Research Overview

What is conversation alignment?
Conversation alignment is an area of research that looks at how people tend to start using the same sorts of language as they talk to each other. For example, if you call my pet a ‘bunny’ rather than a ‘rabbit’, then I am likely to use the word ‘bunny’ as well. I am aligning, or matching, my words to yours. We could match our conversation in different ways: we could match the words we use, or the way we structure our conversation. For example, I might use a ‘passive’ verb: The boy was chased by the dog, rather than an active verb: The dog chased the boy. Those are both good ways of describing something that happened. But just as with the rabbit/ bunny, people tend to copy, or align how they structure their speech.

Why do people show conversation alignment?
Some research shows that this alignment makes conversation easier: if we use the same words and structures as the other person, we don’t have to think so hard to come up with a different way of saying something. Also, by aligning simple things like words, this means that people easily reach a shared understanding about what they are talking about. There are also studies showing that people who align their conversation feel more positive to each other. So, it might be that aligning our conversation helps us understand each other better, and like each other better.
How do people align their conversations?
There are different ideas about why people show alignment. One idea is that we don’t align on purpose: it seems that people do this without noticing. Other ideas are that people align because it helps their conversation partner to understand them better, or because they want to show that they like their conversation partner.

Does everyone show conversation alignment?
Autistic people certainly sometimes report that they find conversation with other people difficult. Could differences in alignment processes be part of this? Some researchers have suggested that aligning conversation might be more difficult between an autistic and a typically-developing person. There are different ideas about the reasons for this. Some people say it is because autistic people do not have the same understanding of other people – a ‘theory of mind’ difference. Other people have suggested that people with autism might not align in the same way as neurotypical people do, because autistic people find it harder to understand the moods and emotions of others, or because their attention processes work differently.

How do we study conversation alignment in the laboratory?
One of the most common ways of studying alignment is a game that Holly Branigan developed. It is a bit like the card game ‘Snap’. The Experimenter (E) and the Participant (P) sit opposite each other, each with a pile of cards showing drawings of different events –for example, a crocodile kicking a knight, or a cow carrying a queen. E and P take it in turns to put cards down, and when they place a card, they have to describe what the card shows. E deliberately describes the pictures in different ways, to see if P copies, or aligns, their language. For example, if the E uses the less common ‘passive’ description ‘The knight is kicked by a crocodile’, will the P also use this to describe their picture? ‘The queen is carried by a cow’
We analyse the results with various different statistical methods, taking into account the possibility that one way of saying something might be more common than another way –the ‘preferred’ form. The papers give the full details of how we do this.

What do our research studies show?
These are basic descriptions of the papers we have published, about the studies we have done. Full versions of most of these are available in the Documents section.

Hopkins, Yuill & Keller 2016 Children with autism align syntax in natural conversation
This paper looked at whether children with autism really do show less conversation alignment than typically-developing children do, for different language structures. We test this in two different ways: looking at a laboratory task and in natural conversation.

Study 1: Snap game in the laboratory
17 autistic children and 17 neurotypical children aged between 6 and 13 played the Snap game for two different language constructions: passive structures (e.g. ‘the boy was kicked by the queen’) and ‘dative alternation’ – for example, the different structure of ‘the tiger gave the ball to the nurse’ and ‘the tiger gave the nurse a ball’. The results showed that the autistic children did show alignment, as the neurotypical children did. This led us to wonder why, then, autistic children often say they find conversation difficult. If the children aligned their language, this should mean that they share understanding with their conversation partner. We thought that perhaps the autistic children might align in a lab study, but perhaps not in their everyday conversation. This is what we looked at in Study 2.

Study 2. Alignment in natural conversation
Researchers have not studied this very often. The reason for this is that it is quite hard to do: natural conversation can be about anything, and people might not use the particular words or structures we are looking for, and we can’t be so sure that someone has used a particular word just because it appeared earlier in the conversation. That’s why so many studies look at alignment in more controlled situations. We developed an automated method to measure whether one person in a conversation was likely to have used a word because the conversation partner just used it in their turn in the conversation.

We recorded conversations between a typically-developing child (who acted as a conversation partner) and an autistic child. Our participants were 17 autistic and 17 typically-developing 6- to 13-year-olds. Each participant had a different typical conversation partner. We asked the children to talk about their favourite pets: this was a topic that everyone was able to talk about. We then analysed their conversations to look for alignment in the words they used.
Our results showed that autistic children, and typically-developing children, both showed alignment in their conversations. We also found that children aligned more in the first, lab study, than in this second natural conversation study. This shows that in the Snap game, we get more conversation alignment than in natural conversation, but we do get some alignment in conversation. We also think that the results show that children were aligning automatically –that is, they were not copying their partner deliberately.

