Up to general Table of Contents

Chapter 4 — The spontaneous speech data revisited

• 4.1 — Introduction
• 4.2 — Licensing restricted to classical negation
• 4.2.1 — The development of candidate licensers
• 4.2.2 — Delay factors
• 4.3 — No correct licenser; negative meaning
• 4.3.1 — Pseudo-licensing
• 4.3.2 — Primitive sentential negations
• 4.3.3 — Missing negation
• 4.4 — No correct licenser; affirmative meaning
• 4.4.1 — Contrastive denial
• 4.4.2 — The development of contrastive denial
• 4.4.3 — Questions
• 4.5 — Pseudo-licensing by nog
• 4.5.1 — Is nog also a pseudo-licenser for NPI-meer?
• 4.5.2 — PPI/NPI pairs and the Principle of Contrast
• 4.6 — Summary and Conclusion

Chapter 3 — The spontaneous speech data revisited

4.1 — Introduction

The spontaneous speech data and the experimental findings discussed in the preceding chapter led to the conclusion that young children already know that NPIs have a restricted distribution, but are still at the beginning of the task of acquiring the variety of correct licensing environments. As was shown by the error patterns in spontaneous speech, this task is not approached with a strictly conservative attitude.

In the light of this conclusion, the ways in which children's NPI usage deviates from adult discourse are to be regarded as intermediate restrictions on NPIs. In this chapter, these deviant patterns in children's spontaneous speech - the ungrammatical utterances with NPIs as well as the extremely limited licenser variability in grammatical utterances - will therefore be reconsidered in terms of a licensing system in development.

The investigation of these patterns will show that NPI licensing and negation are closely connected in acquisition and develop in tandem with each other. Young children's sensitivity to the restricted distribution of NPIs is that these expressions can only appear correctly in environments which are related to a negative polarity. With negative polarity at this young age still being very narrowly defined, NPIs do not yet appear in the full variety of correct licensing environments. Hence, as children's perception of what constitutes a negative environment develops and loosens up, the conditions are created for a more varied distribution of NPIs.

The developing concept of negation thus functions as the forerunner in the development of licensing environments for NPIs. Although this guideline is quite effective in discovering the variety of correct licensers, it also appears to result in patterns of usage which from an adult point of view are ungrammatical. The latter phenomenon, dubbed pseudo-licensing, is very interesting in that it shows that the deviant patterns of NPI usage in child language are closely associated with adult grammar, but do not yet fully conform to these rules. This indicates that children are neither strictly conservative nor overly general in the acquisition of NPIs, but follow a route between these two extremes, in which they are guided by their gradual expanding knowledge of ways to express negative meaning in general.

Up to Chapter 4 Table of Contents

4.2 — Licensing restricted to classical negation

In reanalyzing children's early NPI use, we will start by taking a closer look at the most frequently occurring pattern: the virtual monopoly of niet (not) as a licenser in correct utterances with NPIs. This one-sided use of NPIs in young children's speech as compared to adult speech was visualized in Figure 3.1: niet (not) is the only legal licenser used until the age of two and a half, and from this age until the age of four, only a slight increase in the variability of correct licensers was noted.

As became clear from the previous chapter, a first attempt to interpret this very restricted usage - that children only specify niet (not) as a proper NPI environment (i.e. Hypothesis I) - was not supported by experimental evidence. The ERIC responses to the different test sentence types showed that, besides niet (not), other correct environments (i.e. geen (no, quant.) and alleen (only)) to a certain extent passed as acceptable to the children as well, whereas sentences without a licenser tended to be not accepted. These results point in the direction of the alternative explanation (i.e. Hypothesis IV): children's knowledge goes beyond the particular combination of an NPI with niet (not); they know that there is a variety of licensing environments. In order to give more substance to this alternative explanation, the question now must be answered why children's comprehension of licensing, as apparent from the experiment, exceeds the monotonous pattern in production, as apparent from spontaneous speech.

As the experimental findings show that young children do not consider NPI + niet (not) as the only correct combination, it seems plausible to assume that the absence of other possible NPI + licenser combinations in spontaneous speech is due to a lag between comprehension and production. In production, children might necessarily be limited to niet (not) as a licenser because the wider variety of licensers is not yet available; their productive vocabulary might not yet include other possible licensing expressions.

Up to Chapter 4 Table of Contents

4.2.1 — The development of candidate licensers

In order to assess the viability of this explanation, the object of study must be widened, to include an investigation of the general development of `candidate' licensers. Questions of particular relevance are: when does the variety of possible licensers appear in child speech and, once these expressions do occur, how adequate or comfortable are children in using them?

To work these questions out, the longitudinal studies in the spontaneous speech corpus (cf. Table 2.1) are once more investigated, in order to map the emergence of the possible licensers for hoeven and meer. According to the laws of negative polarity, hoeven and meer are classified as weak NPIs since they are licensed by classical, regular, as well as minimal negations (Zwarts 1995; Van der Wouden 1994a). The search in the children's data thus included the negative expressions in Table 1.4, as well as a range of other possible licensers not listed in this table, such as minder (less), alleen (only), and pas (until). The expressions which were found to occur in the children's speech are shown in Table 4.1, in order of appearance.

It is important to note that the ages given in this table are approximations: the average of the ages at which the first occurrences were found in the data of several children. In data files of spontaneous speech, the presence or absence of certain expressions is subject to various coincidental factors, such as the topic of conversation and the frequency and duration of the speech samples. Most probably then, these expressions emerge somewhat earlier than at the ages mentioned in the table. This qualification obtains in particular for the first possible licenser to appear: niet (not). The emergence of this expression was hard to plot, since in most of the longitudinal studies, niet (not) is already present in the first data file. It is only in the studies of children of whom very early data are available that niet (not) has not yet occurred in the first files. On the basis of these early studies, the average appearance of niet (not) was set at 1;10.

It becomes clear from the table that the classical negation niet (not) does indeed occupy a monopoly position for quite a long period of time. It takes about half a year until another possible licenser finally makes its appearance: the regular negation geen (no, quant.) at the age of 2;04. After this first step towards more variety has been taken, there is a slow but steady increase in possible licensers. At the end of the recordings, at the age of about four, twelve possible licensers for hoeven and meer were found to occur.

These findings show that the solitary use of niet (not) as an NPI environment in early child speech can be explained by the very limited negation vocabulary at this age. Although children may understand a larger variety of negative expressions, they do not use them in production. The fact that NPIs until the age of two and a half are only licensed by niet (not) (cf. Figure 3.1) thus appears to be a direct consequence of limitations on production: at this young age, other possible licensers than niet (not) are just not yet available.

Up to Chapter 4 Table of Contents

4.2.2 — Delay factors

With respect to the period between 2;07 and 4;00, the findings require a closer look. As Table 4.1 shows, several possible licensers are now present in the children's speech, but at the same time, niet (not) is still by far the most frequently used licenser, as is evident from Figure 3.1. This discrepancy can be explained when the relatively low frequency of the newly acquired negative expressions is taken into account (cf. the N column in Table 4.1). To give an example: the proportion of geen (no, quant.) to niet (not) is 1 to 28,5 (compare this to the proportion of geen (no, quant.) to niet (not) in the adults' speech in the same corpus: 1.637/12.567, a ratio of approximately 1 to 7,5). Thus, although other possible licensers now, strictly speaking, are present in the children's speech, their frequency is still only marginal in comparison with niet (not). In this light, the very slight increase in licensing variability between 2;07 and 4;00 appears to be a quite accurate reflection of the development towards more variety of negative expressions in the same period, which also proceeds in rather small numbers.

