The Integration of Stimulus Dimensions in the Perception of Music

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY iFirst,1–28

The integration of stimulus dimensions in the perception of music

Jon B. Prince School of Psychology, Murdoch University, Murdoch, WA, Australia

A central aim of cognitive psychology is to explain how we integrate stimulus dimensions into a unified percept, but how the dimensions of pitch and time combine in the perception of music remains a largely unresolved issue. The goal of this study was to test the effect of varying the degree of conformity to dimensional structure in pitch and time (specifically, tonality and metre) on goodness ratings and classifications of melodies. The pitches and durations of melodies were either presented in their original order, as a reordered sequence, or replaced with random elements. Musically trained and untrained participants (24 each) rated melodic goodness, attending selectively to the dimensions of pitch, time, or both. Also, 24 trained participants classified whether or not the melodies were tonal, metric, or both. Pitch and temporal manipulations always influenced responses, but participants successfully emphasized either dimension in accordance with instructions. Effects of pitch and time were mostly independent for selective attention conditions, but more interactive when evaluating both dimensions. When interactions occurred, the effect of either dimension increased as the other dimension conformed more to its original structure. Relative main effect sizes (| pitch h2 – time h2 |) predicted the strength of pitch–time interactions (pitch × time h2); interactions were stronger when main effect sizes were more evenly matched. These results have implications for dimensional integration in several domains. Relative main effect size could serve as an indicator of dimensional salience, such that interactions are more likely when dimensions are equally salient.

Keywords: Pitch; Time; Music; Dimensions; Integration.

The cognitive mechanisms responsible for the per- dimensions combine in order to form a uniﬁed Downloaded By: [Murdoch University] At: 03:47 19 May 2011 ceptual integration of stimulus dimensions are and coherent mental percept is a primary goal complex and multifaceted. Nonetheless, percep- of cognitive psychology (Treisman & Gelade, tion of multidimensional objects in the natural 1980). Therefore much research has concentrated environment proceeds automatically and without on the central issue of how stimulus dimensions conscious effort. Understanding how these combine in perceptual processing.

Correspondence should be addressed to Jon B. Prince, School of Psychology, Murdoch University, Murdoch, WA 6150, Australia. E-mail: [email protected] Grants from the Natural Sciences and Engineering Council of Canada to Mark A. Schmuckler and William F. Thompson sup- ported part of this research, in addition to grants from the National Science Foundation of the USA awarded to Peter Q. Pfordresher. Portions of this work were presented at the Auditory Perception, Cognition, and Action meeting, Boston, MA, November 2009. The author would like to thank Peter Q. Pfordresher for helpful comments and discussions.

# 2011 The Experimental Psychology Society 1 http://www.psypress.com/qjep DOI:10.1080/17470218.2011.573080 PRINCE

Seminal work by Garner (1974) introduced the Despite considerable advances in the field of concept of integral and separable dimensions, spe- multidimensional visual object perception, less is cifying that integral dimensions cannot be isolated known about integration of multidimensional in perceptual processing, whereas separable properties in patterned auditory stimuli such as dimensions can be processed independently. speech and music. Rather than representing a When attending a stimulus comprised of integral static object, auditory sequences unfold over dimensions, changes along one dimension will time. Accordingly, integrating the dimensions of affect judgements along the other dimension; for such sequences has an additional layer of complex- separable dimensions no such interfering effect ity by virtue of their dynamic nature. In the case of arises. However, the integrality or separability of music, the main stimulus dimensions of pitch and dimensions is influenced by their relative discri- time can be manipulated independently, and there minability, which corresponds to the baseline dif- is a rich background on these two dimensions from ficulty of perceiving a given dimension (Garner & the perspectives of music theory (Aldwell & Felfoldy, 1970) and can be measured using reac- Schacter, 2002; Schenker, 1935/1979) and cogni- ′ tion times, accuracy, or d . A more discriminable tive science (Jones & Boltz, 1989; Krumhansl, dimension will interfere with a less discriminable 1990). Nevertheless, there is still little agreement dimension, even if they have demonstrated separ- on how exactly pitch and time contribute to, and ability under conditions of equal discriminability. combine in, our internal representation of music Yet many dimensions defy a neat classification (this process is referred to hereafter as pitch– into integral and separable dimensions. For time integration). Responses to any stimulus will example, asymmetric interactions may occur even depend on the internal representation of its fea- when dimensions have been matched for discri- tures, and music is no exception to this principle. minability, such that one dimension interferes Thus, understanding how pitch and time integrate with another, but not vice versa (Garner, 1976). is vital to research on musical behaviour (listening, Further research refined the original framework performing, and composing) and can contribute to of integral and separable dimensions into a more the larger literature on dimensional integration in detailed and nuanced view of dimensional inte- other sensory modalities. gration. One such development was the idea of Much research on pitch–time integration has decisional and perceptual separability (Ashby & taken a position supporting either the idea that Townsend, 1986), which proposes that a dimen- pitch and time are independent, or that they sion may interfere with another at either of these are interactive (as reviewed in Ellis & Jones, processing levels when responding to a multi- 2009; Krumhansl, 2000; and Prince, Thompson,

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 dimensional stimulus. Other authors have pro- et al., 2009). Palmer and Krumhansl (1987a) proposed the concepts of dimensional imbalance and vided one of the landmark demonstrations of inde- dimensional uncertainty (Melara & Algom, pendence, where an additive combination of pitch 2003), theorizing that the amount of information and temporal information predicted judgements of present in a given dimension inﬂuences how the melodic phrase goodness, suggesting that pitch perceiver attends to the stimulus. Additionally, and time were processed separately. Other research some stimulus dimensions appear to be privileged proposed instead that pitch and temporal infor- and can exert dominance (i.e., asymmetric mation is processed jointly, because variations in interference) over other dimensions despite rhythmic structure inﬂuenced detection of pertur- careful matching of other factors (Atkinson, bations to pitch patterns (Jones, Boltz, & Kidd, Tipples, Burt, & Young, 2005; Haxby, Hoffman, 1982). Since then, the numerous reports of inde- & Gobbini, 2000; Mullennix & Pisoni, 1990; pendence and interaction of these dimensions Prince, Thompson, & Schmuckler, 2009; have led to much confusion over the nature of Schweinberger, Burton, & Kelly, 1999; Tong, pitch–time integration. It is more likely, therefore, Francis, & Gandour, 2008). that pitch and time function neither as purely

2 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

independent nor as interactive dimensions, but of individual (local) stimulus elements foster that pitch–time integration varies on the basis of observations of independence; conversely, inter- a number of stimulus, task, and cultural factors. actions are more common when attending to In real-life situations, the characteristics of the characteristics of a stimulus on a larger time- music that a listener hears (i.e., the stimulus) scale (global tasks). Because attending to larger, may highlight one dimension, such as emphasizing more global structural relations will require com- the pitch dimension by using a repeating pattern, bining a number of preceding events, such tasks or the time dimension by accelerating the tempo. are likely to require processing of the stimulus in Similarly, listeners may choose to focus on differ- an integrated form. This integrated internal rep- ent aspects of a musical sequence, such as listening resentation would include both pitch and temporal to a particular instrument, a rhythmic figure, or the information; thus these stimulus dimensions are timbral features of the music. Based on what more likely to interact in perceptual processing. aspect is of interest, the task of the listener will Conversely, attending locally to an isolated event involve isolating different dimensions of the can proceed without reference to more global music. Finally, listeners encultured in the properties of a stimulus, not depending on a Western musical style will probably hear music joint (integrated) representation, and therefore of another culture differently from a native listener fostering independent relations. Although typical (e.g., Castellano, Krumhansl, & Bharucha, 1984). listening situations are likely to involve attending Recent work has examined the conditions to the global characteristics of a musical excerpt under which musical pitch and time combine inde- (e.g., the contour of a melody, or the musical pendently or interactively. Among the existing tension and relaxation of a sequence), there are theories on variations in pitch–time integration scenarios in which the listener may focus locally patterns, the most well established are an early- on a single event, such as a particularly striking versus late-stage processing model (Peretz & musical occurrence, or listening for a favourite Kolinsky, 1993; Pitt & Monahan, 1987; chord in a familiar piece. Thompson, Hall, & Pressing, 2001) and local However, not all findings concur with these versus global task characteristics (Bigand, preceding theories. Prince, Thompson, et al. Madurell, Tillmann, & Pineau, 1999; Jones & (2009) reported independent and interactive con- Boltz, 1989; Tillmann & Lebrun-Guillaud, 2006). tributions of pitch and time when responding to Work on stages of processing proposes that different aspects of a probe event (an isolated pitch and time in music are initially processed single tone) following a melody. In one task, lis- independently and integrate at later stages teners rated how well the probe event fitted with

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 (Peretz & Coltheart, 2003). Accordingly, depen- the melody; pitch and time contributed to good- dent measures that rely on early processing (such ness ratings with additive (independent) contri- as perceptual judgements of pitch height or butions. In another task, listeners classified tempo) are likely to demonstrate independence whether the probe event was “on the beat” (if it because the dimensions are still being processed was consistent with the rhythmic framework of separately. Conversely, interactions are more the melody), and the pitch of the probe event likely to be observed when testing later stages of influenced their responses. However, in a homolo- processing (such as melodic completion or rhyth- gous task where the listeners instead indicated mic similarity), when the dimensions have been whether the probe event used a pitch class that integrated. was important in the melody, the timing of the The distinction between local and global task probe event did not affect their responses, thus characteristics also shows promise for reconciling demonstrating an asymmetric interaction the variety of findings for pitch–time integration. between pitch and time. Both of these tasks This account (Tillmann & Lebrun-Guillaud, required relatively late stages of processing and 2006) contends that tasks that favour processing attention to global characteristics of the stimulus,

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 3 PRINCE

yet responses indicated independence in some structure in Western music, loosely categorized tasks and interaction in others. as either surface-based structures or more abstract Prince, Thompson, et al. (2009) proposed organizational principles. Some examples of dimensional salience as a factor to reconcile the surface-based pitch structure include the pattern variety of findings in the literature. The concept of ups and downs in pitch height within a of dimensional salience refers to the tendency for melody known as pitch contour (Dowling, 1978), one perceptual dimension to dominate another, serial patterns (Boltz, Marshburn, Jones, & even when both dimensions are equally difficult Johnson, 1985), and grouping/stream segregation to process (i.e., equated discriminability). (Bregman, 1990). Serial patterns and grouping Instead, interactions may be due to stimulus and also exist in the temporal domain (Garner, 1974; task properties that magnify the importance of Jones, 1987). one dimension at the expense of another. Perhaps the most pervasive abstract organiz- Stimulus dimensions that feature greater informa- ational principle of pitch in Western music is ton- tive value are likely to dominate (cf. Melara & ality, or musical key. Tonality refers to the Algom, 2003; Prince, Schmuckler, & Thompson, hierarchical organization of the 12 pitch classes 2009). Also, task design may attract greater atten- (per octave) used in Western music around a tion to (and thus enhance the salience of) a dimen- central reference pitch, or tonic, that represents sion; for example, tapping tasks may highlight the key of a musical excerpt (Aldwell & temporal information at the expense of pitch Schacter, 2002; Krumhansl, 1990; Lerdahl, 2001; (Pfordresher, 2003; Snyder & Krumhansl, 2001). Schmuckler, 2004). The remaining pitch classes Typical Western music emphasizes pitch over vary in terms of their psychological stability (and time, thus enhancing pitch salience through the frequency of occurrence), in a hierarchical profile properties of the stimulus. Indeed, both music the- commonly referred to as the tonal hierarchy orists and psychological researchers attest to the (Krumhansl & Kessler, 1982). For example, in centrality of pitch in Western music. For the key of C major, C is the tonic and most example, according to Schenker (1935/1979 p. 15), stable pitch. The pitches E and G (used in con- “all rhythm in music comes from counterpoint and junction with C to form the tonic triad of C only from counterpoint”, and Everett (2000 p. 111) major) are the next most stable, followed by the states that “pitch relationships are of central 4 remaining pitch classes used in C major (D, F, importance, forming the core of the structure, A, B). The 5 pitches that are seldom used in that the identity, and even many of the expressive capa- key (C#,D#,F#,G#,A#) are the least stable. bilities of pop-rock music”. Empirical work Music that does not conform to the tonal hierarchy

