COGNITIVE INTERFERENCE IN THE PERCEPTION OF PITCH AND LOUDNESS IN A FIVE-NOTE MUSICAL PATTERN DISSERTATION. Presented to the Graduate Council of the

Transcription

1 3"7<r M8U Ao.V7SV COGNITIVE INTERFERENCE IN THE PERCEPTION OF PITCH AND LOUDNESS IN A FIVE-NOTE MUSICAL PATTERN DISSERTATION Presented to the Graduate Council of the University of North Texas in Partial Fulfillment of the Requirements For the Degree of DOCTOR OF PHILOSOPHY By Gary Thomas Cattley, B.A., M.M. Denton, Texas May 1999

2 Cattley, Gary Thomas, Cognitive Interference in the Perception of Pitch and Loudness in a Five-Note Musical Pattern. Doctor of Philosophy (Music Education), May 1999, 152 pp., 18 tables, 16 figures, references. The purpose of the study was to explore whether musicians experience Garner interference among the auditory dimensions of pitch and loudness. Specifically, the study explored whether the dimensions of intonation and loudness, when presented to musicians in a five-note musical pattern, were perceived as integral or separable in nature. Three convergent tests were employed; each used a five-note musical pattern in which the final note varied in intonation and loudness. Subjects participating in the study were musicians either currently working as professional musicians or possessing significant musical performance experience. In the first test, subjects were required to make speeded sorting judgements on changes in intonation and loudness. Reaction times were measured for blocks of phrase pairs relative to three conditions: a) change on one dimension only (Control); b) changes on both dimensions simultaneously (Redundant); and c) change on the irrelevant dimension only (Orthogonal). The second test used a restricted classification task to assess whether

3 subjects grouped phrases according to the more discriminable dimension, or if both dimensions were equallysalient. The third test explored whether subjective distances between notes varying in pitch and loudness better matched the sum of the individual distances or the Euclidean distance. Analyses of the data showed that subjects' reaction times increased for the orthogonal condition, indicating reduced performance, but showed no facilitation in the redundant condition. 2) Subjects showed pitch and loudness to be equally salient in the restricted classification task, and 3) Subjects ratings of perceptual distances between notes varying in intonation and loudness better conformed to a Euclidean metric rather than a city-block measure. Cognitive interference was evident in the speeded sorting task and in the perceptual scaling of stimulus distances. Since the dimensions remained equally salient in the restricted classification task, and since subjects showed no facilitation in the first test, the dimensions of pitch and loudness did not all match characteristics of Garner interference.

4 3"7<r M8U Ao.V7SV COGNITIVE INTERFERENCE IN THE PERCEPTION OF PITCH AND LOUDNESS IN A FIVE-NOTE MUSICAL PATTERN DISSERTATION Presented to the Graduate Council of the University of North Texas in Partial Fulfillment of the Requirements For the Degree of DOCTOR OF PHILOSOPHY By Gary Thomas Cattley, B.A., M.M. Denton, Texas May 1999

5 TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES Page V vii CHAPTER ONE 1 Background of the Study 2 Rationale 4 Purpose of the Study 16 Statement of Purpose 16 Research Questions 16 Definitions 17 CHAPTER TWO 19 Cognitive Interference 19 Intonation 29 Timbre 31 CHAPTER THREE 34 General 34 Subjects 35 Testing 36 Test 1: Speeded Sorting of Stimuli 37 Test 2: Restricted Classification of Stimuli..41 Test 3: Similarity Scaling of Stimuli Pitch and Loudness 46 CHAPTER FOUR 52 Test 1: Speeded Sorting 52 Test 2: Restricted Classification 58 Test 3: Similarity Scaling 63 CHAPTER FIVE 70 Background 70 The Study 70 Findings 72 Research Questions #1 and #2 72 Question #

6 Question #4 76 Research Question #5 81 Conclusion 82 Discussion 84 Suggestions for Further Research 86 APPENDIX A Forms 87 Approval for Testing of Human Subjects 88 Cover Letter - Subject Testing 89 APPENDIX B Test Instructions and Screen Prompts 90 Test 1 On-screen Instructions 91 Test 2 On-screen Instructions 92 Test 3 On-Screen Instructions 93 Screen Prompts 94 Test 1 94 Test 2 95 Test 3 96 APPENDIX C Raw Data 97 Test 1 98 Test Test 2, retest Test 2, retest Test Test 3, retest 133 REFERENCES 146

7 LIST OF TABLES Page Table 1 Comparison of Test Halves, Test 1 52 Table 2 Main Effects and Interactions, Condition and Attention 54 Table 3 Main Effects and Interactions, Condition and Attention 55 Table 4 Predicted Means of Attention Subgroups 56 Table 5 Predicted Means of Condition Subgroups for Pitch 57 Table 6 Predicted Means of Condition Subgroups for Loudness 58 Table 7 Comparison of Test Halves, Test 2 58 Table 8 Frequency and Percent of Responses for Pitch and Loudness 60 Table 9 Combined Frequency and Percent of Responses 60 Table 10 Frequency and Percent of Responses for Pitch and Loudness (Re-test, equally-discriminable stimuli) 62 Table 11 Frequency and Percent of Responses for Pitch and Loudness (Re-test, stimuli at one-step difference.) 62 Table 12 Frequency and Percent of Responses for Pitch and Loudness (Retest, stimuli at two-step difference.) 63 Table 13 Pearson Product-moment Correlation, Rated vs. Actual 64 v

8 Table 14 Analysis of Variance, Subjective and Geometric Distance 65 Table 16 Pearson Product-moment Correlation, Rated vs. Actual, Test 3 Retest 67 Table 17 Analysis of Variance, Subjective and Geometric Distance, Test 3 Retest 67 Table 18 Test 3 Retest, Means and Sets of Means 68

9 LIST OF FIGURES Page Figure 1. Relationship of Points for Control, Orthogonal, and Redundant Conditions 9 Figure 2. Geometric Representation of Stimulus Relationship in a Restricted Classification Task 10 Figure 3. Geometric Model for City-Block and Euclidean Distance 11 Figure 4. Relationship of Intonational and Dynamic Change 39 Figure 5. The Five-Note Musical Test Pattern 39 Figure 6. Pitch and Loudness Values for Test 2 42 Figure 7. Pitch and Loudness Values for Retest 1, Test 2 45 Figure 8. Pitch and Loudness Values for Retest 2, Test 2 45 Figure 9. Values for Retest 3, Test 2 46 Figure 10. Geometric Representation of Stimulus Relationship, Test 3 47 Figure 11. Array of Pitch and Loudness Values, Test 3 49 Figure 12. Euclidean Distances of Pitch and Loudness used in Additional Testing, Test 3 50 Figure 13. Pitch and Loudness Values, Test 2 59 Figure 14. Actual and Subjective Distances for 4-step city-block stimulus distances 79

10 Figure 15. Actual and Subjective Distances for 6-step city-block stimulus distances 79 Figure 16. Actual and Subjective Distances for 5-step city-block stimulus distances 80 VI11

11 CHAPTER ONE BACKGROUND OF THE STUDY, RATIONALE, PURPOSE OF THE STUDY, DEFINITIONS Human perception has long been the subject of systematic investigation. Within the realm of human perception lies the subject of hearing science. Hearing science explores physiological, psychological, and pathological aspects of hearing. Psychoacoustics is a subset of the area of hearing science, and is concerned with "psychological response to acoustical stimulation" (Durrant and Lovrinic, 1995, p. 256). The hearing mechanisms of humans has been given thorough treatment through systematic investigation, yet many aspects of how humans perceive sound remain unclear. Musical perception is one such area (Rasch and Plomp, 1982, p. 21; Risset, 1978, pp ). Research in musical perception seeks to describe human response to musical stimuli. The responses to such stimuli can be quite different than those obtained by stimuli presented with no musical context (Risset, 1978). The present study explored whether such a

12 difference exists in the cognitive processing of musical pitch and loudness. Background of the Study The first large work in the area of psychoacoustics was Helmholtz's 1885 work, On the Sensations of Tone as a Physiological Basis for the Theory of Music (Helmholtz, Trans. Ellis, 1954). It is interesting to note the extended title, often excluded in bibliographic references, as it shows an early, albeit not the earliest, interest in integrating hearing theory and music theory. As noted by Scharf and Buus (1986), Helmholtz's work along with Fechner's 1860 work Elements of Psychophysics served as an impetus and basis for the large body of research in hearing and psychoacoustics that followed. Seashore's Psychology of Music served to further integrate hearing theory with an understanding of musical behavior (Seashore, 1938). The physical limits of musical acoustics must lie within limits discovered by psychoacoustic research (viz., the stimuli acted upon generally lies well with in detectable limits of human hearing). Risset (1978, p. 522) noted that psychoacoustics defines the boundaries of music

13 perception but does not serve to determine what happens within those boundaries. Terhardt (1978), as well as others, pointed out that musical tones carry precise information, which rely on the performance of the auditory system and thus are not specified solely by the not the physical parameters of the sound. Furthermore, music perception requires one to consider responses beyond those obtained by exposure to isolated tones in non-musical context. Hence, training, construction, and cultural learning affect the perception of music (Rissett, pp. 522,528,535). Considering these views, one may reasonably expect responses to musical stimuli to be different from responses to other presentations of auditory stimuli. The discrimination of the frequency of sound is of central importance to perceiving music. Pitch, the psychological correlate of sound frequency, provides essential information about environmental surroundings. As a component of speech, pitch inflection conveys both broad and subtle meanings. The perception and discrimination of frequency is also essential to the musical behavior of humans. On a macro level, the structure of Western music is based to the largest extent upon the organization of

14 pitches in time. On a micro level, especially to those performing music, the ability to make extremely fine distinctions in pitch is essential to proper intonation in musical performance. Rationale Pitch perception is closely correlated to the physical frequency of a sound; however, pitch perception is also affected by attributes of sound other than frequency. These factors may be inherent in the sound (e.g., whether a sound is pure or complex, duration, intensity, envelope shape) or external to the sound (e.g., the presence of other sounds, previous exposure to a tone close in frequency to a current tone). Studies that have investigated these interactions have been useful in describing and supporting models of pitch perception. Many of these studies explore effects that would in all likelihood be of little or no consequence to musical pitch, either because the pitch differences would be too slight or the duration of stimuli too short (e.g. Hartmann, 1978; Rossing and Houtsma, 1987). However, there

