Psychological Bulletin

Porges, Stephen W.; Bohrer, Robert E.; Cheung, Michael N.; Drasgow, Fritz; McCabe, Philip M.; Keren, Gideon

doi: 10.1037/0033-2909.88.3.580pmid: N/A

The relationships among physiological systems and between physiological and behavioral systems have frequently been evaluated with traditional descriptive statistics. These methods may not be sensitive to underlying rhythmic relationships. Although cross-spectral analysis provides a method for assessing rhythmic co-occurrence, it is only capable of describing the covariation of 2 systems at specific frequencies. Since physiological and behavioral systems tend not to be manifested at a constant frequency, a new statistic is described that is capable of evaluating the shared rhythmicity of 2 systems across an entire band of frequencies. (13 ref)

Bentler, P. M.; Bonett, Douglas G.

doi: 10.1037/0033-2909.88.3.588pmid: N/A

Factor analysis, path analysis, structural equation modeling, and related multivariate statistical methods are based on maximum likelihood or generalized least squares estimation developed for covariance structure models (CSMs). Large-sample theory provides a chi-square goodness-of-fit test for comparing a model (M) against a general alternative M based on correlated variables. It is suggested that this comparison is insufficient for M evaluation. A general null M based on modified independence among variables is proposed as an additional reference point for the statistical and scientific evaluation of CSMs. Use of the null M in the context of a procedure that sequentially evaluates the statistical necessity of various sets of parameters places statistical methods in covariance structure analysis into a more complete framework. The concepts of ideal Ms and pseudo chi-square tests are introduced, and their roles in hypothesis testing are developed. The importance of supplementing statistical evaluation with incremental fit indices associated with the comparison of hierarchical Ms is also emphasized. Normed and nonnormed fit indices are developed and illustrated. (43 ref)

Wike, Edward L.; Church, James D.

doi: 10.1037/0033-2909.88.3.607pmid: N/A

The Type I error rates for K. J. Levy's nonparametric pairwise tests for 1- and 2-way layouts were found in 2 Monte Carlo studies. In Exp I, Levy's test was contrasted with the Wilcoxon rank-sum test at |a = .05 and .01. Exp II compared Levy's test with the Wilcoxon signed-ranks, sign, and stepped-down sign tests at |a = .05 and .01. Both Levy tests displayed excellent experimentwise error rates. But when Levy's tests were used as "protected" tests after a significant overall Kruskal-Wallis or Friedman test, their experimentwise error rates were inferior to the Wilcoxon rank-sum test in the 1-way layout and to the Wilcoxon signed-ranks and the stepped-down sign tests in the 2-way layout. (18 ref)

Jay, Timothy B.

doi: 10.1037/0033-2909.88.3.614pmid: N/A

A. Haas found that little evidence exists to distinguish between males' and females' use of dirty words. In contrast, the present author's review found several empirical studies comparing the sexes. It is argued that establishing whether differences exist does not necessarily require recording colloquial speech, nor would the phenomenon have to be studied entirely outside of the laboratory setting. (41 ref)

Nesselroade, John R.; Stigler, Stephen M.; Baltes, Paul B.

doi: 10.1037/0033-2909.88.3.622pmid: N/A

Examines regression effects in longitudinal sequences of observations by formulating expectations (Es) for later observations conditioned on an initial selection score value. The Es are developed for several variations of classical test theory and autocorrelation models. Es based on the general concept of regression are seen not only to depart from those depicted in the psychometric lore but to vary considerably from one underlying model to another, particularly when moving from the 2-occasion to a multiple-occasion measurement framework. Regression toward the mean is not a ubiquitous phenomenon, nor does it always continue across occasions. It is necessary to specify the characteristics of model assumptions to understand the when, how, and extent of regression toward the mean. Past interpretations have been incomplete and to an extent incorrect because they focused largely on a limited circumstance: 2-occasions of measurement and simplexlike correlation matrices. (23 ref)

Welch, Robert B.; Warren, David H.

doi: 10.1037/0033-2909.88.3.638pmid: N/A

Reviews the literature on the intersensory bias (IB) resulting when the information available to different sensory modalities is caused to be discrepant. Characteristics for which discrepant sensory information has been provided include spatial location, shape, size/length, and orientation. Stimulus variables affecting the magnitude of IB include structural and cognitive factors. Response variables affecting intersensory bias are the type and timing of the response required of Ss. Evidence on developmental trends and individual differences in the susceptibility to IB is both sparse and conflicting. Also unclear is the relationship between IB and adaptation to perceptual rearrangement. Since none of the various theories––modality precision, directed attention, and modality appropriateness––accounts for all of the intersensory discrepancy results, a new model is proposed, based on the supposition that IB is a result of an attempt by the perceptual system to maintain a perceptual experience consonant with a unitary event. (2½ p ref)

Berman, Phyllis W.

doi: 10.1037/0033-2909.88.3.668pmid: N/A

Reviews studies that compared responses of male and female children and adults to infants, young children, baby animals, baby cries, or pictures of the young. Males' and females' responses and therefore sex differences varied greatly depending on the physical and social qualities of the situation, the response required, and the Ss' experimental or prior role relationship with the young. Demographic comparisons showed considerable variability between cultures. Within the American culture, responsiveness changed throughout the life span for both sexes, but these changes do not reflect life span hormonal changes, at least in any simple, straightforward manner. Implications for a theory of the social determinants of responsiveness to the young are discussed. (2½ p ref)

Wallace, William P.

doi: 10.1037/0033-2909.88.3.696pmid: N/A

Procedures for testing recognition memory have almost invariably required that Ss discriminate old items from new items. At the process level, recognition refers to the capacity of knowing or feeling that a given event has been experienced before. Several previous studies have used a discrimination procedure to test this capacity as a way to prevent Ss from indiscriminately claiming to recognize all test events. It is argued that valid objective tests of recognition memory can be conducted without formal discrimination-testing procedures. Testing recognition memory without distractor items on the test is a viable laboratory procedure, and it is more relevant to the basic concept of recognition memory than the standard discrimination-testing procedures. (45 ref)

Hsu, Louis M.

doi: 10.1037/0033-2909.88.3.705pmid: N/A

R. Rosenthal and D. Rubin proposed tests for contrasts of p values reported in independent studies. The present author indicates that although for studies of equal size these tests may be used to test hypotheses about equality of J. Cohen's (1977) effect sizes, for studies of unequal size these tests are generally not robust. An alternative test is described that may be used to test hypotheses of equality of Cohen's effect sizes in the case of unequal-sized studies. (2 ref)