Purchase this article with an account.
David Hoppe, Stefan Helfmann, Constantin Rothkopf; Learning when to blink: Environmental statistics guide blinking behavior.. Journal of Vision 2017;17(10):1154. doi: https://doi.org/10.1167/17.10.1154.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Our eyes blink 15 – 17 times per minute and during this time the stream of visual information is interrupted for 100-400 ms leading to perceptual gaps every two to three seconds. There are numerous situations in which these gaps can lead to negative outcomes including motor control, fight and flight scenarios, and social interactions. Hence, choosing carefully when to blink should be advantageous compared to blinking at random. Various studies indicate a connection between the current behavioral situation and blinking. Blinking rates have been found to decrease during reading but to increase during conversations and when fatigued. Further, blinking behavior is influenced by task difficulty and whether the visual input is meaningful to a person. While there exists a lot of empirical work pointing to the connection between blinking and the visual environment, the environmental regularities are usually complex and unknown. We present a controlled blinking experiment with parametrically generated environmental statistics. In our study, subjects directed their gaze to a grey dot moving on a circular trajectory (100 laps per block) in order to detect events (50 ms in duration). Hence, a normal blink could lead to missing an event. By probabilistically drawing events from spatial probability distributions we could investigate the relationship between event-statistics and blinks. Our results show a clear connection between blinking rates and environmental statistics. Subjects were able to learn regularities in the event generating process and as a consequence adapted their behavior. In addition to the behavioral results we investigated the blinking process by developing a computational model. We show that uncertainty about the statistics as well as costs for blink suppression are sufficient to reproduce key characteristics of the blinking behavior. Remarkably, our computational model predicts various aspects of the visual behavior such as the distribution of time intervals between blinking.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only