19
THE ACQUISITION OF THE DATIVE ALTERNATION BY SECOND LANGUAGE LEARNERS AND LINGUISTIC THEORY Irene Mazurkewich Concordia University The research discussed in this paper attempts to demonstrate that evidence based on the acquisition of dative structures in English by second language learners provides support for a theory of markedness. Within this theory, it is claimed that unmarked structures, which are postulated to be part of core grammar, will be acquired before marked structures, which are considered to be peripheral rules of the core grammar. Using a test that elicited intuitive judgements, we obtained data from native French- speaking and Inuktitut-speaking (Eskimo) students as well as native English-speaking students whose judgements were used as the norm. Dative structures in which the dative noun phrase appears either in a prepositional phrase or as the first noun phrase of a double object construction were tested. Within the framework of markedness we assume, it isargued that the former represents the unmarked structure whereas the latter is the marked one. The results we obtained show that the unmarked structures were acquired first as predicted and the marked structures followed. The theoretical implications of markedness regarding learnability of the complement structures of alternating and nonalternating dative verbs are discussed. In this paper we will focus on the acquisition of to- and for-dative structures in English by second language learners.' We will look at the acquisition of the class of verbs which allows the dative alternation, as in: (1) a. William sent a memo to Isabel. (2) a. William sent Isabel a memo. b. William bought a present for Isabel. b. William bought Isabel a present. We will also look at the acquisition of dative verbs that do not allow the alternation, as in: (3) a. Sam described the film to Joan. b. Sam collated the article for Joan. '1 wish to thank Alison d'Anglejan and Rajendra Singh for their help and advice throughout the course of this research. I would also like to thank John Goldsmith for making his paper available to me as well as two anonymous reviewers for their comments. I am indebted to Anne Barkman who did the statistical analyses and to Jack Upshur for valuable discussion of them. The research discussed in this paper is part of my doctoral dissertation which was supported in part by a grant from the Ministere du QuCbec (FCAC) to A. d'Anglejan et al. for which I am grateful. An earlier version of this paper was presented at the Sixth Annual Boston University Conference on Language Development, October 198 I. 91

THE ACQUISITION OF THE DATIVE ALTERNATION BY SECOND LANGUAGE LEARNERS AND LINGUISTIC THEORY

Embed Size (px)

Citation preview

THE ACQUISITION OF THE DATIVE ALTERNATION BY SECOND LANGUAGE LEARNERS AND

LINGUISTIC THEORY

Irene Mazurkewich Concordia University

The research discussed in this paper attempts to demonstrate that evidence based on the acquisition of dative structures in English by second language learners provides support for a theory of markedness. Within this theory, it is claimed that unmarked structures, which are postulated to be part of core grammar, will be acquired before marked structures, which are considered to be peripheral rules of the core grammar. Using a test that elicited intuitive judgements, we obtained data from native French- speaking and Inuktitut-speaking (Eskimo) students as well as native English-speaking students whose judgements were used as the norm. Dative structures in which the dative noun phrase appears either in a prepositional phrase or as the first noun phrase of a double object construction were tested. Within the framework of markedness we assume, it isargued that the former represents the unmarked structure whereas the latter is the marked one. The results we obtained show that the unmarked structures were acquired first as predicted and the marked structures followed. The theoretical implications of markedness regarding learnability of the complement structures of alternating and nonalternating dative verbs are discussed.

In this paper we will focus on the acquisition of to- and for-dative structures in English by second language learners.' We will look at the acquisition of the class of verbs which allows the dative alternation, as in:

( 1 ) a. William sent a memo to Isabel.

(2) a. William sent Isabel a memo. b. William bought a present for Isabel.

b. William bought Isabel a present. We will also look at the acquisition of dative verbs that do not allow the alternation, as in:

(3) a. Sam described the film to Joan. b. Sam collated the article for Joan.

'1 wish to thank Alison d'Anglejan and Rajendra Singh for their help and advice throughout the course of this research. I would also like to thank John Goldsmith for making his paper available to me as well as two anonymous reviewers for their comments. I am indebted to Anne Barkman who did the statistical analyses and to Jack Upshur for valuable discussion of them. The research discussed in this paper is part of my doctoral dissertation which was supported in part by a grant from the Ministere du QuCbec (FCAC) to A. d'Anglejan et al. for which I am grateful. An earlier version of this paper was presented at the Sixth Annual Boston University Conference on Language Development, October 198 I .

91

92 Language Learning Vol. 34, No. I

(4) a. *Sam described Joan the film. b. *Sam collated Joan the article.

It was originally proposed by linguists working within the classical transformational model that the dative alternation was governed by the Dative Movement Rule (cf. Fillmore 1965; Jackendoff and Culicover 197 1; Emonds 1972, 1976). However, Oehrle (1976) has argued convincingly that the dative alternation is best accounted for by a lexical rather than a transformational rule. Within this framework, dative verbs that optionally allow the alternation, such as sendand buy in ( 1 ) and (2), would be assigned the subcategorial features [NP PP] and [NP NP] in the lexicon. Oehrle suggests that these features could be related by a lexical redundancy rule, perhaps along the line of Jackendoff (1975). Nonalternating verbs, such as describe and collate in (3), would only be assigned the feature [NP PP]. The lexicalist analysis does not a priori rule out overgeneralizations of the kind shown in (4)a and (4)b, but it might explain why they are not very common in the speech of native speakers at the acquisitional stage when they begin to learn verbs of this class. In other words, it may be that constraints applying to lexical rules would be more effective than those holding for transformational rules. Once a rule of the latter type had been acquired, e.g., the Dative Movement Rule in the classical model, it would apply across the board every time the structural description was met so that one would expect to find a proliferation of overgeneralizations of the kind shown in (4)a and (4)b, which is not the case. Overgeneralizations are found but they are not found in abundance. A further problem with the transformational approach is that it must resort to formal devices in the form of negative features to rule out these kinds of overgeneralizations and this is generally considered to be an unsatisfactory solution.

