A cross-linguistic comparison of the acquisition of why-questions by young children

Table of Contents

1. Introduction

2. Theoretical Background

3. The delayed acquisition of why-questions

4. Predictions
4.1. The semantic account
4.2. The usage-based account
4.3. The Continuity Hypothesis
4.4. The Maturation Hypothesis

5. Method
5.1. Danish
5.2. German
5.3. Classification of Danish and German

6. Results
6.1. Danish
6.2. German

7. Evaluation
7.1. The semantic account
7.2. The usage-based account
7.3. The Continuity Hypothesis
7.4. The Maturation Hypothesis

8. Discussion
8.1. The CP-modifier Hypothesis
8.2. The Extended Continuity Hypothesis
8.3. Parameter settings reducing processing load
8.4. The Initial Hypothesis of Syntax
8.5. Referential questions and the wh-criterion
8.6. The Variational Model of Language Acquisition
8.7. Maturation and Very early Setting of the V2 Parameter
8.8. Triggering inverted why-questions

9. Three possible solutions for uninverted why-questions in Early English
9.1. An approach based on the Continuity Hypothesis
8.2. An account based on the Variational Model
9.3. An approach based on the Maturation Hypothesis
9.4. Comparison and Discussion of the three approaches

10. Conclusion

11. References

1. Introduction

It is a well-documented phenomenon that English-speaking children experience greater problems with the acquisition of wh-questions than do children acquiring, for instance, Italian or V2 languages (Guasti 1996, 2000). Among other deviations from adult language like questions lacking an auxiliary (Guasti and Rizzi 1996) or negative questions with non- adult negations (Guasti, Thornton, and Wexler 1995), the optionality of Subject-Auxiliary-Inversion (SAI) is one of the peculiarities of Early English. While such an optionality has not been observed with learners of other lan- guages, English-speaking children generally apply SAI on only around 90 per cent of their wh-questions containing an auxiliary until roughly their fourth birthday. However, with the beginning of their fourth year of life, Eng- lish-speaking children, in general, master SAI. Having accomplished the acquisition of SAI, learners of English produce a nearly perfect amount of adult-like questions.

Yet it has been observed (Brown 1968, Ravem 1974, Labov and Labov 1978, de Villiers 1991, Rowland and Pine 2000, Berk 2003, Thornton 2004) that learners of English produce adult-like questions except those with why. For this wh-constituent, a high percentage of questions are left uninverted until about the child's sixth birthday. SAI thus seems to be an option for why -questions in Early English while it is obligatory with all other wh-elements.

This delay in the acquisition of SAI for why -question has re- cently been used as evidence both in favour of the theory of Universal Grammar (UG) (Crain and Pietroski 2002) as well as against it (Rowland and Pine 2000). Opponents of UG, on the one hand, argue that the differ- ences in the development of English interrogative elements demonstrate children's lack of adult syntactic knowledge and hence such rules as SAI. Questions have to be learned on a word-by-word basis and the delay of why -questions simply shows that children have less input of this type of questions compared with more frequently heard questions containing, for instance, the wh-elements what or where.

Advocates of UG, on the other hand, have been given the op- portunity to use the delay of why -questions as an argument in favour of UG by Rizzi's (1999) account of Italian questions containing perch è (why). This claim would lend strong support to the Continuity Hypothesis (Pinker 1984), which states that child grammars deviate from adult grammars only to such an extent by which adult grammars differ from one another and thus never violate a principle of UG. Studies on the acquisition of wh-questions in V2 languages (Clahsen, Kursawe, and Penke 1995, Santelmann 1999) show as a gener- al tendency the early mastering of both movement of the wh-element and the verb. From the beginning of their uttering of questions consisting of at least three constituents, learners of V2 languages hardly seem to produce any non-adult questions. However, since in none of the studies carried out so far a totally perfect mastery of wh-questions has been observed, the possibility of a delayed acquisition of why -questions in V2 languages, too, cannot be excluded. Although learners of V2 languages demonstrate nearly perfect mastery of I-to-C movement earlier than English-speaking children do, the latter still do not perform SAI in a high percentage of their why -questions at a stage when they invert virtually all of their questions with other wh-elements. This phenomenon may, in principle, also occur with learners of V2 languages.

