Title. The use of conversational analysis: nurse–patient interaction in communication disability after stroke.
Aim. This paper is a report of a study to explore how nursing staff and patients with aphasia or dysarthria communicate with each other in natural interactions on a specialist stroke ward.
Background. Nursing staff often talk with patients in a functional manner, using minimal social or affective communication. Little nursing research has been carried out with patients who have communication difficulties. Conversational analysis, used in other healthcare settings, is a way to explore these interactions in depth in order to gain further understanding of the communication process.
Method. An observational study was carried out in 2005 and the data were 35·5 hours of videotape recording and field notes with 14 nursing staff and five patients with aphasia or dysarthria. The recordings were analysed using conversation analysis.
Findings. Nursing staff controlled the conversations by controlling the topic and flow of conversations, creating asymmetry in all interactions. Patients had very little input because of taking short turns and responding to closed questions. These behaviours are related to the institutional context in which they occur.
Conclusion. In rehabilitation, the focus for interaction may be thought to be patient goals, worries or plans for the future, but in this study nursing staff controlled the conversations around nursing tasks. This may be because they do not have the confidence to hold conversations with people with communication problems. Nursing staff need to receive training to reinforce communication rehabilitation programmes and to engage more fully with patients in their care, but also that a wider institutional culture of partnership is developed on stroke wards.
• Communication is essential for the formation of therapeutic nurse–patient relationships, but communication problems after stroke are common, with between 19% and 38% of stroke patients having aphasia or dysarthria.
• Most communication studies of stroke exclude people with aphasia or dysarthria, mainly because of difficulties in gaining consent and meaningful data as a result of their communication problems.
• Previous studies of nurse–patient interaction with patients without communication deficits have shown that the quality of communication depends on the individual nurse, but that overall communication is mainly based on function and nursing tasks.
What this study adds
• Conversations tended to be controlled by nursing staff, with little opportunity for the patient to contribute to, or influence, the conversation flow (asymmetry).
• This asymmetrical interaction may be due to nursing staff not feeling confident in their skills with people with communication problems following stroke, as well as focusing the interaction on nursing tasks alone.
Implications for practice and/or policy
• Training to reinforce communication rehabilitation programmes and to engage more fully with patients in their care could improve the quality of care for stroke survivors.
• Managers and clinicians should promote an institutional culture of partnership on stroke wards.
People who experience communication problems following stroke not only have to manage the usual physical consequences of stroke but also have wider issues to resolve during their rehabilitation, related to difficulties in being able to interact meaningfully with other people. These difficulties can have a direct effect on physical recovery if people do not understand instructions. They can also create a sense of isolation from family, friends and the wider community which could result in social isolation and depression if not addressed (Astrom et al. 1993, Parr et al. 1997).
An integral part of stroke rehabilitation in nursing is negotiation with the patient about their rehabilitation goals, developing therapeutic relationships, providing strategies to aid independent function with activities of daily living, and providing information and emotional support to help the patient come to terms with their stroke (Long et al. 2001). This requires effective communication between nurse and patient. Stroke survivors with aphasia may not have the ability to comprehend oral or written information, and those with aphasia or dysarthria may not be able to express themselves, challenging the efficacy of the usual routes of communication during rehabilitation. Healthcare professionals, including nursing staff, have a role in providing an enabling environment for meaningful interaction in hospital rehabilitation settings.
In western countries, there is expected to be >500 new strokes per year in a typical population of 250,000 people (Sudlow and Warlow 1996). Communication deficits after stroke are common, affecting up to 38% of stroke survivors (Pederson et al. 1995, Sellers et al. 2002). Aphasia is a reduced ability to understand and/or formulate written or spoken language and is caused by damage to the brain (Rosenbeck et al. 1989). Dysarthria is a difficulty in articulation during speech, leading to slurred or incomprehensible speech because of impairments in the speech musculature (Sellers et al. 2002).
Experiences of people with aphasia or dysarthria can be frustrating because, even if they can get their basic ‘message across’, there is still poor interaction and poor formation of relationships because of communication inadequacies. This is because conversation is seen to have two roles: the transfer of information, opinions and feelings, which can be described as transaction (Kagan 1995) and as a vehicle for the social construction of relationships and situations, described as interaction (Psathas 1995). The latter leads to the development of relationships, and enables the person to have social connection. For example, ‘small talk’ has little exchange of information, but is a basis for forming new relationships (Kagan 1995).