Moreover, once other possible licensers have been added to the productive vocabulary, it may take some time before children completely master the syntax and semantics of these newly acquired negative expressions, which explains why they so hesitantly appear as licensers in utterances with NPIs. This appears to be particularly the case with geen (no, quant.) (see also Kaper (1975) for a discussion of problems in the acquisition of geen). In the spontaneous speech corpus, there is one study (i.e. the Peter files) in which geen (no, quant.) appears quite early, at the age of 1;11. In the beginning, geen (no, quant.) occurs only in one- and two-word utterances, apparently as a rejection negation with scope over propositions as a whole, rather than as a negator of NPs, as would be correct:

It is only months later that the child starts using geen (no, quant.) as a negator of NPs:

The latter example also illustrates another delay factor: the fact that the use of niet (not) still tends to be overextended to utterances in which geen (no, quant.) would have been correct, long after geen (no, quant.) has made its first appearance in the children's speech. [1] See the following examples:

Compare the last example to the following one, from the same child five months later, in which the formerly used niet (not) eventually has given way to geen (no, quant.):

The reflection of this overextended use of niet (not) can be observed in utterances with NPIs, in which children appear to relapse into using the better known negation niet (not) as licenser, rather than geen (no, quant.), which would have been correct:

Due to these factors, it may take a while before children's expanded knowledge of negative expressions becomes apparent in language production. Yet, the few instances in which other negative expressions than niet (not) now are used as licenser make it clear that expansion of the negation vocabulary is necessary for children to unfold their basic knowledge of the restrictions on NPIs: there is a clear correspondence between the order of appearance of negative expressions and the order in which these forms appear as licensers for NPIs. Thus, the regular negations geen (no, quant.) and niks (nothing), the first negative expressions to appear after the classical negation niet (not), also occur as the first other licensers for NPIs:

Niemand (no one) and nooit (never), regular negations which appear only months later in the children's speech, also occur correspondingly later as NPI licensers:

And finally, in diary data, there are some examples of NPIs with the minimal negation maar (just) as a licenser, an expression which also appears relatively late:

It thus seems to be the case indeed that the expressions which at a certain age are used by children as NPI licensers are a reflection of the progression seen in the development of negation. Expansion of the negation vocabulary gives children the opportunity to unfold the already present sensitivity to the restricted distribution of NPIs, and the one-sided use of niet (not) gradually gives way to more variety in licensing, thus approaching the adult model of licensing more closely.

In the next section, it will be shown that this analysis of a tandem development of negation and licensing sheds a new and interesting light on an error pattern in NPI use which first seemed to be in conflict with the very restricted distribution of NPIs from the onset: the utterances with a negative meaning but without a correct licenser.

Up to Chapter 4 Table of Contents

4.3 — No correct licenser; negative meaning

The spontaneous speech data showed children sometimes using NPIs in negative utterances lacking a correct licenser (cf. section 3.3). Some of the examples given earlier are repeated here for convenience:

The clue to the negative meanings of these utterances is the shaking of the head, as in (18), the affixed negative morpheme nee (no, non-quant.), as in (19), or the context, as in (20). In the light of the experimental findings and the proposed analysis in the previous section, this error pattern will now be reconsidered.

Up to Chapter 4 Table of Contents

4.3.1 — Pseudo-licensing

In the spontaneous speech examples given above, headshaking and nee (no, non-quant.) occur in combination with structurally affirmative sentences. In adult grammar, such combinations are not allowed. As was set forth in section 3.3, deictic headshaking and anaphoric nee (no, non-quant.) cannot carry the total weight of negation for an accompanying utterance; they can at best function to underline a negative meaning, thus accompanying propositions which already contain some negative expression. This led to the assumption that in child utterances like (18) - (20), the NPIs themselves might function to express negation. If, at a certain stage in acquisition, NPIs for the child were expressions with an inherent negative meaning, a separate negation marker would be redundant. In this view then, the negative meaning of an utterance like (20), without any overt negation marker, would stem from the NPI itself. In the same way, headshaking and nee (no, non-quant.) in utterances like (18) and (19) would function as natural underliners of the negative meaning already expressed by the NPI. This hypothesis, that NPIs in the children's grammar might function as inherently negative expressions (i.e. Hypothesis II), was experimentally tested. The results, as discussed in the previous chapter, convincingly showed that this hypothesis was not viable: the *hoeven sentences were not acted out as having a negative meaning, and the ERIC responses showed that, in general, these sentences were not considered as correct.

In the light of these experimental results, a reinterpretation of this error pattern is now suggested: in utterances like (18) and (19) above, headshaking and nee (no, non-quant.), rather than being underliners of a negative meaning inherent in the NPI, must be intended to have the function of licensers. Thus, a licensing expression is far from redundant and is present in these utterances, only in a form which does not conform to adult rules of licensing. Headshaking and nee (no, non-quant.) are then reinterpreted as pseudo-licensers - that is, NPI environ- ments with an obvious `flavor' of licensers, but unacceptable as such in adult grammar. In this view, pseudo-licensers thus are to be regarded as representations of a licensing system in development in the children's speech.

Up to Chapter 4 Table of Contents

4.3.2 — Primitive sentential negations

In the light of the analysis proposed in section 4.2 - that the environments for NPIs in the children's speech are reflections of negation development - the puzzle now seems to fit. It is known from various studies about the acquisition of negation, in both English and Dutch, that headshaking, no/nee, and not/niet are the earliest negation markers present in child speech (Bellugi 1967; Bloom 1970; 1991; Hoekstra and Jordens 1994; Kaper 1975; Lord 1974; Pea 1980a). Remarkably, it is precisely in these environments that the first NPIs occur in early child language. This is an obvious corroboration of the analysis that the development of NPI licensing is embedded in negation development: the expressions which at a certain age are known to children as negative markers are also used as licensers and pseudo- licensers for NPIs.

This analysis can be further checked by investigating how, in young children's spontaneous speech, headshaking and nee (no, non-quant.) are more generally used in utterances without NPIs. The line of thought is as follows: if headshaking and nee (no, non-quant.) in young children's speech appear in the role of NPI licensers every now and then, then their status in child grammar must be different from adult grammar. Headshaking and nee (no, non-quant.) at this young age appear to be on an equal footing with niet (not), the most frequent occurring licenser and the first sentential negation to appear in the children's speech. If this indeed is the case, then we would expect to find more general examples in the children's speech where headshaking and nee (no, non-quant.) are used instead of niet (not), thus functioning as sentential negations.

In the spontaneous speech corpus, a search was carried out for utterances in which headshaking and nee (no, non-quant.) combine with structurally affirmative sentences to form utterances with a negative meaning. As for headshaking, the investigation was inhibited by the fact that non-verbal context is not systematically coded in the recorded speech corpora. Therefore, we have to rely on diary data and on studies in which particular attention is paid to both verbal and non-verbal ways to convey negative meanings.

