Overview of the literature on initial and final negation in child speech

It has been observed that children during a period in language acquisition produce utterances with an initial or final negative element. This phenomenon has been observed in different languages (Bellugi 1967; Bloom 1970; 1991; De Villiers and De Villiers 1979; Drozd 1995; Hoekstra and Jordens 1994; Kaper 1975; McNeill 1971; McNeill and McNeill 1968; Wode 1977) and has been reported from American Sign Language as well (Newport and Meier 1985). Examples from English child speech (from Bellugi 1967) include no I see truck, no go back, not Fraser read it, and chew mouth # no. Several positions have been taken as to what this phenomenon might represent.

Sentence-external negations

Bellugi (1967) considers such utterances to be representations of an early stage in the syntactic development of negation, in which negative sentences are formed by attaching a negative element to a `sentence nucleus' - the structure of which is left undisturbed - in one of the following ways:

According to Bellugi, this pattern represents a primitive abstraction, unrelated to the adult model, which in a later stage is replaced by a much more complex system, consistent with sentence-internal negation as it occurs in adult speech.

Based on Bellugi's findings, McNeill (1971) assumes these early negative utterances to represent the not yet abstract deep structure of negative sentences as given by Klima (1964). According to McNeill, such utterances reflect one of children's innate capacities for language, an ability to deny a proposition by attaching a `minus sign' to the proposition denied (McNeill 1971: 33). Utterances with sentence-external negation occur - universally, as McNeill assumes - before any of the transformations involved in English negation have been acquired and sentences necessarily have the form NEG + S or S + NEG. Moreover, negation of this kind persists, despite changes elsewhere, until a child has acquired an order-changing transformation for negation (McNeill 1971: 33). In this view, the disappearance of the structure NEG + S or S + NEG is causally connected with the emergence of sentence-internal negations. As soon as the transformations involved in negation are acquired - in the case of English children, when negation is affixed to the auxiliary verb within the sentence, during the stage described by Bellugi (1967) as Period C - sentence-external negative markers die out.

Reduced surface forms

Bloom (1970; 1991) has challenged this interpretation. She argues that children's early negative utterances are in essence quite similar to adults', only they represent a `much simpler, fragmented, and far more generalized' version of the adult system. Specifically, she contests Bellugi and McNeill's claim that the acquisition of negation involves the replacement of sentence-external with sentence-internal negation, by stating that there exists no such stage of sentence-external negation in early child speech. Bloom noted that the majority of children's early negative utterances is subjectless, which makes it hard to decide whether the subject is deleted from the left or the right of the negative element - that is, whether it is deleted from a NEG + S structure or from a Subject + NEG + Predicate structure. Bloom's interpretation is that these utterances are the reduced surface forms of more complex underlying structures, in which the initial position of the negative element is a consequence of the omission of other constituents, among which the subject. For instance:

Bloom assumes the underlying structure of the child's utterance to be roughly, `X no Y window' (where X and Y represent inferred underlying category symbols with the grammatical functions of sentence-subject and predicate) (Bloom 1970: 155). According to Bloom, it is precisely the inclusion of negation which is responsible for the omission of these categories, as negation constrains sentence length. The operation of negation increases the syntactic complexity of sentences to such an extent that it has a limiting effect on production. Therefore, some constituents of negative sentences are not realized in the surface structure.

The basic contrast of Bloom's approach with that of Bellugi and McNeill is that, whereas Bellugi and McNeill assume that early negative sentences are formed by adding a negative element outside an affirmative nucleus or proposition, Bloom assumes that inclusion of negation within a sentence reduces its surface structure by nonrealization of other constituents, leaving the negative marker in initial position.

A problem with Bloom's approach is that it appears applicable to almost any non-adult-like negative utterance, as it does not specify exactly what reductive transformations are involved in children's early negative sentences. See for instance the following utterance:

It now appears not incompatible with Bloom's approach to assume that the child's utterance is the reduced surface form of an underlying structure with a sentence-internal negative element: ('t jongetje heeft) nee blonde haar. [(the little-boy has) no (non-quant.) blonde hair.] There is, however, no independent criterion by which to decide whether this is really the correct interpretation of the above utterance (see also Brown 1973).

