ASSESSING RELATIONAL LEARNING DEFICITS IN PERSPECTIVE-TAKING IN CHILDREN

：芭芭拉

ASSESSING RELATIONAL LEARNING DEFICITS IN PERSPECTIVE-TAKING IN CHILDREN WITH HIGH-FUNCTIONING AUTISM SPECTRUM DISORDER
Ruth Anne Rehfeldt, Jeffrey E Dillen, Megan M Ziomek, Rhonda K Kowalchuk. The Psychological Record. Gambier: Winter 2007. Vol. 57, Iss. 1; pg. 23, 25 pgs

Abstract (Summary)
Because autism spectrum disorder is characterized by deficits in the ability to form reciprocal social relationships (Klin, Volkmar, & Sparrow, 1992), some researchers have suggested that deficits in perspective-taking may be closely tied to, if not the basis of, the social deficits commonly observed in autism. In order to further demonstrate the possibility that perspective-taking is generalized O p e r ant behavior, future research should expose participants to the same protocol as was used in Experiment 2, but also test for generalization to novel stimuli and novel tasks, particularly in real-world social situations that require the speaker to change his or her perspective with regards to different references of person (Heagle & Rehfeldt, 2006).

：芭芭拉

Full Text (6903  words)

Copyright The Psychological Record Winter 2007

[Headnote]
Perspective-taking, or the ability to demonstrate awareness of informational states in oneself and in others, has been of recent interest in behavioral psychology. This is, in part, a result of a modern behavioral approach to human language and cognition known as Relational Frame Theory, which views perspectivetaking as generalized O p e r ant behavior based upon a history of reinforcement for relational responding. Previous lines of research have developed a behavioral protocol for assessing relational learning deficits in perspective-taking and have implicated the lack of perspective-taking as a basis for the social deficits observed in children with autism. However, no empirical investigations have been conducted on relational learning deficits in perspective-taking with autistic populations. The present pa-per reports 2 experiments that investigated whether children with autism spectrum disorder demonstrated relational learning deficits in a perspective-taking task as compared to their age-matched typically developing peers. We also investigated whether accuracy in perspective-taking correlated with scores on standardized instruments commonly used in the assessment of autism spectrum disorder, and whether relational responding in perspective-taking improves following a history of reinforcement for such responding. Results of Experiment 1 demonstrated statistically significant differences in errors as a function of type of relation, while visual inspection revealed that participants with autism spectrum disorder made more errors than typically developing children on 2 of the 3 types of relations examined. Results of Experiment 2 illustrated that a history of reinforced relational responding improved performance on the perspective-taking task.


Traditional developmental psychologists have defined perspectivetaking as an individual's awareness of informational states in oneself and in others (Baron-Cohen, 1995; Premack & Woodruff, 1978). Frequently investigated under the rubric of "Theory of Mind," results from developmental studies have been interpreted to suggest that children do not demonstrate perspective-taking abilities until they reach certain developmental milestones. For example, in a study by Dixon and Moore (1990), 5-, 7-, and 10-year-old children were read a series of short stories about a child emptying his toy box. The stories varied in terms of the boy's intentions for emptying the box and the information his mother had about his intentions. Only the older children were able to correctly identify discrepancies in the boy's and the mother's perspectives on his informational states, thus demonstrating perspective-taking skills. A similar study by Taylor, Cartwright, and Bowden (1991) required children to respond to a series of questions regarding the knowledge possessed by several characters in a story about the contents of a picture. Responding correctly to the questions required the children to change perspectives between the characters in the story. Children of 4 years of age were unable to pass the test, but children of 6 years of age did pass the test but made more errors than adult participants. Research by Baren-Cohen, Tager-Flusberg, and Cohen (2000), which exposed children to five levels of understanding about the informational states in other people, found that children excel on perspective-taking tasks once they are approximately 5 years of age. This body of findings contributes to the notion that perspective-taking skills emerge over the course of typical child development.

