Linguistics Anonymous

20 April 2006

A Lateral Kick to the Shins: Part 7

I promised more data, so here it is. The figure to the left shows FFT spectra for the fricatives of Western Apache, a language that has both of the sounds involved in the Akkadian sound change I've been discussing (from Gordon et al., "Phonetic Structures of Western Apache," International Journal of American Linguistics, 2001). As you can see from the labels, the fricatives were pronounced by male speakers before [a], so these examples, like those from Kabardian, are not perfect matches for the environment of the Akkadian change (in medial clusters before coronal obstruents). However, it is well known that consonants tend to be reduced in various ways in clusters, so demonstrating similarities between consonants in highly salient contexts such as syllable-initially should be sufficient to suggest that they could be confused when reduced in clusters.

Note that the two segments (the top two squares in the diagram) are very similar. They both have spectral peaks around 3000 Hz, although the peak for [š] is slightly higher in both frequency and amplitude. The main difference is that [š] has lower intensity in the low-frequency area, although the energy distribution is still remarkably similar, with a steady decline up to around 1000 Hz followed by a fairly sharp rise to the spectral peak. The two are particularly similar when compared to the other two fricatives in the language, whose spectra look very different, with different energy distributions in different frequency ranges. I think this evidence supports my contention that the two sounds were similar enough acoustically to be confused by listeners when heard in medial clusters.

The main issue remaining is also the most baffling: why did this change only happen before coronals? After all, /l/ could easily have been devoiced before all voiceless obstruents, and the same perceptual confusion I have posited here could have occurred in those environments. So why was there no sound change there? Why was it limited to just this narrow context? And why did it spread so far within that context without spreading further?

I don't have answers to these questions, important though they are. I only have suggestions, which could explored in further research. The main idea I have, in keeping with the acoustic explanation I have given in these posts for the sound change as a whole, is that there was something about the transitions going in to the coronal segments that made a perceptual confusion more likely. Transitions are very important for perception, so I think this is most probably the direction to go to figure out why the change was so restricted. One way to investigate this, which I didn't do for this project due to time constraints, would be to find native speakers of a language with these segments (for best results, one in which /l/ was phonologically devoiced in comparable contexts) and record tokens with medial clusters like those of Akkadian. Examining those recordings would go a long way toward finding a solution to the issues mentioned above; any particular acoustic features of the clusters could be tested for salience through perception experiments or other techniques of experimental phonetics. My presentation is on Saturday, though, so that will have to wait for another time.


Post a Comment

<< Home