Order preferences in the acquisition of verb clusters
Thesis rMA Linguistics
Student: Sabine van Reijen Studentnumber: 0518298
Supervisor: Prof. dr. F. P. Weerman Second reader: Dr. J. Don
2 1 Introduction
The present study is concerned with the acquisition of verb clusters in Dutch. A verb cluster is usually defined as a group of two or more verbs at the end of a clause that cannot be interrupted by other elements (e.g. Evers 1975, Zwart 2011). Here I will focus on the acquisition of clusters consisting of two verbs. An example of such a bipartite cluster is given in (1) below.
1 2
(1) a. (...) dat Jan een ijsje heeft gegeten.
2 1
b. (...) dat Jan een ijsje gegeten heeft. ‘(...) that John has eaten icecream.’
Example (1a) shows the so called 1-2 ordering, example (1b) shows the opposite 2-1 ordering. The numbering of the verbs is based on the underlying structure; the deeper the verb is embedded, the higher the number. In Dutch, verb clusters show a high degree of variation. That is, the order in which the verbs appear is to a large extent optional. For instance, both the 1-2 order in (1a) and the 2-1 order in (1b) are perfectly acceptable in Dutch. They carry the same semantic content, and there does not seem to be a rule that decides which of the two orders should be used. In other words, the choice between (1a) and (1b) seems completely arbitrary.
The question is, how children acquire verb clusters in Dutch when there is such a high degree of optionality.
Previous studies on the acquisitional pattern of verb clusters in Dutch show somewhat different results. Zuckerman (2001) finds that three-year-old children prefer the 2-1 order, as in (1b). Contrary to this, Meyer (2012) finds that four-to-six-year-old children prefer the 1-2 order, as in (1a).
In order to account for this difference, Meyer (2012) proposes the so called Red Flag
Hypothesis. The claim is that children start out with the 2-1 order as a result of the Dutch OV
word order rule. Crucially, in this stage the child has not yet acquired the notion of 'verb
cluster'. The 2-1 order gegeten heeft (‘eaten has’) merely reflects the Dutch word order rule
3
'red flag', triggering the child to analyse (1a) and (1b) in a similar way. In other words, the variety in order within verb clusters results in the acquisition of cluster hood.
The Red Flag Hypothesis makes clear predictions with regard to the acquisitional pattern, which is compatible with the findings in both Zuckerman (2001) and Meyer (2012): children in the first stage will show a preference for the 2-1 order, as this is compatible with the Dutch word order property. Only later, when they have acquired (or begin to acquire) the notion of cluster hood, will they start using the additional 1-2 order.
The present study investigates the acquisition of verb cluster hood by looking at order preferences of 71 monolingual Dutch children in bipartite verb clusters. Two different age groups are tested here. The first consists of 31 three-year-old children. The second consists of 40 four-to-six-year-old children. The question this study is concerned with, is whether three-year-old children are in a different acquisitional stage than the older children. For this purpose, the results of the younger children will be compared with the results of the year-old children reported in Meyer (2012), as well as with the results of the four-to-six-year-olds from the present study.(
In the next section, I first discuss the theoretical background. In section 3, the methodology is discussed, in section 4 the results of the experiment are presented. Section 5 covers the analysis. Finally, in section 6, I turn to the conclusions and discussion.
2 Theoretical Background
In this section, I first discuss the notion of verb clusters. The question I try to answer is how we define verb clusters. I provide an overview of possible bipartite verb clusters in Dutch and different approaches to the variety attested. Next, I discuss the acquisition of verb clusters.
2.1 Verb clusters
A verb cluster is usually defined as a verbal group that allows no other elements to interrupt it (Evers 1975, Zwart 2011). This phenomenon is attested in many West Germanic languages (Wurmbrand 2005). The fact that a verb cluster cannot be interrupted by other elements, is illustrated in (2) below.
(2) a. Ik zei dat Jan een ijsje gegeten (*snel) heeft. b. Ich sagte dass Jan ein Eis gegessen (*schnell) hat. c. I said that John has (quickly) eaten icecream.
4
The examples in (2) show that the adverbial snel/schnell (‘quickly’) cannot interrupt the verbal groups in Dutch (2a) and in German (2b). In English however, the adverbial can appear in between the two verbs ‘has’ and ‘eaten’, illustrated in (2c). Dutch and German are said to have verb clusters, whereas English does not exhibit this property (Wurmbrand 2005).
The reason a verbal group is impenetrable is attributed to verb raising (Evers 1975). In this approach, verb clusters are considered a surface phenomenon, derived from an underlying structure. The underlying structure is represented by (3a), the surface structure is represented by (3b). The structure in (3b) is derived from (3a) by means of verb raising.
(3) a. b.
VP VP
VP heeft VP heeft gegeteni
gegeten ti
een ijsje een ijsje
‘icecream eaten has’ ‘icecream has eaten’
The verb gegeten (‘eaten’) is raised out of its initial position and attaches to the governing verb heeft (‘has’) in (3b). Because the verbs are joined in the structure in (3b), they form an impenetrable cluster. Within this approach, the verb gegeten in (3a) can adjoin to the right of the governing verb heeft, resulting in the 1-2 order (illustrated in 3b). Crucially, the verb can also adjoin to the left of the governing verb, resulting in the opposite 2-1 order. This is illustrated in (3c) below. Thus, verb raising can result in both the 1-2 and the 2-1 order.
(3) c. VP VP gegeteni heeft ti een ijsje
5
It must be noted that it remains an open question what exactly is the trigger for verb raising. To solve this issue, Evers (1982) proposes that verbs are raised in order to be associated with TENSE. He claims that all verbs need to be associated with tense, and that verb raising functions to connect a verb unmarked for TENSE to a verb that is marked for TENSE.
English lacks the surface structure in (3b) or (3c), and therefore, verb groups allow other elements to interrupt it. German does have the structure in (3b), and Dutch has both the (3b) and the (3c) structure. Therefore, in Dutch and German no elements can appear in between the two verbs. It follows that Dutch and German have verb clusters and English does not.
Verb clusters show a great amount of diversity, both between and within langugaes. Even though German and Dutch share the property of verb raising, there is still a lot of variation between these two languages. German shows a strict 2-1 ordering (4a), the opposite 1-2 order yields an ungrammatical utterance (4b). In Dutch on the other hand, the order of verbs within a cluster is to a large extent optional, as both the 2-1 and the 1-2 order are grammatical, see (5a) and (5b).
2 1
(4) a. Ich sagte dass Jan ein Eis gegessen hat.
1 2
b. * Ich sagte dass Jan ein Eis hat gegessen ‘I said that John has eaten icecream.’
2 1
(5) a. Ik zei dat Jan een ijsje gegeten heeft.
1 2
b. Ik zei dat Jan een ijsje heeft gegeten, ‘I said that John has eaten icecream.’
The difference between (4) and (5) illustrates the diversity between languages when it comes to verb clusters. The reader is referred to Wurmbrand (2005) for an extensive empirical overview of different patterns attested across West Germanic languages. Verb clusters can vary even within one language. The examples in (5) show that for Dutch, contrary to German, the ordering of verbs in verb clusters is optional.