Viewing a situation from the perspective of another individual would seem to contribute greatly to an individual's success in social situations. For example, reciprocal conversation, coO p e r ative play, and the display of sympathy and empathy for others are all social abilities that require an effective repertoire of perspective-taking. Because autism spectrum disorder is characterized by deficits in the ability to form reciprocal social relationships (Klin, Volkmar, & Sparrow, 1992), some researchers have suggested that deficits in perspective-taking may be closely tied to, if not the basis of, the social deficits commonly observed in autism. Indeed, several studies have found that individuals with autism show marked deficiencies on perspective-taking tasks. A task known as the "Sally Anne task" (see Wimmer & Perner, 1983) was administered to typically developing children, children with autism, and children with Down syndrome in a study conducted by Baron-Cohen, Leslie, and Frith (1985). Frequently used in the study of perspective-taking, this task involves two dolls named Sally and Anne. The task begins with Sally placing a marble in a basket and then leaving the scene, at which time Anne enters the scene, removes the marble from the basket, and places it in a box. Sally returns to the scene, and participants are asked where Sally would look for the marble. Baron-Cohen et al. (1985) found that children with autism showed considerable difficulties with the task, whereas typically developing children and children with Down syndrome could easily answer the experimenter's questions about the Sally doll's perspective (the reader is also referred to Baren-Cohen, 1-9-8-9). LeBlanc, Coates, Daneshvar, Charlop-Christy, Morris, & Lancaster (2003) similarly found that children with autism were strikingly deficient in their performance on a version of the Sally Anne task. The children were only able to pass the task after intensive behavioral intervention involving video modeling and positive reinforcement, and even then, the children's generalization of perspective-taking skills to new situations was limited. Finally, Dawson and Fernald (1987) found that the performance of children with autism on a perspective-taking task was positively correlated with the children's social competence, as measured by two standardized scales. In other words, those children who performed worse on the task were clinically evaluated as being less socially competent.

Although the topic of perspective-taking is not new to developmental psychology, behavior analysts have only recently begun to investigate the topic both theoretically and empirically. A behavioral perspective on perspective-taking is important, for rather than assuming that a perspective-taking repertoire emerges as a function of development, behavior analysts would contend that specific learning histories give rise to the higher order skill. It thus follows that establishing the necessary learning history may help remediate deficits in perspective-taking for persons with autism and related disorders. Relational Frame Theory (RFT), a contemporary behavioral account of human language and cognition (Hayes, Barnes-Holmes, & Roche, 2001), has recently inspired theoretical analyses and experimental investigations of perspectivetaking. Proponents of RFT assert that derived relational responding constitutes the basis of much, if not all, of complex human behavior, including perspective-taking. RFT contends that verbal humans come to display arbitrarily applicable relational responding via a history of responding to multiple exemplars. In other words, an individual may experience specific reinforcement for responding relationally to some subset of stimuli, and then prove capable of responding relationally with other stimuli in the absence of reinforcement. Relational responding, then, can be regarded as a generalized, overarching response class (Hayes, Fox, Gifford, Wilson, Barnes-Holmes, & Healy, 2001). With regards to perspective-taking more specifically, RFT proponents would suggest that perspective-taking is a form of generalized O p e r ant responding involving "deictic" relations, or relations between stimuli that cannot be traced to the formal dimensions among the stimuli (Barnes-Holmes, Hayes, & Dymond, 2001). The perspective-taking repertoire emerges following a reinforced history of responding relationally to questions such as, "What would you do if you were me," "what are you doing then," "what will I be doing there," "what would you do there if you were me, " and other questions which require the speaker to change perspective between different references of person (i.e., I versus you), place (i.e., here versus there), and time (i.e., now versus then). Thus, Barnes-Holmes et al. (2001) suggest that learning to respond appropriately to questions that require the child to change perspective is critical in establishing frames that specify the relationship between stimuli in terms of the perspective of the speaker, the frames of I-You, Here-There, and Now-Then being the most critical. Barnes-Holmes et al. (2001 ) suggest that the relational properties of the frames of I versus You, Here versus There, and Now versus Then are abstracted through learning to talk about one's own perspective in relation to the perspectives of other individuals (Barnes-Holmes et al., 2001). Thus, a reinforced history of relating I versus You, Here versus There, and Now versus Then may lead to the emergence of a sophisticated repertoire of generalized perspective-taking.

An innovative study which paved the way toward an understanding of perspective-taking as derived relational responding was reported by McHugh, Barnes-Holmes, and Barnes-Holmes (2004). These researchers report the use of a comprehensive protocol for evaluating perspectivetaking skills in terms of the three frames of I-you, here-there, and nowthen. Coined the Barnes-Holmes protocol, this protocol presents a variety of relational perspective-tasks involving sim-ple, reversed, and double reversed I-you, here-there, and now-then relations. The protocol was administered to a large group of participants spanning fve age groups, ranging from early childhood through adulthood. The protocol was originally administered in a conversation format between the experimenter and the participant, in which the participant had to respond relationally to correctly answer questions such as, "I have a green brick and you have a red brick. If I was you and you were me, which brick would you have? Which brick would I have?" No consequences were delivered for correct or incorrect responses. It was found that errors on the task decreased as a function of the participant's age. These findings lend support to the notion that derived relational responding may be the basis of a perspectivetaking repertoire, as well as set the stage for further research on relational learning and perspective-taking.

