How do phonologists use the terms "phonological phrase" and "intonational phrase"?

Question

I would like to get a handle on the difference between the concepts of phonological phrase and intonational phrase, as used in contemporary phonological theory. How do phonologists define these two terms and differentiate between them? What would be a clear example of the distinction? Are there agreed-upon criteria for what constitutes a phonological phrase or an intonational phrase? How much currency do these have as terms of art -- that is, regardless of theoretical orientation, if one phonologist talks to another about one of these two things, can it be expected that they will both be talking about the same thing or are there substantial disagreements about definition? Is there a view that these terms are not meaningful or necessary, and if so how common is it?

Answer 1

It sometimes appear that the two concepts are interchangeable in actual usage, but I think that is a consequence of indeterminacy where you have evidence for grouping, but you don't have clear evidence for what level that grouping is. The standard answer is that an Intonational Phrase is a superior level in the prosodic hierarchy, generally composed by one or more Phonological Phrases. See Nespor & Vogel (1986) Prosodic phonology for the basic theory, and various examples.

A P-Phrase is composed of some number of P-Words (or, Clitic Groups though that level of analysis has fallen by the wayside). The big question is, what exact rule defines that grouping of "some number"? Analogously, an I-Phrase is composed of some number of P-Phrases (with the same big question about how they are built). P-Phrases seem to built in two main ways, centered around a head word of a syntactic constituent – adding one complements in that phrase, or all complements – "one complement" being the most common option, with "all" being a less-common option. The latter basically reconstructs Xmax; the former might reconstruct X'. However, there are also a lot of quirky options, like "and all following words in the S". The basic motivation for the PP is finding contrastive phonological behaviors of phrases like A B C, where for instance [B C] interact and A B don't in case the structure is Subject Verb Object (V+O = VP), versus where [A B] interact and B C don't in case the structure is N Adj V (N+Adj = NP).

The IP is more used in studies of intonational phonetics, and relies on things like "intonational breaks" and the number of intonational contours for its justification. Overall, there isn't much by way of crisp grammatical generalizations involved in grouping sequences of phrases into IPs, instead IP phrasing seems to be quite fluid and pragmatically governed, unlike PPs which more or less reconstruct standard syntactic structures.

A good study that illustrates analysis in terms of these elements is Kanerva's 1990 Stanford dissertation Focus and phrasing in Chicheŵa phonology. Certain phonological processes involve lexical heads and its modifiers, for instance Pre-H Doubling. (I should point out that Kanerva unfortunately diverges from standard terminology in labeling this the Focus Phrase, since one of the factors governing where this constituent ends is whether the word is focused -- a focused word ends the PP, where it wouldn't if the were s not focused. He has some discussion of FP vs PP at the end of ch. 4 which is, shall we say, inconclusive). Then, the IP is motivated by looking at intonational phenomena, where it is shown that intonational overlays span sequences of PP (FP) phrases, but utterances can have multiple IPs.

There are various problems in this model having to do with higher-level syntactic units: for example, nothing reconstructs "S", or conjunctions of phrasal constituents (NP & NP). The IP can be pressed into service to handle such cases, but there are now alternatives such as allowing recursion in the PP. Since phonologists don't generally deal with intonation, the IP is not much used.

This paper by Selkirk gives a general state of the art account of phrasal relations, with some of her extensions of the original constrained Nespor & Vogel model.

As far as definitions of IP and PP are concerned, there is the fairly unhelpful purely formal statement that IP is immediately dominated by U (utterance) and immediately dominates PP, and PP immediately dominates CG (if there is such a thing) or W. There aren't any particular properties of PP or IP that are specific to that node, so if there is recursion (as is not assumed) then PP and IP are recursive; if rules refer only to Xmin, Xmax and X (neutral as to min or max) then that's a general property of all phrasal nodes.

Originally there were attempts to set out necessary and sufficient conditions for creating PP (little effort was put into IP), basically in order to express the generalization that {X, Y} interact iff X is the head of the phrase that Y is a modifier of, but since that is not the only way in which grouping is defined (just the most frequent, it seems), the "definition" i.e. nature of the rules for building PP and IP are up in the air. The only grouping fact that prosodic theories (still, apparently) insist on is the strict hierarchy of the groupings: e.g. a parsing {[A B] [C } {D] [E F]} would be illegal. I would say that there aren't any clear a priori definitions of IP and PP, instead there are probabilities that a grouping effect would be handled as PP in some conditions (basically, "if the domain looks like syntactic Xmax") and as IP in some other conditions (mostly having to do with the relevance of focus, variability in size, connection to overlays, and on the syntactic side grouping together a number of Xmax's)