Language Facility and Theory of Mind Development in Deaf Children

Transcription

1 Empirical Articles Language Facility and Theory of Mind Development in Deaf Children A. Lyn Jackson University of Cardiff, Wales Deaf children with signing parents, nonnative signing deaf children, children from a hearing impaired unit (HIU), and oral deaf children were tested on three first-order theory of mind (ToM) tasks a subset was also given a second-order task (Perner & Wimmer, 1985). A British Sign Language (BSL) receptive language task (Herman, Holmes, & Woll, 1999) and four nonverbal executive function tasks were also administered. The new BSL task allowed, for the first time, the receptive language abilities of deaf children to be measured alongside ToM abilities. Hearing children acted as controls. These children were given the same tasks, except the British Picture Vocabulary Scale was substituted for the BSL task. Language ability correlated positively and significantly with ToM ability, and age was correlated with language ability for both the deaf and hearing children. Age, however, underpinned the relationship between ToM and language for deaf children with signing parents and hearing children but not for nonnative signing, HIU, or oral deaf children. Executive function performance in deaf children was not related to ToM ability. A subset of hearing children, matched on age and language standard scores with signing deaf children, passed significantly more ToM tasks than the deaf children did. The findings are discussed with respect to the hypotheses proposed by Peterson and Siegal (1995, 2000) and Courtin (2000). I thank the staff and pupils at the schools for all their help with this study; Ros Herman for granting me permission to use the BSL task, thereby making this research possible; and Professor Hadyn Ellis, Professor Marc Marschark, and two anonymous reviewers for their helpful comments. The research forms part of my doctoral thesis, supported financially by Cardiff University. Part of this work was presented at the BPS Developmental Section Annual Conference (1999) and The Child Language Seminar, London (1999). Correspondence should be sent to Lyn Jackson, School of Psychology, Cardiff University, Tower Block, Park Place, PO Box 901, Cardiff, CF10 3YG, UK ( jacksonl@cardiff.ac.uk) Oxford University Press Theory of mind (ToM), the ability to attribute mental states to oneself and others (Premack & Woodruff, 1978), has been a growing area of research over the last 15 years. Many researchers have been investigating how, when, and why ToM abilities develop in young children and whether these abilities, in fact, are unique to humans (Call & Tomasello, 1999). Much interest has also focused on the relative inability of children with autism, and more recently deaf children, to pass falsebelief tests. The ToM performance of deaf children is the focus of this study. Peterson and Siegal (1995) were the first to report that deaf children of hearing parents were delayed in their ability to pass a false-belief test. The false-belief test used was the unexpected transfer test originally devised by Wimmer and Perner (1983). This test, often referred to as the Sally-Ann task, involves a doll (Sally) who places an object in a basket, and out of the room, and therefore unbeknown to Sally, the object is moved by another doll (Ann) to a new location. The child is asked to predict where Sally will look for the object. Peterson and Siegal reported that not only was the performance of deaf children (mean age: 10 years) delayed compared with hearing children on this task, 83% of deaf children failed the test, but the performance of the deaf children was of a similar level to that of children with autism, reported by Baron-Cohen, Leslie, and Frith (1985). Based on these findings, it seemed that the ToM hypothesis of autism (Baron-Cohen, 1995; Baron-

2 162 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 Cohen et al., 1985; Leslie, 1991, 1994) could no longer hold, given the lack of autistic behaviors in deaf children. Peterson and Siegal (1995) posited an alternative and somewhat more parsimonious explanation of the findings for both deaf children and children with autism: the early conversational hypothesis, whereby language, specifically exposure to and discussion about mental states, is one of the main underpinnings of ToM development. Both children with autism and deaf children experience conversational deprivation, albeit for different reasons, which may hinder their ToM development. Further support for ToM ability in deaf children of hearing parents (DH) being delayed owing to lack of early conversational experience appeared when Peterson and Siegal (1999) tested deaf children from deaf families (DD) on three false-belief tasks: an unexpected transfer task, an unexpected contents task, and a false photograph task (originally designed by Zaitchik, 1990; see Method for a full description). The performance of the DD children, Peterson and Siegal s native signers, did not differ from that of hearing preschoolers or oral deaf children, all of whom were performing significantly better than the deaf children from hearing families (DH) and children with autism. This finding suggests that, with access to language from birth, children, irrespective of hearing ability, will develop a normal ToM. There is, however, one problem with the Peterson and Siegal study: the study compares the results from DD signers, whose mean age was 10 years 3 months, with those of hearing children who were 4 and 5 years old. To substantiate the claim that native signing deaf children develop a ToM at the same age as hearing children, around 4 years old, one should complete a comparative study testing 4-year-old native deaf children and hearing children on both false-belief and narrative (Marschark, Green, Hindmarsh, & Walker, 2000) tasks. The results of such a study would clarify whether DD children develop a ToM around the same age as hearing children. Recently de Villiers, de Villiers, Schick, and Hoffmeister (2000) found that deaf children (age: 4 7 years) who had learned sign language early and naturally from deaf parents were significantly better in their