Because perspective-taking deficits have been identified in populations of persons with autism, a worthwhile endeavor would seem to be to determine if children with autism spectrum disorder display deficits in relational learning on the Barnes-Holmes protocol, as compared to their typically developing peers. If indeed deficits in relational learning are identified, exposing such individuals to a reinforced history of responding relationally to I versus you, here versus there, and now versus then may well prove to ameliorate these deficits and permit for the generalization of perspective-taking to other perspective-taking tasks or situations. The purpose of Experiment 1 was to determine if children with autism spectrum disorder, specifically, high-functioning autism or Asperger syndrome, would perform significantly worse on the Barnes-Holmes protocol than their age-matched typically developing peers. We also investigated whether accuracy on the Barnes-Holmes protocol correlated with scores on standardized instruments commonly used in the assessment of autism spectrum disorder. If in fact accuracy on the protocol was shown to correlate with assessment scores, this would suggest that relational learning deficits in perspective-taking may underlie the social deficits commonly observed in persons with autism. The purpose of Experiment 2 was to determine if in fact performance on the Barnes-Holmes protocol improved following specific reinforcement for responding relationally. If so, it could be argued that perspective-taking involves derived relational responding.

：芭芭拉

☆Experiment 1

Method

Participants

A nonprobability convenience sample was used to obtain two groups of 9 children to serve as participants. Participants were recruited via newspa-per ads and flyers distributed at local schools and clinics in the greater southern Illinois region. Parents and children were financially compensated for their time and travel. The experimental group consisted of 9 males ranging in chronological age from 6 years and 8 months to 13 years and 4 months. All had been previously diagnosed with highfunctioning autism or Asperger syndrome; proof of diagnosis was provided upon the parent and child's arrival for the experiment. Participants in the control group consisted of 9 typically developing children (5 females and 4 males) ranging in chronological age from 6 years and 5 months to 13 years and 8 months. Control participants were matched to experimental participants on the basis of three developmental age bands: Middle childhood (age range 6-8 years), late childhood (age range 9-11 years), and adolescence (age range 12-14 years). This was done in attempts to create groups that were generally homogeneous in chronological age. Table 1 shows diagnosis, chronological age, and developmental stage for each experimental participant and a matched control participant. Children who did not read at their grade level were screened out of the experiment via discussion with parents and practice trials prior to the experiment.

Setting and Apparatus

Experimental sessions were conducted in quiet, secluded areas of either the Center for Autism Spectrum Disorders or the Rehabilitation Institute at Southern Illinois University. The perspective-taking task was presented upon a laptop PC and was created in Microsoft® PowerPoint® with program macros controlled by Microsoft® Visual Basic Editor. The program was created by Ruth Anne Rehfeldt and Jeffrey Dillen. All aspects of the task were automated. Participants were allowed brief breaks from the task at any time, during which they engaged in a fun activity with the experimenter (i.e., jumping on a trampoline, playing a computer game, etc.). While their child was completing the task, the child's parent completed two clinical assessment interviews with a graduate student in a nearby room.

Procedure

Clinical assessments. Two clinical assessments, the Vineland Adaptive Behavior Scales - Interview Edition (Sparrow, Balla, & Cicchetti, 1984) and the Social Communication Questionnaire (SCQ) - Current Form (Rutter, Bailey, & Lord, 2004) were administered to one parent of each participant by a Master's level graduate student trained in clinical interviewing and in the scoring and interpretation of both instruments. All interviews were videotaped. The Vineland Adaptive Behavior Scales - Interview Edition was designed to be administered in a semistructured interview format to a parent or caregiver of an individual between the ages of birth and 18 years and 11 months or an adult with low levels of cognitive functioning. The Vineland scales measure an individual's adjustment to the demands of everyday life, providing a profile of those skills in which an individual habitually engages on a day-to-day basis (Sparrow et al., 1984). Scores are provided in each of the following domains: Communication, Daily Living Skills, Socialization, Motor Skills, and Maladaptive behaviors. This study assessed participants' functioning levels in the first three domains only. The Vineland scales provide scores that are indicative of the chronological age at which the individual is currently functioning in each domain. The SCQ is a parent-report screening tool that provides a dimensional measure of autism spectrum disorder symptomatology, providing a cutoff score that indicates the likelihood that an individual has an autism spectrum disorder (Rutter et al., 2004). The score can also be used as an indication of the severity of ASD symptomatology. The SCQ Current Form is concerned only with behavior that has occurred during the past 3 months of the child's life, and assesses three areas of functioning: Reciprocal Social Interaction; Communication; and Restricted, Repetitive, and Stereotyped Patterns of Behavior. A cutoff score of 15 or greater suggests that the individual may have an autism spectrum disorder. Both assessment interviews took between 1-2 hours total.

