Psychometric Modeling of response speed and accuracy with mixed and conditional regression

Human performance in cognitive testing and experimental psychology is expressed in terms of response speed and accuracy. Data analysis is often limited to either speed or accuracy, and/or to crude summary measures like mean response time (RT) or the percentage correct responses. This paper proposes the use of mixed regression for the psychometric modeling of response speed and accuracy in testing and experiments. Mixed logistic regression of response accuracy extends logistic item response theory modeling to multidimensional models with covariates and interactions. Mixed linear regression of response time extends mixed ANOVA to unbalanced designs with covariates and heterogeneity of variance. Related to mixed regression is conditional regression, which requires no normality assumption, but is limited to unidimensional models. Mixed and conditional methods are both applied to an experimental study of mental rotation. Univariate and bivariate analyzes show how within-subject correlation between response and RT can be distinguished from between-subject correlation, and how latent traits can be detected, given careful item design or content analysis. It is concluded that both response and RT must be recorded in cognitive testing, and that mixed regression is a versatile method for analyzing test data.I am grateful to Rogier Donders for putting his data at my disposal.     

Distinctive processing underlies skilled memory

Research on skilled memory has focused on organizational processes to the exclusion of item-specific processes, although theories of skilled memory do acknowledge the importance of both kinds of processes. Using the isolation methodology, we presented lists of American football team names to participants who had either a high or a low level of knowledge about American football. An isolation effect (greater recall of the target in the isolate list than in a homogeneous control list) was observed only with high-knowledge participants. When standard lists were used, an isolation effect was observed with both groups. These findings empirically validate the importance of both organizational and item-specific processing as the basis of distinctive processing underlying skilled memory performance.     

Human cognition and 1/f scaling

Ubiquitous 1/f scaling in human cognition and physiology suggests a mind-body interaction that contradicts commonly held assumptions. The intrinsic dynamics of psychological phenomena are interaction dominant (rather than component dominant), and the origin of purposive behavior lies with a general principle of self-organization (rather than a special neurocognitive mechanism). E.-J. Wagenmakers, S. Farrell, and R. Ratcliff (2005) raised concerns about the kinds of data and analyses that support generic 1/f scaling. This reply is a defense that furthermore questions the model that Wagenmakers and colleagues endorse and their strategy for addressing complexity.     

Individual pattern representations are context independent, but their collective representation is context dependent.

We studied context dependency of the representations underlying perceptual "goodness". Three experiments used a same-different task with classical Garner 5-dot patterns presented with an interstimulus interval (ISI) of 500 ms. Same patterns were allowed to be rotated or reflected versions of each other. Pattern goodness was varied according to rotation and reflection equivalence, using Garner's equivalence set size (ESS) measure. The ESS of both first and second patterns affected reaction time and accuracy. A model based on assumptions that Garner's equivalence sets constitute the generic representation of these patterns and that items within these sets are accessed serially was fitted to the data. Excellent fits were obtained, which were robust against frequency-induced bias at the level of the individual pattern, but sensitive to such bias at the level of the equivalence set. It was concluded that individual pattern representations are context independent, whereas their collective representations are context dependent. Simplicity and likelihood principles, therefore, seem to apply to different levels of a representation hierarchy.     

Compatibility effects based on stimulus and response numerosity

Four choice reaction time experiments documented a stimulus-response (S-R) compatibility effect involving the numbers of stimuli and responses. In Experiment 1, the stimulus consisted of one or two tones, and the correct response was either one or two taps of a response key. Responses were much faster with a compatible S-R assignment, in which the number of taps matched the number of tones, than with an incompatible assignment in which these numbers mismatched. Experiments 2 and 3 replicated this effect, using visual stimuli and bimodal stimuli, respectively, suggesting that auditory/manual rhythmic compatibility is not essential to it. Experiment 4 showed that an analogous but smaller effect is obtained when stimuli are the digits 1 and 2. This new numerosity-based compatibility effect has general theoretical implications regarding the mechanisms responsible for compatibility effects and practical implications for interface design.     

From interpretation to segmentation

In visual perception, part segmentation of an object is considered to be guided by image-based properties, such as occurrences of deep concavities in the outer contour. However, object-based properties can also provide information regarding segmentation. In this study, outer contours and interpretations of object configurations were manipulated to examine differences between image-based and object-based segmentation in a visual search task. We found that locating a two-dimensional object configuration with deep concavities in the outer contour depends on the type of outer contour of the surrounding distractors. In addition, locating a three-dimensional object configuration was harder when it was surrounded by object-based-disconnected distractors, as compared with object-based-connected distractors, regardless of image-based connections in these distractors. We conclude that segmentation based on the outer contours of a target facilitates its localization. However, when three-dimensional information is available, segmentation strongly depends on object-based properties, rather than on image-based properties.     

APP23 mice as a model of Alzheimer's disease: an example of a transgenic approach to modeling a CNS disorder

Animal models are considered essential in research ensuing elucidation of human disease processes and subsequently, testing of potential therapeutic strategies. This is especially true for neurodegenerative disorders, in which the first steps in pathogenesis are often not accessible in human patients. Alzheimer's disease is vastly becoming a major medical and socioeconomic problem in our aging society. Valid animal models for this uniquely human condition should exhibit histopathological, biochemical, cognitive, and behavioral alterations observed in Alzheimer's disease patients. Major progress has been made since the understanding of the genetic basis of Alzheimer's disease and the development and improvement of transgenic mouse models. All present Alzheimer's disease models developed are partial but nevertheless essential in further unraveling the nature and spatial and temporal development of the complex molecular pathology underlying this condition. One of the more recent transgenic attempts to model Alzheimer's disease is the APP23 transgenic mouse. This article describes the development and assessment of this human amyloid precursor protein overexpression model. We summarize histopathological and biochemical, cognitive and behavioral observations made in heterozygous APP23 mice, thereby emphasizing the model's contribution to clarification of neurodegenerative disease mechanisms. In addition, the first therapeutic interventions in the APP23 model are included.     

Inferring the structure of latent class

Present optimization techniques in latent class analysis apply the expectation maximization algorithm or the Newton-Raphson algorithm for optimizing the parameter values of a prespecified model. These techniques can be used to find maximum likelihood estimates of the parameters, given the specified structure of the model, which is defined by the number of classes and, possibly, fixation and equality constraints. The model structure is usually chosen on theoretical grounds. A large variety of structurally different latent class models can be compared using goodness-of-fit indices of the chi-square family, Akaike’s information criterion, the Bayesian information criterion, and various other statistics. However, finding the optimal structure for a given goodness-of-fit index often requires a lengthy search in which all kinds of model structures are tested. Moreover, solutions may depend on the choice of initial values for the parameters. This article presents a new method by which one can simultaneously infer the model structure from the data and optimize the parameter values. The method consists of a genetic algorithm in which any goodness-of-fit index can be used as a fitness criterion. In a number of test cases in which data sets from the literature were used, it is shown that this method provides models that fit equally well as or better than the models suggested in the original articles.     

A blank look in reading: the effect of blank space on the identification of letters and words during reading

Two experiments investigated the effect of visual blank space on reading by varying the amount of interletter and interline blank space in prose passages. Increasing interletter blank spacing slowed the reading process overall, presumably because it disrupted the unitization of words and word identification, but it also improved the identification of the letters within words. By contrast, increasing interline blank spacing sped up the reading process overall, while also improving the identification of words and the letters within words, presumably because the extra spacing reduced the amount of visual information that was processed during reading. The latter finding supports the conclusion that information from surrounding lines of single-spaced text may interfere with reading.