Copyright 1999 Psychonomic Society, Inc. 1624Perception & Psychophysics1999, 61 (8), 1624-1645Often, people cannot ignore irrelevant information whenthey classify a stimulus attribute. For example, whenpeople are asked to classify auditory tones according topitch, their performance is worse (slower and with moreerrors) when the timbre of the tone varies randomly be-tween trials than if the timbre remains fixed between tri-als (Crowder, 1989). Irrelevant variation in loudness simi-larly interferes with performance when tones are classifiedaccording to pitch (Lockhead, 1992b; Melara & Mounts,1994). This result—the fact that irrelevant variation on onedimension interferes with classification on another dimension—is called Garner interference (Garner, 1974).Whenever there is such interference, the stimulus dimen-sions interact, which is contrary to what should occur ifpeople could abstract a feature from a stimulus and pro-cess it independently of other features, as Stevens (1975)proposed. There have been many attempts to understandthis apparent interference (Garner, 1970; Lockhead,1966, 1992b; Melara & Marks, 1990; Melara & O’Brien,1987; Pomerantz & Garner, 1973; Shepard, 1964), noneof which was fully successful. Each of these attempts ad-dressed characteristics of individual stimuli, and all ofthese experimental approaches compared average perfor-mance in one condition to average performance in anothercondition. A factor that cannot be addressed by such studies of av-eraged data is any effect of the sequential structure of thestimuli, because sequential information is lost when dataare averaged across trials. Sequence should be consideredfor at least two reasons: stimulus sequences are differentin tasks that have irrelevant stimulus variation than in taskswhere there is no such variation, and, as Felfoldy (1974)showed, performance depends on sequence. Even so, Gar-ner interference essentially has not been examined for se-quence effects beyond Felfoldy’s study. Some such ex-aminations are reported here.Another factor that is not examined in Garner tasks butshould be studied is how the range of variation on the ir-relevant dimension affects performance. Range affects per-formance in absolute judgment tasks when stimuli differon only one dimension (Durlach & Braida, 1969; Gravet-ter & Lockhead, 1973; Parducci, 1965; Pollack, 1953)and in a bivariate Garner task when the range of the irrel-evant dimension is increased (Lockhead, 1992b; Melara& Mounts, 1994). In each case, performance is poorerwhen range is larger. One might expect sequence effectsand the range of the irrelevant dimension to be related. Thisis because performance is poor when the irrelevant rangeis large and because successive differences betweenstimuli within a task are large, on average, when the rangeis large. This fact allows the suggestion that both rangeeffects and Garner interference are associated with trial-to-trial sequence effects. To consider this, effects of bothstimulus range and stimulus sequence are examined.This research was partially supported by a National Science Foun-dation fellowship to S.A.H. Correspondence should be addressed toS. A. Huettel, Department of Psychology, Duke University, Durham,NC 27708 (e-mail: [email protected]).—Accepted by previous editor, Myron L. BraunsteinRange effects of an irrelevantdimension on classificationSCOTT A. HUETTEL and GREGORY R. LOCKHEADDuke University, Durham, North CarolinaIn univariate classification tasks, subjects sort stimuli on the basis of the only attribute that varies.In orthogonal classification tasks, often called filtering tasks, there additionally are trial-to-trial varia-tions in irrelevant attributes that the subjects are instructed to ignore. Performance is generally slowerin filtering tasks than in univariate control tasks. We investigated this slowing in experiments of howthe range of irrelevant trial-to-trial variation affects responses in pitch/loudness classification tasks.Using two levels of pitch and of loudness as stimuli, Experiment 1 replicated prior work showing thatresponses are slowed more when the range of the irrelevant dimension is made larger. Also in Experi-ment 1, sequential analyses showed that response time depends both on sequence and on the stimulusset independent of sequence. Experiments 2 and 3 used several levels on the irrelevant dimension andshowed that responses to categorize loudness are slowed more by larger trial-to-trial pitch differences,but only on trials when the response repeats. When the response changes, performance is essentiallyunaffected by trial-to-trial irrelevant variation. This interaction supports the conclusion that slowed av-erage performance in orthogonal classification tasks, which is known as Garner interference, is notdue to difficulties that subjects have in filtering stimulus attributes. It is due to how subjects processsuccessive stimulus differences. We call for more frequent reports of sequential analyses, because thesecan reveal information that is not available from data averages.RANGE EFFECTS 1625Theoretically, one reason for examining sequence andrange effects in Garner tasks is to pursue a proposal byLockhead and King (1983) that describes processes thatseem to be involved when people classify stimuli in mag-nitude estimation and absolute judgment tasks. The pro-posal for these univariate tasks is that people judge astimulus by comparing it with their memory of the priorstimulus, and this comparison both requires less time andis more precise when successive stimuli are more similar;the comparison task is then easier. This thesis describesmuch of the univariate judgment data. The experimentsreported in this paper examined whether the thesis alsodescribes performance in Garner tasks. Range and sequence effects are appropriate empiricalmeasures to use to examine this trial-to-trial comparativejudgment thesis. This is because it states that performanceis less precise when successive stimuli are more different.Successive stimuli are more different, on average, whenthe level of the irrelevant dimension changes between tri-als than when it remains fixed, and the magnitude of thischange will be greater when the range of irrelevant vari-ation is greater. Hence, the predictions are that perfor-mance between tasks is poorer in conditions with a largestimulus range than in conditions with a small stimulusrange and that performance within tasks is poorer on tri-als when successive
View Full Document
Unlocking...