Microsoft word - i.doc

The Acquisition of Swahili Verbal Morphology
Abstract
Recently, much attention has focused on the so-called Root Infinitive(RI) phenomenon, where children in languages such as German use infinitival verbs in root context, seemingly optionally. English has been argued to be an RI language (Wexler 1994), though English speaking children use bare stems instead of infinitives. Languages such as Italian, however, have been shown not to exhibit RIs in early child language. I present results from a study of Swahili, a Bantu language with rich agglutinative morphology. Swahili is an SVO language in which the verbal clause has the following order of morphemes in an affirmative indicative sentence: Subject Agreement – Tense/Aspect – (Object Agreement) – Verb Root – (derivational Wexler (1994) argues that English children also suffixes) – Mood Vowel. If we assume that the linear produce RIs, just like their German, Dutch and French order of morphemes is a reflection of the hierarchical counterparts. He shows that although they do not order of heads (Baker’s (1985) Mirror Principle), then produce infinitives per se, they produce bare verbs (2), the morphology of Swahili allows us to test various which are missing all obligatory inflectional affixes. theories of the dropping of functional morphemes. The Wexler argues that these forms are the English speech of four Swahili speaking children ranging in age from 1;8 to 3;1 was recorded in Kenya, transcribed in the CHAT format and coded. The results show that Swahili children do not produce RIs, but do omit Agr and T/A markers. Thus the verb may surface with a Person agreement marker alone, a T/A marker alone, or neither Person nor T/A. In the latter case the verb Interestingly, children acquiring Italian (or Spanish or appears as a bare stem, analogous to what we find in Catalan) do not to go through this stage (Grinstead English. These results are discussed in the light of 1998; Guasti 1993/1994). Instead, Italian children, for various theoretical analyses of RIs, specifically the example, seem to converge on the adult grammar Metrical Omission Model (Gerken, 1991), the Small extremely early with respect to this phenomenon. This Clause Hypothesis (Radford 1986; Radford 1990), is surprising since Italian, unlike English, does have a Rizzi’s (1994) Truncation hypothesis, Wexler’s (1994) true infinitive. Thus European languages can be underspecification of tense theory, Hoekstra & Hyams’ classified into two groups: root infinitive languages and (1998) underspecification of number theory, and non-root infinitive languages. Root infinitive languages Schütze & Wexler’s (1996) ATOM model. The basic can further be divided into true RI languages and bare clause typology is found to be most compatible with verb languages. Note that the only well-studied ATOM, although ATOM fails to account for many of language that fits into the bare verb category is English. 1. Introduction
Root infinitive languages
Non-root-infinitive
Children acquiring certain European languages go languages
through a stage in which they use infinitival verb forms in root context. This stage, called the Root Infinitive Stage, or the Optional Infinitive Stage, is characterized by the optional use of utterances such as the those in (1). This phenomenon has been heavily studied in several languages, most notably German (1a. and b., from Poeppel & Wexler (1993)), Dutch (1c. and d., This typology of languages is a problem in our field. from Weverink (1989)), and French (1e. and f., from What is the cause of this typology and how do other languages fit into this typology? Part of the problem is that relatively little is known about children acquiring non-European languages. The goal of this paper is to The verb is embedded within a verbal complex, the broaden the inventory of languages that is used in structure of which is given in (6). I will briefly describe developing acquisition theory, and show that limiting each element in this complex, except for the optional our focus to the well-known European languages leads derivational suffixes, which are not relevant to this to conclusions which do not truly reflect the constraints on child language. I will present results from a longitudinal study of four Swahili speaking children. I (6) Subject – Tense – Object – Verb–suffixes– Mood will first outline some important features of Swahili, and then present results showing how Swahili children use verbal morphology. I will show how Swahili fits into the typology in (3), and discuss the difficulties Subject agreement marks person and number as shown these data pose for several influential theories of early in (7), and there are several tense/aspect markers, given child language : the Metrical Omission Model, the Small Clause Hypothesis, the Truncation Hypothesis, the Underspecification of Number Hypothesis, the Underspecification of Tense Hypothesis, and ATOM (mimi) nilianguka
(Agr-Tense Omission Model). I will show that while (wewe) ulianguka
the data are most compatible with ATOM, the alianguka
complexity of the Swahili data can not be accounted for tulianguka
(ninyi) mulianguka
(hawa) walianguka
2. Swahili
Swahili is an eastern Bantu language spoken primarily in Tanzania, Kenya, Uganda and neighboring areas. Morpheme
The dialect of Swahili in this study is that spoken in and around Nairobi, the capital of Kenya. It is an SVO language; the subject and the object may be optionally null. The verb occurs in a verbal complex which contains inflectional material as well as grammatical (4) Subject - Verbal complex -
