You are on page 1of 15

J. Child Lang.  (), –.

#  Cambridge University Press


DOI : .\S Printed in the United Kingdom

NOTE

The development of subject–auxiliary inversion in


English wh-questions : an alternative analysis*
ROBERT D. VAN VALIN, JR.
University at Buffalo, The State University of New York

(Received  May . Revised  May )


Rowland & Pine () present an analysis of the development of
subject–auxiliary inversion in wh-questions in the speech of Adam from
the Brown corpus. They show that there is an uninversion period in
which the child fails to invert the subject and auxiliary in wh-questions,
and they argue that this is a function of the frequency of wh-
wordjauxiliary collocations in the input : the more frequent a particular
collocation is in the input, the more likely it is to be inverted in the
child’s speech. In this note an alternative analysis is proposed : the initial
position of the tensed auxiliary signals interrogative illocutionary force,
and the auxiliaries which are most reliably inverted are those that are
overtly tensed morphologically. This analysis not only accounts for
Rowland & Pine’s data but also extends to inversion in yes–no questions.
The analysis predicts three different patterns for the development of
inversion in both types of questions, and it is shown that all three are
attested.


Wh-questions have been a significant topic in syntactic theory and the study
of language acquisition since the advent of the generative era, and English-
style wh-questions with the wh-word displaced at the beginning of the clause
have been particularly important. An important feature of English wh-
questions is the inversion of the subject and auxiliary verb in non-subject
questions. When children learning English begin to produce wh-questions,

[*] An earlier version of this paper was presented at the th Annual Meeting of the Chicago
Linguistic Society. I would like to thank Jeri Jaeger, Jean-Pierre Koenig and two
anonymous referees for comments on an earlier draft. Address for correspondence :
Robert D. Van Valin, Jr., Department of Linguistics, , Baldy Hall, University at
Buffalo, Buffalo, NY  USA. e-mail : vanvalin!acsu.buffalo.edu

 
the general pattern is the following. First, they produce auxiliary-less
questions with an initial wh-word, e.g. what you doing?, and then when
auxiliary elements start to occur, they appear in both inverted, e.g. what are
you doing?, and non-inverted forms, e.g. what he can do? Rowland & Pine
() present an analysis of the development of subject–auxiliary inversion
in one child, Adam from the Brown corpus. The data are summarized in
Figure , from Rowland & Pine ().

100
wh-questions inverted (%)
wh-questions uninverted (%)
90
wh-questions with missing
auxiliaries (%)
80

70
Total number wh–questions (%)

60

50

40

30

20

10

0
6) 2) 8) 4) 0) 6) 2) 8) 5)
1– –1 –1 –2 –3 –3 –4 –4 –5
1( 2(
7
(1
3
(1
9
(2
5
(3
1
(3
7
(4
3
(4
9
3 4 5 6 7 8 9
Data point (transcript number)
Fig. . Data from Rowland & Pine ().

The data are from  one-hour transcripts of Adam from age  ; . (MLU
.) to  ; . (MLU .). Each data point contains the data from six or
seven consecutive transcripts.
There are several interesting features of the data. When auxiliaries first
begin to occur, they are more likely to be inverted than not (data points  and

    
). Beginning with data point , however, the percentage begins to decline,
and at data point , the percentage of non-inverted questions exceeds the
percentage of inverted questions. At data point , there is a dramatic rise in
inverted questions and a sharp decline in auxiliary-less questions, and from
data point  on the percentage of inverted questions steadily increases while
the percentages of auxiliary-less and non-inverted questions declines. Row-
land & Pine characterize the interval summarized in data points – as the
‘ uninversion period’, and it is the focus of their analysis. There are two major
shifts in the data between data points  and  : () the uninversion period
ends, and () the percentage of inverted questions is greater than that of
auxiliary-less questions for the first time. By the end of data point , Adam
is for the first time producing more wh-questions with auxiliaries than
without and is inverting them the vast majority of the time. Something major
has happened in his grammar concerning wh-questions : how can this be
described and explained theoretically? In this paper the analysis proposed in
Rowland & Pine will be reviewed, and then an alternative account in terms
of     (Van Valin, , Van Valin & LaPolla,
) will be given ; it will be argued that it provides a principled explanation
for the data in Figure .

