Adam Reeves¹; ¹Northeastern University

We measured the ‘Perky effect’, the change in vision consequent on having a visual image (Perky, 1910), in a test of visual acuity. Subjects reported whether an upper line was left or right of a lower line (2AFC), or reported if one line was left, center, or right (3AFC). Accuracy was converted to d’. Maintaining an image of vertical lines around the vertical line stimuli increased d’ at low levels (d’<0.6) but lowered d’ at high levels (d’>1), relative to d’ in no imagery. The change from ‘reverse Perky’ (helping) to the regular Perky effect (hindering) occurred in two separate experiments with 54 and 135 subjects. It was modeled by assuming imagery both (a) boosts the signal by 2 to 4 times (helping) and (b) adds increasing amounts of signal-dependent multiplicative noise, Nx, to the intrinsic noise, No (hindering). In imagery, Nx eventually dominates as signal strength increases, reducing sensitivity by up to 0.9 d' units relative to no imagery. Adding external noise reduces this effect to ~0.3 d' units (Dijksra et al, 2022), we think by partially swamping the multiplicative noise. The basic constraint that d’ must increase with signal level can be met if No is Gaussian and Nx is exponential, the total noise being exGauss. The model was fit by least-squares with 2 parameters and this constraint in place; RMS errors were low, 0.08 < 0.17. ExGauss noise implies that the ratio SD(Imagery)/SD(No imagery) increases predictably with signal level, as indeed we found (r>0.93).

Acknowledgements: Thanks to Severine Merand and Amedeo D'Angiulli