Introduction[1]
Two views have been expressed on the Yorùbá relative and focus constructions. The first, which we regard as the traditional interpretation, is presented by Owolabi[2] in which the relative clause is regarded as a Noun-Phrase (NP) and the focus construction is analysed as a sentence. The second is the transformational approach discussed by Awobuluyi[3]. In the papers, the relative and focus constructions are regarded as NP’s. In this work, we shall discuss the syntactic status of the Yorùbá focus and relative constructions. To do this, we also need an apparatus. Our choice of grammar is the Generalized Phrase Structure Grammar (GPSG) and the version of GPSG[4] we are interested in is the one presented by Cann[5].
The Syntax
GPSG is a deliberate reaction to Chomsky’s dismissal of non-transformational phrase structure grammars. The grammar is monostratal in that it posits only one level of syntactic representation. What is contained in the syntactic component of GPSG can be summarized as follows:
(a) A set of syntactic FEATURES which, together with a set of FEATURE COOCCURRENCE RESTRICTIONS, define the set of SYNTACTIC CATEGOREIS used in the grammar, and
(b) A set of IMMEDIATE DOMINANCE (ID) RULES which defines the possible HIERACHICAL RELATIONS within PHRASE STRUCTRUE TREES, and
(c) A set of LINEAR PRECEDENCE (LP) RULES defining the LEFT-TO-RIGHT ORDERING OF words/phrases within a tree, and
(d) A set of METARULES that capture regularities between LEXICAL ID RULES by regularly giving as output a set of ID rules for some already extant input of such rules, and
(e) A RULE TO TREE ALGORITHM to define the set of admissible PHRASE STRUCTURE TREES as defined by the set of ID and LP rules together with,
(f) A set of FEATURE INSTANTIATION PRINCIPLES which put restrictions on the way the barely specified ID rules admit acceptable LOCAL TREES, and
(g) A set of FEATURE SPECIFICATION DEFAULTS which give the UNMARKED values of any features i.e. its values when no other principle or rule determines this.
Together, these define an infinite set of phrase structure trees or syntactic structural descriptions of expressions in some language, the LEXICON providing the LEXICAL ITEMS (i.e. words) to TERMINATE THE SYNTACTIC STRUCTURES.
X-bar theory Versus Constituent Incompleteness
As stated above, we shall work within the framework of GPSG presented by Cann which is a variant of the one introduced by Gazdar et al.[6]. Though both share the same conviction, the latter differs from the former in that, while it uses the X-bar theory, the former’s analysis is based on the notion of ‘constituent incompleteness’.
In Gazdar et. al., a syntactic category is analysed as a set of feature specification each of which consists of a feature name and a value for that name. For instance, a NP that is second person singular could be represented as in (1)
(1) [BAR 2]
[N, +]
[V, -]
[per 2]
[PLU, -]
Gazdar et al. adopt a two-level X-bar theory in their framework. The [BAR, 2] in (1) states that the category is a maximal projection of NP. The other features should be fairly self-explanatory. Double-bar categories as maximal projections also apply to VP’s and S’s. The three can be represented as follows:
(2) NP = [+N, -V, BAR 2]
VP = [+V, -N -SUBJ, BAR 2]
S = [+V, -N, + SUBJ, BAR 2]
It should be noted that a syntactic category in this framework does not have to be a fully specified category. So, [BAR 2] may be understood as referring to any maximal projection.
In Cann, on the other hand, a bar level approach to phrase structure is rejected in favour of a two feature approach that defines phrases in terms of the primitive syntactic notion of ‘constituent incompleteness’ and ‘lexicality’. The two Boolean features introduced to reflect the information necessary to determine phrasal type are MAXIMAL and LEXICAL, we can have the features in (3) which, because they take the values + or -, define four possible categories in terms of phrasal type in (4):
(3) {<MAXIMAL, {+,-} >< LEXICAL, {+,->}
(4) (a) {<MAXIMAL, + > < LEXICAL, ->} :XP (complete non-lexical)
(b) {<MAXIMAL, - >, < LEXICAL, ->} :XI (incomplete non-lexical)
(c) {<MAXIMAL, - >, < LEXICAL,+>} :XI (incomplete lexical)
(d) {<MAXIMAL, + >, < LEXICAL,+>} :XM (complete lexical)
To account for feature distribution within headed paths in lexical structures, there is a Head Feature Convention which states that:
(5) The HEAD features of a head in a local tree must be an extension of the HEAD features of its non-lexical mother, while a lexical mother’s HEAD features must be an extension of those of its HEAD.
