The production of tangential language during conversations was studied with people who have fragile X syndrome, autistic disorder, or mental retardation not caused by fragile X. Tangential language was found to be more prevalent among those with fragile X, compared to the control groups, especially within unsolicited comments. These results support our hypothesis that the tangential language seen in fragile X is not the result of either general developmental delay or undiagnosed autistic disorder. We offer a possible interpretation of these results based upon such phenotypic characteristics as social anxiety, hypersensitivity to social and sensory stimuli, and inhibitory control deficits.
Fragile X syndrome is considered to be the most common inherited cause of mental retardation, affecting an estimated 1 in 4,000 males (Turner, Webb, Wake, & Robinson, 1996). It is caused by a triplet repeat mutation on the X chromosome that interferes with the production of FMR-1 protein (Hagerman, 1996) and is associated with various nervous system anomalies, including increased volume of the caudate nucleus and the lateral ventricle (Reiss, Abrams, Greenlaw, Freund, & Denckla, 1995). In addition to physiological evidence of impaired nervous system development, individuals with fragile X syndrome typically present with a behavioral phenotype that includes both verbal and nonverbal anomalies.
Many of the atypical nonverbal behaviors seen in fragile X syndrome have been attributed to anxiety disorders, especially social anxiety (Bregman, Leckman, & Ort, 1988; Levitas, McBogg, & Hagerman, 1983). The socially avoidant behavioral profile associated with fragile X syndrome is characterized by gaze aversion; tactile defensiveness; overt turning away of the body during face to face social interaction; and stylized, highly ritualistic forms of greeting. People with fragile X syndrome generally try to avoid eye contact and find it very aversive when it is forced upon them (Bregman et al., 1988; Cohen et al., 1988; Cohen, Vietze, Sudhalter, Jenkins, & Brown, 1989; Maes, Fryns, Van Walleghem, & van der Berghe, 1993; Wolff, Gardner, Paccia, & Lappern, 1989). In addition, individuals with fragile X syndrome have been shown to demonstrate hyperactive behavior (Bregman, Dykens, Watson, Ort, & Leckman, 1987; Hagerman, 1991; Reiss & Freund, 1992), impulsive behavior, and poor self-control (Lachiewicz, Spiridigliozzi, Gullion, Ransford, & Rao, 1994).
In addition to this apparent relationship with social anxiety, many of the behaviors that characterize fragile X syndrome have been attributed to undiagnosed autistic disorder. Approximately 8% of individuals with autistic disorder have been shown by DNA analysis to be positive for fragile X syndrome (Li, Chen, Lai, Hsu, & Wang, 1993). In addition, several researchers have found that approximately 17% of individuals with fragile X syndrome also have a diagnosis of autistic disorder (Baumgardner, Reiss, Freund, & Abrams, 1995; Brown et al., 1986; Hagerman, Jackson, Levitas, Rimland, & Braden, 1986). Even those individuals with fragile X syndrome who have not been diagnosed with autistic disorder often exhibit many autistic-like behaviors (Hagerman, 1996).
Atypical speech and language production has also been shown to characterize the verbal behavior of males with fragile X syndrome (Howard-Peebles, Stoddard, & Mims, 1979; Levitas et al., 1983; Newell, Sanborn, & Hagerman, 1983; Paul, Cohen, Breg, Watson, & Herman, 1984; Wolf-Schein et al., 1987), and many clinicians and diagnosticians attribute this atypical language in males with fragile X syndrome to undiagnosed autistic disorder. Indeed, according to DSM-IV criteria (American Psychiatric Association, 1994), atypical language production is one of the hallmarks of a diagnosis of autistic disorder. Specific types of atypical language associated with autistic disorder include echolalia (Fay, 1969; Kanner, 1946; Prizant, 1983), prosodic abnormalities (Baltaxe, 1981; Goldfarb, Goldfarb, Braunstein, & Scholl, 1972), and neologisms (Volden & Lord, 1991). People with autistic disorder are also generally insensitive to conversational pragmatics. For example, they do not understand the facial or body cues that serve to modify spoken discourse, signal requests for additional information, and control turn-taking. Similarly, because they cannot distinguish between what they know and what their conversational partner knows, they will use pronouns with unclear referents and refer to specific people or events without providing necessary background information (Baltaxe, 1977; Baron-Cohen, 1988; Paul, 1987; Tager-Flusberg, 1989). One of the theories that has been proposed to explain atypical language production in autistic disorder is that such individuals exhibit “theory of mind” problems (Baron-Cohen, 1988), which result in their failure to appreciate the linguistic and contextual cues needed by their conversational partners for understanding the flow of a conversation.
