The salience of pulse perception in listeners attending to musical rhythms has beenmodeled as a simple resonator with a resonance frequency around 2 Hz (Van Noorden & Moelants, 1999). The resonance model explains the distribution of perceptual responses to a variety of rhythmic patterns and to large collections of music from different styles. However, in many individual pieces the tempo is ambiguous, allowing several metric levels to serve as the beat. It is not exceptional that the most salient metric level for an excerpt is quite far from the resonance frequency.

In a series of experiments, listeners were asked to tap to the beat of a large variety of musical excerpts. Group responses show that in most cases at least two beat periods were perceived by the subjects. The relative strength of the perceived periods varied widely from near equality to a one dominating the other. The ratio of the two determines the ambiguity of the tempo and can be considered an important element in the perception of the meter. For more ‘conventional’ musical styles, the dominant periods usually correspond to the predictions of the resonance model. However, in some excerpts, very ‘slow’ or ‘fast’ periods turn out to be dominant. If we look at the properties of those fragments, it seems that an important factor in determining the choice for ‘extreme’ metric levels, is the distribution of (dynamic) accents. This finding allows us to get a better image of how we exactly process the metric content of music and to build a model that can be applied in automated models for beat induction.