161 A UNIFIED ANALYSIS OF RELATIVE CLAUSES AND QUESTIONS IN RAINY RIVER OJIBWA Alana Johns University of Ottawa In this paper I will examine the structure of relative clauses an questions in the Rainy River dialect of Ojibwa. Of special interest will be the syntactic status of the morphemes /ka:-/, and the changed form. In contrast with recent analyses (Lees 1979; Pagotto 1980), I will argue that these morphemes display properties more characteristic of relative pronouns than of complementizers. This argument leads to an analysis of questions in Ojibwa in which the internal structure is identical to relative clauses. In Ojibwa all subordinate constructions, including simple sentential complements such as those in (1), are marked with the conjunct (subordinate) theme sign /-it/ to the right of the subordinate verb: (1) a. ngikenima: ikwe akozit I-know-(l-3) woman sick-3c I know that the woman is sick b. ngikenima: ikwe ki: 9 akozit I-know-(l-3) woman past-sick-3c I know that the woman was sick c. ngikenima: ikwe ciakozit I-know-(l-3) woman future-sick-3c I know that the woman will be sick In (la) the present tense is null (as it is in matrix clauses). In regular past tense morpheme /gi: 9 -/ (/M: 9 -/ word-initially) is prefixed to the left of the subordinate verb. In (lc) the morpheme /ci-/ marks the future tense of the subordinate verb. The examples of relative clauses in (2) demonstrate a different set of prefixes from those in (1): (2) a. inini kamagamat kino:zi man WH-sing-3c tall-3 The man who is singing is tall b. ngikenima: inini ka:gi: 9 nagamat I-know-(l-3) man WH-past-sing-3c I know the man who sang
162 Alana Johns c. ngikenima: inini ka:wi 9 nagamat I-know-(l-3) man WH-future-sing-3c I know the man who will sing d. ka:nagamat kino:zi WH-sing-3c tall-3 The one who is singing is tall The present tense form contains a morpheme /ka:-/ prefixed to t left of the verb as in (2a). The past tense form (2b) contains the prefix /ka:-/ plus the past tense morpheme /gi: 9 -/. The future form (2c) contains the /ka:-/ plus the regular future tense morpheme /wi 9 -/. I will leave aside the question of why the future form of the simple sentential complement (lc) contains /ci-/ rather than the regular future tense /wi 9 -/. Note that the antecedent position of relative clauses may be left empty as in (2d) where the relative clause does not have a lexical head, i.e., it is some form of free relative. Relative clauses may contain the changed form in place of the morpheme /ka:-/ as in (3). The changed form is a morpheme which manifests its presence by a phonological change in the first vowel of a verb-phrase. (3) a. ngikenima: inini negamat I-know-(l-3) man WH-sing-3c I know a man who is singing b. ngikenima: inini ka: 9 nagamat I-know-(l-3) man (WH-past)-sing-3c I know a man who sang c. ngikenima: inini wa: 9 nagamat ke:nagamat I-know-(l-3) man (WH-future)-sing-3c I know a man who will sing d. negamat kino:zi WH-sing-3c tall-3 The one who is singing is tall In (3a) the first vowel of the string /nagamat/ has been change /a/ to lei, as in /negamat/. In (3b) the morphophonological process has affected the past tense /gi: 9 -/ since it contains the first vowel in the verb-phrase, changing it to /ka: 9 -/. Again in (3c) the process affects the first vowel of the verb-phrase, changing the future tense
Questions in Rainy River Ojibwa 163 morphemes /wi 9 -/ and /ka-/ (the latter not to be confused with /ka:-/) to /wa: 9 -/ and /ke:-/. Note again that the relative clauses do not require a lexical antecedent, as in (3d). The addition of either the morphemes /ka:-/ or the changed form will alter the meaning of a simple sentential complement to a relative clause, as can be seen in (4): (4) a. ngikenima: ikwe i:za:t I-know-(l-3) woman go-3c I know that the woman is going b. ngikenima: ikwe ka:i:za:t I-know-(l-3) woman WH-go-3c I know the woman who is going c. ngikenima: ikwe e:za:t I-know-(l-3) woman WH-go-3c I know a woman who is going The semantic distinction between /ka:-/ and the changed form can expressed as definite versus indefinite, as in (4), where (4b) translates as 'the woman', while (4c) translates as 'a woman'. The distribution of /ka:-/ and the changed form is summarized in (5). (5) Present Past Future Definite ka:- ka:-gi: 9 ka:-wi: 9 - In definite changed form (verb stem) ka: 9 - wa: 9 - ke: For convenience's sake, the past and future forms of the changed form are presented as single morphemes since the phonological var iation is constant, unlike the present tense form where the lack of a present tense morpheme results in a variety of stem alternations, e.g., /i:za:t/ becomes /e:za:t/, /nagamat/ becomes /negamat/, etc. I assume the phrase structure rules of (6) for Ojibwa relative clauses: (6) Rule 1: NP * S~ Rule 2: S -» COMP S Rule 1 from Ross (1967) states that a noun-phrase may be expanded into a noun-phrase plus a sentential complement. Rule 2 from Bresnan (1970) states that a sentential complement consists of a COMP
