This thesis has followed the guidelines of the Finnish Advisory Board on Research Integrity (2012) for responsible conduct in designing and implementing research. In following these guidelines, an ethical review was administered by the ethics com-mittee of the University of Eastern Finland, and written consent to participate was obtained from the children’s parents and the adults that were involved in the study.
However, it should be stressed that when the participating children cannot provide written consent themselves, there is a risk of ignoring their view on taking part in the research (Porter, 2013). Thus, the children’s willingness to participate was monitored throughout the sessions. As the sessions in original articles II-IV took place during school days, the educators encouraged and prompted the children to participate in the activities. That said, if the children showed reluctance to do so, they were al-lowed to skip a specific activity or the entire activity group session and return to their classroom. Original article I took place as a free-time activity and the children were free to continue or discontinue their participation with similar freedom as in original articles II-IV. The participants received no financial compensation for their involve-ment but it can be seen as beneficial to them to have taken part as the children were able to practise social, motor, and cognitive skills together with their parents, tutors, educators, and researchers.
Respecting participants has been emphasised when reporting the research.
This entailed using considerate language when discussing the participants and in acknowledging their right to anonymity. The names of people and places have been changed to prevent the participants from being recognised. Photo stills (with par-ticipants’ facial features anonymised as much as possible without losing relevant information, such as head direction) were used in some of the original articles for illustrative purposes, with the participants’ consent. The participants who might re-main recognisable by those who know them gave permission to use their photos without anonymisation, and their consent was respected. The research resulted in no reported harm to either the children or their social partners, excluding the reported unpleasantness of the video recording activity for some of the educators (discussed in detail in section 5.6). As the recordings took place during group activity sessions organised by the EvTech and CASCATE projects, the actual teaching situations in the classrooms were not disturbed by researcher influence. Participation remained optional throughout the study, and feedback from the children’s families and educa-tors was predominantly positive (Kärnä, Parkkonen, Tuononen, & Voutilainen, 2013).
The children gave feedback using picture symbols after playing at each action station, which allowed them to communicate their feelings to the researchers and educators.
According to the feedback, the children enjoyed the activity group sessions, and the families were excited to have their children take part.
5.6 reliaBilitY, validitY, and generalisaBilitY of the research
The reliability and validity of this research can be assessed in multiple ways. The relia-bility of research considers how independent of accidental circumstances the research findings are (Kirk & Miller, 1986). Psychological and cognitive research commonly uses inter-rater reliability to assess the reliability of a measurement (i.e. how well
independent researchers categorise observed behaviours under the same category).
Inter-rater reliability was used to assess the reliability of the coding scheme in original article I during its development process. The scheme was also discussed with other researchers and experts from the field of Special Education during this process. This procedure was limited, however, as the inter-rater reliability was not assessed for the final scheme due to limited availability of a second coder at this phase. Original article III also utilised inter-rater reliability (using intra-class correlation coefficient) for the reliable identification of pointing gestures from the data.
In qualitative social scientific research, the questions of reliability and validity are conceptualised differently than in quantitative research. Social scientific research examines the social world with the aim of representing the reality rather than repro-ducing it (Hammersley, 1992). Rather than imposing researcher-led interpretations on data, CA is concerned with participants’ orientations to each other’s conduct, exactly as they are visible in the data. CA is thus particularly rigorous in its requirement of descriptions being strictly empirically grounded before being accepted as valid (Peräkylä, 2004), and, as such, is a distinctive methodological approach within quali-tative research.
According to Peräkylä (2004), CA recordings and transcriptions are ‘raw mate-rial’, comparable to field notes in ethnography. Thus, the quality of recordings and transcripts are crucial for the reliability of CA research. It is this use of audio and video recordings that allow repeated and accurate access to the phenomena under study, not only for researchers but also for the public. This differentiates CA from many other qualitative approaches that primarily rely on researchers’ field notes. For Peräkylä (2004), the key aspects of reliability in CA studies involve the selection of what is recorded, the quality of recordings, and the adequacy of transcripts. One crucial considera-tion is the amount of data that needs to be collected in order to grasp the complexity of the phenomena. As this thesis was part of a larger research project, the amount of data exceeded the needs at hand. However, it would have been interesting to collect data from other settings, perhaps during fully authentic classroom interactions, but this was beyond the scope of this study.