Object agreement also marks person and number, but unlike Subject Agreement, is optional. The optionality Most verb roots in Swahili are monosyllabic and of Object Agreement is dependent on the intended contain a single vowel, but as the table in (5) shows, interpretation of the direct object – if the direct object is other verb root structures are not uncommon. The final specific, Object Agreement is obligatory, and if the vowel in the examples in this table are the obligatory Direct Object is non-specific, Object Agreement is final vowel in Swahili, without which the verb cannot obligatorily absent. Because of this optionality and the surface. Each example has the indicative final vowel. complications it poses, I will only discuss Subject Agreement, Tense and Mood in this paper. Mood is marked as a suffix, and is always the final vowel in the verbal complex. This final vowel alternates three ways between the indicative [a], the subjunctive [e] and the negative [i]. The indicative final vowel occurs with on-going actions/states, present habitual actions, past actions/states, future actions/ states, imperatives, etc. The subjunctive is used to express desires, possibility, necessity, requests, etc. Primary stress in Swahili is on the penultimate syllable of any multisyllabic word (Ashton 1947; Maw and Kelly 1975; Myachina 1981; Vitale 1985). Thus in a disyllabic verb, the stress falls on the first syllable of the stem, as in (9). Furthermore, secondary stress usually falls on the subject agreement marker. Examples are given in (10). (10a) is a transitive indicative sentence. The subject is Juma and the object is Mariam, both proper names. The verbal complex shows 3rd person singular subject agreement with Juma, contains the past tense ‘li’, and 3 rd person singular 3. The Data
object agreement with Mariam. The verb root ‘fuat’ is Turning now to the acquisition data, the data collection followed by the Indicative final vowel ‘a’. was conducted over a period of 11 months in Nairobi, Kenya. Biweekly recordings were made of naturalistic speech in the homes of four children of differing ages. The data were audio recorded and transcribed using Juma a – li – m – fuat – a Mariam CHAT format (MacWhinney & Snow, 1985). Because Juma SA3sg-past-OA3sg-follow-IND Mariam of various social and economic difficulties, it was not possible for all four children to remain in the study for the duration of the project. The table in (12) provides the basic descriptive facts about each child. Tafadhali ni – pat – i – e kalamu A – na – tak – a ku – fu – a dafu Because of the complexity of the data and because naturalistic data is notorious for showing great fluctuations between individual data points, we pooled the data in order to stage the children relative to each other. We did not want to use age as an indicator of (10)b is a polite request using the subjunctive final grammatical development as this is unreliable. vowel. When the subjunctive final vowel is used, quite Similarly MLU by itself has been noted to have several expectedly, the tense marker is absent. (10)c is an drawbacks (see Valian 1991). Therefore a composite example of a complex sentence, with an embedded method was used to stage the children, the criteria for infinitive. Note that the Swahili infinitive marker is a prefix which occurs in the same position as other tense markers in tensed clauses. The final example is an imperative form, which is a bare stem, that is, the verb b. Verbs/Total Utterances (Valian, 1991) c. Percent of proto-syntactic place holders Following Chomsky (1993) and Pollock (1989), and (Bottari, Cipriani, and Chilosi 1993/1994) further following Demuth & Gruber (1995) who analyze Sesotho, a southern Bantu language in a similar According to these three criteria, four stages were manner, I assume a structure for Swahili as in (11). identified, with the files from each child being assigned The verb raises to Mood, with all else being base accordingly, shown in (14). Throughout the rest of this generated in their respective projections2. paper, for ease of exposition, I will be using this staged data rather than data from the individual children. Recall that children in other languages such as English omit certain obligatory inflectional elements. Given this, we might expect Swahili children to do the same. Focusing only on the prefixes, and ignoring Object Agreement because of the complications discussed earlier, there are several logically possible clause types that Swahili children might produce. These, along with Examples of each clause type are given in (17). my terminology for them, are listed in (16). Example (a) is a full clause which is adult-like and fully grammatical. Example (b) is a [-SA] clause, that is, it is missing the subject agreement marker, but has a tense marker, in this case a future tense marker. Example (c) is a [-T] clause, that is, it is missing a tense marker, but has subject agreement. Example (d) is a bare stem missing all prefixes, and example (e) is an RI. (15)a. is a full clause, which contains both Subject Agreement and T. (15)b. is a [-SA] clause, containing tense, but missing subject agreement. (15)c. is a [-T] clause containing Subject Agreement but missing tense, and (15)d. is a bare stem, which is missing all prefixes. Let me be clear here on a terminological point: the term ‘bare stem’ is used to refer to a verb root plus an indicative