You are on page 1of 22

Inner Form and Aesthetic Judgment

397

G ROWING O RANGES ON M OZART S A PPLE T REE : I NNER F ORM AND A ESTHETIC J UDGMENT
Z OHAR E ITAN Tel Aviv University, Tel Aviv, Israel R ONI Y. G RANOT The Hebrew University, Jerusalem, Israel
MUSIC THEORISTS OFTEN PRESUME

that the sections of a musical masterwork match organically, enhancing unity and value. This inner form should be distinguished from coherence associated with inter-opus constraints, such as conventional forms. Studies indicate that violating inter-opus constraints hardly affects listeners aesthetic judgments. Here we examine how violating inner form affects such judgments. Musically trained and untrained listeners heard the intact opening movements of Mozarts piano sonatas, K. 280 and 332, as well as hybrids mixing sections from these two movements while maintaining overall form and tonal structure. Participants rated originals and hybrids on aesthetically relevant scales (e.g., liking, coherence, interest), after a single hearing and following extended exposure. Results show no significant preference for originals, even after repeated hearings. Music training tended to enhance preference for hybrid over original. Thus, inner form and its supposed organic unity, presumed tenets of musical genius, may not affect listeners evaluation. Received August 3, 2006, accepted January 11, 2008.

Key words: musical form, music theory, inner unity, aesthetic evaluation, concatenationism

and controversial proponent of thematic unity in music, recalled that the first impetus to his work in music analysis was a question that he submitted to one of his professors, which was left unanswered: Why is it that we cannot produce a convincing musical composition by taking a group or section from one work and linking it to that of another even assuming an affinity of key, rhythm and tempo? (Rti, 1951, p. 348).

UDOLPH RTI, THE PIVOTAL

Though Rtis own attempt at answering this question his theory of the thematic process in musichas been widely contested, the premises underlying the question have been at the core of diverse approaches to music theory and criticism for more than two centuries. Two widely shared premises seem to motivate Rtis question. First, that in a convincing musical work (let alone an acknowledged musical masterpiece) the different sections are somehow unified into a coherent whole. Hence, as 18th century music theorist and pedagogue Francesco Galeazzi maintains, The art of the perfect composer does not consist of the discovery of a galant motive or agreeable passages, but it consists of the exact conduct of an entire piece of music. (Galeazzi, 1796/1998, p. 86). Second, and more specifically, Rtis question implies that in order to generate real musical unity, broad affinities such as common tempo or key are not sufficient. Nor, one may add, are external generic constraints on the relationships among sections, like the stipulations of conventional musical forms. Rather, unity should be based upon distinct features of the piece in question: it should be inner, rather than merely external unity.1 The 18th century Classical style presents the issue of inner versus external unity most patently, as it stipulates clear normative procedures defining musical forms (e.g., allegro-sonata form, minuet and trio) and constrains the ordering of cadences, overall tonal structure, thematic return, or phrase structure. Musicians and aestheticians have nonetheless repeatedly suggested that the value of a masterwork does not chiefly rely on such external constraints, but on inner form and inner unity, stemming from the structure and characteristics of the specific materials of the particular musical work. Thus, in 1807, Koch (quoted in Bonds, 1991, p. 126) maintains that, If one is speaking of the form of art-works in the sense in which the

1 The sources of these notions of artistic (not only musical) unity go further back than the 18th century, and may be traced to Aristotles Poetics (in particular Chapters 7 & 8).

Music Perception

VOLUME

25,

ISSUE

5,

PP.

397417,

ISSN

0730-7829,

ELECTRONIC ISSN

1533-8312 2008

BY THE REGENTS OF THE UNIVERSITY OF CALIFORNIA . ALL

RIGHTS RESERVED. PLEASE DIRECT ALL REQUESTS FOR PERMISSION TO PHOTOCOPY OR REPRODUCE ARTICLE CONTENT THROUGH THE UNIVERSITY OF CALIFORNIA PRESS S RIGHTS AND PERMISSIONS WEBSITE , HTTP :// WWW. UCPRESSJOURNALS . COM / REPRINTINFO. ASP.

DOI:10.1525/MP.2008.25.5.397

398

Zohar Eitan & Roni Y. Granot

content is appropriated to beauty, then one is not referring to that external form of art-works by which genres differ, but rather to the particular manner in which variety is bound to unity. In different ways, such appeals for an inner bind unifying the specific piece of music continued to pervade music aesthetics and criticism in the 19th and 20th centuries. Thus, more than a century later, evoking a favorite metaphor of organicist aesthetics sourced in Goethe and Schlegel, Arnold Schoenberg suggests an even stronger inner bind as a prerequisite for musical value: an organic relationship in which the entire piece stems from a single and unique generative musical idea. A real composer does not compose merely one or more themes, but a whole piece. In an apple trees blossoms, even in the bud, the whole future apple is present in all its detailsthey have only to mature, to grow, to become the apple, the apple tree, and its power of reproducing (Schoenberg, 1947/1984, p. 161). Though the premise that inner unity is a prerequisite for musical value has been widely shared, views regarding its nature and its musical determinants are diverse, and often derive from contrasting aesthetic tenets, music-theoretical systems, and analytical methodologies (see also Bonds, 1991). For instance, while early 18th century sources emphasize the overall unity of musical idea or of affect (and consequentially the need for an unithematic conception of a musical composition), in the latter part of the century musical unity is perceived to be generated by the pattern of relationships among ideas that may themselves be highly dissimilar in both affect and musical structure (Schwartz, 2001). Thus, unity and artistic value are achieved by maintaining a unifying pattern of restraint, balance, reciprocity, or complementarity among conflicting passions or events. This brings into conjunction themes and motives of widely differing character, bringing about contrasts of mood or activity at every structural level. (Schwartz, 2001, p. 74). Both approachesunity by resemblance and unity through relationships among contrastsare integrated in 19th century notions of organic unity which, as Schoenbergs quotation indicates, have carried on well into 20th century music analysis and theory. Organicist music criticism maintains that, the form is given as necessaryparts cannot be removed, added, rearranged without . . . marring or even destroying the whole (Solie, 1980, p. 149). However, these widely shared organicist tenets have themselves yielded, in the last century, a wealth of contrasting views as to the actual musical structures and processes that generate organic unity. Music theorists relate unity, for instance,

to a strict harmonic and voice-leading hierarchy stemming from a single high-level structure (Schenker, 1935/79), to the hidden repetition of a few voice-leading or pitch configurations at different structural levels (Schenkerian analysis, e.g., Burkhart, 1978; see also Schenker, 1935/1979), to a small number of initial motivic cells, generating an entire musical work through transformational procedures (Rti, 1951), or to the production (at the works outset) and resolution of a generative imbalance or unrest, represented by the pieces Grundgestalt (Carpenter, 1983; Schoenberg, 1995).2 In recent decades, the validity of the notion of organic, inner unity for music analysis has come under attack, as proponents of the so-called new musicology have revealed its ideological underpinning and called for alternative models for music criticism (e.g., Kerman, 1980; Maus, 1999; Solie, 1980; Street, 1989). Nevertheless, as recently published debates among leading music theorists attest, concepts of musical unity and their implications for music analysis are still a central issue in music-theoretical discourse.3 Furthermore, any survey of recent publications in music analysis would reveal thatideological debates notwithstandingthe search for hidden inner unities in musical masterworks still underscores, explicitly or implicitly, many recent works of music analyses, vastly differing from each other in methodology and subject matter. Thus, for instance, four of the five articles in a recent issue of the highly influential Journal of Music Theory (Vol. 47/1, 2003; actually published 2005) try to demonstrate how unity is generated by elements or structures characteristic of the musical work in question. James Bakers study of Haydns String Quartet, Op. 76, No. 6, for example, suggests that motivic materials of the opening theme of the quartet are the basis for an extraordinarily unified four-movement cycle, and that the entire quartet is in fact a series of variations on its opening theme (Baker, 2003, p. 85). Kathryn Whitney suggests that the particular expressive character and overall unity in the first scene of Schoenbergs Erwartung (an athematic, seemingly chaotic composition) is

2 For a discussion and critique of the way the organicist credo shapes influential 20th century approaches to music analysis (Schenkers and Rtis), see Solie (1980). 3 See, for example Robert Morgans defense of the role of unity in music analysis (2003), and a series of responses by Agawu, Chua, Dubiel, Korsyn, and Kramer, published in Music Analysis 23/2 (2004).

