Range and sequence effects in judgment

Abstract

Listeners classified three tones that differed in loudness. Two tones were always similar in intensity (2 dB separation). The third tone was either similar to or different from these two tones. Performance depended on this stimulus range: The greater the difference between two tones fixed in intensity and the third tone, the less precise was the discrimination between the two fixed tones. Performance also depended on sequence: Successive responses were positively correlated. The results show that measures of discriminability depend on stimulus range, and that measures of criterion placements change from trial to trial and depend on stimulus sequence.When humans classify stimuli, their performance on any particular stimulus becomes more variable when a greater number of different stimuli are added to the set. For example, a tone is more precisely identified when it is a member of a set of 6 tones, each separated by 2-dB intervals, than when it is a member of a set of 10 tones, each separated by 2-dB intervals.Such results are consistent with the concept of channel capacity, the idea that there is a performance limit in human discrimination that is analogous to the proven limit on the amount of information that can be transmitted electronically along a wire (Shannon & Weaver, 1949). Many studies show that the amount of information transmitted in judgment tasks increases linearly as the number of stimuli in the set increases, up to some limit, after which the amount of information transmitted is constant. This asymptotic level is considered the channel capacity of the system (Miller, 1956).It is known that performance on individual stimuli in these tasks depends on the range over which stimuli vary and also on the particular stimulus or response that occurred on the most recent trial. In this study we examined the ways in which these two factors, range and sequence, were important to the task of deciding which stimulus had just been presented. Range EffectsIn 1952, Irwin Pollack reported a study in which listeners identified eight tones that were equally spaced on a log frequency scale. In one condition the tones ranged over 400 Hz; in another condition they ranged over 8000 Hz. Unlike many studies in which range was increased by adding more stimuli to the set, here the numThis research was supported in part by AFOSR Grant 85-032. Address correspondence to G. R. Lockhead, Department of Psychology, Duke University, Durham, NC 27706. 53 ber of stimuli was held constant as stimulus range varied. The result was that the listeners' ability to differentiate between the eight tones was not measurably different between the 4OO-Hz and the 8000-Hz conditions. This means that the variability of judgments of any particular stimulus was greater when the stimuli varied over a larger range. Pollack's (1952) result is consistent with the conclusion of a channel capacity for a stimulus continuum. Such a limit on performance would occur if people attempted to locate each stimulus along a memory scale and if the ability to do this wa...

An integrated model of choices and response times in absolute identification.

“…To illustrate how the mapping operates and also how it naturally handles unevenly spaced stimuli, consider an experiment from Lockhead and Hinson (1986). In one part of this experiment, the stimuli were tones of intensity 58 dB, 60 dB, and 66 dB.…”

“…7 For the Lockhead and Hinson (1986) example, suppose the 60-dB stimulus was presented. On this particular trial, the selective attention stage might produce a magnitude estimate of .3, which is quite close to the long-term average value of 2/7 for this stimulus.…”

“…We think this distinction is not as useful nor as clear cut as others believe. Even the most relative models (such as Stewart et al's, 2005, RJM) employ absolute knowledge about the global nature of the experiment: For example, when the RJM is used to model unequally spaced stimuli, such as in Lockhead and Hinson (1986), the (global, absolute) scaled magnitudes of the stimulus differences must be taken into account in setting the spacing of cutpoints, ensuring that the global stimulus setup is captured in the model.…”

The relative importance of perceptual and memory sampling processes in determining the time course of absolute identification.

“…It is well established that increasing range, while keeping set size fixed, lowers response accuracy (Lockhead and Hinson, 1986). Thus, it may be possible that the set size effects observed were due to range, not set size per se.…”

“…Here, allowing c to vary by set size, captured the set size effect (in accuracy) best, with c decreasing with larger set sizes (indicating lower discriminability). As increasing set size also increased stimulus range and increasing stimulus range is known to decrease accuracy (Lockhead & Hinson, 1986), the decrease in c with increasing set size may also allow the model to capture part of the range effect, although note that the model is not able to capture the range effect Table 2 for fit and parameter values). This model had eight free parameters: tres (residual time); cn6 , cn8, cw6, and cw8 (discrimination constants for the four spacing and set size conditions produced by the 2 x 2 manipulation); qp (the processing rate for perceptual information elements); λ (the base of the diminishing return function); and θ (response determinism).…”