social-cognitive reasoning than deaf children from hearing families. This more recent evi- dence does suggest native signers develop a ToM earlier than nonnative signing children. Unfortunately, the data collection for the hearing control group has yet to be completed, making it impossible now to compare directly native signers and hearing children. Since Peterson and Siegal s primary paper, a number of published studies have supported and extended their original finding. Deaf children from deaf families pass significantly more ToM tasks than deaf children from hearing families (Courtin, 2000; de Villiers et al., 2000; Peterson & Siegal, 1997, 1998, 1999; Russell et al., 1998). Furthermore, the ToM performance of DD children has been found to be no different from that of hearing preschoolers (Peterson & Siegal, 1997, 1998, 1999). Several studies have produced conflicting findings particularly with respect to oral deaf and DH children. P. de Villiers and Pyers (2000) found that oral deaf children showed a ToM performance delay of 3 to 4 years with respect to hearing children. De Villiers and Pyers (2000; de Villiers & de Villiers, 2000) suggest that mastery of complementation, with both communicative (e.g., She tells him X ) and cognitive states (e.g., He believes X ), provides the necessary representational abilities for understanding a false-belief task. This linguistic hypothesis does not claim that complementation understanding is a prerequisite for ToM development, rather that it is only necessary for success on a false-belief task. Two recent studies have produced conflicting findings with respect to signing deaf children (Courtin, 2000; Marschark et al., 2000). Courtin found that DD children, but not DH children, exhibited superior performance on traditional false-belief tasks when compared with hearing children. Courtin noted that because the perspective presented when signing is most often the signer s, visual perspective changes are required for the recipient if the intended content of a conversation is to be understood; that is, the recipient must put himself or herself in the other person s position to grasp the true perspective. Given that the ability to attribute false beliefs is rooted in the mastery of visual perspectives (p. 267), Courtin postulated that signing deaf children should develop a ToM earlier than those who speak because of their extensive exposure to perspective taking.

3 Language and Theory of Mind 163 In the work of Courtin (2000) DD, DH, and oral deaf children, between 5 and 8 years old, and hearing children, between 4 and 6 years old, were given two unexpected transfer tasks and one unexpected contents (Smarties) task. All three tasks required the prediction of behavior dependent upon an understanding of mental states. In accordance with his hypothesis, Courtin found that DD and DH children outperformed oral deaf children on these tasks, whereas DD children outperformed DH children. Courtin also found that DD children outperformed hearing children, whereas DH children performed no differently from hearing children. This finding contrasts with the results of Peterson and Siegal (1999), who found that the performance of both oral deaf children and native signing deaf children was at a level equal to that of hearing preschoolers. Another finding from the Courtin (2000) study at odds with the previous literature is that of poor performance by the hearing children. Only 39% of the hearing children between 4 and 8 years old passed the false-belief tests. The DD and hearing children were matched on parental socioeconomic status (SES; lower rather than middle class), and Courtin cites this, along with the use of three false-belief tasks, as a possible explanation for the relatively poor performance of the hearing children. Courtin (2000) proposed a fundamentally different linguistic hypothesis to that of Peterson and Siegal (1995, 2000). He suggested that deaf children s experience with sign language and perspective taking leads to a change at the representational level that allows signing deaf children to form representations earlier and more easily than hearing children. This early formation of representations leads to an optimization of false-belief performance at younger age. In contrast, Peterson and Siegal (1995, 2000) propose that language contributes to false-belief performance via a change at the communicative level, enhancing conversations and social interactions, which subsequently leads to a greater understanding of mental states and, consequently, false beliefs. The findings of Marschark et al. (2000) support Courtin (2000); however, they are interpreted differently. Marschark et al. explored ToM in DH signing children by examining narratives produced under the auspices of telling a story in order that it may be shown to other children with lesser language skills. They found that DH children (between 9 and 15 years old) were no different from age-matched hearing children with respect to the number of statements of true and false beliefs produced. This finding is at odds with most previous research and led the authors to conclude that the false belief task and attributions of mental states are not equivalent abilities (Marschark et al., 2000, p. 1070). Indeed, the deaf children produced more mental state attributions, leading Marschark et al. to conclude that (1) deaf children from hearing families have an understanding of mental states and do possess a theory of mind and (2) the predictive element of the false-belief task may mask the ToM abilities of DH children. The Marschark et al. (2000) study highlights the danger of relying on a single task in deciding whether children should be labeled deficient in a particular domain (p. 1072). Furthermore, Marschark et al. s study, in conjunction with Courtin (2000), suggests that children who are deaf have a fully functional ToM but have been hampered in their opportunities to display it. One criticism of the Marschark et al. study remains that no false-belief task was administered alongside the narrative task, leaving open the question of whether the deaf children were attributing mental states independently of false-belief task success, as their hypothesis suggests. The possibility that mental state attribution coexisted with false-belief success cannot be excluded. The above studies add weight to the hypothesis that language history and modality contribute to mental state attribution and false-belief task success. The language hypothesis, in one guise or another, is able to account for ToM task failure in children with autism and deaf children, thereby bringing into question the primacy and efficacy of ToM as causally implicated in autism. The language hypothesis has not been used to explain the executive function (EF) deficits found in children with autism, so the alternative executive function hypothesis (Ozonoff, 1995; Ozonoff, Pennington, & Rogers, 1991; Pennington & Ozonoff, 1996) cannot be discounted. EF refers to the ability to maintain an appropriate problem-solving set for the attainment of a future goal (Luria, 1966, cited in Ozonoff, 1995, p. 201). Studies

