Modeling melodic perception: An examination of notational accuracy by
populations of differing musical experience
This model has been applied to data gathered from pretests and post-tests administered to several ear-training class groups ranging in experience from undergraduate music students to entering graduate students. Tests included a melodic dictation task as well as a harmonic dictation task which required the transcription of the top and bottom lines of a four-voice homophonic piece of music. Identical tasks appeared on the pretest and post-test. Between each pretest and post-test was a semester of formal classroom instruction as well as optional training aids such as audio and video instructional supplements and computer assisted instruction.
Preliminary analysis has shown significant levels of pretest to post-test improvement in notational accuracy. Application of the previously reported regression model to these data provides an indication as to whether the model can generalize across differing populations and differing musical tasks. Use of the regression technique also shows differences in perceptual profiles between different musical populations as well as across their semester of aural training.
Difficulty factors in the perception of melody by skilled listeners
The entire project includes three controlled experiments using advanced musicians as subjects. The main experiment, which will be the focus of this poster presentation, examined four melodic features, labelled "difficulty factors": contour (direction chances), interval difficulty, tonal ambiguity, and nonadjacent step relations. Thirty subjects were tested, via a specially designed HyperCard program, on their ability to encode aural stimuli consisting of 16 5-note pitch patterns. Patterns were selected from among hundreds of computer-generated patterns to reflect an even representation of the four difficulty factors and combinations of such factors within each pattern. Subjects were required to encode stimuli using two different encoding tasks: listing consecutive intervals and providing scale degree numbers or solfege syllables. Four different grading methods were employed in evaluating each subject's test results: number of pitch errors, degree of pitch error, number of interval errors, and degree of interval error.
The subject pool as a whole expressed no strong preference for any particular mode of response (interval or scale degree). Approximately one-third of the subject pool felt most comfortable responding with the scale response mode, one-third felt most comfortable with the interval response mode, and one-third felt that their response mode comfort was dependent on the particular pattern being encoded. Actual test results generally correlated with these findings. In comparing the four methods of grading, there were patterns that obtained pitch and interval grades that were quite similar, and there were other patterns that produced quite contrasting pitch and interval grades. These results suggest that in pedagogical situations, different grading methods may result in marked differences in grades, depending on the pitch material being evaluated
Expert expectations: professional theorists' Continuations compared with a computer model of musical forces
My presentation to the 1996 SMT meeting will report the results of that experiment. Briefly, analysis shows that participants' continuations agreed note-for-note often enough to claim that they are governed by some underlying mechanisms. Many theorists and psychologists (e.g., Narmour, Lake) have suggested elaborate theories to account for such perceived regularities in listener expectations. But my work offers a much simpler theory that not only explains those regularities, but also, when implemented as a computer model, will generate them. In fact, the computer model matches the participants' responses note-for-note as often as most of the participants do. The theory, described in a series of recent publications, claims that listeners of tonal music hear melodic motions as purposeful action within a dynamic field of musical forces -- which I call "gravity, magnetism, and inertia" -- and that melodic expectations are based on the operation of those forces. The distribution of responses not only offers striking support for the theory underlying the computer model, but also illuminates important theoretical and methodological issues in music-cognition research.
The Binary Bias of metric subdivision and the relative complexity of various meters, or, why is 9/8 so rare?
This research is part of a larger international study being conducted at the Universities of Keele and Sheffield (United Kingdom) on the relationship of fingering to musical expression and the cognitive bases of fingering strategies. Using the Beethoven Sonatina, Wo O 50 (bars 1-16) and the Chopin Nocturne, Op. 48, no.2 (bars 1-28) as case studies in flowing right hand melody, this poster draws on empirical data retrieved from subjects' performances of three alternative fingerings of the above excerpts as videotaped and recorded on a Yamaha Disklavier.
Pianists of various levels of expertise (novice to artist), were asked to perform these excerpts with three different fingerings, two of which were provided. In the Beethoven sonatina, one of the fingerings proposed was the composer's. In the Chopin excerpt, the two alternative fingerings were constructed so as to examine whether fingering could contribute to the articulation of phrase structure. The third fingering consists of subjects' own preferred fingering for the passage. Subjects were asked to perform each excerpt three separate times, using each of the three fingerings (the two designated fingerings and their own fingering) at least once.
This study examines expressive timing profiles associated with each set of fingering. Perceptual data will also be examined to determine whether listeners can differentiate between the different fingerings.
Aural cognition skills of university level trumpet students: a comparative analysis
of melodic recognition skills from varying timbral sources
A pre-recorded test was presented was presented to the subjects that contained ten short melodies consisting of intervals common to the overtone series of the trumpet. Five of the melodies were recorded on trumpet and five were recorded on piano. Each melody was presented twice. The students were given a single sheet of manuscript which contained pairs of melodies numbered 1 through 10. The test was presented in the form of an error detection exercise. One version of the written melody represented the recording exactly while the other version contained slight variations of certain intervals within the melody. The students were instructed to signify the version of the melody as it was presented on the recording.
The data retrieved from this study suggests that the trumpet players tested, whether consciously or sub-consciously, reference their musical experiences as brass players during ear-training: specifically, use of the overtone series as a tonal reference structure. As a group, the students more accurately identified melodies performed on the trumpet than melodies performed on the piano.