An Adversity Passive Analysis of Early Sesotho Passives

An adversity passive analysis of early Sesotho passives: Reanalyzing a counterexample to Maturation 1. Introduction One of the goals of language acquisition research is to discover what linguistic abilities children bring to the language learning process. In their Maturation Hypothesis, Borer and Wexler (henceforth B&W 1987) suggest delayed production of verbal passives and other Amovement constructions in child English and child Hebrew is an indication that grammar is not fully mature at birth and that over time it matures to become adult-like. As they believe it to be a biological deficiency, B&W claim the passive-like constructions children employ before the age where this ability matures (around 4;0) will not be verbal passives, but other constructions that do not have A-movement, such as lexical adjectival passives in English. More generally, children will produce passive-like constructions at an early age only if the language has a syntactic homophone (s-homophone), a construction homophonous with verbal passives, but without A-movement syntax (Babyonyshev et al 2001). Experimental results such as those found in Babyonyshev et al (2001) show that when no s-homophones are available for Amovement constructions in the language, children perform poorly on elicitation tasks. While crosslinguistic acquisition studies showing delayed have supported the Maturation Hypothesis (Japanese, Sugisaki 1998; Spanish, Pierce 1992), other crosslinguistic studies have shown the opposite case. One such counterexample comes from the study of the Bantu language Sesotho (Demuth 1989, 1990, 1992). This acquisition study of spontaneous production data suggests Sesotho-speaking children may acquire the passive as early as 2;8. To provide support for this claim, Demuth shows verbal passives in Sesotho have no s-homophones with adjectival passives, and that children productively alternate between active and passive forms of the verbs. Recent studies, however, have shown that although Sesotho verbal passives may contain Achains, the evidence for productive passives in Sesotho is not particularly strong (Crawford 2004). As the evidence is not strong and there is no experimental data to corroborate the findings in the spontaneous production data, it is possible Demuth may have had too narrow a view of what can be considered an s-homophone in this language. This paper explores the suggestion (Wexler 1999, Babyonyshev et al 2001) that early Sesotho passives are similar to Japanese adversity passives, a construction which Japanesespeaking children acquire early (Sugisaki 1998). The idea that early Sesotho passives are adversity/malefactive constructions and can be acquired early is supported by the literature; early passives in other Bantu languages express situations where the patient is negatively affected (e.g. Zulu; Suzman 1987), and adversity constructions in Japanese do not contain the A-movement that is believed to be the difficulty for children in other verbal passives (Miyagawa 1989, Kubo 1990). Importantly to the present analysis, Pylkkänen (2002) suggests that Japanese adversity passives are applicative (ditransitive) constructions with passive morphology. As I will show, an adversity construction exists in adult Sesotho, and these constructions fit Pylkkanen’s typology in a similar way to Japanese adversity constructions. Furthermore, I will show that depending on the interpretation of an ambiguous perfective/applicative morpheme, the Sesotho acquisition data contains similar constructions, indicating adversity constructions may be a source of homophony for the child. 1 The organization of the paper is as follows: Section 2 gives a brief review of the evidence for maturation and s-homophones, followed by a review of the Sesotho data and its critiques in Section 3. Section 4 will address the suggestion Sesotho passives are adversity passives, showing the relevant Sesotho adversity construction and the child data that supports an adversity analysis. Section 5 concludes the paper, providing some discussion of how the finding relates to recent versions of the Maturation Hypothesis. 