Rowland & Pine’s account in terms of input frequencies


Rowland & Pine argue against the rule-based analyses of de Villiers ()
and Valian, Lasser & Mandelbaum () and propose that during the initial
phase of wh-question acquisition, the child’s grammar consists of a limited
number of simple construction-like lexical formulas consisting of a particular
wh-word and specific auxiliaries :

Correctly inverted wh-questions will be produced when the child has


learnt a wh-wordjauxiliary marker around which to base his\her question
frame. Uninversion errors will only occur when the child has not learnt the
particular wh-wordjauxiliary marker around which to base the question
s\he wishes to ask. (Rowland & Pine  : ).

In other words, the child learns to invert an auxiliary only in combination


with a specific wh-word. No general rule of any kind has been learned. The
motivation for analysing the data in terms of wh-wordjauxiliary collocations
is that during this period the distribution of wh-words and different
auxiliaries in inverted and non-inverted wh-questions is far from random :
some combinations, e.g. what are, always appear in inverted questions, while
others, e.g. why don’t, never appear inverted. Only three combinations occur
in both inverted and non-inverted forms (how can, what ’is and why is) ;
otherwise each of the other combinations occurs either consistently inverted

 
or consistently non-inverted. Except for who, which only occurs in inverted
questions, the wh-words occur in both types of question.
The main prediction Rowland & Pine make is that the wh-wordjauxiliary
pairs that the child uses will be those for which there are sufficient examples
in the input. In other words, the child will learn and use those that are well
documented in the caregiver input. Those wh-wordjauxiliary combinations
which are of low frequency in the input will not be inverted by the child. For
example, their explanation for the strong tendency for why questions to be
non-inverted is low frequency in the input.
An analysis in terms of input frequencies appears to be problematic. As
reported in Rowland & Pine’s Table , some non-inverted combinations
occur much more frequently in the input sample than many of the inverted
combinations. What is striking is that the combination why don’t occurs 
times in the input sample, the same number as where did and more than all
of the others listed in the inverted whjaux column in Table  except for
what are and what do, and yet it occurs non-inverted  % of the time, while
all of the combinations in the inverted whjaux column occur inverted  %
of the time. In addition, the top  combinations of whjauxiliary in the
inverted column are inverted  % of the time in Table , but they have an
input frequency of  % ; that is, they are unattested in the input sample.
Similarly, the top  combinations of whjauxiliary in the uninverted
whjaux column appear non-inverted  % of the time in Table , but they
too have an input frequency of  % ; that is, they are unattested in the input
sample. While the input data are only a sample, if they are at all representative
of the overall pattern, then it is very difficult to see how the low input
frequency of the top  combinations in the inverted whjaux column
explains why they are consistently inverted, while the same low input
frequency of the top  combinations in the uninverted whjaux column
explains why they are consistently not inverted. Thus, Rowland & Pine’s
analysis can describe the pattern of inverted and non-inverted wh-questions,
but it leaves the pattern of data in Figure  and Rowland & Pine’s Table 
unexplained.

A Role and Reference Grammar account


Role and Reference Grammar [RRG] takes a rather different perspective on
these data, and an RRG analysis offers a principled account of the pattern of
inverted and non-inverted wh-questions in Figure . By way of background,
RRG is a monostratal syntactic theory that posits a direct linking between the
syntactic and semantic representations of a sentence, as summarized in
Figure .
Since each sentence is given only a single syntactic representation, there
are no syntactic rules of the traditional transformational type or the more
recent ‘ move α ’ type allowed in the theory. The syntactic representation is

    
 

Linking
Algorithm

 

Fig. . General organization of Role and Reference Grammar.

not based on X-bar principles ; simplified representations of English clause


types relevant to this discussion are given in Figure .
The core of the clause contains the predicate, typically a verb, together
with its semantic arguments, as in (a) and (d). Operators like tense and
illocutionary force [IF] are represented in a separate projection, with linear
precedence statements governing the elements in the two projections. A wh-
element, regardless of whether it is semantically an argument of the predicate
or not, appears in the pre-core slot [PrCS] in wh-questions in languages like
English, as in (b) and (c). In the RRG analysis of English-style wh-questions,
the wh-word is mapped directly from the semantic representation to the
PrCS ; note the lack of an empty NP position in the core corresponding to the
wh-word in (b) and (c).
The key to the RRG account of the data in Figure  is its analysis of IF
marking in English, which is presented in ()."