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 demonstrates that pitch is the fundamental orga- (thus atonal) is perceived as disorganized and nizing principal for musical memory (Dowling, unpleasant by listeners regardless of training 1978; He´bert & Peretz, 1997), recognition (Dibben, 1994; Gagnon & Peretz, 2000) and (Dowling & Fujitani, 1971; White, 1960), simi- accordingly is heard extremely rarely (Everett, larity judgements (Cousineau, Demany, & 2000). Pressnitzer, 2009; Eerola, Ja¨rvinen, Louhivuori, There are abstract organizational structures in & Toiviainen, 2001), segmentation (Dawe, Platt, time as well, most notably metre. Metre refers to & Racine, 1993, 1994, 1995), perceived stability the hierarchical pattern of alternation between (Bigand, 1997), and the overall formation of a strong and weak moments in time (Hannon, mental representation of music (Krumhansl, Snyder, Eerola, & Krumhansl, 2004). Notes 2000). occur more often on strong beats, and such pos- But why is pitch so important in Western itions in time are more psychologically stable music? One of the reasons for the dominance of than weak beats, forming a metric hierarchy in a pitch in typical Western music may be its struc- similar manner to that of the tonal hierarchy tural complexity. There are many forms of (Palmer & Krumhansl, 1990). Rhythmic ﬁgures,

4 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

or repeating patterns of durations within a musical irrelevant when the sequence was tonal, presum- sequence, are a primary factor in establishing the ably because the dimension of pitch was suffi- accented (i.e., strong) moments in time that con- ciently salient as to overwhelm the influences of tribute to the construction of metre (Lerdahl & time. Conversely, when the sequences were Jackendoff, 1983). Metre in Western music is atonal (lacking tonal pitch structure), pitch was overwhelmingly in binary (strong–weak pattern) no longer more salient and enabled effects of or ternary (strong–weak–weak) metre, but may time to emerge. have nested levels of both to create a compound Dimensional salience has the potential to metre (e.g., strongest–weak–weak–strong– reconcile many of the seemingly contradictory weak–weak). These strong/weak cycles repeat in findings on pitch–time integration. But how units called measures, such that the earliest poss- might dimensional salience be manipulated? ible temporal position within the measure is the Given that the presence of tonality in a sequence strongest metrical position (the downbeat). affected pitch salience (Prince, Schmuckler, Rhythm and metre are indispensible components et al., 2009), varying the degree of conformity to of music; however, in terms of sheer complexity, a predefined structure (surface based or abstract) tonality has by far the most informative value. in a given stimulus dimension might be an effec- Whereas tonality involves the multilevel hierarch- tive method of manipulating dimensional salience. ical organization of 12 pitch classes, typical For example, changing the presence of metre may Western music uses only 2–3 unique durations influence the salience of time (e.g., Jones, (Fraisse, 1982) and a metrical framework that Moynihan, MacKenzie, & Puente, 2002). rarely uses more than two levels (Temperley, However, the existing work on dimensional sal- 2001). ience occurred only in the context of responses to Stimuli that resemble typical Western music single tones (Prince, Schmuckler, et al., 2009; (i.e., tonal) should invoke pitch salience, whereas Prince, Thompson, et al., 2009). Listeners do atonal sequences (not representing typical not hear only single notes in a musical experience, Western music) should be less likely to invoke but rather more complex musical units such as pitch salience, due to the lack of conformity to phrases and entire melodies. On this larger scale, tonal pitch structure. Prince, Schmuckler, et al. nuanced patterns and structure emerge in both (2009) tested this possibility by varying the the dimension of pitch (e.g., melody, contour, degree to which the stimuli were tonal and conse- and tonality) and that of time (e.g., rhythm and quently observed changes in dimensional salience. metre). As the existing work on dimensional sal- Participants compared the pitch height of a stan- ience has focused on the role of abstract structural

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 dard and comparison tone (which pitch was principles in music (tonality and metre), the exper- higher?), but had to ignore a sequence of tones iments in the present paper continue along that that intervened between the standard and com- line and aim to determine whether findings comparison. This sequence was either tonal or atonal. parable to those found in single-tone rating The temporal interval between the onset of each studies can result from responses to tone tone was identical except for the interval between sequences. However, future work may focus on the final tone of the sequence and the subsequent exploring this possibility in other forms of struc- comparison tone. That is, the comparison tone ture (both surface based and abstract) and in could be early, on time, or late by virtue of altering other domains such as language and vision. or preserving that final temporal interval. When Because the present research focuses on the the intervening sequence was tonal, this timing abstract organizational principles of tonality and variation had no effect on accuracy; however, metre, in the context of this paper the terms when the sequence was atonal, participants per- “pitch” and “time” refer to the sequences of pitch formed better when the comparison tone was on and temporal elements that form these larger time. In other words, the timing information was emergent patterns.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 5 PRINCE

The primary goal of these experiments is to test linearly to predict goodness ratings of the original whether systematically altering the degree of con- melody. formity to structure within the pitch and temporal Experiment 1 used a new approach in both the dimensions (specifically, tonality and metre) stimuli and task instructions. First, the stimulus affects their relative dimensional salience. sequences varied in their degree of conformity Experiments 1 and 2 consist of goodness ratings to tonality and metre (these manipulations are of sequences that vary in the degree of conformity detailed in the Method section). Second, partici- to tonality and metre; Experiment 3 has dichoto- pants’ instructions were to make judgements of mous conformity judgements (i.e., classification) “goodness” while attending selectively to pitch of whether a sequence adheres to the tonal and/ information (ignoring time), temporal infor- or metric structure (using the same sequences as mation (ignoring pitch), or both pitch and time those in Experiments 1 and 2). All experiments simultaneously. The two selective attention use selective attention conditions (respond based instructions resemble more recent work on judge- only on pitch, or time) and a more holistic con- mentsofsingleisolatednotesfollowingamusical dition (respond based on both pitch and time). context (Prince, Thompson, et al., 2009). In their In terms of the stages of processing and local/ study, judgements of goodness of fit of a probe global theories, these tasks can be characterized event preceded by a melody also showed indepen- as requiring a fairly late stage of processing and dent contributions of both the tonal and metric attention to global characteristics of the stimuli. hierarchy to ratings, even under instructions to Therefore, both of these theories predict inter- ignore the pitch dimension. However, the action between pitch and time regardless of the present study focuses on larger musical sequences, degree of conformity to tonality and metre. From and not single notes isolated from a preceding the perspective of dimensional salience, the context. In turn, this task requires attention to pattern of pitch–time integration should vary if more global aspects of the stimulus than individ- the manipulations of stimulus structure and task ual notes. Another difference concerns the selec- influence the relative salience of both dimensions. tive attention manipulation. In the present study, If one dimension is more salient than another, instructions to attend to pitch and/or time was a then asymmetric interactions are likely, whereas a within-subjects variable instead of a between- more even matching of salience may result in subjects variable. There is also the issue of more global interactions. musical training; both Prince, Thompson, et al. (2009) and Prince, Schmuckler, et al. (2009) used trained performers as participants. To pre-

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 serve continuity with these earlier ﬁndings, only EXPERIMENT 1 trained performers participated in Experiment 1. However, to address potential issues of external The rationale of Experiment 1 was to investigate validity in generalizing results from trained per- how ratings of pitch and temporal goodness formers to all listeners, untrained participants varied as a function of the degree to which a were tested in Experiment 2. sequence adhered to a coherent tonal and metric structure. Goodness ratings are not a new task. Method For example, Palmer and Krumhansl (1987a, 1987b) collected ratings of complete musical Participants phrases, and also for versions of these phrases There were 24 participants in Experiment 1, with that separated the pitch or temporal patterns, an average age of 19.4 years (SD ¼ 1.0); all of yielding either equitonal rhythmic sequences or them had at least 8 years of musical training in isochronous pitch sequences. The ratings of these private lessons (M ¼ 10.5, SD ¼ 2.4). separated pitch and temporal versions combined Participants were recruited using a ﬂyer posted

6 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

on campus and an online experiment database for an introductory psychology course. Recruitment took place at the University of Toronto Mississauga (n ¼ 17) and the University at Buffalo (n ¼ 7). Compensation was either course credit or $10.

Stimuli The stimuli used in Experiment 1 were based on 48 sightsinging melodies (from Berkowitz, Figure 1. Example variants of a single seed melody: (a) original Fontrier, & Kraft, 1997; and Ottman, 1986) that seed melody (Pitch Variant Level 1, Time Variant Level 1); (b) reordered duration (Pitch Variant Level 1, Time Variant Level epitomized the principles of typical Western 2); (c) reordered pitch and duration (Pitch Variant Level 2, Time music (including establishing both tonality and Variant Level 3); (d) random pitch, reordered duration (Pitch metre). These melodies functioned as starting Variant Level 4, Time Variant Level 2). points, or “seed melodies” for variation and factorial recombination of tonal and metric conformity. variants and save them as midi files. Further Their average length was 34.2 notes ( ¼ SD scripts converted the midi files to .wav files. 12.7), and average duration was 9.1 seconds (SD ¼ 1.8). For each melody, four levels of structural conformity in both pitch and time were con- Pitch variant levels structed. The first level was the sequence of Because Level 1 was the original pitch sequence, it elements (pitch classes or durations) found in the required no modifications; however, the extent to original melody. The second level consisted of a which it adhered to the tonal hierarchy was randomly reordered sequence of the original assessed for comparison with the other levels. elements, which nonetheless preserved the existing The Krumhansl and Schmuckler key-finding tonal or metric structural conformity. The third algorithm (Krumhansl & Schmuckler, 1986; level was a randomly reordered sequence of the described in Krumhansl, 1990) was used to find original elements, but which perturbed the tonal the best fit key of the original sequence. This or metric structural conformity. The fourth level algorithm also provides a maximum key-profile was a randomized sequence that included elements correlation (MKC)—the correlation coefficient not found in the original sequence (pitch classes or of the tonal hierarchy of the best fit key—which durations). These manipulations occurred separ- was stored for comparison with other variant

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 ately in pitch and time, such that there were 16 levels. The average MKC was .88 (SD ¼ .06), factorial recombinations (hereafter variants) for indicating that the melodies were strongly tonal. each seed melody. Figure 1 depicts 4 of the 16 var- The order of the pitches in the original iants for a single seed melody. Starting from 48 sequence was randomized to construct Level 2 seed melodies, creating all 16 variants yielded pitch variants. The specified criterion of the 768 stimuli. second pitch variant level was that the MKC To construct all 16 variants of each seed of the reordered sequence had to remain within melody, the original sequence was loaded into +.05 of the MKC of the Level 1 (original) MATLAB (Mathworks, 2004) as a midi file, sequence. This criterion ensured that the new and then the pitch and duration sequences were sequences were decidedly tonal, while allowing randomized separately. A script programmed in slight variation as a result of reordering the orig- MATLAB randomly permuted the order of the inal sequence. Average MKC was .87 (SD ¼ pitches and durations until they met the specified .07). Because this variant level consisted of reor- criteria, detailed forthwith. After randomization, dering the existing elements, the frequency of the script recombined the sequences to form the occurrence of each pitch did not change.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 7 PRINCE