2.0 Background The phenomenon addressed in this paper is the acquisition of the verbal passive. In the verbal passive construction given in (1), the object is moved to subject position by means of an A-chain. The passive morpheme absorbs accusative case, and the original subject is demoted to an oblique position known as the by phrase. Verbal passives with the by phrase are known as long or full passives, and passives without them are known as short or truncated passives. In contrast to verbal passives, adjectival passives (1c) are formed lexically and do not have syntactic movement. (1) a. The sandwichi was eaten ti by John. b. The sandwichi was eaten ti. c. The eggs were hard-boiled/scrambled/fried Early studies (e.g., Horgan 1978, Maratsos, Fox, Becker, and Chalkey 1985) showed that the verbal passive was acquired relatively late, from age 4;0 to as late as the school years. During the same age frame as delayed passives, children were also misanalyzing unaccusative constructions, which also have A-movement. These findings, in addition to observing that adults were not systematically withholding such constructions from the input, led B&W to propose children’s early grammar lacked object to subject A-movement, and that the ability matured as the child did, around age 4;0. In addition to the English data, evidence of delayed verbal passives in Japanese (Sugisaki 1998) and Spanish (Pierce 1992) supported this hypothesis. Despite these findings, passive constructions were not completely absent from the child data. Passives were found in English diary studies (Bowerman 1990), but Horgan (1978) noted that the types of passives used were of a certain type: mostly short passives with actional verbs that described the state of things. The stative semantic nature led B&W to suggest that these early passives were not verbal, but rather adjectival in nature. Adjectival passives were allowed in early grammar because they do not have the movement that causes problems for the child. The two types of passives are easily mistake in English because although they have different syntactic structures, truncated verbal passives are homophonous adjectival passives. Babyonyshev et al (2001) later formalized this notion, calling it an instance of syntactic homophony: (2) Syntactic homophone: A is a syntactic homophone of B if A and B have different syntactic constructions but similar pronunciation. The s-homophone account explains two phenomena in the child data, namely when early passives can occur, and when they cannot occur. Experimental studies show that when no shomophones are available for A-chain constructions in a language, children perform poorly on 2 tasks eliciting them (See Babyonyshev et al’s (2001) study on the Russian genitive of negation on unaccusative NPs). In short, B&W claim children lack the grammar to form verbal passives, and if passive like constructions are present in production data, it is an indication children are employing an s-homophone for verbal passives in the language that does not have A-chain syntax. Despite the evidence for delayed passives and s-homophones, acquisition data from the Bantu language Sesotho challenges these findings. Demuth (1989, 1990, 1992) claims Sesotho speaking children acquire the passive as early as 2;8, and that the data differs from English in that Sesotho verbal passives do not have s-homophones with adjectival passives. The Sesotho data poses a problem for the maturation hypothesis, since, if analyzed correctly, it would indicate that children have access to the verbal passive construction at an early age. Crawford (2004) shows that the Sesotho evidence is not strong enough to contradict the maturation hypothesis. As suggested, although Sesotho passives are not s-homophones with adjectival passives, children may be using other homophonous constructions to form early utterances with passives. The following section reviews the original Sesotho passive data and its critiques in turn. 3.0 Demuth (1989) study and Crawford (2004) critiques Before we address the Sesotho child data, a brief description of Sesotho is necessary. Sesotho is a southern Bantu language of the Nguni group with SVO word order. The Bantu verb is made up of a complex of several elements, shown in (3): prefixed to the verb root in a fixed order is a subject marker (SM), which agrees with the noun class of the subject, followed by tense and aspect (T/A, with the exception of perfective aspect), and object agreement (OA), which is optional. Verbal extensions, such as the causative (CAUS), applicative (APPL), reciprocal (REC), and passive (PASS) extensions, perfective aspect (PERF), and a final vowel (FV) indicating mood, are suffixed to the root. (3) SM-T/A-OA-verb root-(CAUS-APPL-PERF-REC-PASS)-FV It is important to distinguish between the original study and the reanalysis. Demuth’s original study examined data from 4 children, ranging in ages from 2;1 to 4;1. The children’s data was pooled randomly by age. The relevant passive data is presented in Table 1. Table 1: Passive data used in Demuth (1989) Interval 1 2 3 4 5 6 Age (y;m) 2;1-2;3 2;4-2;6 2;7-2;9 2;10-3;2 3;9-3;10 4;0-4;1 # of utterances # of passives 1704 2925 3307 3159 1520 1603 386 6 11 33 27 32 30 23 .4% 1% .9% 2.1% 1.9% 6.0% Passives/total .4% utterances 3 Adult caregivers Crawford (2004) had two methods of reanalysis: A reanalysis of the validity of the 139 tokens shown in the above table, and an individual examination of 2 of the 4 children, Child H, and Child L1. Table 2: Sesotho Child and Adult passives in Crawford (2004) Child H # of utterances # of passives passives/total utterances Child L Adult 1093 1585 3138 38 63 138 3.5% 4% 4% First, Demuth gives an A-movement analysis for Sesotho verbal passives, showing that the data is relevant for examining the maturation hypothesis. Demuth describes two types of passives: those that have A-movement, and those that do not. She claims Sesotho verbal passives have the same A-movement as verbal passives in English: She proposes that the object is promoted to subject position by means of an A-chain, illustrated in (4b). The passive morphology in Sesotho is transcribed in two ways: orthographically as –o- or -uo-, or phonetically as [w] or [uw]. The logical subject may optionally appear as the object of a by phrase, marked in Sesotho by ke. Demuth’s diagnostic for movement is that the subject marker agrees with the noun class of the promoted object. The second type is the impersonal passive, where the subject is an expletive from noun class 17. Adding an expletive in subject position for the impersonal passive precludes promotion of an argument to subject position, and therefore does not involve A-movement. (4) a. Thabo o-pheh-il-e lijo name SM-cook-PERF-FV food Thabo cooked some food b. Lijoi li-pheh-il-o-e ti (ke Thabo) Food SM8-cook-PERF-PASS-FV (by T) The food was cooked (by Thabo) (5) Ho-pheh-il-o-e lijo SM17-cook-PERF-PASS-FV food There has been cooked food Demuth 1989:59 Demuth 1989:59 Since impersonal passives do not require A-movement, they should not be considered in the early A-chain passive counterevidence2. Crawford shows that 30 of the 139 tokens (22%) in the 1 It is important to keep in mind that the two methods of analysis use different amounts of passive tokens for both children and adults. As will be seen, Demuth’s methods for choosing relevant productive child passives is suspect, so Crawford included all utterances that contained a passive morpheme. Crawford increased the amount of adult utterances to match closer the number child utterances, and included both child directed speech and adult to adult speech. Results from the original study, reanalysis of the original study, and results from the Crawford study using the different method will be indicated as much as possible. 2 There is also the question of whether or not Sesotho impersonal passives have covert A-movement or long distance agreement. As Babyonyshev et al (2001) show, Russian children have difficulties with constructions that 4 Demuth study are impersonal passives and should not have been included as counterevidence to B&W’s hypothesis. Crawford further suggests non-movement passives may be more prevalent in the early data than passives that involve A-movement; an individual analysis shows that Child L uses impersonal passives over half the time. Table 3: Number of impersonal passives (Crawford 2004) H L Total passives 38 63 Impersonal passives 4 28 % impersonal/total passives 8% 44% Adult 138 66 48% Secondly, Demuth (1989:60) claims the early passive data does not have s-homophone with lexicalized or adjectival forms. In other words, with the exception of the impersonal passives, the passive forms Sesotho-speaking children are producing are productive forms, and are not lexicalized forms without A-chains. Crawford shows that although this does not permit an adjectival analysis for the early passives, this does not mean that a