제를 유지하는 방식에서 부적절하게 지각하는 것으로 나타났다. 준언어적 요인의 경우에는 두 평가 방법 결과 간 상관을 보이지 않아 행동분석을 통해 객관적으로 측정된 수치와 상관없이 전문가 집단은 준언어적 요인을 다르게 지각할 수 있다는 가능성을 보여주었다. 비언어적 요인에서 행동 분석의 시선과 자세는 각각 지각적 평가의 눈맞춤 빈도 및 적절성, 자세와 부적상관을 보였다. 이러한 결과는 행동분석에서 대화 상대방을 향해 자세를 취하거나 바라보는 시간이 적을수록 전문가 집단은 이와 동일하게 대화 시 상대방을 향한 시선이 적을 뿐 만 아니라 바라보는 방식에서도 부적절하다고 지각하였고, 상대방을 향해 자세를 취하는 경우도 적다고 지각하는 것으로 나타났다. 연구 Ⅰ과 연구 Ⅱ를 종합하여 보면 고기능 자폐스펙트럼장애 아동은 행동 분석에서 대화차례를 주고받으며 대화에 참여할 수 있었으나 일반아동에 비해 소극적인 자세로 대화를
시작하였고, 상대방의 대화차례에 끼어드는 등 대화 규칙을 위반하는 모습을 보였으며, 대화상대방과 공통된 대화를 유지하지 못하고 갑자기 다른 주제로 변경하는 등 미숙한 주제관리능력을 보여주었다. 또한 고기능 자폐스펙트럼장애 아동은 음향학적 분석을 통해 객관적으로 측정한 준언어적 요인에서는 일반아동과 차이를 보이지 않았다. 고기능 자폐스펙트럼장애 아동은 대화 시 일반아동과 유사한 빈도로 몸짓과 표정을 사용할 수 있었으나 자신이나 상대방이 말하는 동안 대화 상대자를 향해 자세를 취하고 바라보는 시간이 일반아동에 비해 유의하게 적어 대화에 관심과 집중을 표현하지 못하고 적극적으로 대화에 임하지 못한다는 것을 보여주었다. 준언어적 요인을 제외한 다른 요인에서의 행동 분석 결과는 전문가에게서도 고기능 자폐스펙트럼장애 아동이 일반아동과 다르게 지각하는 것으로 나타났다. 즉 두 평가 방법 간 결과에서 공통적으로 언급된 대화차례중첩, 주제유지, 주제변경, 시선, 자세에 대한 특성은 고기능 자폐스펙트럼장애 아동에게 두드러지게 나타나는 행동 특성으로 다른 사람과의 대화에 영향을 미칠 수 있다고 해석할 수 있다. 행동 분석과 지각적 평가에서 측정된 대화 요인 중 두 집단을 판별해주는 요인은 공통적으로 주제변경인 것으로 밝혀졌으며, 이러한 요인은 고기능 자폐스펙트럼장애를 고기능 자폐스펙트럼으로, 일반아동을 일반아동으로 구별해 줄 수 있는 것으로 나타났다. 다만 두 가지 대화 평가방법 간 차이를 보이는 대화 요인이 일부 존재하였는데, 이는 특정 대화 요인 중객관적으로 측정된 수치에서의 차이가 지각적 평가에서는 유의하게 다르게 지각되지 않을 수 있으며, 이와 달리 수치에서의 미미한 차이는 지각적으로는 유의하게 다르다고 평가할 수 있다고 해석할 수 있다. 또한 행동 분석과는 달리 지각적 평가에서는 단순히 수치에서의 차이보다는 적절성에 더 중점을 두고 지각할 수 있으며, 특정 행동에 영향을 미치는 다른 요인까지 배제하지 못할 수 있다는 것을 보여주었다. 본 연구 결과는 고기능 자폐스펙트럼장애 아동의 언어적, 준언어적, 비언어적 대화 특성을 포괄적으로 살펴보았고, 이 중 고기능 자폐스펙트럼장애 집단을 판별하는 요인을 밝혔다는 점에서 의의가 있다. 또한 본 연구 결과를 통해 양적으로 행동 분석을 통해 평가한 결과와 지각적 평가 간의 관계를 살펴보고 공통적으로 차이를 보이는 대화 요인과 다르게 나타난
대화 요인을 확인하여 대화 평가 및 중재 시 우선적으로 고려해야 되는 정보를 제공할 수 있었으며, 행동 분석을 통해 객관적으로 측정된 대화 행동 중 다른 사람에게도 지각되어 대화에 영향을 미칠 수 있는 대화 요인을 확인하였다는 점에서도 의의가 있다.

