Empirical bayes estimation of coefficients in the general linear model from data of deficient rank

Empirical Bayes methods are shown to provide a practical alternative to standard least squares methods in fitting high dimensional models to sparse data. An example concerning prediction bias in educational testing is presented as an illustration.The authors would like to thank the referees for several useful comments.The analysis of the data discussed in this report was part of a study funded jointly by the Graduate Management Admission Council and Educational Testing Service.     

Duration judgment and the segmentation of experience

Two experiments tested an interval segmentation explanation of duration judgment. In Experiment 1, a 170-sec time interval was filled with 27 unrelated words and three high-priority events (HPEs). These HPEs were clustered at the beginning of the interval (unsegmented condition) or distributed throughout the interval (segmented condition). Both recognition and recall of list information were measured, in addition to duration estimates. While no differences in memory performance were found, duration estimates were greater in the the segmented condition. Experiment 2 also tested the effects of interval segmentation but used 36 words and eight HPEs, a longer clock duration, and also measured the remembered number of events (RNE) in the interval. As in Experiment 1, the segmented condition produced longer duration estimates in the absence of memory performance and RNE differences. A segmentation hypothesis seems a better explanation of these results than previously proposed storage size or amount-of-processing models of duration judgment.     

Inadequacy of voice recognition as a demonstration of self-deception

Gur and Sackeim (1979) argued that subjects deceived themselves when they failed to recognize their own voices on playback from a tape recorder. This claim is based primarily on the observation that subjects showed a heightened galvanic skin response when their own voices were present regardless of whether recognition took place. The authors suggest that even though subjects may not consciously recognize their own voices, a heightened physiological response implies that true recognition did in fact occur at some other level of cognitive processing. This article describes an experiment demonstrating that results similar to those arrived at by Gur and Sackeim can also be produced when subjects attempt to recognize the voice of a familiar "other." These results suggest that self-deception is not the main factor operating to produce the heightened physiological response.     

Career development of special populations: A framework for research

In view of the recent interest in the career behaviors of diverse groups of individuals, this article presents a model for conducting research on the career development of special populations. Available research paradigms are reviewed in terms of their potential impact on such efforts. An alternative framework, designed to make maximum use of available data sources while minimizing potential theoretical bias, is offered. An illustration of research conducted within the recommended framework is presented in which the impact of a variety of career-related variables on the attitudinal and cognitive aspects of vocational maturity is examined from the perspective of two special population groups, disabled and disadvantaged college students. Drawing upon previous research findings, variables under consideration included level of scholastic achievement and endorsement of rational, intuitive, and dependent decision-making styles. Also included were two population-specific variables related to the characteristics of membership in each of the two special groups. The regression analyses conducted to explore the relationships between vocational maturity and these predictor variables indicate, in general, that variance in the attitudinal and cognitive factors of vocational maturity can be at least partially explained by the variables considered. However, the relative contributions of the predictors and the resulting proportions of variance explained differs according to which population and which criterion factor is under examination. These findings are compared with those emerging from previous research, and implications for future efforts are discussed.     

Computer packages that contain the FFT

This paper critiques a set of software (i.e., SAS, BMDP, LSP) that contain a Fourier analysis program. In addition, this paper discusses both general-purpose software that runs a Fourier analysis and special Fourier analysis hardware. Some examples of the efficacy of commercially available software and hardware are given.     

Taste-taste,odor-odor,and taste-odor mixtures: Greater suppression within than between modalities

Subjects judged the intensities of the component qualities of two taste stimuli [sucrose (S) and NaCl (N)], two odor stimuli [citral (C) and anethole (A)], and the mixtures of these stimuli [taste-taste (SN), odor-odor (CA), and taste-odor (SC, NC, SA, and NA)]. The within-modality mixtures, taste-taste and odor-odor, produced greater suppression of perceived intensity than did the between-modality mixtures, taste-odor. However, taste-odor mixtures also produced a significant amount of suppression of perceived intensity when compared with the individual taste and odor stimuli. The data are discussed with respect to peripheral and central mechanisms in mixture suppression.     

Effects of reward on delayed reaction time task performance of hyperactive children

This experiment was designed to investigate the effects of continuous, partial, and noncontingent schedules of reward, as well as the withdrawal of rewards, on the performance of hyperactive and normal control children on a delayed reaction time task. Although noncontingent reward resulted in faster reaction times for control subjects, performance of hyperactives deteriorated under noncontingent reward and improved when it was withdrawn. Also, reaction times of controls during extinction remained superior to baseline, whereas performance of hyperactives returned to baseline level. It was suggested that these and other findings reviewed point to an unusual sensitivity to rewards in hyperactive children.     

Effects of reinforcement on concept identification in hyperactive children

The effects of continuous and partial reward on the performance of hyperactive and normal children on a concept identification task were compared. Because reduction in information feedback is usually associated with partial reward schedules, the study was designed to yield information regarding the contribution of reduced feeback to performance in the partial reward condition. Previous findings of a performance deficit in hyperactives under partial reward were replicated. The findings help rule out an information feedback explanation for this deficit. The authors suggest that a motivational factor involving the elicitation of frustration when expected rewards fail to appear may be responsible for the poor performance of hyperactives on the partial schedule.This research was supported by Grant No. 943-03-34 from the Social Sciences Research Fund, McGill University.