In this paper, we will show that data gathered from second language learners of English support the lexicalist approach. Furthermore, we will argue that the acquisitional sequence of the dative complement structures follows from a theory of markedness which provides the explanatory criterion for these acquisitional facts.

Markedness and the Dative Alternation

We assume the notion of “core grammar” outlined by Chomsky (198 1 a, 1981 b, 1982), the principles of which are provided by Universal Grammar (UG). UG is postulated to be the common biological endowment that children bring to the language learning task. Although the core grammar of

Mazurkewich 93

a language is determined by a restrictive set of principles of UG, the parameters of individual languages are open in that they are set by experience depending on the particular language that a child is exposed to. Unmarked rules are rules of the core grammar and, as such, it is predicted not only that they are easy to learn but also that they can be acquired on the basis of minimal exposure to primary data. Marked rules are the peripheral rules of the core grammar and they are predicted to be relatively more difficult to learn. Furthermore, they must be learned on the basis of positive evidence of their existence in a particular language since they could not be assumed to exist in that language.

We propose that the lexical feature [NP PP] represents the unmarked feature for dative verbs and that the feature [NP NP] is the marked one. This assumption is based primarily on a criterion of productivity since the vast majority of dative verbs in English take [NP PP] complements while only some are subcategorized to take the [NP NP] feature. The latter group is, in effect, a subgroup of the former. Based on this theory of markedness, the prediction is that the prepositional phrase complement will be acquired before the double-NP complement and this prediction is supported by the developmental studies that have been reported (cf. Fischer 1971, 1976; Stayton 1972; Roeper et al. 1981). There are, however, a few exceptions to this acquisitional sequence.

It was noted by Brown (1973) that one of the children in their study produced the utterance “give doggie paper” before utterances containing prepositional phrase complements were produced (p. 205). Also, it was reported in a study carried out by Potts et al. (1979) that young children demonstrated a preference for double-NP complements in utterances they produced which contained the verb give. We suggest, however, that examples involving give are the exception. It may be that such utterances containing this verb, which occurs with a very high frequency, are part of unanalyzed routines in the speech of young children. The use of holistic or gestalt strategies by young children has been demonstrated by Peters (1977), who studied native English speakers, and Hakuta (1974), who studied second language learners.

A second premise for proposing that the [NP PP] feature is the unmarked one is based on the claim that morphological and semantic factors govern the alternation, but these restrictions apply only to the occurrence of the [NP NP] feature. Insofar as the morphological constraint is concerned, it has often been noted that dative verbs which alternate are mainly monosyllabic and of native origin, while nonalternating verbs are

94 Language. Learning Vol. 34, No. I

polysyllabic and of nonnative or Latinate origin. There are exceptions to both classes, for example:

(5) a. They named the child Ben.

(6) a. He called Jerry a liar. b. *They named Ben to the child.

b. *He called a liar to Jerry. Verbs of this class are rare. The fact that such verbs take only double-NP complements would presumably have to be learned as an idiosyncratic fact of English. There are, as well, some polysyllabic verbs of nonnative origin which permit the alternation, for example, promise, offer, allow, etc. Dysyllabic native verbs have stress that falls on the first syllable and Green (1974) suggests that it is possible for verbs, historically of one class, to change classes. This means that a verb likepromise, which has stress on the first syllable, could be treated as a native verb, thereby permitting it to alternate.

With respect to the semantic constraint which is also needed to restrict the dative alternation, Goldsmith ( 1 980) proposes an integrated theory which is applicable to both to- and for-datives. There are, as he points out, some contexts in which the alternation is constrained with verbs that would otherwise permit the alternation, for example:

(7) a. I owe this example to Joe Smith. b. I owe five bucks to Joe Smith. c. *I owe Joe Smith this example. d. I owe Joe Smith five bucks.

Furthermore, there are native verbs like give which alternate in certain contexts, but permit only the double-NP complement in contexts that involve inalienable property, as in:’

(8) a. Pat gave Mike a kick. b. *Pat gave a kick to Mike.

20ehrle (1976) also discusses cases in which the alternation is restricted with certain verbs depending or. the context. He argues that the interpretation of an utterance involving “causality” only obtains with double-NP complements. For example, he claims that the following sentence is ambiguous:

One interpretation is that the ownership of the book passed from Nixon to Mailer. O n another reading, the interpretation is that Nixon merely handed the book to Mailer and questions of ownership are irrelevant. There is a third reading which holds for the situation which allows for the interpretation that Mailer wrote a book he would not have been able to write if it had not been for Nixon. The last reading can only be obtained from the sentence containing the double-NP complement. It is not available in the sentences containing the prepositional phrase complement, although the alternation is possible with the first two readings.

( i ) Nixon gave Mailer a book.

Muzurkewich 95

(9) a. Pat gave Mike a black eye.

In order to account for the contrasts illustrated in (7), (8), and (9), Goldsmith proposes that only verbs which present the indirect object as the “prospective possessor” of the direct object will be compatible with the double object construction. He admits that the notion of prospective possessor must be extended in order to handle possession of abstract objects associated with verbs such as sing, read, etc., but this is a problem that any theory attempting to explain these facts must somehow deal with.

Goldsmith’s analysis also accounts for a number of other cases involving monosyllabic verbs which do not allow the alternation, as the following sentences illustrate:

b. *Pat gave a black eye to Mike.