() Illocutionary force in English is signalled by the position of the tense-


bearing morpheme :
a. In declarative utterances, tense appears core-internally.
b. In interrogative utterances, tense appears core-initially.

The placement of tense in declarative utterances is illustrated in (a) in Figure


, while its placement in interrogative utterances is exemplified in (b)–(d) in
Figure . In the subject wh-question in (c), the occurrence of the subject NP
in the PrCS has as an automatic consequence the occurrence of the tense-
bearing element in core-initial position ; hence there is no inversion in subject
questions. In non-subject wh-questions and yes–no questions, on the other
hand, the tense-bearing morpheme must occur before the initial NP in the
core, in order to signal interrogative IF, as in (b) and (d). This is why the

[] While this formulation is unique to RRG, the idea of correlating the position of the tense
marker and the signalling of IF could be expressed in other syntactic theories as well.

 

(a) (b)

(c) (d)
Fig. . RRG syntactic representations (simplified).

tense morpheme in the sentence is linked to both the tense and IF operators
in the operator projection of these clauses.#
In terms of the acquisition issues we are concerned with, the child must
learn to place wh-words at the beginning of the clause in the PrCS,

[] For an RRG analysis of the English auxiliary system, see Foley & Van Valin (), §.,
also Van Valin & LaPolla (), §...

    
something for which there is plenty of evidence in the input, and must also
learn the correct placement of the tense-bearing element in order to signal
IF. A crucial precondition for getting this right is recognizing that a verbal
element carries tense. Auxiliaries like is, does, did, has and had, despite being
morphologically irregular, are clearly tensed, as they enter into clear-cut
tense oppositions, e.g. do\does vs. did, am\are\is vs. was\were, have\has vs.
had. Despite their irregularity, is, does and has all end in the present tense
-s, while did and had end in the past tense -d. But it is not so obvious to the
child that modals like can, could, shall, will, may and might are tensed. Based
on the RRG account of IF marking in English and the tense properties of the
auxiliaries, the following prediction can be made :

() In wh-questions, children will initially place only those auxiliaries


which are explicitly tensed in core-initial position.

This predicts that is, are, was, do, does, did, have, has and had should occur
in core-initial position, i.e. in inverted questions, while can, could, may,
might, shall, should and will should not appear core-initially, i.e. they should
occur non-inverted, during the initial phase of the development of inversion.
There is a further prediction regarding the negative auxiliaries can’t,
couldn’t, didn’t, don’t, doesn’t and won’t : because they end in -n’t, which does
not signal tense, the negative auxiliaries will be not be analysed as tensed ;
recall Slobin’s () Operating Principle A : ‘ Pay attention to the ends of
words ’. Therefore they should not be inverted during this period. This is
stated in ().

() Negative auxiliaries will not be inverted during the initial phase of the
development of inversion.

The inversion pattern of all of the auxiliaries in the uninversion period are
given in Rowland & Pine’s Table , and the results are summarized in Table
.

      . Summary of inversion patterns from Rowland & Pine (‚€€€),


Table „

Auxiliary types No.  % inverted\total Percent  % inverted

Tensed \ 
Modals \ 
Negative \ 
Auxiliary tokens No. inverted\total Percent inverted
Tensed \ 
Modals \ 
Negative \ 


 
These results bear out the predictions in () and (), which are based on
(l), to a striking degree. There are twelve types of whjtensed auxiliary
combinations, and eight ( %) occur inverted  % of the time. There are
seven types of whjmodal combinations, and only one ( %) appears
inverted  % of the time. Finally, there are six types of whjnegative
combinations, and none ( %) of them are inverted ; this group includes all
auxiliary combinations with a negative element, regardless of whether the
other element is a tensed auxiliary or a modal. The results are even more
striking when the actual whjauxiliary tokens are examined. There are 
whjauxiliary combinations, and the predictions derived from the RRG
principle in () are correct for  ( %) of them ; there are only twelve
exceptions. When the two unambiguous groups, tensed vs. negative, are
contrasted, the prediction is strongly confirmed :  % of the tensed auxi-
liaries occur inverted, while none of the negative auxiliaries ( %) occur
inverted. The prediction regarding modal auxiliaries is also supported, as
 % of these auxiliaries occur in non-inverted wh-questions.
The exceptions are very revealing. They are summarized in Table .

