Identification of transposed and distorted melodic fragments
Pierre L. Divenyi, Ira J. Hirsh · The Journal of the Acoustical Society of America · 1976
Listeners were trained to identify one of six possible orders of three tones, equal in duration (mostly 18 msec), at frequencies spaced equally (log) around 1 kHz (f3/f2=f2/f1= ;9/8). When mixtures of these patterns included also transpositions (with same frequency ratios) to as many as 13 different frequency regions, pattern identifiability remained unchanged. When four “background” tones, randomly selected from a 2/3-octave region around 1 kHz, preceded or followed the three-tone pattern [see also Divenyi and Hirsh, J. Acoust. Soc. Am. 57, S10(A) (1975)], identification was impaired only for patterns in the 1-kHz region. Performance improved as pattern frequencies were transposed to regions more remote from the background. Further experiments on isolated patterns increased the pattern “vocabulary” by changing the frequency ratios f2/f1=f3/f2. Pattern identification was best when the span between f3 and f1 was between 1/4 and 1/2 octave and was very difficult for spans less than 1/6 or more than 1 octave. The relations between performance and span for both common (Western music) and uncommon intervals were identical. When frequency ratios were changed asymmetrically (f3/f2≠f2/f1) the degree of asymmetry affected performance when the frequency span was smaller than 1 octave. [Supported by Grant NS03856 from NINCDS to C.I.D.]