4 164 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 have revealed that children with autism perform poorly compared with controls on EF tasks such as the Wisconsin Card Sorting Test (WCST) and the Tower of Hanoi (Benetto, Pennington, & Rogers, 1996; Ozonoff et al., 1991). The WCST requires set-shifting ability, the ability to sort cards according to one rule and to switch to a different rule when the previous rule is no longer effective, as well as the ability to inhibit prepotent responses. The tendency to persevere by continuing to sort according to the previous rule is characteristic of the performance of children with autism (Ozonoff, 1995). Successful performance on the Tower of Hanoi task also requires the inhibition of prepotent responses but also involves the ability to plan in order to reach a predetermined goal. As with the WCST, children with autism perform poorly on this task. The poor performance by children with autism on both EF tasks and false-belief tasks led Ozonoff et al. (1991) to propose EF as the primary deficit in autism (Ozonoff, 1995; Pennington & Ozonoff, 1996). By contrast, Perner and Lang (1999) question the primacy of EF and ToM abilities. Thus, while acknowledging the substantial data in support of a relationship between EF and ToM in children with autism, they argued that EF requires ToM abilities as opposed to the more traditional view of ToM requiring EF abilities. A detailed description of Perner and Lang s theory is beyond the scope of this article; however, it does highlight the lack of conclusive evidence in this area. Given the finding that ToM and EF deficits cooccur in children with autism, and that DH children also have problems with false-belief tasks, I ask whether EF deficits are evident in deaf children and whether ToM and EF abilities are linked in this group. With respect to the former, it has been shown that EF has a working memory (WM) component (Pennington & Ozonoff, 1996). In light of the tentative evidence for a link between WM and ToM (Davis & Pratt, 1995; Ottem, 1980) and the possibility of a WM/ short-term memory deficit in deaf children (Marschark, 1993), I hypothesize that deaf children will show EF deficits when the tasks are of a verbal nature. The goals of this study are, first, to investigate the relationship between language and false-belief task performance, incorporating the language modality and experience of deaf children. Several studies have found a positive correlation between language and ToM performance (de Villiers & de Villiers, 2000; Peterson & Siegal, 1999). De Villiers and de Villiers used the Index of Productive Syntax (IPSyn) to measure the grammatical complexity of deaf children s language and, more specifically, the number of complements produced. The IPSyn complement score correlated significantly and positively with success on three false-belief tasks. A criticism of Peterson and Siegal (1999) is that they do not include an objective measure of language ability for deaf participants, allowing a more direct and controlled test of their theory. When the performance of deaf children on false-belief tasks has been compared with that of hearing children, in most studies the verbal mental age (VMA) of the hearing children has been measured by the British Picture Vocabulary Scale (BPVS). Unfortunately, an equivalent test of language ability for the deaf population, thus far, has not been available. Previously, Peterson and Siegal examined verbal ability using a subjective system whereby the teacher rated each child s language on a five-point scale. They found a significant positive correlation between ToM performance and verbal ability in oral deaf, nonnative signers and autistic children. Assessed language ability was positively but not significantly correlated with ToM performance in native signers and hearing preschoolers. There are limitations associated with using a subjective measure of language ability and a standardized, objective measure of sign language ability is required. Recently Herman, Holmes, and Woll (1999) developed such a measure. Their measure of British Sign Language (BSL) ability contains both a receptive and production test. Unfortunately, the production test was still in the development stage while the data were being collected for this study; hence, only the receptive measure was used. The receptive test involves a vocabulary check and then a measure of receptive BSL ability (see Method section for a complete task description). The receptive BSL task is similar in format to the BPVS in that they both require the child to choose from an array of four pictures the one that best describes the word or sign/ sign sequence they have just heard or seen. Neither task tests explicitly for mental state knowledge, al-

