Production of change-of-state, change-of-location and

Preparing to load PDF file. please wait...

0 of 0
Production of change-of-state, change-of-location and

Transcript Of Production of change-of-state, change-of-location and

This article was downloaded by: [Harvard College] On: 10 February 2012, At: 04:16 Publisher: Psychology Press Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Language and Cognitive Processes
Publication details, including instructions for authors and subscription information:
Production of change-of-state, change-of-location and alternating verbs: A comparison of children with specific language impairment and typically developing children
Susan H. Ebbels a b , Julie E. Dockrell c & Heather K. J. van der Lely d e f a Moor House School, Mill Lane, Hurst Green, Oxted, RH8 9AQ, UK b Divison of Psychology and Language Sciences, University College London, Chandler House, London, WC1N 1PF, UK c Institute of Education, Psychology and Human Development, London, WC1H 0AL, UK d Department of Psychology, Harvard University, Cambridge, MA, USA e Laboratoire de Neuropsychologie Interventionnelle, Départament Etudes Cognitives, École Normale Supérieur, Paris, France f Psychological Sciences Research Institute, Université Catholique de Louvain, Louvain de Neuve, Belgium
Available online: 09 Feb 2012
To cite this article: Susan H. Ebbels, Julie E. Dockrell & Heather K. J. van der Lely (2012): Production of change-of-state, change-of-location and alternating verbs: A comparison of children with specific language impairment and typically developing children, Language and Cognitive Processes, DOI:10.1080/01690965.2011.605598
To link to this article:

Full terms and conditions of use:
This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Downloaded by [Harvard College] at 04:16 10 February 2012


Downloaded by [Harvard College] at 04:16 10 February 2012

Production of change-of-state, change-of-location and alternating verbs: A comparison of children with specific language impairment and typically developing children

Susan H. Ebbels1,2, Julie E. Dockrell3, and Heather K. J. van der Lely4,5,6
1Moor House School, Mill Lane, Hurst Green, Oxted RH8 9AQ, UK 2Divison of Psychology and Language Sciences, University College London,
Chandler House, London WC1N 1PF, UK 3Institute of Education, Psychology and Human Development, London WC1H
0AL, UK 4Department of Psychology, Harvard University, Cambridge, MA, USA 5Laboratoire de Neuropsychologie Interventionnelle, De´partament Etudes Cognitives, E´ cole Normale Supe´rieur, Paris, France 6Psychological Sciences Research Institute, Universite´ Catholique de Louvain,
Louvain de Neuve, Belgium

Correct use of verb argument structure relies on accurate verb semantic representations whose formation depends partly on use of reverse linking. We predicted that children with Specific Language Impairment (SLI), who have difficulties with reverse linking, would have inaccurate semantic representations for verbs and hence difficulties with verb argument structure. Fifteen participants with SLI (mean age: 13;1), grammar-matched (GM) (8;3), vocabulary-matched (VM) (8;8), and chronological age-matched (CAM) controls (13;1) described 24 video scenes involving four change-of-state, four change-of-location, and four alternating verbs. All groups performed worse on change-of-state than change-of-location verbs. The participants with SLI performed significantly worse than VM and CAM but not GM controls on change-of-state verbs. However, they did not differ from any group on alternating or change-of-location verbs. We concluded young people with persistent SLI have difficulties with aspects of verb argument structure into their teenage years.
Keywords: Verb argument structure; Specific language impairment.
Verb argument structure is at the interface of syntax and semantics and includes information about which participants in an event are obligatorily expressed and the syntactic positions in which they should appear. Thus, some verbs can only appear in particular syntactic constructions. Consider for example the verbs eat versus devour

Correspondence should be addressed to Susan H. Ebbels, Moor House School, Mill Lane, Hurst Green, Oxted RH8 9AQ, UK. E-mail: [email protected]