final vowel. Other prefix-less verbs also occur in child (and adult) Swahili, namely subjunctive clauses. These are not discussed in this paper3, although see Deen & Hyams (2001) for details on subjunctive use in Swahili child language. Additionally, it should be noted that bare stems in Swahili are different from bare verbs in English, the latter showing a total absence of inflection, while the former show mood marking. Returning to the possible clause types in (15), putting aside theories of agreement, root infinitives, and theories of infinitives for the moment, and given that Swahili has an infinitive morpheme, root infinitives should be possible in child The verbal utterances in the Swahili corpora were (MacWhinney & Snow, 1985), excluding imperatives, imitations, repetitions and non-indicative utterances. The results are presented in (16). Interestingly, all five (although, the child could have intended ‘I want to possible clause types are attested, although in varying Note that while RIs are attested, they are exceedingly rare (16/1342 = 1%). In fact, the rate of RIs in Swahili is even lower than that reported for languages such as Italian (for which some report rates as high as 16%, So the errors that we see in child Swahili are errors of e.g., Guasti (1993/1994); see Sano (1995) for a review omission, not errors of commission. This is entirely in of proportions of nonfinite clauses in various line with what we know of other languages. languages). This lack of use of the infinitive cannot be 4. Discussion
attributed to the fact that the Swahili infinitive is a These results are intriguing in many respects. They prefix (as opposed to a suffix) because other show a clear developmental trend in the use of the inflectional prefixes are used. Thus, with respect to the inflectional prefixes in Swahili, the likes of which we typology of languages in (3), Swahili falls into the have not seen in any language studied thus far. There same category as English: it is a bare verb language, have been several theories accounting for children’s not a true RI language. Swahili provides us with omission of inflectional material and use of RIs, mostly another example of a bare verb language, making based on languages which lack independent marking of English no longer unique in this respect. As for the SA and T. I will now use this Swahili data to evaluate remaining four clause types, I have plotted their various proportions in a line graph, given in (18). The first is a metrical model of omission, first proposed We see from (18) that in stage 1, all four clause types by Gerken (1991), and later developed in Gerken & are produced at or above 18%. [-SA] clauses and bare McIntosh (1993) and adapted for Sesotho by Demuth stems are more frequent than [-T] or full clauses, but (1994). Gerken found that English children are more not significantly so. In stage 2, [-SA] clauses increase, likely to omit weak unstressed syllables in iambic feet as do full clauses, but bare stems and [-T] clauses as in (19), or weak unstressed syllables preceding decrease significantly. In fact, by stage 2, [-T] clauses trochaic feet (as in 20), but not the weak unstressed occur at a rate of less than 10%. By stage 3, full syllables within trochaic feet (as in 21). clauses become the most prevalent form, followed by [- SA] clauses. Bare stems fall below the 10% mark, as [- T] clauses drop to 5%. This trend continues in stage 4, with the relevant proportions approaching adult norms. Subject to drop The only significant difference between adults and children in stage 4 is the proportion of [-SA] clauses, which is almost 30%. This points to a very clear difference in the grammar of Swahili children between [-SA] clauses on the one hand and [-T] and bare stems on the other. The latter clause types cease to be a possibility relatively early and fade out rather rapidly, while [-SA] clauses remain a possibility even into stage This, she claims, accounts for the omission of determiners and other unstressed functional material in English. Demuth (1994) applies this theory to the omission of noun class marking in Sesotho, and suggests that it may also explain the omission of verbal prefixes. However, the Swahili data are incompatible with a simple metrical model. Recall that stress in the Swahili verbal complex is always on the penultimate syllable, and secondary stress in the verbal complex usually falls on the subject agreement marker. Taking a leftward parsing of the string produces a trochaic foot It should be noted that while subject agreement and on the right edge containing the verb and the mood final tense are used optionally, when they are used errors are vowel. Since most verb stems in Swahili are disyllabic, extremely rare. In all the verbal utterances included in this is the most common pattern. Furthermore, the (16), there were a total of 9/1342 errors: less than 1%. Subject Agreement and Tense prefixes are analyzed together as a trochaic foot , yielding two trochaic feet. others. Such a hypothesis claims that the adult axiom According to the Metrical Model, in cases such as (22) of CP=root has not developed in young children, and so when the verb stem is disyllabic there should be no they can optionally choose to specify the root as some omission. If we now consider trisyllabic verb stems lower projection in the tree. When AgrOP, for example, is specified as the root, then everything above AgrOP is omitted, and if TP is specified as the root, everything above TP is