It has been hypothesized that under time pressure decision makers increase the weight given to the most important information cues at the expense of other available relevant information. Three studies were performed to test this hypothesis. Subjects saw information displayed in graphical and tabular formats. There was no evidence for greater use of the most informative cues, but there was a tendency for more even weighting of cues in the graphical display format high time pressure condition. Subjects appeared to make decisions in this condition using a quick, global information gathering and integrating strategy.

Perceptual organization is determined partly by the interaction among dimension and features of stimuli entering the perceptual system. Automatization of two dimensions (i.e., the ability to attend to both simultaneously without decrement) requires prolonged practice at conjunctive search. When conjunctively defined stimuli are processed automatically, unitization (perceptual recoding of the underlying dimensions into distinct units) is said to occur. If two separable dimensions were trained to automaticity in a search task, the stimuli composed of these two dimension should be perceptual redefined as nominal dimensions (stimuli without subject-defined dimensions). This hypothesis was tested with dimensions of color and form. Subjects automatically processed the stimuli, but no perceptual redefinition was obtained as measured by performance in a subsequent speeded classification task. Because the speeded classification task is sensitive to dimensional structure, this experiment provided no evidence that unitization occurs.

The current research was designed to study knowledge development while minimizing methodological problems common to research in that area. First, a procedural network methodology compatible with one commonly used to study declarative knowledge was developed in Experiment 1. In Experiment 2, multiple longitudinal knowledge measures revealed dissociations among the developmental courses of different forms of knowledge. The findings contribute to a growing pool of evidence of the dissociation of declarative and procedural knowledge development. Further, they contradict the majority of knowledge development theories which hypothesize that procedural knowledge development is dependent on the development of declarative knowledge. These findings also indicate that much work needs to be done to determine the relationships among multiple forms of knowledge and to determine how they could be optimally trained. In order for research in these areas to progress, efforts to minimize methodological confounding problems must continue.

Eighteen right-handed subjects (Ss) between 19 and 33 years of age participated in this EEG study of hemispheric specialization. Ss performed a battery of tasks in three phases which required a range of cognitive activity. The Phase I task involved passive reception of multiple presentations of four auditory stimuli (a pure tone, the vowel /a/, the stop consonant /b/, and a frequency and amplitude modulated tone sweep). Signal averaging was used to derive average evoked responses (AERs) to 28 presentations of each Phase I stimulus at four electrode locations (F(,3), F(,4); P(,3), P(,4); reference: linked ears). In Phase II, Ss performed two auditory attention tasks. In one task the stimuli were six different tones; in the other, they were the six stop consonant speech sounds. In both tasks Ss were instructed to attend to the stimuli and compare the sound following the appearance of a signal light to the one which preceded it and indicate by pressing a hand-held microswitch whether the sounds were the same or different. Five comparison trials were presented during each task; no EEG analysis was performed during the comparison trials. AERs to 34 stimulus presentations in each task were derived at the four electrode sites. In addition, frequency aspects of the EEG during Phase II tasks were examined by computing power spectra. All stimuli in Phases I and II were generated by computer, were randomly ordered, 300 milliseconds in duration, with an interstimulus interval randomized between 2, 3, and 4 seconds. Phase III was the most cognitively demanding Phase wherein Ss performed verbal, baseline, and spatial tasks. The verbal and spatial tasks were performed twice with an alternate form of each task used for the repetition. The baseline task was performed prior to each cognitive task. Items for the verbal task were nouns of low imagery and concreteness value while the spatial task involved mental rotation of geometric shapes. The baseline task elicited sensory and motor activity similar to that required by the verbal and spatial tasks, but did not require problem solving. Power spectra were computed during each of the Phase III tasks. Measures of the AER and power spectra data from the three phases were statistically analyzed one at a time using univariate analysis of variance with an Epsilon correction. These analyses showed many task and lead main effects in all three phases of the study, indicating that the Ss did have different EEG responses to the various stimuli and tasks. In Phase III there was a task by lead interaction in the alpha band (8-13 Hz) analysis of parietal activity in which the right parietal power was greater than the left parietal power during the verbal task. There was a difference in the same direction and of the same magnitude during the baseline task. During the spatial task the relative difference between right and left parietal leads was at its maximum, but its direction was counter to the view of the relation between the EEG and hemispheric specialization wherein one would predict relative alpha suppression in the right hemisphere during the performance of a spatial task. The results of this experiment are not consistent with the notion that the EEG reflects hemispheric specialization.

