is also suggested that earcons containing up to 6 notes in 1 second can be used as patterns. Intensity. The relative volumes of patterns should be kept within a narrow range (between 10 and 20dB above the background threshold) and indeed intensity should not be considered as means of differentiating between earcons. Spatial Location. Spatial location can either be defined by a standard left-centre-right panning attribute, or by more complex spatialisation hardware or software as available. Although earcons themselves are not directly employed, the principles of musical pattern design detailed by the above guidelines are essential to any robust method. In creating a set of musical patterns conforming to these guidelines, it was intended to produce contrast with patterns including the high level feature of melodic contour in their design. 3 Melodic Contour Melodic contour has been considered by many musicologists such as Toch (1949) as a means of defining relative changes in pitch (with respect to time), rather than the definition of absolute values. In this manner, the shape, direction and range of a melody can all be summarized by its overall contour. Graphical contour representations were considered by composers such as Schoenberg (1967) as a means of supplementing a musical score. Schoenberg regarded these contours as waves, a sentiment echoed by Toch who described melodic patterns as combinations of waves and breaks of differing amplitude. This use of graphical notation to compliment (and analyse) the traditional score was taken further by the likes of ethnomusicologist Charles Adams (1976), who used contour as the principle classification in his study of Native American melodies. Adams suggested that a contour could be defined in terms of 4 minimal boundary pitches: initial pitch (I), highest pitch (H), lowest pitch (L) and final pitch (F). The relations between these 4 boundary pitches were summarised in 3 categories: Slope, S. Slope defines a comparison between the initial (I) and final (F) pitches as either ascending, level or descending. Deviation, D. Changes in direction between boundary pitches specify levels of deviation. Thus if all four pitches are equal then the deviation is zero, with subsequent disparities between any of the 4 giving different levels of deviation. Reciprocal, R. The direction of the first deviation (either I to H or I to L) is referred to as its reciprocal, dictating the direction of the overall contour. Using these features as a template, Adams defined 15 basic contour shapes for melodic classification. This approach performed well for defining melodies that had been reduced to groups of 4 salient pitches, and allowed Adams to define the similarities and differences between music from 2 separate Native American tribes. Contour can be considered an important part of musical memory. Dowling (1978) suggests that contour information functions separately and independently from scalar information in memory. Experiments by Edworthy (1983) showed that single pitch alterations in a melody could be detected by subjects as changes in contour- even when they were unable to define what pitch had been actually altered in the pattern. This capability is believed to be present in infancy (Chang and Trehub 1977), at a stage of development where changes in pitch cannot be recognised. It has also been shown that different brain cells are used in the processing of melodic contour (Weinberger and McKenna 1988) than are used in the detection of temporal or harmonic (Sutter and Schreiner 1991) components of music. This aspect of neural activity would again suggest that different parts of the brain are used (Zatorre 1999) in the detection and recognition of musical events: rhythmic factors being paramount in detection, while melodic contour and range (Massaro et al 1980) and (Dowling 2005) being more important in the recognition of familiar and recently learned melodies. With this in mind, the use of contour was investigated in relation to musical pattern design. Although rhythmic factors (alongside pitch and timbre) are vital to the detection of an individual pattern, it could conceivably aid subsequent recognition if factors used by long-term memory were also employed. 675

Page  00000676 4 Low Level Pattern Design Initial pattern design guidelines are taken from those used in the design of earcons (Brewster et al 1995), although the patterns produced contained no actual earcon functionality. Instead, the pattern design guidelines are used to demonstrate the lack of high level cognitive features found in patterns created in this manner- features that contour icons seek to provide. A set of low level patterns is used for comparison with contour icons, with particular focus being given to the rhythm of each pattern. By making the rhythm of each pattern as distinct and unique as possible, it is intended to create a set of patterns that could be discerned individually by all test participants. 4.1 Rhythm Pattern Matrix All patterns are designed using an overall rhythm pattern matrix (Table 1) which ensures that no two patterns have the same rhythmic signature. Pattern Time Interval (minim with semiquaver resolution) 1 2 3 4 5 6 7 8 m...em | 2 *i * i n e m of 3 -. -i~ 4 I 5*. 6 * * 8 Table 1. Rhythm pattern matrix (reference pattern design) The aim of the matrix is to provide means of distinguishing the rhythm of each pattern, removing the possibility of two patterns having the same (or similar) rhythm. This is an essential element of the earcon design guidelines, which state that a patterns rhythm is the most important factor in its detection. 