
Multidimensional stimulus identification I HOWARD EGETH2 AND ROBERT PACHELLA THE JOHNS HOPKINS UNIVERSITY In an absolute judgment task, total information transmitted obtained by Klemmer and Frick (1953) and Osborne, about a set of stimuli increases as the dimensionality of the Quastler, and Tweedell (1955) on channel capacity for stimuli is increased. However, as the number ofdimensions is judgments of position within a two-dimensional area. increased, information transmitted about each component Information transmission in the unidimensional study was 3.2 dimension decreases. Four factors were proposed that might, bits and in the bidimensional studies about 4.4 bits-or only singly or in combination, account for this decrement: stimulus about 2.2 bits per component dimension. However, in the duration, interdimensional interference, distraction, and unidimensional study no limit was put on stimulus duration response complexity. The results of four experiments whereas in the bidimensional studies the stimuli were established that each of these factors may be operative under presented tachistoscopically. It seems likely that more time appropriate circumstances. It was also concluded that would be needed to extract information about several component dimensions were processed simultaneously rather dimensions than about a single dimension. Therefore, in than sequentially. However, the processing was not truly comparing information transmission of unidimensional and parallel because the input channels were not independent of multidimensional judgments one must be careful to take one another. stimulus duration into account. In any event, it does not seem appropriate to compare long exposures of unidimensional The ability of humans to make accurate categorical stimuli and short exposures of multidimensional ones and then judgments about stimuli that vary along a single continuum observe a "decrement" in the multidimensional conditions. such as wavelength or sound pressure level has been shown to (2) lnterdimensional interference. This factor is best be quite limited. Using the method of absolute judgment, in described with a hypothetical example. Imagine a set of which S must identify each stimulus as it is presented, stimuli defined by the joint values of two dimensions, X and investigators have shown that the channel capacities for a Y. Typically, information transmission for X is determined by variety of dimensions fall in the range from 2 to 3.2 bits, requiring judgments on X as X is varied at some fixed value of values which correspond to perfect discrimination among four Y. Similarly, Y is judged with a set of stimuli in which X is to nine categories. [See Garner (1962) and Miller (1956) for held constant. However, when bidimensional judgments are reviews of this literature.] 3 required 'both X and Yare varied and both must be judged on This finding seems to be strikingly at variance with the every trial. In this case, X judgments are made at many observation that in their everyday experience humans can different values of Y, and Y judgments are made at many easily and accurately identify thousands of words, faces, and different values of X. The interference argument is simply objects. Miller (1956) pointed to the probable explanation of that: (l) X values may not be equally discriminable at all this discrepancy when he noted that stimuli such as words, values of Y, and vice versa, and (2) unidimensional judgments faces, and objects differ from one another in many ways are usually obtained at that value of the other dimension that whereas the studies mentioned above were concerned with leads to optimal performance. Thus an optimal unidimensional stimuli that differed from one another in only one respect. In condition (say for X) is compared with a bidimensional support of this argument Miller drew together the data of condition in which Y varies, sometimes being at a level that several different experiments and compared results obtained makes X maximally discriminable, but at other times being at with unidimensional and multidimensional stimuli. These a level that leads to some degree of interference. comparisons clearly indicated that as the dimensionality of a (3) Distraction. In the hypothetical example above, we may stimulus set was increased, total information transmission also conceive of an S's report in the bidimensional condition as the increased, just as expected. However, these data also made it result of the following kind of process: When the S is abundantly clear that the increase in overall capacity to render examining X he must temporarily ignore or filter the Y-value, absolute judgments was accompanied by a concomitant and conversely, when he is examining Y he must ignore the decrease in the ability to identify values on each of the X-value of the stimulus. With respect to the identification of component dimensions of the complex displays. In other X, Y is an irrelevant variable, and similarly Y is irrelevant to words, there was a trade-off between what might be called the identification of X. Not only do X and Y serve as depth and breadth in information processing; it was possible to irrelevant dimensions for one another, but they vary in value render fairly accurate judgments about a single dimension, but from trial-to-trial, and previous research has shown that Ss only relatively crude judgments could be made about that may have great difficulty in ignoring such variation along dimension when several other dimensions also had to be irrelevant dimensions and selectively attending to the judged. It is this still unexplained trade-off that forms the appropriate information (Egeth, 1967). focus of the present series of experiments, which have been It is interesting to speculate about the possibility of an designed to explore some factors that may underlie the interaction between the factors of distraction and duration. It decrement in accuracy of identification that occurs when a may be possible for Ss to filter irrelevant information, but stimulus dimension is judged as part of a multidimensional only at a cost in time. Thus distraction may be important at display rather than alone. short durations but not long ones. An examination of the research on absolute judgment (4) Response complexity. In most studies requiring revealed four factors that might, singly or in combination, multidimensional judgments, Ss are requested to specify the account for the observed decrement. A description of these value of each relevant dimension on every trial. Almost four factors follows. inevitably this means that they must respond serially, one (I) Stimulus duration. Garner (1962) and Miller (1956) dimension at a time, and thus order-of-report effects must be compared data obtained by Hake and Garner (1951) on expected to assert themselves. Lawrence and LaBerge (1956) channel capacity for judgments of position on a line with data established that when three attributes of a tachistoscopically Perception & Psychophysics, 1969, Vol. 5 (6) Copyright 1969, Psychonomic Journals, Inc., Austin, Texas 341 presented display had to be judged in a specified order, the represented the factorial combination of five amounts of NaCI first was identified significantly more accurately than the (0, .3, 1.0, 4.8, and 34.7 g) and two amounts of sucrose (0 second or third. The relevance of this finding to the present and 115 g) dissolved completely in 100 ml of tap water. investigation is clear. In a multidimensional absolute judgment Procedure. The two experimental conditions were defined task, average information transmission per dimension reflects a by the two levels of sucrose concentration. Each S was run in combination of relatively accurate and inaccurate perfor­ four sessions on separate days. On the first three days they mances (i.e., 'first dimension vs later dimensions), and thus served in both experimental conditions and on the fourth day ought to be lower than transmission in a unidimensional task just the 115 g condition. Each condition was presented as a where only a "first" dimension is ever reported. single block of 50 trials and the order of presentation of There are several reasonable explanations for the finding conditions was counterbalanced over days. Within a block of that accuracy is greater. for the first dimension reported than trials each salt concentration was presented equally often in a for later ones. For example, it seems likely that, in the absence random order. of instructions to the contrary, Ss will tend to encode the The solutions were presented to the blindfolded Ss on values of stimulus dimensions in the same order as the order in plastic spoons. There was no time pressure and S was allowed which the dimensions must be reported. Harris and Haber to taste the solution in any manner he chose. After rendering a (1963) have shown that with tachistoscopically presented judgment as to which of the five saline concentrations had stimuli, order-of-encoding was a powerful determinant of been presented, the S rinsed his mouth with distilled water. accuracy of report. Since visual afterimages fade rapidly, Feedback was provided after every trial, and before beginning adequacy of recall for a particular dimension may depend a test series of 50 trials 10 practice trials with feedback were upon the state of the image at the time when information given to minimize transfer effects. concerning the dimension is encoded. Thus, it would seem Subjects. The two authors served alternately as S and E reasonable to expect that
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages6 Page
-
File Size-