Multidimensional Fechnerian scaling: basics (Q5953437)

This beautiful paper derives an 'intrinsic' metric on stimulus space, solely from assumptions on perception -- in particular, on discrimination probabilities. The building block is thus the \(x\)-indexed family of functions \(F(y;x)=\text{Prob}[y\) is discriminated from \(x]\), for stimuli \(x,y\). The main assumption is, roughly, that the differences \(F(x+su;x)-F(x;x)\) for real s and stimuli \(x,u\) are all of the same order of magnitude as \(s\) tends to zero. From this, an internal (Finsler) metric on the stimulus space is constructed, and the connection between its geometry and empirical measurements is explained.