Given a permutation $σ$, the Robinson-Schensted correspondence determines a certain partition called the shape of $σ$. Famously, the shape measures the longest unions of increasing and decreasing subsequences, thus giving global information about $σ$. In this paper, by contrast, we ask how prescribing a shape collectively controls local behavior: namely, if $σ$ is a random permutation of shape $λ$, then what is $P^λ_{ij} :=$ the probability that $σ(i) = j$? Using tableau-theoretic methods, we derive explicit formulas for $P^λ_{ij}$ when $λ$ is a hook, two-row, or rectangular shape. We use these formulas to depict and analyze the intricate diffraction-like patterns in the matrices $(P^λ_{ij})$. As a surprising application, we show that for both hook and two-row shapes, as the largest part of $λ$ tends to infinity with the remaining parts fixed (summing to $m$), the expected proportion of fixed points in $σ$ approaches the Wallis integral $\int_0^{π/2} \sin^{2m+1} x \: dx = (2m)!! / (2m+1)!!$.