# 2012 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business

DOI: 10.1080/01690965.2011.605598

Downloaded by [Harvard College] at 04:16 10 February 2012

and pour versus fill: eat does not require a direct Object (the man is eating) whereas devour does (*the man is devouring is ungrammatical); for pour and fill the participants need to appear in different syntactic positions depending on which verb they follow, so the girl is pouring the water in the cup is acceptable whereas *the girl is filling the water in the cup is not because the cup needs to appear in the direct Object position rather than the water. Several researchers have proposed that in order to use verb argument structure accurately, children need to have detailed semantic representations of verbs (Gropen, Pinker, Hollander, & Goldberg, 1991b; Pinker, 1989). Others claim that for such detailed semantic representations to be learned in the first place, syntax plays a crucial role (Fisher, Hall, Rakowitz, & Gleitman, 1994; Gleitman, 1990; Gillette, Gleitman, Gleitman, & Lederer, 1999). Thus, children with language impairments affecting semantics and/or syntax are likely to make more errors in the production of verb argument structure than typically developing (TD) children with better semantic and/or syntactic knowledge.
Children with Specific Language Impairment (SLI) have difficulties in acquiring language despite adequate intelligence, hearing, physical development, and exposure to language. SLI is estimated to affect approximately 7% of kindergarten children (Tomblin et al., 1997), and for some children, language impairment persists into early adolescence (Beitchman, Wilson, Brownlie, Walters, & Lancee, 1996; Botting, Faragher, Simkin, Knox, & Conti-Ramsden, 2001) and beyond into late adolescence (Conti-Ramsden, 2008) and adulthood (Clegg, Hollis, Mawhood, & Rutter, 2005; Mawhood, Howlin, & Rutter, 2000). Persisting language impairments have negative effects on children’s educational achievements (Conti-Ramsden, 2008; Dockrell & Lindsay, 2008; Mawhood et al., 2000) and social adjustment (Clegg et al., 2005; ContiRamsden, 2008; Howlin, Mawhood, & Rutter, 2000), although a recent study reveals better social outcomes than earlier studies (Carroll & Dockrell, 2010).
Given the long-term impact of language impairments on educational and social development, it is important to study children with SLI of all ages, not only young children. There are surprisingly few studies of older children but these studies have found that impairment in specific areas of language remains even when compared with controls matched on other language abilities. These areas are: use of tense and agreement (e.g., Leonard, Bortolini, Caselli, McGregor, & Sabbadini, 1992; Rice, Hoffman, & Wexler, 2009; van der Lely & Ullman, 2001), comprehension and production of syntax (e.g., Bishop, 1979; Leonard, 1995; van der Lely, 2005), and use of reverse linking or syntactic bootstrapping (e.g.,O’Hara & Johnston, 1997; Shulman & Gudeman, 2007; van der Lely, 1994). This study focuses on the possible consequences of difficulties with reverse linking/syntactic bootstrapping, particularly considering change-of-state verbs.
Reverse linking and change-of-state versus change-of-location verbs
‘‘Reverse linking’’ (Pinker, 1989, 1994), or ‘‘syntactic bootstrapping’’ (Fisher et al., 1994; Gleitman, 1990), is the process whereby we can use the syntactic construction in which a verb appears as a cue to the verb’s meaning. Thus for an unfamiliar verb, we can work out whether it includes the meaning X acts or X acts on Y or X causes Y to move to Z, by noting the number of arguments appearing with the verb (Fisher, 1996, 2002; Naigles, 1990; Naigles & Kako, 1993). Thus, a verb appearing with just a Subject (e.g., the boy zaigs) is likely to describe X acting; whereas a verb appearing with a Subject, direct Object and Prepositional Phrase (e.g., the boy zugs the girl to the lady) is likely to describe X acting on Y, causing Y to move to Z. Reverse linking also has a

