Kohler's experiments concern the mapping between the acoustic/phonetic detail of spoken utterances and their underlying phonological structure. A core problem in describing spoken language is that some patterns that are almost identical are perceived as being quite different, while at the same time other patterns that differ immensely are nevertheless classified by listeners as being the same. Kohler addresses two specific manifestations of this problem in the relationship between F0 contours and the utterances that carry them.

Macro F0

His first experiment utilizes the minimal stress pair UMlagern and umLAGern (meaning roughly “to relocate” and “to surround,” respectively). When an F0 peak occurs between the first and second syllables – on the boundary between the /m/ and the / I / – then this can indicate either a “late peak” associated with UMlagern or an “early peak” associated with umLAGern. Hence two different underlying intonational categories, when combined with two different lexical stress locations, will result in apparently the same F0 pattern. Yet listeners are able to recognize the intended word. Kohler explores the question of how listeners recover the underlying forms (specifically where the primary lexical stress falls) from the combination of F0 with other characteristics of the acoustic signal; he uses linear predictive synthesis technology to produce versions of the utterances in which the F0 peak occurs earlier and later than the syllable boundary. The main result seems to me to challenge the widespread assumption that perception of F0 proceeds in parallel with and more or less independently of perception of the rest of the speech signal.