(10) a. Tom proved the theory to Max.

( 1 1) a. Tom stirred the sauce for Max.

(12) a. Tom washed the dishes for Max.

(13) a. Tom drove the car for Max.

b. *Tom proved Max the theory.

b. *Tom stirred Max the sauce.

b. *Tom washed Max the dishes.

b. *Tom drove Max the car. The above contrasts show that although Max may have benefitted in some way from these activities, he cannot be considered the prospective possessor of the direct object.

Oehrle (1976), on the other hand, distinguishes between to- and for- datives in his formulation of the semantic constraint governing the dative alternation. He argues that if a to-dative verb occurs in the double object construction and has the property of transference, then it occurs in the prepositional phrase as well. For-datives are problematic because there does not seem to be a uniform semantic characterization of its domain. Oehrle adopts the proposal made by Williams (1974) thatfor is essentially a marker of thematic goal which falls within the scope of an intention marker. The claim made is that an intention can exist on the part of the subject which involves a relation (of a restricted sort) between the indirect and direct objects. Oehrle postulates that the intention holds at a specified time referent for the prepositional phrase structure and it holds not only at that time but subsequently as well for the double object construction. His hypothesis is less compelling than Goldsmith’s because it is not an unified one; the constraints operating on to-datives are different from those operating onfor-datives. Also, from the point of view of learnability, it is

96 Language Learning Vol. 34, No. 1

not at all clear how such complex constraints, such as those pertaining to for-datives, as outlined by Oerhle, might be acquired.

Working within a theory of markedness, we consider Goldsmith's semantic analysis to be preferable, not only because it is an integrated hypothesis but also because it puts restrictions on the marked rather than the unmarked structures.

The Acquisition of the Constraints Governing the Dative Alternation

It appears that both morphological and semantic constraints are required to account for the dative alternation in English. The question that this raises is how a learner, either first or second, could acquire constraints as subtle as these seem to be. That is to say, how would learners come to know that dative verbs such as describe and collate are exceptional in that they do not allow the double-NP complement? It is not likely that sentences such as (4)a and (4)b would be included in the linguistic input that native English speaking children are exposed to, which means that no occasion would arise for correction to indicate to them that such structures are ungrammatical. In any case, it is well known that any corrections that children do receive mainly concern the truth value of the utterance (Brown and Hanlon 1970), and children tend to ignore any corrections that are made about their syntax (McNeill 1966). The lack of evidence in the available literature demonstrating that children do produce overgeneral- izations of this kind has led Baker (1979) to postulate that children learn language solely on the basis of positive evidence. Baker considers that the dative alternation is lexically governed, but hedoes not state what form the relationship between the two complements might take. His claim implies that since certain dative verbs appear only with prepositional phrase complements while others appear with double-NP complements, their occurrence in the linguistic environment of a learner is a sufficient criterion for learnability.

Baker's proposal, that language is learned on the basis of positive evidence, is plausible in light of the apparent absence of overgeneralization of the dative alternation; however, we would like to suggest that it is not tenable.3 He is correct in noting the absence of overgeneralization in the

'It is suggested by Chomsky (1981 b) that negative evidence 0fan"indirect"nature might be needed in the language learning process. He defines negative evidence as the nonoccurrence of

Mazurkewich 97

literature but the reason for this absence is probably due to the class of verbs involved. To the best of our knowledge, acquisitional studies have concentrated only on the class of verbs that optionally permits the alternation and this is quite understandable since the studies concentrate on young children and these are the kinds of verbs that are learned first. For example, young children would learn an alternating verb like give before they would learn the closely related nonalternating verb donate. Similarly, buy would be learned before purchase, find before discover, teN before report, show before demonstrate, etc. Thus, the absence of overgeneral- izations may simply be due to the absence of nonalternating verbs in the speech of young children. This would mean, however, that once they do start to learn nonalternating dative verbs, we would expect to find examples of overgeneralizations. We did, in fact, find some in thedata that we elicited from our native English-speaking control subjects and they will be discussed when we present the results of our testing.

It is not known when children fully master the dative alternation. Fischer (1971, 1976) shows that children start to acquire the alternation when they are 4 or 5 years old, but Roeper et al. (1981) report that some of the children in their study had not mastered the alternation by the time they were 9 or 10 years old. The morphological constraint is needed to restrict the alternation of nonnative dative verbs, as shown in (4)a or (4)b, but it is not clear whether children are aware of this constraint in the initial stages of their acquisition. However, in order to rule out ungrammatical utterances involving monosyllabic verbs that do not permit the alternation, as in (7)c and (8)b to (13)b, children would have to acquire the semantic constraint which appears to involve the notion of possessor. The results of some further research that was carried out (Mazurkewich, unpublished) indicate that the semantic constraint is acquired before the morphological one, but it is not evident which one is accessed in cases of ~ v e r l a p . ~

a structure or rule. This proposal also raises serious problems for learnability. We discuss in detail the issues raised by Chomsky and Baker (1979) in our paper (Mazurkewich and White 1983) and propose how native English speakers might acquire the constraints governing the dative alternation. We argue that overgeneralization results from an overapplication of the lexical redundancy rule that relates the two subcategorical features of alternating verbs and suggest how they might become “unlearned.”