lexical analysis is entirely ruled out; it is possible that the early passives are lexical forms in the sense of being repetitions or being rote produced, where the child does not analyze the passives as having active counterparts. Demuth suggests that children may initially acquire passives as lexical items, separate from their active counterparts, before they realize them as a form that involves movement. The Sesotho verbs –fa- ‘give’, -rekela- ‘buy for someone’, -etsa- ‘do/make’ and shapa- ‘lash’ are commonly passivized in everyday discourse and constitute a large proportion of the passives that children use before age 2;7 (1989:65). Additionally, there are idiomatic passives in Sesotho which do not have active counterparts. Passive forms of the verb –tswara- ‘grab’ are used to describe a physical state, where the physical state carries some ‘agent’ properties that ‘grab’ a person. For instance, a person is ‘grabbed by sleep’, which is the way to express that a person has become sleepy (6)3. This expression does not have a counterpart with the active verb ‘grab’. (6) Ke-tswer-w-e ke boroko SM-grab-PASS-FV by sleep I’m sleepy (lit., I am grabbed by sleep) Demuth correctly did not include idiomatic passives as a part of the tokens in the original study. However, Demuth did not control for other possible rote forms in the study. For example, Crawford notes some of the passives in the original study were passives previously modeled in the discourse4. Moreover, an analysis of Demuth’s passive tokens shows that these verbs like ‘give’ ‘buy’ and ‘lash’ are quite common not just before 2;7, but in the passive data at all ages. involve covert A-movement. In locative constructions in Sesotho, impersonal ho must be in subject position. It does not agree with the noun class of the locative NP. It is ungrammatical to put the locative subject marker on the verb. This needs further analysis, but it is likely that Sesotho impersonal passives are not an example of long distance agreement. 3 Other grab expressions in the child data include ke tswerwe ke matlafa, ‘I am grabbed by urine’ (I have to pee) and ke tswerwe ke boroko, ‘I am grabbed by excrement’ (I have to defecate). 4 Those marked with (M) in the Demuth appendix (1989:68-79). 5 22% (31 out of 139) of the passives in her child data are passives using these verbs5. It is difficult to know with spontaneous production data whether passive utterances with common verb forms or those based on previous discourse constitute productive forms or if the passives are acquired as separate lexical items. Regardless, another 22% of the original passives are suspect as far as being relevant and productive counterexamples. Demuth’s third claim is that Sesotho-speaking children are using verbal passives productively. Her test of productivity is demonstrating a passive/active alternation, i.e., by showing that a child can use both the passive and active forms of a verb in alternation to one another. In (7) and (8), a Sesotho-speaking child (age 2;8) uses the verb -shapa- ‘lash’ in both active and passive forms. She is telling her little brother that he will be beaten: (7) (8) ho-th-o-e o-tla-shap-uo-a! SM17-say-PASS-FV SM-FUT-lash-PASS-FV It is said that you will be lashed! Demuth 1989:64 ke 'me o-re o-tla-o-shapa (ke 'me ea-re-ng o-tla-o-shapa) COP mother REL-say-REL SM-FUT-OA-lash-FV It's mother who says she will lash you Demuth 1989:64 Crawford shows that this example is perhaps anecdotal. An analysis of Child L’s passive verb forms6 shows that she does not systematically alternate between the active and passive forms of the verb7. Table 4: Active/Passive Alternation L (=50 total passive utterances) TYPE ENGLISH SESOTHO cut off Khaola Sin against sit(ets)a non alternating teach Ruta dress ten(ts)a win Hlola close Kwala possible alternation tear Taboa TOTAL PASSIVE ACTIVE 3 1 3 1 7 1 2 18 0 0 0 0 0 1 2 3 Finally, Demuth considers an alternate hypothesis to explain her data: a frequency analysis. Demuth suggests the frequency of the passive construction in adult Sesotho is greater than that of English-speaking adults, and this explains why Sesotho child passives are acquired earlier than 5 The tokens I am referring to in the Demuth appendix (1989: 68-79) are as follows: Child H: 1c, 1d, 2a, 3a, 4a, 4b, 4c, 4h, 6e, 6o Child L: 1a, 2a, 2c, 5c, 7a, 7b Child K: 1b, 1c, 2e, 2f, 2g, 2i, 2k Child T: 1d, 1f, 2d, 2v, 3f, 4e, 4f, 4o I include –etsa- ‘do/make’ as one of the verbs that is used frequently in the discourse. The list in Crawford (2004) has a list that includes all of the above verbs as well as the number of uses of idiomatic –tswara- ‘grab’. 