5 Language and Theory of Mind 165 though one sign sequence does include not-like ; however, a measure of receptive language ability is a useful tool in investigating ToM performance. Previous studies with hearing children have found significant positive correlations between receptive language ability, as measured by the BPVS, and ToM performance (Charman & Baron-Cohen, 1992; Happé, 1995; Leekam & Perner, 1991; Ruffman, Perner, Naito, Parkin, & Clements, 1998). Thus, because both hearing and deaf children were to be tested in this study, a receptive measure of BSL ability was considered appropriate. A second goal of this study is to extend the level of false-belief tasks administered. To the best of my knowledge, no second-order task has been given to children who are deaf. A second-order task requires the attribution of a first-order false belief (Sally thinks X) to another person (Ann thinks Sally thinks X ). The task used in this study was devised by Baron-Cohen (1989; see Method for a full description of the task). The testing of second-order ToM abilities in hearingimpaired children is an important step in helping to discover the developmental profile of ToM in this population and its link with language abilities. A third goal of this study is to investigate the EF performance of deaf children using four nonverbal EF tests (McEvoy, Rogers, & Pennington, 1993). The use of nonverbal EF tasks should remove any performance decrement due to linguistic demands. The first two tasks simply require the child to remember under which of two cups a sweet was hidden and to retrieve it either immediately (Task 1) or after a 6-second delay (Task 2). One could argue that these two tests are testing only object permanency. While this is quite probably true given the age of the sample used, the tests were originally designed for use with infants, and the results are still of interest as the unexpected transfer task requires object permanency. Failure on Task 2 could go some way toward explaining a poor performance on the false-belief tasks. The two other EF tasks require planning and inhibition of past strategies. For example, in Task 3 the site of hiding of a sweet remains the same until four correct searches have been made; it then switches without warning and remains until a further four correct searches have been made and so on. The strategy (e.g., search left ) must be inhibited in order to search con- sistently on the right. Task 4 requires the formation of an alternating search strategy (e.g., search left, right, left right ) as the site of hiding alternates with every correct search. Finally, this study aims to rectify a criticism levied at most previous published studies investigating ToM in children who are deaf (Courtin, 2000; Peterson & Siegal, 1995, 1999; Russell et al., 1998; Steeds, Rowe, & Dowker, 1997): the inclusion of two experimenters, one of whom is an interpreter interpreting from sign language to spoken language or vice versa for only one group of children. In addition, the language modality used has been dependent on the children tested and has varied from Australian Sign Language (Auslan; Peterson & Siegal, 1995) and Paget-Gorman (Russell et al., 1998) to BSL (Steeds et al., 1997). This disparity in task language administration and the use of two experimenters presents potential confounds when comparing the performance of deaf children with that of hearing children. In this study, therefore, only one experimenter, who is proficient in BSL, was present. Thus, the hearing controls and deaf children experienced the tasks in their own language, in a consistent way, and via the same experimenter. For example, Rhys-Jones and Ellis (2000) tested hearing and deaf children on a task requiring the correct sequencing of pictures alongside the telling of a story and found that hearing children were less likely to use mental state terms or tell long stories. Rhys- Jones and Ellis attribute this to the experimenter being deaf and hearing children feeling uncomfortable communicating via an interpreter. This may be how the deaf children feel with hearing experimenters and an interpreter. Thus, it is of vital importance that the children accept the experimenter, feel relaxed and able to converse in the same language, and express themselves freely (Hindley, Hill, & Bond, 1993). To summarize, the aims of this study are to compare the performance of deaf children, from a variety of linguistic backgrounds, with that of hearing children on three false-belief tasks, a second-order false-belief task, and a set of nonverbal EF tasks. I predicted a positive relationship between receptive language abilities and false-belief task success, irrespective of linguistic modality. Furthermore, I predicted that (1) deaf children of signing deaf parents (DSP) will outperform

6 166 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 nonnative signing deaf children (Courtin, 2000; Peterson & Siegal, 1997, 1999) as well as severe and profoundly deaf oral children (Courtin, 2000; de Villiers et al., 2000) and (2) oral deaf children will perform at a preschool level (de Villiers & Pyers, 2000). Concerning the results for the DH children, there are two possible hypotheses. In line with Peterson and Siegal (1999), I predicted that DH children would perform below the level of hearing preschoolers. An alternative prediction, in line with Courtin (2000), is that performance of DH will not differ from that of hearing preschool children, especially given the older age of the DH children in this study. Performance on the second-order task will be at a level comparable to that of hearing preschool children and will follow the same developmental progression (i.e., no child should pass the second-order ToM task without first passing the first-order ToM tasks). Finally, I predicted that the EF performance of deaf children would be similar to that of hearing children because of the reduced verbal demands of the nonverbal EF tasks. Whether EF performance and ToM success correlate remains speculative. Method Participants The deaf sample consisted of 51 severe-to-profoundly deaf children from deaf schools and hearing-impaired units (HIU) around England and Wales. There were (1) 11 children who had parents who signed (deaf children of signing parents, DSP) (2 boys and 9 girls, mean age 7;1, range: 5;2 10;10); (2) 12 nonnative signers, although one boy failed to attend to the BSL task and was subsequently excluded from the study, leaving 11 children (7 girls and 4 boys, mean age 6;6, range: 4;10 8;3); (3) 14 children from an HIU (8 girls and 6 boys, mean age 9;2, range: 5;9 12;11) who used simultaneous BSL signs and spoken English; and (4) 14 oral deaf children (4 girls and 