The Vineland Adaptive Behavior Scales - Interview Edition is to be conducted over the course of a semistructured interview with the interviewer asking a series of general open-ended questions. Alternatively, the SCQ requires the parent to simply respond in the affirmative or negative as to whether or not their child has displayed a particular symptom in the last 3 months. For this reason, reliability was established for the Vineland interviews only. Interobserver agreement was calculated for 50% of the interviews by having a second Master's level graduate student who was also trained in clinical interviewing skills and in the scoring and interpretation of the Vineland Scales independently watch a videotaped interview and score items on the behavior scales. The scores were then compared on an item-by-item basis to the scores derived from the interviewer. An agreement was scored if both the interviewer and the observer scored an item the same. Percentage of agreement was obtained overall and for all three domains separately. Interobserver agreement was calculated by dividing the total number of agreements by the total number of agreements plus disagreements, multiplied by 100. Interobserver agreement for the Communication domain was 95% (range 86%-100%); interobserver agreement for the Daily Living Skills domain was 95% (range 91%-97%); and interobserver agreement for the Socialization domain was 93% (range 70%-100%). Overall interobserver agreement was 94% (range 88%-96%).

Perspective-taking. All participants were exposed to the same procedure. A modified version of the Barnes-Holmes protocol, as reported by McHugh et al. (2004), was presented in an automated format. The protocol was presented as a test, with no feedback presented for correct or incorrect responses. However, the experimenter provided intermittent general praise for compliance with the task (i.e., "good working;" "you are doing a great job paying attention."). Prior to the experiment, participants were presented with a series of practice trials to ensure that they had sufficient reading comprehension skills to complete the task. Practice trials consisted of word problems that would not be presented in the perspectivetaking task but were in the same format as the questions in the perspectivetaking task (e.g., "if the sky is yellow and the sun is blue, what color is the sky? What color is the sun?"). Participants were told that their job was to read the question and select one of two command boxes below the question to select their answer to the question. They were then oriented to the questions presented on the screen and to O p e r ating the computer mouse to answer the questions. Participants were required to read the practice trials aloud. Participants who did not read the practice trials correctly or who answered the questions incorrectly did not complete the perspectivetaking task. Although some children were dismissed from the experiment because of insufficient reading skills, all of the children in the reported study read at grade level and displayed no reading comprehension problems. The participants were required to intermittently read the trials aloud as they worked through the perspective-taking task to ensure this.

The modified version of the Barnes-Holmes protocol that was used in this study consisted of 57 total trials. Each trial consisted of two questions (e.g., "which brick do I have?" "Which brick do you have?"). A participant had to answer both questions correctly in order for the trial to be scored as correct. If a participant asked the experimenter a question during the task, the experimenter reminded the participant that his or her job was to answer the questions and he or she was not allowed to help the participant. As was the case in McHugh et al. (2004), three types of relations were presented in the protocol. These included sim-ple relations, reversed relations, and double reversed relations. Within each of these three types of relations were trials that evaluated responding to three different perspective-taking frames (l-you, here-there, and now-then). Eight trials for the sim-ple relations were included in the protocol, including 2 l-you, 2 here-there, and 4 now-then trial types. Thirty-six trials for the reversed relations were presented, including 8 l-you, 12 here-there, and 16 now-then relations. Thirteen trials for the double reversed relations were presented, including 4 l-you/here-there and 9 here-there/now-then trial types. The number of trials for each relation and each trial type followed closely that used by McHugh et al. (2004). Trials for all relation types and trial types were presented in a random order. Table 2 shows the questions that were presented for each of the three relations and for each of the trial types within each relation tested. (The reader is also referred to McHugh et al., 2004, for additional detail on the Barnes-Holmes protocol. Correct answers to test trials for the sim-ple relations required responses that were identical to the arrangements specified in the question. Correct answers to test trials for the reversed relations required the participant to reverse the l-you, now-then, or here-there arrangements specified in the question. Correct answers to test trials for the double reversed relations required the participant to simultaneously reverse the l-you and here-there or here-there and now-then arrangements specified in the question. Correct answers for each trial type for each relation type are shown in Table 2.

Figure 1 shows on-screen representations of a sim-ple l-you trial, a reversed now-then trial, and a double reversed here-there/now-then trial. Participants were required to click the computer mouse upon one of the two command buttons presented for each question in order to indicate their answer. Participants' responses to one of the two command boxes presented for the second question advanced to the next trial. The left-right location of the correct and incorrect command boxes was randomly determined across all trials.