On the other hand, the very early setting of the V2 parameter (Wexler 1998) may have the effect of preventing children from producing why -questions without applying I-to-C movement. Since children knew even before their first utterances with a length of at least three constituents that a finite verb has to occupy the second position in both a declarative and an interrogative sentence, the option of leaving why -questions uninver- ted might be ruled out. Such an interconnection between the V2 parameter and the option of uninverted why -questions could provide important in- sights into the nature of UG.

Analyses like Rizzi (1999) and Ko (2003, 2005) point to the ob- servation that why seems to occupy a different position within the Left Peri- phery of the clause than other wh-elements do. While this is not the case in adult English, in Early English, on the other hand, why indeed seems to behave differently. Since child grammar differs from adult grammar, at least in English, studies on children's acquisition of why -questions might be able to contribute to the theoretical discussion about both the structure of such questions as well as about the cross-linguistic differences between them.

In this study, I will, first, summarise the theoretical background behind the content of this paper. Then, I will summarise the results of other work which show the existence of a delayed acquisition of why -questions with English-speaking children. For the study carried out here, I will, then, formulate the predictions that different approaches toward the acquisition of wh-questions make for the data examined here. This predictions will be followed by the results of the analysis of wh-questions produced by chil- dren learning either German or Danish. The resultswill, then, be discussed in connection with several theoretical frameworks that might be able to ex- plain the observations above before, finally, three accounts of the empirical data will be stated and discussed.

The data from children of two different V2 languages allow a more profoundly based statement on the interconnection between the V2 parameter and the acquisition of wh- and especially why -questions. The concentration on only one V2 language carries in itself the possibility of be- ing led astray by certain language-specific peculiarities and ascribing them to the V2 phenomenon. But properties of the acquisition of wh-questions that can be observed both in German and in Danish should be more likely to be a result of the V2 parameter, even when some other influences might be shared by these two languages and not caused by the V2 phenomenon.

2. Theoretical Background

The formation of wh-questions is subject to two crosslinguistic para- metrical variations: First, there are languages in which the wh-element ne- cessarily has to be fronted like, for instance English, German, and Italian and, on the other hand, languages like most East-Asian languages in which the wh-elements remains in situ, that is, the wh-element is the com- plement of VP into which it is merged. Second, there languages in which the wh-element needs to be adjacent to the verb like in German, Italian, and Spanish. On the other hand, there languages in which the subject in- tervenes between the wh-element and the verb like in most Slavonic languages. A very special case in French since it allows both wh-fronting as well as in situ questions and, moreover, it also allows questions with or without adjacency between the wh-element and the verb.

In English, wh-questions are derived by two operations: wh-move- ment of the wh-element to CP and Subject-Auxiliary-Inversion (SAI). The first operations satisfies the requirement for wh-fronting while the latter op- eration meets the need of a language that requires adjacency between the wh-element and the verb. English wh-questions thus have the structural representations like in (1). Note that both operations leave a trace at the site into which the respective constituent has ben merged prior to the movement operations.

illustration not visible in this excerpt

The traditional view according to which there is one single projection named CP was changed by Rizzi (1997) who, similar to Pollock (1989), split up CP into several different constituents. These constituents are Force, Topic, Focus, and Finiteness. Within this left periphery or articulated CP, wh-elements and auxiliaries of English questions are hosted by Focus Phrase (FocP) as in (2).

illustration not visible in this excerpt

It is indicated in (2) that the auxiliaries carries [wh] features. These features are important for the so called wh-criterion (Rizzi 1996) which constitutes a principle of UG. This criterion needs to be satisfied by a Specifierhead relation between the wh-element and the verbal head which carries the wh-features. In (1), the wh-element what is the wh-element being in a Spec-head relation with the auxiliary will.

