A rational model of the effects of distributional information on feature learning

Most psychological theories treat the features of objects as being fixed and immediately available to observers. However, novel objects have an infinite array of properties that could potentially be encoded as features, raising the question of how people learn which features to use in representing those objects. We focus on the effects of distributional information on feature learning, considering how a rational agent should use statistical information about the properties of objects in identifying features. Inspired by previous behavioral results on human feature learning, we present an ideal observer model based on nonparametric Bayesian statistics. This model balances the idea that objects have potentially infinitely many features with the goal of using a relatively small number of features to represent any finite set of objects. We then explore the predictions of this ideal observer model. In particular, we investigate whether people are sensitive to how parts co-vary over objects they observe. In a series of four behavioral experiments (three using visual stimuli, one using conceptual stimuli), we demonstrate that people infer different features to represent the same four objects depending on the distribution of parts over the objects they observe. Additionally in all four experiments, the features people infer have consequences for how they generalize properties to novel objects. We also show that simple models that use the raw sensory data as inputs and standard dimensionality reduction techniques (principal component analysis and independent component analysis) are insufficient to explain our results.     

A rational model of the effects of distributional information on feature learning

Most psychological theories treat the features of objects as being fixed and immediately available to observers. However, novel objects have an infinite array of properties that could potentially be encoded as features, raising the question of how people learn which features to use in representing those objects. We focus on the effects of distributional information on feature learning, considering how a rational agent should use statistical information about the properties of objects in identifying features. Inspired by previous behavioral results on human feature learning, we present an ideal observer model based on nonparametric Bayesian statistics. This model balances the idea that objects have potentially infinitely many features with the goal of using a relatively small number of features to represent any finite set of objects. We then explore the predictions of this ideal observer model. In particular, we investigate whether people are sensitive to how parts co-vary over objects they observe. In a series of four behavioral experiments (three using visual stimuli, one using conceptual stimuli), we demonstrate that people infer different features to represent the same four objects depending on the distribution of parts over the objects they observe. Additionally in all four experiments, the features people infer have consequences for how they generalize properties to novel objects. We also show that simple models that use the raw sensory data as inputs and standard dimensionality reduction techniques (principal component analysis and independent component analysis) are insufficient to explain our results.     

The effects of test-wiseness materials on standardized test performance of preschool disadvantaged children

Evidence was sought to test the claim that many young children lack the abilities prerequisite to taking standardized tests and, therefore, perform poorly. After identifying a number of prerequisite abilities, curricular materials were designed and implemented in an effort to increase the test-wiseness of children unfamiliar with standardized tests. Children from Head Start classes worked with the materials twice a week for six weeks. The experimental group improved significantly on Total Score and the Matching subtest of the Metropolitan Readiness Test administered immediately after completing the instruction. However, group differences, when measured four months later, were not significant.     

Effects of reinforcement on standardized test performance

The effects of two different motivational conditions upon standardized test performance were explored for two student populations. The first study involving 12 trainable retardates showed a significant increase in score on the Metropolitan Readiness Test given under reinforcement conditions when these results were compared with scores taken under standard testing conditions. In a second study, these same results were obtained with a group of 30 normal fourth-graders. An additional study was conducted to determine the effect of different experiences with token reinforcement procedures on test performance. It was found that a group of children with six weeks' exposure to reinforcement for daily academic performance scored higher under both conditions of test administration (standard and reinforcement) than a control group. However, in a single exposure to token reinforcement for correct performance on the Metropolitan Test, both the experimental group and its match control showed a parallel increase in test performance. These findings offer a procedure that yields a more representative assessment of a student's academic achievement than does testing under standard conditions.     

The effects of token reinforcement,feedback and response cost on standardized test performance

The primary purpose of the present study was to compare the differential effects of token reinforcement, feedback, and response cost on the test performance of delinquent boys. Eighty students were randomly assigned to three experimental groups and one control group of 20 subjects each. Each experimental group received a standard and a modified administration of the verbal section of the WISC. For the token reinforcement group, the modified WISC administration permitted students to earn tokens contingent on correct responses; the response cost group forfeited tokens contingent on incorrect responses; and the feedback group simply received information regarding the accuracy of each response. The control group received two standard WISC administrations. The primary measure was the difference in verbal I.Q. scores between the standard and modified WISC administrations. Results indicated that the token reinforcement and response cost groups achieved significantly higher scores than the feedback and control groups. No significant differences were found between the token reinforcement and response cost groups nor between the feedback and control groups. The implications of these findings for clarifying the relationship between motivational condition and test performance are discussed.     

A brief introduction to unified test theory

This article introduces McDonald's unified treatment of test theory, which merges the major contributions of Spearman (True Score Theory and Common Factor Theory) with aspects of Item Response Theory. The fundamentals are first presented. followed by elaboration of selected aspects of the treatment. An SAS program is given that estimates relevant parameters.     

A cautionary tale of two statistics: partial correlation and standardized partial regression

Partial correlation and standardized partial regression are widely used to control for other variables in cross-sectional and longitudinal studies. Two aspects of these statistics can radically alter the interpretation of data and so deserve attention. First, their sign will be opposite to that of the original zero-order correlation when this correlation is of the same sign as and smaller than the product of the other 2 correlations. Second, their size will be considerably larger than a weak zero-order correlation when either that correlation is positive and the product of the other 2 correlations is negative or that correlation is negative and the product of the other 2 correlations is positive. The first aspect is illustrated with data from J. M. Gottman and L. J. Krokoff (1989) on initial conflict behavior and initial and later marital satisfaction, and the 2nd aspect is illustrated with data from C. L. Heavey, C. Layne, and A. Christensen (1993) on the same issue. Their interpretation of their findings is questioned.     

