Much of our understanding of linguistic structure was necessarily first built up on the basis of isolated sentences, either constructed by analysts or speakers consulting their intuitions, or from written documents. The availability of corpora of unscripted speech is now allowing us to see what speakers actually do, opening up vast new areas for exploration. Some of the advantages described here are the access to prosody and context. The question addressed is how closely traditional syntactic structures, particularly constituent structures, are matched by prosodic structures. Points are illustrated with corpus material from Mohawk, a language indigenous to the North American Northeast.
Muito do nosso entendimento sobre a estrutura linguística foi construído inicialmente a partir de frases isoladas, propostas por especialistas ou por falantes, tendo por base suas intuições, ou a partir de documentos escritos. A disponibilidade de corpora de fala espontânea agora nos permite observar o que os falantes realmente fazem, abrindo novas áreas para investigação. Algumas das vantagens descritas aqui são o acesso à prosódia e ao contexto. A questão abordada neste artigo diz respeito a como as estruturas sintáticas tradicionais, particularmente as estruturas constituintes, são correspondidas por estruturas prosódicas. Os pontos são ilustrados com material de corpus do Mohawk, uma língua indígena do nordeste da América do Norte.
As is now generally recognized, speech is typically not produced in a continuous stream, but rather in spurts, termed variously intonation units, prosodic phrases, tone units, etc. Such units are generally identifiable by a convergence of cues. The cues used here are those set forth by Chafe in a series of works spanning the decades from the 1970’s through the first two decades of the 21st century, listed at the end of this paper. They fall into three main categories.
Mohawk is a member of the Iroquoian family indigenous to northeastern North America. There are six main Mohawk communities. Material cited here comes from a conversation between speakers from Kahnawà:ke, Quebec, Mrs. Josephine Kaieríthon Horne (JH), and Mrs. Charlotte Kaherákwas Bush Provencher (CB). They were discussing a funeral procession they had watched the day before, when the body of a lacrosse player was taken to his father’s house for a traditional ceremony.
Mohawk is considered a prototypical example of a polysynthetic language. Words often contain many morphemes. It is holophrastic in a narrow sense: verbs can constitute complete sentences in themselves, with an obligatory pronominal prefix identifying the core arguments, a verb stem, and an aspect suffix.
Morphological verbs can also function as referring expressions, and are often lexicalized as such.
Such words, tightly grammaticalized and often lexicalized expressions, package together elements of one concept: an event or state, complete with given participants, as in (2) and (3), or a referent, as in (4). If a significant new referent is to be introduced, additional words are used.
The prosody of single words is not manipulable for information structuring. Though one Mohawk verb may correspond to a multi-word sentence in languages like English, individual morphemes cannot be given extra prosodic salience. Pitch and rhythm are determined purely mechanically: stress is basically penultimate, with pitch contours and vowel length based on syllable structure, as described in Section 2.
It is easy to see matches between larger syntactic structures and prosodic structure. Consider the sentence in (5).
This sentence was bounded by pauses and showed a coherent pitch contour, with a high pitch reset at the beginning on
The intensity, shown in Figure 2, also shows a gradual fall to the end.
It is important to note that the cues distinguishing intonation units are not categorical. Pitch, intensity, and rhythm all vary along continua. Pauses vary in length and may not even be present at all. The three sentences in (6) were spoken with essentially no pauses separating them. Mrs. Bush had just noted that the newspaper was forbidden from publishing a description of the service.
Nevertheless, the distinct intonation units are easily identifiable from the pitch contour. Each began with a pitch reset on a stressed syllable. (Mrs. Bush, who made the second comment, generally spoke on a higher pitch than Mrs. Horne, who made the first and third comments.)
Basic intonation units usually show decreasing pitch, or declination, over successive stressed syllables. Mohawk stress is basically penultimate. (Epenthetic vowels are not counted.) Stressed syllables show several pitch contours. In closed stressed syllables, which are short, there is simply high pitch, as in
Declination over successive intonation units can be seen in the pitch trace in Figure 4. Stressed syllables are indicated with ▼.
Intonation units thus appear to match syntactic sentences.
Chafe distinguished three basic types of intonation units.
Examples of each can be seen in the passage in (8). Each line represents a separate intonation unit. Punctuation here reflects pitch contours: comma for a non-terminal contour, period for a terminal contour, dash for truncation.
