• Nie Znaleziono Wyników

1.1 The structure of utterances for prosodic analysis

1.1.1 Higher-level utterance constituents

In line with most phonological approaches, we will regard the utterance as the highest-level domain, a relatively autonomous speech unit in terms of prosody, syntax and discourse, often corresponding to a sentence.

The segmentation of utterances in the context of pronunciation teaching has traditionally been based on the Nuclear Tone Approach, associated with the British School (Kingdon 1958, Crystal 1969, O’Connor and Arnold 1961, 1973, Gimson 1980, Cruttenden 1986). This approach was indeed designed for foreign learners of English. The fundamental unit of an utterance has been called a sense group (O’Connor and Arnold 1961), tone-group (O’Connor and Arnold 1973) intonation-group (Cruttenden 1997) or a tone-unit (Crystal 1969, Halliday 1970). The central element of such a unit is the nucleus (Palmer 1922), defined as “the stressed syllable of the last prominent word in a sense group” (O’Connor and Arnold 1961: 271). The part of the utterance from

1.1 The structure of utterances for prosodic analysis 15

the first pitch-accented syllable to the nucleus is referred to as the head (Crystal 1969, O’Connor and Arnold 1961) and the accented syllables preceding the nucleus are known as prenuclear accents (Cruttenden 1997).

Unaccented syllables preceding the head are called prehead (Crystal 1969, Halliday 1970, O’Connor and Arnold 1973, Cruttenden 1997). Finally, the nucleus is followed by an optional tail (O’Connor and Arnold 1961), also called nuclear tail by Crystal (1969). The tail, like the prehead, consists of unaccented syllables alone. The described parts of the tone unit are further divided into feet. Abercrombie (1967) defines the foot as a tone unit constituent comprising a stressed syllable and the following unstressed ones up to the next stressed syllable. According to this view, a word boundary does not automatically terminate the foot, which can thus become a unit larger than the word (a cross-word foot — cf. White 2002).

A similar model of prosodic structure, proposed by Jassem (1952), distinguishes narrow rhythm units (NRUs), which, combined with anacruses, constitute total rhythm units instead of the feet. The narrow rhythm unit is also initiated by a stressed syllable but, in contrast to the foot, it only includes syllables within the same lexical unit, which actually implies a claim that word boundaries in connected speech may influence the prosody. For instance, the phrases “summer dresses” and “some addresses” would receive identical description in Abercrombie’s approach, whereas according to Jassem, the first syllable of “addresses” forms an anacrusis, a constituent outside the narrow rhythm unit (cf. Bouzon and Hirst 2004, Hirst and Bouzon 2005).1

Although the British School model of prosodic structure is still popular, a lot of recent research is based on the Autosegmental-Metrical (AM) approach, proposed by Pierrehumbert (1980) and developed by Beckman and Pierrehumbert (1986) and Ladd (1986, 1996). The AM theory differs from the British School not only in the approach to tones as discrete units, or pitch targets (H or L) rather than contours, but also in the approach to phrasing, which has been based on the studies by Selkirk (1980) and Nespor and Vogel (1986). These researchers postulated a hierarchy of prosodic units, distinguishing two basic levels of phrasing, viz. the full intonational phrase (IP) and the intermediate phrase (ip). The latter is further divided into prosodic words (or clitic groups) and feet (cf. Turk and White 1999). The prosodic domains, except for the lowest constituents, are made up of at least one constitutent of immediately lower level (Selkirk 1978, 1995b, Nespor and Vogel 1986, Hayes 1989). The slight differences between individual authors are presented in Table 1.1.

1 The difference between Abercrombie’s and Jassem’s models, investigated by Bouzon and Hirst and closely related to other considerations of the role of word boundaries, will be taken into account in calculating the timing proportions in our empirical study, although insufficient amount of data will not allow to draw reliable conclusions in support of any of the two approaches.