There is one peculiarity about English wh-questions. While in lan- guages like German, French, or Italian all verbs move to CP, only auxiliar- ies leave VP in English. This difference results from the setting of the verb- movement parameter which governs the movement out of VP. The negat- ive setting of this parameter in English leads to different word order, com- pared with French and German. While in (3) the adverb precedes the verb that stays in Verb, the adverb follows the verb in French and German since it has moved out of VP.

illustration not visible in this excerpt

The setting of the verb-movement parameter in English has important consequences for English wh-questions. While in French and German verbs move to CP to satisfy the wh-criterion and carrying the wh-features with them, this is only possible for English auxiliaries but not for lexical or full verbs. This difference can be seen in (4) where the English question with German word order is ungrammatical.

illustration not visible in this excerpt

The solution to this problem is demonstrated in (4c): Since movement of the lexical verb buy is not licit in English, the dummy verb do is inserted and carries the [wh] features to the head of CP in order to satisfy the wh- criterion. Without this operation named do -support, wh-questions in lan- guages with a negatively set verb-movement parameter would simply not be possible.

Rizzi (1999) adds one more projection to the articulated CP. This projection, labelled Interrogative (INT), hosts the complementiser se (if). More important for the purpose of the current paper are the wh-elements that are hosted by INT in an interrogative clause.

In Italian wh-questions, the Spec-head relationship required by the wh-criterion is satisfied by adjacency between the wh-operator and the verb carrying the wh-feature (5a). If the Subject intervenes between the wh- operator and the verb (5b), the sentence becomes ungrammatical.

illustration not visible in this excerpt

However, the adjacency requirement does not hold for the wh-ele- ment perch è and the semantically very closely related come mai (how come). These two interrogative operators do not require I-to-C movement of the verb carrying the wh-feature because, with these two wh-elements, uninverted questions are not ungrammatical as in (6).

illustration not visible in this excerpt

While in (5a), the wh-criterion is satisfied by movement of the wh-op- erator to the Spec of Focus (FOC) and of the verb carrying the wh-feature to its head, this constellation cannot hold for (6) in which these two con- stituents are not in a Spec-head-relationship. Therefore, Rizzi introduces the position INT which differs from FOC by being intrinsically endowed with the wh-feature. Furthermore, the wh-elements in (6) are not moved to the Spec position but merged there. Since, in this case, the wh-feature is not carried by the verb but by the head of INT, the wh-criterion is satisfied. This analysis accounts for the extraordinary status of Italian questions with per- ch è and come mai which do not require adjacency between the wh-ele- ment and the verb .

3. The delayed acquisition of why-questions

The first study in which the different behaviour of why -questions among all other wh-questions was pointed out is Brown (1968). In his recordings of questions uttered by the three children Adam, Eve, and Sarah, Brown mentions a large number of uninverted why -questions that were produced by Adam. Some of these sentences are given in (7).

illustration not visible in this excerpt

Early English uninverted why -questions are also attested by Ravem (1974). In his study, Ravem recorded the English wh-questions uttered by his daughter and son who are native speakers of Norwegian and learned

English as their second language. At the same time when these two chil- dren produced a robust number of wh-questions with SAI, Ravem's daugh- ter Reidun, who is two years and nine months younger than her brother Rune, uttered a large number of why -questions without SAI. Unfortunately, Ravem does not provide any statistics but, however, some examples from his recordings are given in (8).

illustration not visible in this excerpt

In their longitudinal study on their daughter Jessie carried out up to the age of 4;9, Labov and Labov (1978) recorded nearly all of their daugh- ter's wh-questions. Throughout the corpus, they found very significant dif- ferences between the rates of inverted questions for the different wh-ele- ments (Table 1).

illustration not visible in this excerpt

Table 1, Labov and Labov 1978

Moreover, the percentage of uninverted why -questions is not only significantly lower but the acquisition of full mastering of this type of wh- questions is delayed by two years compared with all other wh-words. While Jessie reached a near total rate of inverted wh-questions by the age of 4;0, the very low percentage of only 15% inverted why -questions lasted until the age of 4;6. Only after this age, Jessie produced an increasing percentage of inverted why -questions and reached 80% by the end of the study at 4;9. Some example why -questions by Jessie are given in (9).

illustration not visible in this excerpt

Since the acquisition process of why -questions is considerably different from all other wh-elements, Labov and Labov suggested that why has a different status within their daughter's grammar. Formulated within their contemporary theoretical framework of rule-based transformational grammar, they suggested for the period between 4;0 and 4;6 that Jessie performed SAI after all wh-words except why.