The technical quality of the recordings was ensured by the research team so that the cameras captured a wide-angle view of the child-adult dyad and the screen where the tasks were presented. Furthermore, the eye tracking glasses made it possible to examine children’s eye movements with a level of accuracy and detail beyond any video recording. The adequacy of transcripts was ensured by annotating the selected segments in great detail using ELAN, transforming them into transcriptions follow-ing Jeffersonian conventions (Jefferson, 2004), and then addfollow-ing the information that was difficult to manage with ELAN, such as intonation. It is important to note that transcriptions are always selective representations of the original recordings and, as such, an important part of the actual analytical process (Bezemer & Mavers, 2011;
Ochs, 1979). They are never neutral representations of reality. However, it is useful to make sure that other people are able to read the transcripts and that they agree on the transcripts’ ability to represent the data. To this end, the transcripts have been examined with the thesis supervisors, as well as other researchers and fellow PhD students, who have commented on their readability and accuracy. Such data sessions are a crucial part of the CA tradition, whereby recordings and transcriptions are ex-amined with other researchers.
The validity of research deals with the accuracy of interpretation of the observa-tions made and whether they are called by ‘the right name’ (Kirk & Miller, 1986). In
CA, these concerns include the transparency of analytical claims, validation through ‘next-turn proof procedure’, and deviant case analysis (Peräkylä, 2004). The transparency of analytical claims requires that the claims are evidently rooted in the transcripts (and authentic recordings) and are accessible to the readers. On the other hand, validation through the ‘next-turn proof procedure’ is a crucial criterion for validity in CA research (Peräkylä, 2004) and was the guiding principle in the analyses of this thesis. Deviant case analysis deals with the examination of cases that do not follow a suggested pat-tern and makes up an important resource for testing hypotheses in CA work.
Peräkylä (2004) has also discussed the danger of importing an institutional context into data. For instance, the educators’ conduct could have been treated as showing authority without carefully examining whether student-teacher categories are actu-ally present in the data for the participants themselves. As this thesis focuses on social interaction more broadly (albeit in an educational setting) rather than student-teacher interaction, I have withdrawn from making such assumptions.
A final consideration of validity relates to the research activity itself. It has been argued that video recording and the participants’ awareness of being studied have an influence upon the interactions recorded (‘observer’s paradox’; Labov, 1972), and thus risk the validity of the research. In our interview study (Kärnä et al., 2013), some of the educators participating in the CASCATE project commented that despite the fact that they understood the need for video recording, it made them feel unnatural.
One of the interviewed educators commented:
‘... you know it is probably good [videotaping] but it is not a natural [situation]. It is a kind of formal [situation]... the devices are running and thus, one is not able to be natural [in the situation].’ (Kärnä et al., 2013, 125)
However, this ‘unnatural’ behaviour was rarely visible in the data except for a few occasions in which some of the educators actively avoided the camera by staying out of the camera view. Such instances have not been taken into consideration as it would be disrespectful to the educators’ anxiety. It is apparent in this study that the interactions involving some of the educators were affected by the video recording activity to some extent, but participants noticing video cameras and orienting to the recording taking place is not especially an issue per se. As Speer & Hutchby (2003) point out, the fact that participants are aware of the recording activity does not neces-sarily hinder their interactions. Rather, such data that includes participants’ orienta-tions to the recording activity can provide rich material for analysis (see also Tuncer, 2016). In discursive psychological readings of social constructionism (Edwards &
Potter, 1992), there has never been a wish for ‘neutral’ data that is not ‘disturbed’ or
‘contaminated’ by the researchers’ presence or by the recording activity (Potter, 2002;
Potter & Hepburn, 2007; Speer & Hutchby, 2003). In this research, the video recording activity was an explicit part of interacting and playing in the EvTech and CASCATE environments, and, as such, the interactions were as natural as they could be in those specific settings. It is also worthy of note that the video recording became very familiar to the participants considering the timeframes of the projects: The EvTech project lasted for 5 years and CASCATE for 4 years, during which time the children and their social partners were recorded at the very least for some months, and mostly for many years. The participants thus became increasingly well-acquainted with the video recording activity, and there was no discomfort reported by the children or their families regarding its use.