4This testing was carried out on 10 native English speaking children with a mean age of 9.0 years. The stimulus sentences included both grammatical and ungrammatical forms of nonalternating verbs, as in (3) and (4). as wellas thegrammatical and ungrammatical forms of the type of nonalternating verbs shown in (10) to (13). Data on their intuitional judgements of these sentences were elicited. The results demonstrate that the subjects more accuratelyjudge

98 Language Learning VOI. 34, No. I

To restate the issue, we propose that the dative alternation is best accounted for by a lexical rule which base generates two subcategorial features, [NP PP] and [NP NP], for alternating verbs, and one feature, [NP PP], for nonalternating verbs. The evidence indicates that children acquire alternating verbs before nonalternating ones, and a complete mastery of the alternation is a late learned accomplishment. Precisely how the semantic and morphological constraints interact in the acquisitional process remains an open question. Anecdotal evidence indicates that second language learners, especially adults, seem to have a great deal of difficulty mastering these constraints and the complexity underlying that interaction may well be the source of the problem.

Although the debate of whether acquisition of a first language proceeds in the same way that second language learning does is far from being settled, we assume, as far as the acquisition of the dative alternation is concerned, that the acquisition process is similar in the two groups. As any review of second language materials will readily demonstrate, the alternation is seldom taught. Some recent materials do include a few drills on the alternation but none of the ones we have seen deal with nonalternating verbs. This is probably because there has not been a rule available until recently that could be used for pedagogical purposes. Therefore, if the dative alternation were taught at all, it would have to be taught as an idiosyncratic fact of the language, which is how lexical rules are learned in any case. The second language acquisition of these structures would not appear to be different. Input that second language learners are exposed to, however, does differ from that received by first language learners in that the former group may get negative evidence in the form of corrections from their teachers and peers. However, it has often been pointed out that second language learners are distressingly immune to

ungrammatical sentences of the category shown in (I0)b to (13)b in comparison to the ungrammatical sentences shown in (4). This evidence suggests that the children acquire the semantic constraint involving possession before the morphological constraint. This is consistent with Brown’s claim (1973) that the semantic notion of possession is already acquired by the time young children reach the two-word stage. The problem of sorting out whether the morphological or the semantic constraint is operating on the alternation in cases where there is overlap, as in (10)b to (13)b, is complicated by the fact that simpler native verbs such as drive, which do not involve the semantic constraint, are probably learned before more complex verbs of Latinate origin such as describe, which do involve the morphological constraint. The results drawn from a cross-sectional sampling of this kind must remain tentative, as more research, especially longitudinal studies, is clearly needed.

Mazurkewich 99

correction. Based on the theory of markedness we assume, the prediction for second language learners is that they would acquire the dative prepositional complement before the double-NP complement.

Elicitation Procedure Used

We carried out a n Intuitive Judgement Test in order to demonstrate the acquisitional sequence of dative structures. Our testing is based on the assumption that one aspect of linguistic competence includes the ability of native speakers of a language to make introspective judgements about the grammaticality of sentences in that language (cf. Chomsky 1965, 1980; Peters 1972), and we assume that second language learners would acquire similar intuitions that would be reflected in their developing competence as learners approach native ability to speak that language. There is a good deal of controversy in the literature on the interpretation of intuitional judgements (cf. the papers in Greenbaum 1977). It has been pointed out, for example, that judgements of grammaticality may be influenced by the context in which sentences appear, the order in which they appear, whether or not test sentences are interpretable, etc. Nevertheless, we consider that grammaticality judgements do reflect linguistic competence. For example, it might be argued by those working within a discourse grammar that (14)d, in the following paradigm, is acceptable because it is interpretable or for some other metalinguistic factor:

(14) a. Jerry handed a book to Anne. b. Jerry handed Anne a book. c. Jerry recommended a book to Anne. d. *Jerry recommended Anne a book.

However, it would still not render the sentence grammatical and it is this distinction that we are insisting on here. In this study, we have used naive, native speakers of English as a control group and it is their collective judgements which determine the norm.

The Subjects Tested

The second language groups we tested were made up of native French and native Inuktitut (Eskimo) speaking students. We chose subjects from two unrelated language families in order to be able to isolate what might be

100 Language Learning Vol. 34, No. I

instances of transference.' The French group was made up of45 students at the high school and collegial levels whose average age was 18.0 years. The Inuit group was made up of 38 students at the high school level whose average age was 17.0 years. We used two control groups of native English speakers. One group was made up of 6 grade 7 students whose mean age was 12;3 years, and 6 grade 10 students whose mean age was 15;6 years. We had originally started with the younger control group subjects because we expected that the acquisition of the syntax would have been accomplished by this age, but on reconsideration of the kinds of testing we were doing, we decided to use a second group who were closer in age and grade level to the second language subjects. We did find a variation in the results obtained from the two control groups and we decided, therefore, to report the results obtained separately.

We used the results of a cloze test as an index to determine the level of achievement in English for our subjects since it is well known that one cannot rely on chronological age or years of schooling to make this assessment. This was especially true of the two experimental groups in this study. The French students were taught English as a second language while their main education was carried out in French. The Inuit, on the other hand, were educated in English, although they received instruction in Inuktitut throughout their primary and secondary school years. Based on previous studies (Oller 1972, 1973; Irvine, Atai, and Oller 1974; Stubbs and Tucker 1974; Porter f978), we adopted the scoring procedure of counting the number of exact words restored to the context as opposed to scores based on contextually acceptable responses. Using the results obtained from this test we divided the subjects into three levels-Group 1 (Beginners), Group 2 (Intermediate), and Group 3 (Advanced). Of the 45 French speakers, 23 were classified as Beginners, 7 as Intermediate, and 15 as Advanced. Of the 38 Inuktitut speakers, 12 were classified as Beginners, 8 as Intermediate, and 18 as Advanced. The subjects in the two control

'Inuktitut is a polysynthetic language in which the functions of nouns are indicated by case marking suffixes, in which there are no prepositions. and in which word order is relatively more free than it is in English. It also has a n alternation involving dative nouns, but the constraints operating on it are quite different to those found in English. In lnuktitut the alternation appears only with ergative constructions so that either the direct object loses its morphological case marker and ends with a zero morpheme, or the indirect object loses its case ending and takes on the zero morpheme. In both instances, the subject carries the ergative case suffix. The constraint governing the loss of case endings in these constructions has not yet been worked out, but it does not appear to involve the kind of semantic constraint operatingin English.