10 boys, mean age 7;6, range: 4;11 11;1) from an HIU where spoken English is the first and preferred method of communication. Both the DSP children and nonnative signers used BSL as their first and preferred method of communication. In the DSP group, only four children came from truly native signing households where one parent or older member of the family was deaf. Seven children came from families where at least one parent had signing skills at BSL Stage II or above. The two groups did not differ on any measure, so they were grouped together under the DSP banner. The DSP and nonnative signing children were all profoundly deaf with a loss of 90 db or more in the better ear. Degree of hearing loss was not available for the HIU or oral children; however, they were all labeled by the schools they attended as severe or profoundly deaf. One child in the DSP group and one in the nonnative signing group had cochlear implants; each had had them for less than a year. Three children from the HIU group had cochlear implants, but the duration of implant was not known. The hearing data were taken from a previous experiment (Jackson, in preparation) in which the same false-belief tests were given. It consisted of 120 children between 3 and 7 years old from primary schools in South Wales. The hearing subjects used in this study were 48 preschoolers (24 3-year-olds [mean age] and 24 4-year-olds [mean age]) and 24 7-year-olds (mean age). There was also a subset of 24 children (six 4-, 5-, 6-, and 7-year-olds) who were given the EF tests and the Raven s Coloured Progressive Matrices approximately a year after the ToM battery. Tasks: BSL Receptive Test (Herman et al., 1999) This task, which has recently been developed, is the only test available that can be used with deaf children to give an estimation of their verbal, or signing, age, by way of raw and standard scores. Participants were initially given a vocabulary test that examined their ability to produce, and understand, signs that were subsequently used in the video presentation of the test. The participants were shown pictures depicting an object, person, or animal used in the test proper and were asked to give the sign for that particular object. If the sign did not correspond to the one used in the video, participants were retested at the end of the vocabulary test with the test sign. For example, if the participant had used a different sign for boy, this picture, along with three other pictures chosen at random, would have been placed before the child. The experimenter

7 Language and Theory of Mind 167 would then sign boy using the test sign and the child had to point to the correct picture, in this case the boy. No child failed to give a sign for a picture or to understand the test sign. Having finished the vocabulary test, the child was instructed to watch the television monitor and was informed that a video would be shown in which a woman would explain what he or she must do. For children younger than eight, the experimenter explained the procedure and then started the tape whereupon they viewed a woman explaining the procedure again. The tape was paused after the procedural explanation to ensure that everything had been understood. Despite prompting, no child asked any questions, and all demonstrated having understood the instructions by completing three practice items. The test proper began, and the child was required to watch a woman signing; the picture faded and the child looked at four pictures and pointed to the one he or she thought depicted what had been signed. Testing continued until the child produced four consecutive incorrect responses or until the 40 items were completed. Theory of Mind Tasks The false-belief tests chosen are the most commonly used tests within the literature. The original or most widely used variations of a test were employed, and, to prevent fatigue, a single version of each test was presented. The unexpected transfer task. The Sally-Ann task (Baron- Cohen et al., 1985, adapted from Wimmer & Perner, 1983) involved two rag dolls. Both dolls were the same height but differed in hair color, skin color, and clothes. A small red container and a large yellow container were used as hiding places for the marble. The test followed the original format in which Sally hides a marble in the red container, and while she is out the room, Ann moves the marble to a new location, the yellow container. The child is then asked the test question, Where will Sally look for her marble ; the reality question (if test question answered correctly), Where is the marble really ; the memory question, Where did Sally put her marble in the beginning ; and a naming question, Which doll is Sally. Pointing, signing, or otherwise indicating the red container, in conjunc- tion with correct responses to the control question, resulted in a pass and a score of one. Indicating the yellow container and/or failing a control question scored zero. The unexpected contents task. The Smarties test (Perner et al., 1989) required a tube of Smarties and two small pencils. Participants were shown a tube of Smarties and asked what they thought was inside. All subjects replied Smarties, chocolates, or sweets. They were then shown that the tube really contained pencils. They were asked what they thought was originally in the tube (own initial belief), what was really in there, and the test question, What will X say is in here (X being their friend). As above, a pass was awarded if the subject answered the belief and control questions correctly. The False Photograph Test. This test (Zaitchik, 1990; Leekam & Perner, 1991) used a toy camera, a piece of card approximately 5 inches square, a doll, and two different colored dresses. Leekam and Perner (1991) modified the False Photograph Test, originally designed by Zaitchik (1990). The test used here was an amalgamated version. A toy camera was used instead of a real Polaroid as Zaitchik, Experiment 2, had shown that the initial findings of a real camera helping the participants was in fact a spurious result and there was no advantage in having a real photo present. The children were introduced to the toy camera and asked whether they had a camera at home or whether their parents had one. They were then asked about pictures and whether they knew what they were. All the children except one had a camera in the house, and all knew what a camera did and what pictures were. The children were then shown the doll in a blue dress and they took the pretend picture of the doll. A piece of card was used as the pretend photo (as using a real Polaroid camera would have produced a picture), and the dress was then changed for a pink one. They were then asked, What color will Judy s dress be in the photo. If the child answered blue, he or she was awarded a score of one. The second-order ToM task. The ice-cream man story (Baron-Cohen, 1989, adapted from Perner & Wimmer, 1985) involved constructing a toy village. Two houses, a church, some trees, an ice-cream van, and four play