In an analysis proposed by de Villiers (1991), it is not only why that is produced relatively late with SAI but both adjunct wh-words why and how. Adjunct questions, de Villiers claims, initially are not derived by movement but by a topicalised wh-element adjoined to IP. Once children learn from embedded questions that why and how have to occupy the Spec of CP, they start to produce adult-like main questions, too.

To test her line of argument, de Villiers analyses the speech corpora of seven children available on the CHILDES database and determines the ages at which they start to produce embedded questions and inverted main questions. The data seem to support her claim because, for all wh-words investigated, the point at which the children started to apply SAI to main questions is the same or immediately follows the point at which they start to produce embedded questions that contain this wh-element. While this ana- lysis of wh-words as occupying the Spec of CP instead of being adjoined to IP is accomplished very early for wh-words asking for arguments, the same process is claimed to be remarkably delayed for adjunct questions. Some example sentences from de Villiers's study are given in (10)

illustration not visible in this excerpt

However, de Villiers' proposal was first challenged by Rowland and Pine (2000, cf. Table 2) who find additional support for their doubts in the data collected by Labov and Labov (1978). These studies clearly show that de Villiers' claim is not empirically supported since questions containing how do not develop in a different way than all other questions except those containing why. Instead, how -questions even had the second-highest in- version rate in Labov and Labov's study and a higher inversion rate than what -questions both in this corpus as well as in the data collected by Row- land and Pine. Therefore, these data suggest that the developmental delay of inverted why -questions does not correlate with an assumed parallel delay of how -questions, since this delay actually does not exist. It is why alone, which behaves differently, and this behaviour cannot be traced back to a special status of adjunct questions in general.

Finally, in their examination of the Adam corpus available on CHILDES, Rowland and Pine (2000) found very significant differences in the number of inverted and uninverted questions with six wh-elements (Table 2).

illustration not visible in this excerpt

Table 2, Rowland and Pine 2000

In a study on 16 children between the ages of 1;11 and 4;2, using the Triggered Natural Speech method in order to elicit wh-questions, Berk (2003) clearly demonstrates the different status of why among English questions word. Out of a total of 293 questions, their subjects produced 18 of 74 or 24,3% uninverted why -questions. Being a relatively low ratio of un- inverted why -questions compared with Labov and Labov's results, this per- centage becomes significant in comparison with the remaining questions in Berk's study. Of these interrogatives with other wh-words than why, only one single question, containing where, was used without SAI being applied. Some why -questions produced by Berk's subject are given in (11).

illustration not visible in this excerpt

Moreover, in her analysis of three different age groups, Berk shows the existence of a developmental delay for inverted why -questions. While the 16 children reliably applied SAI to questions from an early age on, only the oldest of Berk's subjects reached an inversion ratio of around 90% (Table 3). However, not only seems their percentage of inverted why -ques- tions to be higher than in Labov and Labov's study but the children particip- ating in Berk's study also seem to have acquired adult-like why-questions at an earlier age than Jessie did.

illustration not visible in this excerpt

Table 3 (Berk 2003)

Thornton (2004) recorded a corpus of over 900 why -questions in a di- ary study on her daughter A.L. Other types of wh-questions were not recor- ded systematically, but Thornton states that her daughter reliably per- formed SAI by the age of 3;5. For why -questions, a near perfect rate of in- version was not achieved by A.L. until the age of 5;0 for affirmative ques- tions and not at all until the end of the diary recordings for negative why - questions (Table 4). Examples from Thornton's study are given in (12).

illustration not visible in this excerpt

Table 4, Thornton 2004

illustration not visible in this excerpt

Thornton study reveals a very significant difference between the ac- quisition of affirmative and negative why -questions. While the development of affirmative questions resembles the data collected by Labov and Labov, the sudden decrease in the inversion rate of negative questions at the age of four is a phenomenon that calls for an explanation. Moreover, the massive gap between affirmative and negative questions at the age of five is one of the most striking findings in Thornton's study. However, the delayed acquisition of why -questions is confirmed by Thornton's data since her daughter failed to reach an inversion rate of around 90% until her fifth birthday.