In Pea's (1980a) study on the acquisition of negation, it is explicitly mentioned that negative meanings are first expressed non-verbally, by means of headshaking, before they are expressed in speech. Pea explains this gestural priority as emerging from a natural reaction of aversion. The same explanation for the early appearance of headshaking is given in Kaper (1975), who describes a child expressing her aversion against more food by turning her head away. At a later age, the shaking of the head from left to right became a more general sign of rejection or protest.

In such studies on the development of negation, it is indeed mentioned that children sometimes combine headshakes with structurally affirmative utterances. For instance, Bellugi notes: `the following sentences were probably negative sentences for Sarah (based on mother's interpretation, other signals such as shaking her head, or some clear context: He bite me yet; Why need them more? I play either, huh?; I got any; I want any; You want any' (Bellugi 1967: 38. My emphasis).[2] More specific examples are given by other authors:

As for nee (no, non-quant.) in combination with a structurally affirmative sentence, far more data are available. This is not only because nee (no, non-quant.) is linguistically expressed and thus can be checked in corpora of recorded speech, but also because this negative morpheme, in different languages, has been the subject of extensive study, precisely in its function as a sentential negation in early child speech (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). In the literature, different positions have been taken as to what this phenomenon might represent, ranging from the conjecture that it is children's own, primitive abstraction of negation to the hypothesis that it reflects young children's competence in using an independent, well-formed class of discourse negation. As an exposition of this matter is of no direct relevance to the present section, I will not go into details here. An overview over the literature on this topic is given in Appendix B. For the present discussion I will just consider some Dutch examples of the phenomenon.

Kaper (1975) and Hoekstra and Jordens (1994) both report structurally affirmative utterances which are preceded or followed by nee (no, non-quant.). In Hoekstra and Jorden's data - longitudinal diary notes of one child - the use of nee (no, non-quant.) as a sentential negation dies out at 1;11. Data from other sources, however, indicate that this stage may last considerably longer for other children. For instance, Kaper (1975) cites the following example, from a two and a half year old child:

In other diaries and in the recorded speech samples of the Dutch spontaneous speech corpus, numerous examples of both sentence- initial and sentence-final nee (no, non-quant.) were found:

And also a few examples were found, in the diary data of one child, with nee (no, non-quant.) occurring in sentence-internal position:

These examples of headshaking and nee (no, non-quant.) in the role of sentential negations once more corroborate the analysis that the development of NPI licensing must be embedded in negation development. Negation - more specifically, what constitutes negation for young children - appears to be the forerunner of environments which are used as NPI licensers in the children's speech. This shows that children, at a young age already, have a basically correct insight into the licensing conditions on NPIs. In general, negation proves to be a quite successful guide for NPI licensing: it generates many correctly licensed utterances with NPIs, while at the same time preventing children from making overgeneralization errors. In tandem with expansion of the negation vocabulary, there is a step by step development towards more variety in licensing environments. Interestingly, however, this guide also appears to lead the child astray sometimes, generating utterances which from an adult point of view are incorrect. The question as to how these errors might disappear again from the children's speech is taken up in Chapter 5.

These utterances with headshaking and nee (no, non-quant.) as pseudo-licensers provide very valuable information about the basic thread running through the acquisition of NPI licensers: children use NPIs neither strictly conservatively nor totally unrestrictedly, but follow an intermediate path, guided by their developing knowledge about what constitutes linguistic negation. [3]

Up to Chapter 4 Table of Contents

4.3.3 — Missing negation

Pseudo-licensing provides an explanation for ungrammatical utterances like (18) and (19), in which NPIs are accompanied by headshaking or nee (no, non-quant.). There is, however, another subcategory of the `no correct licenser; negative meaning' error pattern which now still has to be dealt with: utterances in which not even a hint of negation is present and where the only clue to a negative meaning is the context. The example given earlier as (20) is repeated below:

The negative expression which would have been a correct licenser seems to be simply left out. This is also indicated by the original transcription of this utterance: ik hoef (geen) wijkoek (I need (no) honeycake).

An explanation for this phenomenon again points in the direction of the development of negation. A missing negative element appears not to be unique for utterances with NPIs, but is more commonly observed in young children's utterances with a negative meaning. Several examples of this phenomenon are given in Bloom (1970) and Lord (1974), and were also found to occur in the Dutch corpus of spontaneous speech. For instance:

Interestingly, Lord (1974) noted that - at least for the child she studied - such superficially affirmative utterances with a negative meaning systematically had an atypical intonation contour in which the verb, and often the whole utterance, was produced with an elevated pitch. Example (35), for instance, was uttered with specific stress on wan' put it. Lord's conclusion is that intonation contour may function as an early sentential negation. If we follow her conclusion, we might now venture that `intonational negatives', as they are dubbed by Lord (1974), might be used as another pseudo-licenser for NPIs. Unfortunately, as stress patterns are not systematically coded in the recorded speech corpora, this interpretation - while seeming to be quite plausible - cannot be corroborated.

Another qualification needs to be made as well. Lord noted that negative meaning indicated by intonation contour occurred only during a short period, around the age of two. But utterances in which NPIs appear without any obvious negation marker also occur at a later age. Moreover, such utterances do not always have a particularly marked intonation contour, like for instance example (33). For these cases then, it seems more reasonable to assume that the absence of a licenser is due to underarticulation. This is also indicated by the occurrence of quasi-errors of this type: utterances with NPIs which are not quite grammatical but not blatantly ungrammatical either, where underarticulation seems to be responsible for the fact that only a hint of licensing is present, expressed by a dummy-element ne or e:

It is important to note that in either way - that is, whether negation is missing because of underarticulation or because intonation contour fulfils its function - these utterances are only superficially incorrect; the semantic representation is unaffected. Hence, when hoeven and meer occur in such utterances with clear negative intent, the basic condition on correct NPI use is not violated.

Up to Chapter 4 Table of Contents

4.4 — No correct licenser; affirmative meaning

Two patterns in children's early NPI use have now been reanalyzed: the supremacy of niet (not) as a licenser, and the seemingly unlicensed utterances with a negative meaning. The reanalysis of these patterns, which on the surface appear to be incompatible with each other (cf. section 3.7), revealed an underlying uniformity in terms of negation development. The young child's small inventory of linguistic and paralinguistic means of expressing a negative meaning is called upon when NPIs are used, resulting in licensed and pseudo-licensed utterances. Somewhat later, when the negation vocabulary expands and comes to include other negative expressions, such as geen (no, quant.) and niks (nothing), there is a parallel increase in the variety of correct licensers.

One spontaneous speech pattern is now left to be explained: the utterances without a correct licenser and with an affirmative meaning. Here, there appears to be an incongruity between the spontaneous speech findings and the experimental results. As was pointed out earlier when the spontaneous speech examples were presented, the meanings of these utterances seem to suggest that hoeven is used as a regular verb without any particular restrictions.[4] If this really were the case, these examples would run counter to the conclusion reached so far, that young children use NPIs in a restricted manner and are in the process of acquiring the variety of correct licensing environments. The experimental evidence, however, showed that the regular verb hypothesis (i.e. Hypothesis III) was not viable: the children, more than half of the time, changed the *hoeven sentences in their ERIC responses, inadvertently correcting them by exchanging hoeven for another verb or by adding a licenser.

