2. Radford's Minimal Projection Hypothesis
2.1 Stage One
2.1.1. Null Subjects
2.1.2. Negations and Questions in Stage One
2.2. Stage Two
2.3. Stage Three
2.4. Final Remarks on Radford's SCH
3. The Lexical Learning Hypothesis by Harald Clahsen et al
3.1. Underspecificity and Verb Placement
3.2. Lexical Learning and Syntax
3.2.1. Subject-Verb Agreement
3.2.2. Determiner Phrase and Case Marking
3.3. Final Remarks on Clahsen et al.'s LLH
4. Critical Points and Possible Way Out
Language acquisition seems to be subject to a paradox like probably all matters connected with the human language: every-one acquires his or her first language with relative ease in a few years – it is a matter of course (except if disabilities or social deprivations obstruct this natural process) - but even the most learned scholars cannot offer a satisfactory and all-encompassing explanation about how this process works. A number of different approaches exists which can be subsumed under the headings "behaviourist", "interactionist", "nativist", "cognitivist" or the like, and each of these major currents bifurcates into sub-theories which often enough seem to be incompatible with each other again.
In this essay I am going to present and discuss two hypotheses of language acquisition which follow the generative approach, i.e. which presuppose the existence of an innate language acquisition device (LAD) helping us solve the task of acquiring the complicated linguistic system by providing a "ready-made" underlying fundamental structure, the Universal Grammar (UG). Accordingly Andrew Radford, and Harald Clahsen et al. assume that the rules of UG are with the child from the very beginning, yet that the final build-up of the internal grammatical structure is triggered by lexical learning processes. Radford identifies three major stages of developmental phases with children acquiring English syntax, whereas Clahsen et al., studying children acquiring German, find plausible explanations for the developmental process by a notion of underspecified functional categories.
2. Radford's Minimal Projection Hypothesis
Radford starts from the assumption, that "syntactic structures are projections of the lexical items they contain". If this is correct and if the observations fit that a) languages differ in their lexical items and b) that children typically learn some types of words earlier than others, language acquisition must necessarily be a step-by-step building process which is in line with the possibilities laid out by UG. Radford holds, on the basis of evidence by other studies on language acquisition, that children first learn content words, i.e. words which clearly refer to things and bear semantic significance, like nouns, verbs, adjectives and prepositions, whereas grammatical words, like determiners, complementizers, auxiliaries etc. are usually learned at a later stage. It is supposed that the latter items are more complex in their meaning and usually less phonetically salient to be as easily perceived by the child as their lexical counterparts.
A further presupposition adopted by Radford is the "Minimal Projection Principle" – an application of the "principle of economy" as postulated by Chomsky in 1989. This principle has it that "a clause is only as big as it needs to be", which means that it is not required to always project up to the highest possible structure but only to the actually needed one, i.e. a normal declarative sentence in English is sufficiently described by a phrase marker projecting up to the functional category IP, whereas only an embedded clause with a complementizer or a question will project to the highest functional category CP.
In view of these assumptions, Radford postulates that children's initial clauses (CIC), which lack any sign of inflectional features, only project as high as VP, i.e. a phrase marker that has no functional category and whose structure is described by:
This is what Radford calls his stage 1 in syntax acquisition, or his Small Claus Hypothesis (SCH). Stage 2 and 3 are characterised by an ordered acquisition of the functional categories of finiteness first (inflection = Infl/ IP) and then of complementizer phrases (C/ CP) in accordance with the acquisition of the English words like auxiliaries or complementizers openly requiring and manifesting these functions.
If we follow Radford's SCH, a number of plausible explanations for linguistic phenomena in early child English can be given.
2.1. Stage One
2.1.1. Null Subjects
Before going into detail of Radford's approach, a further preliminary has to be established – the nature of null subjects. It is obvious that many early sentences of children lack a subject. "Want go", "sit lap" etc. are not uncommon utterances. In terms of the classification of null constituents as proposed by Rizzi, who makes a distinction between three different null constituents (pro, PRO, nc), Radford opts for the null constants (nc) to be applicable here. Pro is ruled out as it has to be licensed by inflectional morphology. PRO, as laid out by the Binding Theory of English adult grammar, is only possible as long as simple declarative sentences are concerned and the null subject occurs only with non-finite verbs, i.e. if this phenomenon were only observable in the very first stage. However, PRO is not licensed in utterances like "don't know", "can't" etc. where we have a clear indication of finiteness thanks to the auxiliary employed. Therefore Radford favours the third type of null subject which can also be found in adult speech, in the so-called "diary style". Thus continuity with adult grammar is established as a side-effect of this interpretation, as it were. This sort of null subject is characterised by its being discourse-identifiable thanks to the root specifier position, i.e. it is not c-commanded by any other identifier. By judging CIC null subjects to be nc, none of the actually observed utterances has to be dismissed as an odd speech error but all of them are in keeping with UG principles. Thus sentences with null subjects can be accounted for, be they negated, interrogative (as will be discussed in the next chapter), finite or non-finite.
2.1.2. Negations and Questions in Stage One
If we adopt the SCH, i.e. that the first utterances of children project only as high as VP and do not yet contain any functional categories, the question arises of how to account for the positions of negation and question words. In adult grammar wh -words are supposed to occupy the specifier position in CP. Negation, however, is apparently a little more complicated to characterise – it seems to be a hybrid of a functional and lexical word, or in other words, its assignment to the one or the other category is problematic as Atkinson points out. Radford treats not as a lexical word although it is not obvious which theta-role might be assigned by it (i.e. one of the defining criteria for lexical words).
