Why Behavioral Observation Is A Treat to Couple Research – And How The Sparse Data Problem Can Be Solved

Session To Be Scheduled

Short Abstract

Inferences about couple interactions are typically based on behavioral observation data. But codeable incidents are rare – and, as a consequence, codeable incidents are summed up across entire interactions. However, this removes the sequential nature of the interaction, undermining the field’s ability to study this fundamental phenomenon. We demonstrate how spares coded behavior is in a sample of 189 couples - and how engineering technology can be used to extract behavioral data in high temporal resolutions to overcome this problem. Results show that couples influence each other in 95% of all talk turns and even change the way they influence each other during the curse of a single interaction.

Full Abstract

Intimate relationships rely heavily on interactions – it’s how partners connect, argue, plan their future, provide support, and solve problems. Inferences about interactions are typically based on behavioral observation data, which entail the interaction’s sequences. Scholars intuitively assume that this method can be used to study how partners influence each other during interactions. But codeable incidents are generally rare – and, as a consequence, forced the field to sum-up codeable incidents across interactions. This is problematic as it removes the sequential and temporal nature of the interaction and undermines the field’s ability to examine how people influence each other during interactions – arguably the essential mechanism for intimate relationships. The aim of this presentations is twofold. Based on an existing data set of 189 couples (N=378 participants), we (i) demonstrate that codeable incidents (positive and negative non-verbal behavior) are so rare that there is no other option than to sum up each individual’s behavior over the course of an 8 minutes interaction, supporting our assumption that this is a fundamental problem with substantial consequences. In order to overcome this problem, (ii) we show how engineering methods can be used to extract behavior in high temporal resolutions. This in turns allows us to use a latent differential equation model (LDE) to examine how changes in non-verbal vocally encoded arousal of one person influence the non-verbal vocally encoded arousal. We found even evidence that the dynamic how people influence each other during an interaction a change within an interaction. This is another step towards tackling this fundamental problem. We argue that the engineering tools have now been developed over the last years so that the field of couple research can collectively confront this problem. Not only will this allow us to study how human mutually influences each other during an interaction but how this provides new possibilities for inter-disciplinary collaborations with engineers and computer scientist.

Our Additional Authors

Name Affiliation Email
Matthew Vowels University of Surrey