We don’t know why the autistic children showed this alignment. We did test the children on their basic ‘theory of mind’ ability, and children showed good understanding. We don’t know if they would be able to understand a more complex theory of mind task. This means we can’t say definitely whether people need ‘theory of mind’ to align their conversation. Also, we don’t know if there are differences between the autistic children and the typical ones in why they align. There could be different reasons, but our results don’t tell us about these. The results tell us about aligning at the level of structure (for example, the passive and dative structures we looked at). But we don’t know whether aligning at this level makes understanding at a higher level easier. So children might align in a short sentence, but does this make it easier for them to understand more broadly what the other person is talking about?

Hopkins, Yuill & Branigan, 2017 Inhibitory control and lexical alignment in children with an autism spectrum disorder
This study looked at two different questions. First, we wanted to know if autistic children showed conversation alignment in their choice of words. If I describe my pet as a ‘bunny’ and not a ‘rabbit’, does this mean you are more likely to describe it as a bunny as well? Second, we wanted to look at the ‘conflict inhibition’ and ‘theory of mind’ explanations of alignment.

The ‘conflict inhibition’ idea is that people have an idea on their head of what word they usually use (e.g. ‘rabbit’), but if they hear their conversation partner say ‘bunny’, they may tend to align, or use the same word. That means they have a conflict between using ‘bunny’ and ‘rabbit’. They have to stop themselves saying ‘rabbit’ as they usually do, and say ‘bunny’ instead. If children find it hard to stop themselves saying the usual word, they will be less likely to align, or copy, the word their partner uses.

The ‘theory of mind’ idea says that people align the words they use because they put themselves in their conversation partner’s shoes. If they hear their partner say ‘bunny’, then they know that the person understands that word. That means it will help to use the same word when talking to that person, because we know they will understand it.

We also wanted to see whether our results depended on children’s age, and their language level. This means that we compared autistic children with typical children of the same age, and typical children who had the same language level. This second group of typical children were younger than the autistic children. This method is often used to see whether it is years of experience, or language level, that is more important in affecting our conversation ability.

In the first study, 12 autistic children aged 8-12, 12 typically-developing children of the same age, and 14 typically-developing children with the same language level as the autistic group, played the Snap game. We also gave the children a game to test their theory of mind (ToM) understanding, and another task to look at how quick they were at stopping themselves from saying something (conflict inhibition). The ToM task was more complicated than the one we used in Hopkins, Yuill & Keller paper because we wanted to see if conversation alignment needed children to understand some harder questions about ToM.

We found that all three groups of children showed the same amount of conversation alignment. Also, children who did less well on the conflict inhibition task did not do any differently on alignment. That means we did not find any evidence that alignment is because of conflict inhibition. We thought there were some problems with the conflict inhibition task we did, so in Study 2, we looked at the same question, but with a different task. In this study too, we did not find that conversation alignment was better in children who did better on the conflict inhibition task. Conversation alignment did not link to the ToM task either. We said that the results do not support the two explanations of alignment, conflict inhibition and ToM. Instead we suggested that ‘priming’ might be a good explanation. This explanation says that just hearing a particular word in a conversation (e.g. ‘bunny’) is enough to get us to choose that word when we reply, even without realising it. Sometimes, priming can lead to a good conversation, but sometimes it might lead us to using a word or structure that is hard to understand.

Branigan, Tosi and Gillespie-Smith Spontaneous lexical alignment in children with an Autistic Spectrum Disorder and their typically developing peers
This paper looked at conversation alignment with words (e.g. rabbit and bunny) in autistic and typically-developing children. Like our other studies, this showed that autistic children did show conversation alignment with word choices. It is important not to rely on one single study to decide about conversation alignment. That’s why in science we look at the same questions in different ways and in different studies, to make sure that our results are reliable and can be repeated. This study also looked more closely at ideas about priming. With the autistic children, children who aligned more (imitated their partner’s word choice more) did not show higher social functioning, or better theory of mind. By ruling out these factors as affecting alignment, this leads us to think that alignment might happen because of priming: your partner’s use of a particular word makes that word more readily available for you to use.