omitted. This theory predicts that nothing from the middle portion of the tree will be omitted. For Swahili it predicts that children may produce full clauses (that is, no truncation because the we see that the rightmost trochaic foot includes the root is CP or AgrSP), they may produce [-SA] clauses second syllable of the verb stem and the mood final (when the root is TP), they may produce Bare stems vowel. The first syllable on the verb stem forms (when the root is either AgrOP or MoodP), or they may extrametrical information, and so should be subject to produce Bare Verb Roots (when the root is VP). omission. However, crucially, the subject agreement and tense markers form a trochaic foot, and should not (24) Predictions of Truncation for Swahili: The vast majority of verbs produced by the children in this corpus were disyllabic or trisyllabic, and so a metrical account falls short of explaining the omissions. Furthermore, children tend to center on the verb stem and produce the verb stem correctly, irrespective of As mentioned earlier, verbs missing the final mood syllabic structure5. What they seem to have problems vowel are unattested in child Swahili, and so (24) d is with more than anything are the inflectional prefixes. problematic for truncation. This could be accounted for Metrical theory cannot account for this specific by postulating a lower limit to truncation, although no such proposal exists for any other language that I am Several researchers (such as Radford (1986), Lebeaux aware of. However, a much more serious problem for (1988); Guilfoyle & Noonan (1988), etc.) looking truncation is the [-T] clauses. These are problematic mostly at English speaking children’s bare verbs and because they contain material from low in the tree and omission of determiners have proposed that very young material from high in the tree, but are missing material children lack functional structure entirely. Given the from the middle portion of the tree – precisely what Swahili structure in (11),we expect Swahili speaking children to produce no prefixes and no mood final In addition to these theories, there have been several vowel. However, interestingly, Swahili children never theories of underspecification which attempt to account omit the final mood vowel. Furthermore, from the for RIs. One such theory was proposed by Hoekstra & earliest stages they use the final vowel appropriately. Hyams (1998), in which they propose that only number This dichotomy between Subject Agreement and Tense is optionally underspecified in the grammars of young on the one hand and Mood on the other is evidence children. With number underspecified, children which a No Functional Structure Hypothesis simply acquiring number-marking languages such as Dutch or cannot accommodate. Furthermore, such a hypothesis English use an infinitival form or a bare stem does not account for the differential omissions of the respectively. Languages such as Italian are person marking languages, and so do not exhibit RIs or bare Similarly, a theory that claims a single functional verbs. According to their definitions, Swahili is a projection above VP, such as that proposed by Clahsen person marking language, and so an underspecification et al.(1996) in which there is a single unitary of Number theory predicts that Swahili children should inflectional projection which later splits into two or not omit any functional elements. This is clearly not more projections, predicts that Swahili children should use one and only one prefix at any one time. However, A second underspecification theory is that proposed by as we saw in (18), Swahili children produce full clauses Wexler (1994), in which tense is underspecified. which contain both Subject Agreement and T. However, such a theory predicts that only tense should Furthermore, they produce bare stems, which contain be omitted, while we see that in Swahili agreement is neither Subject Agreement nor T. This is not predicted The final underspecification theory I will consider is Let us turn now to a Truncation-type hypothesis such as proposed by Schütze & Wexler(1996) and in more that proposed by Rizzi (1994), and subsequently by detail in Schütze (1997). They claim that children can optionally omit either Agr, tense, both or neither. This difference from other inflectional heads. This model, called ATOM (Agr-Tense Omission Model) interesting fact will be the focus of Deen & Hyams predicts that Swahili children may produce full clauses, [-SA] clauses, [-T] clauses and bare stems, as shown in One conclusion we can draw from this study is that our previous understanding of the typologies in child language were overly simplistic and based on languages that confounded various inflectional properties. Languages like Swahili that mark inflectional heads distinctly can be enormously useful in our understanding of the development of child grammar and how inflection is used in their grammars. ATOM appears to make the correct predictions with respect to the various possible clause types, but several Acknowledgments