15 are instances where musical pitch might be influenced byone or more of these attributes. One such attribute, which has been shown to be of importance to pitch as well as to other aspects of sound, is loudness (Stevens, 1935; Fletcher, 1934). However, studies that have explored the effect of various attributes on musical pitch often disagree as to whether loudness significantly effects pitch perception. Swaffield (1974) reported that fine tuning adjustments around 440hz are "highly dependent" on intensity, as well as other factors. Leonard (1967) also found intensity to affect pitch discrimination. Tekman found that pitch and loudness exhibited a trading relationship, in that manipulations could substitute for each other in a discrimination task (Tekman, 1992, pp ). On the other hand, Haack (1975) reported results "not congruent with the conclusions of earlier studies that indicated that frequency sensitivity varies with sensation level" (Haack, p.70). Differences in conclusions as to the effect of loudness on pitch judgements may well be due to presentation of stimuli. For example, Haack utilized tone pairs near 500hz; Swaffield presented stimuli in a melodic

16 context. It is also possible that differences in conclusions among pitch-loudness studies may also arise from the manner in which the stimuli are processed cognitively by the perceiver. In reviewing the body of literature concerning interaction of pitch and loudness, much is known about how the auditory system responds to pitch and loudness (for example, see Shower and Biddulph (1931); Verschuure and van Meeteren (1975); Stevens; Fletcher;) but far less in known about how the listener processes pitch and loudness information. The pitch and intensity of sound are readily measured and quantified acoustically, but it may not be proper to assume that these parameters always exist as two distinct dimensions in human cognitive processing. Garner (1970) has pointed out that the nature of the stimulus is too often ignored in studies intended to draw conclusions as to how information is processed. He wrote "One of the most important variables in determining how the organism will process information is the nature of the input itself, the very thing that has received so little attention in our research." (Garner, 1970, p.351).

17 He stressed that asking when dimensions of a stimulus should be considered a single entity or separate components is an important question, the answer to which must not be assumed a priori. In consideration of this, Garner concluded that the nature of the stimulus is properly an experimental question. Thus, it is important to explore the nature of stimuli used in research concerning musical behavior. To be more specific, it is important to ask whether the physical attributes of pitch and intensity are to be considered unidimensional or multidimensional to the perceiver. Answers to this question may be relevant to studies investigating musical behavior, both in consideration of stimulus presentation in studies and in understanding tendencies of musicians. The question of stimulus dimensionality, however, has rarely been raised with regard to effects on loudness and intonation in music. According to Garner (1974), dimensions of a stimulus may be either integral or separable in nature. A stimulus with separable dimensions (e.g. brightness and size) requires little effort to evaluate the component dimensions. In this case, the stimulus structure may be directly perceived. A stimulus having integral dimensions

18 8 (e.g. color and saturation) requires considerable effort in order to attend to its overall structure. In the case of the integral dimensions, correlated changes along both dimensions will enhance perception of the stimulus. In addition, changes on an irrelevant dimension will cause interference with the dimension being attended to. This increased effort in attending to a single dimension among integral dimensions has been termed "Garner interference" (Pomerantz, 1983, pp. 1-30). Whereas others investigated the nature of cognitive processing of multidimensional stimuli (e.g. Attneave, 1950; Lockhead, 1966; Handel and Imai, 1972), Garner formalized a set of converging operations which, when used together, distinguish integral stimuli from separable stimuli. These converging operations are comprised primarily of three separate tasks: Speeded Sorting of Stimuli, Free or Restricted Classification of Stimuli, and Similarity Scaling of Stimuli (Garner, 1974). A brief description of each operation follows.

19 The Speeded Sorting task measures reaction times of subjects sorting sets of stimuli according to each component dimension. There are three ways in which the stimuli may be structured: 1) the stimuli may change only on a single dimension (For example, B vs. D and A vs. C) o «D o Dimension Y A C Dimension X Figure 1. Relationship of Points for Control, Orthogonal, and Redundant Conditions. 2) the stimuli may change on a single dimension while the other (orthogonal) dimension is varied (For example, A vs. B and C vs. D while attending to Dimension X). 3) the stimuli may have a correlated change on the other dimension. (For example, A vs. D and C vs. B while attending to Dimension X.) Garner found that increased

20 10 reaction time in task 2 and decreased reaction time in test 3 imply integrality of the dimensions in the stimuli. The Restricted Classification task requires subjects to perform an unspeeded sorting task on stimuli presented in the relationships shown in Figure 2. If for each dimension subjects group stimuli according to the dimension having the larger change in value (A with B and E with F) o C o F X X Figure 2. Geometric Representation of Stimulus Relationship in a Restricted Classification Task. the dimensions are considered to be integral. If subjects show dimensional preference irregardless of the dimension that varies to a larger extent, the dimensions are both salient and the stimuli are considered to be separable. One pair of stimuli have the same value on one dimension yet are very different on the other (A and B in Figure 2). The other pair differs slightly but perceptually on both

21 11 dimensions (B and C). According to Garner, stimuli having integral dimensions will be grouped by overall similarity (B grouped with C) and stimuli with separable dimensions will be grouped according to shared dimensional value. Similarity scaling of stimuli asks subjects to judge the magnitude of difference between pairs of stimuli Dimension Y Dimension X Figure 3. Geometric Model for City-Block and Euclidean Distance. arranged as in Figure 3. The stimuli vary along two component dimensions. The important issue in this test concerns which metric best matches the data obtained for magnitude ratings between stimulus points B and C. If the geometric distance best matches the sum of the stimulus distances {dx+dy), as is called the "city-block distance", then the dimensions are considered as separable: each component dimension is assessed and perceived separately.

22 12 If subjects give stimulus ratings that better conform to the Euclidean distance (-Jdxxdy ) then the stimuli are considered as integral, since the dimensions appear to be perceived in a more unitary fashion. Garner's research into the integrality and separability of stimuli used visual stimuli; however, other studies have provided evidence for generalizing the same concept across modalities. Most relevant to the present study are those studies using auditory stimuli. Wood (1973) was perhaps the first to investigate the integrality of auditory dimensions. He utilized a speeded sorting task to measure reaction time and evoked potential of subjects' responses to a syllable varying in pitch and loudness. The stimuli was randomly varied in control and orthogonal presentations, in a manner similar to Garner's speeded sorting task (cited earlier). Results of Wood's study showed interference in the cognitive processing of the stimulus in the orthogonal condition, which supported the integrality of pitch and loudness. Grau and Kemler-Nelson (1988) conducted Garner's set of converging operations on auditory stimuli varying in pitch and loudness. Their stated purpose was to

23 13 investigate "the reality and generality of dimensional integrality" (Grau and Kemler-Nelson, p. 347). Undergraduate students, with no reported musical background, participated as subjects. Grau and Kemler- Nelson concluded that pitch and loudness were psychologically valid dimensions of the stimuli employed in their study, and were processed holistically rather than by their component dimensions. Thus, pitch and loudness were shown as integral in nature. Furthermore, although Grau and Kemler-Nelson did not employ musical stimuli (for example, the pitch values did not correspond to any precise musical pitch) the tones used were in a musically significant range (approximately C 5 ) and varied the musical equivalent of approximately 37 cents (566Hz to 554 Hz). Melara and Marks (1990a) explored the cognitive interaction of timbre and pitch, and timbre and loudness. In their study, stimuli varying on these dimensions showed "substantial Garner interference" (Melara and Marks, 1990a, p. 169). Although the study did not employ musical tones, loudness varied by + /-6db and pitch varied between 920Hz and 900Hz (approximately a 38-cent change). As with the range of stimuli used by Grau and Kemler-Nelson, these variations are large enough to be significant in musical performance.

24 14 In proposing a model of dimensional interaction, Melara and Marks (1990b) included as a sub-test the speeded sorting of stimuli varying in pitch and loudness. Garner interference was observed among the two dimensions (Melara and Marks, 1990b). Again, the change in loudness and pitch would certainly be significant musically ( + /_10 db, and approximately 94 cents). In summary, psychological studies have employed stimuli having two component dimensions (e.g., brightness and size, brightness and saturation, loudness and pitch) in order to evaluate the separability or integrality of the components in cognitive processing. Initially, studies employed visual stimuli; later studies explored auditory stimuli. In studies that used auditory stimuli, the dimensions of pitch and loudness were shown to be integral in nature. A change in one dimension can interfere with judgements concerning the other dimension. Since musical perception requires analysis and construction, it may be possible that Garner interference observed in these studies also has a significant effect on pitch judgements in musical settings.

25 15 Rissett (1978) stated that the effect of loudness on pitch is of little significance in musical settings. One would have to agree with this statement in a global sense. Indeed, it has been observed that if this were not the case, it would be an impossible task for an orchestra to play in tune unless every instrument was of uniform intensity (Durrant and Lovrinic, 1995, p. 279). However, intensity, due to interference in cognitive processing, may play an important role by affecting the ability to make fine judgements in intonation. Based on the body of literature concerning pitch perception of isolated tones, pitch perception of musical tones, and the effects of loudness on the pitch of both isolated and musical tones, additional research would contribute to the body of knowledge concerning the perception of musical pitch. Specifically, it is proposed that since the dimensions of pitch and loudness have been shown to be integral in nature (see Wood; Grau and Kemler-Nelson; Melara and Marks (1990a, 1990b), musical situations involving sudden changes in loudness should also exhibit the same cognitive interference. An exploration of interaction among intonation and dynamic level within a melodic pattern may

26 16 contribute towards a better understanding of musical pitch perception. Purpose of the Study Statement of Purpose The purpose of this study is to explore whether the dimensions of intonation and loudness, when varied within a five-note musical pattern, are integral or separable in nature, and whether there is evidence of cognitive (Garner) interference among the dimensions. In addition, this study will compare any characteristic effects to findings of existing studies using isolated tones and cross-modal stimuli. Research Questions 1) Do musicians exhibit signs of performance interference, in a speeded sorting task that requires discrimination among five-note musical phrases possessing a final note changing in either intonation or loudness, in the face of a change on an irrelevant dimension? 2) Do musicians exhibit signs of performance facilitation in a speeded sorting task requiring evaluation of five-note musical phrases possessing a final note that changes simultaneously in intonation and loudness?