164 Alana Johns node (which is a category that may be filled either with a complementizer or a WH form) plus a sentence. I also assume a WH-movement analysis for relative clauses in Ojibwa. (For a discussion of the criteria for a WH-movement analysis see Chomsky 1977). Although Lees and Pagotto assume /ka:-/ and the changed form to be complementizers, I shall demonstrate that they are rather W H elements. The fact that these morphemes are never found in simple sentential complements where there has been no WH-movement such as (1) would indicate that they are not complementizers. Moreover, the fact that the addition of these morphemes to simple sentential complements results in relative clauses undermines an analysis which assumes them to be complementizers. Were they complementizers, one would expect to find them in all subordinate constructions rather than only in WH constructions. Were they complementizers one would also expect that in cases where WH has moved twice, as in (7), these morphemes would be prefixed to both subordinate verbs. (7) ngikenima: awe inini Bill ka:inendang Mary ki: 9 wabama:- I-know-(l-3) that man Bill WH-think-(3-it)c Mary past-see- I know the man who Bill thinks that Mary saw nit otena:ng (3-obviative)c town-locative in town In (7), the morpheme/ka:-/is prefixed only to the highest verb- Further evidence that /ka:-/ and the changed form are not complementizers is the fact that they occur in relative clauses without lexical antecedents, as in (2d) and (3d). To the best of my knowledge, there exists no language where such constructions contain only a complementizer in COMP. All of these facts indicate that rather than complementizers, /ka:-/ and the changed form are WH elements, and thus I have translated them as such throughout this paper. That they carry the feature definite versus indefinite might be reason to consider them to be some sort of relative pronoun. A complementizer would not carry such a feature. This analysis of /ka:-/ and the changed form as WH forms presents an apparent problem for a traditional analysis of direct questions as consisting of an interrogative (WH) in COMP. The examples in (8) show that direct questions in Ojibwa contain an interrogative morpheme followed by a subordinate sentence containing the changed form.
(8) a. wenen bemipato:t who WH-run-3c Who is running? b. wenen ka: 9 bimipato:t who (WH-past)-run-3c Who ran? Questions in Rainy River Ojibwa 165 c. wenen wa: 9 bimipato:t ke:bimipato:t who (WH-future)-run-3c Who will run? An analysis where both the interrogative morpheme /wenen/ and the changed form (WH) were in COMP would violate the rules of WH-movement since only one WH element may be in COMP. This is because only one WH element in COMP may correspond to the empty category from which WH is moved. In other words, an analysis such as (9) would be impossible, given that the changed form is a WH element. (9) *[ wenen changed form] COMP The obvious answer is that the interrogative morpheme is COMP. Thus direct questions in Ojibwa are constructed exactly the same as relative clauses, i.e., they consist of a noun-phrase followed by a subordinate sentence which has undergone WH-movement. Rule 3 of (10) correctly generates the desired structure while at the same time demonstrating that the entire construction is a sentence rather than a noun-phrase. (10) Rule 3: S^NPS The crucial difference between direct questions and relat in this language is that the former require an obligatory lexical antecedent, i.e., an interrogative morpheme. Evidence that the interrogative morpheme is not in COMP is (11) where the interrogative morpheme /o:gonen/ 'what' is separated from the changed form by a lexical NP. (11) o:gonen inini bekite 9 ang what man WH-hit-(3-it)c What is the man hitting?