6 With the exception of the verbs just mentioned: idiomatic grab, buy, do/make, give, and lash. 7 Child H shows a similar pattern. See Crawford (2004). 6 English-speaking children’s. She shows for example that in Sesotho, subjects may not be questioned in situ (9). Because of this topical subject constraint, passives play a very important role in Sesotho grammar: to question subjects or answer subject questions, she claims one must use a passive construction, as in (10) (1989:68, emphasis mine). (9) (10) *Mang o-o-shap-il-e? who SM-OA-lash-PERF-FV Who lashed you? Demuth 1990:72 O shap-il-o-e ke mang? SM-lash-PERF- PASS-FV by who You were lashed by who? Demuth 1990:72 Demuth suggests children hear a lot of passives from child directed questions: 73% of passives in the original adult sample were subject wh questions. Crawford points out that this claim about the adult data may be exaggerated, and furthermore, the obligatory passive analysis is not quite correct. There are three other ways to form a subject wh question, all of which use active verbs. For example, subject questions can be formed as a cleft relative construction as in (11) (Demuth 1990)8. Sesotho speakers also allow for an additional focus type construction as in (12) and (13). The copula with question word is postverbal, in a similar word order to the passive construction in (10), but an active form of the verb is used. Furthermore, subject questions can be a series of topic/comment type constructions with only the copula, with no inflected verb at all, like in (13). (11) (12) (13) Ke mang ea o-shap-il-e-ng? COP who REL OM-lash-PERF-FV-REL It’s who that lashed you? Demuth 1990:72 he u-tla-theol-o ke mang ? IJ SM-FUT-bring.down-FV COP wh ? Hey, who will take you down ? Adult question in H’s transcript ke ya mang ratiyo ? COP POSS wh radio Whose radio is it? Lit., it’s of who, radio? Adult question in L’s transcript With a larger sample of adult data, it appears as if adults do not use the passive in subject whquestions as frequently as Demuth reported. In fact, Crawford shows that 85% of subject wh questions adults use are types that employ active verbs. Children, accordingly use almost none. 8 The remark in Demuth (1989) is contradicted by the remark in Demuth (1990). Demuth (1990) does not acknowledge these other forms as forms with active verbs, nor that adults and children have the option It is important to notice that the by phrase marker ke in (55) is homophonous with the copula ke in (56). 7 Table 5: (from Crawford 2004) Adult H L Copular subject wh questions 52 61% 3 60% 12 60% Focus subject wh questions 12 14% 0 0% 4 20% 9 11% 1 20% 2 10% 12 14% 0 0% 1 5% 0 0% 1 20% 1 5% Cleft relative subject wh questions Passive subject wh questions Gloss Unclear Total subject wh questions 85 100% 5 100% 20 100% To summarize, Demuth’s original data does not provide a strong counterexample to the maturation hypothesis. The passive tokens do not support movement, productivity, and frequency arguments, and do not contradict the maturation hypothesis. As the literature suggests, children will form passives at an early age only if the language has a construction homophonous with verbal passives but without verbal passive syntax. The Sesotho case shows that children are using passive constructions, but not in a productive manner. Demuth may have had too narrow a view of s-homophones for verbal passives in Sesotho. The early utterances are likely homophonous with some other construction, and the next section describes one suggestion as to what these early passive like constructions might be. 3.0 Adversity Passives 3.1 Background As an early explanation of the Sesotho data, Wexler (1999) suggested early Sesotho child passives were not verbal passives of the relevant type, but forms similar to Japanese adversity passives constructions. Japanese adversity constructions usually introduce a malefactive (as opposed to benefactive) argument, one that is interpreted as being adversely affected by the action. These constructions usually use verbs like hit and kick. The adversity construction in (15) usually has one more