Alphabetic confusion: A test of models for individuals

An earlier article reported extensive analyses of confusion data compiled from group averages (Townsend, 1971). The present study provided for essentially the same analyses with different long-term data obtained with two individuals, the primary intent being to examine the ability of recognition and scaling models to explain data at the individual level (which the recognition models purport to describe) and to compare the confusion characteristics of the English uppercase alphabet between the two Ss and between the individual Ss and the group-averaged data. The choice and overlap models were superior to the all-or-none model in predicting the empirical confusion matrices and tended to explain the data structure in a similar manner. Multidimensional scaling analysis again supported a Euclidean metric and suggested four or five underlying stimulus dimensions. However, as before, there were no overriding intuitively appealing psychological dimensions corresponding to these, and possible reasons are discussed. The choice and overlap models appeared to fit as well or better at the individual level than at the group level and the all-or-none model to fit worse. In the present study, probability correct was fit even better by the all-or-none model than in the group study and replicated the result of being better here than the overlap and choice models. Individuals and the group were consistent in their sensory confusions as represented by similarity parameters in the choice and overlap models but differed in their response biases. A simple measure of physical similarity explained 50% of the variance of the similarity structure in the confusion data.     

The equivalence of target and nontarget processing in visual search

A comparison of a forced-choice visual search task with an item recognition task did not support Neisser’s (1967) hypothesis of a preattentive stage that processes targets and nontargets differentially. In the forced-choice condition, Ss indicated which of two items in a visual display was a target; in item recognition, Ss determined whether or not the single item in the visual display was a target. The size of the memorized set of possible targets was varied from one to six items for both tasks. Latencies increased linearly with memory set size in both conditions; the slopes for forced choice and item recognition were 41.8 and 27.9 msec per item, respectively. The ratio of 1.38 between the two slopes was well fit by Sternberg’s (1967) item recognition model, which predicts a ratio of 1.50.     

Tests of product and additive scaling axioms

A product axiom for ratio judgments and an additive axiom for difference judgments were tested for six Ss judging brightness. The product axiom was rejected for all six Ss whereas the additive axiom was accepted for two Ss. Power functions did not fit the data well, although somewhat better for difference judgments than ratio judgments. A two-stage model fitted somewhat better, but failed to satisfy one important implication of that model, previously confirmed for lifted weights. The data gave stronger support for a theory of brightness estimation based on an additive axiom for difference judgments than a product axiom based on ratio judgments.     

Judgment stability and pseudorecognition of random and patterned configurations

Thirty six simple patterns were judged as “random” or “patterned” by adult Ss. Category judgments were consistent whether Ss were told that 1/4, 1/2, or 3/4 of the patterns were random or if no proportion was specified. In a second experiment, Ss viewed a series of slides previously judged patterned followed by a recognition series with two additional patterned slides inserted in a series of otherwise random slides. Although slides were never labeled random or patterned, Ss made the most false recognitions for the two patterned slides. In another condition where random and patterned slides were reversed, most false recognitions were made for the two random slides. Both Gamer’s notion of hypothetical set size and an analysis of balanced elements fit the results, but the adequacy of both hypotheses can be questioned when applied to previous results with more complex patterns.     

Confirmatory Factor Analysis of the Structure of Adult ADHD Symptoms

Previous research has examined the factor structure of ADHD symptoms in adults using confirmatory factor analysis and there is mixed support for two-factor, three-factor, or more complex bifactor models. The current study re-examines these factor structures on the basis of not only overall model fit, but also criteria such as chi-square difference tests, invariance, reliability, and validity. Three datasets were pooled to form a total sample of 430 adults (50% female). Participants completed questionnaires assessing their ADHD symptoms, education, psychopathology, and parenting (the latter as validity variables). Although the bifactor models demonstrated the best fit, the reliability and construct replicability of specific factors in these models was poor. The three-factor model had good fit and demonstrated good reliability, validity, construct replicability, and some invariance across gender and dataset. Our results replicate previous studies that find superior fit for bifactor models. However, the superior statistical fit may reflect only the increased complexity of these models. The more parsimonious three-factor model may demonstrate stronger validity and clinical utility.     

Visual recognition memory in reflective and impulsive children

Twenty-nine reflective and 29 impulsive fifth-grade boys were tested in a forced-choice visual recognition memory task. In three of the experimental conditions (1FD, 2FD, 4FD) the number of visual feature differences between the correct and incorrect test stimuli was 1, 2, or 4, and correct response could not be based on the name of the stimulus; in the fourth condition (DO) the correct and incorrect test stimuli had different names. As predicted, performance on DO and 4FD was equivalent and was superior to that on 1FD and 2FD. Although reflective Ss made more correct responses than impulsive Ss in all four conditions, only the performance difference in Condition 1FD was significant. Mean correct response latencies mirrored the correct response data. These results were consistent with the Selfridge-Neisser feature-testing model of recognition memory, and it was argued that the primary underlying basis for the dimension of reflection-impulsivity was that reflective Ss tend to engage in a more detailed visual feature analysis of stimulus arrays. Strong inferential evidence was provided that visual feature analysis independent of verbal labeling was responsible for successful recognition performance in these Ss.     