Mazurkewich 101

groups, the first being the younger subjects and the second the older ones, were classified as Advanced.

Test Sentences and Their Classification

We obtained intuitive judgements of simple declarative sentences containing a set of to-dative verbs that optionally take the dative alternation-give, lend, read, send, and throw-as well as three to-dative verbs which obligatorily take prepositional phrase complements-explain, report, and suggest. A parallel set of stimulus sentences containingfor- dative verbs that optionally take the dative alternation-bake, bu.y, choose, make, and save-and three for-dative verbs which obligatorily take prepositional phrase complements-rupture, create, and design-- were also included. We added a set of distractor stimulus sentences that did not contain dative verbs. The stimulus sentences used in this testing were classified in the following way:

Type I This set of sentences contains dative verbs that optionally permit the alternation and the dative noun phrase (NP) appears in the prepositional phrase: (15) a. Peter threw a football to Philip.

b. Diane baked a cake for Nicole. Type 2 This set of sentences contains dative verbs that optionally permit the alternation and the dative N P appears as the first NP of a double-NP construction: (16) a. Peter threw Philip a football.

b. Diane baked Nicole a cake. Type 3 This set of sentences contains dative verbs that obligatorily take prepositional phrase complements and the dative NP appears in the prepositional phrase: (17) a. David suggested the trip to Ruth.

b. Anne created a costume for Sarah. Type 4 This set of sentences contains dative verbs that obligatorily take prepositional phrase complements but the dative NP appears at the first N P of a double-NP construction: (18) a. David suggested Ruth the trip.

b. Anne created Sarah a costume.

I02 Language Learning VOL 34, No. I

Type 5 This set of sentences is made up of distractor sentences which do not contain dative structures: (19) a. Dennis annoyed Karen yesterday.

b. Patrick rescued Lisa from drowning. The alternating dative verbs used were monosyllabic native verbs and the

nonalternating verbs were polysyllabic verbs of nonnative origin. Both the direct and indirect objects in all the test senteixes were full noun phrases. We did not use pronouns because many people d o find acceptable sentences that contain double-NP complements in which the indirect object is pronominalized, but unacceptable where it is a full noun phrase, for example:

(20) a.

(21) a.

Karen repeated Robert the Answer.

Karen selected Robert a book. b. ?Karen repeated him the answer.

b. ?Karen selected him a book. We assume that cliticization is responsible for the difference in acceptability in these cases.

The test sentences were randomized along with a series of dative questions and dative passives, the results of which are reported in Mazurkewich (1982, 1983a). The subjects were asked to read the sentences and to put an X next to any that they thought was not expressed in good English. They were instructed that there were no right or wrong answers but that it was their first decision that we were interested in. There was no time limit and the subjects were allowed to work at their own pace. A copy of the test used is shown in the Appendix.

Results and Discussion

The responses, depending on whether they had been judged to be grammatical or ungrammatical were estimated on the basis of the performance of the subjects in each group according to the five stimulus sentence types described above. The percentage of confirming judgements obtained as well as the means and standard deviations are shown in Table 1.

The results show that both control groups as well as the experimental groups generally judged Type 1 sentences that contain monosyllabic alternating verbs and the dative prepositional phrase complement to be grammatical. With Type 2 sentences that contain alternating dative verbs

Tabl

e I

Per

cent

age,

mea

ns, a

nd

sta

ndar

d de

viar

rons

qJ

con

firm

ing

resp

onse

s

Con

t-E

Ex

p-F

Fxp-

I C

ont-

E

Lxp

-F

txp

-1

~~

__

__

__

_

I(N

=h)

2(N

=6)

I(N

=23)

2(

N=7

) 3(

N=1

5)

l(K

=12)

2(

N=8

) 3(

N=1

8)

I(N

=6)

2(

Y=6

) I(

N=2

3)

?(Y

=7)

3(N

=15)

I(

N=

12)

?(N

=8)

3(Y

=IX

,

Perc

ent

Mea

n S.

D.

Perc

ent

Mea

n S

.D.

Perc

ent

Mea

n s

D.

Perc

ent

Mea

n S.

D

To-d

ativ

e T

ype

I (k =

5):

Pete

r thr

ew a

foot

ball

to P

hili

p.

100

100

98.4

94

.3

96.0

96

.7

5.0

5.0

4.91

4

71

4.8

4.83

0

0 .4

2 .7

6 .4

1 .5

8

To-d

ativ

e Typ

e 2

(k =

5).

Pete

r thr

ew P

hili

p a

foot

ball

. 90

.0

100

47.9

62

.9

81.3

80

.0

4.50

5.

0 2.

39

3.14

4.

07

4.0

.55

0 1.

62

1.46

1.

71

1.04

To-

dati

ve-r

ype 3

(k =

3):

Dav

id s

ugge

sted

a tr

ip to

Rut

h 10

0 I0

0 92

.8

95.2

10

0 94

.4

3.0

3.0

2.78

2.

86

3.0

2 83

0 0

60

38

0

39

To-d

ativ

e T

ype

4 (k

= 3

). *D

avid

sug

gest

ed R

uth

a tr

ip.

66.7

88

.9

64.8

66

.7

40.0

47

.2

2.00

2.

