Do patients with schizophrenia do dialogue differently? Christine Howes, Mary Lavelle, Patrick G.T. Healey, Julian Hough, Rose McCabe JAM 2017 26th July 2017
Background 1 Method 2 Results 3 Conclusions 4 2 / 29
Background 1 Method 2 Results 3 Conclusions 4 3 / 29
Disfluencies. . . such as self-repairs, filled pauses (‘erm’, ‘uhh’) and unfilled pauses. . . are pervasive in dialogue (Schegloff et al., 1977) are regarded as symptomatic of problems with communication (Levelt, 1983) – issues with self-monitoring or production are also interactive (Goodwin, 1979) – they occur in direct response to feedback are associated with increased gesture Seyfeddinipur and Kita (2014); Healey et al. (2015) – people use additional multimodal resources 4 / 29
Patients with schizophrenia. . . . . . experience problems in interaction have difficulty monitoring their own behaviour (Johns et al., 2001) are less effective at ‘meshing’ turns (Mueser et al., 1991) use fewer speaking gestures (Lavelle et al., 2013) have mismatched gesture use and speech (Millman et al., 2014) use fewer self-repairs (Caplan et al., 1996) . . . or more self-repairs? (McCabe et al., 2013; Howes et al., 2012) . . . or the same amount? (Leudar et al., 1992) 5 / 29
Studies on patients. . . usually rely on non-interactive data (e.g. monologues, offline tasks) or dialogues with known interlocutors (e.g. therapist, family members) or ‘dialogues’ with experimenters who know the patient’s diagnosis What happens in social interactions with unknown partners? 6 / 29
Studies on interactions. . . show that the presence of a patient with schizophrenia influences the non-verbal behaviour of their interacting partners: in clinical contexts (Lavelle et al., 2015) during first meetings with healthy controls, when the patient’s diagnosis is undisclosed (Lavelle et al., 2013, 2014) Does interaction itself play a crucial role in patients’ deficits in dialogue? Do their healthy partners modify their own behaviours? 7 / 29
Background 1 Method 2 Results 3 Conclusions 4 8 / 29
Data Triadic conversations of approximately 5 minutes 20 patient interactions one patient two healthy controls who were unaware of the patient’s diagnosis 20 control interactions three healthy participants 9 / 29
Who would you throw out? William Harris: Balloon pilot Susanne Harris: 7 months pregnant wife Robert Lewis: Cancer research scientist Heather Sloan: Musical child prodigy 10 / 29
Data (2) Motion capture (1) Video 11 / 29
(3) Transcriptions A The lady’s, the lady’s pregnant, so we’re saying we don’t want to throw her out C [Well, I don’t know, we can chuck her] A [And the little child], the little child, I don’t think she’s any use C Yeah, [who let a nine year old child go on a hot air balloon?] A [But, it it’s just because she’s young] B [Yeah]. A [But, sh-] even if they did throw her out, she’d be quite light B Yeah A So � pause 220ms � she wouldn’t really [make a difference] C [So you] should throw out the pregnant woman, [because erm, she’s probably gonna be the heaviest, and]= A [She’s two people] C =I’m guessing they’re all like normal weight A Ummm � pause 3200ms � I think the husband B Yeah, but he’s he’s the only one with any flying experience 12 / 29
Analysis Gesture (Lavelle et al., 2012) derived from MoCap hand movements data hand movement speeds > 1 sd above an individual’s mean hand movement speed Self-repair (Hough and Purver, 2014) STIR (STrongly Incremental Repair detection) state-of-the-art incremental disfluency detector Filled pauses (Howes et al., 2014) Find-and-replace er , eerrrrmm , uhmmm etc → er Unfilled pauses (Zellner, 1994) Speech-free spaces > 200ms between segments of speech by the same speaker Rates per turn and per 100 words calculated for each individual 13 / 29
Automatic annotations A The lady’s, the lady’s pregnant, so we’re saying we don’t want to throw her out C [Well, I don’t know, we can chuck her] A [And the little child], the little child, I don’t think she’s any use C Yeah, [who let a nine year old child go on a hot air balloon?] A [But, it it’s just because she’s young] B [Yeah]. A [But, sh-] even if they did throw her out, she’d be quite light B Yeah A So � pause 220ms � she wouldn’t really [make a difference] C [So you] should throw out the pregnant woman, [because erm, she’s probably gonna be the heaviest, and]= A [She’s sh- sh- two people] C =I’m guessing they’re all like normal weight A Ummm � pause 3200ms � I think the husband B Yeah, but he’s he’s the only one with any flying experience 14 / 29
