EL DIA QUE LA MISERICORDIA DE DIOS SE HIZO CARGO
CUANDO EL CIELO BAJA A LA TIERRA
The last sections showed that in EA prosodic phrases at the M aP level are consistently large, and argued that a preference for binary MiPs can account for sparse M aP phrasing. N either M aP nor M iP are routinely co-extensive with the PW d in EA. In this section we explore whether these findings are reproduced in the thesis corpus data.
M uch of the corpus (described in chapter 3 section 3.2.1) consists of read sentences which are neither prosodically long nor syntactically complex. A survey was thus made of selected portions of the narratives database, chosen because the narrative contains sequences of text w ithout punctuation marks in the written text, to which speakers are at liberty to assign w hatever prosodic phrasing they choose. Each speaker read the
narrative three times, as well as being asked to re tell the story later from m em ory102. The second reading of the narrative by each speaker, together with their retold sem i' spontaneous rendition, was prosodically transcribed by the author, with reference to FO track and spectrogram extracted using Praat 4.2 for the corpus survey described in chapter 3. The most common cues to phrasing in this subset of the narratives dataset are boundary tones, such as H- or L-, and pauses, and as discussed above, these are
assumed to mark the edges of MaP level constituents.
The phrasing facts observed in this corpus subset are consistent with the findings of section 5.3.2 above. There are examples of moderately long/complex monoclausal sentences which are phrased into a single MaP. Example (5.40) below shows the auditory transcription and phrasing o f a 5PW d monoclausal sentence103.
(5.40) Speakers’ read speech phrasings of a 5PWd monoclausal sentence. speaker guHa kaan Tuul 9um r -uh 9aayis fi -1- ?ariyaaf
fn a LH* — LH* LH* < LH* < < !LH*
M LH* < < LH* < LH* < < !LH* L-L%
meh LH* < (LH*) LH* < LH* < < LH* H-
m iz LH* < LH* LH* < LH* < < LH* H-
inns LH* < < LH* < LH* < < LH* L- guHa kaan Tuul 9umr -uh 9aayis fi - 1- ?ariyaaf Guha was all life- his living in the villages
[[NP ] AUX [AdvP ] [V [PP ]]]s
‘G uha had lived all his life in the countryside.’
102 The full text o f the narrative is provided in A ppendix A . 13
1(0 In exam ples 5 .4 0 -5 .5 3 MaP boundaries are show n enclosed within vertical lines: IMaPI, and a gloss and syntactic analysis is provided; the transcription conventions used are those set out in chapter 3 section 3 .2 .2 ).
There is evidence from the phrasing in the narratives corpus to confirm that the right edges of VP-internal X P boundaries do not trigger MaP phrase boundaries. This confirms the finding in Hellmuth (2004) that the effects of the interface constraint AlignXP are obscured in EA. For example, (5.42) below shows speakers’ read narrative phrasings o f a 4PW d monoclausal sentence with VP-internal X P s104.
(5.41) Read speech phrasings of a com plex 4PW d monoclausal sentence. speaker ?ana 9awz ?awzin lak kiilu bi balaas
fn a > TLH* LH* < LH* < !LH* L-L%
M > LH* LH* < LH* < !LH* L-L%
meh > LH* LH* < LH* < !LH* L-L%
miz > LH* LH* < LH* = < LH* L-L%
inns > LH* LH* < LH* < !LH* L-L%
?ana 9awz ?awzin lak kiilu bi balaas
I want I-weigh for-you a-kilo for- free
[[N P ] [V [V [PP] [NP] [PP ]] V p ]v p ]
‘I will weigh you out a kilo for free!’
This treatment of com plex XPs is m aintained in speakers’ spontaneous (retold) narrative productions. For example, (5.43) and (5.44) below show the same sentence, in
spontaneous retellings by speakers f s f and meh, which are phrased into a single MaP.
(5.42) Single M aP phrasing (retold by speaker/?/) of a complex clause. ?ana mumkin ?akuun baddiik kiilu bi balaas
> LH* LH* LH* H* < !LH* L-L%
I maybe I-could I-give-you a-kilo for free
[[NP] ADV AUX [V [NP] [PP ] ]v p]s
T could maybe give you a kilo for free.’
(5.43) Single M aP phrasing (retold by speaker meh) of a complex clause.
?eh ra?y-ak ?addiik kiilu mooz bi balaas
LH* LH* -+LH* —>LH* LH* < !LH* L-L%
what your-opinion I-give-you a-kilo bananas for free
[N P NP t [C [V [NP ] [PP ] ]v p]c p]s
‘How about if I give you a kilo for free?’
104 N ote that speaker m iz lengthens the word [kiilu] ‘k ilo ’ slightly, w hich m ight suggest phrase-final lengthening. H ow ever in accordance with the m ethodology o f Hellmuth (2 0 0 4 ), a phrase boundary w as transcribed only w hen at least tw o acoustic or tonal cu es to phrasing were observed (see 5 .3 .2 above).