27 17 3) Do musicians prefer to sort sets of five-note musical phrases, wherein the discriminability of the intonation and loudness of the last note is varied, according to preference for the more discriminable dimension, or are both dimensions equally salient? 4) Do musicians, in scaling the differences among five-note phrases having a final note that changes along the dimensions of pitch and loudness, produce a pattern of responses consistent with a "city-block" or a Euclidean metric? 5) Are findings of procedures employed to answer questions 1 through 4 supportive either of integrality or separability of the dimensions of musical intonation and loudness? Definitions Pitch, as used within the present study, refers to the psychological correlate of frequency, and not to any categorical levels of musical pitch. Musical Pitch, as used within the present study, refers specifically to the categorical levels of musical pitch classes.

28 18 Non-Musical Pitch refers to tones presented without musical any context or setting, such as isolated test tones or tones presented without regard to musically relevant intervals. Musical Context refers to tones arranged in musically traditional frequency ratios and rhythmic sequences. The term, as applied in the current study, does not refer to a broad sense of musical context, which include features such as multi-timbral sounds, broad musical form, or performance situations. Decibel Level (db), when referring to stimuli specified in the present study, refers to sound level SPL at the standard reference of 20uPa (2 x 10" 5 N/m 2 ).

29 CHAPTER TWO REVIEW OF THE RELATED LITERATURE Central to the topic of this paper are those studies and methods which sought to provide an understanding of the cognitive processing of pitch and loudness rather than the physiological responses to those stimuli. Studies concerning interference effects among dimensions of a stimulus form the immediate background for the present study. Therefore, a summary of the procedures described by Garner for the investigation of integrality and separability of stimulus dimensions, as detailed in Chapter 1, will be followed by reviews of studies that used some or all of the procedures outlined by Garner. Investigations that explored intonation in musical phrases and studies that concerned the choice of timbre in intonation matching investigations are also discussed. Cognitive Interference Garner suggested that the nature of multidimensional stimuli could be better understood by the convergence of

30 20 results gleaned through varied approaches rather than through a single test for a single concept (Garner, 1974, p. 188). The various methodologies used in exploring cognitive interference in multidimensional stimuli are discussed in the context of the various studies reviewed in this chapter. According to Garner, for dimensions to be considered integral, they must meet certain criteria in test results. First, the stimulus must show interference in classification performance in the face of changes in an irrelevant dimension. The same stimulus will also exhibit facilitation of performance when presented redundantly in a classification task. Secondly, free and restricted classifications should show dimensional preference for each dimension when discriminability is increased. Thirdly, perceptual similarity judgements should best fit the Euclidean distance between stimulus points that change on both dimensions, rather than the sum of the changes of both dimensions. Grau and Kemler-Nelson (1988) conducted a study that explored Garner interference between the auditory dimensions of pitch and loudness. Their study is the most detailed investigation of those exploring the integrality-

31 21 separability issue concerning pitch and loudness, and as such employed a set of converging operations as outlined by Garner, as well as certain psychophysical scaling tests which reinforced their selection of stimuli for the tests. The first test employed speeded sorting of stimuli. In preparation, they conducted a test for psychophysical scaling of pitch and loudness. In the areas of variation across subjects (n=10) and tendency, the preparatory test produced results in agreement with others such as Stevens (1935). All effects were reported to be "quite small". There was little reported variation across subjects, and an increase in intensity brought about a slight decrease in perceived pitch while an increase in pitch brought about a very slight increase in loudness. Subsequently, the first test was carried out on twelve subjects by means of a computer program that presented stimulus pairs in blocks, and recorded reaction times by means of keyboard response as well as error rate. The redundant variation facilitated subjects' performance while the orthogonal variation hindered performance; an ANOVA showed a significant (p <.001) difference between these two groups. Although the only summary of data was provided were tables showing mean reaction time and mean percent errors, the researchers

32 22 reported that there was no significant difference in performance across the dimensions of pitch and loudness. They reported findings supportive of the integrality of pitch and loudness. The second test conducted by Grau and Kemler-Nelson (1988) was designed to assess whether subjects classified tones varying in pitch and loudness according to a shared dimensional value or according to their overall similarity. Eight subjects participated in an untimed restrictedclassification experiment. The task was to select which of three stimuli in each test item did not belong in the group. In each presentation two of the three stimuli shared the same value on one dimension and differed to a large degree on the other. The third tone differed slightly but perceptually on both dimensions. As with the first test, complete summary tables were not provided for the data obtained; however, reported results of the test showed that classifications across the two dimensions (triads of tones that shared the same loudness and triads of tones that shared the same pitch) were not statistically different (ANOVA, p >.05). Subsequently, the researchers pooled the data from both conditions. The researchers reported that subjects classified the stimuli on the basis

33 23 of overall similarity rather than by shared dimensional value (ANOVA, p <.01) and that very few anomalous classifications were made. The findings were consistent with results expected for integral dimensions. The next test given by Grau and Kemler-Nelson (1988) was designed to determine which geometric model best-fit subjects' ratings of similarity among tones varying in pitch and loudness. As described in Chapter 1, the geometric distance between stimuli, based on similarity (or dissimilarity) ratings may best fit the Euclidean distance between the points. Or, the distance between the points may best fit the simple addition of the component ratings (called a city-block metric, since the distance between the points is similar to adding the distances to travel to the opposite diagonal corner of a city block). Ten subjects rated dissimilarity of pairs of tones, which varied in pitch and loudness according to equal steps. As with the other experiments, complete summary data was not provided for the reader. The researchers show graphical evidence that the dissimilarity ratings better fit the Euclidean metric than the city block metric. The value for the Euclidean distance between the diagonal points on the center square was 2.83, and the city-block distance was 4.

34 24 The obtained value for the perceptual distance between the diagonal points, which was described as being an estimate, was 1.6. While this value is lower than optimal, it is even farther in distance from the value expected for the city-block metric. In addition, is in line with results obtained in other studies (e.g. Handel and Imai, 1972). Although taken somewhat cautiously, the results of the third test also supported integrality of the dimensions of pitch and loudness (Grau and Kemler-Nelson, 1988, pp ). In consideration of the intended convergent nature of the tests, there is strong evidence for the integrality of the stimulus dimensions of pitch and loudness. Studies by other researchers corroborate the findings of Grau and Kemler-Nelson (1988). Melara and Marks (1990a) investigated interaction among the auditory dimensions of timbre and pitch, and timbre and loudness. They concluded that when classifying timbre-loudness and timbre-pitch pairs, subjects suffered Garner interference, whereby subjects were unable to attend selectively to a single dimension. Thus, Melara and Marks (1990a, pp. 169,174) concluded that the dimensional attributes were processed jointly by the perceptual system.

35 25 While the data reported in this study supported their conclusions, the study utilized only one aspect of the set of converging operations proposed by Garner. Since the set of converging operations is designed to accommodate for other possible contingencies that may also explain the pattern noted in subjects' reaction times, one must view the results of the study as supportive of, but not conclusive of, integrality among the dimensions tested. Use of the entire set of converging operations would provide stronger support for any conclusions drawn in an investigation concerning Garner interference. The present study will use each of the three tests posed by Garner. In another paper, Melara and Marks (1990b) explored the interaction of pitch-loudness, pitch-timbre, and loudness-timbre. The proposed model of dimensional interaction outlined in this study is different than the view posed by Garner and Grau Kemler-Nelson (1988). Melara and Marks suggested that interference among dimensions does not imply an inability to extract dimensional information from the stimulus. Rather, they stated that interaction among two dimensions can occur only if the dimensions are initially represented as distinct perceptual dimensions: if dimensions were not perceived as such, they could not

36 26 interact. They further reasoned that interference among dimensions in perceptual processing is influenced bycontext. The details of the model presented by Melara and Marks, while different from the model assumed by Grau and Kemler-Nelson, does not alter the underlying premise of interaction among the auditory dimensions of pitch and loudness. Their study was in support of other studies showing interaction among the auditory dimensions. Wood (1973) investigated the perceptual processes occurring between a stimulus signal and the identification of phonemes in the signal. As a part of his study, Wood varied a syllable (bae) in frequency from 104 to 140 Hz and with an attenuation difference of 20 db. It was not possible to determine the actual SPL of the stimulus, as the study reported levels only in terms of electronic signal level. Results showed a statically significant level of interference (ANOVA, 0 <.001) between the control and orthogonal conditions. No significant difference was reported between the two dimensions. Thus, irrelevant variations on each unattended-to dimension produced substantial interference. This result provides supportive evidence for the integrality of the dimensions of pitch and loudness.

37 27 Fiske (1984) used reaction time as a measure in determining if tonal and rhythmic information in melodic phrases involved sequential or simultaneous processing. Using a single ten-note diatonic melody, Fiske independently altered a single note rhythmically and intervalically and recorded reaction times of two groups of subjects. One group was given advance information as to the nature of the altered note; the other group had no such advance notice. Fiske reasoned that if discrimination of the tonal or rhythmic differences is processed serially, then the group with advance warning should be able to detect a discrepancy more quickly than the group given no cue as to what to attend to. However, if the reaction times were not statistically different among the groups, then the subjects' search for discrepancies must occur simultaneously and thus serve to indicate parallel processing of the stimuli. Fiske concluded that subjects attended to tonal and rhythmic elements simultaneously, and that subjects responded to discrepancies as soon as the discrepancy occurred rather than waiting until the melody ended. Even though Fiske (1984) concluded that simultaneous processing of tonal and rhythmic elements took place, his

38 28 conclusion seems to be based on a rather high (though unspecified) significance level. Although raw data was not provided in his report, a t-test performed by this writer on the summary data showed statistical significance at the.05 level. This would have supported a serial model of processing for tonal and rhythmic elements rather than the proposed parallel-processing. Fiske's study did not directly investigate Garner interference. However, a model of serial processing of musical stimuli could be supported by a slightly different interpretation of the data in this study. Such an interpretation would lend support to the expectation of Garner interference among musical stimuli. In addition, is not surprising that subjects made decisions regarding melodic and rhythmic discrepancies immediately upon their occurrence; the elapsed time in waiting until the end of a ten-note melody would reasonably exceed the time required for a musician to make such a decision. In the present study, only the last note of the melody will be manipulated, thereby providing a consistent reference point from which to measure reaction time. Furthermore, if Garner interference is exhibited between pitch and loudness of a musical note, it would, according to studies cited elsewhere in this paper, be evident even

39 29 when the discrepancies occur at a point that is predictable to the subjects. Since the present study concerns the processing of musical pitch and loudness in context, it is relevant to review studies that explore musical intonation. Studies addressing musical intonation may be conveniently sorted into to main areas: 1) Studies that address intonation performance and 2) studies that investigate perception of intonation. Relevant to the present study are those studies which explore perception of intonation rather than intonation performance. Most directly relevant to the present study, in terms of presentation of stimulus, are those studies that employ solo phrases rather than interval or ensemble presentations of stimuli in assessing intonation preferences. Intonation Researchers have investigated the effects of manipulating various aspects of notes in musical intervals and phrases in order to assess preferences in intonation. A study by Terhardt and Zick (1975) used six musical sounds presented with stretched, contracted, and normal tunings. Fifty subjects (25 musicians and 25 non-musicians) evaluated the tunings in paired comparisons. Their

40 30 findings indicated that no optimal tuning was applicable to every musical setting. Rather, ideal intonation required flexibility and adaptability to the musical situation at hand. Ward and Martin (1961) adjusted the tunings of the third, sixth, and seventh degrees of musical scales in order to find which mis-tunings were most discriminable. They found flatness in tunings to be more discriminable than sharpness, although it is somewhat difficult to lend strong consideration to this finding: In one test, subjects had difficulty in matching a third scale to either of two preceding scales. In another sub-test, the researchers observed large differences among individual subjects, and opted to include only the results of two contrasting subjects. Rasch (1985) studied the perceived quality of intonation of melodic fragments, whereby he de-tuned twopart musical phrases both harmonically and melodically. While the present study uses an unaccompanied phrase rather than two part phrases, Rasch's emphasis on musical context serves well as a model, and is supportive of the use of musical phrases in intonation studies.

