A general model for the analysis of multilevel data

A general model is developed for the analysis of multivariate multilevel data structures. Special cases of the model include repeated measures designs, multiple matrix samples, multilevel latent variable models, multiple time series, and variance and covariance component models.We would like to acknowledge the helpful comments of Ruth Silver. We also wish to thank the referees for helping to clarify the paper. This work was partly carried out with research funds provided by the Economic and Social Research Council (U.K.).     

AN ANALYSIS OF THE EFFECTS OF MULTIPLE SETTING EVENTS ON THE SOCIAL BEHAVIOR OF PRESCHOOL CHILDREN WITH SPECIAL NEEDS

We examined the effects of four combinations of setting events on the social interactions of 7 preschool children with social delays. In Study 1, the status of the teacher, activity materials, and peer varied across conditions. In Study 2, the status of the teacher and materials varied across conditions. Within the combinations of setting events, we also examined teacher behavior. Teacher presence and absence was varied in both studies. The type and rate of teacher prompting were varied in Study 2. The four combinations of setting events produced different rates of social behavior by the children with social delays. The optimal combination of setting events for promoting peer interaction and reducing teacher—child interaction included teacher absence from the activity, a limited number and form of materials, and children paired with a socially skilled partner.     

The effects of family composition,family health,parenting behavior and environmental stress on children's divorce adjustment

The traditional family composition model of children's divorce adjustment was compared to a family competence model. Subjects included 81 children whose parents remained married, were divorced, or were divorced and subsequently remarried. Divorced families had been separated for between 24 and 48 months. Custodial parents completed questionnaires regarding family functioning, occurrence of stressful life events, and child's psychosocial adjustment. Children completed questionnaires regarding parenting behaviors and self-esteem. A trained examiner conducted a structured diagnostic interview of the child. Multivariate analysis of covariance showed that family composition had a significant effect on the occurrence of stressful events and change in income but not children's adjustment. Hierarchical multiple regression correlations indicated the most powerful predictors of child adjustment were family competence variables which accounted for 21% of the variance in child adjustment scores. In contrast, family composition accounted for only 8% of variance in child adjustment scores. Further, hierarchical multiple regression correlations indicated that higher levels of family functioning were associated with families where parental hostility was low and parents displayed few rejecting behaviors while practicing consistent and appropriate discipline.     

Preventing antisocial behavior in the schools

Multiple correlates and determinants of antisocial behavior within the home, community, and school are reviewed. Due to the school's pivotal role in our society, an emphasis is placed on how our schools contribute to antisocial behavior, and what educators can do to prevent anti-social behavior and related attendance problems. A variety of contextual factors and setting events within our schools appear to be major contributors to antisocial behavior, and some of the same factors identified within the schools also have been identified within the home. These setting events, rather than quick restrictive fixes, must be given more attention if we are to provide safe school environments-environments that durably prevent antisocial behavior and related attendance problems.     

Cognitive sequelae of repeated head injury in a population of intravenous drug users

The relationship between closed head injury and performance on neuropsychological (NP) tests was investigated in a group of intravenous drug users (IVDUs). Subjects with repeated head traumas involving loss of consciousness (LOC) performed worse than both a control group without LOC and reference group with only a single episode of LOC. There were no significant differences between the last two groups. Performance on tests of memory, attention, and motor performance was significantly worse in the group with repeated head injury. The average time since the last episode of LOC was more than 11 years. We conclude from these findings that a single episode of LOC does not result in significant cognitive impairment in this population. Two or more episodes, however, are more likely to produce chronic cognitive impairment.     

ADVANCING BACKWARDS

The proposal of Friman, Hayes, and Wilson (1998) that the concept of emotion be incorporated into behavior analysis is questioned. Their proposal is undermined by the problem of the verification of private events by others. A focus on private events, as exemplified by the proposal of Friman et al., can lead to a reversion to notions of agency and the autonomous individual. Also, the anchoring of hypothetical constructs to the behavioral data language is ill advised.     

Drug effects on repeated acquisition: comparison of cumulative and non-cumulative dosing

Pigeons acquired a different four-response chain each session by responding sequentially on three keys in the presence of a sequence of four colors. The response chain was maintained by food presentation under a fixed-ratio schedule. Errors produced a brief timeout but did not reset the chain. Each day there were four 15-minute sessions, with a 10-minute inter-session interval. Cumulative dose-effect curves for phencyclidine, pentobarbital, and d-amphetamine were obtained by giving an injection before each of the four sessions; successive injections increased the cumulative dose in equally spaced logarithmic steps. For comparison, non-cumulative doses of each drug (i.e., doses not preceded by other doses on the same day) were also tested. As the cumulative dose of each drug increased, the overall response rate decreased, the percent errors increased, and there was less within-session error reduction (acquisition). With phencyclidine and pentobarbital, the rate-decreasing and error-increasing effects tended to be greater with a non-cumulative dose than with the corresponding cumulative dose. In contrast, with d-amphetamine, the effects were considerably greater with the cumulative doses. The results indicate that although the cumulative-dosing procedure saved a substantial amount of time in determining dose-effect curves, there were quantitative differences in effects between cumulative and non-cumulative doses.     

An experimental analysis of the effects of d-amphetamine and cocaine on the acquisition and performance of response chains in monkeys.

In one component of a multiple schedule of food presentation, monkeys acquired a different four-response chain each session by responding sequentially on three keys in the presence of four geometric forms (learning). In the other component, the four-response chain was the same each session (performance). Both d-amphetamine and cocaine, at the higher doses, disrupted the behavior in the learning component; the overall response rate decreased, the overall accuracy was impaired (i.e., percent errors increased), and there was less within-session error reduction. The performance component was generally less sensitive than the learning component to the disruptive effects of both drugs on rate and accuracy. After pre-feeding or during an extended session, the response rate decreased in both components, but accuracy was generally unaffected. When the four discriminative stimuli in both components were removed, the behavior was disrupted to a greater extent in the performance component. The disruptive effects of both drugs on behavior in the learning component were attenuated when the drugs were administered during the session after the response chain had been acquired. It was concluded that the greater sensitivity of the learning component to disruptive drug effects is related to the relatively weak stimulus control and/or the lower rate of reinforcement associated with that component.     

Effects of cocaine and d-amphetamine on the repeated acquisition and performance of conditional discriminations.

The acute and chronic effects of cocaine and d-amphetamine on food-reinforced behavior were investigated in pigeons responding on a two-component multiple schedule. In one component, the behavioral task consisted of the same chain of conditional discriminations each session (performance). In the other component, the chain of conditional discriminations was changed from session to session (learning). In comparison to control sessions, both acute cocaine and d-amphetamine increased errors in each component of the multiple schedule. Responding in the learning component, however, was generally disrupted at lower doses than those that affected responding in the performance component. At high doses, both drugs produced pauses in responding in each component in three of the four subjects. Pausing engendered by d-amphetamine was approximately twice as long as that under cocaine. Upon chronic administration, both the pausing and error-increasing effects of each drug diminished. Drug-induced changes in timeout responding, however, did not decrease during chronic administration. Redeterminations of the d-amphetamine dose-effect curves following chronic cocaine administration suggested the existence of cross-tolerance between cocaine and d-amphetamine. Both the acute and chronic data are consistent with the view that conditions of stimulus control may modulate the behavioral effects of drugs.