© Koninklijke Brill NV, Leiden, 2013 DOI: 10.1163/15685373-12342098
Journal of Cognition and Culture 13 (2013) 347–365 brill.com/jocc
Prediction Error During Functional and Non-Functional
Action Sequences: A Computational Exploration of
Ritual and Ritualized Event Processing
Kristoffer L. Nielboa
,
* and Jesper Sørensena
,
b
a Department of the Study of Religion, Aarhus University,
Tåsingegade 3, 8000 Aarhus C, Denmark
b MINDLab, Aarhus University, Tåsingegade 3,
8000 Aarhus C, Denmark
* Corresponding author, e-mail: kln@teo.au.dk
Abstract
By means of the computational approach the present study investigates the difference between
observation of functional behaviour (i.e., actions involving necessary integration of subparts) and
non-functional behaviour (i.e., actions lacking necessary integration of subparts) in terms of
prediction error. Non-functionality in this proximal sense is a feature of many socio-cultural
practices, such as those found in religious rituals private and social, as well as pathological
practices, such as ritualized behaviour found among people suffering from Obsessive Compulsory
Disorder (OCD). A recent behavioural study has shown that human subjects segment non-
functional behaviour in a more fine-grained way than functional behaviour. This increase in
segmentation rate implies that non-functionality elicits a stronger error signal. To further explore
the implications, two computer simulations using simple recurrent networks were made and the
results are presented in this article. The simulations show that non-functional action sequences
do indeed increase prediction error, but that context representations, such as abstract goal
information, can modulate the error signal considerably. It is also shown that the networks are
sensitive to boundaries between sequences in both functional and non-functional actions.
Keywords
Ritualized behaviour, cultural rituals, computer simulation, prediction error, event perception
Introduction
During a single day, humans execute a considerable amount of action
sequences. The majority of these will most likely be instrumental or functional
in proximal terms (in an evolutionary model, non-functional action sequences
might naturally serve a function, such as generating or maintaining group
cohesion (e.g., Bulbulia, 2004; Alcorta and Sosis, 2005), but ultimate explanations