41 31 Timbre It also should be noted that Rasch (1985) used complex tones in playing the phrases in his study. The use of complex waveforms has been shown to be an important consideration in intonation studies. Swaffield (1974) tested twenty-five undergraduate music students in their ability to match intonation of a tone to the tonic of a melodic fragment (scale degrees four to seven). The melodic fragment was varied in timbre, intensity, and note duration. Swaffield reported that context had significant influence upon subjects' fine-tuning of a matching tone. This study relates to the present study in two ways. First, loudness was shown to play a role in pitch perception. It was not possible to know whether Garner interference contributed to observed effects of loudness on pitch. However, one must consider it as a possible explanation for the results. Secondly, timbre was reported to have a slight effect on intonation. It should be noted that the values given in the data summaries show a greater effect of timbre than was reported in the conclusions. The implications of the latter on the present study are to give

42 32 consideration to the choice of timbre in the presentation of musical stimuli. Greer (1970) also emphasized the importance of the timbre used for the stimulus. Greer investigated the effects of timbre on intonation in musical performance. While his study emphasized the performance aspect rather than the perceptual aspect of intonation, it holds relevance to the present study in that it addressed "external intonation" whereby subjects are required to match an external reference (as is the case when performing with an other musician). Among other findings, Greer found that instrumentalists had far more difficulty in matching intonation for a melody played with an oscillator than played by either a piano, an organ, or the subjects' own instruments. Others expressed similar opinions: Lundin (1953) stated that subjects were less able to recognize correct pitch with a tuning fork than with a piano. Riker (1946) found that subjects more accurately named a pitch of piano tones then oscillator-generated pitches. Ely (1992) investigated the effect of timbre on intonation perception of 27 undergraduate and graduate woodwind majors. Timbre, which consisted of live

43 33 performance recordings of various woodwind instruments, was found to affect subjects' ability to detect accurately deviations in intonation in both performance and assessment of performance. The studies investigating the effect of timbre on intonational accuracy suggest that timbre can play a role in accuracy of intonation judgements. Although there is general agreement against the use of sine tones for pitch matching, no clear indication is given to support the use of any specific timbre. Considered together, findings of the various studies suggest that certain complex timbres may enhance the ability to make intonation judgements. For the present study, the timbre for the listening tasks was chosen both in consideration of the above studies and somewhat arbitrarily based on the performance experience of the investigator. The tone used in the study was comprised of a fundamental sinusoidal frequency plus the first two harmonic sinusoids, in strengths of 0.4 and 0.2 relative to the fundamental.

44 CHAPTER THREE METHODS AND PROCEDURES General All tests in this study were administered on-line bymeans of a computer program specifically designed for each test. On-line instructions were provided to subjects prior the start of each test. Subjects were given opportunity to ask for clarification of procedural instructions after reading the instructions but before beginning the actual testing. Data obtained from subjects' responses was stored to a file on the computer for subsequent data analysis. All stimuli were generated using the Csound software sound synthesis language (Vercoe, 1997). Each note had a sound envelope comprised of a rise and decay time of 100ms, and a harmonic structure comprised of the fundamental frequency and the first two partials at relative strengths of 0.4 and 0.2, respectively. The stimuli were produced by means of a Yamaha OPL3-SA3 digital audio system, using a sample rate of 22,050 samples per second. A Mackie 1202 audio mixer, set to flat frequency response, was used to 34

45 35 amplify the audio output from the computer. Stimuli were administered monaurally to subjects through supraaural Sony MDR-V6 earphones. Earphone sound level intensity was calibrated prior to each administration of the test by means of an ADC Model SLM-2 sound level meter coupled to the earphones through a 6cm 3 dummy cavity. The use of such a cavity approximated a typical human ear canal volume, and is routinely used in calibrating the sound pressure level of supraaural earphones (Durrant and Lovrinic, 1995, p. 54). The dummy cavity was also used to verify that the headphones produced a uniform sound intensity level throughout the frequency range of the stimuli. Subjects All subjects participating in the study were musicians who either a) were currently working as professional musicians in classical ensembles (viz., orchestra, wind ensemble, brass and woodwind quintet) or b) have had significant performance experience (viz., regular rehearsal and performance with an established wind ensemble or orchestra, or solo performance experience on an orchestral instrument. Subject participation was voluntary, and inclusion in the study was based solely on invitation by the researcher. Twenty-five subjects participated in test

46 36 1, thirty-one subjects participated in test 2, and sixteen subjects participated in test 3. Each subject was tested individually, and completed the three procedures in a single session. Test length for subjects completing all three procedures was typically thirty-five minutes. Due to personal time constraints of certain subjects, not all subjects participated in all tests; however, all sixteen subjects taking test three first took tests 1 and 2, and all twenty five subjects participating in test 1 also took test 2. Testing Testing was completed using stimulus values obtained in prior research by others, with special consideration to those values used by Grau and Nelson (1974). This was done for two reasons. First, the use of stimuli similar in range to other studies allowed direct comparisons to other findings. Secondly, these stimuli served well as a point of departure in the present study. Subsequent analysis of the data for tests two (Restricted Classification of Stimuli) and three (Similarity Scaling of Stimuli) indicated that additional testing with new stimulus values was necessary. A description of each of the three tests is given below. The initial round of testing is presented

47 37 first within each of the following test descriptions. Reasons and procedures for the additional testing in tests two and three are provided at the end of each section. A more detailed discussion of the additional testing is provided in Chapter Four. Test 1; Speeded Sorting of Stimuli Twenty-five subjects participated in the first test. In this test, subjects sorted groups of phrases on the basis of perceived change in loudness and perceived change in intonation. Each group contained pairs of phrases whose last note either varied along the attended-to dimension, varied on both dimensions, or stayed the same on the attended-to dimension but changed on the irrelevant dimension. The goal was to assess the ease in which subjects could selectively attend to either intonation or loudness in the face of a change in the unattended-to dimension, and to assess whether subjects' performance improved for correspondent changes on both dimensions. According to Garner (1974), if the dimensions of the stimuli are integral, the reaction times of subjects will increase when the unattended-to dimension varies and decrease for stimuli that vary on both dimensions. Thus,

48 38 if intonation and dynamic change are integral dimensions of a musical sound, subjects' reaction times will be greater when attending to pitch during a change in dynamics and when attending to dynamic change on an out-of-tune note. Conversely, subjects' reaction times should decrease when intonation and dynamic level change simultaneously. Reaction times of subjects selectively attending to either loudness or pitch were recorded for the Control conditions, Orthogonal conditions, and Redundant conditions of the stimulus pairs. Referring to Figure 4, the Control conditions for pitch were A vs. B and C vs. D, and for loudness A vs. C and B vs. D. The Orthogonal conditions were A vs. C and B vs. D for pitch, and A vs. B and C vs. D for loudness. Lastly, the Redundant conditions for pitch were A vs. D and C vs. B, while A vs. D and C vs. B comprised the Redundant conditions for loudness. As described above, there were six possible pairings of stimuli. Thus, there were 12 stimulus pairs used in the test. These twelve pairs of stimuli were used for attending to the dimension of pitch and for the dimension of loudness, creating 24 data points. The 24 data points were repeated in order to assess reliability by split-

49 39 halves comparison, resulting in 48 recorded data records for each subject. In Tune Out of Tune Dynamic Change A B No Dynamic Change C D Figure 4. Relationship of Intonational and Dynamic Change. The test stimuli consisted of a single unaccompanied musical phrase, the last note of which varied on the two test dimensions. The phrase used in this test is shown in Figure 5. The variation in pitch of the last note was be 37 cents, which was the equivalent in musical pitch deviation to the change in pitch used by Grau and Kemler- Nelson (1988). The dynamic level of the last note varied 120 Figure 5. The Five-Note Musical Test Pattern.