4.2 Pitch and Timbre Although timbre is considered a very effective means of distinguishing between patterns, this distinction is made from the perspective of concurrent auditory display. As a result, changes in timbre would hinder assessment of pattern design methods involving contour (by providing additional contrasts). Instead, the use of a simple piano timbre was considered as a reliable method of testing pattern fitness. If a pattern is truly memorable, it would still be so in the absence of other distinctions such as instrumental timbre. Pitch is also an important method of distinction (particularly in conjunction with contour), though earcon guidelines state that distinctions of pitch are not reliable discriminators in isolation. For this reason, the melodies used in each low level pattern vary in pitch purely as a means of avoiding similarity. Each pattern is also defined by boundary pitches (5.1), to avoid any confusion about the beginning or end of any pattern. 4.3 Other Design Guidelines Earcon guidelines state that register is not singularly effective as a means of pattern discrimination. As a result this feature is not employed during testing. Also, as single patterns ae used in testing, distinctions of intensity and spatialisation were not considered relevant. Although important features in concurrent pattern Sonification (McGookin 2004) the location or volume of a particular pattern would not serve to segregate it from its neighbours on grounds of pattern design and so these features are not used. These determinations led to the creation of a set of low level reference patterns (Figure 1) for the purposes of comparison during testing. The next stage of the process concerns the production of a second set of patterns including the factor of melodic contour. M IN F 1 e t IT Figure 1. Low Level patte. set 676

Page  00000677 5 Contour Icon Pattern Design Contour icons are designed in a similar manner to low level patterns, with the additional specification of a melodic contour based on a simple shape. The work of Schoenberg (1967) and Adams (1976) has shown that high level graphical representations of melodic patterns can be employed in musical description and analysis. Although not singularly defining as a factor, the relational aspects of a visual representation of musical patterns suggest definite advantages- particularly for non-musicians. In this manner, a more robust framework for pattern design could be considered which would ideally be transparent to all listeners (regardless of musicianship skills). 5.1 Boundary Pitches The 4 boundary pitches used by Adams in his contourbased classification system are used to give definite anchors to the contour of the overall patterns being analysed. These boundary pitches allow the shape of the pattern to be accurately specified from point to point- a useful framework for contour design. This use of boundary pitches also suggests benefits when seeking to create a set of patterns as individual from each other as possible. It was considered best practice to use the idea of boundary pitches as part of a design template for any contour icon set (Figure 2). pitches were also made for the low level patterns as considered in section 4. 5.2 Gestalt Grouping Checklist Bregman (1999) suggests that visual representations of audio are of great benefit in description and analysis, with many of the gestalt laws of grouping by proximity being equally applicable to both visual and audio events. Indeed, the gestalt groupings can be considered as a checklist (Brewster 1994) of pattern design which is particularly useful for high level patterns. If a pattern can be said to conform to any or all of these categories then it may be fairly considered as being distinctive and memorable. Familiarity. By using simple shapes, contour icons can take advantage of the previous exposure to visual shapes that are taught from a scholastic level (Sanford 2005). This familiarity with the visual shape can thus be used as a frame of reference for the shape of the melodic pattern of the contour icon. Good continuation. Contour icons follow melodic shapes, an attribute considered important in audio pattern recall (Herbert and Peretz 1997). The melodic shapes used change in an expected manner, and so exhibit the good continuation required for an audio stream. Belongingness. The shape of a contour icon is its melodic definition, and so the pattern can be said to exhibit belongingness when considered in terms of that shape. Articulation. Patterns which can be considered familiar are also arguably articulate, in that they can be detected more efficiently than a new source. These categories show that a well designed contour icon can potentially provide higher level features which would allow its effective stream segregation. The specification of simple visual shapes as the basis of musical patterns thus led to the design of the set of contour icons. The initial earcon design guidelines (Brewster et al 1995) employed for the low level patterns are used alongside a melodic template defining the overall shape of each contour icon. The design of each pattern is performed in stages, to ensure that gestalt categories of grouping are adhered to without producing confusing patterns. The pattern set utilises simple contour shapes, with amendments being considered subsequent to full testing of the principle. The final contour icon pattern set is defined using the low level pattern detection features of rhythm and boundary pitches, alongside the high level recall feature of melodic contour. It was hoped that patterns designed using higher level melodic features would prove easier to recall during testing. With a