Although atypical language is associated with both autistic disorder and fragile X syndrome, we have demonstrated that the types and prevalence of atypical language differ between individuals with fragile X syndrome and no autistic disorder and individuals with autistic disorder and no fragile X. A comparison of males with fragile X syndrome to chronological age (CA) and language-age matched individuals with autistic disorder or mental retardation not related to fragile X syndrome revealed that the males with fragile X syndrome produced significantly more perseverative language than did those without fragile X syndrome and that individuals with autistic disorder produced significantly more echolalic language than did those without autistic disorder (Sudhalter, Cohen, Silverman, & Wolf-Schein, 1990). We believe that not only does the predominant type of atypical language produced (e.g., perseveration vs. echolalia) differ between fragile X syndrome and autistic disorder, but also that the underlying mechanisms responsible for the production of the atypical language differ between these populations as well. Specifically, it is our belief that the language problems exhibited by males with fragile X syndrome are directly related to the hyperarousal and inhibitory processing deficits that have been associated with this syndrome. Males with fragile X syndrome can be very social and often exhibit a clear desire to interact with others. Because of an inherent inability to modulate sensory input in the same manner as those without the disorder, however, such social interaction may cause them to become overstimulated and hyperaroused, which they experience as being very uncomfortable. We believe it is the anticipation of such discomfort that leads to their socially avoidant behaviors.
This view is consistent with our clinical observations that the social demands of conversation often interfere with their production of language. Previous work in our laboratory has demonstrated an association between eye gaze and the probability of atypical language production in males with fragile X syndrome. We found a significant difference in the probability that during interactions with both familiar and unfamiliar adults, males with fragile X syndrome would produce atypical language when an adult looked at them while they talked compared to when the adult turned away (Sudhalter & Cohen, 1989). More recently, we found both behavioral and physiological indications that males with fragile X syndrome become more aroused during conversations than do males without fragile X syndrome but with otherwise similar developmental disabilities and that the heightened arousal elicited by making eye contact with these individuals is accompanied by an increased production of atypical language (Belser & Sudhalter, 1995).
Several models of language production that rely on excitatory and inhibitory neural processing have been described (Anderson, 1983; Dell, 1986; Dell, Burger, & Svec, 1997; Levelt, 1989), and we briefly review how the atypical language seen in males with fragile X syndrome can be explained within such a framework. Although inhibitory control deficits are certainly not unique to fragile X syndrome, we believe that the disruptive effects of such deficits are heightened when combined with the physiological hyperarousal that has been shown to accompany social interaction in this population.
Our model assumes the presence of a lexicon, a store of information about the words that comprise one's language (Levelt, 1989), which is arranged as a network of nodes containing both semantic and syntactic information. In addition, the network itself is arranged according to a system of semantic and associative relationships (Collins & Loftus, 1975; Noordman-Vonk, 1979; E. Smith & Medin, 1981; M. Smith, Theodor, & Franklin, 1983). An example of a semantic relationship is “boy–girl,” and an example of an associative relationship is “happy–birthday.”
When an individual is presented with a word, the lexical node corresponding to that word becomes activated. This nodal activation then causes a spread of activation to semantically and associatively related nodes. The activation of the lexical representation, and the associated spread of activation, is assumed to be an automatic process (Rafal & Henik, 1994) that is referred to as semantic priming. During conversation, words (as well as phrases and whole topics) are selected according to relative activation of lexical nodes.
There are at least two ways in which inhibitory processes play a necessary role in semantic priming within the lexicon. One is to limit the duration of nodal activation to ensure that previously activated nodes do not interfere with new input. Once the node associated with a “to be produced” word has been accessed, the activation associated with that node must rapidly decay in order to prevent its being continually re-selected. Thus, the failure of nodal activation to decay at a sufficient rate could result in the same word or thought being produced again. We have previously demonstrated a tendency for people with fragile X syndrome to inappropriately perseverate on words and topics (Sudhalter et al., 1990) and have suggested that the same mechanism might also account for that behavior (Belser & Sudhalter, 1995).
