A Theoretical Account for the Undergeneration and the Overgeneration in Japanese Complex

A Theoretical Account for the Undergeneration and the Overgeneration in Japanese Complex Predicates1 1. Introduction In this paper, we discuss the undergeneration and the overgeneration observed in the acquisition of Japanese complex predicates. In particular, we focus on the errors children make for causatives and potentials. It has been reported that Japanese-speaking children, at around 2 to 4 years of age, produce causatives without the causative morpheme -(s)ase as in (1) (Suzuki 1987, Ito 1990, Murasugi and Hashimoto 2004, Murasugi, Hashimoto and Fuji 2007, among others).2 (1) Mama Akkun non -de. (Akkun, 2;8) mommy drink-Req Literal meaning: ‘Mommy, (please) drink Akkun(/me).’ Intended meaning: ‘Mommy, (please) feed Akkun(/me) (with milk).’ (adult form: nom-(s)ase-te) (Murasugi and Hashimoto 2004) In (1), the causative form nom-(s)ase-te should be used in this context, but the child does not phonetically realize -(s)ase and produces non-de instead. Another type of error is also observed in the acquisition of causatives as in (2) (Ito 1990, Arai 2003, Murasugi and Hashimoto 2004, among others). (2) a. b. Nomi -tyatye-te. (-tyatye = -sase) (Akkun, 3;7) drink -Cause-Req Intended meaning: ‘(Please) feed (me with miso soup.)’ Kuruma -o too-si -sase -ru. (Taatyan 3;10) car -Acc pass-Cause -Cause -Pres Intended meaning: ‘(I’ll) let the car pass through.’ (adult form: nom-(s)ase-te) (Murasugi and Hashimoto 2004) (adult form: too-s-(r)u) (Arai 2003) In (2a), the causative morpheme -(s)ase should be attached to the verb stem nom ‘to drink.’ However, the child “erroneously” attaches -tyatye (meaning -sase) to the preverbal form nomi, and produces nomi-tyatye-te. (2b) is an example of doubled causatives. Although the transitive verb too-s-(r)u ‘let … pass’ itself can be a causative verb, the child additionally attaches the unnecessary causative morpheme -sase to it, yielding the unacceptable form too-si-sase-ru. The children overgenerate the morpheme -sase in either case. Arai (2006) gives a phonological explanation for the erroneous causative and potential forms productively uttered by a Japanese-speaking child. The data discussed in Arai (2006) and the data of Sumihare (Noji 1974-1977) are further reanalyzed by Yano (2007a) under Murasugi and Hashimoto’s (2004) VP-shell analysis. This paper is an extension of Yano (2007a), arguing that the parallel undergeneration and overgeneration phenomena are observed in the acquisition of -(rar)e potential complex predicates as well as -(s)ase causative complex predicates.(Okubo 1967, Ito 1990, Shibuya 1994, Arai 2006, among others).3 The relevant examples are shown in (3) and (4). 1 We would like to thank Mamoru Saito, Keiko Yano, Tomoko Kawamura, Tomohiro Fujii, Kensuke Takita, Tom Roeper, William Snyder, Jeff Bernath, Jean Crawford, Alison Gabriele, Jim Huang, Takuya Goro, Elena Koulidobrova, Akira Omaki, Seichi Sugawa, and the members of the undergraduate seminar classes of Murasugi, for their helpful discussions and suggestions on this paper. We thank the anonymous reviewers of BUCLD 32 for the comments and suggestions. 2 Abbreviation used in the glosses are as follows: Acc=accusative Case, Cause=causative, Dat=dative Case, Int=interjection, Neg=negation, Nom=nominative Case, Pres=present, Past=past, Pot=potential, Req=request, Top=topic, Trans=transitive 1 (3) Mother: Zenbu tabe -rare -ru kara ne. all eat -Pot -Pres as Int ‘You can eat all.’ Child: Zenbu tabe φ -ru ne. (Sumihare, 2;1) (adult form: tabe-rare-ru) all eat -Pres Int Literal meaning: ‘(I) eat all.’ Intended meaning: ‘(I) can eat all.’ (Noji 1974-1977) (4) a. b. Yar -(r)are do -Pot Zyoozuni well -nai. (Taatyan, 3;5) (adult form: yar-e nai) -Neg Intended meaning: ‘I cannot do.’ mot -e -rare -ta. (Taatyan, 4;2) (adult form: mot-e-ta) have -Pot-Pot -Past Intended meaning: ‘I could bring (this) up very well.’ (Arai 2006) (3) is an example of the undergeneration. According to the observer, the child repeats his mother's utterance without using the -rare form (Noji 1974-1977). (4a) and (4b), on the other hand, are the examples of the overgeneration. In the case of (4a), the child intends to say that he cannot do something, and the potential morpheme -e should be attached to the verbal stem. However, -rare is attached, and the erroneous form -yarare, instead of -yare, is produced. In (4b), the child attaches the morpheme -e correctly to the verbal stem, but additionally attaches the unnecessary morpheme -rare, thereby producing an unacceptable doubled potential form. In either case, the morpheme -rare is overgenerated. The purpose of this paper is to analyze the undergeneration and overgeneration phenomena observed in the acquisition of Japanese causative and potential complex predicates. We argue that the intermediate acquisition stages can be uniformly explained by Murasugi and Hashimoto’s (2004) v-VP frame analysis for the complex predicates. Based on the analysis of longitudinal studies of Japanese-speaking children reported in the previous literatures, e.g., Akkun (Murasugi and Hashimoto 2004), Sumihare (Noji 1974-1977), Taatyan (Arai 2003, 2006), and the subjects observed by Okubo (1967), Ito (1990), Shibuya (1993, 1994) Suzuki (1987), among others, we argue that the undergeneration is due to the Japanese-speaking children’s initial hypothesis that the v does exist but it is phonetically null (just as in English), as Murasugi and Hashimoto (2004) propose. We also discuss that the overgeneration takes place because of the erroneous realization of the v, and/or the use of the undifferentiated verbal form, as Murasugi (2007a, b) proposes. This paper is organized as follows. In Section 2, we discuss Japanese adult grammar of causatives and potentials. We overview the proposals made by Murasugi and Hashimoto (2004) and Bobaljik and Wurmbrand (2005, 2007) that Japanese complex predicates have v-VP or VP-shell structures (Larson 1988, Hale and Keyser 1993, 2003, Chomsky 1995, Harley 1995, 2006, among others). Then, we discuss the undergeneration phenomenon in Section 3, and the overgeneration phenomenon in Section 4. Section 5 concludes this paper. 2. The Adult Grammar of Japanese Complex Predicates 2.1. -(S)ase Causatives Japanese morphological -(s)ase causatives are formed by attaching the causative morpheme -(s)ase to the verb stems. It has been argued that -(s)ase is ambiguous in two ways (Miyagawa 1984, 1998, Harley 1995, 2006; Matsumoto 2000, Murasugi, Hashimoto and Kato 2003, Murasugi and Hashimoto 2004, among others). The sentence in (5), for instance, is ambiguous between (6a) and (6b). (5) Taroo -ga Hanako -ni pan -Nom -Dat bread ‘Taroo made Hanako eat some bread.’ -o -Acc tabe -sase eat -Cause 3 -ta. -Past This paper is an extension of Yano’s MA thesis (2007a) submitted to Nanzan University, which was also presented at Connecticut-Nanzan-Siena Joint Workshop on Linguistic Theory and Language Acquisition at Nanzan University, February 21st in 2007, and in Nanzan Linguistics Special Issue 3.1. 2 (6) a. b. Taroo gave an order to Hanako and Hanako ate some bread. Taroo fed Hanako with some bread. In one reading, as shown in (6a), Hanako is an agent, whereas it is a goal in another reading as shown in (6b). The sentence with the former reading is called ‘the syntactic causative,’ while that with the latter reading is called ‘the lexical causative.’ Based on the VP-shell hypothesis (Larson 1988, Hale and Keyser 1993, 2002, Chomsky 1995, among others), Murasugi and Hashimoto (2004) propose the structures for these two types of -(s)ase causative, as shown in (7a) and (7b), respectively. (7) a. syntactic causative vP causer v’ VP v [+cause] vP V φ causee v’ -(s)ase VP v [+cause] NP V φ b. lexical causative vP causer v’ VP causee V NP v[+cause] -(s)ase V (Murasugi and Hashimoto 2004) As shown in (7), a syntactic causative has the bi-clausal structure, whereas a lexical causative has the mono-clausal structure. According to their analysis, -(s)ase is ambiguous in the adult grammar of Japanese. In one case where it is an independent V, it takes a v-projection as its complement, yielding a complex structure as in (7a). In this case, the dative argument, or the causee, is interpreted as an agent. In the other case, it combines with the V and forms a complex verb, yielding a simple sentence with no embedding as in (7b). The dative argument is then interpreted as a goal. In Murasugi and Hashimoto’s (2004) terms, Japanese causative morpheme -(s)ase is a realization of the [+cause] v.4 Following Hale and Keyser (1993, 2003), Murasugi and Hashimoto (2004) give the v-VP frame structure to Japanese transitive and intransitive sentences. Japanese transitive and intransitive verbal forms are distinguished with their morphemes as in (9), and these morphemes are in the v as shown in (10).5 (9) a. utu-s-(r)u (= photograph-Pres) b. todok-e-ru (= deliver-Pres) c. os-ie-ru (= teach-Pres) d. too-s-(r)u (= let … pass-Pres) (10) / / / / utu-r-(r)u (= be photographed-Pres) todok-(r)u (=be delivered-Pres) os-owar-(r)u (= be taught-Pres) too-r-(r)u (= pass-Pres) a. transitive vP agent v’ VP v [+cause] theme V’ -s location V utu b. intransitive vP v’ VP v[-cause] theme V’ -r location V utu (Murasugi and Hashimoto 2004) 4 Following Miyagawa (1980, 1984, 1998), Zenno (1985), Harley (1995), and Matsumoto (2000) among others, Harley (2006) also argues that -sase is ambiguous between syntactic and lexical causatives, and that syntactic causatives are bi-clausal while lexical causatives are mono-clausal. She further assumes the v-VP structure for Japanese causatives based on Hale and Keyser (1993, 2002). Harley (2006), however, considers -sase to be in the v in both syntactic and lexical causatives. According to her analysis, in lexical causatives, a CAUS vº is adjacent to a root. On the other hand, in syntactic causatives, a CAUS vº is not adjacent to a root and takes another vP complement. 5 Japanese transitive and intransitive morphemes are not always overt. For instance, in the case of todok-(r)u in (9b), the [-cause] v is realized phonetically null like English verbs. 3 Unlike Japanese, the [±cause] v is not phonetically realized in English transitive and intransitive sentences, as in (11). (11) a. b. Mary sank the boat. (transitive) The boat sank. (intransitive) The structures of (11a) and (11b) are represented in (12a) and (12b) respectively. (12)a. vP Mary [+cause]v v’ VP V NP SINK the boat (v[+cause] +SINK =sink) b. vP [-cause]v VP V’ the boati V NP SINK ti (v[-cause] +SINK =sink) (Larson 1988, Hale and Keyser 1993, 2002, Chomsky 1995) In (12a) and (12b), the verb ‘sink’ consists of two abstract verbs: the [±cause] v and V. The [±cause] v is realized phonetically null, and the transitive and the intransitive verbs have the same verbal form. 2.2. -(Rar)e Potentials Japanese -(rar)e potential complex predicates are constructed by adding the morpheme -(rar)e ‘-able’ to the verb stems. The vocalic verb stems take the morpheme -rare, and the consonantal verb stems take the morpheme –e, as shown in (13) (Shibuya 1993, Kinsui 2003, Arai 2006, among others). (13) a. b. Vocalic verbs: stem +-rare ‘eat-Pot-tense’ present: tabe-rare-ru past: tabe-rare-ta Consonantal verbs: stem + -e ‘go-Pot-tense’ present: ik-e-ru past: ik-e-ta ‘see-Pot-tense’ mi-rare-ru mi-rare-ta ‘make-Pot-tense’ tukur-e-ru tukur-e-ta The verbs tabe-ru ‘to eat’ and mi-ru ‘to see’ shown in (13a) are vocalic, whose stems end with vowels, and the morpheme -rare is attached to the verb stems to derive the potential forms, tabe-rare-ru ‘can eat’ and mi-rare-ru ‘can see.’ In contrast, ik-u ‘to go’ and tukur-(r)u ‘to make’ shown in (13b) are consonantal verbs, whose stems end with consonants, and the morpheme -e is attached to the verb stems to derive the potential forms, ik-e-ru ‘can go’ and tukur-e-ru ‘can make.’ The examples in (14) indicate the typical potential sentences. (14) a. b. Taroo -wa hitoride supagetti -Top by oneself spaghetti ‘Taroo can eat spaghetti by himself.’ Hanako -wa hitoride gakkoo-ni -Top by oneself school -Dat ‘Hanako can go to school by herself.’ -o -Acc tabe -rare eat -Pot ik-e go-Pot -ru. -Pres -ru. -Pres There are several important proposals regarding the structure of potentials (e.g., Tada (1992) and Koizumi (1995) for the 4 AGR-based approach, Saito and Hoshi (1998) for the Head-Head Merger approach, and Takano (2003) for the Prolepsis approach). In this paper, we assume Bobaljik and Wurmbrand’s (2005, 2007) analysis that the potential morpheme -(rar)e is in the head of the vP. Given their analysis, the structure of (14a) would be the one shown in (15). (15) vP Taroo v’ VP v hitori-de VP -rare NP VP supagetti-o tabe Thus, under the v-VP frame analysis, the potential morpheme -(rar)e, transitive and intransitive morphemes, and the causative morpheme -(s)ase in lexical causatives are all in the head of the vP. In the following sections, we analyze the undergeneration and the overgeneration in -(s)ase causatives and -(rar)e potentials, and provide some supporting evidence for the hypothesis that children, in the course of language acquisition, fail to realize the correct adult form for the v (Murasugi and Hashimoto 2004, Murasugi 2007a, b). 