Frequency Resolution of Absolute Pitch Identification

Shin, Eunyoung 1 ;  Wong, Willy 1, 2

1. Edward S. Rogers Sr. Department of Electrical & Computer Engineering, University of Toronto; 2. Institute of Biomaterials & Biomedical Engineering, University of Toronto

Autism and autism spectrum disorder (ASD) are complex disorders of brain development.  Past studies have linked prevalence of absolute pitch (AP) within individuals with autism.  AP refers to the ability to identify frequencies of musical notes to an accuracy, far exceeding the general population. However, there has been little study on the limits of AP possessor’s ability to resolve pitch frequency.  A number of issues come up when trying to test absolute pitch beyond the standard musical frequencies, one of the most important being how to deal with differences in individual pitch preference.  In this study we are focused mainly on ways to better quantifying limits of pitch perception for future use in testing autistic individuals.

Our wish is to test pitch identification at resolutions beyond the usual musical notes.  One complication is how to provide meaningful labels of sub-notes, particularly when different subjects will not necessarily agree on tuning.  Hence in our study, we perform absolute identification of pitch using note frequencies defined by the subjects themselves.  The test will involves two stages.  The first is to determine which individuals have absolute pitch.  Both AP and non-AP subjects listen to pure tones via a headphone and then provide a response on a piano keyboard GUI to indicate which note was played. Those who scored above a threshold (i.e. with very few mistakes) were deemed to be AP possessors.  In the second experiment, subjects were asked to tune a one octave scale beginning at middle C (261.6 Hz).  They were given an opportunity to tune once and a limited number of tries to adjust their choices afterward.  We then divided the frequencies into quarter tones and then tested their accuracy in an identification experiment.  The tones (1 s duration) could be listened to multiple times and there was no feedback provided.

In our limited testing, no subject was found to be an AP possessor who did not already know they had AP.  Those who self-reported as AP possessor had an average accuracy of 96% in the first experiment.  In the second experiment, we found AP possessors (n = 10) to be able to identify quarter tones with accuracy exceeding 91%.  There were no systematic errors in identification found in our results.  The results of the experiment also did not seem to be affected by the use of pure tones.  AP possessors demonstrated consistency in their own definition of pitch as subjects tuned note frequencies with standard deviation less than 3 Hz.  However, as was expected, results differed significantly between subjects with an average deviation of 5 Hz, and as large as 12 Hz.

This study shows that the limits of frequency resolution for AP possessors are much greater than the conventional boundaries defined by the western musical scale. We also found tuning is consistent within a single individual, but not across different subjects. Next we wish to apply this tool in the testing of autistic individuals.