¹ The research was originally presented at the Boston University Conference on Language Development in 1988. Similar proposals were developed at around the same time by David Lebeaux and Andrew Radford and appear in Lebeaux (1988) and Radford (1990). The authors wish to thank Lisa Travis and Lydia White for discussion on many of the points of this paper. Thanks also to William O’Grady and to the members of the Argument Structure project in the Department of Linguistics at McGill, and members of the audience at the BULD Conference for helpful comments and suggestions. This research was supported by a doctoral fellowship from the Social Sciences and Humanities Research Council of Canada to Eithne Guilfoyle, and by a McGill Graduate Faculty Doctoral Fellowship to Náire Noonan. In addition we benefitted from FCAR grant 88EQ3630 to Lisa Travis and Lydia White, and SSHRCC grant 410-91-1956 to Eithne Guilfoyle.

² Not all scholars would agree that frequency bears on the issue, claiming that one instance of the triggering data would be sufficient to force a restructuring of the grammar. However, it seems plausible to us that the trigger should be easily accessible, to ensure that all children will be exposed to it. See Hyams (1987), Guilfoyle (1984) and Lebeaux (1987) for discussion.

³ This raises the question of whether all languages have Functional Categories, or whether all languages have the same Functional Categories. Fukui and Speas (1986), claim that Japanese lacks all Functional Categories. In this respect it resembles early child grammars, and that early child grammars thus fall within the domain of possible human languages.

⁴ Abney (1986) claims that prepositions are functional categories. We do not adopt this claim here, though we feel that certain prepositions may be functional. If that is true, than those that are lexical should be acquired earlier than those that are not.

⁵ In fact MacNamara (1982) claims that the child uses an innate predisposition to categorize real world objects as a means to access the grammatical concept noun. While this view is not predicted by our analysis it is compatible with it, insofar as nouns are lexical categories and acquired first.

⁶ See also footnote 8. Our view is compatible with an articulated INFL (Chomsky 1988; Pollock 1989). The affixes -ing and plural -s are assumed to be lexically derived. Their status is thus oblivious to whether the Syntax contains functional categories.

⁷ Fukui and Speas claim that Lexical Categories project only to the level of X’, and can never form a barrier; hence for them the tree in (8) would not contain an N”. For the purpose of this paper we do not wish to take a position on this issue. Note, however, that under a restricted definition of specifier as daughter of XP and sister of X’, there is only one true specifier available even in lexical categories. One might think of the iterated elements in (13) as multiply adjoined modifying elements. We use the non-categorial label SPEC in tree (13) (following Fukui and Speas’s notation) for convenience, as it allows us to remain open on which category the adjoined elements belong to (NP or AP).

⁸ Abney (1986) has claimed that pronouns are generated under the DET node in adult grammar, rather than under the N’ node. If this is correct, it would be expected that pronouns should be absent from early child grammars. While there are fewer pronouns in early child grammars, there is considerable variation in pronoun use among children (Maiatsos 1982:230). Again, the appearance of pronouns in the early stages does not imply that the Functional Category DET has been acquired, because the child could place the pronoun under the N node.

⁹ Note that if the Case Filter is derived from Visibility and the Theta Criterion, our analysis faces a problem since the child grammar we are arguing for would be in violation of UG.

¹⁰ Radford (1990) also assumes (11) for sentences in early child grammars, while Pierce (1989) assumes that subjects are generated within VP, but that functional categories are present.

¹¹ See Pierce (1989) for a proposal that V-movement is present at this stage.

¹² Noonan (1987) suggests that perhaps the child has a sort of combination of COMP and INFL initially, which only becomes separated at a later point. In Section 7 we discuss the role of input data with respect to the instantiation of individual functional categories.

¹³ They also discuss the acquisition of passives in Hebrew, and the acquisition of causatives in Hebrew and English. We do not discuss any of these topics here, however we assume that our analysis of the impossibility of NP-movement in early child grammars could be extended to account for these data too.

¹⁴ It might be possible to claim that the early grammar lacks Move ± however, Borer and Wexler specifically argue against this view, and assume there is wh-movement at this stage. In Section 6 we will claim that there is no wh-movement or NP-movement in early grammars, not because the grammar lacks the rule Move alpha, but rather because there is no suitable landing site.

¹⁵ See Guilfoyle (1990, 1992) for another analysis of passive, based on proposals in Fukui and Speas (1986). There the agent theta-role is not suppressed but assigned to PRO in SPEC/VP. Note that under this analysis also, passive b not possible in early child grammars, because SPEC/IP is unavailable and there is no landing-site for the theme.

¹⁶ Borer and Wexler (1987) also note that Raising is not possible in child grammars.

¹⁷ See Lebeaux (1987) for a discussion of Hyams’s analysis and a different proposal which links the absence of subjects to the position of INFL in child grammars.