50 40 by 6db, which also corresponds to the loudness change used in the same study. Subjects began the test by reading on-line test instructions. Instructions described what to listen for in the test stimuli, familiarized the subject with the screen prompts and keyboard responses, and emphasized the importance of providing a rapid yet accurate response. The on-line instructions and test screen prompts are given in Appendix B. After the test instructions were given and before beginning the test, subjects had opportunity to ask for procedural clarification. The subject then pressed the spacebar to begin. Elements within the pairs of musical phrases were played concurrently, and subjects responded by pressing the "A" or "L" keys for a response corresponding to "change" or "no change". A prompt in the on-screen panel indicated which dimension (pitch or loudness) was to be evaluated. The subjects response of "A" or "L" invoked a delay of 500ms and the subsequent presentation of the next pair of phrases. An on-screen prompt encouraged the subject to respond more quickly if the response time was inordinately long. Phrase pairs were presented twice, in blocks of 18; each block had the subject focus solely on one dimension. The first six stimulus pairs in each block

51 41 were not recorded, in order to allow subjects time to settle in to the task after a change in attention as directed by the on-screen prompt. Thus, each subject responded to a total of 72 presentations, 48 of which were retained for statistical analysis (exclusive of outlying responses, mentioned earlier). Reaction times, measured from the onset of the initial note to the subject's keyed response, were be stored in a data file. Test half data was recorded in order to check for internal test consistency and reliability. Test 2: Restricted Classification of Stimuli Twenty-five subjects participated in the second test. The test assessed whether subjects classified a short musical phrase, containing a single note varying in loudness and/or intonation accuracy, on the basis of dimensional preference, of if both component dimensions were saliable. Two of the test phrases were identical on one dimension and differed considerably on the other; a third test phrase differed slightly but discriminably on both dimensions (See Figure 6). In applying Garner's test of integrality or separability of dimensions, if for pitch and loudness

52 $- o C 70 db o C + 46* -- o B 67 db -- o B Pitch Intensity Dbe o A 58 db o A 62 db 65 db Db : Intensity Pitch Figure 6. Pitch and Loudness Values for Test 2. subjects show dimensional preference irregardless of overall similarity, the dimensions are said to be integral since the dimension with the smaller change is not salient. If subjects show the ability to classify according to either dimension, then the dimensions are viewed as separable. Subjects listened to the musical phrase shown in Figure 5. The last note of the phrase varied in intensity and pitch according to the values shown in Figure 6. The degree of variation on both dimensions was based on the values employed by Grau and Kemler-Nelson (1988). The phrases were presented in sets of three, each phrase in the set being different. The required task was to select which

53 43 single musical phrase did not belong in the set. Subjects began the test by reading on-line test instructions, and were given opportunity to ask for clarification of instructions. The subjects then listened to three phrases in an order corresponding to three response areas in the on-screen test panel. When a decision was made as to which phrase did not belong in the set, the subject recorded his choice by moving the cursor over the desired response area and clicking the left mouse button. A fourth response area allowed subjects to click in order to listen to the three phrases again. This was allowed as many times as desired. Data was recorded when a selection was made, and the next three phrases were played following a 500ms delay. The screen prompts and on-line instructions are given in Appendix B. Referring again to Figure 6, one half of the sets to be evaluated shared phrases ending in a note identical in intonation, the other half shared a final note having the same change in dynamic level. The ordering of phrases was varied throughout the test for each subject; subjects were informed a priori that the phrase order within each set of three was varied, with no discernable pattern. Subjects' responses were stored to a file on the disk drive for subsequent analysis and interpretation.

54 44 Subsequent analysis of the data obtained led to ambiguous conclusions. At first glance, responses showed that subjects grouped stimuli according to dimensional preference, indicating integrality. However, the data also did not rule out the possibility that subjects could attend to both dimensions equally well, but simple chose the one having the more obvious change. In order to understand better these initial results, the restricted classification test was repeated three times. Each test used a different level of discriminability in order to observe trends in subjects' responses. Data provided by the Similarity Scaling Test, described below, was used to estimate onestep differences of stimuli and a threshold of discriminability. The first test used equally discriminable stimuli placed just above the limit of discriminability. The second test presented an equal onestep change in the discriminability of each dimension; the third test increased the discriminability of each dimension by a two-step difference. The stimulus values used in these additional tests are shown in Figures 7 through 9.

55 45 94 O B o C 63.5 db- o B o C Pitch Intensity Db 5 -- o A 62 db o A 62 db 63.5 db Intensity Db Pitch Figure 7. Pitch and Loudness Values for Retest 1, Test : 0B oc 65 db o B o C Pitch Intensity Db 5 o A 62 db o A 62 db 65 db Intensity Db 5 Pitch +18fc Figure 8. Pitch and Loudness Values for Retest 2, Test 2

56 d o B o C 65 db -- o B o C Pitch Intensity Dbc oa 62 db o A 62 db 66.5 db Intensity Db Pitch Figure 9. Values for Retest 3, Test 2 Test 3: Similarity Scaling of Stimuli Pitch and Loudness The third test assessed how subjects perceived differences among pairs of stimuli varying along two dimensions. Subjects rated the magnitude of differences among pairs of stimuli. ThV resultant data was analyzed in order to find the best fit to one of two models, described below. If the perceptual distance conforms to a city-block metric, the dimensions are viewed as being separable; if the perceptual distance better conforms to the Euclidean distance, the dimensions are viewed as being integral. Referring to Figure 10, ratings of separable dimensions will tend towards simply adding the values obtained

57 47 (d A B = d x + d Y ) Ratings of integral dimensions will tend to conform to the Euclidean distance between the points (d AB = d e ) (see Garner, 1974, pp ). The stimuli used in this test were constructed using the musical phrase shown in Figure 8. The ending note of the phrase varied in pitch or loudness, or both. The pitch of the final note was de-tuned in 9 cent increments; the loudness of the final note was adjusted in 1.5 db Loudness Intonation Figure 10. Geometric Representation of Stimulus Relationship, Test 3. increments. The loudness of the first four notes remained at a constant 59.5 db, and the intonation of these notes was based on equal tempered tuning a A = 440Hz. The complete set of values that were used for the last note of the phrase are shown in Figure 8. These values

58 48 approximated the values used by Grau and Kemler-Nelson (1988), and are consistent with the values used in the first two tests of the present study. Subjects rated the similarity of pairs of phrases having end notes that 1) corresponded to any two points along the center box in Figure 8; 2) corresponded to two adjacent points on either axis and 3) were identical. This paradigm resulted in 24 zero-step pairs, 24 one-step pairs, 16 three-step pairs, 12 four-step, 8 five-step pairs and 4 six-step pairs. In order to equalize the number of times each pair is presented, the three- and four-step pairs were presented two times, and the five-step pairs three times and six-step pairs five times. This resulted in a total of 152 presentations. Subjects began by reading online test instructions. Instructions described the task and familiarized the subject with the screen prompts. Subjects were given opportunity to ask for clarification of instructions. Prior to data collection, each subject proceeded through a short trial run in order to acclimate to the degree of differences among stimuli encountered in the test. A pair of phrases, separated by a 500ms gap, was played. The subject moved a slider on an on-screen rating scale to

59 49 match his assessment of the degree of difference of the last note in each phrase. The subject recorded his rating O 0 Intensity (db) " " O" 63.5 " O- \ / / K <> / \ \ -*> -o 62.0 " "" -27 <t -18 <t -9<J) + Db5 I h- +9(t +18<t +27<fc Pitch Figure 11. Array of Pitch and Loudness Values, Test 3. by clicking on a button labeled "Enter". Alternatively, a subject could click on a button to repeat the presentation of the current pair of phrases and re-adjust his rating. After recording his rating, the next pair of phrases was presented following a 500ms pause. The on-line instructions and screen prompts are shown in Appendix B. Data collected from subjects' responses were stored to a file on the disk drive for subsequent analysis and interpretation.

60 50 As with the second test, data analysis suggested that additional testing was necessary. This necessity arose from the need to determine if the perceptual rating of the Euclidean distance held true for other Euclidean distances. Additional testing also corroborated subjects' responses from the initial testing. The test was comprised of eight zero-step pairs, eight one-step through six-step pairs, two six-step city-block pairs, and four five-step 68.0 i.5 " ~ O Intensity (db) ~ <t -18 <t -9<t Dbs Pitch +9<t *18<fc +27<fc Six-step difference, three per each dimension. Five-step difference (pitch = 3 and loudness = 2). Five-step difference (pitch = 2 and loudness = 3). Figure 12. Euclidean Distances of Pitch and Loudness used in Additional Testing, Test 3.

61 51 city-block pairs. The six-step city-block pairs were each presented four times each, and the five-step city-block pairs were presented two times each. The total number of pairs presented in each test was 72. The array points used in rating the Euclidean distances in this test are shown in Figure 12. A complete listing of data points used for establishing the zero- through six-step ratings is provided in Appendix C.

62 CHAPTER FOUR TREATMENT OF THE DATA Test 1: Speeded Sorting The first treatment of the data was to assess whether the measures obtained could be considered reliable. A one way analysis of variance was used to compare the main effect of reaction times and test halves (see Table 1). Analysis indicated that test half did not have a statistically significant effect on subjects' reaction times. Table 1 Comparison of Test Halves, Test 1. USER-EXCLUDED OBSERVATIONS: none Response: Reaction Time Source DF SS MS F Pr > F Test Half Some preliminary processing of the data was undertaken prior to statistical analysis of the main effects and interactions. Response times lying outside of a practical 52

63 53 range, as would occur when a subject responded earlier than the completion of the stimuli presentation or delayed their response to the point of being reminded to respond more quickly, were discarded. Of the total 1200 reaction time data points, 44 were discarded, resulting in an overall error rate of 3.6%. In addition, one subject listened to the test without regard to the screen prompts directing attention to pitch or loudness. The subject indicated that all responses provided were in attendance to pitch and not loudness; therefore, all loudness responses given by this subject were excluded in the data analyses and only pitch response reaction times were used. An analysis excluding this subject altogether did not alter statistically significant results. The total number of subjects participating in this test was 25. An analysis of variance was then used to test main effects and interactions among variables (see Table 2). The means of the response variable for reaction time (RT) were derived from each occurrence of the particular variable being considered. The variable, Condition, refers to the three conditions in which the stimuli were presented; the three conditions presented were Control, Orthogonal, and Redundant. The variable, Attention,

64 54 refers the two variables Pitch and Loudness. The main effects, Condition and Attention, were considered Table 2 Main Effects and Interactions, Condition and Attention. Analysis of Variance: ANOVA Table Type I Sum of Squares Response: Reaction Time Source DF SS MS Pr > F CONDITION ATTENTION COND x ATTN *p< E6 4.48E E6 1.16E * * significant at the pc.ol level of confidence. The interaction Condition by Attention was not considered to be statistically significant at the p<.01 level. Further analysis was required in order to understand the relationships among the levels underlying the main effects. A Bonferroni test was used to illustrate where significant differences appeared in the main effects, Condition and Attention, and to explore whether the means of levels within these main effects formed distinct sets. Results for the main effect, Condition, appear in Table 3; results for the main effect, Attention, appear in Table 4.