questions remain. For example, how does this model My deepest thanks go to Nina Hyams, without whom account for the differing proportions of the four clause none of this would have surfaced in the form that it did. types? Specifically, what accounts for the difference Thanks also to Carson Schütze, for long and detailed between [-SA] clauses on the one hand and [-T] clauses discussions and comments; the UCLA Psychobabble and bare stems on the other? Secondly, Swahili group for their comments and discussion, and children have shown that they do underspecify Agr and Dominique Sportiche for his help and insight into the T, and Swahili does have an infinitive marker, but why syntax of Swahili and Bantu languages. Of course, all do Swahili children produce bare verbs like English children and not root infinitives like German children? Finally, since Swahili is a null subject language which has rich person agreement, why does it not pattern like 1 The paradigm given in the text is for animate (usually of Italian? ATOM does not answer these questions, and classes 1 and 2) nouns. SA in Standard Swahili is somewhat so while the basic range in clause types is predicted, the more complicated in that it has a distinct SA system for each noun class. Nairobi Swahili, on the other hand, has a reduced SA marking system for inanimate nouns. The simplified view I have presented in the text is to avoid this complication, as it 5. Conclusion
has no bearing on the issues in this paper. See Deen Unlike other European languages which either mark just one inflectional property, or conflate several 2 This is the structure and analysis that Demuth & Gruber properties into a single head, Swahili marks Subject (1995) propose for Sesotho. While such an analysis raises Agreement and Tense individually on independent certain problems with the licensing of inflectional elements, I morphological heads. Because of this characteristic, will continue to assume it for the purposes of this paper. For Swahili offers us a new window into child language, a more detailed analysis of Swahili, see Deen (1999) and revealing several interesting facts. First, we see that Deen (forthcoming), chapter 2. See also Ngonyani (1996) for evidence of verb raising and the positioning of Mood in children may optionally and independently omit Subject Agreement and Tense. Secondly, when inflectional 3 The subjunctive occurs with a SA marker, but does not material is used, it is overwhelmingly used correctly. occur with T. Therefore, these clauses differ from [-T] This is not surprising, as this is a hallmark of child clauses only in the final vowel. Methodologically, these pose language. Third, RIs are unattested. This is somewhat a problem if we are trying to ascertain the nature of [-T] surprising since Swahili does have an infinitive marker. clauses, since it is possible that children are producing One possible reason for this is that Swahili, like subjunctive forms with an incorrect final vowel. Apart from English, marks the infinitive as a prefix, and perhaps the fact that such errors of commission are extremely rare in the position of the infinitive leads to a dropping of the child language in general, these clause types are very clearly distinguishable on the basis of context. Subjunctives very infinitive prefix. However, since other inflectional clearly indicate an irrealis-type meaning, while presumably, prefixes are used in varying proportions, why would the the [-T] forms should not since they are marked indicative. infinitive be completely absent if Swahili were indeed This interpretive difference was used as a criterion in counting, with all those utterances that could possibly be Furthermore, despite Swahili being a pro-drop language considered a subjunctive (due to their irrealis interpretation) with rich person agreement, bare verbal stems still occur. This is surprising given what we know about A possible objection is that other parses are possible. Italian and Spanish, where such verb forms do not However, there are two conditions which rule out other occur. Finally, the mood final vowel is used correctly significantly different parses. First, every foot must contain one and only one strong syllable, limiting the string in (23) to without omission from the earliest stages, a marked two feet. The first strong syllable (left-most) can only be Gerken, LouAnn, and B. McIntosh. 1993. The interplay of parsed as trochaic, while the second foot can be parsed as function morphemes and prosody in early language. either trochaic or iambic. However, the second principle rules Developmental Psychology 29:448-457. out an iambic parse: as Gerken points out, children have a Grinstead, John. 1998. Subjects, Sentential Negation and preference for the metrical pattern exhibited in the ambient Imperatives in Child Spanish and Catalan. Doctoral language. So English children prefer trochaic feet (and will parse an ambiguous string as trochaic, not iambic) because Guasti, Maria Teresa. 1993/1994. Verb syntax in Italian adult English is a trochaic language. French children, on the Child Grammar: finite and nonfinite verbs. Language other hand, prefer iambic feet because adult French is an iambic language. Adult Swahili is a trochaic language, hence Guilfoyle, E., and M. Noonan. 1988. Functional categories children prefer trochaic feet. Therefore, the only possible and language acquisition. Canadian Journal of 5 One notable exception is the verb anguka meaning ‘to fall’. Hoekstra, Teun, and Nina Hyams. 1998. Aspects of Root More than one child produced on several occasions utterances Infinitives. Lingua 106 (1-4):81-112. Lebeaux, David. 1988. Language Acquisition and the form but the verb stem itself has been reduced. Such cases, I of grammar. Doctoral dissertation, UMASS. believe, allow for a metrical explanation. MacWhinney, Brian & Catherine Snow. 1985. The Child Language Data Exchange System. Journal of Child References.