When a study does not involve a large number of participants, the generalisability of the findings is often brought into question, and it may be asked whether the findings apply to other settings and to interactions with other people? This raises the issue of what is meant by ‘generalisability’. In quantitative research, prediction and control are important in seeking to produce enduring claims that are seen as context-free, but such an understanding of generalisation is not without challenges (Lincoln & Guba, 2000). For example, Newton-like generalisations are rarely applicable to complex hu-man (inter)actions (Donmoyer, 2000). It is therefore debatable as to whether linking generalisability to the use of large samples that are carefully designed to represent a certain population in studies examining human behaviour is feasible. As Donmoyer (2000) points out, even statistically significant findings yielded from studies with enor-mous sample sizes cannot be directly applied to a particular individual in a particular situation. This is not to characterise generalisability as an irrelevant concept but as a challenge that requires some rethinking, particularly when applied to qualitative research (see e.g. Donmoyer, 2000; Guba & Lincoln, 2000; Gomm, Hammersley, &
Foster, 2000).
The findings of original article I that stemmed from the categorisation approach served primarily to increase understanding of the context-situated behaviours of a specific child with ASD, and, as a case study, cannot be used to generalise in a tradi-tional ‘distributradi-tional’ sense to other children with ASD. However, the findings can be
‘transferred’ to other cases or settings that have adequate ‘fittingness’, i.e. similarity (Guba & Lincoln, 2000; see also Gomm et al. 2000 for generalisability in case studies).
According to Peräkylä (2004), most CA studies based in institutional settings (includ-ing educational interactions) can currently be considered more like case studies as institutional practices often vary significantly between settings. This is likely to change as more research is conducted and understanding of the similarities and differences between the settings develops. However, Peräkylä (2004) suggests that the traditional distributional understanding of generalisability can be complemented with a notion of ‘possibility’ that somewhat aligns with Guba and Lincoln’s (2009) ‘transferability’.
However, for Peräkylä (2004, 297), possibilities of language use ‘can be considered generalisable even if the practices are not actualised in similar ways across different settings’.
In line with Peräkylä’s notions, I do not aim to suggest that the identified ways in which the participating children interacted in this study could form an exhaustive description of ‘interaction in ASD’. Neither do I believe that the findings of this thesis apply to all children with ASD, regardless of the interactional context in which they are examined. Rather, the findings demonstrate some of the ways in which children with ASD interact with their adult social partners in specific interactional contexts. As Donmoyer (2000, 51) puts it, ’research can only function as a heuristic; it can suggest possibilities but never dictate action’. Thus, the ways in which the children with ASD initiated the management of interactional trouble (original article II), in which the adult co-participants facilitated joint attention with the children (original article III), and in which the children used their gaze in responsive and initiating environments (original article IV) can be considered as possible practices across settings.
6 summaRies oF the oRiginal aRticles
This section presents the summaries of the original articles, placing emphasis on meth-odological reflection.
6.1 original article i
The first original article ‘Context-situated communicative competence in a child with autism spectrum disorder’ (Tuononen, Laitila, & Kärnä, 2014) was based on my Mas-ter’s thesis in Psychology (Tuononen, 2012), and set out to bring context into the examination of joint attention. This study originated from an interest in children’s joint attention behaviours within the biomedical paradigm, using a predefined cod-ing scheme. However, the study developed into an investigation of the local contexts where a child with ASD produced joint attention behaviours, reflecting the starting point of this thesis in which the limitations of decontextualised coding became appar-ent and warranted a shift to the interactional approach. Although the study relied on the categorisation approach (i.e. content analysis) common to the biomedical para-digm, it developed a critical stance towards the examination of skills and challenges
‘within individuals’, resonating with understanding from the social constructionist perspective.
