Logo Logo
Switch language to English
Ganzenmüller, Stephanie (2013): Effects of sensory feedback on duration reproduction: a Bayesian approach to characterize sensorimotor temporal estimation. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)



Most studies, investigating human time perception, have demonstrated a difference between subjective and objective timing. Very common are, for example, results showing that visual intervals are judged shorter than physically equivalent auditory intervals. Recent studies have also found differences between motor and perceptual timing. Considering those perceived differences, the idea has been proposed that the brain might employ distributed (modality- specific) timing mechanisms rather than one amodal timing mechanism. Distributed timing mechanisms and therefore independent temporal estimates would be convenient in the computation for reliability-based multisensory or sensorimotor integration, as predicted by Bayesian inference. Several studies have shown that multisensory temporal estimates can be predicted by reliability-based integration models, as for example the Maximum Likelihood Estimation (MLE) model. Reliability-based integration studies in time research are still fairly rare and discussed controversially, and especially studies investigating sensorimotor integration are mostly missing. The aim of this cumulative thesis was to investigate sensorimotor temporal reproduction with a focus on the influence of sensory (mainly auditory) feedback on motor timing. Here fore, in all studies a sensorimotor temporal reproduction paradigm was employed, and sensory and motor estimates were treated as different/independent estimates. First, we investigated the effect of onset and offset delayed sensory feedback on temporal reproduction (Chapter 2.1). Second, perceptual and motor timing were compared explicitly and then a reliability-based model was used to predict the observed sensorimotor reproduction times (Chapter 2.2). In a third study, we manipulated the prior representation of the standard duration, using different adaptation conditions (Chapter 2.3). The findings showed that if the onset of a feedback stimulus was delayed in relation to an action (in contrast to when the feedback signal was started before the action), reproduced durations increased immediately, as soon as a delay is introduced. Offset-delayed sensory feedback, on the other hand, only induced a minor decrease in reproduction times and this effect could only be observed with auditory feedback. In comparison to auditory comparison estimates, which were shown to be fairly precise, pure motor reproduction as well as auditory reproduction was found to be consistently overestimated. The observed overestimation bias in auditory reproduction was reduced, compared to pure motor reproduction. This pattern of result could be shown for various standard durations and different signal-to-noise ratios (SNR) in the compared/reproduced tones. Further, a reliability-based model 4 predicted observed auditory reproduction biases successfully. In the third study, we could show that shifting the temporal range of accuracy feedback, manipulating the SNR of the reproduced tone, as well as introducing a manipulation of the reproduced tone onset, led to significant changes in the prior representation of the standard duration. Only manipulating the reproduced tone onset during the adaptation phase induced a reduction of auditory weights, which could be observed during the test phase. Additional trial-wise analysis confirmed that the adapted prior representation is shifted back to normal dynamically over time, once no accuracy feedback is provided anymore. The differences between observed sensory and motor estimates of time are discussed. We concluded that the finding that onset and offset delay influenced reproduction performance differentially implies that participants rather rely on the sensory feedback as a start- timing signal (at least if a causal relationship between action and sensory feedback can be established), while the motor stop is used as primary stop-timing signal. Observed sensorimotor reproduction biases and variability could be described as the weighted integration of the auditory estimate and the motor estimate. The integration reflects the brain combines multiple timing signals to improve overall performance. The prior knowledge of the standard duration in the reference memory is updated dynamically in that current sensorimotor estimates are constantly integrated with the history of duration estimates. In the end, overall implications of all the results for time perception, as well as sensory integration research are discussed. In summary, this thesis helps to improve our knowledge about sensorimotor temporal integration in a sensorimotor reproduction task on the basis of behavioral findings as well as probabilistic modeling.