Improving adolescents' standardized test performance: An intervention to reduce the effects of stereotype threat

Standardized tests continue to generate gender and race gaps in achievement despite decades of national attention. Research on “stereotype threat” (Steele & Aronson, 1995) suggests that these gaps may be partly due to stereotypes that impugn the math abilities of females and the intellectual abilities of Black, Hispanic, and low-income students. A field experiment was performed to test methods of helping female, minority, and low-income adolescents overcome the anxiety-inducing effects of stereotype threat and, consequently, improve their standardized test scores. Specifically, seventh-grade students in the experimental conditions were mentored by college students who encouraged them either to view intelligence as malleable or to attribute academic difficulties in the seventh grade to the novelty of the educational setting. Results showed that females in both experimental conditions earned significantly higher math standardized test scores than females in the control condition. Similarly, the students—who were largely minority and low-income adolescents—in the experimental conditions earned significantly higher reading standardized test scores than students in the control condition.     

The effects of confederate prompt delivery style in a standardized social simulation test

Role plays or social simulations are presently one of the most common methods for assessing social skills. Although social simulation techniques have become quite popular, little is known about the psychometric adequacy of many of the role-play instruments which have been developed. This investigation was an attempt to determine certain properties of one particular social skills assessment instrument: the Simulated Social Interaction Test. The effects of various confederate prompt delivery styles on the judged social competency level of subjects in the Simulated Social Interaction Test were examined. Two confederates, one male and one female, were trained to portray three confederate prompt delivery styles: (a) unreceptive, (b) neutral, and (c) receptive. In order to determine if judges would compensate for the different confederate prompt delivery styles, two sets of trained judges rated the levels of skill and anxiety generated by the 30 subjects tested. One set of judges was screened from the confererate delivery while observing the simulation; the other set of judges followed the more common rating procedure (i.e., they were not screened from the confederates' deliveries). Data analysis examined variation in subject performance under each condition and according to each set of judges for both overall social skills and anxiety ratings. Confederate prompt delivery style was found to affect subjects' rated performance. There was also suggestive evidence that judges were sometimes able to compensate for different confederate delivery styles. The practical and theoretical implications of the results are discussed.This study was funded in part by a research grant from the Veterans Administration.     

The effects of distributional learning on rats' sensitivity to phonetic information

During the first year of life, infants show decreased sensitivity to phonetic differences not used in their native language and increased sensitivity to the differences that are used. It has been shown that this change in speech perception is a function of the distributional properties of the input. The present study explores whether the mechanism responsible for the developmental changes regarding the organization of phonetic categories is a general mechanism shared with other animals. The results demonstrate that the distributional exposure to a phonetic continuum affects the subsequent discrimination of these phonemes in rats, indicating that the ability to use distributional cues to change the phonetic category structure extends beyond humans.     

Empirical derivation of SVIB-Holland scales: A brief report

The Strong Vocational Interest Blank responses of 93 students were used to construct six empirical scales similar to the scales of Holland's Vocational Preference Inventory. Scores on the empirical scales were correlated with actual VPI scores. The resulting correlations were compared to coefficients obtained from correlating the intuitive scales designed by Campbell with actual VPI scores. It was concluded that (1) meaningful estimates of VPI profiles can be obtained by scoring selected items from the SVIB and (2) further work with the empirical scales is needed prior to settling on a SVIB scoring procedure for estimating VPI profiles.     

Two new test statistics for the rasch model

It is shown that the presently available statistical tests for the Rasch model are insensitive to violation of the unidimensionality axiom. Two new test statistics are presented. The first one,Q ₁, is sensitive to the same effects as the presently available statistics, but has some desirable properties of a nonstatistical nature. The second statistic,Q ₂, is sensitive to violation of local stochastic independence and unidimensionality and thus fills an existing gap.     

Sense of meaning and academic performance: A brief report

This study examined sense of meaning and academic performance among a sample of South African university students (n = 210, mean age = 19.49, SD = 1.99, female = 54.29%). Data on meaning were collected using the Purpose in Life Test. End of semester marks served as the academic attainment measure. Simple regression analysis with post-hoc independent-samples t-tests revealed that meaning predicts academic performance. No significant differences between genders emerged from the data. Sense of meaning is an important factor in supporting students to enhance their academic performance.     

Children's subjective ratings of their performance on a standardized achievement test

Relationships between children's perceptions of performance on a standardized achievement test and the actual test performance were examined. A total of 311 children in grades 4, 5, and 6 were administered the Iowa Tests of Basic Skills on four separate testing days. At the end of the testing session on Day 4, the children completed a questionnaire in which they rated their performance on each of the subtests as well as their overall score on the test. Students also were asked to make judgments about the difficulty of the test, the effort expended, the extent that ability versus luck was a factor in their test performance, and the importance of the test to them, their parents, and their teachers. The results showed that students were quite accurate in predicting their scores on the subtests and that a personal belief that ability decisively influences achievement associated with their test performance. These findings are discussed as they relate to the effect of self-statements on test performance.