Repertory Grid method

These twenty four episodes are the focus of the investigation and they are the objects (elements) upon which the RepGrid performance is based, (Tan & Hunter, 2002). The element’s comparison which represents the element’s interpretation by the analyst leads to the elicitation of constructs, (Tan & Hunter, 2002). Then the differences and similarities between elements and constructs are revealed from links, (Tan & Hunter, 2002). In the present study the linkage is a 7-point scale which was used to rate all elements above all constructs. This provided enough space for discrimination without creating problems in the visual examination of the grid, (Tan & Hunter, 2002).

In other words using the “Repertory Grid method” based on the “personal construct theory”

by George Kelly (1995) the elements (episodes) are compared and certain bi-polar constructs are elicited (Aldridge & Aldridge, 1996, p. 226). Then, after “positioning elements along construct scales” a matrix is generated and is called a repertory grid (Abrams, 2007, p. 94-95).

The matrix formation presents inter- relations among all elements and constructs depending on the relative levels of difference and similarity, (Abrams, 2007).

A common technique for the constructs’ elicitation is “the standardized form of triad elicitation”, (Aldridge & Aldridge, 1996, p. 228). Three randomly selected episodes were compared according to “two that are similar and one that is different” (Aldridge & Aldridge, 1996, p. 228). For example, I compared the following episodes: Episode 19 to Episode 15 and Episode 7. Episodes 15 and 7 shared a similarity which simultaneously differentiated them from Episode 19. From this comparison two opposite poles of a construct were elicited; ‘tonal center’ versus ‘no tonal center’ (see Table 4).

TABLE 4 Eliciting the poles of constructs

DIFFERENCE NO TONAL CENTER E 19

E 7 E15

SIMILARITY TONAL CENTER

In the same manner all the remaining episodes were compared and twenty two constructs were elicited (see Table 5).

Afterwards, all the episodes (melodies) were set in a position between each construct pole.

The episodes (melodies) were ranked in connection with the bi-polar constructs.

Although Aldridge proposes “an ordering of relative position” for the ranking of episodes (melodies), I used a seven-point scale instead for the following reasons: a 7 point scale enables testing “whether the elements really are in the range of convenience of all the constructs-and thus if the grid has been constructed correctly” (Easterby-Smith, 1980, p. 10).

On the other hand, ranking may have an impact on the differentiation between elements which actually share no difference (Tan & Hunter, 2002). However, it is important to mention that the scale’s points in relation to the construct poles are supplied “as reference points only, and are not intrinsically related to the meaning of the polar attributes”, (Abrams, 2007, p. 98).

Additionally, I considered ranking an impractical solution because of the large size of the data.

Although I was quite familiar with the musical material, and I had memorized the 24 melodies, instead of rating every melody separately in chronological order I followed the technique proposed by Aldridge (1996): I rated first every pair of melodies -from the triads used for the elicitation of constructs- which shared a similarity and afterwards the remaining melodies of the triads.

TABLE 5 The 22 constructs elicited from the 24 episodes

Tonal center No tonal center

Staccato Pedal

Free tempo (a piaccere) Strict tempo

Pulse present Pulse absent

Playing in narrow range of pitch Playing in wide range of pitch

High density of notes Low density of notes

Precise pulsation Imprecise pulsation

Linear melodic development Vertical melodic development

One melodic phrase Variations of melodic motifs

Musically supported Solo

Low register High register

Rhythmic patterns No rhythmic patterns

Initiative after a musical pause Induced from previous music

Dialogue Monologue

With accompaniment (two mallets) Without accompaniment (one mallet)

Tempo accelerates No deviation in tempo

Small intervals Large intervals

No finale Clear finale

Underlying harmony No underlying harmony

Development of pitch contour Fixed pitch contour

Chromatic pitch contour Diatonic pitch contour

Fluctuation in dynamics Flat dynamics

As the elicited constructs were mainly musical qualities (i.e. “free tempo” versus “strict tempo”) and not metaphors describing the music, the technique above enabled me to rate the melodies according to one another, thus staying close to the idiomorphic musical nature of the data.

On account of the large size of musical data (24 elements and 22 constructs) a computational analysis was necessary as the emerging grid was also large and there was limitation of time (Easterby-Smith, 1980). I utilized the ‘OpenRepGrid on air’

(http://www.onair.openrepgrid.org/) which is an online free software suitable for conducting this analysis. I used it to generate the ‘Bertin display’ and the dendrograms of clusters and elements (see graphic 1 and 2).

GRAPHIC 1 AU/BP values of clustered Elements

In order to conduct cluster analysis I downloaded and installed the pvclust package which is

“included in CRAN packages”, (Suzuki, & Shimodaira, p. 1540). CRAN is “the official R package archive”, whereas R is “a free software environment for statistical computing and graphics”, (Suzuki, & Shimodaira, p. 1540).

The RepGrid construction “is a whole that transcends the sum of its individual parts (i.e. the elements and constructs) in both scope and depth, and as such can help provide important insights about a given phenomenon beyond what the parts alone may reveal” (Abrams, 2007, p. 94-95). The emerging grid shows patterns in the data, connecting certain elements and constructs together according to the percentage of their shared similarity.

This process of constructing a RepGrid, may reveal musical characteristics which seem

“unrelated, disparate, or paradoxically opposed on their surface as closely aligned in the scheme of the construction” and vice versa (Abrams, 2007, p. 95). Even though certain distortions may happen during the construction an overall distortion of the grid is improbable (Abrams, 2007).

Therefore, the RepGrid structure can be a reliable source for acquiring valuable information about an improvisation (Abrams, 2007).

GRAPHIC 2 AU/BP values of clustered constructs

5 CASE OF TINA