65 55 The means of the Control and Redundant levels of the main effect, Condition, were not shown to be significantlydifferent, while the Orthogonal level was significantly- Table 3 Predicted Means of Condition Subgroups. Bonferroni Comparisons Response: Reaction Time Test Condition Mean Response Time(ms.) Compares Standard Confidence equal to error Interval(95%) Control Orthogonal Redundant Redundant None Control (14.4) (5592 (14.5) (5774 (14.5) ( ) 5844) 5652) Predicted Means: There are 2 distinct sets of means: (Control, Redundant) and (Orthogonal). different from both the Control and Redundant conditions. Two distinct sets of means, (Control, Redundant) and (Orthogonal) were formed. Thus, the Orthogonal condition produced a statistically significant increase in reaction time over the Control and Redundant conditions. The reaction time for the Redundant level decreased numerically

66 56 in comparison with the Control condition. This decrease was not considered to be statistically significant. Table 4 Predicted Means of Attention Subgroups. Bonferroni Comparisons Response: Reaction Time Pitch and Loudness Mean Attended Response Compares Standard Confidence Dimension Time(ms.) equal to error Interval(95%) Loudness 5652 None (12.5) (5624: 5680) Pitch 5715 None (12.3) (5688: 5743) Predicted Means: The means are significantly different. The means of the Pitch and Loudness levels of the main effect, Attention, were significantly different. Mean reaction time for stimuli where subjects attended to pitch was greater than the mean reaction time for Loudness. Although no significant difference was observed for the effect Condition by Attention, the value for Pr < F was Therefore, further analyses was carried out to determine if the means for the Control, Orthogonal, and Redundant levels maintained the same relationships for both levels of Attention (Pitch and Loudness). Mean reaction

67 57 times of the Control, Orthogonal, and Redundant conditions were analyzed separately for Pitch and Loudness. Results for Pitch appear in Table 5; results for Loudness appear in Table 6. Table 5 Predicted Means of Condition Subgroups for Pitch Response: Reaction Time N = 586 EXCLUDED OBSERVATIONS: LOUDNESS Mean Test Response Compares Confidence Condition Size Time(ms.) equal to SD Interval(95%) Control Redundant (303) (5640:5727) Orthogonal None (339) (5787:5875) Redundant Control (293) (5589:5677) Means: There are 2 distinct sets of means: {Control,Redundant} and {Orthogonal}. The data in Tables 5 and 6 show that the ordering of means for Pitch and Loudness considered separately was the same as for when Pitch and Loudness reaction times were pooled together. In both cases the Orthogonal condition exhibited the greatest time and the Redundant condition the shortest time. Similarly, the same two distinct sets of

68 58 Table 6 Predicted Means of Condition Subgroups for Loudness Response: Reaction Time N = 570 EXCLUDED OBSERVATIONS: PITCH Mean Test Response Compares Confidence Condition Size Time(ms.) equal to SD Interval (95%) Control Redundant (239) (5533:5604) Orthogonal None (284) (5751:5823) Redundant Control (229) (5566:5638) Means: There are 2 distinct sets of means: {Control,Redundant} and {Orthogonal}. means, (Control, Redundant) and (Orthogonal) were formed. Therefore, the subjects responses were in effect identical for both Pitch and Loudness. Test 2: Restricted Classification The second test required subjects to select for omission one phrase from a set of three phrases whose last note varied in intonation and loudness. Thirty-one subjects participated in the procedure. In order to ensure that subjects responded consistently throughout the test, data from the first and second halves were compared. The percentages for each possible response were nearly

69 59 Table 7 Comparison of Test Halves, Test 2 Test Cumulative Cumulative Half CHOICE Frequency Percent Frequency Percent j^st A ^St B ^St C nd A nd B nd C identical, indicating that subjects were consistent in their responses throughout the duration of the test, and did not change strategies for selection of the phrase to be excluded from the set (see Table 7). As shown in Table 8, the pattern of subjects' responses clearly showed preference for excluding the + 61*- o C 70 db o C + 46t O B 67 db - OB Pitch Intensity Dbc o A 58 db o A H h 62 db 65 db Intensity H Db 5 b +15fc Pitch Figure 13. Pitch and Loudness Values, Test 2

70 60 Table 8 Frequency and Percent of Responses for Pitch and Loudness. Cumulative Cumulative SET CHOICE Frequency Percent Frequency Percent abc(pitch) A abc(pitch) B abc(pitch) C abc(loudness) A abc(loudness) B abc(loudness) C phrase whose note changed by a large degree on the single dimension of either pitch or loudness (point A in Figure 13). Total response frequencies and percentages are shown in Table 9. The rare selections of point B (Figure 13) as a response was not an expected condition for the model, and thus was considered to be attributable to subject error. Table 9 Combined Frequency and Percent of Responses. Cumulative Cumulative CHOICE Frequency Percent Frequency Percent A and D B and E C and F

71 61 From the data obtained, it was not clear whether subjects grouped data due solely to dimensional preference, or if subjects were merely selecting the larger dimensional change among two equally salient dimensions. Therefore, additional tests were carried out using three different sets of stimuli. The first test used stimuli that were equally discriminable, at a level just at the boundary of discriminability. A summary of the data from the first test appears in Table 10. The data indicated that subjects did not exhibit any dimensional preference. This was expected and verified that the stimuli was in fact at the limit of discrimination. Furthermore, the equal selection of point 'B', the error response, also indicated that the phrases were too similar to be discerned. In the Table 10 Frequency and Percent of Responses for Pitch and Loudness (Re-test, equally-discriminable stimuli) Cumulative Cumulative SET CHOICE Frequency Percent Frequency Percent abc(pitch) A 85 35, ,.4 abc(pitch) B , ,.2 abc(pitch) C 74 30, abc(loudness) A , abc(loudness) B 79 32, ,.7 abc(loudness) C ,.0

72 62 second test the discriminability of the stimuli was increased to equal a one-step difference. Data from the second test is shown in Table 11. Subjects showed a verystrong tendency to keep the two phrases having the same pitch level together, regardless of the loudness level. The occurrence of the error response, 'B', was greatly reduced. Thus, subjects showed dimensional preference for pitch, even in the case where loudness had the greater dimensional change. The third test increased the discriminability of each dimension by a two-step difference. Data collected using this set of stimuli is summarized in Table 12. Subjects exhibited approximately equal preference for pitch and loudness in sorting the Table 11 Frequency and Percent of Responses for Pitch and Loudness (Re-test, stimuli at one-step difference.) Cumulative Cumulative SET CHOICE Frequency Percent Frequency Percent abc(pitch) A abc(pitch) B abc(pitch) C abc(loudness) A abc(loudness) B abc(loudness) C

73 63 Table 12 Frequency and Percent of Responses for Pitch and Loudness (Retest, stimuli at two-step difference.) Cumulative Cumulative SET CHOICE Frequency Percent Frequency Percent abc(pitch) A , abc(pitch) B 11 4., abc(pitch) C abc(loudness) A abc(loudness) B ,.0 abc(loudness) C 84 35, ,.0 phrases. The equal preference occurred irregardless to a larger change on either dimension. The number of error responses ('B 1 ) was very low, indicating that the subjects could easily sort the phrases according to either intonation or loudness. Test 3: Similarity Scaling The third procedure required subjects to rate the degree of difference between the final note among pairs of phrases. The difference among the final notes varied from zero to six steps difference along a single dimension, or by five- and six-step differences split among the two dimensions. A Pearson product-moment correlation was used to determine how well subjects' ratings compared to the actual geometric distances. A moderately high positive

74 64 correlation was observed between subjective and actual perceptual distances (see Table 13). This indicated that Table 13 Pearson Product-moment Correlation, Rated vs. Actual. Correlations N = 1320 (all) ACTUAL RATED ACTUAL < RATED < the subjects' ratings were in fact correlated to the actual stepwise differences among the stimuli. An analysis of variance was then used to determine if there existed a statistically significant difference among the means of subjects' ratings of the magnitude of differences between the final notes at the different levels presented in the model. As shown in Table 14, there was strong statistical evidence of differences among rated means (RATED) at different levels of the actual geometric distance (ACTUAL). The means of subjects' ratings and sets of means are shown in Table 15. There were four distinct sets of means.

75 65 The perceptual distance between the diagonal center points was 1.9. This value was grouped with the set of stimuli ratings for 2 and 3 step differences. The value is also smaller than the actual Euclidean distance of Furthermore, the confidence interval for the perceptual Table 14 Analysis of Variance, Subjective and Geometric Distance. Model: RATED vs. ACTUAL Source DF SS MS F Pr > F Model Error Total ANOVA Weighted for Unequal Variances F Df num Df denom Pr > F distance rating of the diagonal points is inclusive of Grau and Kemler-Nelson's value of 1.6. Therefore, the perceptual distance rating of stimuli varying on both pitch and loudness conform better to the Euclidean metric than to the city-block metric. Additional testing was done to ascertain if ratings of other values of stimuli also conformed to the Euclidean

76 66 Table 15 Test 3 Means and Sets of Means. Response = RATED ACTUAL Response Compares Confidence Group Size Mean equal to SD Interval (95%) (0.854) (0.392:0.828) (1.032) (0.661:0.948) diag (1.307) (1.308:1.623) diag (1.590) (2.204:2.556) None (1.530) (2.841:3.248) (1.440) (3.318:3.816) (1.504) (3.763:4.468) DIAG ,3 (1.120) (1.637:2.167) There are 4 distinct sets of means: (0,1), (2,DIAG,3), (4), and (5,6) metric. This additional testing included five- and sixstep city block distances, as well as pairs changing along a single dimension by zero- through six-step distances. The stimuli consisted of eight instances each of zerothrough six-step pairs, four five-step city-block pairs, and two six-step city-block pairs. This created a total of 72 comparisons for the entire test. The array points used in rating the Euclidean distances in this test are shown in Figure 12. A Pearson product-moment correlation was used to determine how well subjects' ratings compared to the actual

77 67 geometric distances. As in the initial perceptual scaling test, a moderately high positive correlation was observed between subjective and actual perceptual distances (see Table 16). An analysis of variance was used to determine if there existed a statistically significant difference Table 16 Pearson Product-moment Correlation, Rated vs. Actual (Re-test) Correlations N = 1512 (all) ACTUAL RATED ACTUAL < RATED < among the means of subjects' ratings. As shown in Table 17, there was strong statistical evidence of differences among rated means (RATED) at different levels of the actual geometric distance (ACTUAL). The means of subjects' ratings and sets of means are shown in Table 18. There were five distinct sets of means. The exact Euclidean perceptual distances for the five-step and six-step city-block distances are 3.60 and 4.24,

78 68 Table 17 Analysis of Variance, Subjective and Geometric Distance, Test 3 (Re-test) Model: RATED vs. ACTUAL Source DF SS MS F Pr > F Model Error Total Table 18 Test 3 Means and Sets of Means. (Re-test) Response = RATED ACTUAL Response Compares Confidence Group Mean equal to SD Interval (95%) None ( 0.116) ( : 0.663) ( 0.096) ( : 1.357) ( 0.096) ( : 1.777) None ( 0.096) ( : 2.732) ,4.24 ( 0.096) ( : 3.396) ( 0.096) ( : 4.334) ( 0.096) ( : 4.555) DIAG ( 0.098) ( : 3.343) DIAG ( 0.098) ( : 3.790) There are 5 distinct sets of means: : ( 0 ), ( 1, 2 ), ( 3 ), ( 4, DIAG 3.6, DIAG 4.2), and ( 5, 6 ). respectively. The respective subjective distance ratings for these relationships were 3.07 and Thus, subjects rated that differed along both dimensions much closer to

79 69 their Euclidean distance than the sum of the dimensional distances. This is consistent with the results obtained in the first set of values used in the present study and with results obtained by Grau and Kemler-Nelson.