67

1.96

2.

00

1.20

1.

42

63

.X2

1.15

1.

00

1.08

1.

24

97.5

4.

88

.35

87.5

4

38

.9?

95.x

2.

88

.35

79.2

’ ?X

52

Dis

trac

tor s

ente

nces

Typ

e 5

(k =

5):

Den

nis a

nnoy

ed K

aren

yes

terd

a!

Perc

ent

100

100

71.7

77

.1

8X.0

90

.0

97.5

Mea

n 5.

0 5.

0 3.

57

4.86

4.

4 4.

5 4.

88

S.D

. 0

0 1.

38

1.22

.6

3 .6

7 35

For-

datiL

e T

ype

I (k =

5).

Dia

ne b

aked

a c

ake

Tor h

icol

c.

100

96.7

10

0 96

.6

100

98.7

98

.3

5.0

4.83

5.

0 4.X

! 5.

0 4.

93

4.92

0

.4 I

0 65

0

.26

29

For-

dati

\c 1

)pe

2 (h

= 5

) D

iane

hak

ed Y

icol

c a

cake

87

X

93.3

X6

.7

41 X

54

.3

XI

3 75

.0

4.39

4

67

4 33

2

09

2.

71

4.07

3.

75

.7X

.52

.52

1.73

1.

25

1.10

1

29

For-

dati

\e T

ype

3 (k

= 3

): A

nne

crca

tcd

a co

stum

e lo

r Sar

ah.

100

100

100

98.6

90

.5

95 6

94

.4

3.0

3.0

3.0

2.96

2

71

2.87

2

83

0 0

0 .2

1 .7

6 .3

5 5x

F(ir

-dat

i\e -

1 ype

4 (

k =

3).

*A

nne

crea

ted

Sar

ah a

Cm

tum

c 55

6

3X.9

72

.2

69 3

71

.4

48.6

66

.7

1.67

1.

17

21

7

20

9

2.14

14

7 2.

00

I ?X

98

1.

17

x5

.YO

1.06

.9

5

97.5

4.

88

75

92.5

4.

63

52

Y5.X

?.X

X 35

50.0

I

50

I07

I00

5.0 0

;s 2 1.

11

f $-

100

>

X4 4

12

2

F -_

3.0 0

46 3

I

39

1.79

98.9

4.

94

.24

‘U n

grar

nrna

tical

scni

encc

iypc

s 4

I04 Language Learning Vol. 34, No. I

appearing with double-NP complements, the results show an increasing judgement of acceptability across the three levels in the Inuit groups and, to a lesser extent, in the native English-speaking control groups which suggests that the alternation has not been fully acquired by them. The French experimental groups show less systematic acquisition of this structure as compared to the Inuit subjects; however, the standard deviations demonstrate that over one-third of the subjects judge more sentences differently from the group as a whole. In other words, considerable variation within groups is a likely indicator of the acquisition of the alternation. Generally speaking, then, these results bear out our prediction that the unmarked dative prepositional phrase complement would be acquired before the marked double-NP complement. From a formal point of view, if the dative alternation were, in fact, governed by a transformational movement rule, then one could reasonably assume that the rule had been acquired as indicated by the results elicited by Type 2 sentences, since, a t the lower end, a t least 30% of them were judged grammatical. However, if that were the case, one would have predicted an even higher percentage of grammaticality in this category since the rule should have applied across the board whenever the structural description of the rule was met. As one can see, this was not the case, as the results demonstrate a n acquisitional sequence whereby the subcategorical features were learned individually for dative verbs that alternate. It is a process that takes place over time and the results reflect this fact.‘

The control and experimental groups demonstrated very little difficulty in judging Type 3 sentences containing nonalternating dative verbs that appeared with prepositional phrase complements. Like Type I sentences, these were considered to be the unmarked structures. However, the results elicited by the ungrammatical Type 4 stimulus sentences indicate that overgeneralizations about the dative alternation were made, not only by the experimental subjects but by the native English speaking controls. The Cont-1 subjects judged a higher percentage of these sentences to be acceptable compared to the Cont-2 subjects which shows that the younger

‘We also carried out an Auditory Recall Test, using a subset of the dative verbs used in this test which were built into a short story that was taped and played to the subjects (cf. Mazurkewich 1982). The tape was stopped at measured intervals and the subjects were provided with two written sentences, one of which was the actual one heard on the tape and one which was changed in some way. The subjects’task was tochoose which one they thought they had heard on the tape. The results obtained from this test support the same sequence elicited by the Intuitive Judgement Test.

Mazurke wich 105

native speakers found more of the Type 4 sentences to be grammatical than the older native speakers. This may be a reflection of an incomplete acquisition of the constraints governing the acquisition, but the fact that the younger native speakers found more of them to be grammatical than the older ones suggests that these overgeneralizations will eventually be dropped from their grammar. We consider that some of the judgements obtained from the experimental groups on Type 4 sentences might also be attributed to overgeneralizations as well, especially in the data obtained from the Advanced experimental groups. However, we are unable to separate them from judgements that are a result of an incomplete acquisition of the complements involved using only this evidence. Further testing would be required to enable one to analyze the responses in a systematic way.

Judgements made about stimulus sentences containingfor-datives were not as accurate in general as those containing to-datives. This was found in the data elicited from the control groups as well as the experimental groups. Although the acquisition offor-datives seems to lag behind that of to-datives, the results obtained from the for-datives always mirrored that of the to-datives. Fischer( 1971) reports that she had thedistinct impression that the young children she studied found for-datives difficult t o process. Perhaps the problem with for-datives stems from the semantic notion of “benefactiveness” associated with them which may be acquired later than the notion of “goal” that is associated with to-datives.