Given (3, 4 & 5), we may analyse (6a) as in (6b-e) (both word-for-word translations and category abbreviation are given in brackets).
(6) (a) Olú tí ó lọ (Olú REL go) ‘Olú who went’: (non-lexical, complete (NP))
(b) tí (REL) ‘who/which/that…’ (lexical, incomplete (NL))
(c) lọ ‘go’, ó ‘he’ and Olú: (lexical, complete (NM))
(d) tí ó lo (REL he go) ‘who went’: (non-lexical, incomplete (NI))
The question may be asked as to why we prefer to work within the framework of Cann’s version of GPSG? It seems that the major problem with the analysis of both the Yorùbá focus and relative constructions is the lexical status of the item ni and as Cann has rules that introduce lexical items in his theory, we are of the opinion that it will bring out the lexical status of this item more clearly than Gazdar et. al.
All Yoruba grammarians agree that the relative market, tí, as in Olú tí ó lọ (Olú REL he go) ‘Olú who went’ occurs in a NP and will be analysed as a NPFORM i.e. NL[tí] in this work. There is still a controversy over the lexical status of the item ni as in Olú ni ó lọ (Olú FOC he go) ‘It is Olú who went’. In Awobuluyi’s[7] view, the item ni is a verb meaning ‘to be’. Oyelaran[8] regards the item simply as ‘an assertive marker (which) asserts whatever precedes it as new information in the discourse’.
We quite agree with Oyelaran that the item ni cannot be preceded by any auxiliary verb and that it does not nominalize like most verbs but we note that not all verbs are preceded by auxiliary verbs nor are they all nominalizable. For instance, the verb tì in Ó tì (he not) ‘No’ can neither be preceded by an auxiliary nor be nominalized. So, in this work, we shall analyse the item ni as a VPFORM i.e. VL [ni] for the following reasons. First, like such verbs as pè ‘call’ and rán ‘send’, ni subcategorizes for both NP and S as in (7).
(7) (a) Awakọ̀ ni ọ́ (Drivers is you). ‘You are a driver’
(b) Awakọ̀ pè ọ́ (Driver call you) ‘A driver called you’
(c) Awakọ̀ rán ọ (Driver send you) ‘A driver sent you’
(d) Awakọ̀ ni Olú (Driver is Olú) ‘Olú is a driver’
(e) Awakọ̀ pe Olú (Driver call Olú) ‘A driver called Olú’
(f) Awakọ̀ rán Olú (Driver send Olú) ‘A driver sent Olú’
(g) Awakọ̀ pe Olú lọ (Driver call Olú go) ‘A driver called Olú away’
(h) Èkó ni Olú lọ (Lagos is Olú go) ‘Olú went to Lagos’
(i) Awakọ̀ rán Olú lọ (Driver send Olú go) ‘A driver sent Olú (somewhere)’
Second, ni also takes preverbal items (PV) such as sí, tiẹ̀, etc. as in (8). Note pè ‘call’ and rán ‘sent’ in (9) and (10) respectively
(8) (a) Awakọ̀ sì ni Olú
(b) Awakò tiẹ̀ ni Olú
(c) Awakọ̀ dẹ̀ ni Olú
(d) Awakọ̀ sáà ni Olú
Driver PV is Olú
‘Olu is even a driver’
(9) (a) Awakọ̀ sì pe Olú
(b) Awakọ̀ tiẹ̀ pe Olú
(c) Awakọ̀ dẹ̀ pe Olú
(d) Awakọ̀ sáà pe Olú
Driver PV call Olú
‘Olu was even called by a driver’
(10) (a) Awakọ̀ sì rán Olú
(b) Awakọ̀ tiẹ̀ rán Olú
(c) Awakọ̀ dẹ̀ rán Olú
(d) Awakọ̀ sáà rán Olú
Driver PV send Olú
‘Olu was even sent by a driver’
It should be noted, however, that item ni differs from such words as pé ‘call’ and rán ‘send’ in one major way. Pé ‘call’ and rán ‘send’ belong to a word class which is assigned a particular subcategory identified by a numeral and rules of the types of (11) make reference to those numerals.
(11) (a) V à HL[1], ‘sùn ‘sleep’, kú ‘diẹ̀’
(b) V à HL[2], N pè ‘call’, rán ‘send’
(c) V à HL [3], S pè ‘call’, pé ‘say’ ran ‘send’
In 11a, verbs such as sùn ‘sleep’ and kú ‘die’ belong to V[1] while rán ‘send’ and pè ‘call’ belong to VL[2] AND VL[3]. The item ni, on the other hand, as a VFORM, does not belong to any word-class and is introduced as in (12a and b).
(11) (a)V à HL[2], sùn ‘sleep’, kú ‘die’
(b) V à HL[2], N pè ‘call, rán ‘send’
(c) V à HL[3], S pè ‘call’, rán ‘send’
tí, which is also a NFFORM, is introduced as in (12)
(12) N à NL[NFORM ti], S
The Analysis
Consider the set of rules in (14)
(13) (a) S à N, V
(b) V à HM/sùn ‘sleep’, kú ‘die’
(c) V à HL[2], N/pè ‘call’, rí ‘see’ rán ‘send’
(d) V à HL[3], S/pè ‘call’ pé ‘say’, rán ‘send’
(e) V à HL[4], P/lọ ‘go’, wa ‘come’
(f) V à HL[5], N,P/rán ‘send’, rí ‘see’, fún ‘give’
(g) V à HL[6], D/lọ rí ‘went (there) before’
(h) P à HL[7], N/ní ‘in’, sí ‘si’
(i) N à HM/Olú, Adé, Òjó
(j) N à HL[NFORM ti], S
(k) V à HL [VFORM ni], S
(l) V à HL [VFORM ni], N
The rules we have introduced in (14) will assign labeled brackets shown in (17) and (18) to (15) and (16) respectively
(14) Olú ni ó ra aṣọ
Olú FOC he buy cloth
‘It was Olú who bought some clothes’
(15) Olú tí ó ra aṣọ
Olú REL he buy cloth
‘Olú who bought some clothes’
(16) [S[NM Olú][VP[VL[ni] ni][S[NM ó][VP[VL[2] ra][NM aṣọ]]]]]
‘It was Olú who bought some cloths’
(17) [NP[NM Olú][NI[NL[tí] tí][S[NM ó][VP[VL[2] ra][NM aṣọ]]]]]
‘Olú who bought some clothes’
For the examples, in (15) and (16), the resumptive pronouns which function as subject in place of the topicalized and relativized NP’s are not recognized as empty category in GPSG. The question then is how do we account for (19) and (2) where there are some missing constituents?
(18) ‘Òjó ni ó rí –
Òjó FOC he see
‘It was Òjó that he saw’
(19) Òjó tí ó rí –
Òjó REL he see
‘Òjó that he saw’
GPSG uses a feature NULL to encode that a constituent is phonologically empty. The feature NULL defaults to being absent; hence it will only ever appear in a tree when sanctioned by a rule. Such rules will be derived by a metarule – in fact, by two but here we shall be concerned with one which is called the Slash Termination Metarule 1. The metarule is in (21).
(20) X à W, X
X à W, X[+NULL]
The metarule simply says that any maximal category can be null. So, from rules, such as the ones in (14 j and k), we can have (2a and b).