Downloaded by [Harvard College] at 04:16 10 February 2012

role in refining the semantic representations of verbs in terms of which referent is the Patient. The assignment of arguments to the direct Object position distinguishes change-of-state and change-of-location verbs. Consider, for example fill versus pour: for fill, the Noun Phrase (NP) in the direct Object position changes state (becomes full) (e.g., the girl is filling THE CUP with water) whereas, with pour, it changes location (e.g., the girl is pouring THE WATER into the cup). This reflects an underlying difference in the semantics of these verbs where the focus of the meaning of fill is on the ‘‘Goal’’ whereas the focus for pour is on the ‘‘Theme’’ which changes location. According to Pinker (1989) and Jackendoff (1990), this focus is captured by the concept of a ‘‘Patient’’. The Patient is ‘‘affected’’ by the action, regardless of whether it is also a Theme or Goal. Furthermore, Pinker (1989) and Jackendoff (1990) propose a set of linking rules which link verb semantics to syntactic structure, and whereby the Patient is linked to the direct Object position. Children use these linking rules in a process Pinker (1989) calls ‘‘forward linking’’ to work out how to link verbs’ arguments to structural positions in a sentence, such as the Agent to the Subject position and Patient to the Object position. However, children can only do this if they know the semantics of verbs (in terms of which arguments have the Agent and Patient roles), without which errors may occur. For linking the correct referent to the direct Object position, Gropen et al. (1991b) propose an ‘‘Object affectedness rule’’ whereby ‘‘an argument is encodable as the direct object of a verb if its referent is specified as being affected in a specific way in the semantic representation of the verb’’ (p. 118). Thus, such knowledge is needed in order to use verbs accurately in sentences.
Linking errors (e.g., ‘‘Can I fill some salt into the bear?’’(Bowerman, 1982), where the Theme appears in the direct Object position instead of the Goal), presumably occur because the child thinks the Theme is the Patient rather than the Goal. Such errors are found in TD children from 3 to 6 years of age and are more common and persist longer with change-of-state verbs (Bowerman, 1982; Gropen et al., 1991b). The evidence suggests that children are more sensitive to the meaning components associated with motion than with changes of state. Indeed Gentner (1978) found that children aged 5Á7 years were not sensitive to the change of state component of the verb mix (i.e., that the substance must increase in homogeneity). They were just as likely to accept as an example of ‘‘mixing’’ an event in which the homogeneity did not change (e.g., cream being stirred) as one in which it did (e.g., water and salt being mixed together). In contrast, they rejected events as examples of ‘‘stirring’’ where the appropriate action (rotary motion) did not occur. Therefore Gentner (1978) concluded that ‘‘children appear to learn the action components of the mixing verbs before they learn the change of state components’’ (p. 994). Gropen et al. (1991b) found that for the verb fill, some children (particularly those aged 4;6Á5;11) were biased towards the manner meaning components rather than the change-of-state (e.g., when asked to choose which of two pictures best showed filling, they tended to choose the picture showing a pouring manner without a full endstate, over the picture showing a full endstate with no pouring manner). However, the relative sensitivity to changes of state may vary with language ability. Kelly and Rice (1994) showed children with SLI and TD children two video scenes, the first of a single inanimate object spontaneously changing state in a fairly dramatic way (changing color or shape) and the second of the same object moving in a particular way. They then asked them to point to the scene that depicted a novel verb, thus indicating their initial preferred interpretation. They found that TD children aged 4;6Á5;8 preferred to associate the novel verbs with the event where the object changed state rather than moved in a particular way. However, the majority of children with SLI of the same age and younger TD children

Downloaded by [Harvard College] at 04:16 10 February 2012

(aged 2;7Á3;11) had no such preference; indeed those who did have a preference preferred the motion events.
Any bias towards picking up the manner components of meaning and lack of sensitivity to changes of state would particularly affect verbs such as fill and cover, which involve both motion and a change of state. Given that children with SLI have fewer preferences for changes of state than TD peers even for simple, salient changes of state (Kelly & Rice, 1994), it is likely that they will be even less sensitive to changes in state when they occur together with changes in location. Therefore, we predict that younger TD children and children with SLI will make more errors with change-of-state verbs when compared with change-of-location verbs and verbs that combine changes of state and location would be particularly prone to errors. However, errors on changeof-state verbs should reduce as children become more competent in their use of reverse linking because this enables them to revise imprecise semantic representations.
Reverse linking in SLI
A few studies have investigated the ability of children with SLI to use reverse linking. Studies carried out by van der Lely (1994), and later by O’Hara and Johnston (1997) revealed that when novel verbs are presented in transitive, transitive locative or dative sentences, children with SLI are significantly worse than younger language controls at using reverse linking to infer which NP has which role in the sentence. However, some studies have found that children with SLI can use information regarding the number of NPs uttered in a sentence with known (Hoff-Ginsberg, Kelly, & Buhr, 1996) and novel verbs (Oetting, 1999) to infer something about the likely meaning of the verb with respect to whether it has a transitive-causative or intransitive-stative meaning. However, if more precise knowledge of syntactic structure is required to distinguish causative and stative meanings, such as when the number of NPs are the same (e.g., ‘‘the dogs are kolzim the cats’’ vs. ‘‘the dogs and cats are kolzim’’), children with SLI perform at chance (Shulman & Gudeman, 2007). Thus, while some studies (e.g., Hoff-Ginsberg et al., 1996; Oetting, 1999) show that children with SLI can use sentences to identify the correct number of participants, those by van der Lely (1994), O’Hara and Johnston (1997), and Shulman and Gudeman (2007) indicate that they cannot reliably use reverse linking to assign the correct roles to the correct referents.
The ways in which difficulties with reverse linking may affect the production of verb argument structure have not been explicitly studied to date and form the focus of the current study. If children with SLI have more difficulties with reverse linking than TD children, we predict that they would rely more on observational and conceptual cues (such as the salience of changes in the different participants involved in an event) that are outside the linguistic system per se and less on syntactic cues (such as which participant appears in the direct Object position). Therefore we hypothesise they will be less able to use syntax to overcome their already reduced sensitivity to changes in state compared with TD children (cf. Kelly & Rice, 1994) and consequently will make more errors on change-of-state verbs than TD children.
Locative alternation
Errors on change-of-state verbs could occur for reasons other than (or in addition to) difficulties with reverse linking. Overgeneralisation of the locative alternation could also lead to errors. Verbs undergoing the locative alternation (e.g., pack and spread) can appear both in the change-of-location (e.g., he is packing his clothes into