For Pitch Variant Level 3, the original pitches Time variant levels were again reordered (as in Level 2), but had to The method for creating time variant levels was reduce the MKC by at least .2 (average MKC equivalent to that used for creating the pitch was .64, SD ¼ .07). The size of the numerical cri- variant levels. Level 1 was the original sequence terion was chosen on the basis that it would create of durations and thus involved no modifications. a statistically significant reduction in MKC. By correlating the number of onsets at each tem- Reordering the pitches reduced the MKC by alter- poral location with the metric hierarchy as ing the relative cumulative durations of each pitch measured by Palmer and Krumhansl (1990), it class, subsequently leading changes in the dur- was possible to calculate the extent to which the ation-weighted MKC. Given that MKC is calcu- melody instantiated a duple metre (specifically, lated by using cumulative durations of pitch class, 4/4). This metric hierarchy correlation (MHC) this manipulation of tonal strength was therefore for the original sequences was .76 (SD ¼ .17), ultimately based on a temporal change (albeit demonstrating that the sequences were distinctly based on cumulative sum, rather than sequential metric. timings). The implications of this limitation are Time Variant Level 2 consisted of a reordering addressed in the Discussion section. of the original durations. The order of the dur- The sequences in Pitch Variant Level 4 were ations was randomized, with the constraint that atonal, meaning that they did not adhere to tonal MHC (as calculated for Level 1) could not vary pitch structure. Construction of this variant level more than +.05 of the MHC value for the orig- had two main constraints. First, 7 pitch classes inal sequence. Pitch changes did not affect these were selected from the 12 found in Western correlations because MHCs are not weighted music (the same number as that found in the according to pitch class. The average MHC was other pitch variant levels), but comprised an artifi- .75 (SD ¼ .16). cial scale that did not correspond to any scale Another randomization of the order of dur- found in Western music (thus atonal). Starting ations was used for Time Variant Level 3, but in on C, the pitch classes were C, C#,D#,F,G,A, this case the MHC was reduced by .2 (average and B; however, this set was transposed to start MHC was .26, SD ¼ .32). The size of the on a variety of different pitches (e.g., Figure 1d required reduction was chosen to ensure a statisti- starts this scale on D). To address issues of eco- cally significant reduction in MHC. logical validity in the choice of pitch classes, this The sequences of Time Variant Level 4 had artificial scale preserved a near-universal feature entirely random timings, not limited to standard of scales in all musical cultures in that the pitch note duration denominations (e.g., eighth note,

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 intervals between adjacent scale members were of half note, etc.). The number of temporal events variable size (Dowling, 1978; Kessler, Hansen, & (durations) remained the same as that found in Shepard, 1984; Sloboda, 1985), in this case the original sequence. Furthermore, no durations either one or two semitones. Such unequal-step were shorter or longer than the minimum or scales are associated with processing advantages maximum (respectively) durations found in the in infancy and adulthood (Trehub, Schellenberg, original sequence. Lastly, the cumulative (total) & Kamenetsky, 1999). Accordingly, the fourth duration of each sequence could not vary more level of pitch variant constitutes only a removal than +10% from the cumulative duration of the of tonality and is not confounded with the original sequence. Given that this variant level number of pitch classes. The second constraint was by design not quantized to standard metric on the fourth-level pitch variants was that the positions, it was not possible to calculate the pitch range could be no larger than the corre- MHC, nor to depict the variants accurately in sponding original sequence. The average MKC musical notation (thus Figure 1 does not have an was .54 (SD ¼ .11). example of Time Variant Level 4).

8 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

Apparatus differences to attend preferentially to pitch or time A Macintosh G4 running OSX 10.3 was used to will not affect unequally the pattern of results across run the experiment at University of Toronto instructional conditions. A complication of this Mississauga. The experimental environment was design is the possibility of contamination between created in MATLAB (Mathworks, 2004), using blocks as a result of changing instruction. The ana- the Psychophysics Toolbox extensions (Brainard, lyses in the Results section address this issue. After 1997). Participants wore Sennheiser HD 280 Pro the melody had ﬁnished, participants indicated headphones during the experiment and were in a their rating of melodic goodness on a scale of 1– double-walled soundproof booth (IAC). The par- 7, and instructions on the screen reminded them ticipants from University at Buffalo used a on each trial that a rating of 1 meant very bad, Macintosh G5 running OSX 10.4 and were in a and 7 meant very good. Participants were encour- quiet room during the experiment. All other aged to use the full rating scale during the course aspects of the apparatus were identical across of the experiment. location. Participants completed eight practice trials, with the instructional condition changing every Procedure two trials. Each trial began when the participant All participants were tested individually. They pro- pressed the space bar to indicate their readiness. videdinformedconsentandcompletedaback- Prior to beginning a trial when the instructional ground questionnaire on musical experience. condition changed, the program displayed the fol- Participants received instructions to listen to each lowing message: “Now the experiment changes! melody and evaluate how “good” it sounded. The You will now rate the melodies based on a differ- experimenter stressed that participants were not to ent dimension. PLEASE pay attention to the indicate how much they personally liked the change (please please please!). Press the ‘y’ key if melody, but rather how well formed it was—for you understand, or get the experimenter if you example, if it sounded like a normal, typical need help.” The experimenter was present for the melody, conversely if it sounded like there was practice trials and would indicate verbally at this something wrong with the melody, and so on. point that the experiment was changing and that The experimenter also explained that there were now they would rate the melodies based on a three instructional conditions—rate pitch (ignore different dimension. Once the participants indi- time), rate time (ignore pitch), and rate both. The cated their understanding by pressing “y”, a stan- selective attention instructions were that partici- dard trial message appeared (“Please rate this pants were to ignore all aspects of the melody’s melody based . . .”) with instructions describing

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 timing and rhythm when the screen displayed the which dimension to attend. The participant then message: “Please rate this melody based ONLY pressed the space bar to start the trial. Upon com- on its PITCH characteristics.” This explanation pleting eight practice trials, participants had was repeated verbally with the opposite instructions experienced this instructional change three times. (ignore the pitches) when the screen showed: The participants were informed that the instruc- “Please rate this melody based ONLY on its tional condition would change only twice during TEMPORAL characteristics.” For the third the experimental trials: once after 64 trials and instructional condition, participants were told to again after the second 64 trials. The experimenter form their rating based on both dimensions simul- ensured that the participant understood what the taneously, when the screen read: “Please rate this instructions meant, answered any remaining ques- melody based on BOTH its PITCH and tions, then began the experimental trials and left TEMPORAL characteristics.” These instructions the room. There were 64 trials in each of the remained on the screen before, during, and after three instructional condition blocks, yielding 192 each trial. An advantage of varying instruction as total trials, and the entire procedure took approxi- a within-subject variable is that potential individual mately one hour.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 9 PRINCE

The order of blocks (instructional condition) dimension), giving three levels of both pitch and was counterbalanced across participants. Also, time dimensions (nine factorial combinations). the assignment of variants of a given melody to The effects of pitch, time, and block order on participants was determined using a counterba- goodness ratings were tested with three mixed lanced Latin square design, such that all partici- analyses of variance (ANOVAs; one ANOVA pants experienced all 16 variant conditions in all for each instructional condition), each using a 3 blocks, but only heard 4 variants of a given seed (pitch: Level 1, Level 2–3, Level 4) × 3 (time: melody. Which 4 of 16 variants of a given Level 1, Level 2–3, Level 4) × 3 (block order) melody each participant heard was counterba- design. For these analyses, pitch and time variant lanced across participants, to complete the Latin levels were within-subjects variables, and the square. The 48 total seed melodies were divided between-subjects variable was block order—that such that each participant heard 4 variants from is, the block number in which the participant 16 unique seed melodies in each of the three rated that dimension. For the “rate pitch” instruc- blocks (4 × 16 × 3 ¼ 192 trials). Therefore, tion analysis, the block order variable indicated each participant heard 4 variants of every seed the block in which each participant rated melody—a quarter of the 768 total stimuli—such pitch (1, 2, or 3); for the “rate time” instruction that 4 participants were required to complete the analysis, the block order variable denoted in design for all variants of all melodies. Coupled which block each participant rated time, and so with 6 possible instructional block orders (3 factor- on for the “rate both” instruction. These block ial), 24 participants were needed to counterbalance order variables were included because the within- block order and rate all variants of all melodies. subjects manipulation of instruction creates a potential complication in the results. Speciﬁcally, any interference of the ignored dimension observed in the selective attention conditions Results could be due to contamination between blocks as Goodness ratings were averaged across melody, a result of switching the attended dimension, giving 48 data points per participant (16 variants rather than a true failure of selective attention. in each of three instructional conditions). The By including block order as a between-subjects mean intersubject correlation of the ratings was variable, these ANOVAs tested whether the .64 (SD ¼ .07), indicating substantial agreement effects of pitch and time (and any interaction across participants. Initial inspection of the data between them) changed based on when the par- revealed that goodness ratings for Pitch Variant ticipants completed each instruction.

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Levels 2 and 3 overlapped, as well as Time Variant Levels 2 and 3. Bonferroni-corrected pair- Rate pitch instruction wise comparisons verified this pattern, showing no The “rate pitch” instruction 3 (pitch) × 3 (time) significant difference between the second and third × 3 (block order) ANOVA (as described above) pitch variant levels for the instruction conditions revealed a main effect of pitch, F(2, 42) ¼ of “rate pitch”, M ¼ .23, 95% CI [–.01, .47], 208.22, MSE ¼ 1.00, p , .001, h2 ¼ .732, and a p ¼ .07, “rate time”, M ¼ .20, 95% CI [–.02, main effect of time, F(2, 42) ¼ 35.10, MSE ¼ .43], p ¼ .10, and “rate both”, M ¼ .14, 95% CI .63, p , .001, h2 ¼ .078 on goodness ratings. [–.16, .43], p ¼ 1. The second and third time Block order (the block number in which the par- variant levels also did not differ for any instruc- ticipants rated pitch) was not significant as a tional condition, M ¼ .12, 95% CI [–.06, .29], main effect, F(2, 21) , 1, MSE ¼ 2.73, p ¼ .48, p ¼ .45, M ¼ .24, 95% CI [–.03, .51], p ¼ .10, h2 ¼ .001, but there was an interaction between M ¼ .09, 95% CI [–.14, .31], p ¼ 1, respectively. pitch and block order, F(4, 42) ¼ 3.43, MSE ¼ Therefore, in all subsequent analyses, these two 1.00, p ¼ .02, h2 ¼ .024. This interaction was conditions were averaged (within a single due to the fact that the effect size of pitch was

10 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

slightly larger for the participants who rated pitch in their second block, F(2, 14) ¼ 179.60, MSE ¼ .59, p , .001, h2 ¼ .889, than those participants who rated it in their ﬁrst block, F(2, 14) ¼ 37.42, MSE ¼ 1.66, p , .001, h2 ¼ .675, or last block, F(2, 14) ¼ 61.87, MSE ¼ .73, p , .001, h2 ¼ .637. Block order did not interact with time, F(4, 42) ¼ 1.22, MSE ¼ .63, p ¼ .32, h2 ¼ .005. There was no interaction between pitch and time, F(4, 84) ¼ 2.37, MSE ¼ .23, p ¼ .06, h2 ¼ .004, nor was there between pitch, time, and block order, F(8, 84) , 1, MSE ¼ .23, p ¼ .59, h2 ¼ .003. Figure 2a displays the goodness ratings as a function of pitch and time variant level.