It was suggested by one of the anonymous reviewers that perhaps the Inuit subjects may have been treating English prepositions as equivalent to case inflections in their own language, that is, they associated the [NP PP] feature in English with the dative case marked noun phrase in Inuktitut so that transference may have been responsible for the acquisitional sequence. The reviewer further suggested that in another language such as Igbo, a Kwa language of Nigeria in which dative verbs occur only in the [NP NP] pattern and in which there are no case inflections, second language learners might not show the acquisition sequence found in our Inuit subjects. If this indeed turns out t o be the case, then these acquisitional facts would have to be discussed within a theoretical framework which provides an explanation of why the facts fall out the way they do. For example, it would have to be explained in a principled way why, in a language like Inuktitut, learners would not associate the noncase carrying noun phrase with the dative prepositional phrase in English. It was argued elsewhere (Mazurkewich 1983b) that not only is the unmarked feature for datives in English based on

106 Language Learning Vol. 34, No. I

a theory of markedness, but at a more abstract level, this follows from the case theory outlined in Chomsky (1981b, 1982). In this theory all nouns must be assigned abstract case; otherwise the sentences will be ruled out. This hypothesis, which crucially involves the principle of adjacency, identifies, for example, the dative noun phrase appearing within the prepositional phrase as the unmarked option, since prepositions are considered to be one of the case assigners. Presumably, the unmarked option for datives in Inuktitut would be the overtly case marked noun rather than the noncase carrying noun that appears in ergative constructions (fn. 5) . It might be posited that in the learning of a second language, the unmarked option in the first language is associated with the unmarked option in the second. Perhaps it would involve an acquisition process of identifying grammatical relations with semantic roles (cf. Marantz 1982) that is extended in a similar way during the second language acquisition process. This is certainly a plausible hypothesis but one that would require a good deal further study.

Conclusions

The results we obtained in our testing bear out the prediction that second language learners of English will acquire dative prepositional phrase complements before double-NP complements. This sequence of emergence might be posited to result from transference by the French subjects since the dative structure in French is similar to the prepositional phrase in English. The fact that the same acquisition sequence is followed by the Inuit subjects rules out this possibility. Both forms of the dative alternation are available in the linguistic environment that learners are exposed to. The results of this testing demonstrate that second language learners select data for themselves in much the same way that first language learners have been shown to do (cf. Newport et a]. 1977; Roeper 1978). We propose that the emergence of the feature [NP PP], which we argue to be the unmarked one, before the feature [NP NP], the marked one, follows from the theory of markedness outlined in this paper.

Mazurkewich 107

REFERENCES

Baker, C.L. 1979. Syntactic theory and the projection problem. Linguistic Inquiry 10:533-

Brown, R. 1973. A First Language. Cambridge, Mass.: Harvard University Press. Brown, R., and C. Hanlon. 1970. Derivational complexity and order of acquisition in child

speech. In Cognition and the Development of Language. ed. J . Hayes. New York: John Wiley & Sons.

581.

Chomsky. N. 1965. Aspects o f t h e Theory of Syntax. Cambridge, Mass.: MIT Press. Chomsky, N. 1980. Rules and Represenrations. New York: Columbia University Press. Chomsky, N. 1981a. Markedness and core grammar. In Theor-r’of Markednessin Generative

Grammar, eds. A. Belletti. L. Brandi, and L. Rizzi. Pisa: Scuola Normale Superiore. Chomsky, N. I98 I b. Lectures on Government and Binding. Dordrecht: Foris Publications. Chomsky, N. 1982. Some Concepts and Consequences qf the Theor.), of Governmenr and

Emonds, J . 1972. Evidence that indirect object rule is a structure-preserving rule. Founda-

Emonds, J . 1976. A Transformational Approach to English S.vntax. New York: Academic

Fillmore, C. 1965. The Indirect Object Construction in English and the Ordering of

Fischer, S. 1971. The Acquisition of Verb-Particle and Dative Constructions. Doctoral

Fischer, S. 1976. Child language as a predictor of language change. Working Papers in

Goldsmith, J . 1980. Meaning and mechanism in grammar. In Harvard Studies in Syntax and

Green, G.M. 1974. Semantics and S.yntactic Regularity. Bloomington, Ind.: Indiana Univer-

Greenbaum, S., ed. 1977. Acceptability in Language. The Hague: Mouton. Hakuta. K . 1974. Prefabricated patterns and the emergence of structure in second language

Irvine. P., P. Atai, and J.W. Oller. 1974. Cloze. dictation, and the test of English as a foreign

Jackendoff, R. 1975. Morphological and semantic regularities in the lexicon. Language

Jackendoff, R.S., and P. Culicover. 1971. A reconsideration of dative movements. Foundations of Language 7:397-412.

Marantz, A. 1982. The connection between grammatical relations and lexical categories in language acquisition and linguistic theory. In Studies in Generative Grammar and Languuge Acquisition, eds. Y. Otsu et al. Tokyo: International Christian University.

Mazurkewich, 1. 1982. Second Language Acquisition of the Dative Alternation and Markedness: The Best Theory. Doctoral dissertation, UniversitC de Montreal.

Mazurkewich, 1. 1983a. Dative questions and markedness. In LlniversalsofSecond Language Acquisition, eds. F. Eckman, L. Bell, and D. Nelson. Rowley, Mass.: Newbury House ( to appear).

Mazurkewich, I . 1983b. Syntactic markedness and language acquisition. Paper presented at TESOL Convention, March 1983. Toronto.

Mazurkewich, I . , and L. White. 1983. The Acquisition of the Dative Alternation: Unlearning Overgeneralizations. Manuscript.