Social communication impairment among autism spectrum disorder(ASD) children is an important problem as shown by its inclusion in the ASD diagnostic criteria. SC difficulties among these children are evident when conversing with other people. Many people build important interpersonal relationships by sharing information while talking to one another (Heo & Lee, 2012). Conversation usually occurs linguistically, but also using paralinguistic and nonlinguistic expressions may help people to convey their intentions more effectively. As such, because both paralinguistic and nonlinguistic communication are important for effective expression of intent (Knaap & Hall, 2009), considering all of these factors is important when examining features of conversation. Previous studies examining conversation skills of ASD children used behavioral analysis and rating scales. Although behavioral analysis quantifies specific conversational behaviors of interest in an actual conversation sample through direct observation, it does not enable the examiner to confirm whether the measured behavioral characteristic is perceived by other people. In contrast, rating scales enable examination of how children’s conversational behaviors are perceived (Brinton & Fujiki, 1984; Landa et al., 1992), its dependence on subjective impression makes it vulnerable to underestimation or overestimation. Studies that
attempted to investigate the conversational skills of ASD children have used either behavioral analysis or rating scale. As previously mentioned, if behavioral analysis was used, it is difficult to examine whether the measured conversational characteristics are perceived by other people as well. Therefore, this study performed both behavioral analysis and perceptual assessment to observe whether the conversational characteristics of high-functioning ASD (HFASD) measured through behavioral analysis are perceived by other people. To this end, we examined the conversational characteristics of HFASD children in terms of linguistic, paralinguistic, and nonlinguistic factors in study I and investigated the conversational characteristics with perceptual evaluation in study II. Furthermore, we identified the conversational factors that distinguish the
HFASD group and examined the association between the results of studies I and II. In Study I, The participants were total 51 children; 17 HFASD children and 34 Typical development (TD) children (17 matched on the chronological age and 17 matched on the language age). Conversation data was collected for 10-15 minutes
under the topic of “family, school, friends” according to the conversation procedure of the ‘Hallym Conversation & Pragmatic Assessment Protocol’. The conversations were audio- and video-recorded with an audio-recorder and two camcorders to measure the linguistic, paralinguistic, and nonlinguistic factors of
conversations. The collected data were transcribed according to the turn-taking analysis criteria, and linguistic factors, paralinguistic factors, and nonlinguistic factors were measured. The conversational characteristics of the two groups were compared using t-test, and the factors that distinguish the two groups were
identified using stepwise discriminant analysis. The results of study I were as follows. First, among linguistic factors, the HFASD group did not significantly differ in turn-taking frequency, had a significantly lower frequency of initiation, and significantly higher frequency of overlap compared to that of the TD group. These results suggest that HFASD children are able to participate in a conversation appropriate to their age and
linguistic level but are more passive and less skilled in conversation, as demonstrated by the fewer initiations and more violations of turn-taking. Moreover, regarding topic manipulation, the ASD group showed a significantly higher number of topics and rate of change with significantly lower rate of maintenance compared to the TD group. These results show that HFASD children frequently change the topic of the conversation, thereby engaging in a greater number of topics, and are unable to continue a conversation cohesively or further deepen the topic of the conversation. Second, regarding paralinguistic factors of conversation, there were no significant differences in mean pitch, pitch range, intensity, and speech rate
measured with CSL between the two groups. The HFASD and TD groups showed similar descriptive statistics for all parameters, showing that the HFASD group does not differ from the TD group in terms of acoustically measured voice pitch, changes of pitch, and speech rate. However, interpretation requires caution, as
acoustic analysis alone may be inadequate to examine paralinguistic factors. Third, regarding nonlinguistic factors of conversation, the two groups did not significantly differ in the frequency of gesture and facial expressions, but the HFASD group had significantly lower gaze and posture compared to that of the TD