Downloaded by [Harvard College] at 04:16 10 February 2012

his bag, she is spreading butter on the bread) and change-of-state constructions (e.g., he is packing his bag with clothes, she is spreading the bread with butter), thus overgeneralisation of the locative alternation could result in use of change-of-state verbs (e.g., fill) in both the correct (change-of-state) and incorrect (change-oflocation, she’s filling water in the cup) constructions. Bowerman’s (1982) daughters and the children in Gropen et al.’s (1991b) study showed this pattern for several change-of-state verbs. Thus, errors on change-of-state verbs could arise due to overgeneralisation of the locative alternation, and/or due to difficulties with reverse linking. The locative alternation has not been previously studied in children with SLI, but a study of the causative alternation (i.e., verbs which alternate between unaccusative and transitive constructions, e.g., the glass broke vs. the girl broke the glass) found children with SLI did not differ from language or CAM controls in their ability to use (or restrict overgeneralisation of) the causative alternation (Loeb, Pye, Richardson, & Redmond, 1998).
Current study
In this study, we investigated four change-of-state and four change-of-location verbs and four verbs which can undergo the locative alternation. We asked the participants to describe two video scenes for each verb. If a participant has difficulties with the task demands (i.e., has general processing deficits), they would be likely to make errors on all verbs. The alternating verbs should reveal any general preferences for the changeof-state versus change-of-location construction. If a participant overgeneralises the locative alternation to nonalternating verbs, they would use both the change-of-state (she’s filling/pouring the cup with water) and change-of-location constructions (she’s filling/pouring water into the cup). If they have difficulties with reverse linking and hence are more reliant on observational cues such as the salience of changes undergone by participants in the event, they may be particularly poor with changeof-state verbs. They may think the Theme, undergoing a salient change of location (rather than the Goal, undergoing a less salient change of state) is the ‘‘affected object’’ (i.e., the Patient) and hence would use only the (incorrect) change-of-location construction (e.g., she’s filling water into the cup) for change-of-state verbs.
We compared a group of participants with SLI with TD participants matched on three different criteria. Our first control group was matched on chronological age and therefore we hypothesised similar opportunities to hear verbs and observe the situations in which they are used. The second group was matched on receptive vocabulary and was thus likely to be most similar to the children with SLI in terms of lexical development. The third group was matched on sentence comprehension and was therefore likely to be at the most similar level of grammatical development. We also carried out the task with adults to ensure that we had correctly classified the change-of-location and change-of-state verbs as nonalternating verbs for Southern British English.
We predicted that difficulties with reverse linking would result in the participants with SLI in our study making more errors on change-of-state verbs than TD chronological age and possibly also vocabulary controls (who have similar lexical levels but possibly better grammatical and hence reverse linking abilities), but would probably not make more errors than their sentence comprehension controls (who are likely to have similar grammatical and reverse linking abilities). If the participants with SLI also have a general processing deficit and therefore difficulty with the task demands, we predict they would be poorer than TD controls not only