Ashton, Ethel Osteli. 1947. Swahili Grammar. London: Maw, Joan, and John Kelly. 1975. Intonation in Swahili. Surrey, England: Unwin Brothers Limited. Baker, Mark. 1985. The Mirror Principle and Morpho- Myachina, Ekaterina Nikolaevna. 1981. The Swahili syntactic Explanation. Linguistic Inquiry 16:373-416. language: a descriptive grammar. London: Routlege & Bottari, Piero, Paola Cipriani, and Anna Maria Chilosi. 1993/1994. Protosyntactic Devices in the Acquisition Ngonyani, Deogratis. 1996. The Morphosyntax of of Italian Free Morphology. Language Acquisition 3 Applicatives. Doctoral dissertation, UCLA. Poeppel, David, and Kenneth Wexler. 1993. The full Chomsky, Noam. 1993. The Minimalist Program. In The competence hypothesis of clause structure in early view from building 20, edited by K. Hale and S. Keyser. Pollock, Jean-Yves. 1989. Verb Movement, UG and the Clahsen, Harold, Sonja Eisenbeiss, and Martina Penke. structure of IP. Linguistic Inquiry 20:365-424. 1996. Lexical learning in early syntactic development. Radford, Andrew. 1986. Small children' s small clauses. In In Generative perspectives on language acquisition: Bangor Research Papers in Linguistics. empirical findings, theoretical considerations and Radford, Andrew. 1990. Syntactic theory and the acquisition cross-linguistic comparisons, edited by H. Clahsen. of English Syntax. Oxford: Basil Blackwell. Philadelphia PA: John Benjamins Publishing Company. Rizzi, Luigi. 1994. Some notes on Linguistic theory and Deen, Kamil Ud. 1999. Swahili verbal complexes: language development: the case of root infinitives. Morphology from Syntax. Qualifying Paper, UCLA. Deen, Kamil Ud. forthcoming. Inflectional prefixes, mood Sano, Tetsuya. 1995. Roots in Language acquisition: A and null subjects: The acquisition of Swahili verbal comparative study of Japanese and European morphology. Doctoral Dissertation, UCLA. Deen, Kamil Ud, and Nina Hyams. 2001. The form and Schütze, Carson. 1997. INFL in child and adult language: interpretation of non-finite verbs in Swahili. Paper Agreement, Case and licensing. Doctoral dissertation, presented at Boston University Child Language Schütze, Carson, and Kenneth Wexler. 1996. Subject case Demuth, Katherine. 1994. On the underspecification of licensing and English Root infinitives. Paper read at functional categories in early grammars. In Syntactic 20th Annual Boston University Conference on Theory and first language acquisition: cross-linguistic perspectives, edited by B. Lust, M. Suner and J. Valian, Virginia. 1991. Syntactic subjects in the early speech Whitman. Hillsdale, NJ: Lawrence Erlbaum Associates. of American and Italian children. Cognition 40. Demuth, Katherine, and J. Gruber. 1995. Constraining XP Vitale, Anthony. 1985. Aspects of Kiswahili Stress. sequences. In Niger-Congo Syntax and semantics 6, Kiswahili: Journal of the Institute of Kiswahili edited by V. Manfredi and K. Reynolds. Boston: Boston University African Studies Center. Weverink, Meike. 1989. The subject in relation to inflection Ferdinand, Astrid. 1996. The development of Functional in child language. MA thesis, University of Utrecht. categories: The acquisition of the subject in French. Wexler, Kenneth. 1994. Optional Infinitives, head movement, and economy of derivation. In Verb Gerken, LouAnn. 1991. The metrical basis of chidlren' s Movement, edited by N. Hornstein and D. Lightfoot. subjectless sentences. Journal of Memory and Cambridge, UK: Cambridge University Press.

Source: http://sfs.scnu.edu.cn/tblogs/liuhy/attachments/month_1005/g20105701811.pdf