1.1 The structure of utterances for prosodic analysis 17

Table1.1.Theprosodichierarchy Model LevelSelkirk1986Nespor andVogel1986 (non-recursive) Beckman andPierrehumbert 1986 Hayes1989 (non-recursive)Jun1998Fougeron andKeating 1997

Keatingetal. 2003 Higherutteranceutteranceutteranceutteranceutteranceutteranceutterance ­IPIPIPIPIPIPIP ­intermediate phraseaccentualphraseaccentualphrase (forKoreanand French)

intermediate phrase/ phonological phrase

smallerphrase (variesacross languages) ­phonological phrase (major+minor) Selkirkand Tateishi1988

phonological phrase(always correspondswith syntactic boundaries)

phonological phrase———— ­¯prosodicwordcliticgroupprosodicwordcliticgroup ¯footwordfootfootwordwordword ¯syllablesyllablesyllablesyllablesyllablesyllable Lowermora————

A hierarchic structure can display layeredness or recursivity. In the case of prosodic hierarchy, arguments are presented for both approaches. Strict Layer Hypothesis (SLH) postulates that each prosodic constituent is contained in a constituent of the adjacent higher level (Selkirk 1984). However, some linguistic evidence (pitch and boundary strength) questions SLH, pointing at the possibility of nesting constituents within constituents of the same rank in prosodic hierarchy, thus allowing recursive phrasing (Ladd 1986).

Prosodic domains are also to some extent related to syntactic constituents.

Selkirk (2005), in her Syntactic Grounding Hypothesis, proposes the following correspondence:

Table 1.2. Syntactically determined prosodic constituents (Selkirk 2005) Syntactic constituent Prosodic constituent comma phrase (CmmP) intonational phrase (IP)

lexical maximal projection (lexP) major phonological phrase (MaP) branching syntactic constituent minor phonological phrase (MiP)

lexical word (lex) prosodic word (PWd)

However, although the correspondence between syntactic and prosodic constituents is fairly strong, syntactic boundaries are coded in prosodic structure in not more than 65—84% (Fach 1999). These syntactically determined constituents can be identified on the basis of segmental phonological rule application (Nespor and Vogel 1986, Jun 1993) and intonation (Beckman and Pierrehumbert 1986). An Intonational Phrase, for instance, bears a complete intonation contour. Because it often comprises a full sentence, it can be coextensive with the utterance. Nespor and Vogel (1986) found some differences in the application of certain phonological rules, but Wightman et al. (1992) observed no difference in final lengthening, a process depending on the domain’s rank in prosodic hierarchy (to be discussed in section 1.3.1).

Differences in proposed prosodic hierarchy models are also caused by differences between languages. Certain languages might require specific phrasing, which must be reflected in the prosodic hierarchy. For instance, the Accentual Phrase — a unit between IP and Phonological Phrase is introduced for prosodic description of Korean and French (Jun 1998).

As mentioned before, apart from lexical and syntactic information, the phrasing of speech depends on stress and accent, which highlight important elements of the spoken message.2 Both British School and AM theories recognise nuclear accent, but in Pierrehumbert’s view, prenuclear accent

2 The problems of prominence will be discussed in section 1.4.

(“head” in British tradition) does not constitute a separate category, and the same accent inventory is applied in either case.

Autosegmental-Metrical theory has given rise to a specific transcription system called ToBI (Tone and Break Indices) (Silverman et al. 1992), where the prosodic structure and intonation pattern of a phrase is described by means of H (high) and L (low) tones and their combinations, and break indices referring to the degree of cohesiveness between words. ToBI has become a standard tool for describing the prosody of English, but also a wide variety of other languages (e.g. Jun 2005), which often require “national” modifications of the system.

Summing up the two relevant approaches, we receive a possible inventory of a number of speech units capable of forming a multilevel hierarchy.

The hierarchic structure of utterances refers to the relatively independent and “complete” fundamental unit of intonational phrase, at the level of tone/sense/intonation group/unit. If an IP consists of more than one intermediate phrase (Beckman and Pierrehumbert 1986), then the tone group corresponds with the latter unit.

1.1.2 Relations between the word (lexical or functional)