An interactive method for structural analysis of the subjective organization of human memory is developed which is based upon analysis of interitem proximities derived from free recall interresponse times. The proximity analysis is shown to be effective in revealing organizational structure through the use of both cluster and multidimensional scaling analyses. This technique is tested in a series of multitrial free recall experiments in which the presentation order of stimuli on trial N + 1 is determined by a real-time cluster analysis of recall data from the previous N trials. In the Clustered condition, items on the second and subsequent trials are presented blocked by clusters. This condition is contrasted with Random, Yoked and Fixed presentation orders, using both strongly and weakly categorizable lists. The Yoked group learned faster than the Random group with categorized lists and not as fast with uncategorized lists, indicating that the clustering had some effect upon developing organizational structure. Surprisingly, the Clustered group learned only slightly faster than the Yoked and Random groups, and not as fast as the Fixed group with both types of lists. These results are explained in terms of the qualitative nature of the information provided to the learner in each of the experimental conditions. This structural analysis approach to the measurement of memory organization is recommended for future research in semantic memory.

A series of experiments was conducted to elucidate the role(s) of special or strategic effortful processes in the encoding and retrieval of frequentistic information. Previously, investigators have concluded that sensitivity to frequency was the result of an "automatic" mechanism. However, it is difficult to generalize these findings to what people do when processing real-world frequentistic data because most studies have involved highly artificial experimental stimuli. By contrast, the present series of studies employed multidimensional stimuli that had real world significance to the participants. The complexity and realism of the stimuli facilitated the relatively unobscured assessment of the cognitive processes involved in frequency encoding. The purpose of Experiment 1 was to classify individuals according to the ability to "timeshare" in a dual-task setting. Timesharing ability was operationalized in terms of subject responsiveness to a payoff scheme which was designed to induce tradeoffs in the distribution of attentive effort between two concurrent tasks. The thrust was on the classification of prospective subjects for further investigations of frequentistic encoding and retrieval operations. A sample of four subjects in possession of uniformly good to excellent timesharing skills was enlisted for participation in the subsequent experiments. In Experiment 2, a complex rating task was used to test the hypothesis that special effortful processing is required in the formation of frequency records. The task involved the consistent aggregation of the evaluative cues comprising hypothetical candidate profiles. The principal question was whether the encoding of frequency information would be an incidental by-product of the effortful processing associated with the rating task. The negative findings strongly suggest that special (or strategic) effortful processing is required in the formation of reasonably accurate frequency records. In Experiment 3, the tasks of frequency estimation and hypothetical candidate rating were examined concurrently in a dual-task paradigm. A pronounced performance tradeoff was found between the two tasks; that is, improved performance on the frequency task was gained at the expense of degraded rating task performance, and vice versa. This suggests that whatever the processes governing frequentistic encoding operations, they draw upon the same pool of limited cognitive resources as those governing the rating task. This, of course, provides further support for the view that frequency encoding is the result of special effortful processing strategies. Implications of these findings for real-world decision making are examined, and ideas for future research are discussed.

Memory inhibition has been the subject of much thought and, in this century, much research. Laboratory findings of inhibition to date, however, have virtually all involved creating an incongruity between the study and test contexts, the conditions under which a memory is formed and those under which it is evinced. Studies reported here demonstrate what can be construed as a qualitatively different type of memory inhibition, one that is in some sense context-free. When a just-studied word is tested for recall with word-fragment cues of successively increasing strength (e.g., R------P; R----R-P; R-I--R-P; R-I--ROP), each presented for 4 seconds, performance is impaired compared to when only the strongest fragment cue (R-I--ROP) is presented for 4 seconds, even though in the first case there is additional opportunity for subjects to recall the word (RAINDROP). Indeed, level of performance varies systematically with initial cue strength. Thus, for example, if cuing starts with a 3-letter fragment, the word is recalled less often than if it starts with a 4-letter fragment but more often than if it starts with a 2-letter fragment. Moreover, the effect remains essentially undiminished even if subjects are allowed an additional full minute with the critical cue (R-I--ROP). Since the procedure used here involves manipulating only the graphemic aspects of a given word and not the context in which it was learned, the inhibition is context-free. The inhibition is discussed in light of a theoretical distinction (Mandler, 1980) between elaboration and integration processes in list learning. Another finding of interest is that no such inhibitory effect occurs when the to-be-discovered words are not studied before the test. When subjects try to complete the word fragments on the basis of their general knowledge of words, or "semantic memory," performance between the successive and single-level cuing conditions does not differ. The overall pattern of results therefore provides strong support for the episodic-semantic memory distinction formulated by Tulving (1972).