Results and Discussion

Age Equivalence and SCQ

Shown in Figures 2-4 are the scores on the SCQ and the Vineland Adaptive Behavior Scales. Figure 2 shows that the mean score on the SCQ for the experimental group was 20.78 (SD = 6.02), and the mean score for the control group was 3.22 (SD = 2.17). Figure 3 shows that the mean overall age equivalence on the Vineland Adaptive Behavior Scales was 79.44 months for the experimental group (SD = 10.17) and 119.89 months for the control group (SD = 40.42). Figure 4 shows that the mean age equivalence on the communication subdomain of the Vineland was 80.79 months for the experimental group (SD =14.29; and 111 months for the control group (SD = 35.90). These results suggest that the two groups differed substantially on these measures, although considerable variability was present in the results for the control group.

Perspective- Taking

Shown in Figures 5 and 6 is the mean percentage of errors on the sim-ple, reversed, and double reversed relations, for the control group (Figure 5) and the experimental group (Figure 6). Figure 5 shows that there was considerable variability across the test trials for the different relations for the control group. It can be seen from the figure that several participants (MP, CG, and ER) made the fewest errors on the test trials for the sim-ple relations, and the most errors on test trials for the double reversed relations. SG, AS, and KG2 made the most errors on test trials for the reversed relations. Figure 6 shows that the participants with autism typically made fewer errors on test trials for the sim-ple relations, and more errors on test trials for the reversed relations. NW1 (13 years old), an outlier for the group, made fewer errors across nearly all test trials than the other experimental participants, mak-ing no errors on test trials for the sim-ple relations and fewer than 10 errors on test trials for the reversed relations. Overall, the figures show that most of the participants in both groups made more errors on the reversed relations than the sim-ple relations, but it wasn't necessarily the case that the most errors were made on test trials for the double reversed relations. No particular pattern of errors between the l-you, here-there, now-then, l-you/here-there, and here-there/now-then trial types was observed.

Shown in Figure 7 is the pattern of overall mean percentage errors for the sim-ple, reversed, and double reversed relations, plotted separately for the experimental and control participants. The figure shows that the experimental participants committed more errors on test trials for the reversed and double reversed relations; this difference is most marked for the reversed relations. A 2 × 3, between- by within-subjects analysis of variance was conducted on group (experimental, control) by relation (sim-ple, reversed, double reversed) to test for differences between mean percentage errors for experimental vs. control groups across all test trial types, differences between mean percentage errors for the sim-ple vs. reversed vs. double reversed relations for both groups, and for an interaction effect between group and relation (sim-ple, reversed, and double reversed). The main effect for relation was found to be statistically significant, F(2, 15,) = 12.870, p = .001, by the Wilks' Lambda Criterion. Pairwise comparisons revealed a significant difference between the overall mean percentage errors on the sim-ple vs. reversed relations, F(1) = 18.520, p = .000, indicating that substantially more errors were made on the reversed relations than the sim-ple relations for both groups of participants. Pairwise comparisons revealed a nearly significant difference between the mean percentage errors on the sim-ple vs. double reversed relations, p = .045, but this finding was found to be nonsignificant after adjusting for inflated Type I error rate.

The interaction between group and relation (sim-ple, reversed, double reversed) approached statistical significance, F (2,15) = 3.354, p = 0.063, by the Wilks' Lambda Criterion. Because Figure 7 revealed substantial differences between the two groups' performances on the reversed relations, differences between the groups on reversed trials were more closely examined. Figure 8 shows the pattern of mean percentage errors for the trial types within the reversed relations (l-you, here-there, and nowthen) for the two groups. Although the comparisons were not statistically significant, the figure shows that the experimental group made more errors on all reversed relation test trial types than the control group.

Relationship Between Perspective-Taking and Clinical Assessments

Correlations were calculated between scores on the perspective-taking task and the SCQ, overall age equivalence as measured on the Vineland Adaptive Behavior Scales, and age equivalence on the communication, daily living, and socialization domains of the Vineland. A modest correlation between the percentage of errors on the now-then reversed relations and the Daily Living Skills domain of the Vineland was found, r(18) = -.512, p < .05, indicating that lower age equivalence in daily living skills was associated with a greater number of errors on the now-then reversed relations.

Our ability to draw firm conclusions from these findings is limited because of the small sample size and low statistical power; in addition, within-group variability, particularly within the control group, may have masked further differences between the groups. Nonetheless, some interesting trends were apparent in the data, one of the most important of which was the difference between the number of errors on the reversed, versus sim-ple, relations. This difference was most pronounced for the group of participants with autism. These results suggest that the test trials for the reversed relations required a more complex form of relational responding than test trials for the sim-ple relations: Test trials for the reversed relations required the derivation of the deictic relations. The difference in errors between the sim-ple and double reversed relations also approached statistical significance, but fewer errors were observed across all participants on test trials for the double reversed relations relative to the reversed relations. This finding is somewhat surprising, as the double reversed relations required a more complex form of relational responding than the reversed relations. McHugh et al. (2004) did obtain such systematic differences in errors between the sim-ple, reversed, and double reversed relations, with errors most likely on test trials for the double reversed relations. However, correct responses for test trials assessing the double reversed relations were the same as those required for test relations assessing sim-ple relations. Consider, for example, the following problem: "I am sitting here on the blue chair and you are sitting there on the black chair. If I was you and you were me and if here was there and there was here." This problem can be answered correctly simply by reading the first sentence. Thus, while McHugh et al. (2004) contend that the double reversed relations require a more complex form of derivation than the sim-ple and reversed relations, it is possible that participants were not responding relationally at all when presented with double reversed relational tasks. Needed are analyses to confirm that participants' performances on double reversed tasks are in fact indicative of relational responding rather than a participant's reading the first part of the problem only.