Downloaded by [Harvard College] at 04:16 10 February 2012

on change-of-state verbs, but also on change-of-location verbs. A preferred construction for the alternating verbs should indicate any general preference for one construction over the other.
Fifteen participants (four girls and 11 boys) with persisting SLI (mean age: 13;1 years, range: 11;0Á14;11), 15 CAM controls (mean age: 13;1, range: 11;3Á14;10), 15 vocabulary-matched (VM) controls (mean age: 8;8, range: 5;10Á12;2), and 15 GM controls (mean age: 8;3, range: 5;4Á11;3) participated in the study. The participants with SLI were all attending a specialist school in the UK which caters for specifically for children with primary language impairments. All children in the school who were aged between 11 and 15 and met the following criteria were recruited: (1) receptive and expressive language difficulties (at least (1.5 SD below the mean) as measured on the Clinical Evaluation of Language Fundamentals (CELF-3 UK, Semel, Wiig, & Secord, 1995), (2) typical nonverbal performance abilities (not more than (1 SD below the mean) on the mean of Matrices and Pattern construction from the British ability scales II (BAS-II, Elliot, Smith, & McCulloch, 1996), (3) intelligible spontaneous speech (assessed informally), (4) no hearing impairment, neurological dysfunction, structural abnormalities or diagnosis of autism or Asperger’s syndrome stated in their medical records, and (5) written consent given by the parents.
The TD controls were recruited from six mainstream schools in the same geographical region as the school for pupils with SLI. No TD controls had identified special educational needs or English as an additional language. They scored within normal limits on both language abilities and performance IQ (above (1 SD). As for the children with SLI, performance IQ was measured using the mean of the Matrices and Pattern construction from the BAS-II. The tests used to measure language were the Formulated Sentences subtest of the CELF-3, the British picture vocabulary scale*II (BPVS-II, a multiple-choice vocabulary comprehension test, Dunn, Dunn, Whetton, & Burley, 1997) and the Test of reception of grammar (TROG, a multiple-choice sentence comprehension test, Bishop, 1989). Each language control child was individually matched to a child with SLI on the basis of performance IQ (within one standard deviation) and either the BPVS (‘‘vocabulary-matched (VM) controls’’: raw score within 3 points) or the TROG (‘‘grammar-matched (GM) controls’’: matched on exact raw score). They were also required to score within the average range for their age (i.e., not more than 1 SD above or below the mean) on the test with which they were matched to the participants with SLI. The CAM controls were individually matched to the participants with SLI by age (within 3 months) and also scored within the normal range (i.e., within one standard deviation from the mean) on the BAS-II (performance IQ).1 The scores for the four groups on the matching criteria are shown in Table 1.
1 One age control achieved a z-score of Á1.15 on the BAS, but showed no language difficulties, was matched to the child with SLI with the lowest z-score (-0.95) on the BAS and was exactly the same age; he was therefore considered to provide a good match.

TABLE 1 Mean (SD) plus ranges on matching criteria (raw scores for BPVS and TROG, z-scores for BAS,
years; months for age)


91.07 (14.24) 63 to 115 15.40 (2.32) 9 to 18 (0.04 (0.82) (0.95 to 1.55 13;1 (1;3) 11;0 to 14;11

GM controls
87.00 (16.17) 58 to 120 15.40 (2.32) 9 to 18 0.33 (0.60) (0.60 to 1.25 8;3 (1;8) 5;4 to 11;3

VM controls
90.87 (13.84) 65 to 112 17.00 (1.69) 15 to 19 0.53 (0.68) (0.55 to 1.95 8;8 (1;6) 5;10 to 12;2

CAM controls
121.87 (13.45) 99 to 149 18.33 (0.90) 17 to 20 0.06 (0.66) (1.15 to 1.60 13;1 (1;3) 11;3 to 14;10