In the light of these findings, we may now conjecture that the affirmative utterances with hoeven in the children's spontaneous speech are not what they look like at first sight. In the same way as the negative utterances without a correct licenser, they might contain a pseudo-licenser, an element which for children can function as licenser but which is not grammatical as such in adult speech. The presence of such a pseudo-licenser in the affirmative utterances in spontaneous speech, and its absence in the *hoeven items presented to the children in the experiment could be responsible for the incongruity between the spon- taneous speech findings and the experimental results.

Do the spontaneous speech data contain some distinctive element, setting these utterances apart from the *hoeven items presented in the experiment? A closer look at the children's spontaneous speech utterances and their contexts indeed reveals two characterizing features: they either are interrogative or they express a contrastively affirmative meaning. The contrastively affirmative utterances will be dealt with in the next section; the interrogatives will wait until section 4.4.3.

Up to Chapter 4 Table of Contents

4.4.1 — Contrastive denial

Some of the spontaneous speech examples given earlier are repeated below, supplied with contextual information. It now becomes clear that these utterances are emphatically rather than neutrally affirmative. They express a contrast to an earlier - explicit or implicit - negative proposition:

The above examples differ among themselves with respect to how strong and explicit the contrast is. They are ordered from the top downwards according to decreasing explicitness. The top example (41) expresses a very explicit, echoic contrast to an immediately preceding proposition by means of the affirmative intensifier wel (really/indeed). In (42), the echo effect is somewhat less direct, as there is an intermediate action separating the child's negative utterance from its counterbalancing affirmative contrast. Example (43) is strictly speaking not echoic, as there is no preceding negative utterance with hoeven. This utterance, however, counterbalances the implicit meaning of the child's earlier negative answer: nee (dat hoeft niet) [no (that needs not)]. In much the same way in (44), the preceding rejection of a piece of hard clay carries the negative implicature dat hoef ik niet (that need I not). Example (45) differs from the previously mentioned examples in that there is no preceding negative utterance. There is, however, a negative echo, due to the intonation contour of this affirmative utterance. With contrastive stress on drinken (something to drink), the child's utterance in fact is an implicit negative answer to the adult's question. The same implicit negative echo can be heard in the bottom example, (46), in which a contrastive intonation contour counter-balances a negative assumption in the non-linguistic context.

The central and recurring notion in these examples thus is that they reject earlier - implicit or explicit - negative propositions, in contrastively affirmative utterances. The contrasts expressed are, respectively:

• (41) the kettle (negative) versus the blanket (affirmative)
• (42) no more food versus (on second thoughts) more food
• (43) no new gasoline versus (on second thoughts) new gasoline
• (44) hard clay (negative) versus soft clay (affirmative)
• (45) eating (negative) versus drinking (affirmative)
• (46) animals that are allowed to come along versus animals that are not

It was also pointed out that hoeven is not part of the class of semi-NPIs, as this expression is much more strict with respect to such intensifying affirmation. Firstly, hoeven can only occur in contrastively affirmative utterances when these have the strongest possible echo effect. This requires a directly preceding negative utterance - a negative assumption alone is insufficient - to be contrasted with an affirmative utterance which is echoic up to deixis. The operation involved is that of metalinguistic affirmation: reversing a negative proposition ¬p into its affirmative counterpart p. Such utterances with hoeven are only allowed in an informal register and occur extremely rarely. In the corpus of recorded Dutch, one example of metalinguistic affirmation with hoeven was found:

Metalinguistic affirmation is the only possible `affirmative licenser' for hoeven. There is one other possibility for hoeven to be connected with a contrastively affirmative meaning, but that is in a much more implicit way: in contrastive elliptic clauses in which hoeven itself is omitted. See for instance the following examples:

The link with `affirmative licensing' in adult grammar throws a new light on the children's affirmative utterances with hoeven. It once more shows that apparently unlicensed utterances with NPIs in child speech are, in principle, not far from what is allowed in adult grammar. Examples (41) - (46), although they certainly are deviant from what is common use in adult speech, are not blatant violations of the restrictions on NPIs. Rather, they represent a sliding scale from infelicitous to more infelicitous use, where it is hard to draw the exact boundary between where infelicitous use stops and where grammatical use begins. For instance, when examples (41) and (42) are compared to example (48), it is striking that these utterances are virtually identical qua function; there is only a slight - but nevertheless crucial - difference qua form. The other examples, (43) - (46), do not fully conform to the requirement of metalinguistic affirmation. The negative echo in these utterances is not strong enough; the emphatically affirmative meaning does not obtain its force from a direct contrast to an immediately preceding negative utterance, but instead from an implicit negative assumption.

Importantly, the common and central denominator in examples (41) - (46) is their function as contrastive denials of negative propositions. Considering the fact that this, provided certain conditions are satisfied, indeed is a correct environment for hoeven in adult grammar, the contrastive denials as they occur in the children's speech at this age should be regarded as another instance of pseudo-licensing.

This analysis is corroborated by the fact that contrastively affirmative utterances with NPI-meer, which are not allowed in adult grammar, are not found in the children's speech either. For example, to express an affirmative contrast to a sentence like Hij is er niet meer. (He isn't here anymore.), one cannot say Hij is er wél meer. (He really is here anymore.). In this case, affirmative contrast should be expressed by means of the antonym of niet meer (not anymore), the positive polarity item nog (still), as in Hij is er nog. (He is still here.). The positive/negative polarity pair nog (still)/meer (anymore) in the children's speech will be further dealt with in section 4.5.1.

Up to Chapter 4 Table of Contents

4.4.2 — The development of contrastive denial

In section 4.3 it was argued that children's early pseudo-licensers are reflections of the broader development of negation. In much the same way, it appears that the use of contrastive denial as a pseudo-licenser is not a phenomenon exclusively related to children's NPI use, but one which also is embedded in a broader development: children's discovery of thesis and antithesis.

First, it should be noted that contrastive denial is bidirectional: it can be applied in two opposite ways. Either an affirmative proposition is denied and contrasted with a negative proposition (a negative contrast, resulting in a negative utterance), or a negative proposition is denied and contrasted with an affirmative one (an affirmative contrast, resulting in an utterance with no explicit negative element). The operation involved in both is in principle the same.

In the above discussion of contrastive use of hoeven, the focus was on denial expressing an affirmative contrast. In studies on the acquisition of negation, on the other hand, the emphasis has almost exclusively (with the exception of Keller-Cohen and Gracey 1979 and Pea 1980b; 1982) been on denial expressing a negative contrast, as this results in utterances with an explicit negative element.