argument than verbal passives, like the one in (14). (14) Sensei-ga seito-ni ker-are-ta Teacher-NOM pupil-by kick-PASS-PAST The teacher was kicked by the pupil (15) a. Sensei-ga seito-ni kuruma-o ker-are-ta Teacher-NOM pupil-BY car-ACC kick-PASS-PAST The teacher is such that his car was kicked by the pupil Sugisaki 1998:2 Washio, in Sugisaki 1998:2 There are two reasons for this suggestion. First, Suzman (1987) observed that Zulu-speaking children’s early passives tend to be with verbs like hit and kick, where the patient is negatively affected by the action. She proposed that early passives are passive prototypes that can only be used when the patient is negatively affected by the action. The verbs Sesotho children used to make passives seemed to have this affected quality. Second, Sugisaki (1998) found that Japanese children acquire adversity passive constructions earlier than verbal passive constructions. This would be expected under the usual analysis of adversity passives, which 8 holds that they do not involve object-to-subject A-movement. Miyagawa (1989) argues a numeral quantifier and the NP it modifies must be in a mutual c-command relationship with one another. This ‘quantifier float’ is possible from the surface subject of a verbal passive, but not adversity passives. (16) (17) Tomodachi-ga futari Shinjuku-de [vp Tanaka-sensei-ni atta.] friends-NOM 2-cl Shinjuku-in Prof. Tanaka-DAT met ‘Two friends met Prof. Tanaka in Shinjuku.’ Miyagawa 1988:134 *Kodomo-ga [VPame-ni futari fur-are-ta.] children-NOM rain-by 2-cl fall-PASS-PAST Two children were rained on. Kubo, in Sugisaki 1998:3 Miyagawa argues that the difference is captured if one assumes the mutual c-command requirement is satisfied between the numeral quantifier and the trace coindexed with the surface subject. Kubo (1990) takes this observation one step futher, attributing the ungrammaticality of quantified subjects in adversity passives to the fact that adversity passives do not have object to subject movement. 3.2 Further support for Sesotho adversity passive analysis There is other information in the literature that support the idea that Sesotho child passives are Japanese adversity constructions rather than verbal passives. First, adult Bantu languages do have constructions similar to the Japanese adversity passives. Consider the following examples from Kichaga and Sesotho. Both take one more argument than a non-affected verbal passive: (18) M-kà n-1! a-1i-lyì-í -ò k-èlya. CL1-wife FOC-AGR-TNS-eat-APPL-PASS CL7-food The wife is being benefited/adversely affected by someone' s eating the food. Bresnan and Moshi 1990, in Woolford 1995 (19) a. Monna a-sho-ets-e Lineo. Man AGR-die-APPL-FV Lineo ‘The husband has died on Lineo.’ b. Lineo o-sho-ets-o-e ke-monna. Lineo AGR-die-APPL-PASS-FV by-husband ‘Lineo has been bereaved of the husband.’ Woolford 1993:31 Second, while these examples contain the passive morpheme, Pylkkänen (2002) argues that Japanese adversity passives are not verbal passive constructions, but applicative (ditransitive) constructions. This claim is based on her typology of applicative constructions. She argues there exists two types of ditransitive constructions, one that relates an individual to an event described by the VP (high applicatives) and one that relates an individual to the direct object (low applicatives). This difference in event semantics has consequences for the syntax, as in (20): 9 (20) She shows that in English, a benefactive can only be added to a transitive verb, whereas in a language like Kichaga or Sesotho, a benefactive can be added to an unergative verb (from Bresnan and Moshi 1993: 49-50): (21) English: a. I baked a cake. c. I ran. b. I baked him a cake. d. *I ran him. (i.e. I ran for him) (22) Chaga: b. N-˝a-i-zrìc-í- à mbùyà. FOC-1s-PR-eat-APPL-FV9 friend ‘He is running for a friend’ Pylkkänen 2002 This typology accounts for asymmetries in double object passives as well as between Japanese adversity passives and adversity causatives. Because of their ability to work with unergative verbs and the interpretation that the subject is negatively affected by an event, adversity passives are a type of high applicative. The properties attributed to each type are sketched out in Table 6: Table 6 Applicative Type High (symmetric) Applicative Individual to Event Properties Benefactive/malefactive with unergative Benefactive