      . Summary of exceptions to (‚) in Table 

Modals : inverted Tensed : non-inverted

how can () what ’is ()


how could () why is ()
where could () why ’is ()
where shall () why did ()
why ’has ()

Let us look at the unexpected inversions first. The key factor in inversion
is being a tensed word, and obviously not all verbs in English carry -s and
-d to signal tense. One of the defining characteristics of modal verbs is that
they are not inflected the way non-modal verbs are, but nevertheless they
have a tense value. Since this is not indicated morphologically in the regular
way, it should take children some time to learn that they are in fact tensed.
Modals may be divided into two groups with respect to their tense value. The
first group consists of can and could, which can enter into a true tense
opposition when used deontically, i.e. I can do it now (l I am able) vs. I could
do it when I was younger (l I was able). They are the only modals with this
property. The second group would include will, shall, should, would, may and
might, which do not enter into the same kind of tense opposition. Given the
importance of tense marking for inversion, we may formulate the following
‘ tensedness ’ hierarchy for modals :
() can, could  will, would, shall, should, may, might

    
With respect to (), this hierarchy predicts that the first modals to appear
inverted will be can and could, with the members of the second group
appearing inverted later. The correctness of this prediction cannot be
determined from the data in Rowland & Pine, since they do not give the order
in which the individual whjauxiliary combinations first appear in Adam’s
speech, but it is supported by the fact that four of the five examples of
unexpected inversion involve can and could, with the other one involving
shall.$ It appears that Adam has begun to analyse can and could as being
tensed during the uninversion period.
The other seven exceptions are all instances in which an explicitly tensed
auxiliary is not inverted, and in six of the seven cases the wh-word is why. It
has long been noted that why questions resist inversion more than those with
other wh-words (Brown, Cazden & Bellugi, , Bellugi, , Labov &
Labov, , Bloom, ). Of the  why questions in the data,  are non-
inverted, and all three of the inverted ones involve tensed auxiliaries. While
an in-depth analysis of why questions is beyond the scope of this paper, two
points deserve mention. First, there is another interrogative expression, how
come, which is semantically similar to why and which does not trigger
inversion, as illustrated in ().
() a. How come Sally left the party early?
b. Why did Sally leave the party early?
It is possible that children are exposed to questions asking for explanation in
which inversion does not occur. Second, and probably more important,
unlike other wh-words, why is adsentential in nature, rather than being either
a request for an argument like who or what or a verbal modifier like how. Why
by itself following a statement can constitute a felicitous questions with the
meaning ‘ why [previous sentence]? ’, as illustrated in ().
() Adult : Mommy went to the store.
Child : Why? (Interpretation l why [Mommy went to the store]?)
Adult : Mommy went to the store, because she needs to get some food
for dinner.
Unlike who, what or how questions, the answers to which involve some
change in the original sentence, the answer to a why question would normally
be a because-clause which is added to the unchanged original sentence.
Hence, it is entirely plausible that children expand simple Why? utterances
by appending the sentence related to it, in particular the previous speaker’s
sentence as in (). Note that such a simple collocation would be impossible

[] It is perhaps not surprising that the other inverted modal is shall, since it, along with will,
can be used to express what is in effect a periphrastic future tense in English. They enter
into temporal oppositions, e.g. I shall\will play vs. I played, whereas the other modals in
their group do not. Hence even though they are not morphologically tensed, they do enter
into tense-like oppositions.

