Spontaneous and Hierarchical Segmentation of Non-functional Events

Kristoffer Laigaard Nielbo

Publikation: AfhandlingPh.d.-afhandling


The dissertation, Spontaneous and Hierarchical Segmentation of Non-functional Events (SHSNE henceforth), explores and tests human perception of so-called non-functional events (i.e., events or action sequences that lack a necessary link between sub-actions and sequence goal), which typically are embedded in or constitute ritual behavior. SHSNE applies a dual approach in its empirical part that combines behavioral experiments from psychology with computer simulations. This dual approach is chosen with the purpose of constructing a mechanistic model of perception of non-functional events. SHSNE has four main parts that span research history and contemporary theories of perception and cognition (Theory), methodology of the dual approach (Methodology), empirical investigations (Synthesizing and analyzing perception of non-functional events), and a general discussion (General discussion).Part 1 is a review and critique of ritual theories within the study of religion, anthropology and ethology. It is emphasized that the majority of these theories have committed two fallacies: a) the linguistic fallacy (i.e. to model ritual behavior through formalisms borrowed from linguistics); b) the sui generis fallacy (i.e., to model ritual behavior as a unique category that is not subject to general behavioral taxonomies). As an alternative, SHSNE advances a theoretical position, called behavioral formalism, which bases behavioral models on measurable physical features of events, builds simple mechanistic models, and applies general scientific method. To rectify the mentioned fallacies, part 1 then reviews contemporary theories within the fields of perceptual causality, prediction machines, and event perception. This part finally defines SHSNE’s object (i.e., non-functional events) that is used to model ritual behavior. Part 1 concludes with five primary theoretical hypotheses: I) non-functional events will increase the human event segmentation rate; II) transitions between events will increase the cognitive prediction error signal independent of event type, but this signal will be chronically high for non-functional events; III) representations of non-functional events have less hierarchical organization than representations of functional events; IV) context information will modulate event representations; V) representational competition will occur between non-functional and functional event representations to the degree that non-functional events are derived from functional events.Part 2 is the methodological part that has two main sections: a) a general discussion of the concept of a scientific model and the so-called Third Mode of Science position; b) a conceptual and application-oriented introduction to the use of computer simulations and the dual approach. Because the methods applied in SHSNE might not fit into the humanities’ methodological toolbox, issues pertaining to this field of research are emphasized.Part 3 is SHSNE’s empirical part, which presents two sets of behavioral experiments, consisting of four experiments in total, and four computer simulations. The first set of experiments shows that the event segmentation rate increases for human participants that observe non-functional events compared to functional events. Furthermore, it appears that context information does not have the predicted modulation effect of the event segmentation. The second set of experiments show that the hierarchical organization of event representations is reduced by non-functional events. There is a tendency that context information does modulate hierarchical organization, but it is not significant given the sample size. The experiments are simulated through four computer simulations, which show that non-functional events induce a chronically high prediction error signal and that this signal can be used to classify types of events. The simulations further show that context information can have a modulation effect on the prediction error signal induced non-functional events and that representational competition is a central explanation of the signal’s magnitude. Finally, spike synchrony is demonstrated between the prediction error signals elicited by non-functional events with and without context information associations and, further, that the simulations’ ecological validity can be increased through the use of input based on a motion capture device, while the results still support the main findings.Part 4 is a general discussion, which reviews the empirical findings and the corresponding model in relation to ritual studies, the study of religion, and the event perception literature. It is argued that an adequate model of perception of non-functional events and ritual behavior takes the form of a prediction machine based on recurrent wiring, similar to the artificial neural networks used in part 3.
StatusUdgivet - 2012
Udgivet eksterntJa