Three processes, item identification speed, susceptibility to proactive interference, and memory for order, identified for their potential as sources of memory span differences, were the subjects of the present investigation. A fourth variable, item memory, was included as the experimental complement of order memory (Healy, 1974). Performance on each of these four variables and memory span was obtained for 90 college-aged subjects. Moderate to high reliability was evidenced for each of the tasks. Memory span correlated significantly with each source of differences examined. Additionally, using a multiple regression analysis, each variable was shown to contribute significant, independent variance in the joint prediction of memory span. Subjects' scores on the four predictor variables were used in a cluster analysis which identified four subsets of individuals. An ANOVA determined that the four groups differed in their mean memory span scores. The differences among these groups were examined using a multiple discriminant analysis. The groups were found to differ along three dimensions. Evidence from these analyses indicates that an individual's memory span is the result of complex processes involving, at least, components of item identification speed, susceptibility to proactive interference, order, and item memory. Finally, individuals' performance levels across these tasks were not uniform, indicating possible differential contributions of these processes.

The multiple regression model has been applied to the study of judgment primarily through the paradigms known as policy capturing and multiple cue probability learning (MCPL). The former involves modeling the way the decision maker weights predictive information; the latter focuses on the process by which the decision maker acquires this strategy. In the first part of the paper, the logic underlying the two approaches is examined, and the research generated by each is discussed. Based on this comparative review, it is argued that neither approach has yet lived up to its potential because each has concentrated too narrowly on particular kinds of issues, at least some of which present serious logical difficulties. Policy capturing, for example, has sought to isolate judgment processes through multiple regression despite the fact that it is logically capable of capturing only predictions. For its part, MCPL research has fallen short by limiting itself to variables suggested by Brunswik's lens model. Better use could be made of the model if it were applied in a more functional manner--one in which it is used to index performance rather than to infer cognitive processes. The functional approach serves as a guiding philosophy for a series of experiments that are described in the second part of the paper. The primary question of interest in these experiments was how the format of displaying visual information affects subsequent judgments and decisions. Two types of displays, numerical and graphical, were investigated using the policy capturing paradigm in the first three experiments and MCPL in the fourth one. The most consistent finding was that subjects' cue weighting differed reliably with type of format. Numerical policies tended to be less precise than graphical ones, but accuracy of predictions did not differ with display. The implications of these findings are discussed both from a practical and theoretical perspective.

A review of the literature on alphanumeric displays, especially computer-generated displays, suggests that four basic characteristics of display formats affect how well users can extract information from the displays: (1) Overall density--the number of characters displayed, expressed as a percentage of the total spaces available; (2) Local density--the number of other characters near each character; (3) Grouping--the extent to which characters on the display form well defined perceptual groups; and (4) Layout complexity--the extent to which the arrangement of items on the display follows a predictable visual scheme. Objective ways of measuring these display characteristics have been developed and implemented in a computer program. In Experiment 1, 520 computer-generated displays that varied on these display measures were studied. Search times to locate data items on the diplays were measured as well as subjective ratings of ease of use. Regression equations were developed to predict the search times and subjective ratings using the display measures. The results indicated that both search times (R('2) = .508) and subjective ratings (R('2) = .805) could be predicted quite well. In experiment 2, the regression equations developed in Experiment 1 were used to predict, a priori, search times and subjective ratings for a new set of 150 displays. The regression equations generalized quite well, resulting in high correlations between predicted and actual search times (r = .800) and subjective ratings (r = .799). The regression equations indicate that the most important predictors of search time are two measures associated with the grouping of characters: the number of groups on the display and the average visual angle subtended by those groups. Likewise, the most important predictors of subjective rating are a measure of local density, which is essentially how "tightly packed" the display is, and a measure of layout complexity, which is essentially how well the items on the display are aligned with each other.