The passage in (8) consists of a single syntactic sentence ‘They took him to his house and I guess he prayed according to their custom’. It was uttered in multiple intonation units, but there was an overall prosodic coherence, with a declination in pitch over the whole. Each successive intonation unit began with a small pitch reset rather than a full one, and the final unit ended in a terminal pitch contour.
In fact prosodic sentences consisting of multiple intonation units and ending in a final terminal contour are common. Example (9) was uttered in two intonation units, the first consisting of a conditional clause and the second the matrix clause.
The pitch trace is in Figure 6, with an initial high pitch reset, then a smaller pitch reset at the beginning of the following main clause, declination over the whole, and a final terminal contour.
A similar intonational structure can be seen in a sentence with an adverbial clause in (10).
This sentence began with an initial full pitch reset on the stressed syllable
Similar patterns can be seen with complement clauses. The sentence in (11) contains a complement clause within a complement clause, each in a separate intonation unit.
The sentence shows an overall coherent pitch contour with declination, small pauses and small pitch resets on each intonation unit within it.
A similar pattern can be seen with relative clauses. Sentence (11) above ended with a relative clause ‘that happened yesterday’, visible in the last unit on the pitch trace in Figure 8. The sentence in (12) ended with a relative clause ‘who came here’ modifying ‘children’.
Here, too, the relative clause was uttered as a separate intonation unit within the overall prosodic sentence, set off by an initial pause and small pitch reset.
Prosodic structures thus appear to match syntactic structures, with a separate intonation unit for each clause. Smaller constituents are also uttered in separate intonation units. In Figure (9) the bus was introduced in one intonation unit, followed by the verb ‘brought them’ in a second, and the children in a third. A similar pattern can be seen with the adverb ‘to his house’ in (13).
The adverb ‘to his house’ began after a pause with an initial smaller pitch reset.
A more complex construction is in (14), with a matrix clause followed by a complement clause which contains an adverbial clause which itself contains another adverb.
The declination over the whole sentence is clear, with smaller pitch resets on each intonation unit, in some cases separated by a pause. The stressed syllable
On the basis of these examples, it appears that even elaborate syntactic constituent structure is perfectly reflected in prosodic structure.
Further examination reveals that prosodic structure does not actually always match syntactic structure. The sentence in (15) was pronounced as a single intonation unit, with no pauses or pitch resets, though it contains multiple constituents.
The sentence in (16) contained an adverb in addition to the verb, but unlike the clause with an adverb in (13), it, too, was pronounced in a single intonation unit.
The sentence in (17) contained conjoined adverbial clauses.
Even this complex sentence was pronounced as a single, simplex intonation unit.
The long sentence in (18), consisting of a conditional clause, a matrix clause, and a complement clause, was similarly pronounced as a single intonation unit.
The sentence in (19), consisting of a matrix clause and a complement clause with conditional clause, was also pronounced as a single intonation unit.
The patterns make sense given Chafe’s observations that intonation units generally represent one new idea at a time, a single focus of consciousness, be it an event, state, referent, or something else. He distinguished three main types of ideas.
a) given (active in consciousness, generally because they are currently under discussion
or otherwise present)
b) accessible (semi-active in consciousness, either part of previous discussion
or related to something or someone under discussion)
c) new (inactive in consciousness).
A single intonation unit may include one or more given ideas (a) and/or accessible ideas (b), but no more than one new substantive idea (c). The prosodic phrasing seen in the previous examples becomes understandable once we examine the contexts in which they were uttered, now easily accessible with spoken language corpora.
The sentence in (15) ‘Maybe some of them were riding in another car’, with both a verb and separate core arguments, occurred in the context seen in (20). (The full conversation was in Mohawk, but just the free translations are given here for context.)
As can be seen, the children were already part of the ongoing discussion, mentioned in the three immediately preceding comments. Riding in another vehicle was accessible from the mention of the children being brought in in a bus.
The sentence in (16) earlier, ‘I got a bit cold in the morning,’ with verb and adverb in the same intonation unit, occurred in the context in (21).
The cooler weather was already under discussion. The final sentence in (21) above was followed by the longer sentence in (22) below (seen earlier in (17)), containing a matrix clause followed by conjoined adverbial clauses, ‘It’s more pleasant too now that the sun is out and it’s not quite warm.’, all pronounced as a single intonation unit. It continued the discussion of the cooler weather.