Rate time instruction For the “rate time” instruction, the 3 (pitch) × 3 (time) × 3 (block order) mixed ANOVA yielded main effects of pitch, F(2, 42) ¼ 54.44, MSE ¼ .53, p , .001, h2 ¼ .128, and time, F(2, 42) ¼ 76.32, MSE ¼ 1.70, p , .001, h2 ¼ .573, but not block order (the block number in which the participants rated time), F(2, 21) , 1, MSE ¼ 3.21, p ¼ .65, h2 ¼ .001. None of the interactions was signiﬁcant: pitch and block order, F(4, 42) ¼ 1.07, MSE ¼ .53, p ¼ .38, h2 ¼ .005; time and block order, F(4, 42) ¼ 1.83, MSE ¼ 1.70, p ¼ .14, h2 ¼ .027; pitch and time, F(4, 84) ¼ 1.06, MSE ¼ .29, p ¼ .38, h2 ¼ .003; pitch, time, and block order, F(8, 84) , 1, MSE ¼ .29, p ¼ .66, 2

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 h ¼ .004. Figure 2b exhibits these data. Accordingly, for both selective attention conditions, both the dimensions of pitch and time contributed to melodic goodness ratings, despite instructions to ignore either dimension (i.e., the lines of Figure 2a are not ﬂat, and the lines of Figure 2b are not superimposed). Also, pitch and time combined in a linear (additive) fashion, with no interaction between them. The change in relative effect sizes of pitch and time across instructional conditions signiﬁes that these listeners were able to emphasize either dimension in Figure 2. Goodness ratings for instructional conditions in Experiment 1 (musically trained performers): (a) “rate pitch” accordance with the instructions, but not eliminate condition; (b) “rate time” condition; (c) “rate both” condition. completely the contribution of the other Error bars represent standard error of the mean. dimension.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 11 PRINCE

Rate both instruction or down by small intervals). This reordering there- The results of the 3 (pitch) × 3 (time) × 3 (block fore made the pitch sequences atypical and conse- order) mixed ANOVA were more complex for the quently not as well formed, or as good, as the “rate both” instruction, as shown in Figure 2c. The originals. However, the fourth variant level also only significant main effects were, again, pitch, disturbed the contour of the original melody but F(2, 42) ¼ 249.81, MSE ¼ .45, p , .001, h2 ¼ added a violation of the tonal hierarchy by using .583, and time, F(2, 42) ¼ 77.35, MSE ¼ .60, p pitches that did not belong in any one key; this , .001, h2 ¼ .237, whereas block order (the variant level received ratings significantly lower block number in which the participants rated than the other levels. Accordingly, the operation both dimensions) was not significant, F(2, 21) , of randomizing the order of elements cannot be 1, MSE ¼ 3.55, p ¼ .63, h2 ¼ .001. The effect solely responsible for the effect of pitch on good- sizes denote that pitch contributed more strongly ness ratings. to ratings than time. More importantly, however, In fact, the ineffectiveness of the manipulation there was an interaction between pitch and time, between the second and third pitch variant levels F(4, 84) ¼ 7.33, MSE ¼ .19, p , .001, h2 ¼ suggests a different conclusion. Specifically, a .014, showing that the contribution of pitch and musical context can establish a tonality in an all- time in goodness ratings was not purely linear, or-none fashion simply by presenting the collec- but instead varied based on the particular combi- tion of pitch classes that belong in that key, nation of pitch and time variant levels. Block rather than having to arrange the cumulative rela- order did not interact with any variables: pitch tive durations of these pitch classes in close accord- and block order, F(4, 42) , 1, MSE ¼ .45, p ¼ ance with the tonal hierarchy. In turn, listeners’ .68, h2 ¼ .003; time and block order, F(4, 42) ¼ experience of tonality cannot be only a reflection 1.11, MSE ¼ .60, p ¼ .36, h2 ¼ .007; pitch, of its immediate pitch distribution, but instead is time, and block order F(8, 84) , 1, MSE ¼ .19, largely formed by invoking a stored internal rep- p ¼ .90, h2 ¼ .002. resentation of the tonal hierarchy activated by a sufficiently similar context (i.e., collection of pitch classes) in the stimulus they hear. Discussion Additionally, this lack of effect renders moot the The first and most obvious result in these data is potential limitation of differences between Pitch that, reassuringly, ratings of goodness varied in Variant Levels 2 and 3 being due to a manipulation accordance with the conformity to pitch and tem- of cumulative duration (as mentioned in the poral structure (tonality and metre). Interestingly, Method section).

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 goodness ratings did not change between the It is more difficult to explain why there was no second and third variant levels. Both of these difference between the second and third variant levels used the same pitches or durations found levels that altered the metric conformity, com- in the original, but reordered them to preserve or pared to the original sequence. Reordering the perturb (respectively) the tonal or metric hierar- order of durations inevitably makes a musical chy. One possible explanation for this finding is sequence sound more syncopated. Perhaps an all- that the goodness ratings were not linked to the or-none effect (similar to that found described in numerical degree of conformity to musical struc- the pitch variants) of syncopation exists with the ture (i.e., the tonal and metric hierarchy corre- perception of metre and temporal goodness, such lations), but instead were affected only by that the degree of syncopation is not the operative upsetting the original sequential order of elements factor, but its presence or absence. Another possi- found in the first variant level. Certainly in the case bility is that the manipulation of correlation with of pitch, reordering the pitches destroys the the metric hierarchy may not have been sufficiently contour of an original melody, including its pri- obvious so as to cause listeners to differentiate marily stepwise nature (most melodies move up them in terms of a goodness rating.

12 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

For the “rate pitch” condition, there was an which these dimensions combined and the ability interaction between pitch variant level and the to weight one dimension as instructed suggests block in which participants rated pitch. The that on the continuum of independent to interac- effect size of pitch was larger for those listeners tive pitch–time relations, listeners processed pitch who rated pitch selectively in their second block and time more independently than interactively in than for those who completed the “rate pitch” this task. instruction at the beginning or end of the exper- However, a slightly different picture emerges iment (yet the effect was strong in all blocks). It from the “rate both” condition. When instructed is unclear why this interaction arose. There were to incorporate both dimensions into their good- no other interactions involving block order ness rating (i.e., no selective attention instruc- (neither with pitch nor time) in any instructional tions), pitch and time showed a more interactive condition. Furthermore, the critical three-way relation. Thus a simple summation of the contri- interaction between pitch, time, and block order butions of these dimensions was not sufficient to was not significant in the “rate pitch” instruction predict goodness ratings when listeners attended (or any other instruction), showing that block to the sequence as a whole instead of selectively order did not affect the pattern of pitch–time attending one dimension. What was the nature integration. of this interaction? Judging from Figure 2c, the Another interesting finding in these data is that effect of one dimension increased when the both pitch and time contributed to goodness other dimension adhered more to its original ratings regardless of instruction. Listeners were structure—the slope of the lines (indicating the unable to ignore pitch entirely when rating tem- effect of time) was greater as the degree of tonal poral goodness (and time when rating pitch good- conformity more closely approximated that of the ness), instead demonstrating a failure of selective original variant level. Another way of summarizing attention (Figures 2a and 2b). Nevertheless, they the same result is to note that the separation did emphasize the attended dimension in their between the lines (indicating the effect of pitch) ratings, such that the effect size of pitch and was greatest for the original time variant. time varied according to the instructions. The Indeed, overall these findings support well the effect of pitch (h2 ¼ .732) was substantially notion that pitch was more salient than time in larger than that of time (h2 ¼ .078) in the “rate this experiment. Pitch accounted for over twice pitch” instruction, and the effect of time (h2 ¼ the variance that time did when participants .573) easily exceeded that of pitch (h2 ¼ .128) attended both dimensions (h2 ¼ .583 vs. h2 ¼ when participants rated time. The ability to .237, respectively). Furthermore, the advantage

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 emphasize selectively either dimension in accord- of pitch over time in the “rate pitch” condition ance with the instructions suggests that partici- (pitch h2 – time h2 ¼ .654) was considerably pants were largely (although not entirely) able to larger than the corresponding advantage of time process these dimensions independently. If for the “rate time” condition (time h2 – pitch h2 instead the effect sizes of pitch and time remained ¼ .445). Nevertheless, both dimensions contribu- constant for both selective attention conditions ted signiﬁcantly to predict ratings in all three (i.e., listeners could not accentuate either dimen- instructional conditions. sion as per the instructions), then such failures of Musical training imparts the ability to examine selective attention would suggest instead that lis- music analytically, which could potentially help teners processed the dimensions in a more interac- listeners to separate the dimensions of pitch and tive manner. Furthermore, in these selective time. However, the failures of selective attention attention conditions, the effects of pitch and suggest that these musically trained participants time combined linearly to predict goodness were not entirely successful in this attempt. ratings, with no statistical interaction between Nevertheless, musical expertise may have inﬂu- them. Taken together, the additive manner in enced the results in other unpredictable ways.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 13 PRINCE

Thus to ensure the validity of generalizing these of the apparatus were the same as those in findings to other populations (i.e., not only those Experiment 1. with formal musical training), untrained listeners participated in Experiment 2. Procedure The procedure for Experiment 2 was identical to that for Experiment 1. EXPERIMENT 2 Results Neither rating the goodness of a melody, nor understanding the instruction to attend only to Data were analysed the same way as in Experiment the pitch or timing of a musical sequence, requires 1. Goodness ratings were averaged across melody, explicit musical training. Therefore Experiment 2 giving 48 data points per participant; the mean had the exact same stimuli and design as those in intersubject correlation of the ratings was .58 Experiment 1, but the participants were untrained (SD ¼ .09). As in Experiment 1, initial examin- listeners instead of trained performers. If passive ation of the data revealed that goodness ratings exposure to typical Western music (as opposed to for Pitch Variant Levels 2 and 3 overlapped, as formal training) is sufficient to be sensitive to its well as Time Variant Levels 2 and 3. Bonferroni- statistical properties (e.g., pitch salience), then corrected pairwise comparisons showed a signifi- the same pattern of results should occur regardless cant difference between the second and third of the degree of formal musical training. However, pitch variant levels only for the “rate pitch” the familiarity with music afforded by formal instruction, M ¼ .24, 95% CI [.06 .42], p ¼ .01, training does tend to improve the overall reliability but not for “rate time”, M ¼ –.03, 95% CI and accuracy of responses to musical stimuli, [–.34, .29], p ¼ 1, nor for “rate both”, M ¼ .17, especially in tasks that require explicit knowledge 95% CI [–.08, .42], p ¼ .37. Time Variant of musical structure (cf. Bigand, 2003; Bigand & Levels 2 and 3 did not differ for any instructional Poulin-Charronnat, 2006), so some quantitative condition, for “rate pitch”, M ¼ .14, 95% CI (not qualitative) differences may be expected. [–.20, .47], p ¼ 1, “rate time”, M ¼ .09, 95% CI [–.16, .34], p ¼ 1, and “rate both”, M ¼ .08, 95% CI [–.18, .35], ¼ 1. Because only one of Method p these differences reached significance (and only Participants in one instructional condition), and to maintain There were 24 musically untrained participants in continuity with Experiment 1, these two variant

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Experiment 2; none had more than 5 years of levels were averaged (within a single dimension), private music lessons (M ¼ 1.19, SD ¼ 1.70). giving three levels of both pitch and time dimen- The average age of the participants was 23.13 sions for all subsequent analyses. years (SD ¼ 6.00). These participants were Next, three mixed 3 (pitch: Level 1, Level 2–3, recruited at Murdoch University using ﬂyers and Level 4) × 3 (time: Level 1, Level 2–3, Level 4) an online experiment database, and they were × 3 (block order) ANOVAs tested the effects of compensated with either course credit or $10. pitch, time, and block order on goodness ratings, separately for each instructional condition. Pitch Stimuli and time variant levels were within-subjects vari- All stimuli were the same as those in Experiment ables for these analyses, and block order was the 1. between-subjects variable.

Apparatus Rate pitch instruction A PC running Windows XP was used to run the For the “rate pitch” instruction there was a main experiment in a quiet room, but all other aspects effect of pitch, F(2, 42) ¼ 154.22, MSE ¼ .71,

14 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

p , .001, h2 ¼ .586, and a main effect of time, F(2, 42) ¼ 59.49, MSE ¼ .55, p , .001, h2 ¼ .176 on goodness ratings. There was no main effect of block order (the block number in which the participants rated pitch), F(2, 21) ¼ 1.17, MSE ¼ 1.89, p ¼ .20, h2 ¼ .001, but there was an interaction between time and block order, F(4, 42) ¼ 3.23, MSE ¼ .55, p ¼ .02, h2 ¼ .019. This interaction was caused by a larger effect size of time for the participants who rated pitch in their second block, F(2, 14) ¼ 85.87, MSE ¼ .26, p , .001, h2 ¼ .349, than for those participants who rated it in their ﬁrst block, F(2, 14) ¼ 17.44, MSE ¼ .45, p , .001, h2 ¼ .111, or last block, F(2, 14) ¼ 6.94, MSE ¼ .95, p ¼ .01, h2 ¼ .125. Block order did not interact with pitch, F(4, 42) ¼ 1.42, MSE ¼ .71, p ¼ .24, h2 ¼ .011. There was also no interaction between pitch and time, F(4, 84) , 1, MSE ¼ .27, p ¼ .70, h2 ¼ .002, nor was there between pitch, time, and block order, F(8, 84) , 1, MSE ¼ .27, p ¼ .56, h2 ¼ .005. Figure 3a displays the goodness ratings as a function of pitch and time variant level.