In such studies, denial is conceived as one of a variety of functional categories of negation, such as rejection, nonexistence and disappearance (cf. Pea 1980a). The attention is towards describing the sequence in which these categories appear and are correctly used in child speech. Although there is no general consensus about the exact order in which these categories emerge - a problem is that there is no unequivocal classification, as different functions of negation may conflate - there is agreement that the function which is of particular importance here, denial, appears relatively late. In Bloom (1970) denial is defined as a negative utterance which asserts that an actual (or supposed) predication was not the case (Bloom 1970: 173). For the three children in Bloom's spontaneous speech study, it was found that this function appeared only after nonexistence and rejection negation were expressed. Bloom conjectured that this relatively late appearance of denial might be due to the fact that it involves a symbolic rather than a concrete referent: the child had to perceive the referent in something said (Bloom 1970: 219). The same line of argument is found in other studies (Keller-Cohen and Gracey 1979; Pea 1980a): since denial requires an abstract representation, it is considered to represent a cognitively more complex category of negation, which therefore appears later than negative categories which involve concrete referents, such as rejection and disappearance.

Contrastive denial can be regarded as a more elaborate version of plain denial. It is a dual-function negation, a no but, thus a negation `coupled with the expression of alternatives' (Keller-Cohen and Gracey 1979). In McNeill and McNeill's (1968) study of negation development, it is mentioned that there is a distinct form for this dual-function negation in Japanese, iiya, which means that what was just said is wrong and something else is right. The import of iiya is that one alternative (already mentioned or somehow in mind) is false and another is true. (McNeill and McNeill 1968: 53). McNeill and McNeill found that, for the Japanese child they studied, this particular form was acquired last, after three other negative forms, more or less corresponding to plain denial, nonexistence, and rejection, respectively. McNeill and McNeill explain the late appearance of iiya by the fact that it represents a higher-order negation, one which requires a child to hold in mind two propositions at once (McNeill & McNeill 1968: 61). The same argumentation is found in Bloom (1970) and Keller-Cohen and Gracey (1979). According to the latter, contrastive denials even require the simultaneous retention of three propositions: keeping in mind what has been asserted, what is wrong about that assertion, and what is right instead.

In Pea's (1980a; 1980b; 1982) work on the acquisition of negation, the emphasis is on a subset of the negatives captured by the denial category, to wit logical denial: the negation of a proposition given the truth conditions of language. In a sentence verification task, in which children commented on the truth or falsity of statements (including both false affirmatives and false negatives), Pea found that this function of negation emerges around the age of two. He also noted that two-year-olds are highly aware of the contrast between affirmative and negative utterances, and that it is common practice at this age to produce antithetical phrases, in which a statement is immediately contradicted and followed by its opposite (see also Schaerlaekens and Gillis (1987) for examples of thesis/antithesis practicing in bedtime monologues).

The children's contrastively affirmative utterances with hoeven appear to be instances of the partiality for antithetical phrases at this age. This is also illustrated by a comment of the father of the child who uttered (42), which is repeated below:

According to the child's father, the hoeft niet/hoeft wel contrast was repeated time and again during the rest of that evening.

Up to Chapter 4 Table of Contents

4.4.3 — Questions

A few ungrammatical hoeven-sentences with an affirmative meaning are now left to be dealt with. These were given earlier in section 3.4 and are repeated below:

(52)	op hoeft # huh? dis-is opp(l)akke(n )? (2;03.04) on needs # huh? this-is stick-on?
(53)	hoeft er op? (2;06.11) need there on?
(54)	Kim hoeft ook dit he? (2;07.30) Kim needs also this huh?
(55)	hoef jij ook? (2;11.20) need you also?

First of all, it should be noted that it is hard to pinpoint the exact meanings of these utterances, even when their contexts are taken into account. Apparently, the child's conversation partners also had difficulties interpreting these utterances: as will be shown below, two of the above examples gave rise to misunderstanding and were followed by a request for clarification by the child's parent. Yet, these utterances should not be done away with as just performance errors, since they display an intriguing recurrent pattern: they all have interrogative force. This is very interesting in view of the fact that questions indeed can function as licensers for NPIs. In adult language, however, licensing by questions is subject to rather subtle pragmatic restrictions. Apparently, these conditions have not yet been met in the children's early interrogatives with hoeven. Another interesting factor is that this pattern of usage - hoeven in questions - continues in later child data, but then in a form which more closely approaches adult usage.

We will now take a closer look at these early interrogatives and their contexts. The first example is a simple two-word sentence, the interrogative character of which is marked by the tag huh?:

The second example has an unclear meaning and gives rise to misunderstanding by the child's mother:

In fact, the only clue that the child's utterance is meant as an interrogative is its rising intonation, as indicated by the question mark in the transcript. But as a question, this utterance is still ambiguous and open to different constructions. It could have been meant as a wh-question, with the wh-word omitted: wat hoeft er op? (wat needs there on?), or as a rhetorical yes/no question in reaction to the adult's huh?: er hoeft toch iets op? (there needs yet something on?). The conversation which follows does not give any clue to what the child might have meant, since it is about the arisen misunderstanding only. Eventually, the child succeeds in communicating what it is that he wants, in an unambiguous statement with moeten (must/have to).

The next example, like the first one, is a structurally declarative sentence accompanied by the tag huh? Interestingly, the child's mother interprets this sentence as a negative statement - falsely, as becomes clear from the child's fierce reaction:

The fourth example is a syntactically correct yes/no question with inversion of verb and subject. This question elicits an affirmative response by the child's mother:

In fact, the above four utterances quite nicely reflect the general development of questions (De Villiers and De Villiers 1985; Schaerlaekens and Gillis 1987; Frijn and De Haan 1990). Early questions in child speech do not have interrogative syntax, but are characterized by rising intonation only, sometimes accompanied by a tag like huh? It is only later that questions are expressed with wh-words and subject-verb inversion.

The fact that the above utterances with hoeven have interrogative force is interesting in view of the fact that questions in principle are a correct environment for NPIs. Exactly what it is in questions that induces their licensing ability is as yet unclear; questions as licensers are problematic for both theoretical accounts in terms of entailment patterns and negative implicature accounts (Ladusaw 1979; Van der Wouden 1994a; Linebarger 1991). In general, it is assumed that the correct appearance of NPIs in questions is contingent on speaker assumptions and expectations. It is a well known fact that NPIs more readily occur in rhetorical questions carrying declarative negative force or in questions presuming a negative answer rather than in straight informational questions (Borkin 1971). See the following examples (from Borkin 1971):

Some NPIs, however, can also occur in questions which seem to be quite neutral requests for information, for instance any:

Lakoff (1969) and Ladusaw (1979) maintain that even such seemingly neutral questions induce a negative meaning interpretation, hinging on speaker assumptions and expectations. The rather subtle negative interpretation can be illustrated by means of minimal pairs of questions with some and any. See the following pair of sentences (from Ladusaw 1979):

Ladusaw notes that these questions differ in that asking the one with some seems to suggest that the speaker is optimistic about John's finding some unicorns, while the one with any does not (Ladusaw 1979: 189).

Lakoff (1969) uses the following example to illustrate the rather subtle meaning distinction between some and any in questions:

the first question usually is an invitation to have some beans: the speaker assumes that someone will want them. The second is very frequently an expression of scorn, not spoken by the person offering the beans, but by someone to whom they are offered. (...) The first of these questions, then, assumes a `yes' answer; the speaker would probably be disconcerted at a refusal. The second assumes either a negative answer (the speaker might well be surprised if someone accepted his offer), or makes no assumption (Lakoff 1969: 610).