A second way that inhibitory processing influences semantic priming is to constrain the activation of semantically and/or associatively related nodes. The expected consequence of a disruption of this process would be the selection of highly associated words or thoughts, independently of the current conversational context. Because such associations may be personal or highly idiosyncratic, the words, phrases, or topics that are selected and introduced into the conversation may actually be very tangential to the current conversational thread and, consequently, disrupt the flow of discourse. In fact, it has been our clinical experience that, in addition to perseverative language, males with fragile X syndrome do produce more tangential language than do their peers without fragile X syndrome.
In order to clarify the nature of tangential language production by individuals with fragile X syndrome, in this study we have compared their conversations with those of individuals representing two other groups: mental retardation from causes other than fragile X syndrome and autistic disorder. Because we predict that the production of atypical language will be at least partially determined by changes in anxiety and arousal associated with the varying social demands within conversations, we examined participants' conversational language within the three mutually exclusive conversational–pragmatic categories of questions, responses, and comments.
We predicted that, in general, males with fragile X syndrome would produce more tangential language than would members of the other clinical groups and that the incidence of such atypical language would be related to the socially arousing properties of the specific conversational category. In addition, we predicted that conversational category would not affect the production of tangential language by individuals with autistic disorder because they, as a group, have been shown to be insensitive to conversational pragmatics (Baltaxe, 1977; Baron-Cohen, 1988; Tager-Flusberg, 1989; Wetherby, 1986). Such findings should provide additional evidence that the atypical language associated with fragile X syndrome differs both in its nature and its underlying cause from that seen in autistic disorder and other non-fragile X syndrome groups.
This study is part of a larger project designed to investigate linguistic competency of males with fragile X syndrome, and here we specifically address the tangential language produced by this population. The results of a companion study of disfluent speech, based upon the same conversational transcripts used here, have been presented in a separate report (Belser & Sudhalter, 2001).
Ten males with fragile X syndrome, 10 males with autistic disorder, and 4 males and 6 females with mental retardation due to other causes were each engaged in a 10- to 15-minute videotaped conversation. The diagnoses were made by experienced clinicians prior to the participants' recruitment for the study. All participants were tested for fragile X syndrome, and assignment to this group was made on the basis of a positive DNA analysis of a blood sample. The mental retardation group consisted of 1 male and 2 females with Down syndrome, and the rest of the group had mental retardation of unknown or unspecified etiology. The diagnosis of mild/moderate mental retardation was made following a review of intelligence test scores contained in their medical and educational records, and the diagnosis of autistic disorder was made by clinical staff trained in diagnosing autistic disorder using DSM-IV criteria.
The participant groups were matched as closely as possible for CA as well as for language age and for level of functioning (as measured by the communication age equivalent and the adaptive behavior composite score, respectively, of the Vineland Adaptive Behavior Scales—Sparrow, Balla, & Cicchetti, 1984). In order to be included in this study, each participant had to produce between 80 and 100 individual utterances during a conversation with an experimenter, exhibit a language age of between 4 and 8 years, and be within the mild–moderate range of delay. Table 1 presents the age, language age, and adaptive behavior composite score for participants by group.
Conversations were conducted with participants seated and facing their conversational partner, who was a member of the research staff. The conversational partner allowed each participant to control the conversation to the extent that the participant was willing and to determine the topics of discussion according to his or her individual interests. When participants failed to initiate topics or questions, they were asked questions by the partner in an attempt to elicit a sufficient amount of conversational language. These tended to be easy questions that either required a yes/no answer or requested specific information, usually involving school, friends, and any special activities that the participant had recently been involved with. The conversations were videotaped and subsequently transcribed using the CHAT format (MacWhinney, 1995).
The participant's language was divided into individual utterances, defined as the smallest units of language that comprise unitary thoughts. Utterances usually consist of at least a complete verb phrase, although the object of the phrase may only be inferred. There can be, and often are, several utterances within a single “turn” of conversation. In order to investigate the role of conversational–pragmatic category on the production of tangential language, we classified each utterance as belonging to one of the three mutually exclusive categories of question, response to a question, and unprompted comment. Tangential utterances within these categories were then identified.