Inner Form and Aesthetic Judgment

399

chiefly due to the prominence of a few motivic gestures, based on three specific trichords (Whitney, 2003). David Temperleys discussion of end-accented phrases, though not centered on issues of musical unity, concludes with an analysis of Schuberts An die Musik, showing how subtle hyper-metric relationships within and between phrases provide for overall unity and closure in the song (Temperley, 2003); and Byron Almn proposes a number of narrative archetypes (e.g., a theme or motive with a problematic element sheds that element and achieves fuller musical development, Almn, 2003, p. 31) as a basis for music analysis. What these recent examples (all taken from a single issue of a leading music theory journal) indicate is that notions of inner, piece-specific unity still underlie much current analysisperhaps because the search for how elements of a musical piece work together is inherent in the very definition of music analysis (see Bent, 1987). In addition, however, these analyses clearly demonstrate that (just as in previous centuries) there is no general consensus on what such unity is and how it works. Thus, while in two of the four examples above (Baker, Whitney) unity is provided in part by deriving diverse materials from a few motives or pitch sets (thus following the conceptual, though not the technical, footsteps of Rti or Schoenberg), in others it is metric complementation and stabilization (Temperley), or a goal-directed narrative involving a musical element (Almn), that provides for unity.
Investigating the Perception of an Elusive Feature

Our brief survey suggests that notions of unity in music have differed widely, perhaps from their inception, in their underlying aesthetic and music-theoretic premises. Furthermore, even when an aesthetic principle is shared, its application in music analysis is often hotly contested. For instance, the notion that a veiled similarity underneath contrasting surfaces unifies contrasting sections in musical masterworks is shared by thematic transformation theorists like Rti and by some Schenkerian analyses of motivic parallelism (Burkhart, 1978). However, most motives presented in Rti (1951) as a demonstration of such deep unity would not be acknowledged as legitimate motives by Schenkerian theory, and vice versa. Such complex and conflicting notions of musical unity seem to present insurmountable difficulties for the empirical examination of its perception and cognition. How can one test the perception of something where there is neither agreed definition, nor a clear decision procedure that would

determine whether this something actually exists in a musical work? Yet, as mentioned, two tenets seem to be shared by otherwise extremely diverse views of musical unity. First, the assumption that unity is primarily inner generated by distinct relationships between different sections of a specific musical composition, rather than by shared generic or stylistic features. Second, the conviction that such unity is a mark (perhaps even a necessary attribute) of musical value, and particularly characterizes works in the Classic-Romantic music canonpieces widely accepted as masterpieces, by composers such as Mozart, Beethoven, and Brahms. The experiment reported here attempts to examine empirically these two presuppositions, succinctly conveyed by Rtis assertion that one cannot produce a convincing musical composition by taking a group or section from one work and linking it to that of another (Rti, 1951, p. 348). This was done by directly addressing Rtis challenge, as we created hybrids from two Mozart masterworks (the opening movements of the piano sonatas K. 280 and K. 332, both in F major), while controlling for generic constraints, such as conventional form and tonal structure, and compared listeners evaluations of these hybrids with their evaluations of Mozarts original masterworks. If the notion of an inner, organic unity of musical masterworks (whichever way one defines it) has cognitive validity, our blatant interference with Mozarts masterpieces should have affected listeners appreciation of the music, such that hybrids, composed of sections of different works, would be rated lower by listeners than the original masterworks out of which they were assembled. The Classical style of the 18th century provides a particularly apt venue for such examination. First, its clear constraints regarding large-scale tonal structure and shared forms enable control over external form. In addition, theorists and composers of the Classical style were very much engaged with the problem of creating unity within diversity (a problem of lesser importance earlier, in the Baroque, in which notions of the unity of affect reigned). As mentioned, while contrast among themes was accepted and often demanded in the later 18th century, an inner unity of feeling, created through subtle balance and complementation among these contrasting elements, was a chief aesthetic principle voiced by musicians and aestheticians of the period (see Holtmeier, 2000, 2002; Schwartz, 2001). Mozart, perhaps more than any other Classical master, has represented this principle for generations of musicians and critics and, indeed, this study utilizes two of his exemplary piano sonatas to investigate its perception.

400

Zohar Eitan & Roni Y. Granot

FIGURE 2. The arpeggiated motives of the opening of K. 280/I evoked in the 2nd subject.

FIGURE 1. Three statements of an arpeggiated tonic triad motive (x, y, z) in the opening theme of K. 280/I.

An Analytical Interlude: Inner Unity in K. 280

Though there is no generally agreed upon notion of inner unity (and, as we have argued, such a notion is not a prerequisite for our investigation), we shall attempt to flesh out this concept by revealing relationships between different sections that may constitute such unity in one of the two pieces investigated in this study, relationships that are violated in the hybrid created for this experiment. In the analysis below, we shall first discuss some connections and similarities unifying the two main themes of the first movement of Mozarts K. 280 (mm. 1-13, 27-43), divorced from each other in the hybrids we created, and then suggest how the development section of the movement (mm. 57-82) integrates elements of these themes. Our analysis combines observations of surface features and configurations in pitch, rhythm, and articulation with analysis of foreground and lower middleground voice-leading progressions, indebted in part to Beach (1994).4
2ND AND 1ST THEMES

The opening of the 2nd theme (Figure 2, mm. 27-30), also emphasizing the tonic triad motive (in the secondary key, C major), closely emulates two of the above figures. The arpeggiated bass figure y, now ascending, opens the theme (m. 27), evoking the left-hand figure that has initiated the 1st theme not only in pitch structure, but also in rhythm, register, articulation, and octave dou bling. Figure z is evoked by the 51(G-C) descent in mm. 28-30 (Figure 2). Like its predecessor in m. 2, it opens with a 56 5 upper-neighbor figure (m. 28), and pro ceeds with a stepwise descent to 1, now extended through a sequential repetition of its opening upper-neighbor configuration. Note that the sequence in mm. 28-30 itself emulates another part of the opening themethe descending 16th note figures in mm. 7, 9as both progressions are built of the same materials: a descending third (filled by a passing tone in mm. 7, 9) and an upperneighbor figure (Figure 3). In a different way, the structure of the 2nd theme evokes not the opening of the 1st theme, but rather the beginning of the bridge section that follows it (Figure 4c, mm. 13-17).

As Beach (1994) observes, the 1st theme opens with three statements of an arpeggiated tonic triad motive (Figure 1): a right-hand ascending arpeggio (motive x, m. 1); a left-hand broken chord, in quarter notes (motive y, mm. 1-2); and a 531 descent (motive z, 6 5 upper neighbor and pass m. 2), elaborated by a 5 ing tones (anticipating the higher-level progression that underlies the entire 1st theme, mm. 1-13).
4 Note that while adopting some of the analytical insights of Beach (1994), the focus and music-theoretical stance of the present article are very different.

FIGURE 3. Figurations in the 2nd subject of K. 280/I deriving from the


opening subject (mm. 7-9).

Inner Form and Aesthetic Judgment

401

FIGURE 4. The 2nd subject of K. 280/I as a variant of its transition section.

402

Zohar Eitan & Roni Y. Granot

FIGURE 5. A reference to m. 3 (1st subject) in the development section


of K. 280/I.

Figure 4a (indebted to Beach, 1994, Example 1) shows the structure of the opening phrase of the bridge. This 2 phrase projects a stepwise ascent from 1 through V I (repeated twice), to 3, elaborated on a lower level by two I third progressions, 123 (F-G-A) and 234 (G-A-B ). The very same progression establishes the connection between the two opening phrases of the 2nd theme (Figure 4d, mm. 27-30, 31-34), by way of an inner voice articulated by the closing figures of the two phrases (see Figure 4b, adapted from Figure 5 in Beach, 1994). The remainder of the 2nd theme (mm. 35-43) makes the relationship to the bridge even more explicit, by using similar triplet figurations (compare, for instance, the figurations in mm. 36, 38 with mm. 23-25) and chromatic bass progression (mm. 18-23, 35-40). The 2nd theme thus elaborates features and structures prominent in the entire preceding sectionincluding both the opening theme and the bridge passage that ensues.
THE DEVELOPMENT SECTION

FIGURE 6. A repeating figure in the development section of K. 280/I (c) merging the openings of the 1st (a) and 2nd (b) subjects.

The development section of K. 280 is clearly derived from the 2nd theme. However, elements of the 1st theme are subtly integrated into the music, becoming most prominent in the retransition, which concludes the development and leads toward the return of the opening theme. These elements, which may serve to unify both themes and prepare the listener for the return of the opening theme, would lose their sense as unifying features in our hybrid, where the original 1st theme of K. 280 was replaced by that of K. 332. The first stage in this process is the introduction (in an inner voice) of the chromatic quarter-notes descent C-B-B -A (877b6 in the current key of C major), repeated twice (mm. 63-64, 65-66), which alludes to the similar progression (Figure 5, F-E-E -D in mm. 3-4). The second signal of the impending 1st theme is of a different nature: not a pitch motive, but an articulation specifically associated with the very opening of the movement. The upward arpeggiation in the right hand, similar to that which opened the piece, appears

(concurrently with a bass progression derived from the opening of the 2nd theme) in each link of the sequence in mm. 69-74, and in each downbeat of the climactic measures (75-77) that ensue (Figure 6). Following this climax, allusions to the opening theme become even denser. They include the appoggiatura figures deriving from mm. 5-6 of the theme (mm. 80-82); a double neighbor figure around A (see Figure 7, A-B -G A in mm. 78-80), recalling and anticipating the structural double neighbor around C (C-D-B -C) in the 1st theme (mm. 1-6, 83-88); and finally, yet another occurrence of the chromatic C-B-B -A progression (mm. 81-82), now anticipating the similar figure (F-E-E -D, mm. 8586) in the forthcoming 1st theme. Thus, the development prepares the return of the opening theme by the gradually increasing introduction of diverse motivic allusions.

FIGURE 7. The structure of the retransition as a derivation of the structure of the opening subject.

