On learning several simultaneous probability-learning problems.

Abstract

Ss performed 4 simultaneous probability-learning tasks under either speed or accuracy instructions. When the 4 IT values were the same, Ss tended to overshoot matching under accuracy instructions, but under speed instructions overshot only at higher TT values. When the 4 TT values were not the same, there was a tendency to overshoot for lower TT values, but not higher ones. Ss consistently responded faster at higher IT values and the increase in speed was linear as w increased from .50 to 1.00. Ss' subjective estimates of TT suggested that they rather consistently overshot their ir estimates by 5-10 percentage points (except when IT =1.00). These results were discussed with respect to stimulus-sampling theories of probability learning.

Alternative bases for choice in probabilistic discrimination.

“…Departure toward maximizing across 1,000 trials was reported by Edwards (1961) who considered this departure to be the general tendency in probability learning. Erickson (1966) investigated the effects of -n values and of instructions in leading 6"s to go beyond w.…”

The utility of event variability in two choice probability learning

“…If the RI condition could be considered as providing greater amounts of variability than the NI condition (cf. Erickson, 1966). then a theory employing the concept of utility for variability would predict larger marginal response probabilities for each of the RI groups than for the corresponding NI groups.…”

“…The theory thus provides an ordered hypothesis of increases in asymptotic response probability as a function of choice variability. The overshooting commonly found in situation "a" is clearly predicted from this utility theory, and if one assumes that sequential independence of choices provides response variability (cf., Erickson, 1966), then the theory can account for the overshooting obtained with situations "b" and "c. "…”

Subjective probability revision in twelve simultaneous sequences

“…Multiple-cue probability experiments afford an even closer parallel to the present study since both call for probabilistic decisions in several concurrent sequences. Erickson (1966) found that choice proportions in a four-game situation showed more maximizing than is typically observed in a single probability learning game. Similarly, Peterson, Hammond, & Summers (1965) found nearly optimal responding as measured by a linear multiple-regression equation in a three-game probability learning situation.…”