A tangential question was defined as an utterance by the participant that requests information not appropriately related to the current topic. A tangential response was defined as an utterance by the participant that follows a request for information by the conversational partner but does not appropriately answer the partner's question and in which the associational link between question and answer is not clear. A tangential comment was defined as an utterance by the participant that neither requests information nor provides requested information and does not appropriately relate to the current topic. In each case, utterances were not classified as tangential if they were apparently prompted by a distracting stimulus or event within the participant's physical environment. They were also not coded as tangential if the previous topic had obviously been resolved, and the participant's utterance constituted an appropriate initiation of a new topic. The unexpected character of tangential utterances usually causes them to disrupt the normal flow of a conversation, making them evident to an observer.
Each of the transcripts was initially coded by a research assistant who was highly experienced with the coding scheme and was blind to the a priori hypotheses (she could not be blind to diagnosis). The coding was reviewed by both authors, and any disagreements or ambiguous cases were resolved by consensus of all three investigators. Later, in order to establish the reliability of the coding method, half of the transcripts for each group (totaling 1,000 utterances) were re-coded by a second research assistant. This coder attained near perfect agreement with the consensus ratings in the identification of pragmatic categories but did produce a slightly higher count of tangential utterances. The fact that the consensus-derived tangential utterance counts were lower than those of the independent rater suggests that the authors employed somewhat more conservative criteria in rating questionable utterances. An analysis of the reliability between the original and re-coded tangential ratings did, however, yield a significant level of agreement, κ = .89, p < .05. All statistical analyses, except the kappa determination, were performed on an IBM-PC compatible computer using CSS Statistica (StatSoft, Inc.).
An ANOVA revealed no significant differences in communication age equivalent, adaptive behavior composite, or age among the groups with fragile X syndrome, mental retardation, or autistic disorder, indicating that the groups were well-matched on these variables. In addition, a Pearson product moment correlation revealed no significant correlation between age, adaptive behavior composite, or communication age equivalent and the total number of questions, responses, or comments produced. The mean number of questions produced by the fragile X syndrome, mental retardation, and autistic disorder groups were 7.8, 3.6, and 4.1, respectively; the mean number of responses were 39.5, 40.8, and 31.8, respectively; and the mean number of comments were 40.1, 25.9, and 19.2, respectively. A MANOVA revealed no overall group difference among these utterance counts. Table 2 presents the mean number (and SD) of tangential utterances produced by members of each group as well as the percentage of their utterances that were tangential, within each conversational–pragmatic category. Although only one group (those with mental retardation) consisted of a mixed sex sample, t tests for independent samples revealed no significant gender effects for any of our results.
In order to compare the production of tangential language between groups, as well as between conversational pragmatic categories, we performaed a MANOVA with group and conversational category as the independent variables and percentage of tangential language produced as the dependent variable. Percentage scores, which were used in order to control for differences in the amount of language produced, were derived for each participant by dividing the number of tangential utterances produced within each category by the total number of utterances produced within that category. Because one-word affirmative (e.g., “yeah,” “yes,”) and negative (e.g., “nah,” “nope,” “no”), by definition, could not be tangential, they were not included in the total number of utterances cited in the previous paragraph and used to compute percentages within the response category. The MANOVA yielded three significant effects: group, F(2, 27) = 8.74, p < .001, conversational category, F(2, 54) = 5.68, p < .006, and the interaction between group and conversational category, F(4, 54) = 3.22, p < .02. These relationships are illustrated in Figure 1.
Individual between-group planned comparisons revealed that although differences within response utterances did not reach significance, males with fragile X did produce significantly more tangential language within comments than either those with mental retardation, F(1, 27) = 26.71, p < .00002, or those with autistic disorder, F(1, 27) = 16.25, p < .0005, and that males with fragile X produced significantly more tangential language within questions than did those with mental retardation, F(1, 27) = 6.00, p < .02. Very few tangential utterances were produced by participants with mental retardation or autistic disorder in any category, and no significant between-category differences were found for these groups. As stated previously, any gender difference within the mental retardation group was negligible, with the mean percentage of tangential responses being .005 for males and .006 for females.
Individual between-category planned comparisons revealed no significant differences in the amount of tangential language produced across conversational categories by the participants with either mental retardation or autistic disorder, again reflecting the very low incidence of tangential language produced by both of these groups. Males with fragile X, however, produced significantly more tangential language within comments than within responses, F(1, 27) = 31.59, p < .00001, and significantly more tangential language within questions than within responses, F(1, 27) = 8.42, p < .008. The difference between the percentage of tangential language they produced within comments and questions was not significant.