Models for the processing and identification of faces

Representations of human faces were constructed from Identi-Kit in a comparison of several possible models of information processing relating to facial recognition. There was no experimental evidence that faces were treated as unitary Gestalten, or that the component features were processed to any significant extent in parallel, even when the stimuli were presented as photographic positives in the normal upright mode. Instead, the best theoretical model to fit the data involved serial self-terminating processing, without replacement. Task difficulty, determining both processing time and number of errors made, was found to be a function both of the number of critical features present and the orientation (upright or inverted). Photographic negatives were handled in every way the same as positives. The more difficult the task, the stronger the evidence for a serial model and the greater the effects of practice. The latter suggests that, given time, Ss could handle inverted presentations as readily as upright ones, and that inversion may merely increase the difficulty in separating out the individual features for subsequent processing.     

Higher-order approximations to the distributions of fit indexes under fixed alternatives in structural equation models

Higher-order approximations to the distributions of fit indexes for structural equation models under fixed alternative hypotheses are obtained in nonnormal samples as well as normal ones. The fit indexes include the normal-theory likelihood ratio chi-square statistic for a posited model, the corresponding statistic for the baseline model of uncorrelated observed variables, and various fit indexes as functions of these two statistics. The approximations are given by the Edgeworth expansions for the distributions of the fit indexes under arbitrary distributions. Numerical examples in normal and nonnormal samples with the asymptotic and simulated distributions of the fit indexes show the relative inappropriateness of the normal-theory approximation using noncentral chi-square distributions. A simulation for the confidence intervals of the fit indexes based on the normal-theory Studentized estimators under normality with a small sample size indicates an advantage for the approximation by the Cornish–Fisher expansion over those by the noncentral chi-square distribution and the asymptotic normality. The author is indebted to the reviewers for their comments and suggestions, which have led to the improvement of the previous versions of this paper. This work was partially supported by Grant-in-Aid for Scientific Research from the Japanese Ministry of Education, Culture, Sports, Science, and Technology.     

Tests of an exemplar model for relating perceptual classification and recognition memory

Experiments were conducted in which Ss made classification, recognition, and similarity judgments for 34 schematic faces. A multidimensional scaling (MDS) solution for the faces was derived on the basis of the similarity judgments. This MDS solution was then used in conjunction with an exemplar-similarity model to accurately predict Ss' classification and recognition judgments. Evidence was provided that Ss allocated attention to the psychological dimensions differentially for classification and recognition. The distribution of attention came close to the ideal-observer distribution for classification, and some tendencies in that direction were observed for recognition. Evidence was also provided for interactive effects of individual exemplar frequencies and similarities on classification and recognition, in accord with the predictions of the exemplar model. Unexpectedly, however, the frequency effects appeared to be larger for classification than for recognition.     

Dual-process models of associative recognition in young and older adults: evidence from receiver operating characteristics

In 3 experiments, young and older adults studied lists of unrelated word pairs and were given confidence-rated item and associative recognition tests. Several different models of recognition were fit to the confidence-rating data using techniques described by S. Macho (2002, 2004). Concordant with previous findings, item recognition data were best fit by an unequal-variance signal detection theory model for both young and older adults. For both age groups, associative recognition performance was best explained by models incorporating both recollection and familiarity components. Examination of parameter estimates supported the conclusion that recollection is reduced in old age, but inferences about age differences in familiarity were highly model dependent. Implications for dual-process models of memory in old age are discussed.     

The role of scanpaths in the recognition of random shapes

Eye movements of 12 Ss were examined during learning and recognition of two-dimensional random shapes to determine the nature of the memorial representation of a stimulus and the utilization of this memorial representation in pattern recognition. Specifically, the purpose of this study was to test the scanpath model of pattern perception by determining whether scanpaths exist and, if so, how they influence recognition performance. Scanpaths, defined as overlapping fixation patterns in learning and recognition tasks, were observed in over half of all eye-movement records regardless of shape complexity. Presence of scanpaths did not increase recognition performance as measured by errors in recognition and Ss’ ability to reproduce the shapes. Although scanpaths did not influence recognition performance, their occurrence implicates them as a potential factor in the recognition process.     

结构方程模型常用拟合指数检验的实质

拟合指数检验是评价结构方程模型(SEM)的重要环节。从协方差结构分析的角度将SEM与传统的回归模型比较,容易理解为什么SEM需要拟合指数。揭示了目前几种流行的拟合指数检验的实质:基于卡方的绝对拟合指数(如RMSEA)检验的实质是重新设定卡方检验的显著性水平(不同于通常的.05),相对拟合指数(如NNFI和CFI)检验的实质是基于虚模型设定均方(卡方与自由度之比)降低到的比例;在NNFI大于临界值后,报告和检验CFI是不必要的。根据研究结果提出了一些方便实用的拟合检验建议。     