8 168 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 people were placed in such a way as to depict a village. They were placed on a sheet of laminated paper on which a park and road had been drawn (see Baron- Cohen, 1989, for a diagrammatic description). The second-order ice-cream story followed the procedure and wording used by Baron-Cohen (1989). The experimenter set up a toy village on a table directly in front of the child and proceeded to tell a story. John wants to buy ice cream from the man in the park but realizes he has no money and must first go home. When he is fetching his money, the ice-cream man moves to the church in full view of Ann. Ann, John s friend, goes to look for John at his house and is told by his mother that he has gone to buy an ice cream. To find John, Ann must think that John thinks the ice-cream man is still in the park. Thus, to answer the test question, Where will Ann go to find John? correctly, the child must attribute Ann thinks John thinks the ice-cream man is in the park. A pass was awarded if the test and control questions were answered correctly. All children were given the three first-order falsebelief tasks; the second-order task was given to children if time permitted. This resulted in only 28 children being given the second-order task, and it followed presentation of the first-order tasks. The wording for all tasks was translated into BSL for the children who used BSL; the HIU children were given the tasks using simultaneous BSL signs and spoken English. Oral deaf children received spoken English instructions and so were required to lipread in conjunction with using their hearing aids. Aware that the oral children may experience problems with lipreading a person unfamiliar to them, the experimenter questioned any child who looked confused and offered to repeat the instructions. The experimenter stopped immediately if the child looked away during the presentation of the story or questions. The question or part of the story missed by the child was then repeated. The control questions were also in place to ensure understanding of the stories and tasks. Executive Function Tasks The tasks were identical to those used by McEvoy et al. (1993), who concatenated a number of tasks that, together, provide a comprehensive measure of EF ability (see McEvoy et al., 1993, for task descriptions). The tasks required little or no language and were thus ideal for use with young deaf children who may have only a couple of years exposure to language. The EF test consisted of four separate tasks. The first two required children to maintain a representation of where an object was. For the majority of children, performance was at ceiling for these two tests. The third and fourth tasks required the children to develop a simple response set and then to be able to change it when the response was no longer successful. On these tasks, in line with McEvoy et al. (1993), the number of perseverative errors was measured, allowing a comparison with their autistic sample. The number of nonperseverative errors was also measured; this allowed different patterns of responding to be investigated. All the tasks required the child to search for a hidden sweet placed under one of two plastic cups. During Task 1, the child saw where the sweet was hidden and was immediately allowed to search for it. The site of hiding remained the same until the child was successful; once successful, the site of hiding was changed, and so on. There were 10 trials, and if the child searched correctly for the first three trials, a score of 10 was awarded and the next task began. Task 2 was identical to Task 1, except that after the sweet was hidden, in full view, a shield was then placed between the child and the cups for 6 seconds; then it was removed and the child was allowed to search for the sweet. The scoring procedure was identical to the one used in Task 1. During Tasks 3 and 4 the screen was lowered prior to hiding the sweet; thus, the child had to guess where to search initially. On Task 3 the sweet remained in the same position until the child located it correctly four consecutive times (AAAA). The site of hiding then changed until a further four consecutive correct searches had been made (AAAA, BBBB, AAAA, and so on), until 20 trials were completed. In order to pass Task 3, three sets must have been achieved (there were 20 trials and one set constituted four consecutive correct searches). The strict criteria ensured that the children had developed a search strategy and were not simply lucky in their guessing. Task 4 required the child to develop an alternating search strategy (ABABABA); the sweet remained in the same position until the child had successfully located it; it was then switched to the other location until another correct search had been performed and so on