80 CHAPTER FIVE SUMMARY, SUGGESTIONS FOR FURTHER RESEARCH Background When evaluating a single attribute or dimension of a stimulus, subjects may experience cognitive interference induced by a change on another dimension of the stimulus. Other component dimensions of a stimulus may not produce cognitive interference. When a change in one dimension results in interference with another, the two dimensions are said to be "integral" in nature. In cases where no interference occurs, the two dimensions are said to be "separable" (Garner, 1974). Interference among integral dimensions has been termed "Garner Interference" (Pomerantz, 1986). The Study The purpose of the study was to explore whether musicians experience Garner interference among the auditory dimensions of pitch and loudness. Specifically, the study explored whether intonation and loudness, when to presented 70

81 71 musicians in a five-note musical pattern, were perceived as integral or separable in nature. Three convergent tests were employed, each using a five-note musical pattern in which the final note varied in intonation and loudness. The first test required subjects to make speeded sorting judgements on changes in intonation and loudness. Reaction times were measured for blocks of phrase pairs presented in three conditions. These conditions were: 1) the Control condition, where the stimulus changed on a single dimension, 2) the Redundant condition, where both dimensions changed simultaneously and 3) the Orthogonal condition, where the stimulus changed on the irrelevant dimension. The second test used a restricted classification task to assess whether subjects grouped sets of phrases according to their dimensional structure or if both dimensions were equally salient to the subject. The third test explored whether the subjective distance between notes varying in both intonation and loudness better matched either the sum of the changes on both dimensions, or the Euclidean measure of the distance.

82 72 Findings Research Questions #1 and #2 Research question #1 asked whether musicians would exhibit signs of performance interference in a speeded sorting task, when evaluating changes in either intonation or loudness of the last note of a five-note musical phrase, in the face of a change on the irrelevant dimension. Research question #2 asked whether musicians would exhibit signs of performance facilitation in a speeded sorting task, when evaluating five-note musical phrases whose final note changes along both dimensions of intonation and loudness. Both questions were addressed through the same speeded sorting task. The main effect, Condition, encompassed the three conditions of Control, Orthogonal (used to assess performance interference) and Redundant (used to address performance facilitation). Significant differences were observed among the reactions times for the three levels of Condition. A comparison of the means showed the Orthogonal condition to produce significantly longer reaction times than the Control condition. The comparison of means also showed a numerically lower but statistically insignificant reduction in reaction time for the Redundant condition.

83 73 The reaction time means formed two distinct subsets: the Control and Redundant conditions formed one subset, while the Orthogonal condition was in its own set. The main effect, Attention, included the two conditions of Pitch and Loudness. These conditions corresponded to the component dimensions of intonation and loudness, to which the subject was directed to attend. There was a significant difference among reaction times for the main effect, Attention. A comparison of the means showed that the mean reaction time when attending to Pitch was greater than the mean reaction time for Loudness. No significant difference was reported for the case, Condition x Attention (Pr > F 0.083, p<.01). Since this may be considered significant at a less conservative p value, reaction time means were compared separately for the conditions of Loudness and Pitch to see if the order and subgroups differed from the pooled data. The relationship among the means and distinct subgroups were statistically the same whether pooled or considered separately. Evidence from the present study showed interference for the Orthogonal condition through increased mean reaction time. The first research question is therefore answered in the affirmative: Musicians exhibited signs of

84 74 performance interference in the speeded sorting of fivenote musical phrases according to pitch and loudness when faced with a change on the irrelevant dimension. There was no statistical evidence of performance gain with correlated changes of intonation and loudness. Thus, the answer to the second research question is that musicians did not exhibit signs of performance facilitation in the speeded sorting task. Question #3 The third research question asked whether musicians preferred to sort sets of five-note musical phrases, wherein the discriminability of the intonation and loudness of the last note was varied, according to preference for the more discriminable dimension, or if musicians found the dimensions of intonation and loudness to be equallysalient. Although the first round of testing in the restricted classification procedure produced a distinct response pattern, the values used for the initial round did not provide meaningful insight into the perceptual relationships among the stimuli. The cumulative percentages showed that subjects preferred to group the phrases in the same manner as would be expected for integral dimensions. However, the data could not show

85 75 whether subjects could perceive the excluded dimension. Therefore, selection of the preferred dimension may have been only due to its greater difference in magnitude. Therefore, additional testing was carried out. Three stimulus sets were used in retesting. The first used stimuli that were equally discriminable, at a level just at the boundary of discriminability. As expected, subjects' responses appeared to be random in nature, indicating that it was not possible to accurately sort the stimuli in this case. The second test increased the discriminability of the stimuli equal to a one-step difference. Subjects showed a very strong tendency to keep the two phrases having the same pitch level together, regardless of the loudness level. Subjects showed dimensional preference for pitch, even in the case where loudness had the greater dimensional change. The stimulus set in the third test increased the discriminability of each dimension to a twostep difference. Subjects exhibited approximately equal preference for either dimension in sorting the phrases. The equal preference for intonation or loudness occurred regardless of a greater change on either dimension, indicating that both dimensions were quite salient to the subjects. The number of error responses was very low,

86 76 indicating that subjects had little difficulty in performing the sorting task. In considering the results of the additional testing, it was concluded that the results obtained in the initial restricted classification procedure reflected subjects' attention to the large change in magnitude on a single dimension rather than interference. Results of the additional testing suggested that subjects were able to attend selectively to either dimension in the face of a larger discriminable change in the other dimension. Therefore, question #3 is answered as follows: Musicians did not prefer to sort sets of five-note musical phrases, wherein the discriminability of the intonation and loudness of the last note varied, according to preference only for the more discriminable dimension. Both intonation and loudness were equally salient in the restricted classification task. Question #4 The fourth research question explored whether musicians' scaling of differences among five-note phrases would produce either 1) a pattern consistent with the Euclidean distance between two diagonal points representing change on both dimensions, or, 2) a pattern more consistent

87 77 the "city-block" distance between the two points. The same five-note pattern as used in the other tests was employed, wherein the intonation and loudness of the final note was varied. Subjects made judgements as to the magnitude of difference in the final note among pairs of five-note patterns. In order to determine if results remained consistent for various inter-stimulus distances, the similarity scaling test was given two times using different sets of stimuli. The Pearson product-moment correlation comparing the actual distances between points with subjects' ratings was for the first test and for the second test. This showed a reasonable level of positive correlation between subjective ratings and the actual geometrical distances. In both tests, an analysis of variance showed statistically significant differences ( <.001) among means of subjects' ratings of the differences among stimuli and the actual geometric distance. Four distinct sets of means were formed in the first round of testing. Ratings of stimulus differences of zeroand one-step differences were statistically the same, as were ratings of two-step, three-step, and diagonally-

88 78 orientated stimuli (those stimulus pairs with changes on both dimensions). Ratings of stimuli differing by four steps on one dimension stood alone as a set, and ratings of five- and six-step differences formed yet another set. Thus, the response means were grouped as {0,1}, {2,diag,3}, {4} and {5,6}. If subjects had rated the differences among stimuli changing on both dimensions according to the sum of the dimensional changes (the "city-block distance), the mean subjective rating for diagonally-orientated stimuli would be grouped with {4}, since the stimuli varied by two steps on each dimension (see Figure 14). However, the mean subjective rating for the pair of diagonally orientated phrases was included in the set {2,diag,3}. As shown in Figure 14, the actual length of the hypotenuse is 2.83; the subjective rating of the same distance was 1.9, with a 95% confidence interval of (1.637:2.167). The subjective rating obtained was even smaller than the Euclidean distance. Therefore, for the first test the Euclidean metric was a much better fit than the city-block metric distance. The results of a second round of testing reinforced the findings of the first. Means of subjective ratings

89 79 Loudness 2 step change Intonation City-Block distance (AC + AB)= 4 Actual Euclidean distance (AB) = 2.83 Subjectively rated distance = 1.9 Figure 14. Actual and Subjective Distances 4-step cityblock stimulus distances. Loudness 3 step change Intonation 3 step change Intonation City-Block distance (AC + AB)= 6 Actual Euclidean distance (AB) = 4.24 Subjectively rated distance = 3.52 Figure 15. Actual and Subjective Distances for 6-step cityblock stimulus distances.