In original article I, I employed video data pertaining to one child with ASD, Veeti, to focus on his gaze behaviours during three different technology-enhanced activities in the EvTech project (dance mat station, symbol matching station, and LEGO con-structing station) in the presence of a club tutor and Veeti’s parent. The predefined coding scheme used in the study differed from those in many previous studies in its attempt to grasp means of interacting that were common for this particular child.
Thus, the creation of the scheme was preceded by an observational phase. The scheme was also modified to include behaviours that might be ‘atypical’ in their character.
For instance, most studies record whether a child establishes eye contact with a social partner or looks at her eye region, whereas looking at the body parts or following the action conducted by the social partner are rarely of interest. However, as an initial viewing of the data suggested that such gaze shifts were common for Veeti, they were included in the scheme. The categorisation approach was first used to quantify these gaze behaviours (e.g. ‘gaze directed at the co-participant’s face’ and ‘gaze directed at people near the action stations’) so as to examine whether there were differences in their frequency of occurrence between the three activities. The analysis then moved on to transcribe these instances in order to qualitatively examine the contexts where the coded behaviours occurred, as the quantitative analysis was limited in capturing them.
The statistical analysis of the coded gaze behaviours showed that Veeti used his gaze differently during the three activities. For example, Veeti ‘looked at an object or in a direction that was pointed out or otherwise presented’ and ‘directed his gaze at the co-participant’s body or action’ more often at the LEGO constructing station than at the dance mat station. The qualitative content analytic examination showed that these variations were related to the different nature of the activities. The activity at the LEGO constructing station allowed Veeti and his club tutor to sit side-by-side, constructing the LEGO models together. On the other hand, the activity at the dance mat station
involved Veeti independently playing the game while his club tutor and parent were standing next to the station instructing and supporting him if needed. Thus, the co-operative nature of the activity at the LEGO constructing station made it relevant for Veeti to look at objects pointed out by his club tutor and to look at the tutor’s body while she was participating in co-constructing the LEGO model. Overall, the findings of this article suggest that context plays a crucial role in studying children with ASD.
While I was able to map the variations in different gaze behaviours between con-texts and to make inferences on the nature of the interactions in which these gaze behaviours occurred, the analysis was not yet designed to consider the co-participants’
actions and their contributions to the interaction. The focus was exclusively on Veeti, which meant that there was a risk of attributing all occurrences in an interaction to him personally due to a lack of information on his social partners’ concurrent actions.
This problem, or ‘anomaly’ in Kuhnian (1970) terms, which crucially influenced the interpretations made about a child’s ability for social interaction, pushed me to focus on the detailed study of interactions and shifted me towards social constructionist thinking. The article concludes that an interactional approach such as CA could enable one to examine in more detail how a child with ASD attempts to initiate interaction using ‘atypical’ means and to investigate how the co-participants treat these attempts.
This article was thus the basis for moving from seeing the interactional difficulties as being internal to the child, to viewing them as constructed in interactions between children with ASD and their co-participants.
6.2 original article ii
The second original article ‘Building mutual understanding: How children with au-tism spectrum disorder manage interactional trouble’ (Dindar, Korkiakangas, Laitila,
& Kärnä, 2016a) was located within the interactional approach, and set out to bring a new perspective to the study of children’s interactions with their interactional part-ners, and how they manage the troubles that arise in these interactions. Previous research within the biomedical paradigm has focused on how children with ASD
& Kärnä, 2016a) was located within the interactional approach, and set out to bring a new perspective to the study of children’s interactions with their interactional part-ners, and how they manage the troubles that arise in these interactions. Previous research within the biomedical paradigm has focused on how children with ASD