Behavioral Stochastic Resonance within the Human Brain

We provide the first evidence that stochastic resonance within the human brain can enhance behavioral responses to weak sensory inputs. We asked subjects to adjust handgrip force to a slowly changing, subthreshold gray level signal presented to their right eye. Behavioral responses were optimized by presenting randomly changing gray levels separately to the left eye. The results indicate that observed behavioral stochastic resonance was mediated by neural activity within the human brain where the information from both eyes converges.

Figure 1

(a) Experimental setup. Subjects adjusted handgrip force to the slowly changing gray level of the right eye window, while Gaussian white noise was added to the left eye window (contralateral noise) or to the right eye window (ipsilateral noise). (b) Representative recording from one subject in the contralateral noise condition (subject $B$ in Fig. 2; subthreshold signal $\mathrm{\text{amplitude}}=6$, and noise $\mathrm{S}\mathrm{D}=4$). $G$: visual stimulus signal; $F$: handgrip force; $\widehat{G}$: filtered visual stimulus signal; $\widehat{F}$: filtered handgrip force signal; ${\varphi }_G$: instantaneous phase of $\widehat{G}$; ${\varphi }_F$: instantaneous phase of $\widehat{F}$; and ${\varphi }_F\mathrm{\text{−}}{\varphi }_G$: phase differences between ${\varphi }_F$ and ${\varphi }_G$. (c) One hundred sets of surrogate signals were generated by randomly modifying the interval lengths of the original visual signal. The $t$ values for correlation/synchronization measures of the original data were calculated from the probability distribution obtained from these surrogates.

Figure 2

Effects of noise intensity on cross-correlation coefficients and variance of phase differences for three subjects ($A$, $B$, $C$). Two vertical lines in each graph indicate the normalized noise level where noise $\mathrm{S}\mathrm{D}/\mathrm{\text{signal}}\mathrm{\text{amplitude}}=0.5$ (left line) and 2.0 (right line).

Figure 3

Effects of normalized and categorized noise levels on cross-correlation coefficients (a), and variance of phase differences (b). Black and gray bars, respectively, represent group means for contralateral and ipsilateral noise at each noise level; without noise ($\mathrm{S}\mathrm{D}=0$), weak ($0<\mathrm{S}\mathrm{D}\mathrm{/}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{u}\mathrm{d}\mathrm{e}\le 0.5$), intermediate ($0.5<\mathrm{S}\mathrm{D}\mathrm{/}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{u}\mathrm{d}\mathrm{e}\le 2.0$), and intense ($\mathrm{S}\mathrm{D}/\mathrm{\text{amplitude}}>2.0$) noise. Error bars correspond to the standard error of the mean $*$; significant difference between “without noise” and “intermediate noise level” ($p<0.05$).