To return to hoeven: in adult speech, this verb does show up in questions every now and then. In much the same way as any, hoeven can occur in questions expressing varying degrees of expected negativity. The strongest negative assumption is expressed in rhetorical questions, as in the following examples (from the Hoeksema corpus):

Although these sentences have interrogative syntax, the speaker appears not to expect any answer at all. These questions are indirect speech acts and function as negative statements (cf. Geerts et al. 1984).

Hoeven not only occurs in such rhetorical questions, but also in interrogatives which in principle leave open the possibility of an answer. Still, such questions do make a negative assumption. In the same way as was noted for any, it would surprise the speaker if the answer was `yes'. The following utterances (overheard in conversations) illustrate this:

With respect to the issue whether hoeven is also allowed in questions involving no assumption at all - in the same way as was noted for any by Lakoff - judgments and intuitions are unclear and quite subtle. For a fact, it is interesting to note that no such occurrences have been attested in corpora of adult speech, but it is still easy to think of examples which do not appear to be blatantly ungrammatical. See for instance the following question:

True, this is a marked utterance, which certainly would sound better and more natural if a non-polarity verb was used instead of hoeven, for instance willen (want). Interestingly, this question would also improve if the (unstressed) adverb nog (more or less corresponding to still in English) is added:

Notably, nog (still) adds a hint of negativity to the question; by asking the question including nog (still), the speaker expresses a shred of doubt about whether the addressee really would like to eat another pancake. This suggests that hoeven, unlike any, can only felicitously appear in questions with a negative assumption, be it one which is ever so subtle (see also Van der Wouden 1996).

If we now look back at the children's utterances, it is clear what the crucial difference is with adult questions containing hoeven. Whereas adults' questions express a negative assumption - in many cases so strong as to discourage or even exclude answering - the children's questions appear to be completely unbiased. Obviously, the children's questions are uttered with the intention to elicit answers, and there appears to be no particular assumption about the polarity of the expected answer; it could equally well be affirmative as negative.

The question which now needs to be answered is whether children's assumption-free questions with hoeven really represent a violation of NPI licensing. In the light of the above discussion, which showed that licensing of NPIs in questions involves quite subtle pragmatic factors like speaker assumptions and expectations, I am inclined to conclude - once again - that the children's deviant utterances with NPIs are, in principle, correct within their developing system of NPI licensing and, at the same time, quite close to the rules which apply in adult grammar. The basic condition that NPIs need a licensing environment appears to be met, but still needs refinement until it is in complete accordance with adult rules and intuitions.

Interestingly, data from somewhat later ages (diary notes and personal communication) suggest that there is indeed such a gradual improvement towards adult-like usage. The following questions with hoeven all involve assumptions or expectations from the child's side:

Obviously, example (73) is no neutral request for information; it is reminiscent of the rhetorical questions with hoeven which occur in adult speech. Syntactically, this utterance is a question, but it appears to function as a negative statement: the child cannot imagine that someone really wants to eat rice and beans when there is the possibility of eating ice cream instead. The question in (74) is not rhetorical; the child is really posing a question. Interestingly however, there is an allusion to negativity in the expectancy of an answer, expressed by tot hoever (til how far). The implicature of this expression is no more than n, or at most n. So, in this question, the child asks what the maximum number should be, beyond which she no longer needs to count. The question in (75) also carries a clear assumption, expressed by toch wel (yet really). As was pointed out earlier, wel (really) functions to mark affirmative contrast. In combination with the modal particle toch (yet), it expresses the assumption that the addressee maybe does not need the object (a toy block) after all; hence the gloss or don't you need it? Likewise, in example (76), the child is questioning the necessity of the object, introducing a subtle notion of negativity to the utterance.

Up to Chapter 4 Table of Contents

4.5 — Pseudo-licensing by nog

Various patterns in children's early NPI use now have been reanalyzed: the supremacy of the licenser niet (not), the slow development towards more licenser variety, and the utterances which from an adult point of view lack a correct licenser. It became clear that the deviant patterns in children's early NPI use can all be traced back to a principled restricted distribution which closely approaches the restrictions on NPIs in adult discourse.

In this section, the insights from the reanalysis of the spontaneous speech data will be brought to bear on yet another deviant pattern in children's NPI use, which occurs from the age of about three: hoeven and meer in utterances without a correct licenser but with the adverb nog (more or less corresponding to English still). See the following examples:

In the light of the other deviant patterns in children's early NPI use, which could all be tracked down to a polarity-related restriction, we might assume that nog in these utterances functions as a pseudo-licenser. But on the basis of what characteristics would that be? In order to investigate the possible pseudo-licensing capacities of nog, a closer look must be taken at the meaning of this adverb. This is far from easy, as nog is a polysemous expression, with several related and ill-definable meanings. As a first step, a distinction must be made between stressed and unstressed nog. Stressed nog has a well- defined interpretation; it expresses a repetition or - in combination with a comparative - a higher degree of something. See the following examples:

Stressed nog typically occurs early in Dutch children's speech. It is used in a pivot pattern, meaning more, another one, or once again. See for instance the following example:

Apparently, this is not the nog which occurs in the utterances with hoeven given above. These utterances contain unstressed nog, a modal particle (categorized by Geerts et al. (1984) as a `judgment particle') with a range of closely related meanings. The Van Dale dictionary of Dutch (1984) lists nine vaguely distinct meanings for unstressed nog, larded with examples, as it is first from usage in a particular context that the quite subtle meanings of unstressed nog become clear (see also example (72), where nog introduces a hint of negativity to a question). No attempt will be made here to describe the meaning distinctions of unstressed nog in any detail (but, for an exposition, see Vandeweghe 1983, or Löbner 1989 about the German synonym noch). I will only highlight two recurring notions in its various meanings, as these appear to be relevant for the interpretation of unstressed nog in the children's utterances with NPIs. Firstly, nog may express a notion of quantity; in that case, it is accompanied by a noun phrase denoting an amount of something, as in (84). Secondly, nog may express a temporal notion, denoting a continuation in course of time; in that case, nog is the antonym of niet meer (not anymore), as in (85):

If we look back at the children's utterances with hoeven in (77) - (80), it is clear that nog here first of all is used with the notion of quantity. The interesting point to make is that unstressed nog with a quantitative meaning introduces a hint of negativity, as it comes close to the meanings of nog maar (only/just) or nog een beetje (just a little bit more). Unstressed nog with a quantity notion thus alludes to a negative meaning by means of the pragmatic implicatures it induces. When a closer look is taken at the children's utterances, it appears that they indeed must have been uttered with this subtle intention of negativity. The most clear example is (78). After the child's father reacted to this ungrammatical utterance with watte? (what?), the child corrected herself by uttering (86):

Clearly, the child's preceding utterance with nog, (78), meant that there was only/just one more button left to do. The other examples with hoeven + nog also give rise to such subtle implicatures: (77) implicates that the child will go to bed only after he has had something to eat; the implicature of (79) is that the child does want to have something, but only a little bit, as she still wants to have something left for later; finally, (80) gives rise to the implicature that there are only two more books to read (the context makes it clear that reading these two books will not take long), such that the addressee need not wait long for help.

