Irwin Pollack

A non-parametric analysis of recognition experiments

A non-parametric method for evaluating the results of recognition memory experiments and psychophysical detection experiments is presented. The method is based upon an ordinal analysis of recognition performance, which transforms the results of recognition tests into equivalent results for a forced-choice experiment.

An efficient non-parametric analysis of recognition memory

A method for efficient evaluation of the results of recognition memory experiments is illustrated. The method is based upon the area under the iso-mnemonic curve relating correct and incorrect acceptances obtained by means of ratings.

On the comparison between identification and discrimination tests in speech perception

The conversion between identification and three forms of discrimination tests, based on the extreme assumption that discrimination performance in speech perception tests is determined solely by labeling in identification tests, is examined. Identical conversions are obtained for the two-interval same-different test, for the four-interval forced-choice test of pair similarity, and for the three-interval ABX test. Therefore, differences in discrimination performance among these tests are presumably due to their respective task requirements, not to differential prediction.

A nonparametric procedure for evaluation of true and false positives

A procedure is presented for the evaluation of single-point estimates of true and false positives without strong underlying parametric assumptions. The method is based upon the area operating characteristic and the Green area rule. Estimates of sampling error are also available. The procedure is extended to a strong one-parameter relation between true and false positives.

Iterative techniques for unbiased rating scales.

Three iterative techniques for neutralizing the effects of stimulus bias in category rating experiments were examined with a wide variety of stimulus variables. Under all conditions examined, the iterative techniques quickly led to a stable category estimation. This result was obtained for stimulus variables with strong measurement properties, e.g. length and weight; for stimulus variables with only ordinal properties, e.g. emery papers; and for stimulus variables with only nominal properties, where an ordered set is obtained only in the course of the category scaling, e.g. female profiles.

Detection of changes in spatial position: Short-term visual memory or motion perception?

The limiting conditions for short-term visual memory were explored for materials which are not readily codable. Two fields of random dots were presented successively in time. On half of the observations, the second display was identical to the first, and on the other half, a fraction of the dots was displaced relative to the first display. The task of the S was to indicate if any of the dots had moved (displacement detection), and in separate tests, if any of the dots had not moved (stationarity detection). Stationarity and displacement detection are largely related to the same variables, although performance is somewhat better with displacement detection. Discrimination of small displacements is critically dependent upon the interval between displays, with maximum accuracy in the region of 64 msec for briefly presented displays. Maximum accuracy is obtained under conditions which yield good apparent motion. By contrast, displacement discrimination is, within limits, relatively independent of the number of dots displayed, of the fraction of displaced dots.

Visual discrimination thresholds for one- and two-dimensional Markov spatial constraints

Forced-choice discrimination thresholds were obtained for visual displays of a large number (1,000-3,600) of binary encoded elements with one- and two-dimensional Markov spatial constraints. Several factors which delimit such thresholds were explored. The primary determinant of sensitivity was the number of constrained elements within the display. Configurational features, such as the shape of the display have little influence on sensitivity. Thresholds for one-dimensional spatial constraints were consistently lower than for two-dimensional spatial constraints.

Interaction effects in successive visual displays: An extension of the Eriksen-Collins paradigm

The paradigm of Eriksen and Collins, in which one-haft of the elements of a display is assigned randomly to each of two successively presented fields, was applied to the visual discrimination of one- and two-dimensional Markov constraints. In contrast with the two-flash masking paradigm, performance in the Eriksen-Collins paradigmimproves as the interval between the successively presented fields is shortened. Constraint thresholds may increase with interfield intervals as short as 20 msec. In contrast with the results of most visual information-processing tasks, increasing the duration of each field may result in sharplyelevated thresholds. The elevation of thresholds can be reduced through repeated presentation. Sharp asymmetries are also obtained with differential brightness of the two fields.

ON THE PERFORMANCE OF A COMBINATION OF DETECTORS.

The performance of a combination of two hypothetical detectors is first examined. It is shown that the performance of the combination may be as poor as the poorer detector alone when the inflation in false alarm rate of the combination of detectors is considered. Experimental tests were carried out in which listeners attempted to detect the presence or absence of a tonal signal in noise under two conditions: unaided by an external source of information (actually aided by a simulated machine detector set at chance performance); and aided by a simulated machine detector set at the same discriminability as the listener. While the performance of the aided listener was consistently superior to the unaided listener, the improvement in performance fell considerably short of an ideal combination. It is concluded that the gain in performance expected from combinations of detectors must be interpreted cautiously.

Discrimination of third-order Markov constraints within visual displays*

Discrimination thresholds were measured for third-order Markov constraints within visual displays. The method permits the cancellation of adjacent second-order differences and of first-order differences. Excellent discrimination of third-order constraints is obtained despite the fact that the average conditional repetition probability and the average run length were invariant. The need to distinguish between local short-term and average long-term analyses for the visual system is briefly discussed.