¹⁸ In fact Koopman and Sportiche do not use the terms SPEC of IP and SPEC of VP, but rather NP” and NP* (where NP* is not the specifier but an adjoined position). We use SPEC of IP and SPEC of VP here to discuss their proposal for the sake of consistency.

¹⁹ Actually as we assume the subject to be base generated under SPEC of VP this tree is not strictly accurate.

²⁰ Verb-movement is head movement and respects the Head movement constraint. See Travis (1984) and Baker (1988) for discussion.

²¹ deHaan and Jordens have observed that it is mainly verbs of perception that appear in this position. This point will be taken up later (Section 6.4).

²² Note that this view is also incompatible with the subset principle (cf. Wexler and Manzini 1987).

²³ This view, incidentally, makes certain predictions about VSO languages like Irish. Under many analyses, VSO word order is produced from an underlying SVO order by V-movement to COMP (Sproat 1985) or to INFL (Travis 1984; Guilfoyle 1988, 1990; Chung and McCloskey 1987). The prediction would then be that in the early stages child grammar should exhibit SVO order, as there is no I or С and therefore no V-movement. This prediction is supported by the data from the acquisition of Irish reported in McKenna and Wall (1986) and Hickey (1990). See Guilfoyle (1990, 1992) for discussion.

²⁴ This might also explain why there seems to be a delay in the emergence of agreement marking and verb-second. Possibly, the lack of agreement could be related to its absence on English modals, which have also been assumed to be base-generated under I (Lightfoot 1979). Another, more updated explanation involves the split-INFL-hypothesis (Chomsky 1988; Pollock 1989): it might be argued that T(ense)Phrase emerges before A(greement)Phrase.

²⁵ Note, however, that this should be interpreted with caution, as the same kind of tendency might be operating in the production grammar of the child, in which case the evidence is inconclusive, and the reason we never find the elements in question in final position is that embedding does not occur at this stage.

²⁶ In fact, another possibility springs to mind: perhaps the category we have been calling IP is actually either IP or CP. Since in German it is extremely hard to tell whether an IP or a CP is at stake in a root clause (unlike English, where SPEC/CP is restricted to operator-phrases), the child might assume a CP (or CONFLPHRASE? cf. Platzack (1983) as her first guess, in which case the question of the argument status of the SPEC position need not be posed; and only later, when embedding emerges, would the child realize that there must be an intermediate projection, namely that of INFL (which could then be assumed to be final), and thus restructure her grammar. But see the next section for more discussion on the emergence of CP and IP.

²⁷ The precise analysis of acquisition of German word order depends on the analysis of the adult grammar of German. Recent developments in grammatical theory such as the split INFL-hypothesis (Pollock 1989) and the question of what counts as an A or A’ position, the analysis of V2 languages has become unclear and open to further research. See Diesing (1990), Kinyalolo (1990), Koopman and Sportiche (1990), Mahajan (1990), Pesetsky (1989) and others.

²⁸ We have merged Klima and Bellugi’s period 1 and period 2, since the differences between these two periods are irrelevant for our purposes here, consisting of a development from using wh-words only with certain verbs, to more productive use of wh-words.

²⁹ Cf. Radford (1990), who also relates the absence of SAI to the absence of COMP.

³⁰ Cf. Noonan (1989) for a more thorough discussion of the relevant licensing mechanisms.

³¹ Chomsky (1986b) assumes adjunction of the wh-phrase to VP, but he argues against adjunction to IP. There seems to be no strong reason for this, however as only adjunction to arguments i.e., NPs and CPs is blocked in principle.

³² Huang (1988) suggests that there is a Q-marker generated under INFL in his analysis of certain types of A-not-A questions (which correspond to yes/no questions) in Chinese.

³³ That there is an operator moving to SPEC/CP in yea/’no questions, either at S-structure (English, German, etc.) or at LF (Quebec French, French embedded questions, Chinese, etc.) is argued for in Noonan (1992). In fact, some children seem to assume such a marker on INFL, producing yes/no questions like e.g., Is the boy is running? (We thank Lydia White for pointing this out to us.) A similar phenomena can occasionally be observed in German children’s yes/no questions.

³⁴ It is not so clear how they unlearn this, i.e., what would be conclusive evidence to restructure the grammar to set SPEC/IP as an A-position.

³⁵ See Pesetsky (1989) for a related proposal on the adult grammar of English.

³⁶ This follows from their identification: since they are null they have to be identified though a structural position in the tree. In other words, they have to be identified by word order.

³⁷ Cf. Spanish, where VSO order is licensed in questions, negations and under certain matrix verbs.

³⁸ See White (1982) for a discussion of the distinction between the notions of input and intake.