In this study we compared the conversational language of people with fragile X syndrome, autistic disorder, and mental retardation. Rather than looking for group differences in syntactic and semantic competency, we selected the participants to be matched as closely as possible in language age and cognitive functioning, and examined an aspect of pragmatic competency, specifically the production of off-topic, or tangential, language. Semantic and syntactic skills are necessary, but not sufficient, for effective conversational performance. In addition to expressive and receptive vocabulary, and an understanding of the rules of linguistic organization, individuals must be able to manage the flow of a conversation. That is, they must be able to recognize when they are expected to add or clarify information that continues an ongoing unresolved topic, and when it is all right to diverge from a current topic in order to introduce appropriate variation into the conversation. We have already demonstrated that males with fragile X tend to inappropriately perseverate on topics (Sudhalter et al., 1990), and tangential language represents another type of pragmatic conversational problem.
Most people regularly produce some tangential language due in part to their inability to inhibit prepotent thoughts or ideas, particularly in highly arousing situations, and our results revealed the presence of some tangential language in all groups. Males with fragile X syndrome, however, were shown to produce significantly more tangential language than their peers without fragile X syndrome, and their tangential language production was more affected by conversational–pragmatic context. We believe that because of their hyperarousal and related social anxiety, males with fragile X syndrome are more strongly affected by the social demands of conversation than are members of other groups (who may also have inhibitory control deficits) and that this added stress exacerbates the effects of their already impaired inhibitory control system, resulting in the production of increased amounts of tangential language. That is, males with fragile X syndrome are less able than those individuals without fragile X syndrome to exert sufficient inhibitory control to avoid the production of tangential language.
There is both an anatomical basis for the presumed inhibitory control deficits in males with fragile X syndrome as well as diverse behavioral evidence of such deficits. Males with fragile X syndrome have been shown to have larger caudate nucleus and lateral ventricular volumes than do matched non-fragile X syndrome controls (Reiss et al., 1995). The caudate is involved in frontal-subcortical circuits that are important for executive functions, such as inhibitory control (Cummings, 1993). The broad pattern of verbal and nonverbal behaviors that are commonly seen in fragile X syndrome, which include impulsivity, attentional problems, short-term memory difficulties, sequential information-processing, and difficulty coping with transitions (Bennetto & Pennington, 1996; Levitas, 1996) also is consistent with impaired inhibitory control.
Similar explanations have been proposed to explain the language of individuals with schizophrenia and attention deficit hyperactivity disorder (ADHD). It has been suggested, for example, that deficits in inhibitory control processes may underlie the abnormal associative intrusions that appear in schizophrenia (Frith, 1981; Maher, 1983). Maher suggested that, as in our conception of tangential language in fragile X syndrome, such associative language errors are due either to an inability to inhibit the activation of related lexical nodes or to a deficient decay process. Inhibitory processing deficits have also been implicated in many of the behaviors associated with ADHD (Barkley, 1997; Schachar, Tannock, Marriott, & Logan, 1995), and tangential or noncoherent language has been documented in this population as well (Giddan, 1991; Tannock, Purvis, & Schachar, 1993).
This is the first reported study, to our knowledge, of the relationship between tangential language and conversational–pragmatic context in developmentally disabled populations. As we will show, the observed interaction of such contextual demands and specific population characteristics on the production of tangential language is consistent with our arousal/inhibitory control model of tangential language production.
Males with fragile X syndrome were shown to be much less tangential when answering questions than when making unsolicited comments. Because response type utterances often involve answering well-rehearsed questions (e.g., What is your name? Where do you live?), semantic priming and associate responding would be expected to work to the speaker's advantage, helping him or her to answer quickly and, in most cases, correctly. We expected, and found, relatively little tangential language within responses.
Unlike responding to direct questions, asking questions of a partner and making unprompted comments relies less upon conditionalized language, and more upon the participant's own linguistic creativity. Requesting additional information and adding relevant though unprompted information to a topic are ways that people signal to their conversational partners that they want the conversation to continue. As such, they may be considered as examples of approach behavior. The asking of questions, especially, usually involves physically orienting toward one's conversational partner and may also require making eye contact. These activities can be very arousing to individuals with fragile X syndrome and lead to the experience of strong approach/avoidance conflicts that can affect many aspects of their behavior, including their language. As expected, the production of tangential language by males with fragile X syndrome was greater when asking questions and making comments than when answering questions. We believe that the arousal experienced under these circumstances exacerbates the expression of their inhibitory control deficits, rendering them less able to block the production of questions and comments that are not pertinent to the ongoing conversation.