Inner Form and Aesthetic Judgment

403

The Perception of Global Structural Relationships

Empirical studies of the effect of large-scale structure on listeners aesthetic judgments have mostly concentrated on generic, inter-opus constraints. Thus, listeners evaluations of intact musical works were compared with those of scrambled or altered versions of the same pieces that violated conventional formal schemes or long-range tonal closure (Cook, 1987; Karno & Koneni, 1992; Marvin & Brinkman, 1999; Tillmann & Bigand, 1996; Tillman, Bigand, & Madurell, 1998). These studies have repeatedly shown that while listeners are often highly sensitive to structural events and processes, such as tonal closure or modulation, when these occur on a local level (e.g., Thompson & Cuddy, 1992), comparable processes on a larger scale go unnoticed. This was observed for various tasks: distinguishing original from altered pieces (Marvin & Brinkman, 1999; Tillmann & Bigand, 1996), preference or ratings of liking (Karno & Koneni, 1992), expressivity ratings (Tillmann & Bigand, 1996), and ratings of structural features, such as coherence or closure (Cook, 1987). Note, however, that most of the studies cited above focused on the initial impression gained by listeners from a single presentation. As Pollard-Gott (1983) and Krumhansl (1996) show, listeners awareness of structural features (such as thematic variants) may emerge only after repeated exposure. Thus, their sensitivity to global aspects of musical piecesespecially relatively long and complex musical works, which presumably cannot be fully processed in a single hearingmay not be truly evaluated without examining the influence of repeated exposure. Of particular relevance to the present study, and contrasting most other studies investigating the perception of global unity, are the results of two recent experiments that, rather than altering structure within a single piece, created hybrids of several pieces and compared listeners liking and unity-ratings of hybrids with those of the original intact compositions. Tan and Spackman (2005) report that following a single hearing both musically trained and nave listeners gave hybrids lower unity and liking ratings, as compared to those given to intact compositions. This raises the possibility that even though listeners may not be sensitive to external features such as conventional form or to tonal relationships, other global aspects of musical structure, such as inner, thematic unity, are noticeable and may play an important role in listeners aesthetic appreciation. In a follow-up study, Tan, Spackman, and Peaslee (2006) performed a similar experiment, but asked participants to repeat the rating task following two, three, and four hearings of the compositions involved. While results for the first hearing

again show higher rating for originals, further exposure resulted in linear increase of ratings for hybrids, and decrease (though not linear) in ratings of intact compositions, such that after four hearings, ratings for hybrids became higher than those for the originals. Tan et al. interpret these results in light of Berlynes two-factor arousal theory (1971), suggesting that high aesthetic evaluation results from optimal arousal, which depends, among other factors, on perceived complexity. According to their interpretation, the repeated exposure influenced the perceived complexity of both hybrids and originals such that the initial optimal (medium) complexity of intact pieces was reduced to sub-optimal level, while the initial supra-optimal complexity of hybrids was increased to optimal (medium) level. The possible role of inner factors, such as thematic unity, as distinguished from external (inter-opus) cohesive factors in listeners evaluations is not directly addressed in any of the above studies. Most of them manipulate generic structural factors, particularly conventional musical form and overall tonal closure. The two studies by Tan and colleagues (2005, 2006), most similar in their procedure to the present study, do not distinguish between generic and piece-specific aspects of unity, nor do they specify the structural differences between altered and original pieces. Furthermore, the hybrids created for these latter studies involved abrupt changes in key, range, volume, tempo, mood, and style between segments, with little in the musical content to make the three segments sound like they belonged together (Tan et al., 2006, p. 208). Hence, it is questionable as to whether these studies have addressed the perception of large-scale coherence at all, since the ratings of hybrids may have been based on evaluating local transitions between segments, rather than on assessment of global relationships and structure.
Aims and General Design

In contrast to studies manipulating structure in order to examine listeners sensitivity to generic aspects such as overall tonal closure or conventional form, the present study aimed at examining the effect of the supposed inner form, while controlling for such external constraints. Thus, unlike the above studies, we compared listeners judgments of originals (acknowledged Mozart masterworks) with those of hybrids, which combined functionally equivalent segments from these masterworks. In contrast to Tan and Spackman (2005) and Tan, Spackman, and Peaslee (2006), who also compared hybrids to originals, in the current design important external constraints such as conventional form and

404

Zohar Eitan & Roni Y. Granot

TABLE 1. Structure of the Hybrid Composition Used for Group 1. Function in Allegro-Sonata Form Exposition1st group Exposition2nd group Development (up to the retransition) Retransition Recapitulation1st group Recapitulation2nd group Sonata K. 332 K. 280 K. 280 K. 332 K. 332 K. 280 Mm. 1-40 27-56 57-77 125-32 133-77 109-144 Local Initial Harmony C:I Dm:VII7/V Dm:V (or F: III#) F:V F:I Local Final Harmony C:V

tonal structure (as well as a composers style and overall features like tempo, meter, and key) were shared by both original and hybrid. At the same time, the assumed inner unity characterizing the original masterpieces (as Koch, Rti, Schoenberg, and many others would suggest), was violated, by definition, in the hybrid that was created by combining unrelated works. Some relationships that may constitute such inner unity were presented in the analytical interlude above. Rating Mozarts originals higher than hybrids on evaluative scales would thus suggest that such relationships, present in Mozarts original and violated in the hybrid, may have some cognitive validity. In contrast, finding no difference between the ratings given to originals and hybrids would suggest that such supposedly unifying features are not perceived or, at any rate, that they do not affect listeners aesthetic judgments. Alternatively, such a null finding may suggest that the hybrids we created are characterized by new connections, relationships and a delicate balance, which are as unifying as those of the original Mozart piece. Although we will not be able to dissociate these two alternative interpretations, they both have important ramifications for the understanding of music perceptionramifications that may call into question basic, shared assumptions of music aesthetics and analysis (see Discussion section).

While the above design controlled for extra-opus constraints, it did not preclude the possibility that judgments would be based on listeners evaluations of separate segments, in and of themselves, rather than on any perceived overall relationship (either external or internal). For instance, if listeners prefer segment x in composition A over its equivalent x in composition B, then a hybrid that replaces x with x may be rated lower than the intact A simply because x is rated lower than x, and not due to any differences in perceived overall structure. Therefore, in our study, one group of listeners compared the original 1st movement of Mozarts piano sonata, K. 332, with a hybrid, in which some sections of the original movement were replaced with structurally equivalent sections from an earlier Mozart sonata, K. 280 (for brevity we use hereafter the abbreviation G1/K. 332). A second group of different participants compared the intact K. 280 with a hybrid (G2/K. 280), in which the K. 332 segments that were removed in Experiment 1 now replaced structurally equivalent segments of K. 280 (see Tables 1 & 2). Other factors that may affect the perception of musical structure and the ensuing aesthetic judgments are the amount of exposure to the music, the degree to which the music was retained in the listeners memory following this exposure, and their previous music

TABLE 2. Structure of the Hybrid Composition Used for Group 2. Function in Allegro-Sonata Form Exposition1st group Exposition2nd group Development (up to the retransition) Retransition Recapitulation1st group Recapitulation2nd group Sonata K. 280 K. 332 K. 332 K. 280 K. 280 K. 332 Mm. 1-26 41-93 94-124 80-82 83-108 176-229 Local Initial Harmony C:I Dm:VII7/V Dm:V (or F: III#) F:V F:I Local Final Harmony C:V

Inner Form and Aesthetic Judgment

405

training, which may affect their perception of structurally significant qualities, such as harmonic structure. All of these variables were manipulated in the current study, as described next in the Method section.
Method Participants

One hundred and sixteen university students participated in the experiment. Group 1 comprised 61 participants (26 male, 35 female; mean age = 24.1 years, SD = 5.8), 29 of whom were musically trained (>7 years of formal music training). A separate group of 55 participants comprised Group 2 (25 male, 30 female; mean age = 24.0, SD = 4.4), 25 of whom were musically trained. Pianists and composers, who were expected to know the test pieces well, were excluded. Participants were paid approximately $17 for two 40-min sessions and a home listening assignment.
Musical Stimuli

going to listen to two versions of a movement from a Mozart piano sonata, both composed by Mozart but only one approved by him for publication. They then listened to the two pieces (K. 332 and its hybrid or K. 280 and its hybrid), presented to different participant groups in counterbalanced order. After listening to each piece, they were presented with 10 statements about the music they had just heard, and were asked to rate the degree of their agreement with each statement on a numerical scale of 0-4 for statements 1-7, and of (2) to (+2) for statements 8-10, which were converted for the statistical analysis, such that zero was coded as 4, (1) and (+1) as 2 and (2) and (+2) as 0. The statements (presented in random order) were: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. The piece in this version is coherent (unified). The piece in this version is interesting. This version is a masterpiece. This version presents conspicuous contrasts among different sections. I liked the piece in this version. All the musical ideas in this version stem from its initial phrases. The performance just heard is excellent. This version is too long (mark 1 or 2), too short (1 or 2) or just the right duration (0). The performance you have just heard is too fast (1 or 2), too slow (1 or 2) or in just the right tempo (0). The performance you have just heard is too dramatic (1 or 2), too flat and expressionless (1, 2) or possessing just the right measure of expressivity (0).