as subject of passive Theme as subject of passive Languages Kichaga, Sesotho, Japanese Adversity Passive Low (asymmetric) Applicative Individual to other Individual *Benefactive/malefactive with unergative Benefactive as subject of passive *Theme as subject of passive English, Finnish, Japanese Adversity Causative Because the applicative head in (20a) is introduced above the VP, Pylkkänen argues that similar to passives, it introduces an external argument that is not an agent. Pylkkänen suggests the passive morphology on adversity passives is the default spell-out of the verbal category feature of nonactive verbal functional heads which do not introduce an agent. In other words, the passive morphology on adversity passives is a result of the non-agent status of the applicative head, and not because the construction is a passive. Since Chaga and Sesotho passives overtly contain an applicative morpheme, it is likely that Japanese adversity passives and certain Bantu applicative passive share the same syntax. 10 3.3 Child Data To argue that Sesotho child passives are adversity passives, then, one would look for evidence of applicative passives in the child data. Unfortunately, Demuth (1998) indicates that while children produce active applicative constructions in about 3-4% of their utterances, they do not use the passive applicative construction very often. What Demuth notes that they do use, however are perfective passives. This type of passive occurred in 11% of the child passives in the Demuth 1989 appendix, and 16% (child L) and 14% (child H) in individual transcripts (Crawford 2004). While perfectivity does not bear on the discussion of adversity constructions as applicatives do, they are important if one considers that in Sesotho, the applicative and perfective morphemes are homophonous. 3.4 Homophony: Perfective passives and Applicative passives I argue that what Demuth called perfective passives in her data are really applicative passives and therefore adversity constructions. In Sesotho, the perfective morpheme and the applicative morpheme are homophonous: (23) (24) Lijo li-pheh-ets-w-e Food SM-cook-PERF-PASS-FV The food has been/is in a state of having been cooked. Demuth 1998:788 Lineo o-sho-ets-o-e ke-monna. Lineo AGR-die-APPL-PASS-FV by-husband ‘Lineo has been bereaved of the husband.’ Woolford 1993:31 In addition to being homophonous, the perfective is the only aspectual marker that is post verbal, meaning perfective and applicative occupy the same place in the verb stem. There is a historical link between the perfective and applicative morpheme in Bantu, and while some allomorphs differentiate the two in Sesotho, some modern Bantu languages, like Cibemba, the applicative and perfective morpheme are always homophonous (see Hyman 2002). The perfective and applicative morphemes in Sesotho do not reduplicate when used together, i.e., a perfective applicative will not yield a –elil- or –etsets- morpheme. The syntactic mechanisms require further research, but the perfective passives Sesotho-speaking children produce, if analyzed as applicative passives, provide evidence that early passives in Sesotho are adversity constructions, and should not be subject to maturation9. 9 The early passives as adversity passives analysis would satisfy both previous and current versions of the Maturation Hypothesis. McGinnis (1998) argues that high applicatives are complete verbal complexes, and can function as phase heads. Children giving an adversity passive/applicative analysis to these constructions could be an example of Wexler’s (2004) Universal Phase Requirement. Just as children may allow defective v to be a phase head, Sesotho children might also produce adversity passives/applicatives because the applicative head provides an escape hatch for the object NP. In either case, if early Sesotho passives are adversity constructions, then the passive morphology in early child constructions is default spell out, and not a product of verbal passive formation. 11 4.0 Conclusion In this paper I have shown the Sesotho data is not a strong counterexample to the Maturation Hypothesis, and that a broader view of s-homophones in Sesotho may explain the early emergence of passive like constructions. The evidence for an adversity passive as s-homphone analysis is particularly strong, though the syntax requires further research. 12