Rate time instruction The same main effects emerged in the “rate time” instruction: pitch, F(2, 42) ¼ 70.92, MSE ¼ .62, p , .001, h2 ¼ .217, and time, F(2, 42) ¼ 81.62, MSE ¼ 1.18, p , .001, h2 ¼ .474, but not block order, F(2, 21) , 1, MSE ¼ 3.27, p ¼ .92, h2 ¼ .000. None of the interactions was sig-

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 niﬁcant: pitch and block order, F(4, 42) , 1, MSE ¼ .62, p ¼ .50, h2 ¼ .005; time and block order, F(4, 42) ¼ 2.02, MSE ¼ 1.18, p ¼ .11, h2 ¼ .023; pitch and time, F(4, 84) , 1, MSE ¼ .41, p ¼ .41, h2 ¼ .004; pitch, time, and block order, F(8, 84) , 1, MSE ¼ .41, p ¼ .79, h2 ¼ .005. Figure 3b exhibits these data. As in Experiment 1, both the dimensions of pitch and time contributed to melodic goodness ratings for both selective attention conditions, despite instructions to ignore either dimension. Also, pitch and time combined in a linear (addi- Figure 3. Goodness ratings for instructional conditions in Experiment 2 (musically untrained listeners): (a) “rate pitch” tive) fashion, with no interaction between them. condition; (b) “rate time” condition; (c) “rate both” condition. The change in relative effect sizes of pitch and Error bars represent standard error of the mean. time across instructional conditions signiﬁes that

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 15 PRINCE

these listeners were able to emphasize either similarly sized advantage in the “rate time” con- dimension in accordance with the instructions, dition (time h2 – pitch h2 ¼ .256). In the “rate but not completely eliminate the contribution of both” condition the two dimensions were more the other dimension. equally matched (pitch h2 – time h2 ¼ .084), but again tipped in favour of pitch. Rate both instruction In fact, the main divergence between The “rate both” instruction results also mirrored Experiments 1 and 2 is that all effects (main and Experiment 1, as shown in Figure 3c. Pitch and interaction) were quantitatively weaker for time were the only significant main effects, F(2, untrained listeners. This finding may be due to 42) ¼ 117.39, MSE ¼ .74, p , .001, h2 ¼ .420, motivation—musically trained individuals are F(2, 42) ¼ 105.61, MSE ¼ .66, p , .001, h2 ¼ likely to engage more in musical tasks because .335, respectively, but not block order, F(2, 21) they have invested deeply in this activity. ,1, MSE ¼ 1.62, p ¼ .80, h2 ¼ .000. Again, Therefore, they tend to give cleaner and more the effect sizes denote that pitch contributed robust effects than untrained participants who more strongly to ratings than time. More impor- find it more difficult to remain engaged in a tantly, however, there was an interaction between musical task. However, the smaller relative main pitch and time, F(4, 84) ¼ 3.66, MSE ¼ .32, p effect sizes (pitch h2 – time h2) do not suggest ¼ .01, h2 ¼ .011, showing that the contribution that untrained listeners showed more evidence of of pitch and time in goodness ratings was not interactive pitch–time processing than trained purely linear, but instead varied based on the par- performers, as the interaction effect size (pitch × ticular combination of pitch and time variant time h2) was also weaker and only present in the levels. Block order did not interact with pitch, “rate both” condition for both experiments. F(4, 42) , 1, MSE ¼ .74, p ¼ .78, h2 ¼ .003, Indeed, the two populations showed the same with time, F(4, 42) ¼ 2.10, MSE ¼ .66, p ¼ .10, qualitative pattern of effects, suggesting that the h2 ¼ .013, or with the pitch–time interaction, underlying implicit cognitive mechanisms respon- F(8, 84) ¼ 1.59, MSE ¼ .32, p ¼ .14, h2 ¼ .010. sible for pitch–time integration are unchanged across levels of expertise. This common pattern demonstrates selective attention failure, pitch Discussion salience, and more independent contributions The goodness ratings of melodies varying in their in selective attention tasks yet increasingly degree of conformity to the tonal and metric hier- interactive contributions when attending to both archies provided by musically untrained listeners dimensions.

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 (Experiment 2) showed nearly exactly the same There were other differences between the two patterns as those of trained performers groups, however. For musically trained perfor- (Experiment 1). Namely, neither group success- mers, pitch interacted with block order in the fully eliminated the influence of pitch or time in “rate pitch” instructional condition, whereas selective attention conditions; however, in both untrained listeners displayed an interaction of cases the two dimensions made independent con- time with block order in the “rate pitch” instruc- tributions to goodness ratings. Additionally, tional condition. These findings were due to vari- pitch and time demonstrated interactive relations ations in the main effect size of one dimension in the “rate both” condition, for both participant (curiously always largest in the second block) and groups. Finally, as in Experiment 1, pitch appeared did not affect any interactions between pitch and to be more salient than time for Experiment 2, as time. No obvious theoretical interpretation of measured by relative main effect sizes. In the “rate these effects is forthcoming. Also, the difference pitch” condition, pitch enjoyed a substantial between Pitch Variant Levels 2 and 3 reached sig- advantage in effect size over time (pitch h2 – nificance only in the untrained listeners. The fact time h2 ¼ .410), but time did not experience a that this effect was small, was only present in

16 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

one instructional condition, and did not occur in of conformity—of pitch or temporal structure in the trained performers (who, if anyone, would pre- a melody, the stimulus manipulations should, sumably be more sensitive to such manipulations, theoretically, make a dimension truly irrelevant, not less) renders this finding uninterpretable. rather than requiring active cognitive effort to sup- Overall, the results of Experiments 1 and 2 are press its contribution to the participant’s response. remarkably similar, meaning that these findings Classification paradigms have a long history in generalize to all listeners encultured in Western research on dimensional integration (Garner, music. Nevertheless, the pattern of findings is 1974; Pomerantz, 1983) and are still relevant somewhat puzzling: Failures of selective attention today (for recent studies in audition see Dyson & suggest interaction between pitch and time, but Quinlan, 2010; Silbert, Townsend, & Lentz, their contributions to ratings were additive in 2009). Therefore Experiment 3 had the same these conditions, while interactive when rating melodies and the same instructions, but changed both dimensions simultaneously. the task. Specifically, for the “classify pitch” con- One explanation to this pattern of findings has dition, participants indicated whether the melody to do with task design, specifically the nature of the was tonal or atonal (did it establish a key area?). dependent measure. Participants rated the same The “classify time” condition consisted of classify- abstract property of “goodness” regardless of ing whether the melody was metric or random (did which dimension they were attending. it establish a regular beat?). The “classify both” Accordingly, the reliable influence of pitch and condition required participants to evaluate time in both selective attention conditions may whether the melody was both tonal and metric, not indicate how participants processed pitch and or if either dimension was “off” (i.e., atonal, time as separate dimensions, but instead be a random, or both). Measuring the extent to which result of evaluating both dimensions along a scale the irrelevant dimension contributed to partici- of the same basic psychological quality. A task in pants’ judgements in the selective attention con- which information along one dimension is purely ditions comprised the measure of interference irrelevant to the other would provide a context between pitch and time. Given that these instruc- that promotes the ability to process pitch and tions require explicit knowledge of musical struc- time more independently and therefore an oppor- ture, all participants had formal musical training. tunity to test how involuntary this interference is. Fortunately, Experiments 1 and 2 had established To investigate this possibility, and to explore the similarity of pitch–time integration across further the integration of pitch and time, expertise levels, attenuating concerns of generaliz- Experiment 3 complemented the task of goodness ing findings to other populations.

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 ratings with an explicit classification—judging conformity to tonal and/or metric structure. Method Participants EXPERIMENT 3 The participants in Experiment 3 were 24 musically trained performers; all of them had at least The ratings of melodic goodness in Experiments 1 8 years of musical training (M ¼ 9.7, SD ¼ 2.4), and 2 provide a number of interesting findings and their average age was 20.4 years (SD ¼ 2.9). regarding how pitch and time combine in a Participants were recruited using a flyer posted musical context. The goal of Experiment 3 was on campus and an online experiment database for to clarify these findings in a different task an introductory psychology course. Recruitment context and to provide convergent evidence for occurred only at the University at Buffalo. the pattern of pitch–time integration in the per- Compensation was either course credit or $10. ception of melodies. By using a task that required None of the participants from Experiment 1 par- explicit classifications—dichotomous judgements ticipated in Experiment 3.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 17 PRINCE

Stimuli that there was no significant difference between All stimuli were the same as those used in previous the second and third variant levels, for “classify experiments. pitch”, M ¼ .04, 95% CI [–.04, .11], p ¼ 1, “classify time”, M ¼ –.02, 95% CI [–.09, .05], p ¼ 1, Apparatus and “classify both”, M ¼ .04, 95% CI [–.07, .16], The apparatus were the same as that used at the p ¼ 1. Time Variant Levels 2 and 3 were different University at Buffalo in Experiment 1. for the “classify both” condition, M ¼ .11, 95% CI [.04, .18], p ¼ .001; however, they were not for Procedure “classify pitch”, M ¼ –.00, 95% CI [–.06, .06], The procedure was the same as that in the previous p ¼ 1, nor “classify time”, M ¼ .01, 95% CI experiments except for the instructions provided [–.09, .12], p ¼ 1. Accordingly, and as in the pre- for participants. Prior to beginning the practice vious experiments, these two levels were averaged trials, participants received thorough instructions. together (for pitch and time separately) in sub- For the “classify pitch” condition, the on-screen sequent analyses, giving three levels of both pitch instructions read: “Please classify if this melody is and time dimensions. tonal (press T) or atonal (press A).” The “classify Further analyses of Experiment 3 data used the time” instructions were: “Please classify if this same three mixed ANOVAs with a 3 (pitch: Level melody is metric (press M) or random (press R).” 1, Level 2–3, Level 4) × 3 (time: Level 1, Level The term “random” was used instead of 2–3, Level 4) × 3 (block order) design to test “ametric” such that the response key (R) would the effects of pitch, time, and block order on classi- not be the same as the response key for “atonal” fications, for each instruction separately. (A), used in the “classify pitch” block. Nevertheless, the participants received clear Classify pitch instruction instructions to indicate whether the melody was The “classify pitch” instruction ANOVA repli- metric or not (as in previous experiments). The cated the findings of the previous experiments “classify both” condition instructed participants: (see Figure 4a)—there was a main effect of pitch, “Please classify if this melody is BOTH tonal F(2, 42) ¼ 150.64, MSE ¼ .06, p , .001, h2 ¼ AND metric (press B), or if it is EITHER .706, and a main effect of time, F(2, 42) ¼ atonal OR nonmetric (press E).” Particular care 21.15, MSE ¼ .03, p , .001, h2 ¼ .044, but no was given to explaining the “classify both” con- main effect of block order (the block in which par- dition, as it had the greatest potential for con- ticipants classified pitch), F(2, 21) ¼ 1.37, MSE fusion. The same message as that in Experiment ¼ 2.21, p ¼ .28, h2 ¼ .005. None of the inter-

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 1 appeared when the instructions changed (“Now actions was signiﬁcant: pitch and block order, the experiment changes! You will . . .”). F(4, 42) , 1, MSE ¼ .06, p ¼ .45, h2 ¼ .009; time and block order, F(4, 42) ¼ 1.17, MSE ¼ .03, p ¼ .34, h2 ¼ .005; pitch and time, F(4, 84) Results ¼ 1.97, MSE ¼ .03, p ¼ .11, h2 ¼ .008; pitch, Responses of “tonal”, “metric”, and “both tonal and time, and block order, F(8, 84) , 1, MSE ¼ .03, metric” were coded as “yes” responses, whereas p ¼ .67, h2 ¼ .006. “atonal”, “random”, or “either atonal or random” were coded as “no” responses. Data were then ana- Classify time instruction lysed as the proportion of “yes” responses, as a For the “classify time” instruction, the ANOVA function of conformity to pitch and temporal revealed an effect of pitch, F(2, 42) ¼ 41.97, structure, and were averaged across melody. MSE ¼ .08, p , .001, h2 ¼ .240, and time, F(2, Intersubject correlations were again high (mean 42) ¼ 96.83, MSE ¼ .06, p , .001, h2 ¼ .397, r ¼ .67, SD ¼ .07). Bonferroni-corrected pair- but not block order, F(2, 21) ¼ 1.93, MSE ¼ wise comparisons of pitch variant levels showed 2.28, p ¼ .17, h2 ¼ .008. In contrast to

18 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

Experiments 1 and 2, the interaction between pitch and time was signiﬁcant, F(4, 84) ¼ 20.93, MSE ¼ .02, p , .001, h2 ¼ .055 (shown in Figure 4b). Block order did not interact with pitch, F(4, 42) ¼ 1.97, MSE ¼ .08, p ¼ .12, h2 ¼ .022, with time, F(4, 42) ¼ 2.41, MSE ¼ .06, p ¼ .06, h2 ¼ .020, or with the pitch–time interaction, F(4, 42) ¼ 1.11, MSE ¼ .02, p ¼ .37, h2 ¼ .006. Like Experiments 1 and 2, and despite the task design and selective attention instructions, both pitch and time always affected classiﬁcations. Also, participants were able to emphasize either dimension in accordance with the instructions, as evidenced by the change in effect sizes of pitch and time across instruction. Unlike Experiments 1 and 2, where the effects of pitch and time combined linearly in the “rate time” instruction, there was an interaction between pitch and time for the “classify time” instruction (compare Figures 2b and 3b to Figure 4b).