9 Language and Theory of Mind 169 until six consecutive correct searches had been made or 20 trials had been completed, whichever happened sooner. A pass was awarded if the subject produced six correct searches in a row. Procedure I tested all children in a quiet room of the school during one 30- to 40-minute session. I am proficient in BSL and have had BSL Stage II (minimum requirement for teachers of the deaf, Council for the Advancement of Communication with Deaf Persons). Although proficiency in BSL was good, it can in no way compare with the fluency and increased level of understanding between participant and a deaf experiment or native signer. The administration of the false-belief, executive function, and BSL receptive tasks (or BPVS for oral deaf children) was counterbalanced. Within this, the order of the false-belief tasks was also counterbalanced (for children receiving the second-order task, this was administered after the completion of the first-order tasks). Raven s Coloured Progressive Matrices were administered to most of the children at the end of each session; unfortunately, owing to time constraints, not all children could be tested. Receptive language ability of the oral deaf children was tested using the BPVS, and the results must be interpreted with caution. It may be argued that the BPVS tests only oral deaf children s ability to lipread and not linguistic understanding. Unfortunately, no other test was available, and, given that the oral deaf children s first language was English, the BPVS seemed the only appropriate test to use. Scoring One point was awarded for each of the false-belief tasks and zero for a failure; thus, the ToM score ranged from zero to three. A task was scored as passed if the child answered, either in sign or spoken language, the control and test questions correctly. The two exceptions to this were the Sally Ann task and the second-order task in which pointing was accepted as an answer. The BSL score was the number correct out of 40 (these scores were also converted to standard scores for some analyses). The EF tasks were scored within each task, and then an overall EF score was calculated; for Tasks 1 and 2 each correct response was awarded one point, up to a maximum of 10 points. To pass tasks 1 and 2 a score of eight on each test was required. However, for Tasks 3 and 4 a number of different variables were measured: number of correct responses, number of nonperseverative errors, number of perseverative errors, and number of sets achieved. A response was counted as perseverative if it was incorrect and the preceding response was also incorrect (Griffith, Pennington, Wehner, & Rogers, 1999). The total EF score was reached by awarding one point for passing the first task, two points for the second task, three points for task three, and four points for task four, giving a maximum of 10. This scoring procedure allowed the more difficult EF tests to be taken into consideration if passed. Results Of the 50 deaf children, 4 (1 oral and 3 HIU) failed one or more control questions on the Sally Ann task, and 13 (2 oral, 1 DSP child, 1 nonnative signer, and 8 HIU) failed one or more control questions on the Smarties task. Only two hearing children (both 3 years old) failed a control question on the Sally-Ann task. I decided to include the control question failers in the analyses, as it would mean losing over half the subjects from the HIU group (see Discussion). There was no significant difference between the age of those passing and failing the control questions. Control question failers had lower receptive language scores (BSL: 18.8 vs. 24; VMA: 1; 11 vs. 4; 4). No child who failed a control question passed the test question. Table 1 shows the percentage of children in each group who passed each false-belief task. A chi-square Table 1 Percentage of children passing each false-belief task Group (n) Sally Ann Smarties False Photo Native (11) Nonnative (11) HIU (14) Oral (14) Preschoolers (48) (3-and 4-year-olds) year-olds (24)

10 170 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 test revealed no significant difference between any of the groups of deaf children on any task. Table 2 shows the mean age, ToM score, receptive language score, and standard score for the four groups of deaf children and two groups of hearing children. With respect to age, a significant difference was found between the mean ages of the four groups of deaf children (one-way ANOVA), F(3, 46) 4.76, p.01. Scheffé post-hoc tests revealed that children from the HIU were significantly older than the nonnative signing deaf children. Language ability was hypothesized to be important for the development of ToM. Having had access to language at school and at home, the DSP group should have better receptive language skills and higher BSL scores than the nonnative signers, and indeed they did, t(20) 1.8, p.05, 1-tailed. The difference between the BSL scores for the DSP group and the HIU group approached significance, t(23) 1.6, p.06, 1-tailed. Table 2 shows that the DSP group obtained the highest BSL scores, followed by the HIU group and then the nonnative children. With respect to ToM ability, a one-way ANOVA with group (DSP, nonnative signers, HIU, oral, hearing preschool children) as the independent variable and ToM score as the dependent variable produced no significant effect of group, F(3, 46).08, p.05. This was despite the age and receptive language differences among the four groups of deaf children. Theoretical interest in individual group performances led to the analyses including samples with small group numbers. Thus, caution should be expressed when interpreting the results. Table 3 shows the correlation matrix for age, BSL score (or VMA), and ToM score for each group of children. Age and receptive language correlated highly and significantly for all five groups of children. In keeping with the hypothesis, there is a significant positive cor- Table 2 Age, ToM score, receptive language score, and standardized language score for both deaf and hearing groups Age ToM score BSL score Standard Group (n) (yr;mth) (max 3) (max 40) VMA (yr;mth) score Native (11) 7; SD Nonnative (11) 6; SD HIU (14) 9; SD Oral (14) 7; ;11 65 SD Preschoolers (48) 4; ;8 97 SD yr-olds (24) 7; ;4 101 SD Table 3 Correlation and partial correlation matrix for age, language, and ToM score for deaf and hearing children Oral (13) Nonnative (11) HIU (14) Hearing (48) Native (11) Rec. Rec. Rec. Rec. Rec. lang. ToM lang. ToM lang. ToM lang. ToM lang. ToM Age.68**.54*.68**.19.68**.04.67**.45**.75**.76**.01 a.18 a.05 a.36** a.59* a ToM.79** *.60*.69** a.45 a.11 a a One oral deaf child was not given the BPVS; thus, no language measure is available and n 13. a Partial correlations. *p.05. **p.01 (all 1-tailed).