3. The Undergeneration According to Murausgi and Hashimoto (2004), the structure of VP-shell itself is acquired very early,,6 but it takes time for the Japanese-speaking children to acquire the “correct” lexical form of the v. Murasugi and Hashimoto (2004) propose that there is a stage where children hypothesize that the v is realized phonetically null. This stage corresponds to what we call the stage of the undergeneration. 3.1. The Undergeneration in -(S)ase Causatives Children start producing the intransitive and the transitive verbs correctly at around the age of 2. However, they do not always produce the correct ones. The utterances which have the causative meaning (or intention) are produced without the causative morpheme -(s)ase in the early 2-year-old. We call such phenomenon ‘the undergeneration.’ A couple of examples are given in (16a (=(1)) and (16b). (16) a. b. Mama Akkun non -de. (Akkun, 2;8) mommy drink -Req Literal meaning: ‘Mommy, (please) drink Akkun(/me).’ Intended meaning: ‘Mommy, (please) feed Akkun(/me) (with milk).’ Kutyu hai -te. (Sumihare, 2;1) a pair of shoes put on -Req Literal meaning: ‘(Please) put on (your) pair of shoes.’ Intended meaning: ‘(Please) put a pair of shoes on (me).’ (adult form: nom-(s)ase-te) (Murasugi and Hashimoto 2004) (adult form hak-(s)ase-te) (Murasugi, Hashimoto and Fuji 2007) 6 From around 1;5 through 3 years of age, the sentences with suru/sita/site ‘do/did/doing’ as in (i) are often produced by Japanese-speaking children. (i) Mama Akkun hai doozyo tiyu (tiyu = suru) (Akkun, 2;5) mommy yes please do ‘Akkun(/I) will give (it) to Mommy.’ (Murasugi and Hashimoto 2004) Murasugi and Hashimoto (2004) propose that suru/sita/site assigns an agent role to a subject. That is, the v is realized as suru/sita/site in this stage, and children have VP-shell structures. See Murasugi and Hashimoto (2004) and Murasugi, Hashimoto and Fuji (2007) for more data and detailed analysis. 5 As shown in (16a), Akkun intends to say nom-(s)ase-te to ask his mother to feed him, but utters non-de ‘please drink’ instead, without lexically realizing -(s)ase. The same type of undergeneration error is found in the longitudinal data of Sumihare (Noji 1974-1977). In (16b), Sumihare produces hai-te instead of hak-(s)ase-te. He does not produce -(s)ase though it is clear that he intends to ask someone to put a pair of shoes on him. In this stage, children also produce transitive/intransitive alternation errors such as in (17). (17) a b. Akkun ima kaya koe nayab-u. (nayab-u = narab-(r)u) now from this be-in-line Intended meaning: ‘From now, Akkun(/I) will put these in line.’ Nui-ta koko. (Sumihare, 2;1) pull-Past here Intended meaning: ‘(This) is out from here.’ (Akkun 2;11) (adult form: narab-e-ru) (Murasugi and Hashimoto 2004) (adult form: nuk-e-ta) (Murasugi, Hashimoto and Fuji 2007) In (17a), instead of the transitive verb narab-e-ru ‘to put...in line,’ Akkun erroneously produces what corresponds to the intransitive verb narab-(r)u ‘to be...in line.’ In contrast, in (17b), the intransitive verb nuk-e-ta ‘came off’ must be used in the adult grammar. However, Sumihare uses the transitive form nui-ta ‘pulled.’ Murasugi and Hashimoto (2004) argue that both of the causatives without -(s)ase and the transitive/intransitive alternation errors are due to children’s initial hypothesis that the [±cause] v is phonetically null. To be precise, the utterances in (16) are not ‘omission’ errors, but they are ‘the undergeneration.’ Although the children apparently omit -(s)ase, they just assume the v to be null. This also holds for the transitive/intransitive alternation errors. As discussed in Section 2.1, in adult Japanese, the [±cause] v are basically overt, and therefore a transitive verb and an intransitive verb have different forms, such as utu-s-u ‘to photograph’ and utu-r-u ‘to be photographed.’ However, children assume the [±cause] v to be null in this period, and they do not distinguish the two forms. Thus, Murasugi and Hashimoto (2004) propose that this is the stage where children know that either -s or -r should be attached to the verb (probably because they never hear the verb without these morphemes), and hence, children randomly attach these morphemes directly onto the V. This is why children sometimes produce verbal forms correctly and sometimes do not. 3.2. The Undergeneration in -(Rar)e Potentials Like causatives, children start producing potentials without the potential morpheme -(rar)e in the early 2-year-old. We find the same type of undergeneration as causatives. The example (3) is repeated below as in (18). (18) Child: Zenbu tabe φ -ru ne. all eat -Pres Int Literal meaning:‘(I) eat all.’ Intended meaning:‘(I) can eat all.’ (Sumihare, 2;1) (adult form: tabe-rare-ru) (Noji 1974-1977) As we mentioned in Section 1, Sumihare repeats after his mother without using -rare form, despite the fact that the mother speaks to her son using -rare form tabe-rare-ru ‘can eat.’ This type of error is also reported in other literatures, e.g., Okubo (1967) and Ito (1990). Given that the potential morpheme is in the v (Bobaljik and Wurmbrand 2005, 2007), this error can be well explained under Murasugi and Hashimoto’s (2004) VP-shell analysis. That is, children in this stage, at around the age of 2, hypothesize that the v is phonetically null. On the other hand, the “correct” potential forms also appear in the children’s early production (Okubo 1967, Ito 1990, Shibuya 1994, Arai 2006, Yano 2007a, b, c). Shibuya (1994) and Arai (2006) report that children start producing the correct potential forms with the morpheme -e at around the age of 2. Based on her analysis of Noji corpus from CHILDES database (MacWhinney 2000), Yano (2007a, b, c) also finds that -e potentials appear very early, at 2;0. Some examples of Sumihare’s potential sentences are given in (19). 6 (19) a. b. Too -r -e -n. (Sumihare, 2;0) pass -Trans -Pot -Neg ‘(I) cannot pass.’ Kakko -ga hak -e -n. (Sumihare, 2;2) shoes -Nom put on -Pot-Neg ‘(I) cannot put on my shoes.’ (Yano 2007a, b, c) In (19a) and (19b), Sumihare apparently produces potential forms correctly. Too-r-e-n ‘cannot pass’ in (19a) and hak-e-n ‘cannot put on’ in (19b) are the short forms of too-r-e-nai and hak-e-nai. Those are possible truncated forms in some dialects. Then, are potentials really acquired in such an early stage, even at the age of 2? A possible answer for this question is a positive one, i.e., the adult syntax of potentials is acquired very early compared with other complex predicates. The other possibility is, on the contrary, what apparently looks like the adult potential form does not, in fact, have exactly the same structure as adults’. Murasugi (2007a, b) gives an answer to this problem. She proposes that children’s predicates produced at the early age of 2 are the uninflected (undifferentiated) forms, and that the v is phonetically null in this stage. Hereafter, we support the latter possibility and Murasugi’s (2007a, b) analysis, presuming that Sumihare’s potential sentences produced at the early age of 2, in fact, are the undifferentiated forms, hence, this is the stage where the v is phonetically null as well. In other words, in this stage, the children attach the potential morpheme -e onto the V. Some evidence for this proposal is found in Murasugi and Fuji’s (2007) analysis of the longitudinal data of Sumihare (Noji 1974-1977). First, Sumihare, in the stage where the undergeneration in question is observed, produces “erroneous” inflection forms of tense markers as in (20). (20) Tootyan, koko gomi tui -ta yo. (Sumihare, 2;1) father here dust stick-Past Int Intended meaning: ‘Daddy, you have got the dust here (=on your cheek).’ (adult form: tui-te iru (perfect)) (Murasugi and Fuji 2007) The context of (20) is that Sumihare wants to tell his father that his father has dust on his cheek. In this context, the perfect form -te iru7 should be used. However, Sumihare employs the simple past form and says tui-ta ‘stuck’ instead of tui-te iru ‘have been stuck.’ Based on the detailed analysis of other perfect and progressive forms produced by Sumihare (Noji 1974-1977), Murasugi and Fuji (2007) argue that tense markers such as -te(iru), -ta, or -(r)u and the stems of verbs are not differentiated at the very early stage. This analysis stands upon the proposal by Murasugi (2007a, b) that the verbal forms of Japanese-speaking children at around the age of 2 are the unmerged form of the verb and the tense, and that the children’s functional categories such as T and v are not phonetically realized at the very early stage in such agglutinative language as Japanese. The verbs and the tense markers are unanalyzed (undifferentiated), and they are base-generated in the V. That is, Sumihare regards the whole verbal form tui-ta in (20) as V. The second evidence is, as is discussed in Murasugi and Fuji (2007), found in Sumihare’s erroneous negative forms observed in the acquisition stage in question. In adult Japanese, -nai ‘not’ is a verbal predicate which itself carries finite tense (Sano 2000). The structure of negation is represented schematically in (21). (21) IP I’ NEGP I VP NEG -i (present) /-kat-ta (past) Root +irrealis affix na-(=finite) (Based on Sano 2000) The examples in (22), however, indicate that Sumihare, in an early age of 2, does not have the structure shown in (21). 