4. Predictions

4.1. The semantic account

In an early attempt to address the delayed acquisition of why -ques- tions, Ravem (1974) has his proposal based on the relative complexity of the underlying semantic concept of causality. Adopting some ideas roughly sketched by Brown (1968), Ravem claims that the concept of causality is not accessible to children until a certain developmental stage. Uninverted why -questions therefore cannot be analysed in the same terms as other wh-questions because children are not aware of the semantic content of this type of questions.

Syntactically, this means that the wh-operator why is not related to a missing constituent replaced by it. Put in modern syntactic terms, why is not connected to any trace because this would have to be the trace of a subordinate clause introduced, for instance, by because. Since such subor- dinate clauses are too complex for children up to a certain age, uninverted why -questions differ from their adult counterparts insofar as they are not derived by wh-movement. Instead, why is simply preposed in front of a de- clarative sentence the word order of which is kept. Ravem supports his claim about missing knowledge of the concept of causality with a dialogue between the two observed children (13) which illustrates that Rune, the older brother of Reidun, already has this knowledge while his sister is still missing it.

illustration not visible in this excerpt

The interesting fact about Ravem's study is that the two subjects were not native learners of English but of Norwegian, only learning English as their second language. While still producing uninverted why -questions, these children cannot be claimed to not master the concept of causality be- cause, as Ravem admits, at the same time when Reidun did not respond appropriately in (13), she already used the Norwegian equivalent of why.

This circumstance suggests that a semantic account is not very likely to explain English children's delayed acquisition of why -questions. However, given the absence of cross-linguistic studies on why -questions, a semantic explanation cannot altogether be excluded. Since the concept of causality holds universally across languages, the semantic account pre- dicts that both German and Danish children should exhibit the same delay as English-speaking children do because they should experience the same problems with subordinate clauses expressing causality. Consequently, these children should also simply place the equivalent of why in front of a declarative sentence with no connection between the wh-element and its trace. This should result in a considerably delayed acquisition of adult-like why -questions very similar to English-speaking children.

4.2. The usage-based account

In contrast to a theory of language acquisition based on UG, the us- age-based account (Tomasello 2000) does not assume that children know any categories or rules but regards language acquisition as a lexically-spe- cific process highly driven by input. Within this framework, Rowland and Pine (2000) formulate their account of the acquisition of wh-question without the assumption of children's knowledge of such rules as fronting of the wh-element or SAI and without knowledge of such categories like wh- words. Instead, the formulation of adult-like questions has to be learned in- dividually for each single wh-word until, eventually, a certain level of ab- straction will be reached. Consequently, the striking differences between the inversion rates in Adam's questions (Table 2) need to be accounted for by their high or low input, respectively. While a child easily learns to pro- duce adult-like questions with wh-elements more frequently heard, this ac- quisition process will take significantly longer for wh-words with less input. The high degree of non-uninverted why -questions in Adam's speech there- fore is accounted for by Rowland and Pine by the low input of this specific wh-word.

If the usage-based account describes the acquisition of wh-questions correctly, then the rates of the application of I-to-C movement in the Danish and German corpora should differ as significantly between the different wh- words as they do in the English corpus examined by Rowland and Pine. The specific properties of each individual wh-element would need to be learned separately and without any knowledge of categories or rules. Any differences assumed by UG-based accounts like the V2 parameter should not influence the acquisition process in different languages. Therefore, there should not exist any differences between the acquisition processes of wh-questions in any languages.

Moreover, the usage-based account does not regard why to possess a status that is different from other wh-words. The extremely high degree of uninverted why -questions is ascribed to the low input Adam is assumed to have received for this specific interrogative element. Being raised within a certain linguistic environment, such a high rate of non-adult question are possible for every other wh-word as well. Therefore, the usage-based ac- count does not predict that German and Danish children would demon- strate the same delayed mastering of why -questions observed with Eng- lish-speaking children. Rather, the usage-based account predicts that questions with any giveb wh-elements the input of which is rather low for the child, will more often and longer show a non-adult structure. Further- more, these questions will not be deviant within the limits set by UG but will simply be deficient and less articulated compared with adult speech.