General nurse–patient interaction
Effective nurse–patient communication is required to form therapeutic relationships and negotiate care. Researchers have highlighted that nurses tend to focus on physical care and interact in a routine and superficial way, with limited social and emotional interaction (Anderson & Adamsen 2001,Caris-Verhallen et al. 2000,Davies 1992, Nolan et al. 1995, Pound & Ebrahim 2000, Suominen et al. 1995, Wilkinson 1991). Researchers carrying out these observational studies have coded the frequency and type of nurses’ verbal interaction with patients who do not have communication problems in a variety of ward settings. In the literature, there is no overall trend as to the type of interaction, or agreement on influences to the interaction, with wide variations dependent on individual nurse behaviour. Workload and time pressures have been cited as reasons for limited social and emotional communication, as well as the ward culture emphasizing biomedical needs over psychosocial needs (Chant et al. 2002, Pound & Ebrahim 2000).
Social scientists have studied ordinary conversations and what is achieved through them using the methodology of conversation analysis (CA) (Atkinson & Heritage 1984). In this methodology, researchers are not so much interested in the topic of the conversation researched, but rather in trying to understand the social rules of a conversation, such as the rules of taking turns and not interrupting each other. Conversation analysts view talk and body movement not just as a medium by which to communicate, but as a way of constructing social realities and relationships (Heritage 1997).
Conversation analysis in health care
Conversation analysis has not been widely used in healthcare research and has been predominantly used to evaluate speech and language therapy (Oelschlaeger 1999, Beeke et al. 2003). There have been very few studies of the interactions of other healthcare professionals during inpatient rehabilitation. However, several researchers have unpacked details of conversations between patients and other healthcare professionals. Patients’ consultations with doctors have been explored (Maynard 1991, Heath 1992), showing that the physician controlled the conversation, whereas patient interaction was limited or not forthcoming, despite patients given opportunities to speak. This has been described as asymmetrical (or one-sided) conversation. Both researchers described this asymmetry as grounded in the social context of lay people seeking advice from an expert. Parry (2004) found a similar conversation structure during goal-setting in physiotherapy, where patients were reluctant to contribute to goal-setting, despite encouragement from the physiotherapist. It appears that in healthcare settings professionals take the more dominant role in conversations about treatment.
Conversation analysis with people with aphasia or dysarthria
There has been very little research using CA with people with dysarthria, but it has been found that when these people speak their conversations are similar to ‘ordinary’ talk. Conversations are collaborative, and the person with aphasia uses a wide range of resources including gesture, tone and pitch of voice or sounds, and the resource of the speech of others (Goodwin 1995). However, the meaning of the conversation is developed through the context in which it occurs, and the preceding talk before a turn, i.e. the sequential context (Wilkinson 1999). For example, if a conversation partner is attempting to guess what a person with aphasia is wanting on their toast, it will be related to the preceding talk and the context; thus, if ‘butter’ was wrong, the next guess could be ‘jam’ or ‘marmalade’. People with aphasia rely strongly on this sequential context to help their partners understand what they are trying to say, and so it is difficult for them to take the lead in the conversation and change to a new topic (Lesser & Perkins 1999).
A second characteristic in aphasic conversations is that repair (which is the correction mechanism when the speaker makes a mistake) tends to be longer (Milroy & Perkins 1992). Self-repair can occur when the person who makes the mistake corrects themselves, for example if you mispronounce a word, you tend to correct yourself immediately. Those without communication problems prefer self-repair, whereas those with aphasia tend to rely on others to help with the repair (Perkins 2003). If mistakes are protracted, with over-reliance on others, this can lead to the person with aphasia feeling linguistically incompetent (Wilkinson 1999).
Third, people with aphasia tend to have frequent long silences in their speech because of word-finding difficulties. They also exploit non-verbal cues to indicate the end of their turn, for example avoiding eye contact until the end of their turn and then looking to their conversation partner (Lesser & Perkins 1999). People with aphasia use minimal turns, for example ‘mm hm’, ‘yeah’ or ‘aha’ to relinquish their turn and allow their conversation partner to start talking (Perkins 1995). They communicate by using different conversational rules, and it is essential to know how nurses are managing these new rules within ward settings.
The aim of the study was to explore how nursing staff and patients with aphasia or dysarthria communicate with each other in natural interactions on a specialist stroke ward.
