The studies by Ladd and by Inkelas and Leben make an important contribution to the vigorous, continuing debate on foundation issues in intonation theory. They are both concerned with establishing the basic formal mechanisms used to describe intonation, understood here (in a narrower sense than the usual one) as the way in which tonal and accentual patterns are mapped into phonetic F0 contours. What is central to these proposals is the significance of the role played by register. Both papers view register as a fundamental component of a linguistic description of intonation, and propose to characterize it in terms of separate intonational structures built over the tone melody (a sequence of pitch accents or tones), to which rules assigning F0 have access.

There are broad areas of agreement in these papers. Their common strategy is to shift some of the burden of accounting for intonational regularities from phonetic implementation rules to phonological structures defining global intonation patterns. Both papers argue that such an approach can capture generalizations that would otherwise go unexpressed, and allows the phonetic implementation rules to be simplified or constrained in desirable ways. To achieve these results Ladd introduces metrical tree structures, while Inkelas and Leben propose to recognize a new tier of autosegmental representation which they term the “register tier.”

These remarks will focus primarily on Ladd's and Inkelas and Leben's views on how register is to be formally characterized in intonation theory. I will suggest that while we do want a way of expressing register as an independent parameter of pitch assignment rules, we may not require the full power of metrical and autosegmental structures to obtain an adequate account of register shift phenomena.