contour icon pattern set designed (Figure 3), a set of tests was undertaken to assess their effectiveness in relation to the low level reference patterns. I F (L) Figure 2. Example contour icon defined by boundary pitches Bass Initial Final Highest Lowest Pattern (I) (F) (H) (L) Up A2 A4 A4 A2 Down E4 E2 E4 E2 Trough C4 C4 C4 C3 Peak F3 F3 Ab4 F3 Leap down G3 F#2 G3 F#2 Leap up C#3 Eb4 Eb4 C#3 Non-1 down Ab3 F2 Ab3 G2 Non-1 up Bb2 D4 D4 Bb2 Table 2. Boundary pitch table for contour icon pattern set The boundary pitches used (Table 2) ensure that each pattern differs in overall pitch characteristics from its counterparts- alongside its unique melodic contour. In this manner, it is less likely that users would struggle to detect the beginning and end of each pattern used in a Sonification. In order to maintain consistency, definitions of boundary 677

Page  00000678 .............................................................................................................................................................................................................................................................................. AF..................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................... k", 91&-................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................ 7W '................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................... 1st 1st 2nd 2nd Training Testing Training Testing Session Session Session Session Group 1 Low level Low Contour Contour patterns level icons icons patterns Group 2 Contour Contour Low Low icons icons level level I_ patterns patterns Table 3. Testing procedure for the low level pattern vs. contour icon experiment All tests were performed using Sonifications of 2 up to 6 patterns, with the number of patterns identified being the dependent variable. The workload placed on participants by each condition was also of interest, and so NASA TLX questionnaires (Hart and Staveland 1988) were filled in by participants after completing each condition. 6.2 Training Phase The training phase was used to familiarise participants with the musical patterns they would be using. Each training phase also contained a tutorial on the Sonification method as it was employed during testing, followed by a brief period in which participants could ask questions about any aspect of the testing procedure they wished. Participants were allowed to audition the 8 patterns used as many times as they required. Once familiar with the patterns a brief listening test was performed, where all 8 available patterns were played to participants once at random. If a participant was unable to identify all 8 patterns they were allowed further time to audition them again. After this period a second listening test was performed, and if a participant was still unable to identify all patterns they took no further part in the experiment. In this experiment, all participants identified their pattern choices within 2 attempts as required. Participants were played an example Sonification, accompanied by a visual diagram of the patterns they were listening to. In the low level pattern condition, participants were shown a list of numerical icons. In the contour icon condition, participants were given an accompanying list of contour icons denoting the patterns they were hearing. After the example Sonification, a further brief period was allowed for any other questions participants had about the experiment. 6.3 Testing Phase In the testing phase participants were asked to listen to Sonifications of between 2 to 6 patterns. Using the TrioSon application (Cullen and Coyle 2005) and (Figure 4), participants were played various predetermined pattern Sonifications. 6 Contour Icon Testing The aim of testing was to assess the affect on pattern recognition of musical patterns designed with (contour icons) and without (low level patterns) specific melodic contours. The earcon design guidelines (Brewster et al 1995) used to create all test patterns were further augmented by basic contour shapes to aid in their recognition. The test schedule assessed point estimation performance for participants using low level patterns and contour icons. 6.1 Procedure 20 participants took part in the experiment described in this section which was of a within groups design involving two conditions, the low level pattern condition and the contour icon condition. No participants were taken from formal music courses as the tests sought to assess the pattern memory of listeners without prior exposure or musical training. Participants were randomly assigned to one of two groups to determine the order in which they would undertake the experiment. Each group contained the same number of participants, and both conditions consisted of training and testing phases (Table 3). 678