 
with argument wh-words and much more restricted with other adjunct wh-
words. Hence, it is reasonable to assume that it is not an accident that six of
the seven unpredicted non-inversions involve why and that it is the primary
factor in the lack of inversion with the tensed auxiliaries in Table .
Thus, of the twelve exceptions, four out of the five (if not all) of the
unexpected inversions can be explained in terms of the likelihood of a tense
interpretation for the modals in question, and six of the seven unexpected
non-inversions seem to be attributable to properties of the wh-word, why. It
appears, then, that RRG can provide a principled account of the inversions
and non-inversions in Table .
The Rowland & Pine analysis in terms of input frequencies entails the
claim that if the input frequencies had been different, then the inversion
pattern would be different. Furthermore, if different children are exposed to
different input frequencies, then the inversion pattern with each of them
should be correspondingly different. The RRG analysis makes a much
stronger prediction : regardless of input frequencies, the pattern in Table 
should show up, because it follows from the analysis of the role of tense in
signalling IF in (). The RRG analysis invokes general principles from a
syntactic theory and does not tie the major claims to specific lexical items, as
in the Rowland & Pine account. From the RRG perspective, the child has
learned a general rule, namely (l), albeit one which is very different from the
movement rules posited by de Villiers () and Valian et al. ().
Another difference between the RRG proposal and Rowland & Pine’s is
that it readily makes predictions about inversion in yes–no questions. The
principle in () applies to both wh- and yes–no questions, and the predictions
in () regarding the priority of tensed auxiliaries for inversion and ()
regarding the non-inversion of negative auxiliaries can be extended naturally
to include yes–no questions. That is, as inversion develops in yes–no
questions, the first auxiliaries to be inverted should be those that are tensed,
negative auxiliaries should not invert, and among modal verbs, the first to be
inverted should be those highest on the tensedness hierarchy for modal verbs
in (). The summary of the development of auxiliaries in questions in the
children in the Brown corpus presented in Cazden () supports these
claims, but the data are not nearly as detailed as those presented in Rowland
& Pine for Adam. Another prediction is that children will not invert negative
auxiliaries, in either type of question, until they are able to analyse them
morphologically and recognize that there is a tensed verb before the negative
morpheme. Erreich () supports this prediction ; in her study of inversion
in both yes–no and wh-questions, none of the questions with a negative
auxiliary of either type was inverted. However, it is not clear how one could
extend the Rowland & Pine account to inversion in yes–no questions, since
it is concerned specifically with whjauxiliary combinations and not with
inversion as an independent phenomenon.

    
A fundamental issue which must be addressed regarding the data in Figure
 is, why is there an uninversion period in the first place? And why, if the
child uses and inverts auxiliaries, is there a sudden dramatic increase in their
use seemingly all at once? The pattern in Figure  is not like that of
overregularization ; rather, it is the opposite. At data points  and , most of
the child’s wh-questions with an auxiliary are inverted. Note, however, that
this is a small fraction of the total number of wh-questions. At data point 
the percentage of inverted wh-questions has actually declined, and while it
increases again at data point , it is still smaller than the percentage of non-
inverted questions. Then at data point  there is a dramatic increase in the
percentage of inverted wh-questions, with a corresponding decline in the
percentage of auxiliary-less questions. By data point  and then from there
on wh-questions with inverted auxiliaries occur more often than the other
two types. The situation is rather curious : the child (Adam) begins to
produce wh-questions with inversion, and then goes through a period during
which he produces more questions without inversion than with inversion. A
comparison with the well-known cases of overregularization involving the
past tense and plural is instructive : in those cases, the child learns a rule and
overapplies it, whereas in this case the child seems to be learning some kind
of inversion rule and then applying it. What makes this pattern even
more puzzling is the fact that there was a period in Adam’s development
when he, along with the other children in the Brown corpus, was inverting
yes–no questions but not wh-questions (Cazden, ). Hence even at data
point  he was capable of producing yes–no questions with auxiliary
inversion. What explanation could there be for this behaviour? Why would
he not invert the subject and auxiliary in wh-questions at a time he was
producing inverted yes–no questions and then, after starting to invert the
subject and auxiliary in wh-questions, why would he continue to do it but
only in a limited way?
To find an answer it is necessary to go back to () and its motivation. It
concerns the overt morphosyntactic marking of IF in English. In children’s
earliest yes–no questions, IF is signalled solely by prosody, and then later
core-initial tense develops as the morphosyntactic indicator of IF. The
situation is different for wh-questions ; there are two possible morphosyntac-
tic signals of interrogative IF : inversion and the wh-word in the PrCS. It
would not be unreasonable for a child to conclude that the wh-word is the
morphosyntactic indicator of IF in wh-questions. Let us imagine three
hypothetical children and examine their possible developmental patterns
with respect to the application of the principles in () to wh- and yes–no
questions. The first one, Child A, treats (b) as applying to both types of
questions, which is a reasonable conjecture, given that core-initial tense is a
feature of both types of questions. The second one, Child B, treats (b) as
applying to yes–no questions only and treats the wh-word as the indicator of

 
IF in wh-questions. The third, Child C treats (b) as applying to wh-
questions only and treats prosody as the indicator of IF in yes–no questions.
The profiles of the three hypothetical children are summarized in Table .