Classify both instruction The results of the “classify both” instruction revealed a main effect of pitch, F(2, 42) ¼ 122.19, MSE ¼ .04, p , .001, h2 ¼ .374, and time, F(2, 42) ¼ 58.51, MSE ¼ .07, p , .001, h2 ¼ .297, but not block order, F(2, 21) , 1, MSE ¼ 2.24, p ¼ .58, h2 ¼ .003. Figure 4c displays these data. There was an interaction between pitch and time, F(4, 84) ¼ 15.87, MSE ¼ .03, p , .001, h2 ¼ .061. The pitch–time interaction in the “classify

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 time” and “classify both” instructions is similar to that observed in the “rate both” instruction of Experiments 1 and 2. That is, the effect of variation in one dimension was stronger as the other dimension adhered more closely to its original structure. Block order did not interact with pitch, F(4, 42) , 1, MSE ¼ .04, p ¼ .57, h2 ¼ .004, with time, F(4, 42) , 1, MSE ¼ .07, p ¼ .10, h2 ¼ .004, or with the pitch–time interaction, F(8, 84) ¼ 1.16, MSE ¼ .03, p ¼ .33, h2 ¼ .009.

Figure 4. Probability of classiﬁcations for melodies in each Discussion instructional condition of Experiment 3: (a) “classify pitch” condition; (b) “classify time” condition; (c) “classify both” As in Experiments 1 and 2, participants’ responses condition. Error bars represent standard error of the mean. followed the manipulations in pitch and temporal

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 19 PRINCE

structural conformity within the stimuli. In fact, structural conformity, the effect of the other the overall pattern of data in Experiment 3 dimension increased. Accordingly, in these inter- closely resembles that of the previous experiments. active cases, greater structural conformity in one The data from all experiments showed essentially dimension enhanced the ability to detect structural no difference between the second and third conformity in the other dimension, even when it variant levels for pitch and time, with one excep- was irrelevant to the task. tion for each (out of a possible 18). All experiments Overall, the variations in the pattern of findings showed failures of selective attention when partici- across instruction and experiment suggest that pants’ instructions were to ignore either pitch or differences in salience affected how the dimensions time. This result is particularly notable in combine. Specifically, a more even matching of Experiment 3 because the classification task salience across dimensions fostered interactions design made the unattended dimension truly irre- between them; conversely larger mismatches in levant in the selective attention conditions, as dimensional salience prevented such interactions. opposed to effortfully ignored (Experiments 1 Table 1 illustrates this principle by listing the and 2). Therefore, it appears that the contributions absolute difference score in effect size (i.e., of pitch and time are involuntary in these two |pitch h2 – time h2|) along with the strength of tasks. Another overlap in findings across exper- the interaction between the dimensions (pitch iment is that in the selective attention instructional × time h2) for all instructional conditions. For conditions, the effect size of the dimensions varied musically trained performers, when classifying according to the instructions. Additionally, for metric conformity, the advantage in effect size “rate both” and “classify both” instructions, there for time, |pitch h2 – time h2| ¼ .157 was was always an interaction between pitch and smaller than the corresponding advantage when time, and pitch was overall clearly the more influ- rating temporal goodness, |pitch h2 – time h2| ential dimension in all experiments. ¼ .445. Proportionately, the pitch–time inter- Nevertheless, the differences between the action was more powerful when classifying time experiments warrant further inspection. In the than when rating temporal goodness (interaction “rate pitch” instruction of Experiment 1, there was an interaction between pitch and block order, demonstrating that variations in tonal con- Table 1. Differences in main effect sizes for pitch and time formity influenced ratings more for those partici- for each instruction and the corresponding effect size of the pitch–time interaction pants who rated pitch in their second block. In Experiment 2, block order interacted instead Main effect size Interaction effect

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 with time, again for the “rate pitch” condition. difference score size | 2 2| × 2 Neither of these interactions was signiﬁcant in Instruction pitch h – time h Pitch Time h Experiment 3 (in fact, no interactions with block Musically trained performers (Exp. 1) order emerged in Experiment 3), and neither Rate pitch .654 .004 lend themselves well to any particular theoretical Rate time .445 .003 ∗∗ interpretation. Rate both .345 .014 The more interesting divergence in Experiment Musically trained performers (Exp. 3) Classify pitch .663 .008 3 is the presence of an interaction between pitch ∗∗ and time in classiﬁcations of metric conformity, Classify time .157 .055 Classify both .078 .061∗∗ where there was no such interaction in the temporal goodness ratings of the previous exper- Musically untrained listeners (Exp. 2) iments. This interaction replicated the shape of Rate pitch .410 .002 Rate time .256 .004 the other pitch–time interactions (for “rate both” Rate both .084 .011∗ and “classify both” instructions), showing that as ∗ ∗∗ either dimension resembled more its original p , .01. p , .001.

20 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

h2 ¼ .055 vs. h2 ¼ .003, respectively). main effects of both dimensions in all conditions, Additionally, this observation may account for even for selective attention instructions; further- the increased effect size of the pitch–time inter- more, these effects were consistent with the action for the “classify both” instruction of structural manipulations in the stimuli. For con- Experiment 3 (h2 ¼ .061) versus the “rate both” ditions in which the salience was more evenly instruction of Experiment 1 (h2 ¼ .014). The matched between dimensions (as measured by main effect sizes of pitch and time were more main effect size), the two dimensions also evenly matched in the “classify both” condition showed a statistical interaction such that increas- of Experiment 3 (difference score ¼ .078) than ing structural conformity in either dimension in the “rate both” condition of Experiment 1 exaggerated the effect of the other dimension. (difference score ¼ .345), and there was a stronger These interactions appeared in all experiments interaction in Experiment 3. For musically trained when participants responded on the basis of performers, this inverse relation between the effect both dimensions, and also in the “classify time” size difference score and pitch-time interaction instruction of Experiment 3. effect size as shown in Table 1 is strong, r(4) ¼ There were joint contributions of both pitch –.90, p ¼ .01, even for such a small sample. and time in all conditions of these experiments. Because of the replication of quantitative differ- Moreover, the forms of these contributions were ences between musically trained and untrained lis- consistent with the degree of tonal and metric con- teners for explicitly musical tasks, the untrained formity—as the melody adhered more to its orig- listeners were examined separately (Experiment inal pitch and/or temporal characteristics, 2). In this population the same pattern goodness ratings and classifications of structural emerged—as the main effect sizes of pitch and conformity increased. Neither instruction, nor time became more evenly matched, interaction task, nor whether there was a statistical interaction effect sizes grew (see Table 1). Overall, therefore, between the pitch and time variant levels affected the principle of dimensional salience, explored this pattern. This finding is especially intriguing initially in the context of single notes following a in the selective attention instructions, where struc- musical context (Prince, Schmuckler, et al., 2009; tural conformity in the ignored/irrelevant dimen- Prince, Thompson, et al., 2009) also applies to sion nevertheless influenced responses. Similar longer sequences such as entire melodies. results have been reported elsewhere with respect to both dimensions. For instance, the tonal stability of individual events following a musical GENERAL DISCUSSION context can influence temporal change detection

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 (Lebrun-Guillaud & Tillmann, 2007), phoneme Three experiments tested how the dimensions of monitoring (Bigand, Tillmann, Poulin, pitch and time combine in the perception D’Adamo, & Madurell, 2001), and temporal of complex melodies that varied in their degree classification (Prince, Thompson, et al., 2009). of conformity to pitch and temporal structure. Similarly for time, the metric stability of an In the first experiment, musically trained event can affect detection accuracy of a change in performers rated the goodness of melodies based a pitch pattern (Jones, Johnston, & Puente, only on the sequence of pitches, durations, or 2006), and pitch height comparison (Jones et al., both; in the second experiment, musically 2006; Jones et al., 2002)—but only in a context untrained listeners performed the same tasks. lacking conformity to typical pitch structure For Experiment 3, musically trained performers (Prince, Schmuckler, et al., 2009). Therefore, it classified whether the melody adhered to typical appears that irrelevant stimulus structure is auto- pitch and/or temporal structure. Despite these matically processed and can nonetheless affect per- task and participant differences, the results from ception, even when this operation confers no task all experiments were quite similar. There were benefit. Indeed, work in visual perception proposes

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 21 PRINCE

exactly this concept—a coherent object can auto- perception and reproduction of temporal matically attract attention when identifying the sequences can be affected by other factors such as colour of a target object in a visual array serial patterns (Jones, 1976), rhythmic consonance (Kimchi, Yeshurun, & Cohen-Savransky, 2007). (Monahan & Carterette, 1987), and hierarchical An interesting extension of this notion arises simplicity (Povel & Essens, 1985). The reordering from the ineffective manipulation of tonal and of elements in the second and third levels would metric conformity between the second and third destroy such information that presumably was variant levels of pitch and time. That is, tonality present in the first variant level. and metre are automatically extracted from (and Even though listeners did not demonstrate sen- subsequently affect the processing of) stimuli that sitivity to the difference between the second and only weakly conform to such structures. The third variant levels, pitch and time always affected third variant levels were intended to decrease responses in every condition of all experiments. the degree of tonality and/or metre within the Interestingly, there was variation in how these sequence while still retaining the original dimensions combined—sometimes showing inde- elements, and yet participants rated them as pendent and additive properties, and other times good as (and classified them as equally tonal or demonstrating more complex interactive relations. metric as) the second variant levels that preserved As mentioned earlier, a good explanation for these this structural conformity. The remaining vestiges patterns emerges from examining dimensional sal- of tonality and metre in the third variant level were ience, indexed by the relative effect sizes of pitch sufficient to establish fully fledged corresponding and time as main effects. Interactions were more percepts. Indeed, several reports show that to likely as the salience of pitch and time approached invoke these complex structures, only rudimentary parity. Yet pitch remained the more salient dimen- levels of tonality (Cuddy & Badertscher, 1987; sion except when listeners attempted to con- Krumhansl, 1990; Oram & Cuddy, 1995; Smith sciously ignore it, and even in these conditions & Schmuckler, 2004) and metre (Brochard, the advantage of time over pitch never approached Abecasis, Potter, Ragot, & Drake, 2003; Desain the corresponding advantage of pitch over time & Honing, 2003; Palmer & Krumhansl, 1990; when rating or classifying pitch. Previous work Povel & Okkerman, 1981) are necessary. Given (Prince, Thompson, et al., 2009) found that that impoverished levels of structural conformity when classifying the temporal characteristics of a can nonetheless invoke these organizational fra- note following a typical musical context, the meworks, listeners have most likely internalized tonal stability of the note influenced responses. the tonal and metric hierarchies over years of However, variations in temporal location did not