Automatic annotations A The lady’s, the lady’s pregnant, so we’re saying we don’t want to throw her out C [Well, I don’t know, we can chuck her] A [And the little child], the little child, I don’t think she’s any use C Yeah, [who let a nine year old child go on a hot air balloon?] A [But, it it’s just because she’s young] B [Yeah]. A [But, sh-] even if they did throw her out, she’d be quite light B Yeah A So � pause 220ms � she wouldn’t really [make a difference] C [So you] should throw out the pregnant woman, [because erm, she’s probably gonna be the heaviest, and]= A [She’s sh- sh- two people] C =I’m guessing they’re all like normal weight A Ummm � pause 3200ms � I think the husband B Yeah, but he’s he’s the only one with any flying experience 14 / 29
Automatic annotations A The lady’s, the lady’s pregnant, so we’re saying we don’t want to throw her out C [Well, I don’t know, we can chuck her] A [And the little child], the little child, I don’t think she’s any use C Yeah, [who let a nine year old child go on a hot air balloon?] A [But, it it’s just because she’s young] B [Yeah]. A [But, sh-] even if they did throw her out, she’d be quite light B Yeah A So � pause 220ms � she wouldn’t really [make a difference] C [So you] should throw out the pregnant woman, [because erm, she’s probably gonna be the heaviest, and]= A [She’s sh- sh- two people] C =I’m guessing they’re all like normal weight A Ummm � pause 3200ms � I think the husband B Yeah, but he’s he’s the only one with any flying experience 14 / 29
Background 1 Method 2 Results 3 Conclusions 4 15 / 29
Overview Patient Partner Control Turns 43.78 56.22 57.43 Words 247.67 439.03 399.63 total Self-repair 3.56 10.13 10.49 Filled pause 2.11 4.47 8.08 Unfilled pause 14.94 32.92 22.00 Words per turn 5.583 7.993 7.215 per turn Self-repair 0.081 0.186 0.181 Filled pause 0.046 0.081 0.140 Unfilled pause 0.362 0.650 0.426 per 100 Self-repair 1.330 2.203 2.507 words Filled pause 0.774 1.004 1.811 Unfilled pause 6.582 7.691 5.929 16 / 29
Self-repair significant main effect of condition ( F 2 , 96 = 10 . 71 , p < 0 . 01) patients use fewer self-repairs than partners ( t 1 , 96 = − 3 . 48 , p < 0 . 01) patients use fewer self-repairs than controls ( t 1 , 96 = − 3 . 64 , p < 0 . 01 same pattern for self-repairs per 100 words 17 / 29
Filled pauses significant main effect of condition ( F 2 , 102 = 5 . 93 , p < 0 . 01) patients fewer filled pauses than controls ( t 1 , 102 = − 3 . 35 , p < 0 . 01) patients’ partners also fewer than controls ( t 1 , 102 = − 2 . 36 , p = 0 . 02) same pattern for filled pauses per turn 18 / 29
Unfilled pauses significant main effect of condition ( F 2 , 102 = 5 . 34 , p < 0 . 01) patients’ partners use more than patients ( t 1 , 102 = 3 . 24 , p < 0 . 01) patients’ partners use more than controls ( t 1 , 102 = 2 . 48 , p = 0 . 02) no significant effects by 100 words 19 / 29
Gesture patients fewer gestures as speaker repair rate and gesture correlated in controls 20 / 29
Background 1 Method 2 Results 3 Conclusions 4 21 / 29
During social interaction patients with schizophrenia . . . repair their own speech less use fewer filled pauses (‘er’, ‘um’) gesture less when speaking no association between gesture and speech patients’ healthy interlocutors also . . . use fewer filled pauses than controls use more unfilled pauses no association between gesture and speech 22 / 29
Implications for turn-taking? Reduced competition for the floor in patient interactions? turn holding techniques (filled pauses) less necessary increased within-turn pauses for patients’ partners where they expect/encourage the patient to take the floor. patients less responsive to turn taking cues and/or more reluctant to select as next speaker? 23 / 29
Future directions This is a really broad brush approach by participant over the whole conversation but suggests we should look more closely at turn-taking (including gesture and gaze?) and repair and the usefulness of automatically derivable data from speech (potentially live ) (the start of) a fruitful line of research in investigating the difficulties experienced by patients in social interaction . . . 24 / 29
Recommend
More recommend
