Unidirectional influence of vision on locomotion in multimodal spatial representations acquired from navigation

Abstract

Visual and idiothetic information is coupled in forming multimodal spatial representations during navigation (Tcheang et al. in Proc Natl Acad Sci USA 108(3):1152–1157, 2011). We investigated whether idiothetic representations activate visual representations but not vice versa (unidirectional coupling) or whether these two representations activate each other (bidirectional coupling). In a virtual reality environment, participants actively rotated in place to face certain orientations to become adapted to a new vision–locomotion relationship (gain). In particular, the visual turning angle was equal to 0.7 times the physical turning angle. After adaptation, participants walked a path with a turn in darkness (idiothetic input only) or watched a video of the traversed path (visual input only). Then, the participants pointed to the origin of the path. The participants who were presented with only idiothetic input showed that their pointing responses were influenced by the new gain (adaptation effect). By contrast, the participants who were presented with only visual input did not show any adaptation effect. These results suggest that idiothetic input contributed to spatial representations indirectly via the coupling, which resulted in the adaptation effect, whereas vision alone contributed to spatial representations directly, which did not result in the adaptation effect. Hence, the coupling between vision and locomotion is unidirectional.

Appendix

In Experiment 2, to provide stronger evidence that the observed difference ratio in the transportation group is due to no adaptation effect, we also calculated Bayes’ factors measuring the ratio of the likelihood of an adaptation effect to the likelihood of no adaptation effect for the transportation group (see Gallistel, 2009).

In particular, we used the difference ratio observed in the walking group (i.e., 0.14 or 14%) as the real adaptation effect in the transportation group under the alternative hypothesis claiming an adaptation effect, assuming that the transportation group would show the same amount of adaptation effect as the walking group. The likelihood of any possible observed difference ratio under the null hypothesis (i.e., no adaption effect) can be measured by the probability density of the t value of the observed difference ratio ($t=\frac{{ - 0.02 \times \sqrt N }}{{{\text{SD}}}}$, where N = 25 and SD = 0.36 from the transportation group) in a t distribution (df = 24). The likelihood of any possible observed difference ratio under the alternative hypothesis can be measured by the probability density of the t value of the observed difference ratio in a noncentral t distribution (noncentral parameter λ is the t value of the theoretical adaptation effect, i.e., $\lambda =\frac{{0.14 \times \sqrt N }}{{{\text{SD}}}}$, where N = 25 and SD = 0.36 from the transportation group, df = 24). The probability density as a function of the observed difference ratio under the competing hypotheses is plotted in Fig. 9. Results showed that the Bayes’ factor (i.e., likelihood ratio) in favor of the null hypothesis was 10.61, providing strong support for null adaptation effect. The null effect is favored if Bayes’ factor is larger than 3, and strongly favored if the Bayes’ factor is larger than 10, whereas an adaptation effect is favored if the Bayes’ factor is smaller than 1/3, and strongly favored if the Bayes’ factor is smaller than 1/10 (Rouder, Speckman, Sun, Morey, & Iverson, 2009). If the Bayes’ factor is between 1/3 and 3, neither is favored.

Similarly, we calculated Bayes’ factors in favor of no adaptation effect for each turning angle (BF was 0.35, 4.32, and 12.95 for turning angles of 63°, 90°, and 117°, respectively). The null effect was favored for the turning angle of 90° and strongly favored for the turning angle of 117°, although the Bayes factor could not distinguish between the null effect and an adaptation effect for the turning angle of 60°.

Note that when calculating the Bayes’ factors, we used the observed gain effect (i.e., gain difference ratio) in the walking group as the real gain effect and used the observed variances of the gain effect in the transportation group as the variance. Therefore, Bayes’ factors that favored the null effect in the transportation group were calculated as have already considered the influences from the two facts that the observed gain effect was much smaller than the theoretical one and that the pointing response in the transportation group was noisier.