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 passive exposure and/or automatically use what influence pitch classifications, even though both little information there is in the stimulus to boot- tasks were equally difficult. Thus, Prince, strap the full structure. The difference between the Thompson, et al. concluded that in such musical first and second pitch variant levels (both of which contexts, the dimension of pitch was more salient exemplified tonality) is likely due to the disturb- than that of time. In fact, follow-up work ance of contour as a result of reordering, particu- showed that only upon removing tonal pitch struc- larly the lack of stepwise motion in the second ture did time became sufficiently salient as to influ- (and third) variant level as a result of order ran- ence the processing of pitch (Prince, Schmuckler, domization. It is somewhat more challenging to et al., 2009). account for the corresponding difference in the However, many of the stimuli in the present first and second time variant levels. One possible experiments did not represent typical Western explanation is that the technique of correlating music in that they did not demonstrate tonality note onsets with the metric hierarchy did not and metre. Accordingly, why was pitch consist- capture entirely the contributing factors to ently more salient than time? One possibility is temporal goodness/conformity. Indeed, the that the nature of the tasks increased pitch

22 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

salience. In previous work on single-note ratings of development of music perception, regardless of probe events, Prince, Thompson, et al. (2009) formal musical training. Indeed, infant listeners pointed out that responding to a pitched event (who therefore have not had the benefit of years inevitably makes the task inherently pitch based. of exposure to a musical culture) demonstrate In the context of the present experiments, listeners more interactive patterns of pitch–time inte- did not necessarily have to attend to the temporal gration (Trehub & Hannon, 2009). properties of the melodies when attending selec- In fact, listeners encultured in Western music tively to its pitch characteristics (although they may have internalized a listening strategy that apparently did to some degree). However, when emphasizes pitch at the expense of time, thus evaluating the temporal aspects of the stimuli, lis- creating a mental schema of pitch salience. teners had no choice but to listen to the pitches Recent work on pitch–time integration suggests themselves in order to extract and process the tem- that cultural factors in Western music create a lis- poral properties of the melody. Therefore, the tening strategy that enhances the salience of pitch tasks may have offered an intrinsic advantage to at the expense of time (Prince, Schmuckler, et al., the dimension of pitch, because even the temporal 2009; Prince, Thompson, et al., 2009). This infer- instructions required processing of pitch on some ence depends on two assertions: first, typical level. Studies that use tasks that reverse this Western music emphasizes pitch over time; pattern of dependence (e.g., tapping to the beat second, listeners passively learn these statistical of a melody) often show reverse patterns of dimen- properties through long-term exposure and inter- sional salience, such that manipulations based on nalize listening strategies that emphasize pitch. time predominate over pitch (London, Himberg, The first of these assertions is uncontroversial, & Cross, 2009; Pfordresher, 2003; Snyder & as discussed in the introduction. The second asser- Krumhansl, 2001). In these cases, the temporal tion requires some explanation. In any domain, position is of primary importance, and no pitch perceivers learn to attend to the informative processing is necessary. aspects of a stimulus at the expense of other An issue that music cognition research faces aspects through experience provided by passive regularly is the concern that the use of musically exposure (Bhatt & Quinn, 2011). Perhaps the trained participants restricts the generality of the most striking examples of this principle can be conclusions to that specific population. However, found in developmental studies of metre percep- careful examination of the existing research tion (Hannon & Trehub, 2005a, 2005b) and shows that all encultured listeners perceive music speech perception (Nazzi, Jusczyk, & Johnson, remarkably similarly (Bigand & Poulin- 2000; Werker & Tees, 1984). Essentially,

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Charronnat, 2006). With a specific view towards although infants initially are able to learn a wide abstract organizational structures, both musically range of musical metres and are sensitive to all trained and untrained listeners show the same pat- phonemic contrasts, variation that is not musically terns of perception of tonality (He´bert, Peretz, & or linguistically meaningful in their culture is not Gagnon, 1995; Koelsch, Gunter, Friederici, & informative. Over time, these listeners acquire Schro¨ger, 2000; Loui, Wessel, & Kam, 2010; strategies that maximize sensitivity to the mean- Tillmann & Poulin-Charronnat, 2010; Trainor ingful information (separately for music and & Trehub, 1994; van Egmond & Boswijk, 2007) language), at the expense of sensitivity to other and metre (Geiser, Ziegler, Jancke, & Meyer, information. Accordingly, lifelong passive 2009; Hannon et al., 2004). The present studies exposure to Western music, which features pitch extend these general findings by revealing that above other musical dimensions, may cause listen- musical training is not associated with any quali- ers to internalize a strategy (i.e., mental schema) tative differences in pitch–time integration. that reflects this imbalance, thus creating a salience Instead, it is more likely that the passive exposure of pitch. Importantly, this principle of pitch sal- to the statistical properties of music guides the ience via passive exposure applies to all listeners,

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 23 PRINCE

and not only musicians. Indeed, untrained listeners These findings do support another possible uni- are nonetheless strikingly sensitive to abstract fying theory of dimensional integration, investigated organizational principles of music (Bigand & here in the context of pitch–time integration. Poulin-Charronnat, 2006). Also importantly, this Assessing dimensional relations by their salience as listening strategy might only be invoked in the well as relative perceptual difficulty (i.e., discrimin- context of a stimulus that resembles Western ability) shows promise for explaining many of the music (Prince, Schmuckler, et al., 2009). contradictory reports, both in the context of single However, this theory requires cross-cultural work notes and in longer sequences. Essentially, as sal- comparing Western listeners to others encultured ience and discriminability become more evenly in musical systems that do not favour pitch, such matched, global (not asymmetric) interactions as African, Eastern European, or Australian between these dimensions become more likely. It Aboriginal music. is important to note that salience can be dissociable What contributions do these findings offer? The from discriminability—Prince, Thompson, et al. ultimate aim of this research is to provide a theor- (2009) demonstrated that in typical Western etical framework on dimensional integration that music, there was an imbalance in dimensional sal- can explain how stimulus dimensions combine to ience that caused asymmetric interference of pitch form a coherent percept and also to unify the on time, even with equalized discriminability. complex pattern of findings specifically in the Additionally, Garner’s classic research on dimen- pitch–time integration literature. The current find- sional interactions showed that demonstrably inde- ings are relevant to existing theories of pitch–time pendent dimensions can appear to interact when integration. Regarding the local/global proposal discriminability is unbalanced (Garner & Felfoldy, (Bigand et al., 1999; Jones & Boltz, 1989; 1970). There are procedures for measuring discri- Tillmann & Lebrun-Guillaud, 2006), it is impor- minability, such as those provided by signal detec- tant to note that the present experiments always tion theory (Macmillan & Creelman, 1991), required attention to the entire melody (i.e., Stroop interference (Melara & Algom, 2003), and global tasks), yet the pattern of pitch–time inte- Garner interference (Garner & Felfoldy, 1970) on gration varied. Instead, manipulating instructions reaction times, but there are currently no methods affected whether interactions occurred, rather than to measure dimensional salience. However, the find- showing overall evidence of interactions with ings reported here suggest a possible objective these global tasks. Similarly for the early versus measure of dimensional salience, based on compar- late processing stage theory of pitch–time inter- ing main effect sizes in selective attention con- actions (Peretz & Kolinsky, 1993; Pitt & ditions. In this case, relative main effect sizes | 2 2| Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Monahan, 1987; Thompson et al., 2001), these ( pitch h –timeh ) between stimulus dimensions experiments required cognitive processes that in selective attention conditions predicted the occur at a relatively late stage. However, again the strength of an interaction (pitch × time h2)in pattern of pitch–time integration in the present conditions of attending both dimensions jointly. data varied along the continuum of independent Thus, when equalized in terms of discriminability, and interactive processing, not a homogenous the difference in main effect size between stimulus pattern of interactive relations. Accordingly, expla- dimensions could provide an indicator of relative nations of pitch–time integration based on strong dimensional salience. Hopefully, these findings can formulations of the local/global or stages of proces- guide future work on dimensional interactions and sing theories are unwarranted in the face of the increase our understanding of the cognitive mech- present study. However, without further direct anisms of perception. tests of these theories (i.e., finding independent and interactive contributions in local tasks and at Original manuscript received 10 May 2010 early processing stages), the present data alone Accepted revision received 28 February 2011 cannot discount these theories entirely. First published online day month year

24 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

REFERENCES Brochard, R., Abecasis, D., Potter, D., Ragot, R., & Drake, C. (2003). The “ticktock” of our internal clock: Direct brain evidence of subjective accents in Aldwell, E., & Schacter, C. (2002). Harmony and voice isochronous sequences. Psychological Science, 14, leading (3rd ed.). Belmont, CA: Thomson/Schirmer. 362–366. Ashby, F. G., & Townsend, J. T. (1986). Varieties of Castellano, M. A., Krumhansl, C. L., & Bharucha, J. J. perceptual independence. Psychological Review, 93, (1984). Tonal hierarchies in the music of North 154–179. India. Journal of Experimental Psychology: General, Atkinson, A. P., Tipples, J., Burt, D. M., & Young, A. 113, 394–412. W. (2005). Asymmetric interference between sex and Cousineau, M., Demany, L., & Pressnitzer, D. (2009). emotion in face perception. Perception and What makes a melody: The perceptual singularity Psychophysics, 67, 1199–1213. of pitch sequences. The Journal of the Acoustical Berkowitz, S., Fontrier, G., & Kraft, L. (1997). A new Society of America, 126, 3179–3187. approach to sightsinging (4th ed.). New York, NY: Cuddy, L. L., & Badertscher, B. (1987). Recovery of the W. W. Norton & Company. tonal hierarchy—some comparisons across age and Bhatt, R. S., & Quinn, P. C. (2011). How does learning levels of musical experience. Perception and impact development in infancy? The case of percep- Psychophysics, 41, 609–620. tual organization. Infancy, 16, 2–38. Dawe, L. A., Platt, J. R., & Racine, R. J. (1993). Bigand, E. (1997). Perceiving musical stability: The Harmonic accents in inference of metrical structure effect of tonal structure, rhythm, and musical exper- and perception of rhythm patterns. Perception and tise. Journal of Experimental Psychology: Human Psychophysics, 54, 794–807. Perception and Performance, 23, 808–822. Dawe, L. A., Platt, J. R., & Racine, R. J. (1994). Bigand, E. (2003). More about the musical expertise of Inference of metrical structure from perception of musically untrained listeners. In G. Avanzini, iterative pulses within time spans deﬁned by chord C. Faienza, D. Minciacchi, L. Lopez, & M. Majno changes. Music Perception, 12, 57–76. (Eds.), Neurosciences and music (Vol. 999, Dawe, L. A., Platt, J. R., & Racine, R. J. (1995). pp. 304–312). New York, NY: New York Academy Rhythm perception and differences in accent of Sciences. weights for musicians and nonmusicians. Perception Bigand, E., Madurell, F., Tillmann, B., & Pineau, M. and Psychophysics, 57, 905–914. (1999). Effect of global structure and temporal Desain, P., & Honing, H. (2003). The formation of organization on chord processing. Journal of rhythmic categories and metric priming. Perception, Experimental Psychology: Human Perception and 32, 341–365. Performance, 25, 184–197. Dibben, N. (1994). The cognitive reality of hierarchical Bigand, E., & Poulin-Charronnat, B. (2006). Are we structure in tonal and atonal music. Music Perception, “experienced listeners”? A review of the musical 12, 1–25. capacities that do not depend on formal musical Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Dowling, W. J. (1978). Scale and contour: Two com- training. Cognition, 100, 100–130. ponents of a theory of memory for melodies. Bigand, E., Tillmann, B., Poulin, B., D’Adamo, D. A., Psychological Review, 85, 341–354. & Madurell, F. (2001). The effect of harmonic Dowling, W. J., & Fujitani, D. S. (1971). Contour, context on phoneme monitoring in vocal music. interval, and pitch recognition in memory for melo- Cognition, 81, B11–B20. dies. Journal of the Acoustical Society of America, 49, Boltz, M. G., Marshburn, E., Jones, M. R., & Johnson, 524–531. W. W. (1985). Serial-pattern structure and tem- Dyson, B. J., & Quinlan, P. T. (2010). Decomposing the poral-order recognition. Perception and Garner interference paradigm: Evidence for dis- Psychophysics, 37, 209–217. sociations between macrolevel and microlevel per- Brainard, D. H. (1997). The psychophysics toolbox. formance. Attention, Perception & Psychophysics, 72, Spatial Vision, 10, 433–436. 1676–1691. Bregman, A. S. (1990). Auditory scene analysis: The per- Eerola, T., Ja¨rvinen, T., Louhivuori, J., & Toiviainen, P. ceptual organization of sound. Cambridge, MA: MIT (2001). Statistical features and perceived similarity of Press. folk melodies. Music Perception, 18, 275–296.