group. These results show that the HFASD group does use gesture and facial expressions at a similar frequency with that used by the TD group, but they look at the other person or maintain a posture toward the other person during a conversation in a markedly lower frequency, indicating that they may have shortcomings in paying attention to the conversation partner or topic and actively participating in and maintaining a conversation. Fourth, we performed stepwise discriminant analysis to identify the conversational behaviors measured through behavioral analysis that distinguish the HFASD and TD groups. The conversational factors examined in this study distinguished the two groups, and topic change rate, overlap rate, and gaze were found to be the most important discriminant factors. The sensitivity and specificity
based on the final discriminant equation were 76.5% and 88.2%, respectively. In study II, we aimed to examine whether experts can perceive the conversational features of HFASD using a perceptual rating scale by reviewing the conversation data collected in the behavioral analysis in study I. Six speech-language pathologies with at least three years of experience of educating or treating ASD children were designated as the raters. The 51 conversation samples collected in study I were randomly numbered according to a set
proportion between the two groups, and the video recordings made by two camcorders were edited to be played side-to-side so as to observe and assess the linguistic, paralinguistic, and nonlinguistic behaviors that may affect conversation. The perceptual rating scale used was a four likert scale that consisted of linguistic, paralinguistic, and nonlinguistic factors. The raters were first educated and trained for the use of the scale, and each rater was instructed to assess 17-18 conversation samples. Four of these samples were rated by all six raters to measure reliability. As done in study I, the conversational features of the two groups found with perceptual evaluation were compared using t-tests, and the factors that discriminate the two groups were identified with stepwise discriminant analysis. Furthermore, the association between the results in study I and the results in study II were analyzed with Pearson correlation analysis. The results of study II were as follows. First, among linguistic factors, frequency of initiation and breakdown in turn-taking did not significantly differ
between the two groups, but the HFASD group had a significantly higher inappropriate initiation and frequency of overlap compared to that of the TD group. These results suggest that experts perceive that HFASD children are able to take turns in a conversation but inappropriately initiate their turn in disregard to
the other person’s response or state and inappropriate jump in on the other person’s turn more frequently. Regarding topic management, the HFASD group scored higher in topic maintenance frequency and inappropriateness and topic change frequency and inappropriateness compared to that of the TD group. In
other words, experts perceived that the HFASD group cannot expand or inappropriately expand on the shared topic with the partner of the conversation due to a failure of considering the partner’s emotions or circumstances and abruptly change topics in higher frequencies than the TD group. Second, regarding paralinguistic factors, there were no significant differences in speech rate and intensity between the two groups, but the HFASD group scored significantly higher in changes of intensity, pitch, and intonation compared to the TD group. These results show that experts perceive the HFASD group to have similar speech rate and intensity with that of the TD group but have slightly unnatural changes of intensity, higher pitch, and unnatural intonation. Third, regarding nonlinguistic factors, there were no significant differences in gesture frequency between the two groups, but the HFASD group scored significantly higher in posture, expression, inappropriate gesture, and eye-contact frequency and inappropriateness. The experts perceived that the HFASD group does not significantly differ in terms of the frequency of gestures, but that the HFASD group had more inappropriate gestures. Furthermore, experts also perceived that the HFASD group not only looks at the other person or maintains posture toward the other person in conversation significantly less frequently but
also more inappropriately and that the HFASD group also has inappropriate facial expressions. Fourth, we performed stepwise discriminant analysis to identify the conversational behaviors measured through perceptual evaluation that distinguish the HFASD and TD groups. The conversational factors examined in this study distinguished the two groups, and topic change frequency and topic maintenance frequency were found to be the most important discriminant factors. The sensitivity and specificity based on the final discriminant equation were 76.4% and 85.3%, respectively. Finally, the correlations between study I and study II results were as follows. Among linguistic factors, the rate of overlap and topic change found in