      . Profiles of hypothetical learners

(b) applies to : IF indicator for Y\N Qs IF indicator for wh-Qs

Child A Both types Inversion Inversion\wh-word


Child B Yes–no questions only Inversion wh-word\Inversion
Child C wh-questions only Prosody Inversion\wh-word

Child A should show inversion to some degree in both types of questions,


and development should follow the predictions in ()–(). In wh-questions in
which inversion does not occur, the wh-word serves to signal IF. Hence for
this child, inversion is the primary indicator, with the wh-word as a back-up.
Child B, on the other hand, should show inversion only in yes–no questions
and not in wh-questions initially ; development should follow the predictions
in ()–(), as they apply to yes–no questions. Eventually, s\he would figure
out that inversion in wh-questions signals interrogative IF, but given the
child’s analysis of the wh-word as the interrogative IF indicator, inversion is
redundant. Because it is redundant, there is no strong motivation for
employing it. Indeed, because Child B has analysed inversion as redundant,
s\he may use it less, even as s\he learns more auxiliary verbs. As inversion
develops in wh-questions, it would follow the predictions in ()–(). At some
point, however, Child B will encounter evidence that an initial wh-word does
not signal interrogative IF ; the evidence could come in the form of a sentence
like ().
() What Fido did on the rug was bad.
This sentence is a statement, not a question, and therefore the initial wh-
word in this sentence cannot be interpreted as an IF indicator. Now Child B
is forced to conclude that inversion is not a secondary feature of wh-
questions, but rather that it is, as in yes–no questions, the primary
morphosyntactic signal of interrogative IF. At this point the use of inversion
in wh-questions should increase dramatically, as Child B’s grammar is now
the same as Child A’s in this respect. Child C should show inversion only in
wh-questions and not in yes–no questions initially ; development should
follow the predictions in (), () and (). Eventually, s\he would discover that
there is a morphosyntactic indicator of IF in yes–no questions, namely
inversion, and s\he would then begin to invert in yes–no questions as well.
How do the hypothetical Child A, Child B and Child C compare with real
children learning English? Child A shows the same general pattern as the

    
children discussed in Ingram & Tyack () : they showed inversion in both
yes–no and wh-questions from the start. Child B, on the other hand, restricts
inversion to yes–no questions only during the initial phase and shows the
same general pattern as the children discussed in Kuczaj & Maratsos ()
and Labov & Labov () and the children in the Brown corpus, one of
whom is of course Adam, whose wh-question production is summarized in
Figure  and Table . The developmental sequence for Child B outlined
above matches very closely the pattern of Adam’s use of auxiliaries in Figure
 : an initial period of no inversion in wh-questions in which the child takes
the wh-word as the morphosyntactic marker of interrogative IF (data point
l), followed by a period in which as more auxiliaries are used, the actual
percentage of inverted auxiliaries rises and then declines, as the child
analyses inversion as a redundant cue to IF (data points –), which is then
followed by a sudden and rapid increase in the occurrence of inversion, as the
child comes to realize that inversion does mark interrogative IF in both
yes–no and wh-questions (data points  and ), leading quickly to a stage in
which inverted wh-questions constitute the majority of wh-questions (data
points  and ).
Children like Child C do exist. Five of the subjects in the study reported
in Erreich () showed the Child C pattern : they used prosody to signal IF
in yes–no questions and had some degree of inversion in their wh-questions.%
It is impossible to tell from the data reported in the study whether the pattern
development of these children follow the predictions in ()–().
Thus, starting from (), the RRG analysis of morphosyntactic IF marking
in English, we have arrived at an explanatory account of not only the inverted
and non-inverted wh-wordjauxiliary combinations in Table  but also of
the general pattern of the acquisition of inversion in wh-questions in Figure
. Moreover, we have predicted possible patterns of the development of
inversion which are attested in the literature on the acquisition of both wh-
and yes–no questions.


In examining the pattern of development of auxiliary inversion in wh-
questions, we have seen how plausible but ultimately incorrect hypotheses on
the part of the language learner can lead to striking consequences. Because of
their assumption that (b) applies only to yes–no questions and that the wh-
word signals interrogative IF in a wh-question, children like Child B, e.g.
Adam and the other children in the Brown corpus, go through a period in
which they exhibit subject–auxiliary inversion with yes–no but not wh-
questions, and this also plausibly leads to the ‘ uninversion period ’ discussed

[] Anna, a child studied by J. Jaeger, also follows the Child C pattern (personal com-
munication).