Downloaded by [Harvard College] at 04:16 10 February 2012

In order to validate the matching procedures, the groups were compared on age, performance IQ, and raw scores of the language tests used for matching. We found a significant effect of chronological age, F(3, 56) 051.18, pB.001, hp2 00.73, where the participants with SLI did not differ from the CAM controls (p01.0, d 00.004), but differed significantly from both GM (p B.001, d03.35) and VM controls (pB.001, d03.03). The latter two groups did not differ from each other (p01.0, d 00.3), but did differ from the CAM controls (TROG: pB.001, d 03.34; BPVS: pB.001, d03.03). The four groups showed no difference in their performance IQ as measured on the BAS, F(3, 56) 0 2.14, p0.11, hp200.10.
The four groups differed significantly on the BPVS raw score, F(3, 56) 018.874, pB.001, hp2 00.50. Post-hoc tests showed the participants with SLI did not differ from either their VM (p01.0, d00.01) or GM controls (p01.0, d00.28) but scored significantly lower than their CAM controls (pB.001, d02.13), as did both the GM (pB.001, d 02.41) and VM controls (p B.001, d 02.15) who did not differ from each other (p 01.0, d00.27).
The four groups also differed significantly on the TROG raw score (x2(3) 023.46, pB.001).2 Post-hoc tests showed the participants with SLI did not differ from either their GM (W 0232.5, n1015, n2015, p01.0) or VM controls (W074.5, n1015, n2 015, p0.11), but did differ from their CAM controls (W0130, n1 015, n2 015, p B.001). The GM controls differed from the CAM controls (W0130, n1 015, n2015, pB.001) whereas the VM controls did not (W0185.9, n1015, n2 015, p 0.05; the Bonferroni corrected significance value is 0.05/6 00.008). Again, the GM and VM controls did not differ significantly from each other (W 0194.5, n1 015, n2015, p0.11).
We also carried out the task on 10 adults (five aged 25Á50, and five aged 50Á75) who lived in the same geographical region as the school attended by the child/ adolescent participants.
Stimuli, procedure and scoring
The stimuli were part of a larger study of verb argument structure in SLI. The full test consisted of 72 video scenes (each 5 seconds on average) showing adults and children carrying out common actions with everyday objects. However, in this paper we report only on a subset3; those verbs which exclusively use either the change-of-state

2 The non-parametric Kruskal-Wallis and Wilcoxon signed ranks tests were used because the data for the SLI and language control groups were not normally distributed (the SLI and GM groups were positively skewed, while the VM controls showed a bi-modal distribution).

Downloaded by [Harvard College] at 04:16 10 February 2012

construction (fill, build, cover, and rob) or the change-of-location construction (pour, put, spill, and steal) or those which alternate between these two forms (pack, spread, peel, and wipe*the latter two of which involve removing an item from a location). Two video scenes were shown for each verb; these are described in Appendix 1. The stimuli were recorded by the first author and piloted on three TD children (aged 6Á8 years) and four adults to ensure that the events were correctly interpreted. New scenes were recorded where there was any sign of confusion.
The order of the video scenes was randomised but then checked by hand to ensure that there was a gap of at least two items between different scenes involving the same target verb. All participants watched the scenes in the same order. For each scene, the participant was shown the video clip once while the experimenter provided the target verb in the gerund: ‘‘this is VERBing’’. The clip was then repeated and the participant was asked: ‘‘What is happening?’’ Four practice items at the beginning of the test were used to train the participant to use the target verb in a complete sentence. These followed the same format as the test items and used the verbs ringing, dropping, walking, and telling for scenes showing a lady ringing a bell, a girl dropping from a climbing frame, a man walking to a shed and a lady telling a story to a little girl. These items required a varying number of arguments and the participants were encouraged to include a subject and any obligatory postverbal arguments, but were not required to produce optional arguments. Responses were recorded on a DAT tape recorder (TCD-D8) using an external Sony Electret condenser microphone and transcribed later.
Some participants omitted obligatory arguments (usually direct objects or prepositional phrases) and these errors were analyzed elsewhere (Ebbels, 2005). For the purposes of this paper, the responses were coded according to whether the direct Object changed state or location. Hence omissions of Prepositional phrases were ignored, but failure to use a direct Object meant the response could not be coded and was therefore recorded as missing data. The first author carried out all the testing, transcription, and scoring. The third author also coded the responses of 12 (20%) randomly selected participants (four with SLI and eight controls). The Kappa coefficient for inter-rater agreement was .975; disagreements were resolved through discussion.
Alternating verbs
We first consider the participants’ willingness to use alternating verbs in both constructions associated with the locative alternation aiming to establish whether they have any general or verb-specific preferences for a particular construction and whether the participants with SLI differ from their controls. Table 2 shows the mean use of the change-of-state construction for the locative alternation, for all groups, including the adults. Equal use of the two possible constructions for each alternation would result in a score of 0.50. Table 2 shows the overall mean and standard deviation for
3 The other scenes (reported in Ebbels, 2005) investigated the dative and causative alternations. For the dative alternation, this involved the alternating verbs pass and give. Investigation of the causative alternation involved verbs which are obligatorily intransitive (two unergative: jump and laugh and two unaccusative verbs: bubble and fall) and verbs which can undergo the causative alternation (in the transitive form two verbs involve changes of location: hang and roll and two involve changes of state: melt and open).