4.3. The Continuity Hypothesis

Unless proven wrong, the Continuity Hypothesis (Pinker 1984, Crain 2002, Crain and Pietroski 2001, 2002) functions as a very strong support for the theory of UG. The process of language acquisition is regarded as a transitional change between several states before, eventually, the targeted adult grammar is fixed while all principles of UG are available to the child from the beginning of language development. Hence none of these inter- mediate grammars violates principles of UG. Children try out several gram- mars which, as the most important claim made by the Continuity Hypothes- is states, allow constructions that exist in human languages never heard by the respective child. Deviations from the target grammar and hence from children's linguistic input are explained as instantiations of innate possibilit- ies made available by UG and accessible to children without positive evid- ence that will finally be fixed to the settings of their target grammar. Func- tioning as an argument against non-nativist theories of language acquisi- tion, advocates of the Continuity Hypothesis claim that such theories can- not explain why children only deviate from the target grammar in such ways that are allowed in existing target grammars and never violating principles of UG.

On this assumption, learners of English are claimed to entertain the optional inversion of Italian perch è -questions for why -questions even after they already master adult-like questions with all other wh-elements. This claim is supported by Thornton (2004) who examines the speech of one English-speaking child with regard to five distributional properties of perch è that should be observable in the child's speech if the claim made by the Continuity Hypothesis turns out to be correct. For instance, in the ambigu- ous sentence (14), perch è should behave like other wh-elements if the long-distance reading is intended because, in contrast the local reading asking for the reason of his saying instead of his resigning, perch è has clearly moved from the embedded clause.

illustration not visible in this excerpt

Along these lines, Thornton found 21 why -questions with tensed em- bedded clauses. Of these 21 questions, four were not inverted but need not necessarily include long-distance movement but could as well be inten- ded to have a local reading and hence allow for optional inversion. Thornton favours the latter analysis and (15a) is the one of the four ques- tions that most clearly suggests a local reading while (15b) illustrates a case of ambiguity.

illustration not visible in this excerpt

Uninverted why -questions are by no means the only deviation from adult grammar that is claimed to constitute an innate syntactic option mani- festing itself in Early English. Thornton (2002) analyses Genitive subjects (16a) as actually being Nominative subjects with a cliticized form of be (16b) resulting from a phonological constraint on syllables that is overcome by the age of 2;0. Such sentences, Thornton claims, are intended to ex- press contrastive focus as in (16c) and are produced by children roughly until the age of 2;6.

illustration not visible in this excerpt

While in English focus is expressed by stress and movement to FocP only takes place at LF, other languages like, for instance, Hungarian, re- quire overt movement of the focussed element before Spell-Out. In Early English, however, children also apply overt movement because they have, in the absence of experience, the parameter governing overt or LF move- ment to FocP set to a different value than in adult English. But while in Hungarian and other languages the verb moves to the head of Foc in order to have its feature checked, this is not possible in Early English because the parameter governing verb movement has been set to the value of adult English at a very early age (cf. Wexler 1998) and hence does not allow overt movement of lexical verbs. As a result of this restriction, be is inser- ted into IP and moved to the head of Foc to satisfy the feature checking re- quirement and finally cliticizes onto the subject that is moved to the Spec of Foc. As it is claimed for why -questions, children seem to deviate from their target grammar in a way that is permitted by other natural languages and thus seem to behave as predicted by the Continuity Hypothesis, which claims that child grammars differ from adult grammars only in the same ways than adult grammars differ from each other.

Another phenomenon that has been claimed to lend support to the Continuity Hypothesis (Crain and Pietroski 2001, 2002) are so-called medi- al wh-questions (Thornton 1990, Thornton and Crain (1994). In these non- adult long-distance questions, an overt wh-element is left behind in the em- bedded clause instead of a trace as in (17).

illustration not visible in this excerpt

Again, children produce a type of construction that is perfectly grammatical in other languages since some dialects of German and French allow medial wh-questions as in (18).

illustration not visible in this excerpt

'Who do you think who goes home)'

However, in these German dialects, no complex wh-phrases are allow in the intermediate CP as in the ungrammatical (19). In order to test if Early English medial wh-phrases resemble their adult German counterparts, Thornton and Crain (1994) set up an experiment in which they tried to elicit sentences like (20) from their subject. However, no such sentences were produced by the children and hence Thornton and Crain argue that learners of English for a certain period of time optionally use a construction of long-distance wh-questions that is grammatical in adult German but not in English-speaking children's target language.

[...]