Personality and conditioning: a test of four models

Four experiments were conducted to test hypotheses derived from 4 alternative models of individual differences in instrumental conditioning. A standard go--no-go discrimination learning task was used in each of the 4 experiments. The results indicate that individual differences in performance of this discrimination are more consistently and strongly associated with impulsivity and anxiety than with extraversion and neuroticism. In each of the experiments, high anxiety hindered the learning of a go--no-go discrimination more among high impulsive Ss than among low impulsive Ss, and in 2 of the experiments high anxiety actually facilitated learning among low impulsive Ss. These findings are incompatible with Eysenck's and Gray's hypotheses regarding extraversion but are not inconsistent with Newman's. Aspects of these results do support Spence's and Gray's models of anxiety and instrumental conditioning. However, both of these models were contradicted by other trends in the data. A modification of Gray's model of impulsivity and anxiety that emphasizes the role of expectancies was proposed to fit these data.     

Masking misfit in confirmatory factor analysis by increasing unique variances: a cautionary note on the usefulness of cutoff values of fit indices

Fit indices are widely used in order to test the model fit for structural equation models. In a highly influential study, Hu and Bentler (1999) showed that certain cutoff values for these indices could be derived, which, over time, has led to the reification of these suggested thresholds as "golden rules" for establishing the fit or other aspects of structural equation models. The current study shows how differences in unique variances influence the value of the global chi-square model test and the most commonly used fit indices: Root-mean-square error of approximation, standardized root-mean-square residual, and the comparative fit index. Using data simulation, the authors illustrate how the value of the chi-square test, the root-mean-square error of approximation, and the standardized root-mean-square residual are decreased when unique variances are increased although model misspecification is present. For a broader understanding of the phenomenon, the authors used different sample sizes, number of observed variables per factor, and types of misspecification. A theoretical explanation is provided, and implications for the application of structural equation modeling are discussed.     

A Bayesian prediction of four-look recognition performance from one-look data

The hypothesis that a human O’s (S’s) performance in a visual recognition task can be modelled by Bayes’ theorem was investigated. Two Ss were run for 40 experimental sessions each. Their task was to specify the direction of the gap of tachistoscopically presented Landolt rings (Cs). There were four possible gap directions, and two experimental conditions. In one condition, S responded after each stimulus presentation. In the other, a fixed-observation condition, Ss responded after four consecutive presentations of a C. Exposure durations were such that performance under both conditions was greater than chance, but less than unity. Predictions of four-look performance from one-look data were made. Overall hit rates were predicted closely. The entire pattern of each S’s four-look data was also predicted reasonably well. Further tests of the model are currently under way.     

Signal recognition as influenced by presentation schedules

Performance in a recognition task involving two amplitudes of the same tone was investigaled over a wide range of presentation schedules. The task was arranged so that there was no trial-to-trial feedback or other information regarding the relative frequencies of the two tones. The hit and false alarm rates (the proportion of “loud” responses to loud and soft stimuli, respectively) on any given trial were strongly influenced by the stimulus and response on the preceding trial. In general, Ss tended to repeat the last response and were more accurate after a stimulus alternation than after a stimulus repetition. In addition, hit and false alarm rates were inversely related to the presentation probability of the loud tone, in contrast to the direct relation typically found in signal detection experiments and in recognition experiments with trial-to-trial feedback. A mathematical model incorporating three processes (memory, comparison, and decision) was shown to give a good account of these data.     

Model Fit and Item Factor Analysis: Overfactoring,Underfactoring, and a Program to Guide Interpretation

In exploratory item factor analysis (IFA), researchers may use model fit statistics and commonly invoked fit thresholds to help determine the dimensionality of an assessment. However, these indices and thresholds may mislead as they were developed in a confirmatory framework for models with continuous, not categorical, indicators. The present study used Monte Carlo simulation methods to investigate the ability of popular model fit statistics (chi-square, root mean square error of approximation, the comparative fit index, and the Tucker–Lewis index) and their standard cutoff values to detect the optimal number of latent dimensions underlying sets of dichotomous items. Models were fit to data generated from three-factor population structures that varied in factor loading magnitude, factor intercorrelation magnitude, number of indicators, and whether cross loadings or minor factors were included. The effectiveness of the thresholds varied across fit statistics, and was conditional on many features of the underlying model. Together, results suggest that conventional fit thresholds offer questionable utility in the context of IFA.     

The non-monotonicity of the psychometric function in recognition memory

The technique of using the signal detectability theory parameter d’ as a measure of trace strength in recognition memory is extended in a decision-making model that predicts that a S’s psychometric function lp(Y/s) plotted against stimulus intensity / will have a peak for intermediate values of trace strength when the S has a bias towards a “Yes” response. An experiment is reported where Ss were required to recognize previously presented nonsense syllables under a number of pay-off conditions designed to give different groups of Ss different response biases. The predictions of the model were confirmed. It is suggested that a S’s discriminability and response bias parameters may not be as stable as previous studies, that pool large samples of data, have indicated, and the results also provide support for a two-stage model of recognition, where a preliminary estimate of the trace strength of an item is obtained before the location of the response criterion is decided.