Participants in Group 1 listened to two musical pieces. The first was the opening movement of Mozarts piano sonata in F major, K. 332. The second was a hybrid, which begins with the opening section of K. 332 but progresses differently, due to the replacement of two sections with structurally analogous sections from the 1st movement of a different Mozart piano sonata, K. 280 (Table 1). The complementary material was presented to Group 2: the intact 1st movement of Mozarts K. 280 and its hybrid, which begins with the opening section of K.280, but continues with analogous sections from K. 332. This hybrid thus included sections excluded from the hybrid presented to Group 1 (Table 2). The two movements used in this study share meter, tempo (in the performance used), key, tonal structure, and some voice-leading and registral features (Beach, 1994), but differ considerably in character and thematic material. We used Steinbergs Wavelab 4.5 audio editing software to create the hybrids. Three professional classical musicians scrutinized the hybrids prior to the experiments and found the links between their sections smooth and stylistically credible. Mitsuko Uchidas recording of Mozarts sonatas (Philips B00005QDYG) was used for all music materials.
Procedure
SESSION 1

Experimental sessions were conducted in small groups (3-6 participants). Participants were told that they were

These evaluative statements differ in the criteria they apply (e.g., coherence, interest, appropriate duration), in the degree of objectivity they invite (e.g., this version is a masterpiece, vs. I like this version), and in the object of evaluationthe musical work itself or its performance. Participants were given about 5 min to complete the task. They then listened to the second version, and repeated the task. Finally, they were asked two concluding questions: (1) Which version do you prefer? (2) Which version, in your opinion, did Mozart choose to publish? Participants also supplied a free account of their choices in these questions. To examine the effect of exposure, each of the experiments was conducted twice: after a single hearing of each version (Session 1), and following a weeklong exposure to the music (Session 2). Thus, at the end of the first session, participants were presented with a CD containing the recordings of the original and the hybrid they had just heard in the experimental

406

Zohar Eitan & Roni Y. Granot

session. They were asked to listen to the CD frequently and attentively before returning for the second session. Listening was encouraged by telling the participants that their payments would be linked to their success in a recognition test, to be taken during the second session.
SESSION 2

The second session took place about a week after the first session, and started with a recognition test, in which participants heard 16 music phrases and were asked to indicate which of these phrases had been included in their take-home CD. Of the 16 phrases presented to Group 1, nine were included in the CD, three were phrases of K. 280 not included in the hybrid, and four were taken from another Mozart sonata movement (K. 570/I). In the recognition test of Group 2, three of the seven phrases not included in the CD were taken from K. 332. Following the recognition test, participants repeated the task performed in Session 1. At the end of the session they were asked to explain in writing the reasons for their selections.

Hypotheses

We hypothesized that the notion that music masterpieces possess an inner unity would be supported by the following results: 1. Ratings of Mozarts original compositions would be higher than those of hybrids. 2. Tendencies to prefer originals (as indicated by differences in ratings of originals and hybrids) would strengthen in the 2nd session (following repeated exposure to the music). 3. In the 2nd session, tendencies to prefer originals would be stronger for participants whose scores in the recognition test were higher, indicating a higher level of familiarity with the music materials. 4. Tendencies to prefer originals would be stronger for musically trained participants. 5. Ratings for coherence would correlate with overall preferences, in particular with objectively expressed preferences (this is a masterpiece).

Results

The hypothesis that the original masterworks, supposedly possessing superior inner unity, would be preferred by listeners over hybrids was not confirmed by our data.

Tables 3A-3D present results of several comparisons between original and hybrid; each of these four tables refers to one experimental session in one of the two participants groups (Group 1, comparing K. 332 with its hybrid, and Group 2, comparing K. 280 with its hybrid). The distribution of participants selecting the original or the hybrid as either their preferred version or the version they assume Mozart chose is presented in the first two rows of Tables 3A-3D. Overall, chi-square tests (df = 1) indicate that this result does not differ from chance (50%). For example, in the 1st session of Group 1 (N = 61; see Table 3A), 26 participants preferred Mozarts original K. 332, whereas 34 participants preferred the hybrid (result not significantly different from an even distribution). The only result approaching significance was obtained in the 2nd session of Group 2 (Table 3D), in which 28 participants preferred the hybrid version versus only 16 who opted for the original 2(1, N = 51) = 3.27, p = .07; note, though, that this result is opposite to that suggested by our first two hypotheses. We also compared, using paired samples t-tests, the average rating for the 10 individual statements, pooled together, in both K. 332 and K. 280 and in both sessions (row no. 3 in Tables 3A-3D). Consistent with the results of the two summarizing statements, none of the results indicate a preference for the original over the hybrid. If anything, there is again a marginally significant higher pooled rating for the hybrid over the original (p = .086) in the 2nd session of Group 2 (Table 3D). Finally, we compared participants average rating for each of the ten individual statements in response to the hybrids versus the originals. Wilcoxon paired sampled statistics, corrected for multiple testing using the false discovery rate (FDR) procedure (Benjamini & Hochberg, 1995), revealed that only statement 6 yielded a statistically significant difference between the two versions: the hybrid of K. 332 was considered more appropriate in duration, as compared to the original (Z = 4.21, p < .001; Z = 3.94, p < .001 in Session 1 and Session 2 respectively). In effect, the hybrid, which lasted 5 min and 20 s, was 98 s shorter than Mozarts original K. 332, lasting 6 min and 58 s. If the supposed superior inner form of masterworks affects listeners preferences, one would expect that the increased exposure during the week in which the participants were instructed to listen intensively to the pieces would have resulted in increased preferences for the originals over the hybrids (Hypothesis 2). Again, our results do not support this hypothesis. For both experimental groups, there was no significant difference between Session 1 and Session 2 in the distribution of participants who personally preferred the original, the hybrid, or had no preference (~10%). This was

Inner Form and Aesthetic Judgment

407

TABLE 3A. Comparisons of the Original K. 332/I with Its Hybrid (Session 1). K. 332 Session 1 (N = 61) Original 1 2 3 4
**p < .01

Hybrid 35 33 2.90 0.48 3.53 0.94

Statistic x2 = 1.33, df = 1 x2 = 1.79, df = 1 t = 1.66 Wilcoxon = 4.21**

N (chosen by Mozart) N (personally preferred) Mean and SD of ratings across 10 statements Mean and SD for duration ratings (statement 8)

26 23 2.74 0.54 2.47 1.36

TABLE 3B. Comparisons of the Original K. 332/I with Its Hybrid (Session 2). K. 332 Session 2 (N = 59) Original 1 2 3 4
**p < .01

Hybrid 31 28 2.90 0.52 3.44 1.05

Statistic x2 = 0.15, df = 1 x2 = 0.17, df = 1 t = 0.36 Wilcoxon = 3.94**

N (chosen by Mozart) N (personally preferred) Mean and SD of ratings across 10 statements Mean and SD for duration ratings

28 25 2.86 0.46 2.32 1.40

TABLE 3C. Comparisons of the Original K. 280/I with Its Hybrid (Session 1). K. 280 Session 1 (N = 53) Original 1 2 3 4 N (chosen by Mozart) N (personally preferred) Mean and SD of ratings across 10 statements Mean and SD for duration ratings 21 19 2.66 0.50 3.19 1.14 Hybrid 32 27 2.78 0.52 2.65 1.25 Statistic x2 = 2.28, df = 1 x2 = 1.39, df = 1 t = 1.33 Wilcoxon = 2.21

TABLE 3D. Comparisons of the Original K. 280/I with Its Hybrid (Session 2). K. 280 Session 2 (N = 51) Original 1 2 3 4
*p < .1 > .05

Hybrid 31 28 2.85 0.64 2.88 1.28

Statistic x2 = 2.37, df = 1 x2 = 3.27*, df = 1 t = 1.75* Wilcoxon = 1.52

N (chosen by Mozart) N (personally preferred) Mean and SD of ratings across 10 statements Mean and SD for duration ratings

20 16 2.67 0.61 3.27 1.12

408

Zohar Eitan & Roni Y. Granot

TABLE 4. Comparison of the Relative Ratings for Originals and Hybrids before (Session1) and after (Session2) Extended Exposurea. Mean Difference between Rating of Original & Hybrid (SD) in K. 332 Statement Coherence Interest Masterpiece Contrast Like Derived from a single motive
a

Mean Difference between Rating of Original & Hybrid (SD) in K. 280 Z (Wilcoxon) 0.21 0.90 0.55 0.05 0.35 0.40 Session 1 (N = 53) 0.29 1.38 0.14 1.06 0.33 0.99 0.08 1.68 0.37 1.17 0.37 1.52 Session 2 (N = 51) 0.31 1.33 0.20 1.28 0.49 1.19 0.00 1.83 0.53 1.42 0.14 1.17 Z (Wilcoxon) 0.16 0.41 1.03 0.53 0.82 2.76

Session 1 (N = 61) 0.14 1.25 0.00 1.13 0.03 1.23 0.19 1.37 0.02 1.43 0.02 1.12

Session 2 (N = 59) 0.17 1.33 0.10 1.12 0.09 0.84 0.17 1.49 0.03 1.47 0.05 1.16

Negative rating values indicate that mean ratings for the hybrid are higher than for the original.