Page  00000679 Isr~~....... Figure 4. TrioSon Sonification Application All Sonifications were performed on a Compaq NX6100 laptop, using the onboard ADI AC97 soundcard. All tests were carried out using the General Midi (Rigas and Alty 1998) piano sound provided by the ADI AC97 synthesis engine. Participants were asked to identify the number of times they heard each pattern. The independent variable in testing was the method of pattern design (low level patterns or contour icons) with the dependent variable being the number of correct point estimation questions answered. All tests were performed with participants being handed writing materials after listening to a Sonification, to prevent the scoring of patterns prior to testing. As pattern design was the testing variable, it was crucial that all participants determined the patterns used based only on information provided during testing. Participants were allowed to listen to a Sonification once for each part of a question, ranging from 2 passes through to 6 for the last question of each test. After the 5 test questions were answered, participants were asked to answer post-test TLX questionnaires to determine how difficult they had found the process. 7 Overall Results Overall results (Figure 5) show performance improves from 44% in the low level pattern condition to 56.87% in the contour icon condition. This is a significant improvement (T(20) = -3.68, p 0.0007) in point estimation performance between test conditions, suggesting that contour icons are more memorable than low level patterns. 7.2 Effect of Pattern Count on Point Estimation Point estimation tests were performed for 2 through to 6 different patterns (Figure 6), to determine how well participants could recognise the patterns they were asked to detect. V.- C-t by Grp Figure 6. Graph showing average percentage scores by value count (for each test condition), showing standard deviations Test results show that performance in the 2 pattern contour icon condition had reached an average of 87.5% (compared to 68.75% for low level patterns), and had decreased through 80.83% (65% low level) to 66.87% for the 4 value test (55.62% low level). This is considered to be a reasonable level of contour icon recognition given that there had been no exposure to them prior to testing. Having said this, performance in the 5 and 6 pattern conditions dropped from 45.5% (34% low level) to 37.5% (25.83% low level), which is still too low for effective recognition. Future work (section 10) will consider methods of designing larger groups of recognisable contour icon shapes. 7.3 Post Test TLX Results The overall TLX results (Figure 7) show a significant reduction (T(20) -4.53, p<0.0001) in overall workload from 50.33 (low level) to 36.25 for the contour icon condition. 10 -40 Group Figure 5. Graph showing overall average percentage scores (by test condition), showing standard deviations Test Condition Figure 7. Graph showing overall average post-test TLX scores (by test condition), showing standard deviations

Page  00000680 TLX Category Figure 8. Graph showing average post-test TLX scores by category (for each test condition), showing standard deviations Individual workload scores (Figure 8) show that mental demand was considered the most important aspect of testing, with a reduction in average from 14.72 (low level) to 12.57 for the contour icon condition being observed. Though no reduction was significant in any individual category, the scores were lower for the contour icon condition in each case. No problems were encountered during explanation of the test schedule, and participants understood the principle of Sonification as it related to the tests. 8 Discussion The contour icon tests show that significant improvement can be made when melodic contour is employed as a feature in pattern design. Testing had employed non-musicians in a series of point estimation questions, which show that contour icons can achieve average recognition rates of 66.87% in 4 pattern Sonifications. Higher pattern counts have proven less effective, and so further work is needed to improve the pattern set. Participants were comfortable with the use of shapes as descriptors of musical patterns, and often suggested other possible shapes during training. This is considered an indicator of the potential of contour icons, in that no musical knowledge or training is required to recognise a melodic shape. Post test TLX questionnaires show a significant reduction in overall workload due to contour icons, with all individual categories exhibiting a reduction (although not individually significant). This suggests that contour icons are a more effective method of pattern design than low level patterns, providing participants with a straightforward means of pattern recognition and recall. 9 Conclusions Contour icons show significant improvement in performance during testing, with recognition rates of 87.5%, 80.83% and 66.87% for 2, 3 and 4 pattern tests respectively. These results show that musical patterns designed using high level contour features are more memorable than low level patterns when both employ distinctions of rhythm, register and timbre. Contour icon shapes are effective in conveying the melodic contours they represent, with suggestions for other contour icons being made by participants during training. Contour icons also produce a significant reduction in workload during testing, again suggesting greater effectiveness than the low level patterns. 9.1 Limitations of Contour Icons Although contour icons perform well in 2 to 4 pattern tests, this level of recognition was not observed for higher pattern counts. As a result, contour icons cannot be considered any more effective than low level patterns in these cases. This reduction in point estimation performance is considered to be indicative of one the limitations of contour icon Sonification, in that tests with high pattern counts will require greater training and exposure than can be afforded during normal testing. 