A modest correlation was found between participants' performance on the perspective-taking task and scores on one domain of the Vineland Adaptive Behavior Scales. This correlation was found between scores on the now-then reversed relations and age equivalent scores on the Daily Living scale. It would have been expected that performance on the perspective-taking task would correlate with scores on the Communication or Socialization domains, as perspective-taking deficits would seem to contribute to an individual's overall adjustment in such areas. Nonetheless, the present finding alludes to the role of derived relational responding and perspective-taking in adaptive behavior. Future research should explore this relationship further.

The group differences detected in this study raise the question whether indeed individuals with high-functioning autism are deficient on relational learning tasks; more specifically, relational learning tasks in perspective-taking, relative to their age-matched peers. Results from this experiment suggest that such group differences do exist. One might be quick to attribute these differences to differences in reading ability, which might be expected to be lower in children with autism. However, efforts were made to ensure that all of the participants could read and comprehend the practice and test questions, and no obvious delays in reading comprehension were observed on the part of any of the children at any time over the course of the experiment. If there were even minor questions about a child's reading ability, he or she was eliminated from the study. Thus, substantial differences in reading ability between the two groups seems unlikely. Differences in cognitive functioning, however, may well have separated some of the participants with high-functioning autism from their age-matched peers, and lower levels of cognitive functioning would undoubtedly create differences in performance on any form of relational learning task. Thus, the group differences observed may be caused by differences in overall cognitive ability, rather than differences in perspective-taking more specifically. Future research aimed at comparing perspective-taking in children with autism to children without autism might match participants with autism with a typically developing peer on IQ or some other equivalent measure of cognitive functioning.

The present results suggest that perspective-taking does involve derived relational responding, as such a repertoire was required of the task. These results are thus consistent with those of McHugh et al. (2004). In order to further demonstrate a relational learning conceptualization of perspective-taking, it is necessary to show that performance on the Barnes-Holmes protocol is indeed sensitive to reinforcement contingencies, and that the relations can in fact be shaped up through exposure to reinforcement. Such a finding would suggest that perspective-taking is a class of O p e r ant behavior. The purpose of Experiment 2 was, therefore, to further evaluate the notion that perspective-taking may emerge via a history of reinforced relational responding. In this experiment, the Barnes-Holmes protocol was administered to 2 typically developing children who had served in Experiment 1. We specifically questioned whether the sim-ple, reversed, and double reversed l-you, here-there, and now-then relations could be taught via exposure to reinforcement contingencies. Furthermore, we assessed whether the double reversed relations truly emerged as relational O p e r ants for the 2 participants, which would not be the case if the participants responded to the double reversed test trials by answering only the first part of each question. In order to document the emergence of a relational O p e r ant, Hayes and Quinones (2005) suggest depicting the increasing probability of relation-consistent responding as training advances. Training data in the present experiment were thus analyzed in accordance with the recommendations of these authors. Unlike Experiment 1 feedback was delivered for correct and incorrect responses. Following mastery of each set of relations, a posttest, conducted under extinction, was administered. No published study to date has reported the use of this protocol in teaching perspective-taking skills.

：芭芭拉

☆Experiment 2

Method

Participants

Two typically developing children with no known disabilities participated. Both had served as control participants in Experiment 1. CY was a male of 9 years and 2 months of age at the time of his participation. AS was a female of 10 years and 4 months of age at the time of her participation. Neither participant had scored higher than 80% on the sim-ple, reversed, or double reversed test trials in Experiment 1.

Setting and Apparatus

The setting and apparatus were identical to those of Experiment 1.