confirmed using the McNemar-Bowker test (2(3, 59) = 3.72, p > .2 in K. 332 and 2(3, 51) = 2.14, p > .5 in K. 280).5 Similarly, extended exposure did not change the relative proportion of participants attributing either original or hybrid to Mozart, 2(1, 59) = 0.26, p > .6 in K.332 and 2(1, 51) = 0.004, p > .9 in K. 280. The same pattern of results is also evident in the rating of the individual statements. In Table 4 we present the data for the first six statements (excluding the statements that relate to performance features, rather than to structural features of the experimental items). For each statement we compare the differences () between the mean ratings of original and hybrid in Session 1 with those in Session 2 (i.e., following a week-long exposure). Wilcoxon paired samples statistics, corrected for multiple testing using the false discovery rate (FDR) procedure, indicate that the relative ratings of original vs. hybrid in all 10 statements was not affected by exposure in either Group 1 (K. 332) or Group 2 (K. 280). To further explore the effect of familiarity, we compared, for each of the 10 statements, ratings of high scorers (above the mean) in the recognition test performed at the 2nd session, with those given by low scorers

(Hypothesis 3). This measure was taken as an indication of a higher degree of familiarity with the music materials, derived either from more extensive listening between sessions or from higher musical memory skills. Table 5A presents, using the Mann-Whitney U test, the rating differences between the original and the hybrid with respect to the first six statements described above. Results indicate no significant difference between good and bad memorizers with regard to any of these statements. Correspondingly, chi-square tests indicate that both good and bad memorizers responded similarly to the two summarizing questions (Table 5B). In sum, repeated exposure, as well as familiarity (as indicated by performance in the recognition test), did not result in preference for Mozarts masterworks over our hybrids.6 Finally, we tested the hypothesis that familiarity with the relevant musical style would result in a clear preference for the Mozart originals over the hybrid versions. More specifically, we assumed that musically trained participants, who had been exposed to music of the

5 The McNemar-Bowker test extends the McNemar test, so that the measured variable can have more than two possible outcomes (Brace, Kemp, & Snelgar, 2006, p. 127).

6 This is also consistent with the fact that we did not find any significant differences between the rating of participants who indicated they were unfamiliar with the pieces and those few (9 in Group 1 and 10 in Group 2) who attested to prior exposure to the pieces in question. Importantly, none of those familiar with the piece was aware of the structural intervention in the pieces. Hence, we pooled the results of these participants together with those of the others.

Inner Form and Aesthetic Judgment

409

TABLE 5A. Mean Difference (Upper Cell Entry) and SD (Lower Cell Entry) of Rating between Original and Hybrid for Both Levels of Success at Recognition in Session 2. Mean Difference & SD of Rating between Original and Hybrid K. 332 Bad Memorizers N = 29 0.14 1.39 0.00 1.00 0.00 1.00 0.00 2.00 0.10 1.61 0.10 1.45 Good Memorizers N = 23 0.24 1.28 0.16 1.14 0.11 0.94 0.16 1.40 0.08 1.42 0.14 0.98 MannWhitney U 376.0 382.5 368.5 376.5 363.0 352.5 Mean Difference & SD of Rating between Original and Hybrid K. 280 Bad Memorizers N = 37 0.26 1.32 0.17 1.30 0.39 1.12 0.32 1.70 0.30 1.18 0.09 0.95 Good Memorizers N = 21 0.24 1.35 0.28 1.31 0.55 1.24 0.24 1.88 0.76 1.57 0.17 1.31 MannWhitney U 314.0 317.5 289.5 287.5 272.5 324.5

Statement Coherence Interest Masterpiece Contrast Like Derived from a single motive

TABLE 5B. Distribution of Responses to Summarizing Questions Comparing the Original and Hybrid for Both Levels of Success at Recognition in Session 2. K. 332 N-Mozart vs. N-Hybrid Summarizing Questions Personally preferreda Chosen by Mozart
a

K. 280 N-Mozart vs. N-Hybrid Good Memorizers N = 29 8 vs. 17 20 vs. 9 Bad Memorizers N = 23 8 vs. 11 11 vs. 11

Good Memorizers N = 21 8 vs. 9 9 vs. 12

Bad Memorizers N = 37 16 vs. 19 18 vs. 19

2 0.08 0.18

2 0.48 1.88

Analysis excludes participants who had no preference for either version.

Classical style through extended listening, performance, and studies of music theory and history, would be more likely to recognize the subtle unifying features in Mozarts masterworks and thus rate them as superior to the hybrids (Hypothesis 4). Once more, this hypothesis was not supported by our results. Mann-Whitney U tests, comparing the differences between mean ratings of the original and the hybrid in the musicians group with those of the nonmusicians, revealed no significant differences as a function of music training (Tables 6 & 7). Paradoxically, the only significant difference between musicians and nonmusicians is revealed in the second session in Group 1 (K. 332), albeit in a direction that contradicts our hypothesized training effect (see Table 6b).

Musically trained participants, after extensive exposure to the musical pieces, selected the hybrid as the piece chosen by Mozart for publication more than the musically nave participants. Whereas 40.6% of the musically nave participants attributed the hybrid to Mozart, 66.7% of the musically trained assumed that the hybrid had been chosen by Mozart for publication (p < .05).
Correlations Among Evaluative Scales

The multifaceted evaluations used in this study enable an investigation of the relationships among different evaluative measures. Specifically, they afford an examination of how general judgments, such as I like this

410

Zohar Eitan & Roni Y. Granot

TABLE 6A. Mean Difference (Upper Cell Entry) and SD (Lower Cell Entry) of Rating between Original (K. 332) and Hybrid as a Function of Music Training. Mean Difference & SD of Rating between Original and Hybrid (K. 332) in Session 1 Statement Coherence Interest Masterpiece Contrast Like Derived from a single motive Nonmusicians N = 32 0.22 1.36 0.00 1.11 0.03 1.12 0.28 1.44 0.16 1.32 0.06 1.05 Musicians N = 29 0.03 1.21 0.00 1.16 0.17 1.34 0.07 1.28 0.10 1.52 0.07 1.25 Mean Difference & SD of Rating between Original and Hybrid (K. 332) in Session 2 Nonmusicians N = 32 0.06 1.34 0.13 0.83 0.09 0.73 0.38 1.68 0.06 1.48 0.03 1.33 Musicians N = 27 0.30 1.32 0.07 1.41 0.08 0.98 0.07 1.21 0.00 1.49 0.15 0.91

MannWhitney U 413.5 449.0 444.0 424.5 429.5 410.5

MannWhitney U 379.5 408.5 381.5 348.5 399.5 371.5

Note: negative values indicate that mean ratings for the hybrid are higher than for the original.

TABLE 6B. Personal Preference and Attribution to Mozart of the Original K. 332 versus its Hybrid: Results in Both Sessions as a Function of Music Training. Session 1 N-Mozart vs. N-Hybrid Summarizing Questions Personally preferreda Chosen by Mozart Nonmusicians N = 32 11 vs.17 16 vs. 16 Musicians N = 29 12 vs. 16 10 vs. 19 2 0.74 1.50 Session 2 N-Mozart vs. N-Hybrid Nonmusicians N = 32 17 vs. 14 19 vs. 13 Musicians N = 27 8 vs. 14 9 vs. 18 2 1.76 3.98*

* p < .05 a Analysis excludes participants who had no preference for either version.

TABLE 7. Personal Preference and Attribution to Mozart of the Original K. 280 versus its Hybrid: Results in Both Sessions as a Function of Music Training. Session 1 N-Mozart vs. N-Hybrid Summarizing Questions Personally preferreda Chosen by Mozart
a

Session 2 N-Mozart vs. N-Hybrid 2 1.01 0.38 Non Musicians N = 28 9 vs. 14 10 vs. 17 Musicians N = 25 7 vs. 14 10 vs. 14 2 0.16 0.11

Nonmusicians N = 28 12 vs.13 10 vs. 18

Musicians N = 25 7 vs. 14 11 vs. 14

Analysis excludes participants who had no preference for either version.