This was an observational study using video recordings of conversations with patients with communication deficits after stroke and nursing staff on hospital wards, using a conversation analytic approach.
Conversation analysis is a qualitative methodology that is informed by several disciplines: sociology, ethnography, linguistics, social psychology and ethnomethodology. CA focuses on meaning and context, with talk and action being productive of, and reflecting, the circumstances in which it is said (Heritage 1997). Ordinary conversation is structured, with recurrent natural ‘organizations’ within the conversation. Some are related to the place where they occur, while others can be found in any talk (Heritage 1997). For example, you would use a different organization of conversation when giving evidence in a courtroom to when you are recalling the same story to a friend, because the context dictates how you construct your conversation. CA uses a naturalistic, observational method of actual behaviour (both verbal and non-verbal).
An aspect which has been of interest to analysts is how talk is influenced and structured by taking place within a particular institutional context (rather than general everyday conversations). Drew and Heritage (1992) describe three main features of institutional interaction:
• Talk is task focused. The participants organize the interaction around the institution’s task or goals.
• Institutional talk is constrained or restricted. The type of contribution from the speakers is governed by the formal character of the interaction. This may be positive, for example talking openly about incontinence in a healthcare setting, which would not be possible in ordinary social contexts.
• Institutional talk is associated with certain frameworks and procedures particular to specific institutional contexts. An example is that the healthcare environment constrains the professional from showing surprise.
In this study, we were interested in exploring the institutional nature of conversations on the ward.
Participants were recruited from a stroke rehabilitation ward in a community hospital, and an acute stroke unit in a district general hospital. A convenience sample of five stroke survivors with aphasia or dysarthria participated in the study. Inpatients on the stroke wards were identified by speech and language therapists (SLTs), and invited to participate if they were diagnosed by a SLT as having aphasia or dysarthria caused by a primary diagnosis of stroke. Exclusion criteria included: the communication problem not resulting from their recent stroke; the patient was not for active rehabilitation or dying; and the patient’s first language not being English.
All nursing staff working on the two wards were invited to participate through ward meetings. Temporary or agency nursing staff were included if there was adequate time for the consent process. The nursing participants were a convenience sample of those who were on duty and interacted with the patients at the time of data collection.
Video recording and concurrent field notes were used to collect data. Recording occurred for periods of 1–3 hours, over 4 months, in 2005. The time and day varied to capture different atmospheres and routines on the wards. The camera was placed near to the bedside, but was obstructed when the bedside curtains were drawn. Video recordings captured the audio and visual elements of interaction as it arose, and enabled exploration of non-verbal influences, for example the body and physical objects, on talk (Heath 1997).
Conversation analysis is considered to be a reliable method because data are collected through video recording of actual interactions, leading to accurate data of events that can be replayed during analysis. However, the inclusiveness of data is affected by the field of the camera lens, and could lead to inaccurate interpretations. Field notes taken whilst videotaping helped to recover some of the losses incurred and avoid false interpretations (Jordon & Henderson 1995).
The study was approved by the appropriate ethics committee. After gaining consent, patients and nursing staff could withdraw at any time for no reason, and could also review the tapes that involved them and have them erased if desired; however, this did not occur. Invasion of privacy was minimized by not recording intimate care.
The process of gaining informed consent is difficult with patients who have severe communication problems. This was addressed by producing an alternative, simplified information sheet with pictorial reinforcement of key words to aid written comprehension, along with spending time talking through the literature with patients. Assent from their next of kin was also sought. Consent was obtained from all nursing staff who could potentially be involved in the recordings before commencing data collection. If a person who had not given consented came into the camera field, the recording was stopped.
Recordings were viewed to compile content listings of events, incorporating the field notes (Ten Have 1999). All interactions from the first patient were transcribed and analysed to find significant events and patterns. A 5-minute sample of interaction was chosen for the remaining patients’ recordings. These samples were chosen because they were the most data rich, and were analysed to see if they would support or challenge the patterns found with the first patient. Analysis was limited in this way because of the time restrictions of the researcher. Sequential analysis of the interactions was performed, including detailed transcription of verbal and non-verbal communication with eye gaze and body movement. The verbal transcription conventions used were adapted from those developed by Jefferson (2004) and Psathas (1995), and are routinely used in CA research. They are included in Table 1 to act as a reference for readers. The specific areas of focus during analysis were turn-taking organization, sequence organization, conversation repair and turn design (Perkins 1995, Heritage 1997). In this paper, we only discuss the first two of these.