All subjects asked very few questions during the conversations, probably because this language type requires the most initiative. In spite of their social anxiety, males with fragile X syndrome appear to want to talk, interact, and be friendly and will attempt to participate as actively as they can by commenting within conversations. Because this represents an approach behavior that elevates their level of social arousal, however, the resulting hyperarousal would be expected to cause a large percentage of the comments they make to be tangential. In fact, we found tangential language to be most prevalent within this category for all groups, but especially for males with fragile X syndrome, for whom approach behavior is especially stressful.
We do not believe that the production of atypical language by males with fragile X syndrome is associated with undiagnosed autistic disorder because, as this and other studies have shown, the profile of atypical language production differs between these two groups (Ferrier, Gashir, Meryash, Johnston, & Wolff, 1991; Sudhalter et al., 1990). Similarly, the different atypical language profiles exhibited by the groups with fragile X and other kinds of mental retardation suggest that the production of tangential language is not simply a result of developmental delay. The fact that males with fragile X syndrome produced more tangential language than those with autistic disorder or mental retardation, especially during the more socially demanding categories of questioning and commenting, is consistent with our hypothesis that hyperarousal associated with social anxiety exacerbates the effects of inhibitory control deficits. Further, this inhibitory control deficiency rather than undiagnosed autistic disorder or mental delay per se contributes to the production of the atypical conversational language commonly seen in this population.
Although we believe results of the present study support our suggestion that tangential language, as well as other verbal phenotypic characteristics of people with fragile X syndrome, are related to the poor regulation of anxiety and arousal, our conclusions are limited by the fact that no independent assessments of anxiety or arousal were obtained during the conversations. It is not easy to measure anxiety in children, and the generally used self-report measures are simply not applicable for developmentally disabled populations. Although we have had some success with concurrent physiological recording of autonomic nervous system activity in these populations (Belser & Sudhalter, 1995), the surface electrodes necessary to record heart rate or skin conductance activity can be very distracting to participants, and we wanted to examine their conversational language in the absence of any such interference. Although no gender differences were evident in our results, the inconsistent gender composition of the experimental groups also may be considered a methodological weakness of this study. All speakers and their conversational partners incorporate linguistic idiosyncracies that interact to influence each conversation's dynamic character. Therefore, future investigators of conversational language of individuals with fragile X syndrome may benefit from employing a more structured conversational protocol in order to reduce variation in participants' opportunity to produce certain types of language.
Despite these limitations, we believe that our findings have direct implications for the therapeutic intervention of persons with fragile X syndrome. First, realizing that their tangential language is not a result of undiagnosed autistic disorder, it is important that these individuals not be treated as if they have autistic disorder unless they meet all the diagnostic criteria for that condition. Second, in order to reduce the production of tangential language in males with fragile X syndrome, it is important to first help them become as calm and relaxed as possible. This may include the provision of appropriate occupational and, when appropriate, pharmacological therapy, as well as such behavioral considerations as avoiding direct eye contact during conversation. In order to be most effective, therapy should be performed in rooms that are designed and arranged so as to minimize extraneous environmental stimuli, such as voices from hallways or adjacent rooms, bright lights, and other distractions. We believe that training to generally improve self-control may also lead to improved language and social behavior in this population. For example, individuals with fragile X could be taught to wait for a specific signal before responding to a cue or to suppress a behavior for a certain period of time. According to our model, after receiving such training to improve their inhibitory control, and being placed in an environment with little to distract or overstimulate them, the production of tangential language by such individuals should greatly diminish.
NOTE: Requests for reprints should be sent to Vicki Sudhalter, NYS Institute for Basic Research in Developmental Disabilities, 1050 Forest Hill Rd., Staten Island, NY 10314. (firstname.lastname@example.org)
NOTE: This research was supported in part by a grant from the March of Dimes Foundation (12-FY93–0731) and by funds from the New York State Office of Mental Retardation and Developmental Disabilities. The authors thank Lisa Bonura, Shaveta Gujral, and Maria Maranion for their assistance in transcribing and coding the conversational transcripts. Reprint requests should be sent to Vicki Sudhalter, NYS Institute for Basic Research in Developmental Disabilities, 1050 Forest Hill Rd., Staten Island, NY 10314. (email@example.com)