 
by Rowland & Pine. Children who assume that (b) applies to both types of
questions from the start will exhibit inversion in both types of questions ;
they are like Child A, e.g. the children discussed in Ingram & Tyack ().
Finally, children who assume that (b) applies only to wh-questions will use
prosody to indicate IF in yes–no questions and use inversion only in wh-
questions, e.g. some of the children discussed in Erreich (). It seems
implausible that these differences could be due solely to different input
frequencies of whjauxiliary combinations in the language to which the
different groups of children were exposed, as Rowland & Pine maintain.
The study of complex phenomena like the acquisition of syntax requires a
well-defined linguistic theory to characterize the form and content of the
linguistic knowledge to be acquired. Rowland & Pine analyse Adam’s data
without reference to any linguistic theory, and attempt to account for the data
in terms of input frequencies alone. RRG has provided a lens which directs
attention to the crucial importance of the position of tense in signalling IF,
and this turns out, as we have seen, to provide the basis for an explanation
of the inversion data. One of the anonymous reviewers situated this
discussion in the context of the ‘ grammar-based ’ vs. ‘ input-based ’ debate
regarding acquisition. RRG maintains that children construct a grammar on
the basis of their general cognitive endowment (not an autonomous LAD)
plus the data in the speech to which they are exposed.& Hence from this
perspective, the contributions of both the input and grammatical theory are
crucial to understanding language acquisition, since children start from the
input to create a grammar and a grammatical theory like RRG characterizes
what the child creates, thereby pointing to the relevant features of the input,
predicting the course of development, and ultimately providing an ex-
planatory framework for the analysis of the acquisition of syntax.

REFERENCES
Bellugi, U. (). Simplification in children’s language. In R. Huxley & E. Ingram (eds),
Language acquisition : models and methods. New York : Academic Press.
Bloom, L. (). Language development from two to three. Cambridge : C.U.P.
Brown, R., Cazden, C. & Bellugi, U. (). The child’s grammar from I to III. In J. R.
Hayes (ed.), Cognition and the development of language. New York : Wiley.
Cazden, C. (). Children’s questions : their form, functions, and roles in education. Young
Children, March, –.
de Villiers, J. (). Why questions? In T. Maxwell & B. Plunkett (eds), Papers in the
acquisition of wh : Proceedings of the UMass Roundtablel, May, ‰‰€. Amherst, MA :
University of Massachusetts Occasional Papers.

[] See the Epilogue in Van Valin & LaPolla () for a summary of RRG-related work on
language acquisition, also Van Valin (). Van Valin () is concerned with a number
of issues relating to the acquisition of wh-questions.

    
Erreich, A. (). Learning how to ask : patterns of inversion in yes-no and wh-questions.
Journal of Child Language , –.
Foley, W. & Van Valin, R. (). Functional syntax and universal grammar. Cambridge :
C.U.P.
Ingram, D. & Tyack, D. (). Inversion of subject NP and Aux in children’s questions.
Journal of Psycholinguistic Research , –.
Kuczaj, S. & Maratsos, M. (). What a child can say before he will. Merrill-Palmer
Quarterly , –.
Labov, W. & Labov, T. (). Learning the syntax of questions. In R. Campbell & P. Smith
(eds), Recent advances in the psychology of language. New York : Plenum Press.
Rowland, C. & Pine, J. (). Subject–auxiliary inversion errors and wh-question ac-
quisition : ‘ what children do know? ’ Journal of Child Language , –.
Slobin, D. (). Psycholinguistics, nd edn. Glenview, IL : Scott Foresman & Co.
Valian, V., Lasser, I. & Mandelbaum, D. (). Children’s early questions. Paper presented
at the th Annual Boston University Conference on Language Development.
Van Valin, R. (). A synopsis of Role and Reference Grammar. In R. Van Valin (ed.),
Advances in Role and Reference Grammar. Amsterdam\Philadelphia : John Benjamins.
Van Valin, R. (). The acquisition of wh-questions and the mechanisms of language
acquisition. In M. Tomasello (ed.), The new psychology of language : cognitive and functional
approaches to language structure. Hillsdale, NJ : Erlbaum.
Van Valin, R. (). The acquisition of complex sentences : a case study in the role of theory
in the study of language development. In J. Boyle, J.-H. Lee and A. Okrent (eds), CLS Ġ,
Vol. ‚ : The panels.
Van Valin, R. & LaPolla, R. (). Syntax : structure, meaning & function. Cambridge :
C.U.P.



You might also like