The sentence seen earlier in (18) ‘If I’d been home I wouldn’t have noticed what happened, occurred in the context in (23).
The possibility of not noticing had just been mentioned. The new contribution was ‘If I had been home’.
The sentence seen earlier in (19), ‘It’s true you wouldn’t have noticed if you were home’ immediately followed.
In all of these passages, each intonation unit represents just one significant
Much can be learned about grammatical structure from the written word, but some things can be missed. The sentence in (25) appears to consist simply of a subject and predicate.
It is immediately clear from the prosody, however, that it has quite a different structure.
The construction shows a distinctive, recognizable prosodic pattern, with an initial referential expression ‘these White people’, followed by a pause then a full pitch reset on the following nuclear clause ‘they were following weren’t they’. It is a topicalization construction, signaling a shift in topic to ‘these White people’. The preceding context was seen earlier in (8), ‘His father prayed in the way they do in their tradition, according to their custom’.
The same construction topic shift construction can be seen in (26). When just seen in print, it would appear to consist of two simple sentences.
The prosody shows a different structure.
The sentence began with a full pitch reset, then continued with a pause and another full pitch reset on the following clause. The characteristic steep rise then sharp fall of the tone marked with a grave accent can be seen on the verb
Similarly, in print the sentence in (28) appears to be a simple question.
The prosody shows a different structure. It is a focus construction. It began with an extra high pitch on the initial focused element (‘that one’), immediately followed by the remainder of the clause, with no pause or pitch reset. The pitch pattern can be seen in the last intonation unit in Figure 24, along with that of the preceding context.
In print (30) appears to be a simple sentence, with what appears to be a subject, ‘his father’, at the end.
This sentence, too, was pronounced with a distinctive prosodic pattern. It consisted of three intonation units. But the last showed no pitch reset. It was pronounced with a low, flat pitch contour.
The pitch contour can be seen in Figure 25.
This is an antitopic construction. The final nominal, here ‘his father’, serves to confirm the identity of an ongoing topic, a given or accessible participant, especially when multiple participants are in play. These speakers had been discussing the procession leading to the father’s house, where the father was going to pray in the traditional way.
Both syntax and prosody are structured. They often work in concert, but not always. Syntactic structure is more categorical than prosodic structure. Words are in one order or another; markers of subordination and other relations are present or absent. Prosody, by contrast, involves continua of pitch, intensity, and timing.
The two show different degrees of crystallization. Morphological structure is highly crystallized: in Mohawk, morphemes are strictly ordered, and word-level prosody is completely determined by phonological structure: stress occurs basically on the penultimate syllable, and pitch and rhythm depend on the syllable structure.
Syntactic structure is still conventionalized, but slightly less so than morphology. In Mohawk, basic constituent order reflects information structure, with more newsworthy information appearing early in the clause, followed by successively less important and peripheral information. In complement constructions, however, matrix clauses precede their complements, and various kinds of subordinate clauses are usually, though not always, identified by overt markers. Prosodic sentences, ending in a full terminal contour, usually, though again not always, match syntactic sentences.
Information structure in the narrower sense is signaled by both constituent order and conventionalized prosodic structure. Topic shift constructions consist of an initial constituent identifying the topic, pronounced with a full pitch reset, followed by a pause then the nuclear clause, which also begins with a full pitch reset. Focus constructions consist of an initial constituent identifying the focus, pronounced with an extra-high pitch reset, followed immediately by the nuclear clause with not pause or pitch reset. Antitopic constructions, which reiterate the identity of topic already under discussion, consist of a clause followed by a constituent identifying the topic, with no pitch reset and generally a flat pitch, often slower speed, and sometimes creaky voice.
Finally, information packaging in the wider sense (including managing the flow of information beyond considerations of topic and focus) is accomplished with prosodic structure. Intonation units present one new idea at a time. Intonational sentences, which may consist of multiple intonation units showing declination over the whole and ending in a final terminal contour, generally present a single event or state. Intonation units may correspond to syntactic constituents, but not necessarily.
In the end, prosody can tell us things syntax does not, often more directly reflecting the knowledge state of the speaker and what the speaker assumes to be the knowledge state of the hearer.