behavioral analysis were positively correlated with conversation overlap and topic change frequency and inappropriateness found in the perceptual evaluation, respectively. The rate of topic maintenance in behavioral analysis was negatively correlated with topic maintenance appropriateness in perceptual evaluation. In other words, the higher the rate of overlap and topic change in behavioral analysis, the more the experts perceived children to inappropriately jump in during turn-taking and more abruptly change the topic of conversation. Furthermore, the higher the topic maintenance rate in behavioral analysis, the more the experts perceived children to inappropriately maintain a given topic, such as having vague conversation or unnecessarily giving detailed information. For paralinguistic factors, there were no significant correlations between the results of the two assessment methods, suggesting the possibility that experts may perceive
paralinguistic factors differently regardless of the objective measurements taken through behavioral analysis. Among nonlinguistic factors, gaze and posture in behavioral analysis were negatively correlated with eye-contact frequency and appropriateness and posture in perceptual assessment, respectively. These results
suggest that the less the children maintain a posture toward or look at the other person during conversation in behavioral analysis, the most the experts perceived children to engage in less and inappropriate gaze at the other person and take a posture toward the other person during conversation. Study I and study II results showed that HFASD children were able to participant in a conversation by taking turns but was more passive in initiating a conversation and more often violated conversation rules by jumping in on the other person’s turn compared to TD children. Moreover, HFASD children were less skilled at topic manipulation, such as failing to maintain a conversation under the common topic with the partner and abruptly changing the topic of the conversation. On the other hand, there were no significant differences in objective paralinguistic factors measured via acoustic analysis between the HFASD and TD group. HFASD children were able to use gestures and facial expressions at a similar frequency with that used by the TD group, but they gazed at and took a posture toward the partner during the conversation significantly less, suggesting that they have difficulty expressing interest in and concentrating on the conversation and actively participating in the conversation. Experts also perceived factors other than paralinguistic factors to be different between the HFASD and TD groups, as was found in the behavioral analysis. In other words, the overlap, topic maintenance, topic change, gaze, and posture characteristics commonly mentioned in both assessment methods are behavioral characteristics that stand out in HFASD, which may affect their conversations with other people. Topic change was commonly found by behavioral analysis and perceptual assessment to be a discriminant factor between the two groups; that is, topic change characteristic distinguishes HFASD from TD. However, some conversational factors were assessed differently between the two assessment methods, suggesting that factors measured to be significantly different between the two groups in behavioral analysis may not be perceived significantly differently in perceptual evaluation, and conversely, factors measured to be similar between the two groups in behavioral analysis may be perceived significantly differently in perceptual
evaluation. Furthermore, unlike behavioral analysis, in which the focus is on differences of objective measurements, perceptual evaluation may be more focused on the appropriateness of behaviors and that it cannot exclude other factors that may influence a specific behavior. This study comprehensively examined the linguistic, paralinguistic, and nonlinguistic features of conversation in HFASD, and it is meaningful in that it
identified the factor that distinguishes HFASD from TD. This study also examined the correlations among quantitative results of behavioral analysis and results of perceptual evaluation to pinpoint the type of information to be preferentially considered during conversation assessment and intervention by identifying the
conversational factors found to be different between the two groups in both assessment methods or in only one assessment method. Furthermore, this study is also meaningful in that it confirmed conversational factors that are both quantitatively different and perceived to be different between the HFASD and TD groups and therefore may have an impact on conversation.
Key words: High functioning autism spectrum disorder, Linguistic, Paralinguistic, Nonlinguistic, Conversational factor, Discriminant analysis