Limits...
A prototype-based resonance model of rhythm categorization.

Bååth R, Lagerstedt E, Gärdenfors P - Iperception (2014)

Bottom Line: This model is used to simulate the categorical choices of participants in two experiments of Desain and Honing (2003).The model accurately replicates the experimental data.Our results support resonance theory as a viable model of rhythm perception and show that by viewing rhythm perception as a dynamical system it is possible to model central properties of rhythm categorization.

View Article: PubMed Central - PubMed

Affiliation: Lund University Cognitive Science, Lund University, LUX, Lund, Sweden; e-mail: rasmus.baath@lucs.lu.se.

ABSTRACT
Categorization of rhythmic patterns is prevalent in musical practice, an example of this being the transcription of (possibly not strictly metrical) music into musical notation. In this article we implement a dynamical systems' model of rhythm categorization based on the resonance theory of rhythm perception developed by Large (2010). This model is used to simulate the categorical choices of participants in two experiments of Desain and Honing (2003). The model accurately replicates the experimental data. Our results support resonance theory as a viable model of rhythm perception and show that by viewing rhythm perception as a dynamical system it is possible to model central properties of rhythm categorization.

No MeSH data available.


An example of an activation pattern generated by feeding the resonance model a rhythm with IOIs 0.25 s, 0.5 s, and 0.25 s.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4441029&req=5

Figure 4: An example of an activation pattern generated by feeding the resonance model a rhythm with IOIs 0.25 s, 0.5 s, and 0.25 s.

Mentions: Prediction (B) implies that rhythm sequences resulting in similar states when given as input to a resonance model should be categorized similarly in an experimental task. A resonance model does not directly produce a categorization, but this is not required for testing this prediction. It is possible to compare the resulting states of two rhythm sequences by calculating the respective activation patterns and comparing these. A suitable similarity measure is given by considering the activation patterns as points in an n-dimensional space, where n is the number of oscillators in the resonance model, and then taking the Euclidean distance between these two points, where a shorter distance corresponds to more similar states. Considering the twelve most common rhythm categories chosen by the participants in Desain and Honing's study as prototype categories, it is possible to use the rhythm sequences corresponding to these categories to generate prototype activation patterns. For example, to generate the prototype activation pattern for the category 1-2-1 (as shown in Figure 4) the rhythm sequence with IOIs 0.25 s, 0.5 s, and 0.25 s would be used as input to the resonance model. A rhythm sequence's activation pattern can then be compared with these prototypes' activation patterns and the prototype category with the most similar activation pattern can be assigned to that rhythm sequence. In this way, all rhythm sequences can be assigned a category and these categories can be compared with the categories selected by the participants in Desain and Honing's study. Specific hypotheses are then that a resonance model categorization of the stimulus used by Desain and Honing should be similar to the categorization made by the participants in the no meter, duple meter, and triple meter conditions. Furthermore, as the participants' categorizations of the rhythm sequences in the no meter condition were more similar to the categorizations made in the duple meter condition than to the categorizations made in the triple meter condition, the same relation should be present in the categories generated by the resonance model.


A prototype-based resonance model of rhythm categorization.

Bååth R, Lagerstedt E, Gärdenfors P - Iperception (2014)

An example of an activation pattern generated by feeding the resonance model a rhythm with IOIs 0.25 s, 0.5 s, and 0.25 s.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4441029&req=5

Figure 4: An example of an activation pattern generated by feeding the resonance model a rhythm with IOIs 0.25 s, 0.5 s, and 0.25 s.
Mentions: Prediction (B) implies that rhythm sequences resulting in similar states when given as input to a resonance model should be categorized similarly in an experimental task. A resonance model does not directly produce a categorization, but this is not required for testing this prediction. It is possible to compare the resulting states of two rhythm sequences by calculating the respective activation patterns and comparing these. A suitable similarity measure is given by considering the activation patterns as points in an n-dimensional space, where n is the number of oscillators in the resonance model, and then taking the Euclidean distance between these two points, where a shorter distance corresponds to more similar states. Considering the twelve most common rhythm categories chosen by the participants in Desain and Honing's study as prototype categories, it is possible to use the rhythm sequences corresponding to these categories to generate prototype activation patterns. For example, to generate the prototype activation pattern for the category 1-2-1 (as shown in Figure 4) the rhythm sequence with IOIs 0.25 s, 0.5 s, and 0.25 s would be used as input to the resonance model. A rhythm sequence's activation pattern can then be compared with these prototypes' activation patterns and the prototype category with the most similar activation pattern can be assigned to that rhythm sequence. In this way, all rhythm sequences can be assigned a category and these categories can be compared with the categories selected by the participants in Desain and Honing's study. Specific hypotheses are then that a resonance model categorization of the stimulus used by Desain and Honing should be similar to the categorization made by the participants in the no meter, duple meter, and triple meter conditions. Furthermore, as the participants' categorizations of the rhythm sequences in the no meter condition were more similar to the categorizations made in the duple meter condition than to the categorizations made in the triple meter condition, the same relation should be present in the categories generated by the resonance model.

Bottom Line: This model is used to simulate the categorical choices of participants in two experiments of Desain and Honing (2003).The model accurately replicates the experimental data.Our results support resonance theory as a viable model of rhythm perception and show that by viewing rhythm perception as a dynamical system it is possible to model central properties of rhythm categorization.

View Article: PubMed Central - PubMed

Affiliation: Lund University Cognitive Science, Lund University, LUX, Lund, Sweden; e-mail: rasmus.baath@lucs.lu.se.

ABSTRACT
Categorization of rhythmic patterns is prevalent in musical practice, an example of this being the transcription of (possibly not strictly metrical) music into musical notation. In this article we implement a dynamical systems' model of rhythm categorization based on the resonance theory of rhythm perception developed by Large (2010). This model is used to simulate the categorical choices of participants in two experiments of Desain and Honing (2003). The model accurately replicates the experimental data. Our results support resonance theory as a viable model of rhythm perception and show that by viewing rhythm perception as a dynamical system it is possible to model central properties of rhythm categorization.

No MeSH data available.