The early versus late selection issue in attention models was examined by means of a new methodology. Through cues or precues, attention was directed to one location of a multistimulus visual display and, while attention was so engaged, the identity of a stimulus located at a different position in the display was changed. By varying the time after display onset before the stimulus was changed, we controlled the preview time that the original stimulus was represented on the retina. Then, using a marker cue, we directed the subject's attention to the location of the changed stimulus. The subject's response was a timed discrimination between two possible target letters. The data of main interest was the effect of preview time upon the subject's latency in identifying the new target that appeared in the changed location. We found that the preview time of the original stimulus, before RT was affected to the new target, depended upon whether the original stimulus was a neutral (noise) letter or whether it was the alternative target. When the original stimulus was a noise letter, RTs to the new target were just as fast as those obtained in the control condition in which the target was present throughout the preview interval and did not change its identity. Significant effects upon RT were obtained at preview times of 83 msec when the original stimulus was one of the targets that changed to the alternative target. Preview times also varied as a function of precuing. Preview times were correspondingly shortened when the first cue occurred 50 msec before display onset, thus providing an extra 50 msec for attention to be directed to the first display location. The results were interpreted in terms of two separate information-processing systems in the human: an automatic system and an attentional system. Even though a stimulus may have been automatically processed, when the attention system is directed to that stimulus, processing starts at the beginning again.Pashler (1984) has presented a series of experiments that call into serious question a basic assumption of lateselection theories of attention, namely, that stimuli are parallel-processed and identified before attentional processing is engaged. In these experiments, Pashler used a modification of the iconic memory paradigm of Averbach and Coriell (1961). In the paradigm, an overload of visual information is presented for a brief duration and then, at various intervals after the display has been terminated, a bar or cue designates a location in the visual field. The subject is asked to report the stimulus that had occupied that location in the display. The now classic result is that accuracy of report decreases rapidly as the cue is delayed with performance, becoming asymptotic at around 300 msec.In his first experiment, Pashler modified the Averbach and Coriell procedure in several ways. He used displays of eight letters presented in two rows of four characters each. Three probe conditions were used. In the early-probe condition, the probe preceded the display by 200...

If several positions must be attended in a large visual display, does the efficiency of performance vary as a function of the display distance between these to-be-attended positions? Two previous experiments (Podgorny & Shepard, 1983;Shaw, 1978) gave conflicting answers. In the present experiments, eight-letter circular displays were briefly presented. On each trial one, two, or three positions of the display were cued or preeued. The number of noncued display positions intervening between the cued locations varied from zero to three. The subjects' task was to rapidly discriminate between two target letters. Although reaction time was found to increase with increases in the number of cued locations, no significant or suggestive effects were found for the spacing or distance between the cued locations. The evidence strongly suggests that the subjeets serially searched the cued locations, which further implies that attention can index locations in the visual field at a speed that is independent of the distance between these locations.If a subset of stimuli must be attended in a complex large visual display, does it matter which way this subset is distributed throughout the display and intermixed with distractor stimuli? Our purpose in the present experiments was to obtain some evidence on this question and to relate the findings to several current conceptions of visual attention. To preclude the role of saccadic eye movements, we dealt only with displays presented for durations too brief for changes in eye fixation.Although there is a vast literature on visual search, only a few studies have addressed the question we have posed above. Two of these studies came up with conflicting answers (Podgomy & Shepard, 1983;Shaw, 1978;Shaw & Shaw, 1977). The visual display used in Podgomy and Shepard's (1983) experiment was a 3x3 matrix of squares. Prior to a trial, a subject was instructed to distribute his/her attention over a subset of these squares. The squares to be attended were lightly shaded. When the subject felt that his/her attention was deployed over the lightly shaded area, he/she initiated the trial. At this point a dot appeared in one of the nine squares. The subject made a discriminated response as to whether the dot occurred on a shaded or a nonshaded square, and the latency of the response was recorded. Latencies were shortest when the target fell on attended squares. Furthermore, the latencies were reduced when the subregion of shaded squares (attended) constituted a compact region on the nine-square display. Latencies increased as attended squares were distributed over the display with nonattended squares intervening between attended ones. Crassini (1986), in a reanalysis of Podgomy and Shepard's (1983) data, concluded that a benefit in responding occurred only when the attended squares were unitary, that is, when the attended squares were not separated or interlarded with squares that were not to be attended (see Shepard & Podgomy, 1986).The results of these experiments suggest that attention can be dis...

Qualitativeuser-centered design processes such as contextual inquiry can generate huge amounts of data to be organized, analyzed, and represented. When you add the goal of spreading the resultant understanding to the far reaches of a large, multi-site organization, many practical barriers emerge.In this paper we describe experience creating and communicating representations of contextually derived user data in a large, multi-site product development organization. We describe how we involved a distributed team in data collection and analysis and how we made the data representations portable. We then describe how we have engaged over 200 people from five sites in thinking through the user data and its implications on product design.

We investigated the relative efficacy of a variety of spatial maps and analogies as instructional aids for the learning and navigation of hierarchical data bases. Subjects performed single- and multi-node retrieval tasks with novel data bases. The results of Experiment 1 indicated that subjects who studied an analogy prior to performing data retrieval tasks were less affected by increasing search distance than were subjects who studied a spatial map. The results of Experiment 2 suggested that the benefits of analogy relative to other instructional aids increased as time passed between initial instruction and interaction with the data base. A comparison of the findings in Experiments 1 and 2 indicated that “generic” analogies, which can be applied to a number of data bases, are as effective as more specific, literal similarities. The findings are discussed in terms of models of analogy as well as implications for the design of instructional materials for data base navigation.

This poster provides a definition and taxonomy for iconic communication and describes the use of formal psychological tools and methods in the evaluation of icons. The methods that can be usefully applied include:1. Psychophysics2. Scaling3. Recognition/Memory Testing4. Statistical Modeling/AnalysisExamples of some of these approaches are provided from pilot studies currently under way at HP. Analyses used include Multi-Dimensional Scaling (MDS) and Cluster analysis. Results can be applied to development of metrics, standard methods, and design guidelines.