6
In this paper the focus is on bipartite verb clusters in Dutch. Dutch bipartite clusters can consist of an auxiliary and a participle, illustrated in (6a) and (6b), or of a modal and an infinitive, illustrated in (7a) and (7b) below.
(6) a. Ik zei dat Loes een boek heeft gelezen. b. Ik zei dat Loes een boek gelezen heeft.
‘I said that Loes has read a book.’
(7) a. Ik zei dat Loes een boek kan lezen. b. Ik zei dat Loes een boek lezen kan. ‘I said that Loes can read a book.’
For both the auxiliary/participle construction and the modal/infinitive construction, orders 1-2 and 2-1 are grammatical in Dutch. Some argue however, that for constructions with modal verbs, the 1-2 order is almost exclusively used in Dutch (e.g. Zuckerman 2001).
Furthermore, the bipartite cluster can also appear with only non-finite verbs, illustrated in (8) and (9).
(8) a. Loes is trots een boek te hebben gelezen. b. Loes is trots een boek gelezen te hebben.
‘Loes is proud to have read a book.’
(9) a. Loes is trots een boek te kunnen lezen. b. * Loes is trots een boek lezen te kunnen. ‘Loes is proud that she can read a book.’
For non-finite verb clusters, the modal/infinitive construction does not allow the 2-1 order. The auxiliary/participle construction does allow both orders even in the non-finite clusters.
This leaves the language user with a wide variety of seemingly arbitrary options. The question that arises is how the speaker decides which order to use. Are there strategies a speaker uses to determine this? Are there regional differences? Is it the case that one order is more frequent than the other in every day language use?
Coussé, Arfs & De Sutter (2008) provide an overview of different strategies the language user can employ when deciding between the possible orders for verb clusters. They
7
distinguish between five different factors that can influence the ordering of verbs in clusters. The first two factors, production pressure and region, have to do with context. The factor of production pressure refers to the observation that in contexts where the speaker has less time to control his or her language utterances, the 2-1 order is more frequent. The regional factor refers to the fact that the order that is more frequent is dependent on the region. For instance, in Flanders the 2-1 order is more prominent, whereas the 1-2 order is customary in Standard Dutch. The third factor, rhytmh, refers to the fact that speakers prefer a balanced rhythm in their utterances, which can influence the choice of order in verb clusters. Speakers prefer a hat-shaped prosodic rhythm, meaning that a stressed syllable should be surrounded by unstressed syllables. The fourth factor is semantics, which may also be of influence. When the participle in the auxiliary/participle construction is interpreted as being more adjective-like, this triggers the 1-2 order. The final factor is related to the discourse: once a certain order is used in the discourse, speakers are more likely to stick to that particular order.
All of the above reveals that there is a huge amount of variety in the order in Dutch verb clusters. Even though speakers have some strategies at their disposal, the order of verbs within a verb cluster remains to a large extent optional. This brings us to the question the present study is concerned with: how do children acquire verb clusters in Dutch? The fact that there is such a great amount of variation, does not make this task any easier. The next section discusses the notion of verb cluster hood in more detail, as well as the acquisition of verb clusters in Dutch.
2.2 The acquisition of verb clusters
The present study deals with the acquisition of verb clusters in Dutch by looking at order preferences in bipartite verb clusters. In this section, I first discuss previous studies on the acquisition of verb clusters.
Previous work on the acquisition of Dutch verb clusters reveals a difference between older and younger children. Zuckerman (2001) finds that three-year-old monolingual Dutch children prefer the 2-1 order. Meyer (2012) however, finds that four-to-six-year-olds prefer the 1-2 order.
Zuckerman (2001) performs two experiments. The first is an elicitation task, in which two age groups were tested on their order preferences in bipartite verb clusters containing an auxiliary and a participle. The youngest group, consisting of 17 three-year-old children, showed a clear preference for the 2-1 order. Contrary to this, the older age group, consisting
8
of 10 five-year-old children, revealed a preference for the opposite 1-2 order. In addition, 19 adults revealed a preference for the 1-2 order.
The second experiment reported in Zuckerman (2001) consisted of a sentence repetition task, in which the child was presented with a sentence containing a verb cluster, and was then asked to repeat it. The stimuli consisted of both 1-2 and 2-1 ordered verb clusters. A total of 8 three-year-old children participated in this sentence repetition task. The results hereof reveal a preference for the 2-1 order, as the children made fewer errors on stimuli of this order: in the 2-1 order, 85% of the stimuli was repeated correctly, whereas in the 1-2 order only 30% of the stimuli was repeated correctly.
The younger children in Zuckerman’s (2001) study, thus show a preference for the 2-1 order. The older children and the adults on the other hand, show a clear preference for the 1-2 order. This leads Zuckerman (2001) to conclude that the 2-1 order is the unmarked one, as the younger children clearly prefer this order. Therefore, he claims, this order must be the more economical one. This economy principle no longer plays a valid role in the adult grammar, which is claimed to be the reason adults do not show a preference for the 2-1 order.
In sum, Zuckerman (2001) claims that the 2-1 order is the least marked one. Around the age of four, a shift is visible: children start to use the more adult like 1-2 order. The principle of economy no longer plays a valid role at this point.
Meyer (2012) however, suggests that the data in Zuckerman (2001) do not reveal the 2-1 order as the unmarked one, but rather that the preference for this order is the result of the Dutch OV word order property. Dutch is generally taken to be an SOV language with a Verb Second rule, meaning that only in main clauses the finite verb is moved from sentence final to the second position. The basic rule in Dutch however, is that the finite verb appears in sentence final position. Meyer (2012) proposes that the younger children prefer the 2-1 order, because this is the one in which the finite verb appears at the end of the sentence, see example (10) below:
1 2
(10) a. Ik zei dat Loes een boek heeft gelezen.
2 1
b. Ik zei dat Loes een boek gelezen heeft. ‘I said that Loes has read a book.’
9
Only in example (10b) does the finite verb heeft (‘has’) appear at the end of the sentence, thus obeying the OV word order rule. According to Meyer (2012), this leads to a 2-1 order preference for the younger children.
Meyer (2012) investigates the order preferences of four-to-six-year-old children, in both bi- and tripartite verb clusters. For this purpose, two sentence repetiton tasks were conducted, in which a total of 49 monolingual Dutch children, aged 4;0 – 6;2, participated.
She finds that children of this age hardly produce any tripartite verb clusters. Here, the focus is on the bipartite stimuli. A total of 25 children was presented with sentences that contained bipartite verb clusters, in both the 1-2 and the 2-1 order. The clusters were of the type auxiliary/participle, as well as the modal/infinitive type (recall the examples given in section 2.1). Additionally, the stimuli consisted of both finite and non-finite clusters. This resulted in eight different stimuli for the bipartite condition. An example of a complete set of stimuli is given in Table 1 below:
Sentence Finiteness Cluster type Order 1. De beer zegt dat Loes Tim taart heeft gegeven. FIN AUX+PART 1-2
2. De beer zegt dat Loes Tim taart gegeven heeft. FIN AUX+PART 2-1
‘The bear says that Loes has given Tim cake.’
3. Loes is erg trots Tim taart te hebben gegeven. INF AUX+PART 1-2
4. Loes is erg trots Tim taart gegeven te hebben. INF AUX+PART 2-1
‘Loes is proud to have given Tim cake.’
5. De beer zegt dat Loes Tim taart mag geven. FIN MOD+INF 1-2
6. De beer zegt dat Loes Tim taart geven mag. FIN MOD+INF 2-1
‘The bear says that Loes may give Tim cake.’
7. Loes is erg trots Tim taart te mogen geven. INF MOD+INF 1-2
8. Loes is erg trots Tim taart geven te mogen. INF MOD+INF 2-1
‘Loes is very proud to be allowed to give Tim cake.’
10
First of all, the findings show that the children score significantly worse on non-finite verb clusters than on the finite clusters: in the finite condition 70.6% of the answers were correct, as opposed to only 38.33% in the non-finite condition.
Second, the results reveal a significant effect for order, with children scoring better on the 1-2 stimuli as compared to the 2-1 stimuli. In the auxiliary condition, children showed a preference for the 1-2 order, but only in the finite condition. In the modal condition, the preference for the 1-2 order is a bit stronger, as this is evident in both the finite and in the non-finite condition.
Overall, this study reveals that four-to-six-year-old children perform better on stimuli with the 1-2 order than on stimuli with the 2-1 order. If we compare this to the findings in Zuckerman (2001), the results are similar to what he finds for the older children and the adults. However, the younger children in Zuckerman (2001) show the exact opposite order preference, namely for the 2-1 order. Zuckerman (2001) concludes from this that the 2-1 order must be the unmarked one and this is the order that is acquired first, as younger children produce mostly this order. Contrary to this claim, Meyer (2012) suggests that the preference for the 2-1 order in younger children is the result of the Dutch OV word order. She proposes the so called Red Flag Hypothesis, which states that children follow three different stages when acquiring verb clusters in Dutch. In stage 1, children produce mostly 2-1 orders. Crucially, within this approach, they have not yet acquired the notion of verb clusters at this point. The 2-1 orders they produce in this stage merely reflect the OV word order in Dutch, which requires the verb to appear at the end of the clause. In this stage, Meyer (2012) argues, children might interpret (11a) and (11b) below as having similar structures.
(11) a. (…) dat mijn gestolen fiets groen was. ‘… that my stolen bike green was.’ b. (…) dat mijn groene fiets gestolen was.
‘… that my green bike stolen was.’
In the sentence in (11a), which does not contain a verb cluster, the finite verb was (‘was’) appears in sentence final position as a result of the OV property of Dutch. Meyer (2012) argues that for the first acquisitional stage, the same analysis holds for the sentence in (11b): children produce a 2-1 order with the finite verb in final position as a result of the OV property of Dutch. At this point, the child does not produce a verb cluster.
11
However, for a sentence like the one in (11b), the input also confronts children with 1-2 orders (see 11-2b). For a sentence like (11a), this conflicting input does not occur, as (11a) does not contain a verb cluster, and it is thus ungrammatical to switch the order of the adjective groen and the verb was (see 12a):
(12) a. (...) dat mijn gestolen fiets groen was / * was groen
‘... that my stolen bike green was / * was green.’ 2 1 1 2
b. (...) dat mijn groene fiets gestolen was / was gestolen.
‘... that my green bike stolen was / was stolen.’
The Red Flag Hypothesis states that these conflicting input data trigger the child to analyse (12a) and (12b) differently. In other words, the presence of the 1-2 orders in the input functions as a ‘red flag’, causing children to analyse (11b) and (12b) as being a verb cluster, as opposed to (11a) and (12a). Meyer (2012) situates this point somewhere around the age of four, based on the results in Zuckerman (2001), and her own findings which show that four-year-olds already produce more 1-2 than 2-1 orders, indicating that they are producing verb clusters.
It is very likely that this acquisitional pathway starts out in a construction specific way. Note that the participle gestolen (‘stolen’) in (12b) can easily be interpreted as being an adjective ‒ in fact, it functions as an adjective in (12a). Recall that verb clusters can also be of the type modal/infinitive (compare 13a and 13b below). The infinitive in (13b) is more clearly verb-like than the participle in (13a), and is therefore less likely to be interpreted as an adjective:
(13) a. (...) dat mijn groene fiets gestolen was. ‘... that my green bike stolen was.’ b. (...) dat hij mijn groene fiets stelen kan.
‘... that he can steal my green bike.’
The infinitive stelen (‘steal’) in (13b) is clearly a verb, whereas the participle gestolen (‘stolen’) in (13a) can also be interpreted as an adjective. Therefore, the modal/infinitive construction in (13b) obviously contains an actual verb cluster, whereas this is less straightforward for the auxiliary/participle construction in (13a). We can imagine that the
12
language learning child will start out learning verb raising and the notion of verb cluster hood with the (obviously verb-like) modal/infinitive constructions, and only later will analyse the (more adjective-like) auxiliary/participle constructions in a similar way. Thus, we expect that the acquisition of verb cluster hood starts out with the modal/infinitive verb clusters as in (13b), and that the auxiliary/participle verb clusters as in (13a) will follow later on.
It is not until the child reaches stage 3 in the development, that both the 1-2 and the 2-1 verb clusters are completely acquired. See Table 2 below for the three different stages proposed in Meyer (2012).
Stage 1 Stage 2 Stage 3
Mostly 2-1 orders OV, no clusters
Mostly 1-2 orders Clusters
Both 1-2 and 2-1 orders Clusters
Age: < 4;0 Age: 4;0 – 6;0 Age: > 6;0
Table 2: Different stages proposed in the Red Flag Hypothesis (Meyer 2012).
Table 2 shows that the Red Flag Hypothesis can account for the fact that three-year-old children seem to prefer different orders for verb clusters than four-to-six-year-old children do. The underlying analysis causes them to use different orders in different stages of development. This does not necessarily mean that the 2-1 order is the more economical one, as Zuckerman (2001) claims.
The claims presented in Meyer (2012) have consequences for the acquisitional pattern of verb clusters. According to the Red Flag Hypothesis, in the first stage of acquisition sentences (11a) and (11b) seem the same to the child: both are constructions with the inflected verb in sentence final position. In this stage, the 2-1 orders follow from the OV word order in Dutch, whereas the 1-2 orders do not. Only the 1-2 order necessarily is evidence for verb raising and thus for the existence of verb cluster hood. In this first acquisitional stage, the child responds differently to participles, as they are more adjective-like, than to infinitives, as they are more clearly verb-like. The overview in Table 2 above shows that it is predicted that children younger than four years will produce less 1-2 orders, as they have not yet acquired the notion of verb cluster hood. Furthermore, it is predicted that children between four and six years old will show a preference for the 1-2 order. Only in stage 3, after the age of six, will they show adult-like behaviour, using both the 2-1 and the 1-2 orders.
13
The present study aims to shed some light on the acquisition of verb cluster hood, by investigating the predictions the Red Flag Hypothesis makes. I do so by looking at order preferences of bipartite verb clusters in monolingual Dutch children of two different age groups: three-year-old children on the one hand, and four-to-six-year-old children on the other. If the claims in Meyer (2012) are correct, three-year-old children are in a different stage of acquiring verb cluster hood than four-to-six-year-old children. If this is indeed the case, three-year-olds are expected to show a different pattern of order preferences in verb clusters than the four-to-six-year-olds do. The four-to-six-year-old children are expected to show a clear cut preference for the 1-2 order for bipartite verb clusters, both for verb clusters containing a participle and for verb clusters containing an infinitive. Three-year-old children on the other hand, are expected to respond differently to verb clusters containing a participle than to verb clusters containing an infinitive. It is predicted that clusters with a participle lag behind, in that the child will take longer to analyse them as being verb-like and thus as a subject to verb raising. It follows that the 1-2 order should be more prominent in the verb-like modal/infinitive clusters, and less clear cut in the (more adjective-like) auxiliary/participle clusters. Overall, the Red Flag Hypothesis predicts that the three-year-old children will produce less 1-2 orders than the older children.
In order to test these predictions, two experiments were carried out. The first was to test the order preferences of three-year-old children in bipartite verb clusters. The second was to test the order preferences of four-to-six-year-old children. How these two experiments are carried out, is discussed in more detail in the next section.
3 Method
This section elaborates on the implementation of the two experiments conducted for the purpose of the present study. The first experiment is a replication of the experiment reported in Meyer (2012) and focusses on three-year-old children. The second experiment is designed especially for the purpose of the present study, to establish the preferences of four-to-six-year-old children in bipartite (finite) verb clusters. This second experiment is designed to make a fair comparison between verb clusters containing a participle and verb clusters containing an infinitive. It should therefore be able to tell us whether the acquisition of verb clusters is construction-specific (i.e. different patterns are found for different verb cluster types), or whether it is a more rule based process (i.e. all cluster types show a similar acquisitional
14
pattern). The results of the three-year-old children will be compared with the results of the four-to-six-year-old children, both from Meyer (2012) and from the present study.
Both experiments consist of a sentence repetition task, in which the participant hears a sentence and is afterwards asked to repeat it. This kind of task enables us to elicit verb cluster constructions from very young children. Since the stimulus the participant hears triggers a construction containing a verb cluster, it is more likely that the child will produce a cluster. Crucially, the basic idea behind a sentence repetition task is that the sentences are too long for the child to reproduce them by memory alone, forcing the child to use his own linguistic capacities to reconstruct the sentence (Eisenbeiss 2010). Thus, a sentence repetition task enables us to test the participants’ capacities with regard to verb cluster constructions.
In what follows, I first discuss the replication with the three-year-old children. Next, the second experiment with the four-to-six-year-old children is discussed.
3.1 First Sentence Repetition Task
The experiment with the three-year-olds aims at investigating the acquisition of verb cluster hood in Dutch. The Red Flag Hypothesis put forward in Meyer (2012), claims that three-year-old children have not yet acquired the notion of verb cluster hood. This predicts that when it comes to verb cluster constructions, children aged three behave differently from their slightly older peers. Where the four-to-six-year-old children show a preference for the 1-2 order, three-year-old children are expected to produce less 1-2 orders. For the children who have not yet acquired verb raising (and therefore verb cluster hood), the 2-1 orders are not yet analysed as clusters, but as following the OV word order in Dutch. The opposite 1-2 orders, although possible in Dutch, are not yet available to these children, as they would require a cluster analysis. The present study aims at testing this hypothesis by replicating the experiment reported in Meyer (2012), only with slightly younger children, namely three-year-olds. Following Meyer (2012) and Zuckerman (2001), the prediction is that children of this age group will show a preference for the 2-1 order in bipartite verb clusters.
As stated before, the experiment with the three-year-olds is a replication of Meyer (2012). For this purpose, I used the experiment Meyer (2012) designed in collaboration with another study on verbal inflection errors (Buijs, Van Reijen & Weerman, in press) to test childrens’ preferences in both bi- and tripartite verb clusters. The results in Meyer (2012) indicate that children of four to six years old hardly produce the complicated tripartite verb clusters. For this reason, the present study is limited to bipartite verb clusters alone.
15
The bipartite test items contained either an auxiliary and a participle, or a modal and an infinitive. For the auxiliary condition, the auxiliary hebben (‘to have’) is used. For the modal condition, the modal verbs kunnen (‘to be able to’ or ‘can’), mogen (‘to be allowed to’ or ‘may’) and zullen (‘will’) are used. Although the different verb types were kept constant – the child was presented with as many auxiliaries as modals – the individual verbs were not distributed evenly, as the task used only one auxiliary, as opposed to three modal verbs. Therefore, the different modal verbs were not tested to the same extent as the single auxiliary
hebben (‘to have’) was. We do not know to what extent these individual verbs play a role in
the acquisition of verb cluster hood. I come back to this issue in section 3.2, when discussing the newly designed experiment for the four-to-six-year-olds.
The auxiliaries and the modal verbs were combined with eight different lexical verbs. These verbs were selected based on production data from Dutch corpora in the CHILDES database (MacWhinney 2000; Van Kampen 1994; Wijnen & Verrips 1998). This way it is ensured that children of the intended age group are familiar with the verbs. Table 3 provides an overview of the auxiliaries and the modals used in the experiment, and the lexical verbs they are combined with.
Auxiliaries Modals Lexical Verbs
hebben
(‘to have’)
kunnen
(‘to be able to’)
dromen
(‘to dream’)
lezen
(‘to read’)
mogen
(‘to be allowed to’)
geven (‘to give’) spelen (‘to play’) zullen (‘will’) knippen (‘to cut’) vertellen (‘to tell’) lachen (‘to laugh’) zwemmen (‘to swim’)
Table 3: Verbs used in Meyer (2012), replicated here for the three-year-olds.
Every lexical verb was used once in combination with the auxiliary, and once with one of the modal verbs, both in the 1-2 and in the 2-1 orders. In addition, the resulting clusters were presented to the child in both a finite and a non-finite condition. This led to eight testing conditions, as presented in Table 1 in section 2.3, for convenience repeated here in Table 4 below. The table shows an example of a set of test sentences with the verb geven (‘to give’), which was presented with the auxiliary hebben (‘to have’), and the modal verb mogen (‘may’).
16
Sentence Finiteness Cluster type Order 1. De beer zegt dat Loes Tim taart heeft gegeven. FIN AUX+PART 1-2
2. De beer zegt dat Loes Tim taart gegeven heeft. FIN AUX+PART 2-1
‘The bear says that Loes has given Tim cake.’
3. Loes is erg trots Tim taart te hebben gegeven. INF AUX+PART 1-2
4. Loes is erg trots Tim taart gegeven te hebben. INF AUX+PART 2-1
‘Loes is proud to have given Tim cake.’
5. De beer zegt dat Loes Tim taart mag geven. FIN MOD+INF 1-2
6. De beer zegt dat Loes Tim taart geven mag. FIN MOD+INF 2-1
‘The bear says that Loes may give Tim cake.’
7. Loes is erg trots Tim taart te mogen geven. INF MOD+INF 1-2
8. Loes is erg trots Tim taart geven te mogen. INF MOD+INF 2-1
‘Loes is very proud to may give Tim cake.’ Table 4: Set of test sentences with the verb geven (‘to give’).
Each child was presented with all of the conditions three times. For example, the child was asked three times to repeat a sentence such as 1 (see Table 4), which contained the auxiliary
hebben and a lexical verb in the finite condition, in a 1-2 order.
All sentences were presented in the third person singular, to keep variation between the test items to a minimum. Moreover, all sentences were prerecorded by the same female speaker using neutral intonation. This to make sure that all children were presented with exactly the same test items and that possible intonational differences could not influence the results. Sentence lenght was kept constant; every sentence consisted of either 9 or 10 words, and comprised no less than 10, and no more than 12 syllables. A total of 48 test sentences was constructed. These sentences were combined with a supporting picture to make the test more attractive for the children (for an overview of all the test sentences and their supporting pictures used in this first sentence repetition task, see Appendix A).
The test items were distributed over two versions of the test, resulting in 24 test items per version. A 2:1 testitem:filler ratio was maintained, which led to 12 filler items per version of the test. Thus, both versions comprised a total of 36 test sentences. These sentences and
17
their supporting pictures were put together in a slideshow. To be able to control for effects of order and fatigue, two different orders of both versions were constructed. This resulted in four final versions of the test.
Of course, for the experiment in the present study, the test had to be adjusted to fit the capacities of three-year-old children. These younger children have a much shorter attention span than the older children do. The test with 36 items would be much too long for children of this age. Therefore, the sentence repetition task was cut in half to fit the three-year-olds’ attention span. A total of 18 test items per version remained, 12 of which were verb cluster items. The remaining six were filler items. A collobaration was arranged in which the filler items in the present study functioned as test items for a different study on verbal inflection errors (Buijs 2013), and vice versa. An example of a filler item is given below in (14).
(14) Mama zegt dat kleine Tim op zijn kamer leest. Mommy says that little Tim in his room reads
‘Mommy says that little Tim is reading in his room.’
For the three-year-olds, the test was presented as if it were a game. A doll was used to create a situation in which repeating the sentence seemed natural for the child. The experimenter introduced the doll and told the child that the doll was not able to talk to computers, but only with people. So, the child had a very important task, namely to tell the doll exactly what the computer had said. The experiment was executed in a separate corner in the classroom the child frequented.
Every session started with an introduction, in which the child was familiarized with the task. Only if the child understood what was expected, did the actual experiment begin. The experiment was carried out by two experimenters: one interacted with the child, the second scored the childs’ answers. All sessions were audiorecorded. The responses were judged by two people in a second listening session. In case of doubt on how to score a particular answer, a third persons’ judgment was decisive.
A total number of 31 children participated in this study: 6 boys and 25 girls. At the moment of testing, the mean age was 3;7, with ages ranging from 2;8 to 4;0. All the children attended a child day care centre in the western region of the Netherlands. They were all typically developing and monolingual speakers of Standard Dutch. For this first sentence repetition task, it is not necessary to obtain results from an adult control group, as Meyer
18
(2012) already established that for this experiment, adults do not show a preference for either the 1-2 or the 2-1 order. The results of the three-year-old children are presented in section 4.1.
In the next section, the sentence repetition task with the four-to-six-year-old children is discussed.
3.2 Second Sentence Repetition Task
The second sentence repetition task serves to investigate the acquisition of verb clusters by looking at order preferences of four-to-six-year-old children in bipartite verb clusters. The results hereof will be compared with the results of the three-year-old children, in order to establish possible differences in order preferences between these two age groups. Following Meyer (2012) and Zuckerman (2001), it is expected that overall, children of four to six years old will show a preference for the 1-2 order in bipartite verb clusters.
Just as with the three-year-olds, the older children participated in a sentence repetition task. As stated before, the present study is limited to bipartite verb clusters, as the results in Meyer (2012) indicate that children of this age group have trouble producing tripartite clusters. In addition, only finite items were tested here, as the results in Meyer (2012) reveal that four-to-six-year-old children have trouble producing the non-finite clusters.
The test items contain either an auxiliary and a participle or a modal and an infinitive. The present experiment offers a balanced selection of auxiliaries and modals: two auxiliaries – zijn (‘to be’) and gaan (‘to go’) – are compared with two modals: kunnen (‘to be able to’ or ‘can’) and moeten (‘to have to’ or ‘must’). These verbs are selected based on previous studies (Schlichting 1996, Jordens 2002), that show that children of the intended age group have acquired them. This test design enables us to establish to what extent individual verbs can influence the order preference of bipartite verb clusters. In other words, it enables to determine to what extent the acquisition of verb clusters is a construction-specific or rule-based process.
Nine different lexical verbs were used to create verb clusters with these auxiliaries and modals. They were all selected based on production data from the CHILDES database (MacWhinney 2000; Van Kampen 1994; Wijnen & Verrips 19), to make sure that children of four to six years old are familiar with the verbs. Table 5 below provides an overview of the auxiliaries and the modals used in this experiment, and the lexical verbs they are combined with.
19
Auxiliaries Modals Lexical Verbs
gaan
(‘to go’)
kunnen
(‘to be able to’)
fietsen (‘to cycle’) rennen (‘to run’) vallen (‘to fall’) zijn (‘to be’) moeten
(‘to have to’)
klimmen (‘to climb’) rijden (‘to run’) vliegen (‘to fly’) lopen (‘to walk’) springen (‘to jump’) zwemmen (‘to swim’)
Table 5: Verbs used in the second sentence repetition task.
Every lexical verb is combined with both auxiliaries and both modals, in the 1-2 and in the 2-1 order. Thus, every lexical verb yielded eight test sentences. This resulted in a total of 72 test items (see Appendix B for an overview of all the test items used in this second sentence repetition task). Sentence lenght was kept constant; all sentences consisted of either 9 or 10 words, and 12 to 14 syllables. The table below gives an example of a set of test sentences with the lexical verb vallen (‘to fall’).
Sentence Verb Cluster type Order
Mama zegt dat de appel uit de boom is gevallen. zijn AUX+PART 1-2
Mama zegt dat de appel uit de boom gevallen is. zijn AUX+PART 2-1
‘Mommy says that the apple has falenl from the tree.’
(‘to be’)
Mama zegt dat de appel uit de boom gaat vallen. gaan AUX+PART 1-2
Mama zegt dat de appel uit de boom vallen gaat. gaan AUX+PART 2-1
‘Mommy says that the apple is going to fall from the tree.’
(‘to go’)
Mama zegt dat de appel uit de boom kan vallen. kunnen MOD+INF 1-2
Mama zegt dat de appel uit de boom vallen kan. kunnen MOD+INF 2-1
‘Mommy says that the apple can fall from the tree.’ (‘can’)
Mama zegt dat de appel uit de boom moet vallen. moeten MOD+INF 1-2
Mama zegt dat de appel uit de boom vallen moet. moeten MOD+INF 2-1
‘Mommy says that the apple must fall from the tree.’
(‘must’)
20
The test sentences were divided over two versions of the test. In each version, the lexical verb appeared with one modal and one auxiliary, in both the 1-2 and the 2-1 order. Every lexical verb thus appeared four times in each test version. The test design ensured a fair comparison between the verb types: every child was presented with an equal amount of modals and auxiliaries.
All sentences were presented in the third person singular, to keep the variables other than order and verb type to a minimum. In addition, all sentences were prerecorded by the same female speaker. This to ensure that all children heard exactly the same stimuli and to control for possible effects of intonation. Every test sentences was accompanied by a supporting picture.
Of the resulting test items, 48 were used in the experiment presented here. The remaining 24 functioned as filler items in a collaborative study on verbal inflection errors (Buijs 2013). In exchange for this, the present study used test items from the study on verbal inflection errors as filler material. An example of a filler item is given in (15).
(15) Mama zegt dat Tim gisteren bij zijn opa speelde. Mommy says that Tim yesterday at his grandfather played
‘Mommy says that yesterday Tim played at his grandfathers’.’
Finally, the test items were put together in random order in a slideshow. The items were divided over the two versions of the test, leaving 24 test items per version. Together with the 12 filler items, this resulted in 36 items per version of the test. To control for possible effects of order and fatigue, each version was also presented in an alternating order. This resulted in four different versions of the final test.
Each child was seen individually, in a separate classroom. Two experimenters carried out the experiment: one interacted with the child, the other one scored the answers. Every session started with an introduction to familiarize the child with the task. If the child understood what was expected, the actual experiment began. All sessions were audiorecorded and all responses were judges by two people in a second listening session. In case of doubt on a particular answer, a third judgment was decisive.
A total of 42 children participated in this study. Of this group, two children were not included in the final analysis, because they were unable to finish the test. A total number of 40 children remained: 26 boys and 14 girls. Of these children, 20 participated in the experiment presented above, and 20 participated in the study on verbal inflection errors (Buijs 2013). At
21
the moment of testing, the mean age was 5;6, with ages ranging from 4;0 to 6;7. All the children attended school in the western region of the Netherlands. They were all typically developing and monolingual speakers of Standard Dutch.
Additionally, a control group of 10 adults participated in the experiment: 6 men and 4 women. The mean age was 25, with ages raging from 23 to 29. All participants are monolingual speakers of Standard Dutch. The results of the four-to-six-year-old children, as well as the results of the adult control group are presented in section 4.2.
4 Results
In this section, the results of the sentence repetition tasks presented above are discussed. Both aimed at investigating the acquisition of verb clusters. The question the present study is concerned with, is whether three-year-old children are in a different acquisitional stage than four-to-six-year-old children. If this is indeed the case, we expect to find different patterns of order preferences for the younger children as compared to the older children.
Below I first discuss the results of the sentence repetition task with the three-year-old children. Next, the results of the second sentence repetition task, with the four-to-six-year-old children, is discussed.
4.1 First Sentence Repetiton Task with the three-year-old children
The first sentence repetition investigates the acquisition of verb clusters by establishing the order preferences of three-year-old children in bipartite clusters. The results hereof will be compared with the results of the four-to-six-year-old children reported in Meyer (2012), and to a lesser extent with the results of the four-to-six-year-old children from the present study.
A total of 31 three-year-olds participated in this study, 15 of which participated in the experiment designed for verb clusters. The remaining 16 participated in its colloborative study on verbal inflection (Buijs 2013). This resulted in a total of 276 responses. These responses were divided into five categories: ‘correct’, ‘reverse order’, ‘doubling error’, ‘only one verb uttered’ and finally, ‘other’. An answer was scored as correct when the child produced the same verbs in the same order as presented in the test item. When the child uttered the same verbs, but in the reversed order as offered in the stimulus, the answer was scored as ‘reverse order’. An answer was considered a doubling error when the child uttered the finite verb both before and after the main verb: kan zwemmen kan (‘can swim can’). The category ‘only one verb’ refers to the cases in which the child uttered only one of the verbs offered in the
22
stimulus. Finally, the category ‘other’ contains all the remaining errors. A response was considered an error when the verbs produced were different from the verbs offered in the stimulus, when the response was an ungrammatical utterance, when the response did not contain an embedded clause, or when the child did not repeat the presented test item at all. The distribution of the different response types is given in Table 7 below.
Response Type
Correct Reverse order Doubling error Only one verb Other Total
Frequency 66 17 10 87 96 276
Percentage 23.9% 6.2% 3.6% 31.5% 34.8%
Table 7: Spread of response types for the three year old children.
Before taking a closer look within the different conditions separately, the data is analysed together. Table 7 shows that overall, the three-year-olds were not very likely to produce verb clusters; 34.8% of the responses are coded as error. Furthermore, in 31.5% of the cases the children produced only one verb, instead of a verb cluster. This shows that this age group has trouble producing verb clusters. For the remaining responses, 3.6% consisted of doubling errors. This means that only 30% of the answers contained a verb cluster, either in the same order as offered in the stimulus (23.9%), or in the reversed (6.2%). This differs from the results for the four-to-six-year-old children in Meyer (2012), which shows that the older children are already a lot better at producing verb clusters. In Figure 1 below, the results for the three-year-olds and the results for the four-to-six-year-olds in Meyer (2012) are compared.
23
Figure 1: Responses of three-year-olds and four-to-six-year-olds in Meyer (2012) compared.
The figure shows that the four-to-six-year-old children produce verb clusters a lot more often than the younger children do, as their responses are ‘correct’ and ‘reverse order’ more often than was the case for the three-year-olds. Furthermore, Figure 1 shows that the older children make fewer errors, and produce ‘only one verb’ less often than the younger children. Thus, when it comes to verb clusters, the four-to-six-year-olds are much better than three-year-olds.
In Meyer (2012), an effect for finiteness is reported: four-to-six-year-old children score significantly better on finite items as compared to the non finite ones. For the three-year-olds, the same effect is present (X2 (4, n = 276) p < 0.001).
Finally, the results reveal no significant effect for verb type, which indicates that the difference between auxiliary/participle and modal/infinitive constructions is not significant (X2 (4, n = 276) p = 0.754). This is again similar to the findings reported in Meyer (2012).
For the purpose of the present study, we want to find out whether the three-year-old children are more likely to produce bipartite verb clusters with a 2-1 or with a 1-2 order, and compare this with the order preferences of the four-to-six-year-old children. Therefore, we take a closer look at the effect of order. The table below shows the response types of the three-year-old children, for the 1-2 and the 2-1 stimuli separately.
0% 10% 20% 30% 40% 50% 60%
Correct Order Reverse Order Doubling error Only one verb Other error Three-year-olds Four-to-six-year-olds in Meyer (2012)
24
Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 37 6 0 47 47 137
% 27% 4.4% 0% 34.3% 34.3%
2-1 Freq. 29 11 10 40 49 139
% 20.9% 7.9% 7.2% 28.8% 35.3%
Table 8: Distribution of response type of the three-year-olds per order.
This table shows that for both the 1-2 and the 2-1 stimuli, children are the most likely to produce no verb cluster at all, as most responses belong either to the ‘only one verb’ or to the ‘other’ category. If we turn to the remaining responses, these show that the three-year-old children produced more correct answers when presented with the 1-2 order than when presented with the 2-1 stimuli (27% vs. 20.9%). Additionally, reversed orders show the exact opposite pattern; they are more likely to occur with the 2-1 than with the 1-2 stimuli (7.9% vs. 4.4%). Finally, doubling errors occur only in the 2-1 order. This overall picture reveals that the three-year-olds produce more 1-2 orders as opposed to 2-1 orders. A first chi-square test for independence reveales that this difference is indeed significant (X2 (4, n = 276) p = 0.011). This is similar to the results reported in Meyer (2012), who also finds a significant effect for order that shows that children prefer the 1-2 order. This is somewhat unexpected, as the claim that children under the age of four have not yet acquired verb clusterhood (Meyer 2012), predicts that these younger children will produce mostly 2-1 orders. However, a chi square test for independence reveals that the results of the three-year-old children do significantly differ from the results for the four-to-six-year-olds in Meyer (2012): (X2 (4, n = 276) p = 0.011). This indicates that the younger children, although showing a similar pattern for all the results taken together, do behave differently than the older children did. This becomes more clear when I discuss the different conditions seperately below.
In sum, the results presented here reveal a significant effect for finiteness, with children scoring better on finite items, as well as a significant effect for order; the three-year-olds show a preference for the 1-2 order. At first sight, it thus seems that the results of the three-year-olds are quite similar to the results of the older children: both show a preference for the 1-2 order. When we look within the different conditions however, a different picture
25
emerges. Below, all the different conditions are discussed separately and compared with the results of the four-to-six-year-old children in the previous study on verb clusters (Meyer 2012).
The first condition presented here is the auxiliary/participle construction within the finite condition. This contained sentence types 1 and 2: heeft gegeten (‘has eaten’, 1-2 order) and gegeten heeft (2-1 order). Table 9 below shows the distribution of response types within this condition for the three-year-old children. For comparison, Table 10 shows the distribution for the four-to-six-year-old children in Meyer (2012).
AUX/PART finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 15 5 0 12 3 35
% 42.9% 14.3% 0% 34.3% 8.6%
2-1 Freq. 15 4 4 5 7 35
% 42.9% 11.4% 11.4% 14.3% 20%
Tabel 9: Distribution of response types for the three-year-olds.
AUX/PART finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 57 11 0 4 3 75
% 76% 14.7% 0% 5.3% 4%
2-1 Freq. 47 16 6 2 4 75
% 62.7% 21.3% 8% 2.7% 5.3%
Table 10: Distribution of response types for the four-to-six-year-olds in Meyer (2012).
First, we look at the results for the three-year-old children. A chi-square for independence reveals no significant effect for order (X2 (4, n = 70) p = 0.072) for this age group. The three-year-olds do not seem to have a preference for either heeft gegeten (1-2) or gegeten heeft (2-1), as they score almost similar on stimuli with a 1-2 order and stimuli with a 2-1 order. Second, the results of both age groups are compared. The results for the three-year-old children are different fromt the results for the four-to-six-year-old children, in that the older children do show an order preference: they produce significantly more 1-2 orders (Meyer 2012). A chi square analysis reveals a significant effect for group within this
26
auxiliary/participle finite condition: (X2 (4, n = 220) p < 0.001). Thus, the three-year-old children score significantly different from the four-to-six-year-old children within this condition. Children of the younger age group do not reveal a preference for either of the two orders, whereas children of the older age group reveal a preference for the 1-2 order (heeft
gegeten).
The second condition is the auxiliary/participle construction in the non-finite condition. Sentences within this condition were of the type te hebben gegeten (‘to have eaten’, 1-2 order) and gegeten te hebben (2-1 order). Table 11 shows the distribution of responses for the three-year-old children in this condition. Table 12 shows the results for the four-to-six-year-old children.
AUX/PART non-finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 1 0 0 13 20 34
% 2.9% 0% 0% 38.2% 58.8%
2-1 Freq. 0 1 0 17 16 34
% 0% 2.9% 0% 50% 47.1%
Table 11: Distribution of response types for the three-year-olds.
AUX/PART non-finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 21 1 1 24 28 75
% 28% 1.3% 1.3% 32% 37.3%
2-1 Freq. 36 1 1 15 22 75
% 48% 1.3% 1.3% 20% 29.3%
Table 12: Distribution of response types for the four-to-six-year-olds in Meyer (2012).
For the results of the three-year-olds, a chi-square for independence reveals no significant effect for order (X2 (3, n = 68) p = .395). Again, the three-year-old children score similar on stimuli with a 1-2 order as on stimuli with a 2-1 order. This is similar to the results for the four-to-six-year-old children, where no significant effect for order is reported (Meyer 2012). However, the three-year-olds do differ from the older children in this condition, in that they produce far less non-finite verb clusters than the older children did. The compared results of
27
the two groups reveal a significant effect for group (X2 (4, n = 218) p < 0.001). Thus, the three-year-old children produce significantly less verb clusters in this condition than the older children in Meyer (2012) did.
The third condition is the modal/infinitive construction within the finite condition. Sentences of this type are kan zwemmen (‘can swim’, 1-2 order) and zwemmen kan (2-1 order). Table 13 shows the distribution of the response types for the three-year-old children, Table 14 the distribution for the four-to-six-year-old children.
MOD/INF finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 16 1 0 6 11 34
% 47.1% 2.9% 0% 17.6% 32.4%
2-1 Freq. 8 5 6 7 9 35
% 22.9% 14.3% 17.1% 20% 25.7%
Table 13: Distribution of response types for the three-year-olds.
MOD/INF finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 63 0 0 1 11 75
% 84% 0% 0% 1.3% 14.7%
2-1 Freq. 45 12 7 6 5 75
% 60% 16% 9.3% 8% 6.7%
Table 14: Distribution of response types for the four-to-six-year-olds in Meyer (2012).
For the three-year-olds, a chi-square for independence reveals a significant effect for order (X2 (4, n = 69) p = 0.021). The three-year-old children produce significantly more 1-2 than 2-1 orders within this condition. This is similar to the findings in Meyer (202-12) in which a significant preference for the 1-2 order is established. However, the compared results do show a significant effect for group (X2 (4, n = 219) p < 0.001) in this condition. Thus, the results for the three-year-old children do significantly differ from the four-to-six-year-old children. This could be due to the fact that the younger children produce far less ‘correct’ or ‘reverse order’ responses, and relatively more erroneous responses than the older children did. This shows
28
that again, the three-year-old children produced far less verb clusters than the four-to-six-year-old children.
The last condition is the modal/infinitive construction within the non-finite condition. Sentences of this type are te kunnen zwemmen (‘to be able to swim’, 1-2 order) and zwemmen
te kunnen (2-1 order). The table below shows the responses within this condition.
MOD/INF non-finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 5 0 0 16 13 34
% 14.7% 0% 0% 47.1% 38.2%
2-1 Freq. 6 1 0 11 17 35
% 17.1% 2.9% 0% 31.4% 48.6%
Table 15: Distribution of response type for the three-year-olds.
MOD/INF non-finite Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 34 2 0 24 15 75
% 45.3% 2.7% 0% 32% 20%
2-1 Freq. 24 6 0 22 23 75
% 32% 8% 0% 29.3% 30.7%
Table 16: Distribution of response types for the four-to-six-year-olds in Meyer (2012).
A chi-square for independence reveals no significant effect for order (X2 (3, n = 69) p = 0.469) for the three-year-old children. This differs from the results for the four-to-six-year-old children, which reveals a significant preference for the 1-2 order (Meyer 2012). Furthermore, the three-year-olds again produce far less non-finite verb clusters than the older children. The compared results reveal a significant effect for group (X2 (3, n = 219) p = 0.001), indicating that the three-year-old children score significantly different from the four-to-six-year-old children.
In sum, the results for the three-year-olds analysed separately per condition, reveal that in the non-finite conditions, the children score similar on the auxiliary/participle and the modal/infinitive constructions. In both non-finite conditions, the three-year-old children hardly produce any verb clusters. This is different compared to the results for the
four-to-six-29
year-old children in Meyer (2012). The older group produces non-finite verb clusters a lot more often. As for the order preferences, the results for the three-year-old children do not reveal a significant effect of order for neither the auxiliary/participle constructions, nor for the modal/infinitive constructions. Thus, the three-year-olds do not exhibit a clear preference for either the 1-2, or the 2-1 order within the non-finite conditions.
If we look only within the finite conditions, the three-year-old children do produce verb clusters, but still they make a lot more errors than the four-to-six-year-olds in Meyer (2012). Furthermore, in the finite conditions, the results for the auxiliary/participle constructions on the one hand and the modal/infinitive constructions on the other hand, show somewhat different patterns. Only for the modal/infinitive constructions a significant effect for order is found; the three-year-old children show a preference for the 1-2 order. Within the auxiliary/participle constructions on the other hand, children do not seem to have a preference for either of the two orders. This is different from the results reported in Meyer (2012), which revealed that four-to-six-year-old children have a preference for the 1-2 order in the auxiliary/participle condition. This preference however, is not present in the three-year-olds, as they score almost the same on the 1-2 and the 2-1 ordered stimuli. This issue is discussed in more detail in section 5. Below I first turn to the results of the second sentence repetition task with the four-to-six-year-old children.
4.2 Second Sentence Repetition Task with the four-to-six-year-old children
The second sentence repetition task investigates the acquisition of verb clusters by looking at order preferences of four-to-six-year-old children in bipartite verb clusters.
For this purpose, the order preferences of four-to-six-year-old children were established in bipartite verb clusters with four different verbs. The verb clusters that are tested here can be subdivided into two groups; clusters containing a modal and an infinitive, and clusters containing an auxiliary and a participle. Each group consisted of two individual verbs. In the auxiliary condition, the verbs gaan (‘to go’) and zijn (‘to be’) are tested. In the modal condition, the verbs kunnen (‘to be able to’ or ‘can’) and moeten (‘to have to’ or ‘must’) are tested. Below I first discuss the results for the adult control group, before turning to the results of the four-to-six-year-old children.
A total of ten adults participated in the present experiment. Six took the test on verb clusters, the remaining 4 participated in its collaborative study on verbal inflection. This way, all the versions were checked for unexpected things that could have influenced the results. A total of 192 responses was obtained. They were divided over three categories: ‘correct’,
30
‘reverse order’, and ‘error’. Of all the responses, only one was coded as error, in this particular case, the participant repeated only one of the verbs offered in the stimulus. The remaining 191 were all correct repetitions of the stimulus. It thus seems safe to say that the test did not cause difficulties for the adult participants.
A total of 40 children participated in this experiment, 15 took the test on verb clusters, the remaining 15 participated in its collaborative study on verbal inflection (Buijs 2013). This resulted in a total of 720 responses. The responses were divided into five categories: ‘correct’, ‘reverse order’, ‘doubling error’, ‘only one verb’ and ‘other’. These categories are the same as was the case in the analysis for the three-year-olds (see the previous section). The distribution of the different response types is given in Table 17 below.
Response Type
Correct Reverse order Doubling error Only one verb Other Total
Frequency 480 93 14 21 112 720
Percentage 66.7% 12.9% 1.9% 2.9% 15.6%
Table 17: Distribution of response types for the four-to-six-year-old children.
The distribution of responses shows that overall, the four-to-six-year-old children are much better at producing verb clusters than the three-year-olds. Figure 2 below shows the difference between the older and the younger children that participated in the present study.
Figure 2: Responses of three-year-olds and four-to-six-year-olds compared.
0% 10% 20% 30% 40% 50% 60% 70% 80%
Correct Order Reverse Order Doubling error Only one verb Other error Three-year-olds Four-to-six-year-olds
31
For the four-to-six-year-olds, only 15.6% of the cases was coded as error, as opposed to 34.8% in the younger children (recall Table 7). Furthermore, where the three-year-olds produced only one verb in 31.5% of the cases, the four to six-year-olds produce just one verb in only 2.9% of the cases. Thus, compared to the younger children, the older children are already much better at producing verb clusters. For the remaining responses, 1.9% consisted of doubling errors. The correct order was produced in 66.7% of the cases, the reversed order in 12.9% of the cases.
First, all the answers are analysed together. A first chi-square revealed a significant effect for order (X2 (4, n = 720) p <0.001). The corresponding table, with the distribution of responses within the 1-2 and the 2-1 stimuli, is given below:
Response Type
Stimulus order Correct Reverse order Doubling error Only one verb Other Total
1-2 Freq. 291 5 0 7 57 360
% 80.8% 1.4% 0% 1.9% 15.8%
2-1 Freq. 189 88 14 14 55 360
% 52.5% 24.4% 3.9% 3.9% 15.3%
Table 18: Distribution of response type per order.
Overall, the four-to-six-year-old children show a preference for the 1-2 order, as they produce more correct answers for stimuli with a 1-2 order (80.8%), than for stimuli with a 2-1 order (52.5%). Moreover, they produce a lot more reverse orders in the 2-1 order (24.4%), than in the 1-2 order (1.4%).
In order to detect possible different patterns of order preferences between older and younger children, we want to see how the children perform on different types of verb clusters. First of all, the difference between verb types (auxiliary/participle constructions vs. modal/infinitive constructions) turned out not to be significant (X2 (4, n = 720) p = 0.460). The responses per verb type are given in the table below.
