Abstract
Human ability to detect an intensive change decreases with background intensity (Weber's law). Insofar as incremental thresholds reflect a constant S/N ratio, there are two possible extreme accounts for this behavior. The first holds that discrimination is limited by a constant internal noise, sigma (independent of the stimulus intensity, C), and that it decreases with C because of the compressive nonlinearity of the transducer. The second account assumes that, above threshold, sigma increases with the internal response R (and thus with C), while R is linear with C. We present here a psychophysical method to decide between these two views. The method takes advantage of a newly discovered constraint on human decision-making: in a detection task where multiple signals are presented with equal probabilities, observers use a unique decision criterion1. Four observers had to report the presence/absence of a 80 ms contrast increment (chosen to yield a d' close to 1.5) presented on either one of two suprathreshold 3 cpd Gabor pedestals displayed simultaneously within two cue-circles 1.6° to the left and right of fixation. Pedestals had equal or different contrasts ranging from 10 to 60%. One of the two circles persisted after the disappearance of all other items to indicate the event to be reported (partial report). In one case, increments yielding d'-s of about 1.5 and 2.5 presented on identical pedestals where used to confirm the unique criterion constraint (ucc) in all observers. The ucc requires that, when expressed as the z-score of False Alarms (zFA) scaled by the appropriate sigma-units, any two criteria c1, c2 assessed for two jointly presented pedestals, be equal: c1 = c2, with c1 = −zFA1(sigma1 and c2 = −zFA2(sigma2. Because zFAi is experimentally measurable, the noise ratio sigma2/sigma1 can also be estimated. The present results support the first account above: for contrasts up to 60%, sigma is practically constant, while the transducer displays the expected saturation.
1
GoreaSagi(2000) Proc. Natl. Academy USA, 97, 12380–12384