Inner Form and Aesthetic Judgment

411

TABLE 8. Correlation Matrix (Spearmans Correlation Coefficient r) in Group1/K. 332. Like Like in Session 1a Like in Session 2 Masterpiece in Session 1 Masterpiece in Session 2
a

Coherence .31* .18 .23 .19

Interest .68*** .69*** .64*** .57***

Performance (general) .54*** .44** .46** .27

Performance Tempo .43** .40** .02 .13

Expression in Performance .34* .32* .356* .24

Duration .28 .55*** .41** .40**

.50*** .64***

Two statementsthematic contrast (4) and thematic derivation (6)did not correlate with like or masterpiece in either session, and thus do not appear in this table. *p < .05; ** p < .01; *** p < .001

version or this version is a masterpiece correlate with more specific evaluative measures. To examine these relationships, we tested Spearmans correlations (two-tailed) among listeners ratings of the 10 specific statements. Correlations were measured, in each group, with regard to differences between ratings of the two versions presented. We were particularly interested in correlations between each of the two general evaluative statements, I like this version (subjective evaluation; statement 5) and this version is a masterpiece (objective evaluation; statement 3) with statements associated with structural unity, particularly coherence (no. 1), contrasts (no. 4), and derivation (no. 6), as such correlations would support Hypothesis 5. In Group 1 (K. 332; see Table 8), differences in rating the original and the hybrid with respect to the two general statements 5 (liking) and 3 (masterpiece) were not significantly correlated with differences in rating the two pieces with regard to statements 4 (contrasts) and 6 (derivation). Coherence (no. 1) was correlated with liking only in the first session, and not correlated with masterpiece at all. Note that these results do not seem to stem from a lack of understanding of the notion of musical coherence: an indication that participants did understand this notion (though perhaps superficially) lies in the positive correlation found between statement 1The piece in this version is coherent (unified) and statement

6All the musical ideas in this version stem from its initial phrases (r = .33, p < .05 in Session 1, and r = .57, p < .0001, in Session 2), and the expected negative correlation between statement 1 and statement 4This version presents conspicuous contrasts among different sections (r = .45, p < .01 in Session 1, and r = .37, p < .05 in Session 2). In contrast to the generally low correlations between features associated with coherence and the two general evaluations, differences in rating the original and the hybrid with respect to the two general statements I like this version and this version is a masterpiece were highly correlated with the different evaluations of performance (statements no. 7, 9, and 10), or appropriateness of duration (statement no. 8). Like and masterpiece were also strongly associated with each other. Though most correlations in Group 2 (K. 280; see Table 9) are similar to those in Group 1, one noteworthy difference emerged in Session 2. In contrast to Group 1, in Group 2/K. 280 there was a correlation between ratings of the pieces coherence and ratings of both liking and masterpiece. By and large, most correlations were shared by both musicians and nonmusicians. Notably, regardless of music training, performance considerations (statements 7, 9, 10) correlate with the general preferences/ judgments of like and masterpiece better than

TABLE 9. Correlation Matrix (Spearmans Correlation Coefficient r) in Group 2/K. 280. Like Like in Session 1a Like in Session 2 Masterpiece in Session 1 Masterpiece in Session 2
a

Coherence .16 .53*** .32 .37*

Interest .63*** .62*** .64*** .36*

Performance (general) .55*** .55*** .51*** .48*

Expression in Performance .55*** .40* .32 .37*

Duration .36* .37* .31 .35*

.78*** .84***

Three statementsthematic contrast (4), thematic derivation (6) and performance tempo (9)did not correlate with like or masterpiece in either session, and thus do not appear in this table. *p < .05; **p < .01; ***p < .001

412

Zohar Eitan & Roni Y. Granot

TABLE 10. Classification of Free Verbal Responses in Group 1 (K. 332) and Group 2 (K. 280). Category of Response Interest, complexity, richness, originality Example (K. 280): The second version has more musical ideas. It is more varied, has many themes, and it is more surprising and exciting. Expressivity, drama, depth Example (K. 332): The first version is more interesting. It is as if it has subdivisions into scenes and one can actually build around it an interesting story with events which develop but at the beginning it sounds less coherent than the second version. Lightness, delicacy charm, Mozartness Example (K. 332): The first version is more modern in style and in certain points too winding, too long and less unified. The second version is more consistent with the style and epoch of the composer, lightermore logical, more correct. Coherence, continuity, completeness, less contrasts Example (K. 280): The second version presents less contrasts. On my view, in the first version there are low bass notes which create a contrast with the character of the rest of the piece, and therefore the second version in which there are less contrasts is more appropriate to the Classical style...the second version seems to me more delicate and therefore more appropriate to the Classical style. Brevity, faster tempo Example (K. 280): The first version, because it is shorter. Aspects of dynamics Example (K. 280): The performance is more expressive, interesting with more changes in dynamics as compared to the first version in which there are no perceivable dynamic changes, and less expression. 280 Original 1 280 Hybrid 3 332 Original 8 332 Hybrid 3

structural ones, such as coherence, (statement 1) and derivation (statement 6). One surprising training-related difference, however, should be noted: contrary to our expectations, in Group 2, Session 2, higher coherence ratings correlate with ratings of liking for nonmusicians (r = .626, p < .01), but not for musicians.
Free Verbal Responses

In order to gain some further insights regarding participants considerations in evaluating the works, we examined their free accounts of the reasons that led them to select a particular version as their preferred version or as Mozarts preferred version. In Table 10 we summarize these free comments, classifying the motives in the participants statements into six categories: interest or complexity; drama; features of the

presumed Classical or Mozart style; coherence or lack of contrasts; considerations of duration or tempo; and considerations of dynamics, timbre, and register (secondary parameters; see, e.g., Meyer, 1973). We did not perform any statistical analysis on these data, for several reasons. First, about half of the participants chose not to address the free text question. Second, in several comments more than a single consideration is mentioned (e.g., in the second citation both interest and drama appear), in which case we included the same comment in the two or more appropriate categories, thus giving extra weight to some participants views. Third, the categories themselves, and the assignment of participants texts into these categories, result from the authors rather subjective judgments. Nevertheless, we believe that the free texts provide valuable information as to the participants judgments, and we discuss some of their implications below.

Inner Form and Aesthetic Judgment

413

Discussion Perception of Inner Form

Support for Strict Concatenationism?

At the beginning of this article we quoted Rtis rhetorical question, asking why a convincing musical composition cannot be produced by taking a group or section from one work and linking it to that of another. The findings presented above suggest that, music theorists convictions notwithstanding, a convincing musical composition may indeed be produced in precisely that way. In both experimental groups, and for musicians and nonmusicians alike, musical hybrids seemed as convincinginteresting, likeable, preferable, truly Mozarteanfor our participants as Mozarts original compositions, even after extended, repeated hearings. In contrast with our hypotheses, neither extended exposure nor better recognition of the music resulted in preference for the original. Rather (for musically trained participants), repeated exposure even tended to enhance a preference for the hybrid over Mozarts original (Group 1). In other words, our hypothesis that better acquaintance with the piece, gained over repeated hearings, would lead, at least for the musically trained, to acknowledgment of the subtle effects of inner form, and hence to a preference for Mozarts originals, was not confirmed. This result is in line with findings reported by Tan et al. (2006), although in their study exposure tended to elevate the hybrid ratings for both musically trained and untrained participants, while in our study this tendency (itself weaker than in Tan et al.) was confined to the musically trained listeners, probably due to the subtler nature of the alterations we employed. Note that although repeated exposure did not enhance the original over the hybrid ratings, it did have an effect on listeners evaluative criteria. For instance, whereas in the 1st session of Group 2 (K. 280) listeners ratings of coherence did not significantly correlate with overall liking of the music, in the 2nd session this correlation was highly significant; in contrast, the correlation of liking and masterpiece evaluations with ratings of performance expressiveness considerably declined from Session 1 to Session 2 (see Table 9). This suggests that the role of structural features in determining the overall evaluation of the music increased with repeated hearings, a result consistent with findings of previous studies involving repeated exposure to music (PollardGott, 1983; Krumhansl, 1996). Nevertheless, this seeming increase in attention to structural features did not result in a preference for the supposedly better structured, more unified original over the hybrid.

How can the above findings be interpreted? The lack of listeners preference for Mozarts originals seems to concur with previous findings (summarized previously) indicating that violations of large-scale structure have little effect on aesthetic appreciation, and are often unnoticed even by experienced listeners. Such findings have led to the conclusion that large-scale structure has only a minor role in aesthetic response to music (Tillmann & Bigand, 2004), supporting the concatenationist position voiced by philosopher Jerrold Levinson (1997). Levinson, following 19th century musician and psychologist Edmund Gurney, attributes aesthetic response in music chiefly to local effects and relationshipsto the appreciation of individual phrases and sections and to the cogency of succession between them. In fact, while Levinson carefully qualifies this strict concatenationist view, results of empirical studies, including the present one, suggest that these qualifications may be unnecessary. In supporting a concatenationist view, this study complements previous studies in an important way. Earlier studies have examined how conventional external form (such as formal schemas and large-scale tonality), which any moderately trained composer could have successfully applied, affect aesthetic appreciation. Here, in contrast, we examined (albeit in a limited way; see caveats below) the perceptibility of what generations of musicians and music critics have presumed to be the hallmark of musical genius: inner form and the supposed organic unity it entails. This hallmark seems to evade listeners, both trained and untrained, after either a single or multiple listenings.
Distinguishing Distinct and Style Related Features

While the present results lend some support to concatenationism, the listeners lack of preference for Mozarts originals over the hybrids could have stemmed from sources other than a general inability to grasp large-scale or temporally remote musical relationships. The idea of inner, as differentiated from external form, is necessarily related to a notion of distinctive features, that is, features appearing in the relevant composition more frequently or saliently than in a relevant wider corpus, such as Mozarts piano music. In assuming inner unity we presuppose that some features or relationships distinct to the piece in question contribute to a sense of coherence within it. To appreciate inner unity, then, listeners must not only identify and categorize such features across wide time-spansitself a difficult long-term memory and categorization taskbut also differentiate them