Formally and functionally anaphoric negation

As De Villiers and De Villiers (1979) pointed out, the test case is constituted by initial-negative utterances containing a subject. In her own corpus, Bloom found that the initial negation in such utterances was anaphoric in function, referring back to an earlier event or proposition rather than negating the proposition to its right. For example:

Although the intonation of such utterances did not differ from that of negative utterances (i.e. there was no pause after no), their contexts made it clear that they expressed affirmative meanings. A survey of the contexts of initial-negative utterances in the Dutch corpus also revealed several at first sight negative utterances to be affirmative, expressing a positive alternative to a preceding utterance. For instance:

Likewise, when De Villiers and De Villiers (1979) investigated the meaning of initial-negative utterances containing a subject in Bellugi's corpus, they found only very few examples in which the initial negation unambiguously applied to the sentence to its right.

Miscategorization of adult input

In contrast with the findings from Bellugi's corpus, De Villiers and De Villiers (1979) found, in their own corpus, a large number of initial-negative utterances with an apparent subject and a nonanaphoric meaning. For instance:

As the child's utterance is a rejection, it cannot be interpreted as having an anaphoric meaning (no, let Daddy dress me). This appears to be one of the crucial examples, then, in which the negative element precedes the subject. Bloom (1991), however, interprets such rejections as utterances in which the matrix verb want is omitted, thus with the apparent subject actually being part of the complement: (I) no (want) Daddy dress me.

The De Villierses' (1979; 1985) account for the differences found between their corpus and that of Bellugi is that children's use of sentence-initial negation reflects the pattern most commonly encountered in the adult input. The initial-negative utterances from the child in the De Villierses' corpus were rejection negations, corresponding to the most commonly used pattern for rejection in the speech of the child's parents, usually an initial no attached to a polite negative sentence, as in No, let Daddy do it first. De Villiers and De Villiers traced similar correspondences in form and function between child speech and adult input for the children in the corpus of Bellugi (for related observations, see Pea 1980a). Drozd (1995) has challenged this account, pointing out that the De Villierses' account of children's early sentence-initial negations is based on a not necessary assumption.

Metalinguistic negation

An alternative account is given by Drozd (1995), who examined the viability of Horn's (1989) suggestion that children's sentence-initial negations are an early form of metalinguistic negation. In an extensive corpus of English child speech, Drozd investigated whether children's nonanaphoric, initial-negative utterances fit the interpretation don't say X or no way X (where X is a previous utterance), taking into account the characteristic properties of utterances with exclamatory metalinguistic negation: being echoic to a previous utterance and expressing objection or rectification. An example of such an utterance is given below (from Drozd 1995), with Drozd's interpreted meanings in brackets:

Drozd found that the vast majority of children's initial no/never-sentences in the corpus displayed these characteristic properties of exclamatory negation, and less than half of the initial not-sentences. With respect to the latter, Drozd pointed out that these sentences were formally ambiguous between contrastive constituent negation and contrastive sentence negation, which is also a grammatical pre-sentential negation construction. Drozd concluded that the children's pre-sentential negatives, well-formed when seen in their discourse contexts, constitute an independent class of negation rather than an early derivational variant of sentence-internal negation. Therefore, the task for the child is not to learn to produce internal sentence negation but to learn the appropriate lexical forms for expressing exclamatory negation (Drozd 1995: 607).

The Dutch corpus displayed only some exceptional occurrences of sentence-initial negatives which were compatible with a paraphrase as exclamatory metalinguistic negations. For instance:

The low frequency is in accordance with Drozd's contention that, since exclamatory negation is a highly specific use of negation, rarely used in adult discourse, this type of negation should also be infrequent in child speech.