166 Alana Johns If both the interrogative morpheme and the changed form were in COMP, example (11) would be a violation of the A/A constraint (Chomsky 1973) which states that an element of a category may not be moved unless the maximal string of that category would be moved. In other words, one would expect that the changed form would be found alongside the interrogative morpheme, not separated from it as it is in (11). The fact that direct questions usually, but not necessarily, contain the indefinite changed form rather than the definite /ka:-/ can be explained by pragmatic aspects of questions. That is, one does not know who, what, etc., is being referred to. Note that this analysis applies to all interrogative elments, who, what, when, where, and why; however, not all will be treated here. Rule 3 does not specify that the antecedent position will be filled only by interrogative morphemes. Thus one should expect to find examples of sentences generated by Rule 3 where the antecedent position is filled by a non-interrogative morpheme. This is exactly what is found in (12), one of the so-called focus constructions. (12) ikwe ka:nagamat woman WH-sing-3c It's the woman who is singing Focus constructions are composed exactly the same as direct que tions and are similar in structure to relative clauses save that the latter are generated from an NP and do not require a lexical antecedent while direct questions and focus constructions are generated from an S and require an obligatory NP in antecedent position. Finally, when one examines an example of an indirect question in Ojibwa, one finds that it is identical in structure to relative clauses, the interrogative morpheme functioning as antecedent to the subordinate sentence. (13) kawin ngikenima:si: (o:gonen) ka: 9 ata:wet not I-know-(l-3)-negative (what) (WH-past)-buy-3c I don't know what he bought Just as with relative clauses, the antecedent position may opti ally be left empty as is indicated by the use of parentheses in (13). Like direct questions, it is usually the changed form, rather than the /ka:-/ which is found prefixed to the subordinate verb. In summary, both direct questions and focus constructions are generated by Rule 3 and produce structures such as (14). Relative clauses and indirect questions are generated by Rule 1 and produce
Questions in Rainy River Ojibwa 167 structures such as (15). All of these constructions involve Rule 2 which expands the subordinate sentence into a COMP node and a sentence. (14) Direct questions and focus constructions: NP COMP wenen changed form inini ka: - nagamat (15) Relative clauses and indirect questions: NP NP COMP inini ka:- wenen changed form nagamat Relative clauses, indirect questions, focus constructions and direct questions all involve a subordinate sentence containing a WH element (either /ka:-/ or the changed form) in COMP. The only difference is that those constructions generated from an require an obligatory antecedent while those generated by an NP do not. It is likely that this difference can be explained by the fact that relative clauses and indirect questions are always within a matrix sentence containing an inflection (theme sign) on the matrix verb which will bind or govern
168 Alana Johns the empty antecedent position in the subordinate sentence. Since direct questions and focus constructions are themselves matrix sentences, the antecedent position is never bound or governed by anything else and therefore it must always be filled. REFERENCES Bresnan, Joan 1970 On Complementizers: Toward a Syntactic Theory of Complement Types. Foundations of Language 6:197-321. Chomsky, Noam 1973 Conditions on Transformations. In A Festschrift for Morris Halle. S. Anderson and P. Kiparsky, eds. New York: Holt, Rinehart and Winston. 1977 On WH-Movement. In Formal Syntax. P.W. Culicover, T. Wasaw A. Akmajian, eds. New York: Academic Press. Johns, Alana 1980 Relative Clauses and Related Phenomena in Ojibway. M.A. thesis, University of Ottawa. Lees, James 1979 A Mini-Grammar of Cree-Montagnais. Montreal Working Papers Linguistics 12:109-148. Pagotto, Louise 1980 On Complementizer Adjuncts in the Rapid Lake Dialect of Algonquin. Pp. 231-246 in Papers of the Eleventh Algonquian Conference. William Cowan, ed. Ottawa: Carleton University. Ross, J.R. 1967 Constraints on Variables in Syntax. Ph.D. thesis, Massachsusetts Institute of Technology.