4 SISTEMA DE GESTION DE LA CALIDAD
4.2.2 MANUAL DE CALIDAD
The first detailed account on how tonal primitives are phonologically related to metrical primitives is provided in Pierrehumbert and Beckman (1988) for their analysis of Japanese tone structure. In this study, metrical information is represented by means of a prosodic tree, which consists of a hierarchy of phonologically defined units. The idea of prosodic trees has been applied to several languages. However, the number and kinds of levels of phonological structure within the tree varies among these languages. In English, for example, four prosodic levels have been posited, namely, the IP, the ip, the foot and the syllable. The IP is the highest level of prosodic structure whose limits are demarcated by a boundary tone, either L% or H% at the right edge, and no tone or H% at the left edge^. The IP is the prosodic domain within which pitch range is specified and where a large disjuncture, such as a major pause or break, is perceived at the boundary. The ip is a lower constituent, which must include at least one pitch accent. The ip is tonally demarcated by the presence of a phrase accent, either L- or H- at the right edge. It has a medium degree of disjuncture (i.e. no large breaks). In English, the ip is defined as the domain of downstep (Beckman and Pierrehumbert 1986). However, this varies cross-linguistically since in other languages downstep has been shown to occur at the IP level (e.g. Greek: Arvaniti and
^ The default initial boundary is supposed to have a mid or low pitch range and is left unmarked for transcription (Pierrehumbert 1980, Beckman and Hirschberg 1994). For criticisms on this aspect, see Cabrera-Abreu (1996) who proposes a model with no L tones.
Baltazani 1999). A simplified adaptation of Pierrehumbert and Beckman's tree for English is presented in (3.1) below^.
(3.1) PROSODIC TREE Intonation Phrase IP Intermediate Phrase Foot
A A
a a a d Syllable a o PhA BT PA PATONE TIER BT PA PA PhA
PHONEME TIER (e.g.) Singing at the opera must be thrilling
As observed in (3.1), the highest level of the prosodic tree is the IP, which dominates lower daughter nodes representing relevant entities of phonological structure. Domination is schematised by straight lines. Curved lines represent the association between elements at the prosodic tier and at the tone tier. Association between tonal entities and metrical entities operates both at a low-level node, where syllables are linked to pitch accents (PA) and at higher level nodes, where edge tones (PhA and BT) are associated to the edges of prosodic phrases. The association between syllables and pitch accents is not arbitrary but governed by metrical strength (see chapter 1 for more details). Pitch accents are linked to the strong branch of the foot, marked with a thicker black line.
The prosodic tree varies cross-linguistically. For example, languages such as Japanese (Venditti 1999) and Korean (Jun 1996) have no ip level of phonological structure.
There are several aspects related to this hierarchical representation that have not been shown since they are not relevant for the purposes o f this study. For example, pitch accents might have an internal structure (bitonality) as opposed to edge tones, which are monotonal (see Grice 1995a, b for more details).
Instead, the domain below the IP is the Accentual Phrase"^. Alternatively, Jun and Fougeron (to appear) claim that French has both an ip and an Accentual Phrase. The Accentual Phrase is defined by a specific tonal pattern, which demarcates the right edge of the phrase. In Korean and French, the delimitative tones are LH. In Japanese, there is a rise associated to the second mora and a subsequent fall to a low tone at the end of the phrase. No Accentual Phrase level of phonological structure has been proposed for English since there are no delimiting edge tones for such a constituent (Beckman and Pierrehumbert 1986).
Cross-linguistic differences not only apply in the levels of phonological hierarchy but also in the nature, function and distribution of pitch accents. For example, whereas in English the association between pitch accents and metrically strong syllables occurs at a postlexical level (Pierrehumbert 1980), in Japanese the location of pitch accents is specified at the lexical level (Beckman and Pierrehumbert 1986, Pierrehumbert and Beckman 1988 and Venditti 1999). This means that the distribution of pitch accents in Japanese is fairly fixed as opposed to English where the distribution of accents is more flexible and is closely linked to a focussing function.
The intonational structure described so far for English (with two levels of phrase prosodic structure: the IP and the ip) and the postlexical association of pitch accents as a reflex of focussing have been argued in other languages, such as Palermo Italian (Grice 1995a), European Portuguese (Frota 1998), Bengali (Hayes and Lahiri 1991)^, Greek (Arvaniti and Baltazani 1999), German (Benzmtiller and Grice 1999) and Central Catalan (Prieto 1995, 1997), among others. In this study, a similar prosodic representation to the one sketched in (3.1) for English will apply to Central Catalan. Thus, the two levels of phrasing proposed by Prieto will be confirmed by the data as well as the postlexical association of tonal primitives to metrical relevant units. However, a smaller prosodic domain positioned below the ip will also be justified on the basis of the data presented.
^ Pierrehumbert and Beckman (1988) proposed three levels above the word in the prosodic hierarchy of Japanese: the Accentual Phrase, the ip and the Utterance. In the J_ToBI model (Venditti 1999), the ip and the Utterance have merged into a single level o f phrasing, the IP.
^ The levels o f prosodic structure proposed for European Portuguese and Bengali are defined both on intonational grounds and on the application o f junctural phenomena. In both studies, the level below the IP is the Phonological Phrase (as in Selkirk 1980 or Nespor and Vogel 1986).