10 Future Work The development of contour icons suggests means by which musical patterns may be made both distinct and memorable by including high level cognitive features, and further work is required to determine what other additions could be made. Musical contour was chosen as an effective high level descriptor of a pattern, and the shapes used for contour icons are by no means comprehensive (or indeed conclusive). Future work will investigate the use of shape more thoroughly, to consider: What shapes are possible. Basic shapes such as up and down arrows prove effective, but many other shapes are possible. Further work could approach this from 2 directions: the analysis of existing popular melodies to determine which shapes are most common- in a manner similar to Adams (1976)- and the design of contour icons based on recognisable visual shapes. How distinct can contour icons be made from one other. Although development produced a set of contour icon shapes, no study was undertaken to assess how distinct or similar those icons were from each other. Further work could consider how to maximise the distinction between icons, utilising features such as boundary pitches and perhaps timbre to create the most disparity between each icon in a set. Can contour icons be concatenated to create more complex shapes. Earcons are modular in design, with more complex patterns containing greater information being constructed from smaller musical units. The same approach could be considered with contour icons, investigating whether contour icons can be concatenated to form more verbose patterns. Similarly, an overall shape (such as up or 680

Page  00000681 down) could be used as the basis of a family of icons, where changes in other pattern features would denote a new icon within a larger sub-group. References Adams C. 1976. Melodic Contour Types. Ethnomusicology, vol. 20, no. 2, pp 179-215. Blattner M, Sumikawa D and Greenberg R. 1989. Earcons and icons: Their structure and common design principles. Human Computer Interaction, vol. 4, no. 1, pp. 11-44, Bregman AS. 1999. Auditory Scene Analysis, the Perceptual Organisation ofSound. MIT Press 1999, ISBN 0262521954. Brewster SA. 1994. Providing a structured method for integrating non-speech audio into human-computer interfaces. PhD Thesis, University of York. Brewster SA, Wright PC and Edwards DN. 1995. Guidelines for the Creation of Earcons. Proceedings of BCS-HCI 95, vol. 2, pp. 155-159. Brewster SA et al. 2003. Multimodal 'Eyes-Free' Interaction Techniques for Wearable Devices. In Proceedings of ACM CHI 2003 (Fort Lauderdale, FL), ACM Press, AddisonWesley, pp 463-480. Chang H and Trehub SE. 1977. The Audio Processing of Relational Information in Young Infants. Journal of Experimental Child Psychology, no. 24, pp. 324-331. Cullen C and Coyle E. 2005. TrioSon: A Graphical User Interface for Pattern Sonification, The 11th Meeting of the International Conference on Auditory Display, ICAD 05, Limerick, Ireland, 6-9 July. Dowling WJ. 1978. Scale and Contour: Two Components of a Theory of Memory for Melodies. Psychological Review, vol. 85, no. 4, pp. 341-354. Dowling WJ. 2005. Tonal Strength and Melody Recognition after Long and Short Delays. Program in Applied Cognition and Neuroscience, University of Texas, Dallas. Edworthy J. 1983. Towards a Contour-Pitch Continuum Theory of Memory of Melodies. Acquisition of Symbolic Skills, zPlenum Press. Essens PJ. 1995. Structuring Temporal Sequences: Comparison of Models and Factors of Complexity. Perception and Psychophysics, vol. 57, no. 4, pp.519-532. Gaver W. 1989. The SonicFinder: An interface that uses auditory icons. Human Computer Interaction, vol. 4, no. 1, pp. 67-94. Hart S and Staveland L. 1988. Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. Human Mental Workload, pp. 139-183. Herbert S and Peretz L. 1997. Recognition of Music in Long-term Memory: are Melodic and Temporal Patterns equal Partners? Memory Cognition, vol. 25, no. 4, pp 518-533. Kail R and Salthouse TA. 1994. Processing speed as a mental capacity. Acta Psychologica, vol. 86, pp. 199-255. Kramer G(ed) et al. 1997. Sonification Report: Status of the Field and Research Agenda. International Conference on Auditory Display (ICAD). Kramer G. 1994. An introduction to auditory display. Auditory Display: Sonification, Audification and Auditory Interfaces. Addison-Wesley, Reading, Massachusetts, pp 1-77. Massaro DW, Kallman HJ and Kelly JL. 1980. The Role of Tone Height, Melodic Contour and Tone Chroma in Melody Recognition. Journal of Experimental Psychology (Human Learning), vol. 6, no. 1, pp. 77-90. McGookin D. 2004. Understanding and Improving the Identification of Concurrently Presented Earcons. PhD thesis, University of Glasgow. Newell AF. 1995. Extra-ordinary human-computer interaction. Extra-Ordinary Human-Computer Interaction: Interfaces for Users with Disabilites, Cambridge University Press, Cambridge, pp 3-18. Rigas D and Alty J. 1998. How Can Multimedia designers Utilise Timbre? People and Computers XIII, Springer-Verlag, ISBN 3540762612. Schoenberg A. 1967. Fundamentals of Music Composition. Faber and Faber Ltd. Sutter ML and Schreiner CE. 1991. Physiology and Topography of Neurons with Multipeaked Tuning Curves in Cat Primary Auditory Cortex. Journal of Neurophysiology, vol. 65, no. 5, pp. 1207-26. Stockburger DW. 1998. Introductory Statistics: Concepts Models and Applications, v 1.0. Online book, Southwest Missouri State University. Toch E. 1948. The Shaping Forces in Music. Criterion Music Corp. Vargas MLM and Anderson S. 2003. Combining speech and earcons to assist menu navigation. International Conference on Auditory Display (ICAD), Boston, Massachusetts, pp 38-41. Weinberger NM and McKenna TM. 1988. Sensitivity of Single Neurons in Auditory Cortex to Contour: Towards a Neurophysiology of Music Perception. Music Perception, no. 5. Sanford. 2005. Shape Art. Online Document, Available http://www.sanford-artedventures.com/study/g_shape.html. Zatorre RJ. 1999. Brain imaging studies of musical perception and musical imagery. Journal of New Music Research, no. 28, pp. 229-36. 681