The mechanism behind a friendly chat: 'Puzzle gets unravelled bit by bit'
photography: Anna Loh
A friendly chat is more complicated than you might think. As soon as the other person finishes talking, you already have an answer ready. But how do we know when it's time to change turns? University lecturer Johanneke Caspers has been awarded an NWO Open Competition grant to investigate the role of intonation in turn-taking.
'We always take turns and rarely talk over each other. There is a system behind this that makes you subconsciously know who is taking turns and when the turn may change,' explains Caspers. 'This turn-change system works very precisely: there is an average of one-fifth of a second between turns.'
This is in fact too short to respond adequately. 'We know from research that it takes longer to listen to someone, process the message and think of a message yourself than it does to change turns,' says Caspers. 'But you sometimes only recognise a question at the end of the sentence because the intonation rises. "He is coming with us too" is a statement, but rising intonation turns it into a question: "He is coming with us too?" It's a mystery for now how we know that so quickly. We don't know if that intonation is redundant or if we are actually using that information.'
Corpus analysis
The first phase of the research focuses on corpus analysis. Most of the corpus has already been annotated, but Caspers believes it should be even more detailed. 'Since we know that we often don't have enough time to do anything with the information at the end of a sentence, perhaps we should look at other aspects of a conversation, such as pragmatics,' she says. Pragmatics is about interpreting language and it depends on the context of the conversation. For example, a friend who says it is warm in a room may actually be making a request to open a window. Caspers hopes to gain more information about the role of intonation in turn-taking.
Tricky research design
Next, Caspers wants to see in an experimental study whether reality matches the corpus research. That is a tricky job, as turn-taking turns out to be almost impossible to manipulate in a test setting. 'Even if there is only a minimal delay, you continuously interrupt each other. You also saw this during video calling in corona time. The turn-taking system immediately stopped working as it should,' says Caspers.
Caspers is therefore taking an approach that mimics a real conversation. 'Subjects have to listen to a conversation in a booth and indicate when they think a speaker will speak,' she explains. At the same time, Caspers uses eye tracking to see which speaker the subjects are looking at. 'In this way, we hope to see at what moment the information enters the subject's mind so that you know who will take over which turn and what role intonation plays in this, so that we can unravel the puzzle of the turn-taking system bit by bit.'