The results from the narratives corpus are also consistent with the idea that it is a preference for branching prosodic phrases ( Bi nMaPmip) which outweighs the effects of
the interface constraint ( Al i g nX P ) . For example, a 5PWd monoclausal sentence, such as in (5.44a) below (reproduced from (5.40) above), cannot be subdivided at the M aP level because the resulting M aPs would contain less than two MiPs (assum ing that MiPs must themselves be minimally branching at normal speech rates). The sentence is thus rendered by all speakers in a single M aP105.
(5.44) Phrasing analysis: two observed renditions of a 5PWd monoclausal sentence
a. l([guHa]) (ken [Tiiul] [9umr-uh]) ([9aayis] [f-il-?ariyaaf]Pwd)MiP Im«p
b. l([guHa] ken Tul [9umr-uh]) ([9aayis] [f-il-?ariyaaf]Pwd )mip Im«p
Guha was all life-his living in-the-country ‘Guha had lived all his life in the countryside.’
In (5.44) there is slight variation in pitch accent distribution between different speakers: fna and miz accent the word [Tuul] ‘all’, whereas f s f and inns leave it unaccented (there is a possible accent on the word in speaker m eh's rendition)106. In both cases there is insufficient prosodic material to form more than one well-formed MaP.
The narratives also contain multiclausal sentences, which show that short clauses may be phrased independently. The example in (5.45) below involves a short introduction to a fragm ent of reported speech , phrased into two MaPs of few er than 4 PW ds each by three out o f five speakers, in violation of the constraint Bi nMaPm ip.
(5.45) Exam ple of a sequence of very short clauses (introducing reported speech), ( ‘xxx’ indicates that a word was omitted.)
guHa ?aal-Iuh (laa?) bi- talaata SaaG
fn a 2 LH* LH* H- xxx > LH* LH* L-L%
m LH* LH* H- xxx > LH* LH* L-L%
meh2 LH* LH* LH* L- > LH* LH* L-L%
miz2 LH* LH* LH* < LH* !LH* L-L%
mns2 LH* LH* xxx > LH* !LH* L-L%
guHa ?aal-luh laa? bi- talaata SaaG Guha said-him no for- three piastres
[[NP] [V [NP]]VP]s [[PP ]]
‘G uha said to him: “No, three piastres!’”
105 Pitch accents are indicated with an acute accent mark eg [Tuul]; PWd boundaries are indicated by square brackets [PW d], M iP boundaries by round brackets (M iP). and M aP boundaries by vertical lines IMaPI. 1 assum e that PWds align to the right edge o f an M iP. This is explored further in chapter 6 (6.2.4). 106 T he relative accentability o f m odifiers such as /T u u l/ is discussed in chapter 3.3.2.
The option of phrasing such short fragm ents into individual M aPs is, I suggest, due to the influence of an interface constraint regarding mapping of clauses to Intonational Phrases (IPs). Any string which maps to an IP, must also map to at least one MaP, by virtue of the constraint He a d e d n e s s, which requires a constituent at one level of the
prosodic hierarchy to dominate at least one constituent of the next level down, and is assumed by some authors to be unviolated (e.g. Selkirk 1996). In (5.45) above then, the m id-utterance boundary should be analysed as an IP (as well as M aP) boundary.
As discussed above in section 5.2.3, the distinction M aP and IP boundaries was established on the basis of evidence from non-application of epenthesis across IP boundaries. Chahal (2001) distinguished between two phrase levels in Lebanese A rabic (LA) on the basis of gradient, though statistically discernible phonetic correlates to phrasing. These took the form o f increased final lengthening and earlier peak delay at higher boundaries, according to the level of constituent boundary marked; final lengthening was greater in IP-final words than in MaP final words, for example. It is beyond the scope o f the present investigation to establish this gradient distinction empirically in EA, and thus, although the auditory transcription identified these inter clause boundaries as M aP boundaries they are now re-analysed on theoretical grounds as IP boundaries, as in (5.46) below 107.
(5.46) Phrasing analysis: observed renditions of a sequence of short clauses,
a) ip ( Map! M ip ( pwd[guHa] [?aal-luh]) (bi- [talaata] [SaaG ]PWd )iviip iMaP ) ip
b) jp ( Mapl M ip ( PWdtguHa] [?aal-luh])i)ip ip(l(bi- [talaata] [SaaG]Pwd ) m i p Im sP ) ip
Guha said-to-him for- three piastres ‘Guha said to him: “No, three piastres!” ’
This section has shown that the claims set out regarding EA phrasing in sections 5.2.2 and 5.2.3 hold in the thesis corpus: prosodic well-formedness constraints outweigh interface alignment constraints (BinMaP MiP» ALIGN XP,R) and M aP-binarity involves sequences o f 4 PW ds or more. Crucially, these facts are consistent with rejection of M aP and M iP as potential domains of pitch accent distribution in EA, since even in the largest MaPs in the narratives every PW d bears a pitch accent. The next section (5.4) explores accentuation of function and content words in order to determ ine the relationship of PW d distribution to m orphosyntactic categories.
107 This re-analysis is consistent with the notion that parentheticals and other non-root sentences form an independent IP (N espor & V ogel 1986, Selkirk 2005a).