11 Language and Theory of Mind 171 relation between receptive language and ToM score for all but the HIU children. There is also a significant positive correlation between age and ToM score for the DSP, oral deaf, and hearing children. To examine the relationship between ToM score and receptive language more fully, I partialed out age. For the DSP and hearing children, this resulted in the positive correlation between ToM score and receptive language being reduced to almost zero (see Table 3). In contrast, the relationship between ToM and language remained for the nonnative, HIU, and oral children. Conversely, when receptive language score was partialed out, the relationship between age and ToM score remained for the DSP and hearing children (r[8].59 and r[45].36, respectively). The correlation for the oral children was reduced to almost zero (r[13].01). Not all children were given the nonverbal IQ (NVIQ ) test, and the individual group numbers did not permit separate correlations. To investigate the relationship between ToM performance and NVIQ, I combined the four groups of deaf children and found a small positive correlation (r[41].11, ns), indicating that general cognitive abilities are not strongly related to improved ToM performance. To determine how performances varied with age and language ability, I divided the four groups of deaf children into young (age: 4 6 years) and old (age: 7 12 years) groups. The young hearing children were the preschoolers used in the previous analyses; the old hearing children were 7-year-olds from a previous study (Jackson, in preparation). In Figure 1 one can see that the older children obtained higher ToM scores than the younger children in all but the nonnative group. Also, the older DSP and hearing children obtained similar ToM scores, and these were higher than for all the other groups. A 2 (Young vs. Old) 5 (DSP vs. Nonnative vs. HIU vs. Oral vs. Hearing) ANOVA with ToM score as the dependent variable was conducted. There was a main effect of age, F(1, 112) 23.8, MSE.79, p.001, with the older children obtaining higher ToM scores than the younger children (see Figure 1). There was also a main effect of group, F(4, 112) 3.33, MSE.79, p.05). Scheffé post-hoc tests revealed no significant differences between the groups; this is possibly due to the large differences in sample size between the groups even though the Scheffé test can ac- Figure 1 ToM score by age and group. Young children up to and including children six years old. Old included children seven years old and over. DSP deaf children with signing parents; HIU children in hearing-impaired unit. commodate unequal sample sizes (Sheskin, 1997). The interaction between age and group failed to reach significance, F(4, 112) 2.04, MSE.79, p.09. Receptive language ability (BSL score) was analyzedbya2(young vs. Old) 3 (DSP vs. Nonnative vs. HIU) ANOVA, which revealed a significant main effect for age, with the older children scoring more highly than the younger children, F(1, 30) 8.63, MSE 80.9, p.01. There was also a main effect of language modality, F(2, 30) 3.68, MSE 80.9, p.05, but no interaction. As above, post-hoc tests produced no group differences. To allow a direct comparison to be made between DSP and hearing children on ToM ability, I converted raw BSL and BPVS scores to standard scores and matched the two groups for age and standard score. Oral deaf children were excluded because the standard scores were too low to be matched. With ToM score as the dependent variable, the assumption of homogeneity of variances was violated; consequently, the p value was reduced to.025. A t test revealed that the hearing children had significantly better ToM scores than the DSP children (2.6 vs. 1.6, respectively), t(18) 2.19, p.025. Second-Order Task Results Six signing deaf children (all from deaf schools) and 10 hearing children (all age 4 or younger) failed one or more control question on the second-order task. None of these children passed the test question; they were

12 172 Journal of Deaf Studies and Deaf Education 6:3 Summer 2001 Table 4 Mean score for each executive function task Task 3 Task 4 Nonpers. Pers. Nonpers. Pers. Task 1 Task 2 errors errors errors errors Deaf children (n 49) Hearing children (n 24) Autistic children (McEvoy et al. study) Nonpers. nonperseverative; pers. perseverative. For Tasks 1 and 2, the mean score refers to the number of correct searches. One oral deaf child did not complete the EF tasks. not excluded from the analysis but included and classed as failers. Of the 26 deaf children (14 oral deaf children and 11 children from deaf schools [5 DSP children and 7 nonnative signers]) who were given the second-order task, 22% (6, that is, 3 oral deaf and 3 deaf school) passed, a figure comparable to the 25% pass rate of the hearing preschoolers. The first-order ToM scores of the passers were significantly greater than those of the failers (1.8 and 0.9, respectively), t(24) 2.0, p.05. The passers were also older than the failers (7;11 years vs. 6;8 years, ns). Executive Functions From the 120 hearing children, a subset of 24 (six 4-, 5-, 6-, and 7-year-old children) were given the Raven s Coloured Progressive Matrices and EF tasks approximately 1 year after completing the ToM task battery. The mean age of this hearing subset was 6 years, 11 months, comparable to the oral, DSP, and nonnative signers but younger than the HIU children. One oral deaf child refused to participate in tasks 3 and 4, so that child s score was excluded from the analysis. Table 4 shows the descriptive score for each task along with the scores for children with autism (McEvoy et al., 1993). The children with autism made many more perseverative errors than either the deaf or hearing children did in this study. For EF Tasks 3 and 4 and the overall EF score, the data for all groups (DSP, nonnative, HIU, oral, and hearing children) were analyzed using MANOVA. No group differences were revealed. EF performance did not correlate with ToM score for the deaf children (r[49].19, ns) or hearing chil- dren (r[24].11, ns). This analysis was not appropriate for Task 2 because the data were not spherical. A nonparametric test, however, revealed that the hearing children scored significantly higher than the deaf children (U 362.5, p.001). Further analysis of the deaf individuals revealed that performance on Task 2 did not differentiate between control question passers and failers (t(49).91, ns). Data for Task 1 were not analyzed because none of the hearing participants made any errors on this task. Discussion The main aim of this study was to examine the relationship between ToM performance and receptive language and within this to investigate ToM performance in four groups of deaf children compared with hearing preschoolers. For nonnative, HIU, and oral deaf children, receptive language and ToM score were positively correlated, and the significant relationship between age and ToM in the oral deaf group was underpinned by language. Conversely, in the DSP and hearing groups, the positive correlation between receptive language and ToM score was underpinned by age, to the extent that when age was partialed out, the relationship decreased to almost zero. That ToM improvements are related to receptive language skills in the groups of children where there has been a delay in access to language is in keeping with my hypothesis. It suggests that properties specific to language are required for ToM development, over and above age-related experiences and neurological maturation, as recently suggested by Remmel, Bettger, and Weinberg (2000).