Procedure

Participants' scores on the sim-ple, reversed, and double reversed relations served as their pretest scores for the particular relations in Experiment 2. Participants were first trained in the sim-ple relations; once a mastery criterion of 90% correct was achieved, a posttest for the sim-ple relations was presented. If the participant met the criterion for inferring the emergence of the sim-ple relations on the sim-ple relations posttest, training of the reversed relations was introduced. If the participant did not meet criterion for inferring the emergence of the sim-ple relations on the sim-ple relations posttest, training of the sim-ple relations was again instated, and the sim-ple relations posttest was repeated once the relations were again shown to be mastered. This process was repeated for the reversed and double reversed relations. The same number of trials for each relation type and for each trial type within each relation type was presented as occurred in Experiment 1. A mastery criterion of at least 90% correct on the pretests and posttests (7/8 test trials correct for the sim-ple relations, 32/36 correct for the reversed relations, and 11/13 correct for the double reversed relations) was taken as indicative of the emergence of the particular relations. No feedback was presented during pretests and posttests.

All trials presented in the pretests, posttests, and training phases were identical to those presented in Experiment 1. The particular trial types were presented within a random order within each pretest, training phase, and posttest. As was the case in Experiment 1, both questions presented had to be answered correctly in order for the trial to be scored as correct. During training, a variety of 3-s animation clips were presented as reinforcers following correct trials, whereas incorrect trials produced a slide which read, "Try again," and the respective trial was then repeated.

Results and Discussion

CY

Figure 9 presents CY's performance on the pretests, posttests, and training trials for the sim-ple, reversed, and double reversed relations. Although he initially did not meet criterion performance on the pretest for the sim-ple relations, CY required only one training block of eight trials to master the sim-ple relations, after which he responded with 100% accuracy on the sim-ple relations posttest. CY's pretest score for the reversed relations was 81%. He mastered the reversed relations in five 36-trial training blocks, but performed with only 36% accuracy on the first posttest. After six more 36-trial training blocks, CY demonstrated criterion performance on the second posttest for the reversed relations. CY scored 46% accurate on the pretest for the double reversed relations, and required only two 13-trial training blocks to master the relations, after which he performed with 100% accuracy on the double reversed relations posttest.

Shown in Appendix 1 is the percentage of correct trials per training session for CY. These scores are shown separately for each frame (l-you, here-there, and now-then), for the three relational tasks. In order to ascertain the strengthening of each relational O p e r ant over training, it would be expected that the percentage of correct responses would be low for each frame early on in training, but increasing as training advances (Hayes & Quinones, 2005). For the sim-ple and reversed relations, it appears that the three frames were sufficiently strong at the beginning of training. Thus, CY's failure to demonstrate mastery during test sessions may have been caused by the difficulties associated with completing the protocol in the absence of feedback. The double reversed relations, on balance, clearly emerged over the course of training, as CY demonstrated 50% and 56% percent correct during his first training session on training trials for the l-you/here-there and here-there/now-then frames, respectively. These latter results suggest that CY was in fact responding relationally on double reversed relations trials.

AS

Shown in Figure 10 is AS's performance on the pretests, posttests, and training trials for the sim-ple, reversed, and double reversed relations. The figure shows that the participant performed with 50% accuracy on the pretest for the sim-ple relations, and then mastered the relations in three training blocks. She did not meet criterion performance on the second posttest for the sim-ple relations, but did following two more training blocks. AS performed with 47% accuracy on the pretest for the reversed relations, and required 41 training blocks to master the reversed relations, after which time she demonstrated criterion performance on the posttest for the reversed relations. AS did not meet criterion performance on the pretest for the double reversed relations, and required only two training blocks to master the relations, after which she scored 100% accuracy on the posttest for the double reversed relations.

Shown in Appendix 2 is the percentage of correct trials per training session for AS. The appendix shows that all of the relational opérants strengthened considerably over the course of training for the sim-ple, reversed, and double reversed relations. AS responded with only 0% and 33% accuracy during his first training session for the l-you/here-there and here-there/now-then frames for the double reversed relations, but 100% during the second training session for these same frames. This suggests that by the conclusion of training, AS was in fact responding relationally on double reversed relations trials.

Results from Experiment 2 indicate that the ability to change perspective between l-and-you, here-and-there, and now-and-then can be established via a history of reinforced relational responding, as both participants demonstrated criterion performance on the posttest, conducted under extinction, following exposure to reinforcement contingencies for correct changes in perspective. Thus, this protocol was effective in establishing l-you, here-there, and now-then frames, which specified the relationship between stimuli in terms of the perspective of the speaker. Not surprisingly, the sim-ple relations were the easiest for both participants to master, and the reversed relations required the most training trials. These results are consistent with those of Experiment 1. Together, results from both experiments demonstrate that derived relational responding plays an evident role in perspective-taking. In order to further demonstrate the possibility that perspective-taking is generalized O p e r ant behavior, future research should expose participants to the same protocol as was used in Experiment 2, but also test for generalization to novel stimuli and novel tasks, particularly in real-world social situations that require the speaker to change his or her perspective with regards to different references of person (Heagle & Rehfeldt, 2006).

