The Effect Of Decreased Chaos In Melody Generation On Its Conformity To Common Chord Progressions

The purpose of this project and experiment was to determine the effect of increasing and decreasing randomness on the conformity of a randomly generated melody to known chord progressions. This project was developed in Jython, a java-based implementation of Python created to be cross-platform compatible. It was hypothesized that decreasing randomness in the computer-generation of melodies would lead to increased conformity to chord progressions. The rationale behind this hypothesis was the fact that less chaos implies more order, meaning that these melodies would tend to stick to patterns which are known to be pleasing to human ears, and thus, to pleasing and therefore known chord progressions. The project’s procedure involved only the coding of the program, and the sampling of data. The code will be provided in Appendix A. The sampling of data was quite straightforward, the program was run at each ‘independent variable value’, which corresponds to a set of Boolean parameters in the code, and the value returned by the program as the chord conformity score was taken. The score is an average score produced by the score for each note interval in the melody divided by the total number of intervals per melody. For each step up, randomness decreases. Results showed inconclusive numbers, with the average values for each IV as follows: 10.406 for the lowest, 9.9895, 10.0143, 10.5066, 10.363 for the highest. A one-way ANOVA with alpha = 0.05 returned a p-value of 0.002265, which is sufficient to reject the null hypothesis. A post-hoc Tukey test confirmed that of all group combinations, only RemainOctave vs StartOnTonic, RemainOctave vs StepWise, and MiddleOctave vs StartOnTonic were significantly different at alpha = 0.05. This suggests that while there may not be an easily noticeable difference in the chord conformity, certain levels of randomness did account for a significantly different amount of chord conformity