Binding. Cambridge, Mass.: MIT Press.

tions o f Language 8:546-561.

Press.

Transj’ormarions. The Hague: Mouton.

dissertation, MIT.

Linguistics 8:71-104, University of Hawaii.

Semantics 111, ed. S. Kuno.

sity Press.

acquisition. Language Learning 24:287-297.

language. Language Learning 24:245-252.

5 11639-412.

108 Language Learning Vol. 34, No. 1

McNeill, D. 1966. Developmental psycholinguistics. In The Gene.sis of Language, eds. F. Smith and G. Miller. Cambridge. Mass.: M I T Press.

Newport. E.L., H. Gleitman, and L.R. Cleitman. 1977. Mother, I'd rather d o it myself some effects and non-effects of maternal speech style. In Talking to Children. eds. C.E. Snow and C.A. Ferguson. Cambridge: Cambridge University Press.

Oehrle, R.T. 1976. The Grammatical Status of the Dative Alternation in English. Doctoral dissertation. MIT.

Oller. J .W. 1972. Scoring methods and difficulty levels for cloze tests of proficiency in English as a second language. Modern Language Journal 56: I5 I - 158.

Oiler. J.W. 1973. Cloze tests of second language proficiency and what they measure. Language Learning 23: 105- I 18.

Peters, A.M. 1977. Language learning strategies: does the whole equal the sum of the parts? Language 53: 560-573.

Peters. S. 1972. The projection problem: how is a grammar to be generated? In Goa1.s of Linguistic Theory, ed. S . Peters. Englewood Cliffs, N.J.: Prentice-Hall.

Porter, D. 1978. Cloze procedure and equivalence. Language Learning 28:333-341. Potts. M.. P. Carlson. R. Cocking, and C. Copple. 1979. Srructureand Development in Child

Language. Ithaca, N.Y.: Cornell University Press. Roeper, T. 1978. Linguistic universals and the acquisition of gerunds. In Papers in rhe

Structure and Development of Child Language, eds. H . Goodluck and L. Solan, pp. 1-36. Occasional Papers in Linguistics, vol 4. University of Massachusetts.

Roeper. T.. S . Lapointe, J . Bing. and S. Tavakolian. 1981. A lexical approach to language acquisition. In Language Acquisition and Linguisric The0r.v. ed. S . Tavakol ian. Cambridge, Mass.: M I T Press.

Stayton. B. 1972. The Acquisition of Direct and Indirect Objects in English. Manuscript. University of Kansas.

Stubbs. . l .B . . and G.R. Tucker. 1974. The cloze test as a measure of English proficiency. Modern Language Journal 58:239-24 I .

Williams, E.S. 1974. Rule Ordering in Syntax. Doctoral dissertation, MIT.

APPENDIX

Read the following sentences and put an X next t o any sentence that you think is not expressed in good English. There are no right or wrong answers. Your first decision is what we want. Please d o not change any of your answers.

__ 1 . Allan lent Louise a dollar. (Type 2 to-dative) - 2. Bob reported the accident to the police. (Type 3 to-dative) - 3. Isabel made Jerry a sweater. (Type 2 for-dative) - 4. David suggested Ruth the trip. (Type 4 to-dative) - 5. Fred took Joanne to a movie. (Type 5 Distractor)

Mazurkewich 109

__ 6. Paul designed a house for Claire. __ 7. Diane baked Nicole a cake. __ 8. Anne created a costume for Sarah. __ 9. Jack read Kim a story. -10. Cathy gave a book to Kevin. -1 I . Peter threw Philip a football. -12. Anne created Sarah a costume. -13. Tom captured the prize for Canada. -14. John sent a postcard to Carol. -15. Susan explained Jane the problem. -16. Sam walked Muriel home from school. -17. Cathy gave Kevin a book. -18. Nancy saved Mark a seat. -19. Bill bought a present for Mary. 2 0 . Sylvia chose a pattern for Betty. 2 1 . Diane baked a cake for Nicole. 2 2 . Dennis annoyed Karen yesterday. 3 . 3 . Sylvia chose Betty a pattern. 2 4 . Tom captured Canada the prize. 2 5 . Patrick rescued Lisa from drowning. 3 6 . Andrew chased Richard upstairs. 2 7 . Nancy saved a seat for Mark. 2 8 . Peter threw a football to Philip. 3.9. John sent Carol a postcard. 7 0 . Susan explained the problem to Jane. 3 1 . Bill bought Mary a present. 3 2 . Isabel made a sweater for Jerry. 1 3 . Allan lent a dollar to Louise. 1 4 . Bob reported the police the accident. 3 5 . David suggested the trip to Ruth. 1 6 . Jack read a story to Kim. 7 7 . Paul designed Claire a house.

(Type 3 for-dative) (Type 2 for-dative) (Type 3 for-dative) (Type 2 to-dative) (Type 1 ro-dative) (Type 2 to-dative)

(Type 4 for-dative) (Type 3 for-dative) (Type 1 to-dative) (Type 4 lo-dative)

(Type 5 Distractor) (Type 2 to-dative)

(Type 2 for-dative) (Type 1 for-dative) (Type 1 for-dative) (Type 1 for-dative) (Type 5 Distractor) (Type 2 for-dative) (Type 4for-dative)

(Type 5 Distractor) (Type 5 Distractor) (Type 1 for-dative) (Type 1 lo-dative) (Type 2 to-dative) (Type 3 to-dative)

(Type 2 for-dative) (Type 1 for-dative) (Type 1 to-dative) (Type 4 to-dative) (Type 3 to-dative) (Type I to-dative)

(Type 4for-dative)