(21) (a) N à NL [NFORM ti], S/N
(b) V à VL [VFORM ni], S/N
The rules in (14, 20 and 21) taken together will assign labeled brackets shown in (22) and (23) to (18) and (19) respectively.
(22) [S[NM Òjó [VP[VL[ni] ni][S/N[NM ó][VP/N[VL[2] rí]]]]]
‘It was Ojo that you saw’
(23) [NP[NMÒjó][NI[NL[tí] tí][S/N[NM ó][VP/N[VL[2] rí]]]]]
Conclusion
As the analysis in this work shows, Yorùbá focus construction is an S and the relative construction is a NP. The Yorùbá focus marker is a VPFORM which takes an S or a NP as a complement. The relative construction, on the other hand, is a NP. The relative marker is a NPFORM which takes an S as a complement. The analysis, therefore, supports the one earlier presented by Owolabi[9]. The difference between the papers is that the lexical status of both the item ni and tí is clearly specified in our analysis.
Zwicky[10] describes GPSG as ‘a generative, but non-transformation, theory of syntax’. According to him:
GPSG is of recent vintage, less than a decade old, and English is its first language. It is no surprise, then, that the first book-length exposition of the theory, the 1985 volume by Gazdar, Pullum, and Sag, relies on English to illustrate the features of GPSG. Nevertheless, GPSG is intended as a… universal… theory of syntax.
Although this paper does not go very far, it is able to show that GPSG can provide a framework for describing the syntax of what we regard as one of its second languages, Yorùbá. The importance of the focus and relative constructions often referred to in the literature as the un-bounded dependencies to any grammar cannot be over-emphasized[11]. The fact that GPSG can handle some difficult problems of the relative and focus constructions left unsolved by previous writers demonstrates the usefulness of the grammar for providing adequate analyses for Yorùbá syntax.
[1] An earlier version of this paper was published as L.O. Adewole (1991), ‘Heads without Bars: A Solution to the Sentential Status of the Yoruba Focus and Relative Constructions’, Odu: Journal of West African Studies 38: 19-27.
[2] K. Owolabi, ‘More on the Inadequacy of the Analysis of Focus Construction as Noun Phrases with Reference to Yoruba’, Linguistic Analysis 12 (1983), pp.453-471 and ‘Focus Construction as Noun Phrases: A Critique’, Yoruba (N.S.) 1 (1987), pp. 45-62.
[3] O. Awobuluyi, ‘Focus Constructions as Noun Phrases’, Linguistic Analysis 4,2 (1978), pp. 93-114 and ‘Focus Constructions as Noun Phrases: A Reply’, Yoruba (N.S.) 1 (1987), pp. 72-87.
[4] We will ignore the matter of semantic translation in this paper, first, on the assumption that they are derived algorithmically from syntactic rules in the manner described by E. Klein and I.A. Sag in ‘Semantic Type Control’, in Development in GPSG, M. Barlow et. al. (eds), (Indiana, 1982), pp. 1-25, and second, because they never constribute to determining string or tree admissibility. For convenience, we have replaced all references to V’’, N’’, etc. with VP, NP.
[5] R. Cann, ‘Heads without Bars: A Theory of Phrase Structure’, Unpublished Manuscript, (University of Edinburgh, 1986).
[6] G. Gazdar et.al., Generalized Phrase Structure Grammar, (Oxford, 1985).
[7] Awobuluiy, op. cit., 1978, p. 93.
[8] O. Oyelaran, ‘Focus and Anti Focus: A Perspective from the Syntax of Yoruba’, Conference Paper, Ninth Annual Conference of Linguistics Association of Nigeria, 31st July to August 4th 1988 at Obafemi Awolowo University, Ile-Ife, Nigeria.
[9] Owolabi, 1983 and 1987, op. cit.
[10] A.M. Zwicky, ‘Introduction to German Syntax in GPSG’, Linguistics 24,5 (1986), pp. 855-856.
[11] See G. Horrocks, Generative Grammar, (London, 1987) and P. Sells, Lectures on Contemporary Syntactic Theories, (Stanford, 1985).
No comments:
Post a Comment