In a conversational analysis approach the validity of the work is based on interpretation of directly observable data (e.g. turn design), and does not require subjective interpretation of the data (Psathas 1995).
Five patients were recruited, one was recruited from a rehabilitation ward and the others from an acute stroke unit. Their ages ranged from 43 to 84 years, and three were men. The time from stroke onset to recruitment ranged from 12 to 197 days. Three patients had aphasia and two had severe dysarthria. Those with aphasia all had expressive aphasia, and two also had receptive aphasia. Barthel scores (Mahoney & Barthel 1965) indicated that three patients were fully dependent for their activities of daily living (see Table 2). Fourteen nursing staff members participated in the study. Their nursing experience ranged from 6 months to 20 years, and included all levels of roles from healthcare assistant (HCA) to charge nurse (see Table 3).
Table 2. Patient characteristics
Barthel (out of 20)
Type of stroke
Time since onset
Type of ward
Left fronto-parietal haemorrhage
Left frontal infarct
Severe expressive aphasia, reduced comprehension (single stage commands)
Acute Stroke Unit
Right middle cerebral artery and cerebellar infarct
Acute Stroke Unit
Left middle cerebral artery infarct
Dysarthria Good comprehension
Acute Stroke Unit
Left parietal haemorrhage
Severe expressive and mild receptive aphasia. Previous mild expressive aphasia from a Stroke in 1996
Acute Stroke Unit
Table 3. Nurse characteristics
Mean (sd) years experience
Of the 35·5 hours of recording, there were 5 hours 43 minutes of nurse–patient interaction, illustrating that patients spent the majority of their time alone. Only twice were patients seen to be interacting with each other. Patients rarely sought to initiate social conversations with anyone apart from their visitors. Those with dysarthria occasionally attempted to start a conversation with patients or nursing staff, but this was never seen in those with aphasia. The recordings provided a wide range of interactions, from brief passings with eye contact and a wave, to conversations of up to 12 minutes in length. A total of 24 different interactions were transcribed, 19 for the first patient, and one each for the remaining four patients. Twenty-seven patterns emerged that appeared to be regardless of the type of nurse–patient interaction, and the most frequently occurring patterns were chosen to be analysed for all patients.
Institutional interaction had an overwhelming influence on the processes used. There was asymmetry in the interactions, with very limited patient contributions. Evidence can be seen through two analytically distinguishable, but interlinked conversation organizations: (a) turn taking and (b) sequence organization. Because of their interconnecting nature, they will be discussed together through analysis of conversation openings. Although three interactions will be discussed, all of the patterns presented in this paper were found in all patients. Participants’ names have been changed to maintain confidentiality.
The interaction in Figure 1 involves a patient and HCA. The patient had expressive aphasia, and his receptive communication was intact. The opening sequence starts before the HCA has entered the room, with her asking the patient if he has made his bed (line 01). She immediately focuses the conversation on her task. There is no social greeting as one would expect in ordinary conversation, and this was a question–answer sequence that the patient interpreted to be teasing in nature. This turn design, especially with the ‘teasing’ intonation, suggests that the HCA is judging his bed-making skills, because patients do not usually make their own beds in hospital.
There is a 1·2-second pause before the patient answers. This silence could have been construed as negative by the HCA, as silences are unusual in ordinary conversation unless preceding a negative turn; however, they are common in aphasic talk (Lesser & Perkins 1999). This pause is most likely to be for word finding, as the ensuing talk is not negative in context. The HCA has accommodated his aphasia by not interrupting his word-finding silence by either rephrasing the question or interpreting it as a negative answer, as could be expected in ordinary conversation. In line 03 the patient looks toward the bed, indicating non-verbally that he has understood the question, and starts his reply with ‘well’, followed by a pause. His downward intonation and pause could indicate the end of his turn, but he continues his turn with word finding speech, finishing with incomprehensible speech, while chuckling and waving of his hand at the bed. His verbal turn does not directly answer the question, but through his gestures and intonation of humour in his voice, he conveys that he did not make the bed very well. His turn design is also noteworthy as he does not reply with a closed answer that could be perceived as the most direct and easy answer, but manages to convey information with minimal verbal content.