7 Japanese -te iru sentences are ambiguous between the progressive interpretation and the perfect interpretation. 7 (22) a. Father: Sumihare: b. Mother: Sumihare: Sinbun tot -ta? newspaper take -Past ‘Have (you) fetched the newspaper yet?’ Tot -ta -nai. (Sumihare, 2;1) (adult form: tot-te-nai) take-Past-Neg Intended meaning: ‘(I) haven’t.’ Oti -ru yo. fall down -Pres Int ‘(You) will fall down.’ Oti -ta -nai. (Sumihare, 2;1) (adult form: oti-te-nai/oti-nai) fall down -Past-Neg Intended meaning: ‘(I) won’t.’ (Murasugi and Fuji 2007) In (22a), Sumihare is asked if he has already fetched the newspaper, and he intends to say, ‘No, I haven’t.’ For this context, the negative form tot-te-nai ‘haven’t taken’ should be used. However, Sumihare puts -nai on the past tense form tot-ta ‘took,’ and produces the unacceptable negative form tot-ta-nai. (22b) shows the same type of error. Even though his mother speaks to him with the present verbal form oti-ru ‘fall down,’ the child produces oti-ta-nai, putting -nai on the past tense form oti-ta ‘fell down,’ despite the fact that the negative form oti-te-nai or oti-nai should be used for this context. Recall here that in such agglutinative language as Japanese, the functional categories such as T and v are not phonetically realized in the early stage, and the verbal forms of Japanese-speaking children at around the age of 2 are the unmerged form (Murasugi 2007a, b). The examples given above create one of the examples for this hypothesis. The negative marker -nai ‘not’ is attached to the verbal form, totta or otita in this acquisition stage, because the child regards the whole verb containing the tense marker as a rote form. Thus, the discussion so far leads us to presume that Sumihare, at the early age of 2, produces the undifferentiated verbal forms as Murasugi (2007a, b) discusses. Then, it is natural to consider that the potentials found in this period are also undifferentiated verbs. Given this analysis, the examples in (19) would be, in fact, the “lexical potentials.” That is, the child attaches the potential morpheme -e, not as a realization of the v, but as a part of the V. Note here that the child forms are all used with the potential meaning in the “correct” context. Hence, it is plausible that the v does exist and it has the feature at this point, but it is phonetically null in the stage of the undergeneration, as Murasugi and Hashimoto (2004) propose. 4. The Overgeneration In Section 3, we argued that the undergeneration phenomenon observed in Japanese causatives and potentials are elegantly explained under Murasugi and Hashimoto’s (2004) v-VP frame analysis. In this section, we turn to the overgeneration phenomenon. We first report that the parallel pattern of the overgeneration is observed in the acquisition of Japanese causatives and potentials. We, then, point out that there are two types of overgeneration. We argue that they indicate that there is an acquisition stage where the children know that the v has to be phonetically realized in their target grammar, but (i) have not acquired the “correct (adult)” lexical form, and/or (ii) have still the undifferentiated rote verbal form. 4.1. The Overgeneration in -(S)ase Causatives Murasugi and Hashimoto (2004) report that Akkun, their subject, produces lexical causatives at around the age of 3, as in (23). (23) Akkun -ni tabe -sase -tee. -Dat eat -Cause -Req ‘(Please) feed Akkun(/me) (with food).’ (Akkun, 3;6) (Murasugi and Hashimoto 2004) In (23), Akkun is not an agent but a goal, since Akkun is asking his mother to put some food directly into his mouth. Thus, in this stage, children are apparently aware that the [±cause] v must be realized phonetically. However, at the same time, children produce two types of overgeneration. The one is “a verb + -(s)ase” form, and the other is “a causative verb + a causative morpheme” form. The examples of the first type of overgeneration are given in (24a) 8 and (24b(=2a)). (24) a. b. Oki -sasi -te. (Sumihare, 3;1) (adult form: ok-osi-te) get up -Cause -Req Intended meaning: ‘(Please) get (me) up.’ (Noji 1994-1997) Nomi -tyatye -te. (-tyatye = -sase) (Akkun, 3;7) (adult form: nom-(s)ase-te) drink -Cause -Req Intended meaning: ‘(Please) feed (me with miso soup.)’ (Murasugi and Hashimoto 2004) In (24a), Sumihare asks someone to get him up. Here, the imperative form of the lexical causative verb ok-os-(r)u ‘to get … up,’ or ok-osi-te, is expected in the adult grammar, but Sumihare erroneously attaches -sasi (which is sometimes used instead of the standard -sase in his dialect) to the stem of the intransitive verb ok-i-ru ‘to get up.’ (24b) shows the similar type of error. The correct causative form of the verb nom-(r)u ‘to drink’ is nom-(s)ase-ru. However, Akkun erroneously attaches -tyatye, or the child’s phonetic form of -sase, to its preverbal form nomi, and makes nomi-tyatye-te. Thus, both of Sumihare and Akkun overgenerate the causative morpheme -sase in the cases where the transitive morpheme -os or the causative morpheme -ase should be attached. These data indicate that the children know that the v should be phonetically realized, but they fail in choosing the right one among several ways of the realization of the causative morphemes, and choose the unmarked bound morpheme form, i.e., -sase. The second type of overgeneration, “a causative verb + a causative morpheme,” is produced at around later 3 years old to 5 years old. The relevant examples are shown in (25a (=2b)) and (25b). (25) a. b. Kuruma -o too -si -sase -ru. car -Acc pass -Cause -Cause -Pres Intended meaning: ‘(I’ll) let the car pass through.’ Kondo mi -se -si-te ageru kara next time see -Cause -Cause let as Intended meaning: ‘(I’ll) show you next time’ ne. Int (Taatyan, 3;10) (adult form: too-s-(r)u) (Taatyan, 4;6) (adult form: mi-se-te) (Arai 2003) In (25a), the transitive verb too-s-(r)u ‘to let … pass,’ which is a causative verb as well, is erroneously associated with an additional causative morpheme -sase. (25b) is another example of doubly marked causatives. The transitive verb mi-se-ru ‘to show’ or ‘to let … see’ is, in fact, a causative verb containing the transitive morpheme -se in it. However, the child adds the additional causative morpheme -si 8and produces mi-se-si-te “by mistake.” Why do the children produce the second type of overgeneration? We conjecture that this is the stage where children still have difficulty in finding the appropriate stem form of a verb as well as the correct form of a bound morpheme. They assume an undifferentiated causative verb as the stem of a bare verb. They also know that the v should be phonetically realized to make a causative verb. Hence, children attach an additional causative morpheme -sase or -si onto the undifferentiated V, as illustrated in (26). (26) vP v’ VP v V -sase/-si toosi/mise Since children regard the whole verbal forms toosi and mise as V, they add the causative morphemes to realize the v phonetically. This is probably why the children make the causative doubling: the lexical causatives with an additional 8 The causative morpheme -(s)ase and -(s)e are often produced as -(s)asi and -(s)i in the western dialects of Japanese. 9 unnecessary causative morpheme.9 4.2. The Overgeneration in -(Rar)e Potentials Just like the causatives, in the acquisition of potentials as well, we find in Shibuya (1994) and Arai (2006) that the Japanese-speaking children may go through the stages of the overgeneration. And just like the causatives, two types of overgeneration is also found in the acquisition of -(rar)e potentials. One type of overgeneration is “the stem of a verb + -rare.” Shibuya (1994) and Arai (2006) report that children produce the overgeneration as given in (27a) and (27b (=4a)). (27) a. b. Taakun hitori-de tukur -(r)are -ta. (Taatyan, 3;0) by oneself make-Pot -Past Intended meaning: ‘Taakun(/I) could make (this) by himself(/myself).’ Yar-(r)are -nai. (Taatyan, 3;5) do -Pot -Neg Intended meaning: ‘I cannot do.’ (adult form: tukur-e-ta) (adult form: yar-e nai) (Arai 2006) In standard Japanese, -rare is attached to vocalic verb stems, while -e is attached to consonantal verb stems, as is discussed in Section 2. However, Taatyan overgenerates -rare in this stage. According to Arai (2006), the child has his own morpho-phonological rule, and puts -rare on the vocalic verbs, and -are on the consonantal verbs. Thus, in (27a) and (27b), the morpheme -are attaches to the stems of the verbs tukur-(r)u ‘to make’ and yar-(r)u ‘to do,’ though the morpheme -e should be attached in the adult grammar.10 Again, just like the causative -sase, children tend to choose an unmarked potential morpheme. In the case of potentials, -rare seems to be the unmarked potential morpheme attached. Thus, in the stage where the overgeneration in the acquisition of potentials is found, children have knowledge that the v is phonetically realized, but they do not know the “correct” adult form, and hence, they choose the unmarked potential morpheme -rare. Another type of overgeneration, “a potential verb + a potential morpheme,” is also observed at the later 3 years old to 5 years old.11 A couple of examples are given in (28a) and (28b (=4b)). (28) a. b. Gakko-ni ik -e -re -ru yo. (Sumihare, 3;1) school-Dat go -Pot -Pot -Pres Int Intended meaning: ‘(I) can go to school by myself.’ Zyoozuni mot -e -rare -ta. (Taatyan, 4;2) well have -Pot-Pot -Past Intended meaning: ‘I could bring (this) up very well.’ (adult form: ik-e-ru) (Noji 1974-1977) (adult form: mot-e-ta) (Arai 2006) In (28), children attach an additional potential morpheme -re or -rare onto the already-potential-verb (or “lexical” potentials). Sumihare, in (28a), produces an erroneous potential form, ik-e-re-ru for ik-e-ru (the adult potential form of ik-u ‘to go’). The same type of error is found in Taatyan’s data, as in (28b). Taatyan correctly attaches -e to the consonantal verb mot-(r)u ‘to have.’ However, he additionally attaches the potential morpheme -rare, yielding an unacceptable potential form mot-e-rare-ta. As is the case of causatives, this type of overgeneration would reflect the stage where children still use the undifferentiated potential verbs. This is the stage where they know that a potential morpheme should be attached to make a productive potential form in the target grammar, and hence, they put the (unmarked) form -rare on the lexical potential verb. 9 Mamoru Saito brought our attention to this analysis. The overgeneration in (25) can be considered as the morphological one since the number of argument does not increase unlike the famous example of the syntactic overgeneration ‘Don’t giggle me’ (Bowerman 1982). We would like to thank Jean Crawford and William Snyder for also making this point to us. 10 As far as we know, Shibuya (1994) is one of the pioneers who found the productive overgeneration errors in potentials. Arai (2003, 2006) report the overgeneration phenomena in causatives as well as potentials, and attempt to analyze them in a uniform way. 11 In the case of Sumihare, he starts producing two types of overgeneration at the same age, at around the age of 3. 10 5. Conclusion In this paper, we reported that the parallel intermediate stages in the acquisition of Japanese causatives and potentials are found. We argued that these intermediate stages, the undergeneration and the overgeneration, can be uniformly analyzed under the VP-shell hypothesis for the acquisition of the Japanese complex predicates proposed by Murasugi and Hashimoto (2004). First, in the stage of the undergeneration, -(s)ase as the v is not realized, and the transitive/intransitive alternation errors are also found. As for potentials, children do not produce -(rar)e, and employ the unanalyzed adult form as a potential verb. We argued that these errors are due to children’s initial hypothesis that the v does exist in their grammar, but it is phonetically null as Murasugi and Hashimoto (2004) propose. They relate this stage to the English transitive/intransitive alternation (See also (11)). (29) a. b. John passed the ring to Mary The ring passed to Mary Both the transitive pass and the intransitive pass are realized as pass. The children’s alternation errors found in Japanese have, in fact, the adult English-type structure: both the v’s of [±cause] are realized as zero morphemes. Second, children go through the two types of overgeneration stage for causatives and potentials. One type of overgeneration is “a correct (but sometimes inappropriate) stem of a verb + (unmarked) -sase or -rare.”12 Another type of overgeneration is the causative and the potential doubling: “an undifferentiated causative verb + a causative morpheme” and “an undifferentiated potential verb +a potential morpheme.” We argued that the overgeneration takes place, since Japanese-speaking children, at one point of language acquisition, have difficulty in finding the “correct” adult forms of the bound morphemes for causatives and potentials, and those agglutinative language learners, at an early stage of language acquisition, also have difficulty in differentiating the bound morphemes from the stems of verbs (Murasugi 2007a, b). 12 Precisely speaking, this type of error may not be the case of overgeneration, but rather, a morphological error, as Mamoru Saito (p.c.) and Alison Gabriele (p.c.) pointed out to us. 11