90 80 formed five distinct sets: (0),{1, 2},{3},{4, 3.6, 4.24}, and {5, 6}. Referring to Figures 15 and 16, the actual city-block distances for the diagonal relationships in the set of stimulus values were 5 and 6 steps, with corresponding Euclidean distances of 3.60 and 4.24, respectively. Subjective ratings for the same two diagonal distances were 3.07 and 3.52, and were similar in nature to the subjective ratings in the first round of testing. Loudness B 2 step change Intonation C B 3 step change Intonation City-Block distance (AC + AB)= 5 Actual Euclidean distance (AB) = 3.60 Subjectively rated distance = 3.07 Figure 16. Actual and Subjective Distances for 5-step cityblock stimulus distances. Based on the analyses of the subjective ratings of both tests, question #4 can best be answered as follows: Musicians' scaling of differences among pairs of five-note

91 81 phrases, ending with a note that changes along the dimensions of intonation and loudness in stepwise fashion, best fit the Euclidean distance between the two points. Research Question #5 The fifth research question asked whether the results of procedures used to answer research*questions one through four are supportive either of integrality or separability of the dimensions of musical intonation and loudness. The strongest support for the integrality of pitch, in the form of musical intonation, and loudness was provided by the similarity scaling procedure. Musicians consistently rated perceptual distances among stimuli as smaller than the sum of the changes on both dimensions. The speeded sorting task produced mixed results. Interference in intonation and loudness judgements in the face of changes on the irrelevant dimension were consistent with expectations for integral dimensions. Performance improvement for a redundant change on both dimensions is also consistent with expectations for integral dimensions. There was, however, no statistically significant improvement in reaction time for the redundant condition. Results from the restricted classification procedure supported the separability of the dimensions of intonation and loudness. In response to the

92 82 fifth research question, loudness and pitch (in the form of musical intonation) show strong evidence of integrality in certain tasks, but show an equally strong tendency towards separability in other cases. Conclusion The results of the convergent tests showed evidence of cognitive interference in musicians' processing of intonation and loudness information. Some, but not all, results obtained were in agreement with studies that investigated interference in the processing of pitch and loudness. Results of the speeded classification task are in agreement with findings of other studies in that interference occurred in the face of orthogonal change in stimuli. Interference effects similar in nature to those reported by Woods (1973), Melara and Marks (1990b) and Grau and Kemler-Nelson (1988). However, the latter two studies reported performance facilitation for correlated changes along both dimensions. No such facilitation was observed in the present study. The pattern of responses obtained in the original round of restricted classification testing seemed initially to be in agreement with results obtained by Grau and

93 83 Kemler-Nelson (1988). However, the test results of the first round of testing in the present study, as well those reported by Grau and Kemler-Nelson, seemed to allow for dual interpretation: Subjects may have been grouping stimuli according to overall similarity (as Grau and Kemler-Nelson concluded) or subjects may simply have been classifying stimuli according to the larger dimensional change regardless of whether the other dimension was salient. Garner (1974, p ) suggested that use of dimensional preference in classification, regardless of greater discriminability of another dimension, was evidence for separable dimensions. Results obtained herein from additional testing showed that subjects categorized stimuli equally well according to either dimension, without regard to degree of discriminibility, which suggested that intonation and loudness were in fact separable in nature. The strongest evidence of cognitive interference was found in the perceptual scaling of differences in the dimensions of intonation and loudness. In the present study, distance ratings between stimuli varying along both dimensions were less than the actual Euclidean distances. Grau and Kemler-Nelson (1988) reported similar results, using isolated tones and non-musicians as subjects.

94 84 Discussion Central to the issue of the integrality of stimulus dimensions is consideration for the nature of the stimulus being processed. This study shows that, under certain conditions, musicians exhibit signs of cognitive interference in evaluating changes in intonation and loudness. However, musicians did not show evidence of interference in the restricted classification task. Perhaps musical training gave musicians better perceptual strategies in grouping comparisons of intonation and loudness. Another expectation in the present study was increased performance in the redundant condition of the speeded sorting task. Results in this study found no such facilitation of performance. Perhaps musicians, as a result of musical training, made decisions at an optimized rate that could not be improved upon by increasing the overall perceptual change of stimuli. Another possibility is that subjects employed a degree of selective attention in the control task and maintained that selective attention through the redundant condition. Perhaps more contradictory is the comparison of results from the restricted classification procedure with the results from the perceptual scaling test, the former

95 85 suggesting that intonation and loudness are separable dimensions while the latter supports integrality. The restricted classification procedure primarily indicated the ability of subjects to attend to either dimension; it did not require subjects to evaluate the degree of difference among stimuli presented in the sets. In the present study it was found that subjects could easily sort the stimuli according to either dimension; therefore, both dimensions were accessible to the subjects. However, the perceptual scaling test showed that subjects consistently underestimated the magnitude of the change in the stimulus for concurrent changes in intonation and loudness. The physical variations in pitch and dynamic level of the stimulus were not large enough to attribute this underestimation to physiological considerations such as those effects observed by Stevens (1935). Therefore, the underestimation of stimulus distances seems to arise from an integration of intonation and loudness information. Garner (1974) proposed that there likely exists degrees of integrality rather than a simple dichotomy between integrality and separability. Based on the present study and on the findings of other studies, it is likely that there is some degree of integrality among the dimensions of

96 86 intonation and loudness, the occurance of which is taskdependent in nature. Suggestions for Further Research Garner (1974) stressed the importance of the use of converging operations to assess the validity of concepts regarding cognitive processes, rather than to rely on a single procedure. The findings of the present study, through such a set of convergent operations, provided evidence that intonation and loudness behave as integral dimensions under certain conditions. However, further research is needed to better understand the nature of cognitive interference in the processing of musical stimuli, and which conditions may produce such interference. In addition, it remains to be studied whether musical training can affect the degree of cognitive interference among auditory dimensions. Evidence of crossmodality of cognitive interference suggests additional areas where psychological dimensional structure may differ from the energic properties of the stimulus. There remains a need for further investigation in order to gain a better understanding of the cognitive processes involved in musical perception.

97 APPENDIX A FORMS 87

98 88 Approval for Testing of Human Subjects 3 University of North Texas Sponsored Projects Administration November 30, 1998 Gary T. Cattley 13 Independence Drive Bordentown, NJ Re: Human Subjects Application No Dear Mr. Cattley: As permitted by federal law and regulations governing the use of human subjects in research projects (45 CFR 46), I have conducted an expedited review of your proposed project titled "Cognitive Interference Among the Auditory Dimensions of Intonation and Loudness in A Five-Note Musical Pattern." The risks inherent in this research are minimal, and the potential benefits to the subjects outweigh those risks. The submitted protocol and informed consent form is hereby approved for use of human subjects on this project, provided that you closely monitor the loudness level of the stimuli to ensure that the level does not exceed 70dB. The UNT IRB must re-review this project prior to any modifications you make in the approved project. Please contact me if you wish to make such changes or need additional information. If you have any questions, please contact me. Sincerely, Sandra L. Terrell, Chair Institutional Review Board ST:sb cc: IRB Members 2 P.O. Box Denton. Texa* <940> Fax <940i TDD (800) c.'-mtiil: I;>ne(0 abn.uoi.edu

99 89 Cover Letter - Subject Testing COVER LETTER I am a doctoral candidate in the College of Music at the University of North Texas and I am conducting a research project concerning the way in which musicians process intonation and loudness information in musical phrases. I am in need of subjects for a listening test and would greatly appreciate your voluntary participation in this study if you have the time to do so. The test involves making assessments regarding differences among intonation and loudness in musical phrases, and is approximately thirty minutes in duration. Personal information will not be collected regarding individuals; all data recorded during the test will remain anonymous in nature. There is no personal risk or discomfort involved in taking this test and you are free to stop the testing at any time should you feel the need to do so. Should you have any questions or concerns that arise in connection with your participation in this study, please contact Gary Cattley at (609) Thank you for your consideration. Sincerely, Gary Cattley This study has been reviewed by University of North Texas Committee for the Protection of Human Subjects (Phone: ).

100 APPENDIX B TEST INSTRUCTIONS AND SCREEN PROMPTS 90

101 91 Test 1 On-Screen Instructions TEST 1 The test that you are about to take requires you to make rapid comparisons regarding a change in the last note among pairs of five-note musical phrases. In some cases the basis for comparison will be intonation; in other cases the comparison will be based on loudness. The differences do not occur in any set pattern, and the number of pairs that sound the same may not equal the number that sound different. Musical phrases will be presented in pairs. Listen carefully to the last note of each phrase in the pair. Then, following the screen prompt, press the key (A or L) that corresponds to your decision. Please be sure to follow the screen prompt that indicates whether to attend to pitch or loudness. This is especially important, as there will be cases where both the pitch and loudness of the last note will be different. Please make your choices immediately upon the completion of the final note of the second phrase. The test window will display a prompt if a faster response is required. Thank you for your participation!

102 92 Test 2 On-screen Instructions TEST 2 The test that you are about to take will present five-note musical phrases in sets of three. The final note of each phrase will vary in loudness, accuracy of intonation, or both loudness and intonation. You will be asked to select which phrase, as presented in A-B-C order, does not belong in the set. Your decision will be recorded by moving the cursor over the A, B, or C response boxes and clicking the left mouse button. Any set may be repeated by selecting "Play Again" in the onscreen response box. Unlike the first test, the present task does not require rapid responses. Proceed through the examples at a comfortable pace. A sample run of the test will be given first, in order for you to develop a basis on which to make your judgements. Thank you for your participation!

103 93 Test 3 On-Screen Instructions TEST 3 In the test that you are about to take, you will be making judgements as to the magnitude of difference in the last note in pairs of five-note phrases. The final note of each phrase in each pair will vary in loudness, accuracy of intonation, or both loudness and intonation. Your estimation of the magnitude of difference will be recorded by moving the slider on the six-point scale (displayed in the on-screen test panel) to a number that corresponds to how different you think the final notes are. There are three ways that you can move the slider across the rating scale. 1) The slider may be moved by placing the cursor over it and holding the left mouse key down. 2) The slider will also move by placing the cursor over the desired number and clicking the mouse until the slider is in the desired position. 3)You can click on the scale to the left or right of the slider; the slider will move one degree towards the cursor each time you click the left mouse button. This test does not require a rapid response. This is, however, the longest test of the three. Thus, it is desirable to work as efficiently as possible in order to complete the test in a reasonable time. A sample run of the test will be given first, in order for you to develop a basis on which to make your judgements.

104 94 Screen Prompts Test 1 Read the Instructions, then press the spacebar to begin. ml m\ Compare only the last note of each phrase. Attend to: PITCH»Dld the PITCH change?«a - NO CHANGE L = CHANGE Please respond faster! Press A or L to continue.. Compare only the last note of each phrase. Attend to: LOUDNESS m\»did the LOUDNESS change?«a = NO CHANGE L = CHANGE This Test is done m o n Thank you You're Done 0

105 95 Test 2 You will hear phrases presented in sets of three. m Please select ONE phrase that DOES NOT belong in the set of three. B Play again ^Tnis Tesl is done m o n Thank you You're Done H

106 96 Test 3 You will hear two musical phrases. The last note of each phrase may differ in its loudness or intonation. Listen and compare the last notes of each pair of phrases. Use the mouse to move the slider in order to rate how different the ending notes are. 0 = IDENTICAL 6 = MOST DIFFERENT qph= Play Again MMWil KMy 6 Cnfar inwlh p! g^this Test is done o Thank you n You're Done 0