In lines 04 and 05, the HCA overlaps the end of the patient’s turn by starting to laugh, and continues her turn with asking the patient if she should re-make his bed. By her continuing the next sequence in the conversation, it can be deduced that she understood the patient, because if she did not, she would have started a repair sequence, for example, asking the patient to repeat his reply. The HCA chooses another closed question, and the patient replies immediately (line 06) with a minimal turn, ‘yeah’, accepting her suggestion. The conversation then continues about negotiating the task of making the bed.
The second conversation in Figure 2 is between a staff nurse and patient with severe expressive and receptive aphasia. The patient is asleep, and the nurse draws the curtains before starting the conversation. In this extract, the nurse’s first language is not English. She wakes the patient by touch, then initiates talk with a closed question greeting, asking how he is feeling. The patient acknowledges the interaction by eye contact. He attempts a reply, and supplements his verbal response with shaking of his head. While asking for clarification of the patient’s last turn, (lines 04 and 05), the nurse has looked away and started on her task of checking his catheter, lifting up the bed covers. She interprets the non-reply from the patient as a positive reply (line 06) which, in normal talk would be perceived as a negative. She changes the topic to describe the reason for her interaction.
In comparison with those with aphasia, the two patients with dysarthria started three interactions, but only after nurses had already approached them to perform a task. In this extract (Figure 3) lunch is being served, and involves a patient with severe dysarthria.
The patient follows the nurse approaching with his gaze. He opens the conversation with a greeting as the nurse reaches his bedside with his lunch; however, he has dropped his gaze to look at the table, therefore his non-verbal communication could suggest that he is not ready to engage in the interaction despite initiating the greeting. A long 6·8-second pause occurs before the nurse takes her turn. During this pause she is arranging the table, but it could also be perceived as negative because she does not complete his greeting, but changes the talk to relate to her task of helping him with his lunch.
These openings illustrate the conversation organizations of turn taking and the sequential context of pursuing a task. In the entire data, nursing staff members started 88% (n = 20) of the opening sequences analysed. No patient with aphasia started an interaction. Fifty-four per cent (n = 9) of the interactions were related to a nursing task, for example, entering the room to administer medications. There were, however, five openings (19%) that appeared ‘social’ in nature and not directed by a task, but were only seen in interactions with the first patient on the rehabilitation ward. Therefore, the overarching theme in the interactions is asymmetry in the conversation, with nursing staff having the majority of control over the conversational floor. This is illustrated through turn-taking with nursing staff controlling conversation openings and speaker change, and patient staking small turn sizes and minimal turns of ‘yes’ or ‘no’ answers. Second, the sequential organization of the conversation enabled staff control through talk being directed around nursing tasks.
The generalizability of these findings may be limited because of the varying degrees of communication problems among patients, the fact that there were two different settings of acute and rehabilitation wards, and the small sample size. Small sample sizes are standard in CA research, mainly because of resources required to transcribe and analyse the data. Typically, CA studies have fewer than 20 participants, with the number of interactions ranging from single case studies to 100 different interactions. However, there is the possibility of contextual generalization. In CA the generation of social reality is through conversation, and so if similarities are found within these data and the conversational context in which the findings are to be generalized, then the findings could be used to inform those similar interactions. However, it should be noted that excluding personal care interactions may have had an impact on the type of data collected, as more personal and affective communication could occur during times of personal care.
Discussion of findings
The findings from this study, although identified using a new methodology, support the findings of previous research that nurses tend to focus on physical care and to interact in what could be seen to be a superficial way (Caris-Verhallen et al. 1999, Pound & Ebrahim 2000, Anderson & Adamsen 2001). None of the interactions in the data involved discussions that could be expected in a rehabilitation setting, for example about the impact of the stroke on people’s lives, or even topics of interest to patients, but instead related to nursing tasks. Researchers have described the qualities of effective nurse–patient communication as respectful, therapeutic and an individualistic approach with commitment to the communication process (Pound & Ebrahim 2000, Sundin et al. 2000, 2002). The conversations in our study, because of their task focus, appeared to be neither therapeutic nor individualistic. This is illustrated in particular through the nursing staff member governing the topic of the talk, and hence not being responsive to patients’ individual needs. The use of closed questions leading to patients taking short turns also led to a less therapeutic approach, not only with regard to speech therapy but also to therapeutic communication essential to the rehabilitation process.