Nog, by means of an allusion to negativity, thus appears to function as a pseudo-licenser for hoeven. This shows, once again, that children's utterances without a correct licenser are not just neutrally affirmative and therefore are not blatantly ungrammatical; rather, they closely touch a restriction which applies in adult grammar. The long time-span during which such utterances occur - the latest example attested, (80), occurs as late as 7;06 - indicates that nog as a pseudo-licenser is quite persistent. A grammaticality judgment experiment (to be discussed in the next chapter) indeed shows that seven-year-olds are inclined to judge sentences with hoeven + nog as correct, and that the acceptation of nog as an environment for hoeven only gradually decreases with increasing age. Another illustration of the persistence of nog as a pseudo-licenser is the following example (overheard in a conversation), a slip of the tongue of a teenager:

Up to Chapter 4 Table of Contents

4.5.1 — Is nog also a pseudo-licenser for NPI-meer?

The analysis of nog as a pseudo-licenser for hoeven gives reason to return to children's ambiguous utterances with meer, which were mentioned in Note 2 in Chapter 3. It was argued that the unclear meanings of these utterances did not provide a solid basis for deciding which of the two meers was used: quantitative-meer (more), which is not polarity sensitive, or NPI-meer (anymore), which expresses a temporal meaning (cf. section 2.8). In review, it is striking that nog is a recurrent expression in these utterances. The examples concerned are repeated below:

In the light of the analysis of nog as a pseudo-licenser, it may now seem plausible to assume that these utterances contain NPI-meer, pseudo-licensed by nog. A close look at the above and related data, however, shows that there must be more going on.

Let us assume that the meers in the above examples are indeed instances of NPI-meer. Then the only way in which nog and meer can be brought in line with each other is when nog, in the same way as meer, is taken to represent a temporal notion. But then there is a problem in assessing which relation holds between nog and NPI-meer, more particularly, whether this is a pseudo-licenser/NPI relation. As was pointed out above, nog (still) with a temporal notion is pre-eminently affirmative; it is the antonym of the negative expression niet meer (not anymore). The state of affairs with meer + nog thus appears to be completely different from that of hoeven + nog discussed in the preceding section. There, it was exactly the inherent negativity of nog with a quantitative notion which was of crucial importance for its characterization as a pseudo-licenser. The contribution of nog in (88) - (90), however, does not look like that of a licenser for NPI-meer at all. On the contrary, it appears as if nog is capable of carrying the temporal notion in these utterances alone, and can well do without NPI- meer. In fact, they would improve if NPI-meer were just left out. Two questions now arise: why do utterances like (88) - (90) occur in the children's speech, and why is it clear - to adults, that is - that the redundancy in these utterances can be reduced by omitting meer, but not by omitting nog?

Up to Chapter 4 Table of Contents

4.5.2 — PPI/NPI pairs and the Principle of Contrast

As was pointed out above, nog (still) with a temporal notion can be considered as the antonym of niet meer (not anymore). If the negation of NPI-meer is the opposite nog (still), then NPI-meer in its bare form must express the same meaning as still (nog). This explains the redundancy in (88) - (90): the notion of continuation in time is expressed twice in these utterances. Adults' intuition that meer (anymore), rather than nog (still), would have to be omitted in order to reduce this redundancy is prompted by knowledge of rules underlying the nearly complementary distribution of nog (still) and NPI-meer. The children's utterances in (88) - (90) are affirmative, and meer (anymore), due to its sensitivity to negative polarity, is not allowed in such utterances. Nog (still), on the other hand, is pre-eminently the expression to be used in affirmative sentences, as it is a positive polarity item (Van der Wouden 1988; 1994a). Nog (still) and meer (anymore) thus form a positive/negative polarity pair: in their bare form - that is, without taking into account the environments in which they tend to occur - they express the same meaning, but the respective polarity restrictions on their distribution have the effect that their distribution in actual language use barely shows any overlap. According to Zwarts' (1995) laws of negative polarity, meer (anymore) should be classified as a weak negative polarity item, as it is licensed by minimal, regular, and classical negation. Nog (still) is classified by Van der Wouden (1988; 1994a) as a weak positive polarity item, as it is compatible with minimal and regular negation, but not with classical negation. So, although there is some overlap in the distribution of nog (still) and meer (anymore) (compare a) and b) in (91) and (92)), their distribution in straightforwardly affirmative and classical negative sentences is complementary (compare the a)- and b)-sentences in (93)):

The difference between nog (still) and meer (anymore) thus lies in their distributive properties, not in their meaning per se. This points at a possible explanation for why children sometimes produce sentences like (88) - (90), in which both expressions occur. According to Clark (1987; 1993), children acquiring the lexicon of their mother tongue rely on a pragmatic assumption, the Principle of Contrast. This principle predicts that children assume that any new word which is encountered must contrast in its meaning with other words which are already known. As the PPI/NPI pair nog (still)/meer (anymore), setting aside their distinctive distributive properties, can be considered as a pair of synonyms, it might be the case that the co-occurrence of nog (still) and meer (anymore) in (88) - (90) is a manifestation of children's problems with the lack of meaning contrast between these two. Data in support of such an analysis are the following examples (from diary data, unfortunately without contextual information) in which nog (still) and meer (anymore) again co-occur, now in negative sentences:

As these are negative utterances, the roles in the source of redundancy are reversed: this time it is nog (still), rather than meer (anymore), which is superfluous.

Examples such as (88) - (90) and (94) - (95) seem to indicate that children experience some trouble in the special case of a PPI/NPI pair, in such a way that it may be difficult sometimes to decide which one of the pair to use in connection with which polarity. This is in accordance with a prediction of the Principle of Contrast. Although it is in general assumed that this principle offers economy of effort in acquisition and makes the task of acquiring meanings simpler, it appears to be a bit of an obstacle in acquiring the meanings of synonyms (Clark 1987; 1993).

The absence of a meaning contrast between the PPI nog (still) and the NPI meer (anymore) might also be underlying antonymous use of the negations of these expressions. Kaper (1985) cites the following example, in which niet meer (not anymore) is used, but where the context made it clear that the child meant to express its antonym nog niet (still not):

Kaper assumes this to be a case of `semantic fog', an instance of the more general phenomenon that children (and adults) sometimes may say the opposite of what they intend. Kaper's hypothesis is that children and adults, when trying to find the right word to express what they intend to say, sometimes and somehow (I must be very vague here) may make use of oppositions in meaning. It is as if they are scanning a semantic field in which among other things words are associated with each other because they are opposites. Assuming that speakers may proceed in that way (of course quite unconsciously), we can understand that they sometimes grasp an antonym (Kaper 1985: 63).

In the corpus of recorded speech, no examples of niet meer (not anymore) meaning nog niet (still not) were found, but instead some utterances with the opposite pattern of antonymous use: utterances in which the children use nog niet (literally still not, but as a gloss not yet) where they appear to intend niet meer (not anymore). See the examples below:

My conjecture is that the above utterances are not what they look like at first sight. Instead of constituting examples of mere accidental antonymous use, where the child for some reason has grasped an antonym out of the semantic fog, I think these are examples of children expressing exactly the meaning they had in mind, only in a way which is not in accordance with the polarity restrictions on a PPI/NPI pair.