It is precisely this highly specific character of exclamatory negation which makes it seem unlikely that it is the appropriate interpretation for all children's early nonanaphoric sentence-initial negations. With respect to the examples cited in Drozd's study, it seems to me that an interpretation as metalinguistic exclamatory negation sometimes is too strong, given the context of these utterances. Specifically, an objection or rectification meaning - the basis of metalinguistic negation - does not always seem to be the only plausible interpretation. In interpreting the initial-negative utterances in the corpus, Drozd used the following criteria: A negative was coded as consistent with a metalinguistic paraphrase if an exclamatory negative consisting of no way, don't say, etc., followed by the child's negative complement was felicitous in the discourse context. A metalinguistic negative was considered to be felicitous in context if an objection or rectification meaning was consistent with the logic of the discourse (Drozd 1995: 593). By these criteria, the following pre-sentential negative was interpreted as an exclamatory negation (example taken from Drozd 1995):

According to Drozd's interpretation, the child's negative utterance can be paraphrased as don't say it has a hole in it! or no way it has a hole in it! In my opinion, such a paraphrase is too strong, given that the adult utterance the child is reacting to is not a statement but a question (which, moreover, seems to be a quite neutral one). Although it is possible to view the child's utterance as an objection, it is not the only plausible option. A denial interpretation (no, it doesn't have a hole in it.) also appears to be consistent with the logic of the discourse. Another example is the utterance in (7) given above. Drozd interprets the child's utterance as an objection, but it may appear to be equally well compatible with a rejection meaning, in which a reductive transformation (Bloom 1991) has left the negative element in initial position: (I) no (want) Mommy cut it.

The same point can be made with respect to the Dutch data. Only very rarely is an example found which clearly has the function of exclamatory negation (as in (8) and (9) above). Other possible examples are ambiguous between an interpretation in terms of objection or rectification and other negative interpretations. For instance:

The child's utterance might be an objection against the adult's way of saying that a flower fell out of the vase, but it might also be the reduced surface form of (I) not (want) the flower out of the vase.

Since, as was pointed out above, Bloom's reduction analysis is not constrained in a principled way, the interpretation of such utterances may appear indecisive. It should be pointed out, however, that Drozd's interpretation is not solely based on the function of metalinguistic negation, but also on its form - that is, whether it is echoic to a previous utterance. Many of the initial-negative utterances cited in Drozd (1995) indeed have a form which is strikingly similar to an immediately preceding (adult) utterance. Still, I do not consider that to be the decisive factor in interpreting children's initial-negative utterances as well-formed metalinguistic exclamatory negations. An alternative account, which will not be further elaborated here, is that these utterances were not always deliberate metalinguistic negations, but that young children's limited capacities to produce structurally correct negative sentences left the negation of a previous utterance as a strategy to communicate a specific negative meaning.

Initial and final negations: one phenomenon?

In Bellugi's (1967) observations about English children's early negative utterances, mention was made of both sentence-initial and sentence-final negations. In the subsequent studies of early negative development discussed above, the focus of attention has shifted towards sentence-initial negation only, leaving sentence-final negation out of consideration. This state of affairs must be attributed to the fact that final-negative utterances in English child speech occur rarely, compared to initial-negative utterances. The few examples found in Bellugi's corpus were incorporated in a general rule of sentence-external negation, but as such utterances occurred extremely rarely in the corpus studied by Bloom and, presumably, also the De Villierses' corpus, these were not elaborated on any further.

In the Dutch corpus, however, utterances with a sentence-final negation marker - both niet (not) and, less often, nee (no, non-quant.) - alternate with initial-negative utterances. See the examples given below:

As such utterances occur in roughly the same period as sentence-initial negations, it might be conjectured that they are functionally and/or formally related. Yet, the different analyses of sentence-initial negations discussed above do not appear to be applicable to the Dutch sentence-final negations. Firstly, it is highly unlikely that such utterances result from a reductive transformation. Also, they appear to be incompatible, both formally and functionally, with a metalinguistic negation interpretation. And, finally, also an account in terms of anaphoric negation appears to be inappropriate. Although strictly speaking, the final negative element in these utterances is anaphoric, as it applies to a proposition to its left, its function is totally different from anaphoric negation in adult discourse, in that it appears to have scope over the preceding proposition.

It is also interesting to note that there is an intermediate stage in which children produce utterances with double negations - niet (not)/niet (not) or niet (not)/nee (no, non-quant.) - one of which is now in its correct sentence-internal position, the other of which is still attached at the end: