Acquisition of Dutch Aspect

(1)

1

Acquisition of Dutch Aspect

Lexical and grammatical aspect in first language acquisition

Name: Judith van Dijk

Student number: 1319361

Supervisor: Dr. A.M.H. van Hout

Second reader: Dr. W.M. Lowie

Date of completion: 22-05-2010

(2)

2

Acknowledgements

Firstly, I would like to thank Angeliek van Hout for her invaluable guidance, advice, and support. This is the second thesis I have written under her supervision, and I have done so gladly. Her enthousiasm for first language acquisition is undeniably contagious. I thank Wander Lowie for helping Angeliek and me out on the statistics. And more generally, for explaining statistics in such a way that I actually understand it, at a time when I had given up on ever learning what a standard deviation is. I also thank Angeliek and Wander for giving excellent courses in linguistics in the first year of the Bachelor of English Language and Culture, which sparked my interest and led to a Master’s education in this fascinating field. Also, I thank Marjolijn Verspoor and Kees de Bot of the Department of Applied Linguistics for their excellent classes, given with enthousiasm and insight.

Lastly, I would like to thank Liselotte van Velzen for the support she has given me throughout my entire studies, and particularly in writing this thesis. Giving a whole new meaning to the term ‘tentamenstress’, I didn’t make it easy on her, but I’m glad she saw it through.

(3)

3

Abstract

(5)

5

Introduction

(6)

6

1. Theory of aspect

In this chapter, I will explain what grammatical aspect is, and how it works in Dutch and English. Also, I will show what role inherent lexical aspect plays in this study.

Grammatical aspect is marked explicitly by linguistic devices, usually auxiliaries or verbal inflections. It characterises ‘different ways of viewing the internal temporal constituency of a situation’ (Comrie 1976:3). In other words, it distinguishes different ways of describing the speaker’s temporal perspective on an event. We can see these different ways of viewing a situation when we compare sentences such as The man ate an apple and The man was

eating an apple. Both sentences are in the past tense, but there is an important aspectual

difference. The use of the simple past tense in the first sentence reflects perfective aspect and indicates that the apple was completely eaten at the end of the activity, while in the second sentence, the past progressive form, which carries imperfective aspect, has no such completion entailment. Comrie argues that the first sentence views the situation in its entirety (external perspective) while the second views the situation as consisting of phases (internal perspective). These two perspectives are present in many languages, including Dutch. Before we take a closer look at grammatical aspect in Dutch, we need to know more about another type of aspect; inherent lexical aspect.

Inherent lexical aspect or aktionsart is an inherent feature of verbs and is determined by the action described by the verb. Every verb has its own inherent aspectual characteristics which describe the situation. For example, the verb love is stative, while play is active. To map these inherent semantic distinctions into a system of situation types, Vendler (1967) proposed the following system:

“ACHIEVEMENT: that which takes place instantaneously, and is reducible to a single point in time (e.g. recognise, die, etc.).

ACCOMPLISHMENT: that which has some duration, but has a single clear inherent endpoint (e.g. run a mile, make a chair, etc.).

(7)

7

STATE: that which has no dynamics, and continues without additional effort or energy

being applied (e.g. believe, live, etc.).”

Vendler (1967:97-121)

This study uses both achievement and accomplishment verbs in the experimental design. To determine what the exact difference is between those two types of verbs, we need a more detailed verb analysis. Smith (1991:30) proposed a feature analysis for Vendler’s system, using the features [static], [durative] and [telic]. ‘Static’ applies to verbs that describe a state. ‘Durative’ applies to verbs which describe a situation or process which lasts for a period of time. ‘Telic’ verbs involve a culmination point; the action progresses towards a

natural completion or transition point. Smith classified the four verb categories as follows1:

(1) Situations Static Durative Telic

State [+] [+] n.a.

Activity [-] [+] [+]

Accomplishment [-] [+] [+]

Achievement [-] [-] [+]

We can see here that both accomplishment and achievement verbs are [-static] and [+telic]. This means that neither describes a state, while both indicate a culmination point. The difference between the two lies in the [durative] feature; while accomplishment verbs indicate a process that takes a certain amount of time, the processes described by achievement verbs take up virtually no time. We can see this in the following sentences: in

John eats an apple, the process of eating an apple takes up some time. In John wins the race,

the process of winning is instantaneous. Looking at these two sentences, we also distinguish another difference. If John were to stop halfway through eating the apple, he would still have eaten some apple. In the second sentence, however, John cannot stop halfway during the process of winning; he either wins or he does not.

1

(8)

8

The four types of verbs are each used differently in combination with grammatical aspect markers. Stative verbs are generally never used with progressive forms, while activity verbs can be used with every verb form (e.g. simple past, present perfect, past progressive). For accomplishment and achievement verbs, this is also the case. However, there are some important distinctions to be made. Let us have a look at the following three sentences in the simple past.

(2). A. Activity: John ran.

b. Accomplishment: John walked to school.

c. Achievement: John reached the top.

If in these situations, John were to stop halfway during the process, the simple past or the perfective form John has run in case (2a) would be correct, but John has walked to school or

John has reached the top for cases (2b) and (2c) would be incorrect, as accomplishments and

achievements need a culmination point. We can see this in the meanings of progressive forms as well. The progressives of the three verb categories are listed in (3).

(3) a. Activity: John was running.

b. Accomplishment: John was walking to school. c. Achievement: John was reaching the top.

(9)

9

does not imply the coming about of the result state, in spite of the fact that the sentence involves an accomplishment verb, which should be bounded. Instead, the progressive form takes an insider viewpoint, which focuses on a narrow temporal interval which falls inside the event. Thus, it excludes its endpoints from view and consequently makes no assertion about whether the event was completed. The Imperfective Paradox has been widely discussed in literature. DeClerck (1979), for example, notes that it may not be such a paradox at all. He claims that Dowty is incorrect in assuming that progressive sentences are no less bounded than their non-progressive counterparts. He states that “the presence of an imperfective progressive form appears to be one of the many factors that can render a proposition unbounded” (270). In other words, the grammatical aspect overrides lexical aspect. Gonzalez (2003) develops the idea that the Imperfective Paradox relies heavily on lexical aspect being a property of the verb, when it is rather a property of a verb phrase. “Nowadays, it has become clear that it is the verb and its arguments that provide the aspectual information about the aspectual nature of a predication.”(4, emphasis added) In other words, in sentence (3b) for example, it is not the verb alone, but rather the verb phrase, that would receive a bounded reading in perfective sentences, and is assumed to be unbounded through the progressive form.

In Dutch, there are three forms that have past-time reference; the simple past, the present perfect and the past progressive. These forms are aspectually different from their English counterparts. The different aspectual properties of the Dutch past tenses are outlined in (3).

(3) a. Tom bouw-de een huis. Onvoltooid Tegenwoordige Tijd

Tom built(PAST) a house ‘Simple Past’

b. Tom heeft een huis ge-bouw-d. Voltooid Tegenwoordige Tijd

Tom has a house built (PAST PARTICIPLE) ‘Present Perfect’

c. Tom was een huis aan het bouw-en. Aan-het-construct

Tom was a house on the build(PRESENT PARTICIPLE) ‘Periphrastic Past Progressive’

(10)

10

(4) a. Tom bouw-de een huis. Kijk, het is af.

Tom built(PAST) a house. Look, it is off.

‘Tom built a house. Look, it is finished.’

b. Tom bouw-de een huis. Hij maak-te hem niet af.

Tom built(PAST) a house. He made(PAST) him not off.

*‘Tom built a house. He did not finish it.’

While sentence (4b) is acceptable in Dutch, it is not so in English, and would rather be translated with Tom was building a house. He did not finish it. We see in (4a) and (4b) that in Dutch, the perfective and imperfective readings are both possible for the simple past. Van Hout (2005), however, argues that the Dutch simple past is not aspectually neutral, but rather essentially imperfective. She tested fifteen Dutch adults on their interpretation of Dutch past tenses, using a picture-selecting task. She found that with the present perfect, only completed situations were chosen, while the past progressive was only combined with ongoing situations. The simple past tense was accepted for both completed and ongoing situations. When ongoing situations were an option, subjects opted for those; when they were not, subjects opted for the complete situation. Van Hout concluded that the simple past is essentially imperfective, and may alternatively get a perfective reading, if imperfective is not an option. Thus, this study did not support an analysis that claims that the simple past is essentially ambiguous. To avoid this newly addressed problem of the (im)perfectivity of the simple past, the present study used only the present perfect and the past progressive in the comprehension task.

(11)

11

combinations of lexical elements with grammatical or lexical elements.” (3) She tested this using an acceptability judgment task with three age groups; 14-18 year-olds, 20-30 year-olds, and ages 50 and above, and asked the participants to choose between a sentence with a simple form and one including an aan het-construction in different contexts and for different situation types. Also, the participants were asked to grade the answer they did not choose in terms of acceptability. She found that the younger age groups chose and accepted the aan

het-construction significantly more often than the older group. She ascribes this age effect to

the process of grammaticalisation of the aan het-construction.

(12)

12

2. Prior acquisition research

Grammatical aspect has been researched for many different languages and age groups, and with a broad range of methods. Hodgson (2003) summarises the reason for this as follows:

The interaction of grammatical aspect and lexical aspect has been the center of much interest in the field of language acquisition—both, first and second language acquisition—because it stands at the interface between the lexicon and the grammar. Therefore, it provides researchers with information in the manner which semantic meaning assists on the acquisition of grammatical categories. The ability to distinguish between the aspectual classes and to talk about time is of vital importance.

(2003:2)

So far, most studies have focused on production data. This chapter presents an overview of the relevant prior research on this subject, including the few comprehension experiments that have been carried out. Using the findings of previous studies, we can hypothesise on the outcomes of this study. The hypotheses will be listed at the end of this chapter.

(13)

13

study the difference between achievement and accomplishment verbs. However, Chilton (2007) argued that as achievement verbs are more punctual than accomplishment verbs, they are less likely to be used with progressives. Thus, we can assume from the theory on these verb types that in production, speakers will use imperfect or progressive forms more with accomplishment verbs than with achievement verbs.

These findings for activity and achievement verbs are in line with the Imperfective Paradox; there is an extra difficulty of assigning a progressive, incomplete reading to verbs that inherently focus on the completion of the event; achievement and accomplishment verbs. However, not much research has been done on the Imperfective Paradox from the perspective of language acquisition. Kazanina and Phillips (2007) also noted this and studied the role of the Imperfective Paradox in children’s acquisition of grammatical aspect in Russian. They tested 3- to 6-year-olds using four comprehension experiments. The first experiment was a truth-value judgment task, in which stories containing complete and incomplete events were acted out in front of the child, which made it possible to have events that were clearly in the past and permanently incomplete. A hand puppet then made a statement about what had happened in the story, using either a perfective or imperfective verb; e.g. At the castle the monkey built / was building a smurf. The children had to respond to the puppet’s statement. Kazanina and Phillips found that most children gave adultlike responses for the perfective sentences, but had difficulty matching imperfectives to incomplete events. They explained this as a result of their inability to find “a suitable temporal interval against which to evaluate imperfective statements.” (68) In a second truth-value judgment task, they asked children to judge perfective and imperfective sentences that contained a while-clause as an explicit temporal modifier; e.g. While the boy was

watering the flowers, the girl cleaned (all of) / was cleaning the table. With this experiment,

they found that children were only willing to accept imperfectives as describing incomplete events in the past if the sentence contained an explicit temporal modifier, such as a when- or

while-clause. The interval of such a temporal modifier coincides with the internal perspective

(14)

14

One of the first Dutch studies that reported on the acquisition of Dutch aspect was conducted by Bol & Kuiken (1988). They analysed the spontaneous speech of Dutch children, and found that the perfect participle, with or without an auxiliary, develops by age 2;0, while the imperfect past develops later on. The question remained what would happen with comprehension; Bol & Kuiken wondered if the tenses and aspects they studied were also understood target-like. This question was picked up by Van der Feest & Van Hout (2002), who tested three-year-old children as well as adults on three tenses: simple past, present perfect, and simple present. They used a truth-value judgement task, taken from Wagner (2001), to test the three tenses for three types of events; past-time completed, past-time incomplete and ongoing, rendering a three-by-three design. The test involved 18 items; 2 items per cell in the design. They used a puppet, which performed a certain action (e.g. doing a puzzle) at three locations on a road drawn on a table. As the puppet moved along, the actions at the first two locations became events in the past. The puppet finished only one of the first two actions. A blindfolded second puppet (played by a second experimenter) was asked questions about the events at the moment the first puppet was at the third location (so, the third action was going on at speech time). For example, the experimenter asked the second puppet: “where is the girl making a puzzle?”. The puppet would respond with one of the three locations. The participants had to judge whether the blindfolded puppet’s guesses were correct or incorrect.

(15)

15

Feest and Van Hout suggested the use of a ‘when’-clause to function as a temporal antecedent in future research. Such a clause would help clarify if the simple past is indeed aspectually neutral, as a sentence with a simple past could include ‘but she did not finish it’. This incompletion would be impossible in a sentence with a past or present perfective form. When-clauses are included in the experimental design of the present study, as described in chapter 3.

Van Hout (2006) also tested the comprehension of perfectivity in young children. The results of her study contradicted those of Van der Feest & Van Hout (2002). Using a picture-selection task, she tested comprehension of the perfective and imperfective past. The experiment consisted of short stories combined with three pictures each. The experimenter told a story which described the first picture. The second picture of every story showed closed curtains, so the child did not know what would happen next. A hand puppet was used to look behind the curtains, and describe the scene using the test sentence; either perfective or imperfective. For every story, the last picture was not there yet; rather, it was an empty slot in which the final picture, described by the hand puppet, could be inserted. The child’s task was to select the right picture, based on the test sentence. Van Hout found that Dutch three-year-olds know the completion entailment of the perfective. However, they chose the wrong picture (i.e. the completed situation) for imperfectives, which showed that they incorrectly associated imperfectives with completion.

Not much research has been done on the acquisition of aspect or tense in bilingual children. Bozinou & Santiago (1984) tested English monolingual and English-Greek bilingual 5-year-olds on the present progressive and past tenses under varying cue conditions. They tested the children’s language production in picture-describing and picture-selecting tasks. For every verb, there were three visual cue cards; one depicting the activity in progress, the second depicting the activity completed, and the third depicted an unrelated event; the ‘odd’-stimulus. For the production task, the experimenter asked the participant to describe what was on a card. For the comprehension task, the experimenter read out a sentence using either the present progressive form or the simple past, and the participant had to indicate which picture combined with that sentence. Half of the participants were given verbal cues such as right now and already in both the production (tell me what Johnny did

(16)

16

Santiago found that the monolingual children had a better performance on the verbal cue condition than the bilingual children, especially in the past tense. Furthermore, they found that bilingual 5-year-olds made more errors overall than their monolingual peers, and improved only in their comprehension on the verbal cue condition. They concluded that the complex interdependence of structural complexity, context, and level of abstraction of the concept (past vs. present) determines performance. Bilingual children have two separate linguistic systems to acquire. The present study examines the acquisition of aspect in Dutch-English bilingual children, whose two languages differ in terms of aspectual system, but not much. The tense that is most different between the two languages is the simple past, and we expect the bilingual children to have particular difficulty with the production of this tense. Thus, we may expect the monolingual children to perform more adult-like than bilingual children in the present study.

For this study, five hypotheses are formed. These have been formulated based on the aspectual theory and previous research as described above. The first three hypotheses concern the target aspectual grammar of Dutch, and thus focus on adult grammar. The fourth and fifth hypotheses concern the main research question of this study; how do Dutch 4-, 5-, and 6-year-olds acquire aspect? The last hypothesis focuses on the acquisition of aspect in bilingual children. The hypotheses of this study are the following:

(A) Dutch adults expect perfectives to entail event completion, and expect imperfectives to have no such completion entailment. Thus, they

(I) accept perfectives only for completed situations (II) use perfectives only for completed situations

(III) accept imperfectives for completed and incomplete situations (IV) use imperfectives for incomplete situations

(17)

17

event, as an imperfective form leaves the possibility of the event not being completed open. For incomplete situations, only imperfect verb forms are correct, so we expect the adults to accept and use only imperfect verb forms here.

(B) Speakers use progressive forms more with accomplishment verbs than with achievement verbs.

Previous studies by e.g. Shirai & Andersen (1995) and Bloom, Lifter & Hafitz (1980) have shown that progressives are generally used with atelic rather than telic verbs, as progressives focus on the duration of the event. Chilton (2007) argued that as achievement verbs focus on the culmination point of the event, while accomplishment verbs focus on the duration of the event, achievement verbs are less likely to be used with progressives. Thus, we can assume that in production, speakers will use progressive forms more with accomplishment verbs than with achievement verbs.

(C) Dutch adults treat the simple past as aspectually neutral and thus use it for completed and incomplete events alike.

As argued by Boogaart (1999), Van der Feest & Van Hout (2002), and Van Hout (2007), we hypothesise that Dutch adults will use the simple past as an aspectually neutral form. Therefore, we expect them to use the simple past for completed as well as incomplete events.

(D) 4- to 6-year-old Dutch children use both perfect and imperfect verbs, although they do not fully grasp the aspectual features of these forms.

(18)

18

This study compares the children’s performance on the comprehension task with their performance on the production task to determine if they use aspectual forms of which they have not yet fully acquired the meaning.

(E) The development of aspect correlates with general linguistic development.

As aspect is one of many facets of general linguistic development, we expect the acquisition of aspect to go hand in hand with the acquisition of other linguistic skills, such as phonological awareness and the use of articles. These skills are all tested in the Grammar and Phonology Screening. Therefore, we expect the children’s scores on the aspect experiment to correlate with their scores on the GAPS test.

(F) Dutch-English bilingual children show a different development of aspect in their languages compared to monolingual children.

(19)

19

3. Method

3.1 COST aspect experiment

The Aspect test used in this study was developed as part of Action A33 of the European

Cooperation in Science and Technology (COST) network2; “Crosslinguistically Robust Stages

of Children’s Linguistic Performance, with Applications to the Diagnosis of Specific Language Impairment” (Gagarina, van Hout & Dressler, in prep.; Van Hout, Gagarina & Dressler, in prep.). The goal of this COST Action was to identify and describe areas of cross-linguistic uniformity that can then be used for cross-linguistic tests of language impairment. The project focuses on semantic and pragmatic development in first language acquisition. Grammatical aspect is one of the points of focus, and the COST Aspect comprehension and production experiment was designed to investigate this across a number of European languages.

The COST aspect experiment consisted of alternating production and a comprehension tests. The participant watched movies in which a clown did something while music was playing. When the music stopped, the clown had to stop as well. Every movie was about one verb and action, and contained six items (four for comprehension, two for production) with which the clown performed the action. For example, the movie about ‘opening’ contained a jar, a bottle, a toy box, a box, etc. There were six verbs used, all transitive, with telic predicates. Thus, we were able to test the relation between completedness and perfectivity. Three of the verbs were accomplishment verbs (make, draw, build) and three were achievement verbs (open, close, blow out). We used both achievement and accomplishment verbs to examine the influence of lexical aspect on grammatical aspect. Before the experimental items were presented, there was a training session of 3 items during which subjects were made familiar with the procedure and the task. While-clauses were included in both comprehension and production to function as a temporal antecedent. In comprehension as well as production, a while-clause clarifies the temporal window (the playing of the music) on which we focus. In production, it helps to distinguish between the simple past and the present or past perfect. If the simple past is indeed aspectually neutral, this may be clarified

(20)

20

in such an anaphoric context. The implication of completion seems less strong, so that a simple past may be continued with ...maar hij heeft hem niet afgemaakt (‘...but he did not finish it’). The present and past perfect forms on the other hand are necessarily perfective, and cannot be combined with such a sentence (Van der Feest and Van Hout, 2002:11).

The comprehension and production items alternated for two reasons. Firstly, it facilitated priming effects in the production section for the verb forms used by the experimenter in the comprehension section; the clearly perfective and imperfective forms. As the use of one of these two forms was targeted in the production section, the priming effect of the comprehension section was very important. Secondly, it functioned as a way to keep the experiment interesting for the children. Alternating comprehension and production kept them active, as they did not know whether they had to judge a statement or describe the situation themselves, they paid close attention to the films played.

There were three versions of the experiment; list A, B, and C. List A and B only differed in the order of the two parts of the experiment; the experiment was administered in two sessions, so what was session one for list A was session two for list B, and vice versa. List C had the same order of sessions as list B. However, in the comprehension section of list C, all grammatical aspectual forms were switched; for example, an incomplete event was followed by a perfective sentence (to be judged by the participant) in list A and B, and by an imperfective sentence in list C, and vice versa. This was done to examine if there were any item effects. Of the Dutch adults, half was tested on list A, and half on list C. Of the 38 Dutch children, 19 were tested on list A, 9 on list B, and 10 on list C. The bilinguals were tested on list A in both languages.

(21)

21

of grammatical aspect, we used aspectual forms that were clearly defined as perfective or imperfective. The perfective verb forms needed to have clear completion entailments. The English simple past could therefore not be translated to the Dutch simple past, as the Dutch form is not perfective. The Dutch present perfect tense is, and was therefore used in this experiment.

The participants had to respond to the experimenter’s statements with yes or no, and received no feedback on their replies. The possibilities are listed in Table 1 below.

Action completed Action incomplete

“Terwijl je het muziekje hoorde, heeft de clown een huis gebouwd.”

“While the music was playing, the clown built a house.” (perfective)

Code: ComP Target: yes.

Code: IncP Target: no.

“Terwijl je het muziekje hoorde, was de clown een huis aan het bouwen.”

“While the music was playing, the clown was building a house.” (imperfective)

Code: ComI Target: yes.

Code: IncI Target: yes.

Table 1; experimental design of the comprehension test

The codes we see in Table 1 will be used throughout this paper. The first three letters stand for the completedness of the action; Com for completed, Inc for incomplete. The last letter stands for the grammatical form used by the puppet; either P (perfective) or I (imperfective). The comprehension part consisted of 24 items, six items per cell in the design. As three accomplishment verbs and three achievement verbs were used, the experiment had a total of eight cells.

(22)

22

the prompt “Can you say that another way?” or “Can you tell me something about the clown?”. If this did not lead to a target-like sentence structure, the first utterance was recorded. The reader is referred to Appendix I for a more detailed list of the comprehension and production items of the COST Aspect experiment.

3.2 GAPS test

To compare the children’s knowledge and use of grammatical aspect with their general linguistic level, a Dutch translation of the Grammar And Phonology Screening (GAPS) test was used. This is a standardised sentence and non-word repetition test, developed by Gardner, Froud, McClelland & van der Lely (2006) as “a short, reliable assessment of children’s language abilities” (513). It starts with a sentence repetition task in the form of a story. It includes a cardboard cut-out of an alien, Bik, who wants to hear the story. The participant is told that Bik cannot hear adults, only children. So, every sentence of the story has to be repeated by the participant. After that, the participant is told that the experimenter has a list of words in Bik’s language that they could say to Bik. Once again, the participant is required to repeat after the experimenter.

In the sentence repetition section, the participant was instructed to repeat a sentence which the experimenter read aloud. There were eleven experimental sentences. These included some grammatical structures such as reflexives and auxiliaries. If the child repeated the grammatical structures correctly, the sentence was scored as correct. One or more mistakes in the grammatical structures would make this item incorrect. Lexical errors, such as saying

dog instead of cat, were ignored, as well as purely phonological errors. The task started with

(23)

23

3.3 Participants

A total of 38 monolingual Dutch children were tested for this experiment3. Ages ranged from

4;10 to 6;9, with an average of 5;8. These children all went to the same primary school in

The Netherlands.4 For the COST Aspect experiment, all children were tested individually, in

two sessions, which were a week apart at most. If the sessions were on different days, the training items were repeated at the start of the second session. Fifteen children were also tested on the GAPS experiment, which was administered in one session. For all participants, consent was obtained from a parent or guardian. Another six bilingual children were tested, ages 5;6 to 6;9, mean age 6;2. They had all been raised bilingually, with Dutch and English. These six children were tested in both Dutch and English (on separate occasions). They also did the GAPS test in both languages.

As a control group, ten adult native speakers of Dutch were tested on the COST Aspect experiment. They had no background in linguistics. Five of them were tested on List A, and five on List B. The testing of adults could be done in one session, and with the researcher making propositions for the participants to respond to rather than with the hand puppet ‘guessing’.

All testing, including the GAPS tests, was hand-scored on-line by the experimenter. Also, audio recordings were made of every session, to be used in case of doubt about the utterances of the participants.

3

I thank Angeliek van Hout and Margreet van Koert for their work on collecting these data.

(24)

24

4. Results

Our main focus is the comparison of the Dutch children’s scores to those of the adults. To make an SPSS analysis on the COST results possible, the comprehension data are taken together per category. As described in section 3.1, there are four possible combinations of grammatical aspect and (in)completedness; ComP (complete – perfect), ComI (complete – imperfect), IncP (incomplete – perfect), and IncI (incomplete – imperfect). These distinctions are used in a GLM Repeated Measures Analysis of Variance using Aspect (i.e. perfect vs. Imperfect) and Situation (i.e. completed vs. Incomplete) as within-subject factors. Further on in the analysis, we include Verb class (i.e. achievement vs. accomplishment) as another within-subject factor. This renders an eight-cell framework with 3 items per cell. The participants responded with yes or no, which are scored as 1 or 0 respectively. The between-subject factor is age group.

For the production data, the utterances of the participants are identified by the form of the target verb. These utterances are first categorised and scored per category. In Dutch, there are five categories, as listed in Table 2. We see here that there are two perfective tenses, two imperfective tenses, and one ambiguous tense; the simple past. In the analysis, the present and past perfect are collapsed and labelled ‘perfective’. The counts on the periphrastic progressive and the ging Ven-construction are taken together and labelled ‘imperfective’. The ging (or gaan; present tense) Ven-construction is a Dutch verbal construction of which the aspectual features have not been thoroughly analysed in the literature yet. Booij (2002) characterises the gaan Ven-construction as an inchoative use of the verb gaan (to go) in combination with a verbal infinitive. The gaan Ven-construction does not focus on the duration of an event as does the progressive. Rather, it “only expresses that an event will begin” (9). Despite this subtle semantic difference as Booij calls it, the periphrastic past progressive and the ging-Ven construction are both imperfective in nature, and therefore form one aspectual class together.

(25)

25

Tense Form Aspectual class

Present perfect Heeft gebouwd Perfective

Past perfect Had gebouwd Perfective

Periphrastic progressive Was aan het bouwen Imperfective

Ging Ven-construction Ging bouwen Imperfective

Simple past Bouwde Ambiguous

Table 2. Categories of Dutch production.

In English, there are 2 categories, as listed in Table 3. Here, the categorising is much more straightforward; all the utterances containing a simple past on the main verb are scored as perfective, and all past progressives are scored as imperfective.

Tense Form Aspectual class

Simple past Built Perfective

Past progressive Was building Imperfective

Table 3. Categories of English production.

For both Dutch and English, all utterances that do not fall into one of the categories described above, such as het lukte/it worked and het was af/it was finished are categorised as ‘other’.

(26)

26

of those in list A and B, while the children in list C are even younger, with a mean age of 62.7 months. Furtheremore, all other scores, including the GAPS scores, were stable across the three lists. We therefore conclude that the COST experiment has no item effects which interfere with its validity, and we collapse the three groups of Dutch children into one large group.

Using the data scored as described above, we try to find evidence for the hypotheses stated in chapter 2. Below, the relevant results for these hypotheses are listed. These will be discussed in depth in chapter 5.

(I) accept perfectives only for completed situations (II) use perfectives only for completed situations

(III) accept imperfectives for completed and incomplete situations (IV) use imperfectives for incomplete situations

(27)

27 Figure 1; Comprehension scores Dutch adults

In Figure 1, we notice a similar effect for the imperfect forms in incomplete situations. We expected yes-answers here, as we did for the complete situations. Yet in a few cases the participants judged imperfect forms of achievement verbs as incorrect for incomplete situations. For example, they judged De clown was de kaars aan het uitblazen (The clown

was blowing out the candle) as incorrect if the clown had not succeeded in blowing out the

candle. However, this difference was not found to be significant; t(9) = 1.500, p = .168.

Figure 2; Production scores Dutch adults 0 0,2 0,4 0,6 0,8 1 1,2 0 5 10 15 20 25 30 35 acco m p lis h m en t ach ie ve m en t acco m p lis h m en t ach ie ve m en t Complete Incomplete other Vde

(28)

28

Figure 2 shows the production scores of the Dutch adults. We see here that adults use perfectives (heeft geVd) and the ambiguous simple past to describe complete events, both with accomplishment and achievement verbs. For incomplete achievements, they use the periphrastic progressive (was aan het Ven) as well as a number of other constructions, e.g.

hij was klaar (he had finished). For accomplishment verbs, they use the periphrastic

progressive most, followed by the simple past (Vde), other constructions, and the ging V-en construction. They never use perfective forms to describe incomplete events.

The comprehension and production results of the Dutch adults show similar results. The participants do not accept perfective forms for incomplete events in the comprehension test, nor do they use these forms to describe incomplete events. For complete events, they accept imperfective forms as grammatical, but prefer to use perfective forms in production.

The Dutch progressive is formed with the periphrastic aan het-construction. As we can see in Figure 2, this construction is used by Dutch adults in incomplete situations for both achievement and accomplishment verbs. Although we expected to find more uses of the progressive for accomplishment verbs than for achievement verbs, an ANOVA analysis found no significant effect of verb class. This hypothesis is therefore not confirmed by the results of this study. This may be because for the incomplete achievement condition, the participants used many constructions coded as ‘other’ (e.g. It didn’t work, or he didn’t finish in time). An ANOVA analysis showed that the participants used ‘other’ constructions significantly more often for incomplete accomplishments than for incomplete achievements; F(1,18) = 27.000,

p = .000. A qualitative analysis of the adults’ production data shows that most constructions

(29)

29

(C) Dutch adults treat the simple past as aspectually neutral and thus use it for complete and incomplete events alike.

If we look at Figure 2, we see that the Dutch adults often use the simple past for completed events, of both accomplishment and achievement verbs. For incomplete events, they use the simple past only for accomplishment verbs. A GLM Repeated Measures analysis showed a significant effect for situation (completed – incomplete) of F(1,9) = 17.686, p = .002. Adults use the simple past more to describe completed events than to describe incomplete events. We also see in Figure 2 that there is a small difference between completed accomplishments and achievements; the simple past is used slightly more often to describe completed achievements. However, this difference was not found significant. On the other hand, an ANOVA analysis revealed a significant effect of verb class between incomplete accomplishments and incomplete achievements; F(1,18) = 9.000, p = .008. Adults use the simple past more often to describe incomplete accomplishments than to describe incomplete achievements.

(30)

30 Figure 3; comprehension scores Dutch children and adults

We see in Figure 3 that the children are adult-like in their responses for the completed situations, but not entirely so in their responses for the incomplete situations. To determine the reason for this difference, we analysed the comprehension data further and included the lexical aspect verb classes.

Figure 4; comprehension scores Dutch children and adults by verb class

As we see in Figure 4, there are no notable differences between the scores of children and adults for completed actions. For the incomplete actions, we see that the children’s scores

0 0,2 0,4 0,6 0,8 1 1,2

ComP ComI IncP IncI

(31)

31

are less clear-cut than those of the adults. Using an ANOVA analysis, this difference was found significant for the IncP with accomplishment verbs, F(1,46) = 13.965, p = .001. For the achievement verbs in the IncP and IncI situations, the difference was also near-significant. For the IncP this was F(1,46) = 3.784, p = .058, and for the IncI this was F(1,46) = 3.308, p = .065.

In production, we can observe differences between the scores of the children and those of the adults. Figure 5 shows the production scores of the Dutch children. We can see that the children use a wider range of inflections than the Dutch adults. Overall, they use non-targeted constructions more often than the adults. Especially in incomplete situations, they use ‘other’ constructions most of the time, such as het lukte (it worked) and hij was klaar (he

was finished).

Figure 5; production scores Dutch children

The children use as many constructions as the adults did, but they use them in more contexts. The ging Ven-construction for example, is used by adults only for incomplete accomplishments. Children use this construction in all four situations. To analyse the children’s production, we collapsed the verb forms into three categories; perfective, imperfective, and ambiguous, as described at the beginning of this chapter. Using a GLM Repeated Measures analysis, we looked at the use of these three aspectual classes in the four conditions. 0 20 40 60 80 100 120 accom p lis h m en t ach ie ve m en t acco m p lis h m en t ach ie ve m en t Complete Incomplete other Vde

(32)

32

For all three verb classes, there was a significant effect of situation by group, as we can see in Figure 6, 7 and 8. Looking at these Figures, we must keep in mind that the scores by the children are much lower than those of the adults, because the children gave so many responses classified as ‘other’.

Figure 6; production perfective forms Dutch children and adults

(33)

33 Figure 7; production imperfective forms Dutch children and adults

Figure 7 shows the use of the imperfective forms (the past periphrastic progressive and the

ging Ven-construction) by Dutch children and adults. We see a clear difference between the

(34)

34 Figure 8; production ambiguous forms Dutch children and adults

In Figure 8, we see the production of the ambiguous simple past by children and adults. As shown in the results of hypothesis (C), the adults use the simple past almost solely in completed situations. The children on the other hand, hardly make any use of the simple past in both completed and incomplete situations. This difference was found significant at F(1,46) = 29.430, p = .000.

(35)

35 Figure 9; production ambiguous forms by Dutch adults

In Figure 9, we see that adults often use the simple past for achievement verbs in completed situations (situation 1 in the figure). For incomplete achievements (situation 2 in the figure), they hardly use the simple past. For accomplishment verbs, we see the same tendency of more use of the simple past in completed than in incomplete situations, but here, the tendency is less pronounced.

(36)

36

Figure 10 shows the production of the simple past by Dutch children. We see that they use the Simple Past differently compared to adults. In completed situations, they use the simple past in achievement and accomplishment verbs nearly alike. In incomplete situations, they hardly use the simple past in achievement verbs, but use it often for accomplishment verbs. The GLM Repeated Measures analysis shows a significant effect of Verb Class by Situation and Group; F(1,46) = 5.732, p = .021. However, we must keep in mind that the children used the perfective, imperfective and simple past tenses far less often than the adults. We will discuss the implications of this in detail in chapter 5.

The ages of the group of Dutch children ranged from 4;10 to 6;9, with an average of 5;8. We used a Chi-square analysis and an ANOVA analysis (using age groups) to correlate age to the scores on the comprehension and production experiments. No such age effects were found within the group of Dutch children. We will come back to this for hypothesis (E).

(E) The development of Aspect correlates with general linguistic development.

(37)

37

Figure 11; scores Aspect comprehension Incomplete-Perfective and GAPS

We expected a correlation between the GAPS and Aspect scores. As the GAPS targets a high score while the IncP targets a low score (i.e. no-answers), we would expect a negative correlation of IncP and GAPS. However, a Chi-square analysis revealed there was no correlation between the IncP and GAPS results whatsoever, neither when the IncP was broken down by verb class nor when it was taken as a whole.

To analyse the development of aspect in bilingual children, we compare their scores on the comprehension and production experiments to those of their monolingual peers; using

ANOVA analyses, we compare their English to the English of monolingual English children5,

and their Dutch to that of monolingual Dutch children.

5

(38)

38 Figure 12; Comprehension scores bilingual and monolingual children

Figure 12 shows the comprehension scores of the four groups of children; the monolingual Dutch and English children, and the bilinguals’ scores in Dutch and English. We see that all children are adult-like in their responses to the completed situations. In incomplete situations, their results show a tendency to the adults’ responses, but are less clear-cut. If we look at the scores in Dutch, we see that the monolingual and bilingual children have equal performances in completed situations; the ComP (completed-perfective) and ComI (completed-imperfective) condition. In the incomplete condition however, we see that the bilingual children gave more no-responses than the monolinguals. Considering that the adults’ answers were no (0 on the y-axis) on the IncP condition and yes (1 on the y-axis) on the IncI condition, we see that the bilinguals are more adult-like than monolinguals on the IncP condition, and less so on the IncI condition. With an ANOVA analysis, we see that these results are only significant for the IncP condition; F(1,42) = 5.526, p = .024. The difference on the IncI condition is not significant; F(1,42) = 2.275, p = .139.

In English, we see that monolinguals and bilinguals perform similarly in completed situations and in the IncI condition. In the IncP condition however, the bilingual children give more no-responses, and thus seem to behave more adult-like. This difference is significant; F(1,25) = 9.528, p = .005.

To analyse these results in further detail, we will break the comprehension results down by verb class. This is shown in Figure 13.

0 0,2 0,4 0,6 0,8 1 1,2

ComP ComI IncP IncI

(39)

39

Figure 13; Comprehension scores bilingual and monolingual children by verb class

In Figure 13, we see that the difference between monolinguals and bilinguals in Dutch is stable across verb classes; the bilinguals give more no-responses than monolinguals for both accomplishment and achievement verbs in the IncP and IncI conditions. In English, the same is true; for both accomplishment and achievement verbs, bilinguals give more no-responses in the IncP condition. However, due to the small number of bilingual participants, these results were only significant for the IncP accomplishment conditions in both languages; in Dutch, at F(1,42) = 7.406, p = .009, and in English, at F(1,25) = 6.641, p = .016.

In conclusion, we observe that whereas the results for the completed situations are similar between monolingual and bilingual children, we see that for all incomplete situations and in both languages, the bilingual children have more no-responses than their monolingual peers. We will come back to this in chapter 5.

Next, we look at the scores on the production experiment. We compare the bilinguals’ scores in Dutch with those of Dutch monolingual children, and their scores in English with those of their English peers. We have already seen the Dutch children’s production scores in Figure 5. We will compare these to the bilinguals’ scores in Dutch, shown in Figure 14.

(40)

40 Figure 14; Dutch production scores bilinguals

If we compare the Dutch children’s scores in Figure 5 to those of the bilinguals in Figure 14, we see that in general, they have similar results. In both groups, there are many responses scores as ‘other’, which leaves only a small amount of responses for analysis. We see in Figure 14 that the bilingual children make no use at all of the simple past, neither in completed nor in incomplete situations. Also, the bilinguals use the ging Ven-construction less often than monolinguals. These observations will be further discussed in chapter 5.

The last part of the analysis is to compare the English production of the bilingual children to that of monolingual English children. To analyse the results of these two groups, we first need to know what the target performance is; the production of English adults. This is plotted in Figure 15. In Figures 16 and 17, we have plotted the results of the bilingual and monolingual English children.

0 2 4 6 8 10 12 14 16 18 20 acco m p lis h m en t ach ie ve m en t acco m p lis h m en t ach ie ve m en t Complete Incomplete other Vde

(41)

41 Figure 15; Production scores English adults

In Figure 15, we see that the production scores of English adults are very straightforward; the adults nearly always use the simple past for completed situations, and the part progressive for incomplete situations. They never use the simple past for incomplete situations. These results function as a background to the comparison of the English of bilingual children and that of their monolingual peers.

Figure 16; Production scores English children

Figure 16 shows the production scores for the English children. We see that they, like Dutch children, use many verb forms classified as ‘other’, such as It was working, He didn’t finish,

(42)

42

or He didn’t make it. We also see that the children mostly use the simple past as a perfective tense for completed situations, and the past progressive as an imperfective tense for

incomplete situations6. In contrast to the adults, the children sometimes use the simple past

to describe incomplete events.

Figure 17; English production scores bilinguals

In Figure 17, we see that the bilingual children also use the simple past in incomplete events, but only for accomplishment verbs. There, they use it more than monolingual English children. This difference was found near-significant; F(1,25) = 3.306, p = .081. This could be an influence of the aspectual features of the Dutch simple past, as will be discussed in chapter 5. We also see that in general, the bilinguals’ production scores are very similar to those of the monolinguals. The bilinguals make slightly more use of the past progressive in completed situations. However, none of these differences was found to be significant.

6

The very few instances (two, out of a total of 252 recorded responses) in which the English children responded in the past or present perfect. These were scored as perfective; and thus as simple past.

(43)

43

5. Discussion

(I) accept perfectives only for completed situations, (II) use perfectives only for completed situations,

(III) accept imperfectives for completed and incomplete situations, (IV) use imperfectives for incomplete situations.

The Dutch adults performed in both the comprehension and the production experiments as expected. In completed events, they accepted perfective and imperfective forms, but preferred to use perfective forms themselves. This is in accordance with the Gricean maxim of quantity, which states that speakers try to be as informative as possible. Dutch adults know that both perfective and imperfective forms are correct to describe complete events, but use perfectives to be more informative. For incomplete events, they only accept imperfective forms. This is also as expected, as perfective forms imply the completion of the described event. In production, they only used imperfect forms as well. Unexpectedly, we saw an effect of lexical aspect (verb type) on their judgement of perfectives in incomplete situations. Although we expected the participants to answer no in response to sentences with perfective forms in incomplete situations, they sometimes answered yes, but only for accomplishment verbs. The reason for this might be that in the case of incomplete accomplishment situations, the agent does indeed bring about a change of state in performing the action, albeit not a complete one. This is not the case for achievement verbs. In short, all four subclauses of this hypothesis are confirmed by the results of this study.

(44)

44

completed. DeClerck (1979) had a slightly different hypothesis; he argued that grammatical aspect overrides lexical aspect. We can observe the phenomenon described by Dowty and DeClerck in the comprehension results of this study. In incomplete situations, Dutch adults judge imperfective (progressive) forms as correct, even though the achievement or accomplishment verbs are telic. In the production results however, the theory applies to a lesser degree. Especially in describing incomplete accomplishments, speakers do not always use the progressive to describe the situation, even though they are ‘allowed’ to do so, judging from their own responses in the comprehension experiment. This will be further analysed in the discussion of hypothesis (B).

(45)

45

are in line with Chilton’s argument that achievement verbs have a specific focus on the culmination point of the event, making the culmination point a vital part of the lexical aspectual features of achievements. As this culmination point is not reached in incomplete situations, adults often use negations to describe the event, or lack thereof. In describing the less punctual incomplete accomplishments, this is not the case, as there has indeed been an action taking place, which facilitates using an imperfective verb form; in Dutch, the periphrastic past progressive, simple past or ging Ven-construction.

(C) Dutch adults treat the simple past as aspectually neutral and thus use it for completed and incomplete events alike.

The Dutch simple past can be used to describe both complete and incomplete events. Unlike the English simple past, it is not inherently perfective. Instead, it has been argued that it is aspectually neutral (Boogaart, 1999; Van der Feest & Van Hout, 2002; Van Hout, 2007). On the other hand, the study by Van Hout (2005) does not support an analysis that claims that the simple past is essentially ambiguous. She concluded that the simple past is essentially imperfective, and may alternatively get a perfective reading, if imperfective is not an option. The current study sought to shed light on this matter by analysing the use of the simple past by ten Dutch adults in the production test of the Aspect experiment. We found that Dutch adults use the simple past significantly more in completed than in incomplete situations. This goes against the hypothesis and the previous literature; an aspectually neutral simple past should be used for completed as well as incomplete events. On the basis of the results of only ten participants, this study does not suggest a completely new theory. Rather, we conclude that further research into the nature of the Dutch simple past is necessary. This research should include comprehension testing of the simple past.

(46)

46

action, he would still have done part of the action. This renders the simple past, whether ambiguous or essentially imperfective in nature, applicable for incomplete accomplishments.

The main goal of the present study is to examine the use of aspectual grammar by Dutch 4-, 5- and 6-year-olds. We expected the children to use perfective as well as imperfective forms, but differently, i.e. less clear-cut, than adults. Also, we expected that the children’s comprehension of aspect is not at the level of the adults yet.

Bol & Kuiken (1988) were the first to show that Dutch children use perfect forms from age 2;0, and imperfect forms later on. We therefore expected 4- to 6-year-olds to use both perfect and imperfect verbs. The results show that this is correct, the children use all verb forms which are also used by adults; perfectives (the present and past perfect), imperfectives (the periphrastic past progressive and the ging Ven-construction), and the ambiguous simple past. We observe that for adults, the use of these forms depends on the situation; they use perfectives for completed situations, and imperfectives for incomplete situations. The children use many constructions categorised as ‘other’, which included negations and other verbal constructions, such as He didn’t finish, It almost worked, or He

was trying to do it. This resulted in a small number of analysable utterances. That said, we

(47)

47

From all accounts in the discussion about the Imperfective Paradox (Dowty 1977, 1979; DeClerk, 1979; Gonzalez, 2003), we assumed that the acquisition of the aspectual features of the progressive may be difficult for children. We therefore expected them to have particular difficulty mapping the progressive form to its aspectual meaning. In this study, we found that all aspectual classes (perfective, imperfective, ambiguous) were used differently between adults and children. However, we did find that the effect between Situation and Group was strongest for the imperfective forms. These results seem to support the theory; the children had particular difficulty mapping imperfectives to their aspectual meaning. Note that the ‘imperfective’ category in this study also included the ging Ven-construction, which is not progressive in nature, but rather focuses on the beginning of the event (Booij, 2002). This hypothesis should be further tested with production studies on a larger scale.

We found that Dutch children make almost no use of the simple past. This could be explained by the large number of ‘other’ responses in the production experiment. More surprisingly, we found that the children use the simple past as often for completed as for incomplete situations, while adults make more use of it for completed than for incomplete situations. Also, we found that the children use the simple past surprisingly often for incomplete accomplishments. We can conclude that the children have not yet grasped the aspectual features of the simple past that the adults’ use of it indicates it to have. However, the adult group in the present study was fairly small (ten participants) and the production testing of the children rendered very few analysable verb forms. Therefore, more research should be done on the aspectual features of the Dutch simple past and on its production by both adults and children.

In the comprehension experiment, we found that the children performed adult-like for the completed events. In incomplete events, they differed from the adults; they had less clear-cut performances on the IncP-accomplishment condition; the use of a perfective to describe an incomplete accomplishment. In this condition, the adults predominantly responded with

no; about eighty percent of the responses was no. The children gave more yes-responses;

(48)

48

incomplete achievement. Here, about ninety percent of the adults’ responses were yes, versus seventy-five of the children’s. This difference was also near-significant.

The results show that 4- to 6-year-old children have not quite grasped the perfectivity of the Dutch present perfect tense. They incorrectly judge it as applicable in incomplete situations. Also, they did not always judge imperfectives as correct in incomplete situations, which shows that the imperfective tense, the past periphrastic progressive in this study, is not used completely adult-like either.

When we look at the results of both the comprehension and the production experiments of this study, we can conclude that Dutch 4- to 6-year-olds do not yet fully grasp the aspectual meanings of Dutch perfective and imperfective verb forms.

(E) The development of aspect correlates with general linguistic development.

Even though the group of Dutch children seemed to have a broad range of ages, no age effects were found within the group for the development of aspect. The GAPS experiment (Gardner, Froud, McClelland & van der Lely, 2006) tested general linguistic development. Contrary to expectation, no age effects were found in the GAPS scores either. This might be due to strong individual differences in the acquisition of aspect among the children. The acquisition of aspect does not seems to be linear and clear-cut enough to show a clear acquisition pattern between ages 4 and 6.

(49)

49

This study examined the acquisition of aspect in Dutch-English bilingual children. Because there is so little literature on the acquisition of aspect in bilingual children, this part of the current study was explorative in nature. We had a very small number of bilingual participants. As a result, not many observations were found statistically significant.

Dutch and English have different aspectual structures; the form-to-meaning mappings of aspect differ. In Dutch-English bilingual children, their two languages possibly influence one another. This study examined if this results in a different acquisition pattern for both languages, compared to monolinguals.

In comprehension, we saw that the results for the completed situations were similar between monolingual and bilingual children. They all performed adult-like. For incomplete situations, we noticed that in both languages, the bilingual children have more no-responses than their monolingual peers. It might be that the bilingual children are more careful, or more critical, to judge certain verbal constructions as correct. As they are aware they are acquiring two languages which have different grammatical systems, they are careful not to judge too many constructions as correct. Rather, they tend to err on the side of caution. However, more research should be done to investigate these claims.

In production, we observed that in Dutch as well as English, the bilinguals performed similar to monolinguals, but with a few interesting differences. In Dutch, we noticed that the bilinguals made no use at all of the simple past tense. This could be a result of the carefulness described above; the simple past tense has very different uses in Dutch and English. In English, the simple past is a perfective form, while in Dutch, we take it to be aspectually neutral. As a result, the bilingual children avoid this difficult verb form.

(50)

50

(51)

51

6. Conclusion

This paper introduced a newly designed experiment which tested comprehension as well as production. The COST Aspect experiment was developed as part of Action A33 of the European Cooperation in Science and Technology (COST) network; “Crosslinguistically Robust Stages of Children’s Linguistic Performance, with Applications to the Diagnosis of Specific Language Impairment” (Gagarina, van Hout & Dressler, in prep.; Van Hout, Gagarina & Dressler, in prep.). The goal of this COST Action was to identify and describe areas of cross-linguistic uniformity that can then be used for cross-cross-linguistic tests of language impairment. The COST aspect experiment was found to have no list effects on the groups tested here. In the present study, the test was used with Dutch 4-, 5- and 6-year olds to determine their level of acquisition of grammatical aspect. For this end, we compared the results of the children to those of adult native speakers of Dutch. Also, we looked for possible asymmetries in comprehension and production in children and described which processes might cause this. Lastly, we looked at the acquisition of aspect in bilingual children, to see if the development of grammatical aspect was stable across their two languages.

Acquisition of Dutch Aspect