The ambiguity in these utterances is caused by the fact that the combination nog niet in Dutch is used to express the meaning of not yet. When nog niet in the above utterances is interpreted as meaning not yet, then it indeed appears as if the children are saying the opposite of what they mean. When nog niet, however, is interpreted as the negation of nog (still), then it expresses the meaning niet meer (not anymore), and this is precisely what the children appear to intend with these utterances. The reason why these utterances to adults at first sight may look like antonymous use is that nog (still), due to its character as an NPI, is not allowed in a negative environment. Adults, adhering to the polarity rules governing the distribution of the PPI/NPI pair nog (still)/meer (anymore), would not express the negative meaning of the utterances in (97) - (99) by negating the PPI nog, but instead by means of the NPI meer in a licensing environment.

Support in favor of this interpretation comes from two sources. Firstly, it comes from the utterances and their contexts themselves. The preceding context in (97) and (98) shows that the children's utterances indeed are negative reactions to earlier affirmative utterances with nog (still). See also the following example, where this is very obvious:

Also, the utterance in (99) contains a retracement, as if the child is aware of the fact that nog is not the right word to express the intended meaning of this utterance.

The second source of support is very interesting, as it indicates a crosslinguistic parallel in the acquisition of PPI/NPI pairs: among English children, the same type of error occurs with the PPI/NPI pairs some/any and too/either. On the basis of the Brown corpus, Bellugi (1967) already reported that some in young children's utterances often occurs in the scope of negation. A search in an extended corpus of English child speech (consisting of the data compiled by Bloom, Brown, Clark, Kuczaj, MacWhinney, Sachs, Snow, and Suppes; all available in the CHILDES database, MacWhinney 1995) showed that this phenomenon is recurrent in the speech of different children and, moreover, that the PPI too displays the same pattern of usage (Van der Wal 1996). See the following examples:

Children's tendency to use some in the scope of negation has also been observed experimentally. O'Leary and Crain (1994), using a truth value judgment task with an elicitation component, found that 4 and 5 year old children had a tendency to give responses containing the PPIs some or something in a negative environment, for instance He didn't get something to eat and Well, they didn't get some food. They also noted that the opposite pattern - any or anything in affirmative sentences - tended not to occur. This is corroborated by the corpus study of English child speech (Van der Wal 1996). Until the age of about three, the PPIs some and too occur more frequently than their NPI counterparts any and either - even in negative sentences, in which the NPIs would be more felicitous. Only a few examples are found of any occurring in an environment where some would have been appropriate. In all cases, this happens in direct response to a question containing any:

Also, an example was found with a redundant PPI in a negative sentence containing an NPI, in the same way as in the Dutch examples where the PPI nog (still) and the NPI meer (anymore) occurred together in one utterance ((88), (89), (90), (94), and (95)):

Finally, an example was found from the age of three and a half, in which the PPI some and the NPI any again co-occur in one utterance, but now with both expressions restricted to the environment in which they, according to their respective polarity sensitivity, are allowed to occur:

How should the lag between some and too on the one hand, and any and either on the other be explained? According to McNeill (1971), the dominant role of some is another illustration of the principle that unmarked forms (e.g. some) are simpler than marked forms (e.g. any), and therefore occur earlier in development. The Principle of Contrast might do the rest. Just as with the Dutch PPI/NPI pair nog (still)/meer (anymore), there is no clear contrast between the meaning of some, respective too, and that of any, respective either. Although there certainly is a difference between the positive and negative part of these pairs, this primarily regards their sensitivity to a positive or negative polarity, not their meaning per se. Once some and too, as unmarked forms, have appeared in the children's speech, the absence of a clear meaning contrast with any and either may further delay the appearance of the latter.

Up to Chapter 4 Table of Contents

4.6 — Summary and conclusion

The reanalysis of the early spontaneous speech data has shown that there is a uniform pattern underneath the seemingly incompatible patterns which appear on the surface. Although hoeven and meer in the children's speech have a distribution which superficially can be characterized as being too narrow and too broad at the same time, a closer look at these data reveals an underlying continuity. Hoeven and meer are from the onset restricted in a principled way: the children use these NPIs only in environments which, to a greater or less extent, are related to a negative polarity.

The reason why this underlying restriction on children's NPI use is not immediately obvious is that it results in utterances which not always conform to adult licensing rules. Although it appears that children from early on have an understanding of the basic principle underlying the restricted distribution of NPIs, they do not master the full range of complex and sometimes quite subtle licensing possibilities all at once. This is a gradual development which is dependent on progression in another area, taking place at the same time: the acquisition of various ways to express negative meaning. It is the latter development which gives rise to various non-adult like patterns of usage in children's early utterances with NPIs. On the one hand, as the early negation vocabulary is still rather small, children may overextend the at that age well known negative expression niet (not) to utterances in which, from an adult point of view, other licensers would have been more appropriate. On the other hand, as young children sometimes express negative meaning in a way which does not conform to adult grammar (anaphoric nee (no), headshaking, intonation contour, or underarticulation), NPIs sometimes occur in so-called pseudo-licensing environments.

At a somewhat later age, when the children's negation vocabulary expands and comes to include more negative expressions, there is a gradual increase in the variety of correct licensers for NPIs. At about the same time, children make a start with NPIs in utterances which are not negative in a straightforward way but in which assumptions, expectations and pragmatic implicatures come into play: contrastive denial, questions, and the modal particle nog. It was argued in this chapter that also these environments should be regarded as pseudo-licensers, since they closely touch on possible ways of NPI licensing in adult grammar. The possibilities at stake - `affirmative licensing', licensing by questions, and licensing by negative implicature - are just the ones which are problematic for formal theories of polarity sensitivity, as they depend on ill-definable notions such as speaker assumptions and expectations. It is interesting to note that the vague status of such environments - in theory, that is -appears to present no impediment for children to recognize their NPI licensing potentials. This shows, once more, that children's early knowledge of the licensing conditions for NPIs is essentially correct: whatever it exactly is that NPIs are sensitive to, children appear to be sensitive to the same.

Context-dependent licensers are integral to NPI use, as can also be inferred from the fact that adults themselves sometimes can be caught using NPIs in utterances which only allude to negative polarity, but which strictly speaking are no correct licensing environments. See for instance the following examples (overheard in conversations):

The main conclusion which can be drawn from the investigation in this chapter is that children's deviant ways of NPI use can all be traced back to a principle which essentially is correct. In the children's speech, like in adults', NPIs are not free to appear in just any con- figuration; their distribution is restricted to specific environments only. These environments are always connected with linguistic polarity and, hence, to adult licensing rules. This insight, that children already from the onset adhere to a restriction which is basically correct, sheds an interesting new light on the `no negative evidence problem' in NPI acquisition. This issue will be taken up in the next chapter.

On to next chapter: `Adjustment of restrictions'

Up to general Table of Contents

Back to previous chapter: `Apparent Incompatibilities'