Figure 6
Response distribution is shaped by dependence between target angles in adjacent trials P(θn+1|θ n ). (A) Visual working memory protocols typically use a sequence of targets with no trial-to-trial dependence (uniform P(θn+1, θ n ), shown for θ n  ≡ 0°)6,9. Relative responses (r n  − θ n ) are normally distributed about the true target. (B) Current target θn+1 depends on the previous target θ n according to a locally peaked distribution. The response distribution narrows (note decreased standard deviation σ), since the target θn+1 is often close to the previous target θ n . (C) Current target θn+1 is skewed clockwise from previous angle θ n . Responses are thus skewed counter-clockwise towards the previous target (note average response \(\bar{r}\) is shifted). Numerical methods are described in Methods.