If, however, we leave this discussion aside, Radford offers a logical treatment of the phenomena of negation and wh -word position by assuming that at first both categories, negation and wh -constituents, are added to the SC as an adjunct. The child intuitively (thanks to UG) infers that these words must have scope over the whole sentence and puts them in front of the SC. This technique of simply adjoining new constituents to existing structures as a transitory stage until an extended structure emerges in conjunction with further requirements fulfils the principle of isomorphism, i.e. the tendency to employ the same sort of construction plan on whatever level of representation. The notion of adjuncts is also in keeping with the minimal projection principle, as no higher structure has to be projected unless necessary.
no Fraser sharpen it: [VP no [VP [NP Fraser] [V' [V° sharpen][NP it]]]
what Kitty doing?: [VP what [VP [NP Kitty] [V' [V° doing]]]]
Moreover, this analysis accounts for the unchanged word order children use when asking wh -questions. A simple adjunct does not require the raising of the verb – which due to its lack of functional features at this stage of development would constitute a problem in terms of UG.
Furthermore, null subjects pose no problem as the null constituent keeps figuring in root specifier position and can thus be discourse identified. Thus this hypothesis of adjunct mechanisms can account for null subjects in questions and in negations.
what say?: [VP what [VP nc [V' [V° say]]]]
no like celery: [VP no [VP nc [V' [V° like] [NP celery]]]].
2.2. Stage Two
Stage 2 in Radford's theory is characterised by the appearance of finite verb forms, i.e. by the first use of auxiliaries, usually around the age of 2. It is supposed that the first use of auxiliaries does not necessarily lead to a change of the subject position from VP specifier to Infl specifier as in adult grammar. Radford provides examples for apparently inverted word order in early declarative sentences containing an auxiliary. Thus the following analysis seems to be plausible:
is shoes off: [IP [I '[I° is] [VP [NP shoes] [V' [V°] [off]]]]]
The inverted word order is constituted by the subject remaining in VP specifier position although the finite verb "is" raises to I°. A similar mechanism may be responsible for the fact that in early object questions, the finite verb seems to agree with the question word and not with the subject, the question word simply uses the still free Infl specifier position as landing site:
what's the wheels doing?: [IP [NP what] [I' [I° is] [VP [NP wheels] [V' [V° doing]]]]]
From this example evidence may be derived that there is indeed a connection between Infl and agreement. As long as the structure is such that the question word is in Infl specifier position, agreement seems to be striven for with the question word. However, as soon as agreement is sought between subject and the finite verb, the subject can be assumed to occupy the Infl specifier position as in adult grammar.
A further phenomenon co-occurring with the eventual establishment of Infl specifier as subject position, i.e. that the subject raises from the SC position as VP specifier to the adult-like Infl specifier together with the occurrence of auxiliaries, is a coherent nominative case assignment. It is observable with children using SCs that their subjects often are assigned oblique case. Sentences like "me do it" or "my need her" etc. are not rare. Radford offers three hypotheses accounting for this phenomenon by the notions of structural, inherent and default case. However, all three hypotheses need further exploration. What is, however, striking is the fact that once subjects are raised to Infl specifier and thus are case marked by Infl', nominative case is regularly used. Thus case theory finds a plausible confirmation by Radford's stage 2 analysis, or - a little more cautiously formulated – the two theories are very much in keeping with each other.
However, if the subject NP fills the Infl specifier position, the landing site for wh-words is no longer available. If we apply the same mechanism as in stage 1, i.e. if we simply adjoin the wh -word, which would again be in line with the isomorphism principle, another phenomenon of child language can be accounted for – the lack of inversion in questions.
Where the other Joe will drive?: [IP Where [IP [NP the other Joe] [I' [I°will] [VP drive]]]].
The structure of the declarative sentence remains the same, only the question word is adjoined to IP.
2.3. Stage Three
At the age of around 2;6, children will enter stage 3 according to Radford. This stage is triggered by the acquisition of overtly used complementizers in embedded sentences which require a new sort of projection. Having a sentence like "see if swimming water's there", if can no longer be a an adjunct to the IP "swimming water's there". The IP has become a complement to the VP see but adjuncts are not permitted to "selected complements of lexical heads". Therefore if requires a projection of its own: the complementizer phrase which in turn creates an additional landing site for sentence constituents - the CP specifier position. This position is filled by wh -words in direct and indirect questions.
However, the question arises how children know that direct questions imply auxiliary inversion whereas indirect questions do not. From yes/no questions we may derive that the auxiliary occupies the C° position as the following example shows – the auxiliary in the last sentence occupies the same position as the complementizer if in the third one.
 cf. Crystal p. 234 f., Clahsen p. xv
 Radford (1996) p. 43
 However Atkinson (1992) draws our attention to the fact that the exact nature of the defining criteria which functional categories should fulfil has not yet been fully determined (cf. p. 258)
 Radford (1996) p. 44
 Grimshaw as quoted in Radford p. 44
 However, these assumptions are not shared by all scholars and stand in contrast to the Full Competence Hypothesis (FCH) as professed by e.g. Hyams or Harris & Wexler who, in their contribution to the same volume, favour the view that child grammar and adult grammar are more or less structurally identical. Atkinson (1996), however, shows that the methodological as well as the empirical base for FCH arguments is rather contestable.
 all examples in chapter 2 are taken from Radford (1996) if not indicated otherwise
 Atkinson (1992) p. 262
 In respect to negation in adult English, the theory underwent a change thanks to Pollock's (1989) introduction of NegP. The change becomes visible in the different Radford publications. In his 1996 article, Radford still refers to his 1990 publication in which he posits negation as an adjunct to VP also in English adult grammar, whereas in his 1997 book, negation is treated as a phrasal head on its own taking VP as its complement (cf. p. 231 ff).
 Radford (1996) p. 71