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 25 PRINCE

Ellis, R. J., & Jones, M. R. (2009). The role of accent preexisting representation and musical expertise. salience and joint accent structure in meter percep- Canadian Journal of Experimental Psychology–Revue tion. Journal of Experimental Psychology: Human Canadienne De Psychologie Experimentale, 49, Perception and Performance, 35, 264–280. 193–210. Everett, W. (2000). Pitch down the middle. In W. Jones, M. R. (1976). Levels of structure in reconstruc- Everett (Ed.), Expression in pop-rock music: A collection of temporal and spatial serial patterns. Journal tion of critical and analytical essays (pp. 111–174). of Experimental Psychology: Human Learning and New York, NY: Garland. Memory, 2, 475–488. Fraisse, P. (1982). Rhythm and tempo. In D. Deutsch Jones, M. R. (1987). Dynamic pattern structure in (Ed.), The psychology of music (1st ed., music—recent theory and research. Perception and pp. 149–180). New York, NY: Academic Press. Psychophysics, 41, 621–634. Gagnon, L., & Peretz, I. (2000). Laterality effects in Jones, M. R., & Boltz, M. G. (1989). Dynamic attend- processing tonal and atonal melodies with affective ing and responses to time. Psychological Review, 96, and nonaffective task instructions. Brain and 459–491. Cognition, 43, 206–210. Jones, M. R., Boltz, M. G., & Kidd, G. (1982). Garner, W. R. (1974). The processing of information and Controlled attending as a function of melodic and structure (Vol. 203). Oxford, UK: Lawrence Erlbaum temporal context. Perception and Psychophysics, 32, Associates. 211–218. Garner, W. R. (1976). Interaction of stimulus dimen- Jones, M. R., Johnston, H. M., & Puente, J. (2006). sions in concept and choice processes. Cognitive Effects of auditory pattern structure on anticipatory Psychology, 8, 98–123. and reactive attending. Cognitive Psychology, 53, Garner, W. R., & Felfoldy, G. L. (1970). Integrality of 59–96. stimulus dimensions in various types of information Jones, M. R., Moynihan, H., MacKenzie, N., & Puente, processing. Cognitive Psychology, 1, 225–241. J. (2002). Temporal aspects of stimulus-driven Geiser, E., Ziegler, E., Jancke, L., & Meyer, M. (2009). attending in dynamic arrays. Psychological Science, Early electrophysiological correlates of meter and 13, 313–319. rhythm processing in music perception. Cortex, 45, Kessler, E. J., Hansen, C., & Shepard, R. N. (1984). 93–102. Tonal schemata in the perception of music in Bali Hannon, E. E., Snyder, J. S., Eerola, T., & Krumhansl, and in the west. Music Perception, 2, 131–165. C. L. (2004). The role of melodic and temporal cues Kimchi, R., Yeshurun, Y., & Cohen-Savransky, A. in perceiving musical meter. Journal of Experimental (2007). Automatic, stimulus-driven attentional Psychology: Human Perception and Performance, 30, capture by objecthood. Psychonomic Bulletin & 956–974. Review, 14, 166–172. Hannon, E. E., & Trehub, S. E. (2005a). Metrical cat- Koelsch, S., Gunter, T. C., Friederici, A. D., & egories in infancy and adulthood. Psychological Schro¨ger, E. (2000). Brain indices of music proces-

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 Science, 16, 48–55. sing: “Nonmusicians” are musical. Journal of Hannon, E. E., & Trehub, S. E. (2005b). Tuning in to Cognitive Neuroscience, 12, 520–541. musical rhythms: Infants learn more readily than Krumhansl, C. L. (1990). Cognitive foundations of adults. Proceedings of the National Academy of musical pitch. New York, NY: Oxford University Sciences of the United States of America, 102, Press. 12639–12643. Krumhansl, C. L. (2000). Rhythm and pitch in music Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. cognition. Psychological Bulletin, 126, 159–179. (2000). The distributed human neural system for Krumhansl, C. L., & Kessler, E. J. (1982). Tracing the face perception. Trends in Cognitive Sciences, 4, dynamic changes in perceived tonal organization in 223–233. a spatial representation of musical keys. He´bert, S., & Peretz, I. (1997). Recognition of music in Psychological Review, 89, 334–368. long-term memory: Are melodic and temporal pat- Krumhansl, C. L., & Schmuckler, M. A. (1986, July). terns equal partners? Memory and Cognition, 25, Key-ﬁnding in music: An algorithm based on pattern 518–533. matching to tonal hierarchies. Paper presented at the He´bert, S., Peretz, I., & Gagnon, L. (1995). Perceiving 19th Annual Meeting of the Society of the tonal ending of tune excerpts—the roles of Mathematical Psychology, Cambridge, MA.

26 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) STIMULUS DIMENSIONS

Lebrun-Guillaud, G., & Tillmann, B. (2007). Inﬂuence Peretz, I., & Coltheart, M. (2003). Modularity of music of a tone’s tonal function on temporal change detec- processing. Nature Neuroscience, 6, 688–691. tion. Perception and Psychophysics, 69, 1450–1459. Peretz, I., & Kolinsky, R. (1993). Boundaries of separ- Lerdahl, F. (2001). Tonal pitch space. New York, NY: ability between melody and rhythm in music dis- Oxford University Press. crimination—a neuropsychological perspective. Lerdahl, F., & Jackendoff, R. (1983). A generative theory Quarterly Journal of Experimental Psychology, 46A, of tonal music. Cambridge, MA: MIT Press. 301–325. London, J., Himberg, T., & Cross, I. (2009). The effect Pfordresher, P. Q. (2003). The role of melodic and of structural and performance factors in the percep- rhythmic accents in musical structure. Music tion of anacruses. Music Perception, 27, 103–120. Perception, 20, 431–464. Loui, P., Wessel, D. L., & Kam, C. L. H. (2010). Pitt, M. A., & Monahan, C. B. (1987). The perceived Humans rapidly learn grammatical structure in a similarity of auditory polyrhythms. Perception and new musical scale. Music Perception, 27, 377–388. Psychophysics, 41, 534–546. Macmillan, N. A., & Creelman, C. D. (1991). Detection Pomerantz, J. R. (1983). Global and local precedence— theory: A user’s guide. New York, NY: Cambridge selective attention in form and motion perception. University Press. Journal of Experimental Psychology: General, 112, Mathworks. (2004). Matlab [Computer software]. 516–540. Natick, MA: Author Povel, D. J., & Essens, P. (1985). Perception of tem- Melara, R. D., & Algom, D. (2003). Driven by infor- poral patterns. Music Perception, 2, 411–440. mation: A tectonic theory of Stroop effects. Povel, D. J., & Okkerman, H. (1981). Accents in equi- Psychological Review, 110, 422–471. tone sequences. Perception and Psychophysics, 30, Monahan, C. B., & Carterette, E. C. (1987). The inﬂu- 565–572. ence of temporal contour and interval information on Prince, J. B., Schmuckler, M. A., & Thompson, W. F. the discrimination of rhythmic patterns. Journal of the (2009). The effect of task and pitch structure on Acoustical Society of America, 81, S91–S92. pitch–time interactions in music. Memory and Mullennix, J. W., & Pisoni, D. B. (1990). Stimulus Cognition, 37, 368–381. variability and processing dependencies in speech- Prince, J. B., Thompson, W. F., & Schmuckler, M. A. perception. Perception and Psychophysics, 47, 379–390. (2009). Pitch and time, tonality and meter: How Nazzi, T., Jusczyk, P. W., & Johnson, E. K. (2000). do musical dimensions combine? Journal of Language discrimination by English-learning 5- Experimental Psychology: Human Perception and month-olds: Effects of rhythm and familiarity. Performance, 35, 1598–1617. Journal of Memory and Language, 43, 1–19. Schenker, H. (1979). Free composition (Der freie satz), Oram, N., & Cuddy, L. L. (1995). Responsiveness of E. Ostler, Trans. New York, NY: Longman, Western adults to pitch-distributional information (Original work published 1935). in melodic sequences. Psychological Research, 57, Schmuckler, M. A. (2004). Pitch and pitch structures. In

Downloaded By: [Murdoch University] At: 03:47 19 May 2011 103–118. J. Neuhoff (Ed.), Ecological psychoacoustics Ottman, R. W. (1986). Music for sight-singing (3rd ed.). (pp. 271–315). San Diego, CA: Elsevier Science. Englewood Cliffs, NJ: Prentice-Hall. Schweinberger, S. R., Burton, A. M., & Kelly, S. W. Palmer, C., & Krumhansl, C. L. (1987a). Independent (1999). Asymmetric dependencies in perceiving temporal and pitch structures in determination of identity and emotion: Experiments with morphed musical phrases. Journal of Experimental Psychology: faces. Perception and Psychophysics, 61, 1102–1115. Human Perception and Performance, 13, 116–126. Silbert, N. H., Townsend, J. T., & Lentz, J. J. (2009). Palmer, C., & Krumhansl, C. L. (1987b). Pitch and Independence and separability in the perception of temporal contributions to musical phrase percep- complex nonspeech sounds. Attention, Perception & tion—effects of harmony, performance timing, and Psychophysics, 71, 1900–1915. familiarity. Perception and Psychophysics, 41, 505–518. Sloboda, J. A. (1985). The musical mind: The cognitive Palmer, C., & Krumhansl, C. L. (1990). Mental rep- psychology of music. Oxford, UK: Clarendon Press. resentations for musical meter. Journal of Smith, N. A., & Schmuckler, M. A. (2004). The per- Experimental Psychology: Human Perception and ception of tonal structure through the differentiation Performance, 16, 728–741. and organization of pitches. Journal of Experimental

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0) 27 PRINCE

Psychology: Human Perception and Performance, 30, Trainor, L. J., & Trehub, S. E. (1994). Key membership 268–286. and implied harmony in Western tonal music: Snyder, J. S., & Krumhansl, C. L. (2001). Tapping to Developmental perspectives. Perception and ragtime: Cues to pulse finding. Music Perception, Psychophysics, 56, 125–132. 18, 455–489. Trehub, S. E., & Hannon, E. E. (2009). Conventional Temperley, D. (2001). The cognition of basic musical rhythms enhance infants’ and adults’ perception of structures. Cambridge, MA: MIT Press. musical patterns. Cortex, 45, 110–118. Thompson, W. F., Hall, M. D., & Pressing, J. (2001). Trehub, S. E., Schellenberg, E. G., & Kamenetsky, S. B. Illusory conjunctions of pitch and duration in unfami- (1999). Infants’ and adults’ perception of scale struc- liar tone sequences. Journal of Experimental Psychology: ture. Journal of Experimental Psychology: Human Human Perception and Performance, 27, 128–140. Perception and Performance, 25, 965–975. Tillmann, B., & Lebrun-Guillaud, G. (2006). Influence Treisman, A. M., & Gelade, G. (1980). Feature-inte- of tonal and temporal expectations on chord proces- gration theory of attention. Cognitive Psychology, sing and on completion judgments of chord 12, 97–136. sequences. Psychological Research, 70, 345–358. van Egmond, R., & Boswijk, M. (2007). The perception Tillmann, B., & Poulin-Charronnat, B. (2010). of key: The role of music training. Music Perception, Auditory expectations for newly acquired structures. 25, 31–41. Quarterly Journal of Experimental Psychology, 63, Werker, J. F., & Tees, R. C. (1984). Cross-language 1646–1664. speech perception: Evidence for perceptual reorgan- Tong, Y. X., Francis, A. L., & Gandour, J. T. (2008). ization during the first year of life. Infant Behavior & Processing dependencies between segmental and Development, 7, 49–63. suprasegmental features in mandarin Chinese. White, B. W. (1960). Recognition of distorted melo- Language and Cognitive Processes, 23, 689–708. dies. American Journal of Psychology, 73, 100–107. Downloaded By: [Murdoch University] At: 03:47 19 May 2011

28 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 0000, 00 (0)