414

Zohar Eitan & Roni Y. Granot

from the style related features characterizing the wider corpus in question, and use this distinct in evaluating the piece (see also Huron, 2001). Added to the other tasks of long-term memory, categorization, and schematization demanded of the attentive listener, this cognitive demand may be a tall order for most listeners, even those familiar with the relevant musical idiom (here, 18th century Classical style). Listeners may either fail to distinguish between distinctive and style related features or simply do not care to do so, and thus base their evaluation of musical unity primarily on the latterin the present case, on features of the Classical style present in the hybrids as well as in the originals. Furthermore, given the constraints of musical form, genre, and style, unifying features or relationships, including those distinctive of the composition in question, may accidentally arise even in a random assemblage of fragments of the same musical style. Indeed, striking unifying relationships of that sort can be observed in our hybrids. For instance, in the hybrid provided to Group 1, whose exposition combines the first key area group of K. 280/I with the second key area of K. 332/I, the opening theme, from K. 280 (Figure 8a), shares with the secondary theme, taken from K. 332, (Figure 8b, mm. 41-42) a remarkable combination of features: the exact opening rhythm, the melodic contour (a triple repetition of the opening note, followed by a descent), as well as the ascending arpeggiation, which initiates both themes. The same rhythm and melodic profile are also featured in the next thematic group, also from K. 332 (see Figure 8c, mm. 71-72). In the hands of a competent advocate of organic inner

form (and perhaps in the mind of similarly inclined listeners as well), such relationships would surely provide triumphant evidence for the supreme unity of Mozarts compositionand of Mozarts greatness.
Inner Unity and Perceived Coherence

FIGURE 8. Derivation of two themes of K. 332/I from the opening of


K. 280/I.

As Table 9 reveals, following repeated exposure, participants in Group 2 associated the evaluation of the hybrid vs. the original with ratings of coherence. However, most participants also preferred the hybrid over Mozarts original. We have suggested above some possible sources for this seeming incongruityan inability to distinguish distinctive features from style related ones, or the coincidental emergence of unifying features in the hybrids. Yet another possible basis for this unsettling disparity may derive from the listeners diverse conceptions of what musical coherence is. As Levinson or Gurney would suggest, the participants sense of incoherence may be primarily based on the presence of local contrasts between adjacent events, rather than on global or remote relationships. Moreover, as several empirical studies have indicated, listeners judgments of thematic similarity often rely on musical dimensions such as dynamics, pitch register, and texture, rather than on the dimensions that most music theorists view as the bases of musical unity, such as tonal relationships and pitch intervals (Eitan & Granot, 2007; Lamont & Dibben, 2001; Ziv & Eitan, 2007). Simple auditory cues, not specific to music, thus play a central role in similarity perception and categorization in music, even in those musical styles (such as the Classical style) where, according to established music theories, they are supposed to be almost irrelevant to such tasks. As a result, listeners perception of thematic structure may be at odds with music theorists notions and composers strategies, particularly since thematic variants, as described by music theorists, often exhibit considerable contrasts with their thematic source in such surface features, while maintaining underlying pitch or rhythmic structure. Several of our listeners free comments indeed suggest such disparity, as they cite contrasts in dynamics or register as the central source of incoherence in the music. Thus, while analyses of K. 280 (including Beach, 1994, and our analysis above) would point out the bass figure in m. 27 as a source of unity in the piece, as it stems from the initial arpeggiation (m. 1-2), several listeners pointed out the very same figure as a source of incoherence, due to the sharp local contrasts in pitch register and dynamics it introduces.

Inner Form and Aesthetic Judgment

415

Coherence vs. Expressiveness?

An even more radical approach to the aesthetic role of coherence is suggested by the correlation analysis for Group 1 (see Table 8). Here, after repeated exposure (Session 2), none of the three statements concerning coherence or unity significantly correlated with liking of the piece, or with agreement with the assertion that it is a masterpiece. In other words, it would seem that, by and large, the participants in this group did not view coherence or unity as important considerations in judging the hybrid versus the original (K. 332indeed, a hallmark of striking thematic contrasts in the Classical style). This may, of course, reflect an aesthetic preference typical of the post-modern age, very different from aesthetic considerations of the Classical epoch. Several of the participants free texts (in both groups) shed some light upon these aesthetic considerations. Following are a few examples (emphases added by the authors):7 In Mozarts time, richness and strong impression were preferred over coherency and orderly structure. Hence, it is reasonable that [the original], in which playing is more powerful and sections more heterogeneous was chosen by Mozart. (Group 1) [I prefer the original since it] is more interesting. It is as if it has subdivisions into scenes and one can actually build around it an interesting story with events which develop, though at the beginning it sounds less coherent than the second version. (Group 1) Mozart chose [the hybrid] since [the original] exhausts the musical ideas too much. (Group 1) I preferred [the hybrid], since it has more musical ideas: it is more varied, has many themes, and it is more surprising and exciting. (Group 2) Mozart was not that imbecile (sic), he wouldnt choose the [original], since the 2nd subject is utterly stupid (sic), and is actually a copy of the 1st subject. (Group 2) It would thus seem that, for some listeners, coherence, orderly structure, exhausting the musical material, and unifying similarity of different musical ideasprinciples of organicist musical aestheticsseem to stand in contrast to richness, expressiveness, and even interest. Quite a few of the participants evaluated the latter over the former, or even assumed that such evaluation was the standard of Mozarts or of Classical aesthetics.

Thus, even the very attributes that to Rti or Schoenberg would seem to offer evidence of sublime inner unity, such as a derivation of a secondary subject from the principal one, or exhaustion of the musical material, served for some participants as proofs of lesser musical value, even musical imbecility, since they supposedly work against the demands of richness and interest. Even when inner unity was perceived, therefore, it may have been considered an aesthetic weakness rather than an asseta result that in itself may attest more to our own Zeitgeist than to cognitive strategies or abilities.
The Work and Its Performance

While structural aspects such as coherence, derivation from initial material, and the degree of contrast only occasionally correlated with general judgments and preferences, ratings of different aspects of the pieces performance (overall performance assessment, assessment of tempo selection, and appropriateness of expressivity in performance) correlated strongly and consistently with preferences for the piece itself in both experiments. Notably, such correspondence was found not only for nonmusicians (where the notion of the performanceindependent musical work may not be clear and stable) but also in the group of musically trained participants, even following repeated listenings. This suggests that the notion of an autonomous musical work, independent of any specific performance, may not be reflected in listeners actual evaluation of music. Rather, listeners (perhaps since many of them are accustomed to musical genres where a distinction between composition and performance is not clear cut) may freely mix evaluations of composition and performance without making any clear distinction between the two. Paradoxically, this implication of the results is particularly noteworthy since in fact all the music in the experiment was recorded by the same performer (M. Uchida, noted for a distinct style of performance), in the same recording session. Nevertheless, listeners seem to have attributed perceived differences between the two versions to their performance (though they were in fact identical), rather than their structure, and this perceived difference in performance quality strongly affected the overall evaluations of the compositions.
Open Questions and Suggested Follow-Ups

7 In these quotes from the free texts we have replaced the terms first version or second version with [original] and [hybrid].

This study calls for further research not only due to its rather unsettling results but also due to the nature of the musical materials it used. The need for additional research arises from characteristics of the Classical style, of Mozarts specific idiom, and of the particular sonatas

416

Zohar Eitan & Roni Y. Granot

used in the experiments, all of which may restrict the import of the present study. As mentioned in the introduction, the constraints of tonality and form, which were patently defined and understood within the Classical style, allow for considerable freedom in combining specific musical materials, provided such constraints are followed. Indeed, our experiment may be viewed as a latter day version of the ars combinatorial dice games, developed by composers and theorists such as Mozart, Haydn, and Kirnberger, in which the dice selected each structural unit from a number of choices, analogous in their harmonic and structural functions (Ratner, 1970). Furthermore, Mozartperhaps more than Haydn and Beethovenfrequently uses dramatic juxtaposition and extreme contrasts as a compositional principle, such that Mozartian themes and forms appear with a certain accidental quality, as if something different could sound in their place, as if they were exchangeable (Holtmeier, 2002, p. 307). As mentioned, the piece often seen as a paradigm of such dramatic juxtapositions is actually the opening movement of K. 332 itself. These characteristics of the Classical style, Mozarts in particular, made the choice of Mozarts music for this experiment almost inevitable. However, they also suggest that examining whether and how inner form is perceived in other musical styles and epochs is clearly called for. Finally, a tentative yet intriguing aspect of the results, suggesting further examination, should be noted: the different weighting of evaluative criteria in the two experiments. Most importantly, in Group 2 (K. 280 vs. hybrid 2) evaluations of the pieces coherence correlated strongly with overall judgments (this version is a masterpiece,I like this version), but in Group 1 (K. 332 vs. hybrid) such correlation was hardly found, and instead other factors (such as appropriate duration) were prominent (see Tables 8 & 9). This diversity may suggest that even within the same style and genre, listeners apply an ad hoc weighting of the various aesthetic criteria available to them (perhaps guided by the conspicuous facets of each piece), rather than using a stable set of aesthetic criteria. Such a conclusion is of course premature; since it is based on comparing two different participant groups, the results may be due to inter-participant differences. It is, however, intriguing and deserves further study, examining within-participant correlation of different aesthetic criteria in the evaluation of diverse musical works.
A Tentative Conclusion

only. We disassembled two musical masterpieces so that sections that (according to that credo) should associate through distinct, unifying relationships were severed from each other; these sections were then reassembled into newly composed musical hybrids, which do not possess (as exemplified in our analytic interlude) these unifying relationships. That cruel operation did not, however, affect listeners evaluation of the musical pieces; rather, many of them, including musically trained ones, preferred our hybrids over the original masterpieces. Perhaps more important than the results themselves are the questions that they raise as to the role of largescale unity in aesthetic perception of music. As we suggested in the (admittedly speculative) discussion above, the results may stem from different grounds, or from their combination and interaction: an inability to perceive (or a disregard for) large-scale structure, as also suggested by previous studies; the emergence of accidental similarities and unities in the hybrids; the prevalence of notions of unity based on surface connections and contrasts, particularly in secondary parameters such as dynamics and timbre; and even disdain of unity and coherence themselves as aesthetic criteria. Investigating these possible sources of musical judgment empirically may further our understanding of how actual listeners aesthetic perception is related to, and in what ways it is distinct from, music theorists beliefs.
Author Note

