| dc.description | In the present work, two eight-stage experiments, using a new on-line assessment, were conducted to assess how listeners parse melody. In Part 1 of Stages 2, 4, 6 and 8 (Stage 2 served as practice), subjects listened to a melody and pressed a key any time they detected the end of one unit (or the start of a new unit). In each stage, subjects heard the same melody three times and produced a boundary profile for each repetition; Stage 4 used a familiar nursery rhyme, Stages 6 and 8 used the same unfamiliar tonal piece. To model the data, Group Preference Rules 2 and 3 of Lerdahl and Jackendoff (1983) were quantified and then compared to the empirically determined boundaries. Results indicated that Rules 2b (Attack-Point), 3a (Register) and 3d (Length) had predictive validity. Rule 2b was the most important in both the familiar and unfamiliar melodies. The theory of Lerdahl and Jackendoff was extended to Rule 4 (Intensification) and Rule 6 (Parallelism). Rule 4 was the optimal combination of Rules 2 and 3. A new method of analysis, based on asymmetric error terms and non-linear regression, was developed to properly assess the contributions of the different rules. Rule 4 demonstrated that only Rule 2b was necessary to predict boundaries; Rule 3a and 3d played a much less important role. Rule 6, Parallelism. was quantified as two forms of Pitch Pattern Parallelism (pitch pattern including duration and pitch pattern ignoring duration) and two forms of Time Pattern Parallelism (time pattern for the veridical timing of breaks between notes and time pattern for the encoded representation of durations). Finally, Stages 1, 3, 5 and 7 assessed the sensitivity of subject to tonality using a modified probe-tone task. Several detailed analysis relating the tonality profile to the boundary profile. and to the boundary efficacy tasks, indicated that the tonality profile was only minimally related to boundary profiles or the boundary efficacy tasks. Overall, the present work indicated the effectiveness of the new on-line boundary assessment task. The present work also revealed that the model of Lerdahl and Jackendoff's could partially predict melody parsing. However, the model of Lerdahl and Jackendoff could be refined and extended, particularly in the implementations of Rules 4 and 6, possibly using the extensions that were developed in this work. In addition, boundary efficacy tests need to be refined if they are to be used within naturalist studies. (Abstract shortened by UMI.) | en_US |
