Abstract
We have shown that in visual detection[1] and discrimination[2] tasks where two signals were presented with equal probabilities, observers use a unique decision criterion in accordance with a model whereby their decisions are based on a unique internal representation. Here we test an alternative account[3] of the data, according to which observers (Obs) equate the false-alarm (FA) rates for the different targets, rather than the internal response criteria. In the experiments, Obs detected a Gabor target in the presence of external noise, using a Yes/No paradigm. The noise gray-levels were sampled from a Gaussian distribution with a mean equal to the screen luminance. Noise SDs were set to values higher than the equivalent internal noise so that target detection was limited by the external noise. Stimuli consisted of two circular noise patches (D=1°), left and right of fixation (1.6°), presented before (400 ms), during (80 ms) and after (400 ms) target presentation. Targets appeared with 0.5 probability independently in each of the noise patches. A post-cue marked the location of the target to which the Obs responded. d' and criterion for target detection were estimated in two conditions: Dual Same (DS) and Dual Different (DD), referring to the SDs of the two noise patches. We hypothesize that different noise SDs (DD condition) produce different zFAs (or FA rates) for the two targets if Obs adopt an equal response criterion, but not if they equate FA rates. More specifically, we predict that zFA1/zFA2=SD2/SD1. The results support this prediction when SD2/SD1 < 4, showing no effect on d' relative to the DS condition. For stimuli with larger noise ratios, zFA ratios are smaller, and d' is reduced. We conclude that observers use a single response criterion when involved in a multi-target detection task.
[1]
GoreaSagi(2000) PNAS 97,1238012384
[2]
GoreaSagi(2001) Nature Neurosci 4,11461150
[3]
Kontsevich(2002) Nature Neurosci 5,707; Gorea & Sagi (2002)ibid.