We thank Richard Ashley, Lola Cuddy, Stephen McAdams, and two anonymous reviewers for Music Perception for helpful suggestions, David Steinberg and Ilana Galenter for assistance in statistical analysis, and Motti Adler and Golan Gur for help in data collection and design of music examples. This study was supported by an Israel Science Foundation Grant no. 800/02-27.0. Findings reported in this article were presented at the 9th International Conference on Music Perception & Cognition (ICMPC9), Bologna, Italy, August 2006, and reported at the conference proceedings (Eitan & Granot, 2006). Findings were also presented at the 3rd International Symposium on the Science of Musical Language (SLM3), Bologna, Italy, February 2006. The SLM3 proceedings are currently (June 2008) in press. Correspondence concerning this article should be addressed to Dr. Zohar Eitan, School of Music, Tel Aviv University, Tel Aviv, Israel 69978. E-MAIL: zeitan@ post.tau.ac.il, and Roni Y. Granot, Department of Musicology, The Hebrew University of Jerusalem, Jerusalem, Israel 91905. E-MAIL: rgranot@huji.013.net.il

This study has examined a major credo of generations of musicians and music aestheticians, and its results suggest that, insofar as listeners cognition is concerned, this credo may be based on deep (perhaps admirable) faith

Inner Form and Aesthetic Judgment

417

References
A LMN , B. (2003). Narrative archetypes: A critique, theory, and method of narrative analysis. Journal of Music Theory, 47, 1-40. B AKER , J. M. (2003). Chromaticism, form, and expression in Haydns Quartet Op. 76, No. 6. Journal of Music Theory, 47, 41-102. B EACH , D. (1994). The initial movements of Mozarts piano sonatas K. 280 and K. 332: Some striking similarities. Intgral, 8, 125-146. B ENJAMINI , Y., & H OCHBERG , Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B, 57, 289-300. B ENT, I. (1987). Analysis (Norton/Grove handbooks in music). New York: Norton. B ERLYNE , D. E. (1971). Aesthetics and psychobiology. New York: Appleton-Century-Crofts. B ONDS , M. E. (1991) Wordless rhetoric: Musical form and the metaphor of the oration. Cambridge, MA: Harvard University Press. B RACE , N., K EMP, R., & S NELGAR , R. (2006). SPSS for psychologists: A guide to data analysis using SPSS for Windows. London: Routledge. B URKHART, C. (1978). Schenkers motivic parallelisms. Journal of Music Theory, 2, 145-175. C ARPENTER , P. (1983). Grundgestalt as tonal function. Music Theory Spectrum, 5, 15-38. C OOK , N. (1987). The perception of large-scale tonal closure. Music Perception, 5, 197-205. E ITAN , Z., & G RANOT, R. Y. (2006). Growing oranges on Mozarts apple tree: Inner form and aesthetic judgment. In M. Baroni, A. R. Addessi, R. Caterina, & M. Costa (Eds.). Proceedings of the 9th International Conference on Music Perception & Cognition (pp. 1020-1027). Bologna: Alma Mater Studiorum Universita Degli Studi Di Bologna. E ITAN , Z., & G RANOT, R. Y. (2007). Intensity changes and perceived similarity: Inter-parametric analogies. Musicae Scientiae, Discussion Forum 4A, 99-133. G ALEAZZI , F. (1998). Theoretical-practical elements of music (excerpts) (B. Churgin, Trans.). In W. O. Strunk & L. Treitler (Eds.). Source readings in music history (pp. 85-91). New York: Norton. (Original work published 1796) H OLTMEIER , L. (2000). Versuch ber Mozart. Juxtaposition und analytische Collage: KV 332. In W. Gruhn & H. Mller (Eds.), Wahrnemung und Begriff (pp. 109-74). Kassel: Gustav Bosse Verlag. H OLTMEIER , L. (2002). Reconstructing Mozart. Music Analysis, 21, 307-325. H URON , D. (2001). What is a musical feature? Fortes analysis of Brahmss Opus 51, No. 1, revisited. Music Theory Online, 7. Retrieved October 1, 2006 from http://www.societymusictheory.org/mto/issues/mto.01.7.4/mto.01.7.4.huron.html K ARNO, M., & KONENI , V. J. (1992). The effects of structural interventions in the first movement of Mozarts symphony in G minor, K.550, on aesthetic preference. Music Perception, 10, 6372. K ERMAN , J. (1980). How we got into analysis, and how to get out. Critical Inquiry, 7, 311-31. K RUMHANSL , C. L. (1996). A perceptual analysis of Mozarts Piano Sonata K. 282: Segmentation, tension, and musical ideas. Music Perception, 13, 401432. L AMONT, A., & D IBBEN , N. (2001). Motivic structure and the perception of similarity. Music Perception, 18, 245-274. L EVINSON , J. (1997). Music in the moment. Ithaca, NY: Cornell University Press. M ARVIN , E. W., & B RINKMAN , A. (1999). The effect of modulation and formal manipulation on perception of tonal closure. Music Perception, 16, 389-408. M AUS , F. E. (1999). Concepts of musical unity. In N. Cook & M. Everist (Eds). Rethinking music (pp. 171-192). Oxford: Oxford University Press. M EYER , L. B. (1973). Explaining music. Berkeley: University of California Press. M ORGAN , P. R. (2003). The concept of unity and musical analysis. Music Analysis, 22, 7-50. P OLLARD -G OTT, L. (1983). Emergence of thematic concepts in repeated listening to music. Cognitive Psychology, 15, 66-94. R ATNER , L. (1970). Ars combinatoria: Chance and choice in eighteenth-century music. In H. C. Robbins Landon (Ed.), Studies in eighteenth century music: A tribute to Karl Geiringer on his seventieth birthday (pp. 34363). Oxford: Allen & Unwin. R TI , R. (1951). The thematic process in music. New York: Da Capo. S CHENKER , H. (1979). Free composition (E. Oster, Trans.). New York: Longman. (Original work published 1935). S CHOENBERG , A. (1984). Folkloristic symphonies. In L. Stein (Ed.), Style and idea: Selected writings of Arnold Schoenberg (pp. 161-166). London: Faber. (Original work published 1947) S CHOENBERG , A. (1995). The musical idea and the logic, technique, and art of its presentation. P. Neff & P. Carpenter (Eds. & Trans.). New York: Columbia University Press. S CHWARTZ , J. L. (2001). Conceptions of musical unity in the 18th century. Journal of Musicology, 18, 56-75. S OLIE , R. A. (1980). The living work: Organicism and musical analysis. 19th-Century Music, 4, 147-156. S TREET, A. (1989). Superior myths, dogmatic allegories: The resistance to musical unity. Music Analysis, 8, 77-124.

418

Zohar Eitan & Roni Y. Granot

TAN , S., & S PACKMAN , M. P. (2005). Listeners judgments of the musical unity of structurally altered and intact musical compositions. Psychology of Music, 33, 133-153. TAN , S., S PACKMAN , M. P., & P EASLEE , C. L. (2006). The effects of repeated exposure on liking and judgment of musical unity of intact and patchwork compositions. Music Perception, 23, 407-421. T EMPERLEY, D. (2003). End-accented phrases: An analytical exploration. Journal of Music Theory, 47, 125-154. T ILLMANN , B., & B IGAND, E. (1996) Does formal structure affect perception of musical expressiveness? Psychology of Music, 24, 317. T ILLMANN , B., & B IGAND, E. (2004). The relative importance of local and global structures in music perception. Journal of Aesthetics and Art Criticism, 62, 211-222.

T ILLMAN , B., B IGAND, E., & M ADURELL , F. (1998). Local versus global processing of harmonic cadences in the solution of musical puzzles. Psychological Research, 61, 157-174. T HOMPSON , W. F., & C UDDY, L. L. (1992). Perceived key movement in four-voice harmony and single voices. Music Perception, 9, 427438. W HITNEY, K. (2003). Schoenbergs single seconds of maximum spiritual excitement: Compression and expansion in Erwartung, op. 17. Journal of Music Theory, 47, 155-214. Z IV, N., & E ITAN , Z. (2007). Perceiving thematic structure: Similarity judgments versus categorization tasks. Musicae Scientiae, Discussion Forum 4A, 39-75.

You might also like