We observed that nurse–patient dialogue is institutional in its organization, talk being focussed on the goal of one of the participants, in this case the nursing staff member. However, in this study this was more pronounced than in other studies examining healthcare professional–patient dialogue (Heath 1992, Maynard 1991, Parry 2004). This was probably because of the patients’ communication deficits. The findings suggest that dialogue in a setting which should be enabling and therapeutic both for development of the patients’ speech in accordance with their rehabilitation goals and the support and information that they receive during their stroke recovery, did not occur. Patients without communication problems can use their conversation skills to direct the talk away from nursing tasks to what they want to discuss, which then the nursing staff member can respond to appropriately. This is not possible for patients with aphasia/dysarthia, which means that there could be inequalities in stroke rehabilitation care for patients with communication deficits.
In previous studies, the rationale for ineffective nurse–patient communication included high workloads and time pressures, as well as tasks preventing nurses picking up patients’ cues (Caris-Verhallen et al. 1999,Dennison 1995, Pound & Ebrahim 2000). Our study did not provide data to support or contradict these rationales; however, we suggest that the reason behind nursing staff controlling the interaction, including when and about what they talk, is complex and multifactorial. First, healthcare professional interactions tend to occur in the context of the professional having ‘superior knowledge’ compared with patients, which is thought to be the foundation for asymmetry (Heath 1992). Due the nature of aphasia and dysarthria, patients’ turns are unpredictable and problematic, and may even be seen as a threat to superior professional knowledge. Problems with understanding undermine nursing staff members’ competence as communication partners, and therefore they may find it uncomfortable conversing with patients to whom they feel ill-equipped to talk. This may explain why they appear to control the interaction excessively, and is compounded by the fact that those with aphasia rely on the sequential context of talk to aid understanding, resulting in little opportunity to influence conversations unless others provide the openings.
Previous researchers who have investigated the education of nurses to improve communication with patients have shown little benefit (Caris-Verhallen et al. 2000, Wilkinson 1991). A reason for this could be that the institutional influence demonstrated in our study was not addressed, as the ward culture may have a major impact on conversations. If the cultural emphasis were to move away from the hierarchy which can be seen in ‘traditional’ medical settings and more towards partnership working, this might diminish some of the institutional influences. The degree of asymmetry demonstrated in this study could reflect the fact that the majority of the data were collected on an acute stroke unit; this could be said to have a more biomedical emphasis than on a rehabilitation unit.
Effective patient communication is vital for participation in the rehabilitation process, and we suggest that there needs to be improvement in the dialogue between nursing staff and patients with communication deficits after stroke. Currently, the institutional influences on conversations do not seem to be widely recognized in nursing practice. Increased awareness of the extent of the institutional influence on day-to-day dialogue could foster more conversations with patients which are not solely linked to nursing tasks. A cultural change within the ward environment, with a commitment from the team to consider the institutional dynamics that may currently hinder effective communication, could enable nursing staff to improve their communication during their day-to-day clinical practice.
Further specific education in how to be an effective communication partner might also help to improve the interaction between staff and patients with aphasia/dysarthria, and enhance the skills necessary to be able to follow speech therapy guidance. Nursing staff could gain more confidence in their own conversational skills if they were supported to learn how to manage hesitations, silences and the differing rules related to people speaking with aphasia. Communication needs to be balanced with the practical tasks required of the rehabilitation nursing role. This paper highlights the importance of giving communication rehabilitation equal importance to the physical rehabilitation of stroke survivors. This includes recognition of the increased time resources required for patients with communication problems.
There is a need for further research describing interaction between nursing staff and patients with aphasia/dysarthria. Using larger sample sizes from both acute and rehabilitation ward settings would improve the validity and reliability of the findings. It might be useful to carry out interviews with the observed nursing staff members and patients to help to ascertain the underlying influences that lead to interactional asymmetries. Evaluation of educational strategies to improve communication would also be useful, as well as comparison of patients with or without communication disorders to describe the differing degree of asymmetry between the two groups.
We would like to thank the patients and nursing staff who participated in this study, and the speech and language therapists who helped with the practicalities of conversation analysis and recruitment.
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
CG, AA and CEH were responsible for the study conception and design. CG performed the data collection, data analysis and was responsible for the drafting of the manuscript, and provided administrative, technical or material support. AA and CEH made critical revisions to the paper for important intellectual content and supervised the study.