[Sidebar]
This investigation was supported by a research grant from TAP (The Illinois Autism Project, funded by Illinois Department of Human Services), awarded to Ruth Anne Rehfeldt. Portions of this pa-per were presented at the annual Southern Illinois University Phi Kappa Delta & Phi Kappa Phi Research Day, Carbondale, Illinois, April 2005, and at the annual meeting of the Association for Behavior Analysis, Chicago, Illinois, May 2005. We acknowledge Rocio Resales, Yors Garcia, and Holly L. Bihler for their help with data collection and subject running. In addition, we thank the Center for Autism Spectrum Disorders and the Rehabilitation Institute at Southern Illinois University for the use of space for conducting experimental sessions. Finally, we thank Steve Hayes for comments and suggestions on a previous version.

[Reference]
References
BARNES-HOLMES, D., HAYES, S. C., & DYMOND, S. (2001). Self and selfdirected rules. In S. C. Hayes, D. Barnes-Holmes, & B. T. Roche (Eds.), Relational frame theory: A post-Skinnerian account of human language and cognition (pp. 119-139). New York: Plenum.
BARON-COHEN, S. (1-9-8-9). The autistic child's theory of mind: A case of specific developmental delay. Journal of Child Psychology and Psychiatry, 30,285-297.
BARON-COHEN, S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge, MA: Bradford Books, MIT Press.
BARON-COHEN, S., LESLIE, A. M. & FRITH, U. (1985). Does the autistic child have a "Theory of Mind"? Cognition, 21, 37-46.
BARON-COHEN, S., TAGER-FLUSBERG, H., & COHEN, D. (2000). Understanding other minds: Perspectives form developmental cognitive neuroscience (2nd ed.). Oxford: Oxford University Press.
DAWSON, G., & FERNALD, M. (1987). Perspective-taking ability and its relationship to the social behavior of autistic children. Journal of Autism and Developmental Disorders, 17(4), 487-498.
DIXON, J. A., & MOORE, C. F. (1990). The development of perspective taking: Understanding differences in information and weighing. Child Development, 61, 1502-1513.
HAYES, S. C., BARNES-HOLMES, D., & ROCHE, B. (Eds.). (2001,). Relational frame theory: A post-Skinnerian account of language and cognition. New York: Plenum.
HAYES, S. C., FOX, E., GIFFORD, E. V., WILSON, K. G., BARNES-HOLMES, D., & HEALY, O. (2001). Derived relational responding as learned behavior. In S. C. Hayes, D. Barnes-Holmes, & B. T. Roche (Eds.), Relational frame theory: A post-Skinnerian account of human language and cognition (pp. 21-49). New York: Plenum.
HAYES, S. C., & QUINONES, R. M. (2005). Characterizing relational O p e r ants. Latinamerican Journal of Psychology, 3, 277-289.
HEAGLE, A., & REHFELDT, R. A. (2006). Teaching perspective-taking skills to typically developing children through derived relational responding. The Journal of Intensive Early Behavioral Intervention, 3, 8-34.
KLIN, A., VOLKMAR, F. R., & SPARROW, S. (1992). Autistic social dysfunction: Some limitations of the Theory of Mind hypothesis. Journal of Child Psychology and Psychiatry, 33, 861-876.
LEBLANC, L. &Aacute;., COATES, A. M., DANESHVAR, S., CHARLOP-CHRISTY, M. H., MORRIS, C., & LANCASTER, B. M. (2003). Using video modeling and reinforcement to teach perspective-taking skills to children with autism. Journal of Applied Behavior Analysis, 36, 253-257.
MCHUGH, L., BARNES-HOLMES, Y., & BARNES-HOLMES, D. (2004). Perspective-taking as relational responding: A developmental profile. The Psychological Record, 54, 115-144.
PREMACK, D., & WOODRUFF, G. (1978). Does the chimpanzee have a 'theory of mind'? Behavioral and Brain Sciences, 4, 515-526.
RUTTER, M., BAILEY, A., & LORD, C. (2004). SCQ: Social Communication Questionnaire. Western Psychological Services: Los Angeles, CA.
SPARROW, S., BALLA, D., & CICCHETTI, D. (1984). Vineland Adaptive Behavior Scales (Survey ed.). Circle Pines, MN: American Guidance Service.
TAYLOR, M., CARTWRIGHT, B. S., & BOWDEN, T. (1991). Perspective Taking and Theory of Mind: Do children predict interpretive diversity as a function of differences in observers' knowledge? Child Development, 62, 1334-1351.
WIMMER, H., & PERNER, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13(1), 103-128.

[Author Affiliation]
RUTH ANNE REHFELDT, JEFFREY E. DILLEN, MEGAN M. ZIOMEK, and RHONDA K. KOWALCHUK
Southern Illinois University