Analysis of group differences in processing speed: Where are the models of processing?

Recently, Myerson, Adams, Hale, and Jenkins (2003) replied to arguments advanced by Ratcliff, Spieler, and McKoon (2000) about interpretations of Brinley functions. Myerson et al. (2003) focused on methodological and terminological issues, arguing that (1) Brinley functions are not quantile-quantile (QQ) plots of distributions of mean reaction times (RTs) across conditions; that the fact that the slope of a Brinley function is the ratio of the standard deviations of the two distributions of means has no implications for the use of slope as a measure of processing speed; that the ratio of slopes of RT functions for older and young subjects plotted against independent variables equals the Brinley function slope; and that speed-accuracy criterion effects do not account for slowing with age. We reply by showing that Brinley functions are plots of quantiles against quantiles; that the slope is best estimated by the ratio of standard deviations because there is variability in the distributions of mean RTs for both older and young subjects; that the interpretation of equality of the slopes Brinley functions and plots of RTs against independent variables in terms of processing speed is model dependent; and that speed-accuracy effects in some, but not all, experiments are solely responsible for Brinley slopes greater than 1. We conclude by reiterating the point that was not addressed in Myerson et al. (2003), that the goal of research should be modelbased accounts of processing that deal with correct and error RT distributions and accuracy.     

Psychometric and information processing properties of selected response time models

This paper discusses the compatibility of some response time (RT) models with psychometric and with information processing approaches to response times. First, three psychometrically desirable properties of probabilistic models for binary data, related to the principle of specific objectivity, are adapted to the domain of RT models. One of these is the separability of item and subject parameters, and another is double monotonicity. Next, the compatibility of these psychometric properties with one very popular information processing approach, the serial-additive model, is discussed. Finally, five RT models are analyzed with respect to their compatibility with the psychometric properties, with serial-additive processing and with some alternative types of processing. It is concluded that (a) current psychometric models each satisfy one or more of the psychometric properties, but are not (easily) compatible with serial-additive processing, (b) at least one serial-additive processing model satisfies separability of item and subject parameters, and (c) RT models will more easily satisfy double monotonicity than the other two psychometric properties.     

Face gender recognition in developmental prosopagnosia: Evidence for holistic processing and use of configural information

Face identification deficits in developmental prosopagnosics (DPs) have been thought to be due to general difficulties with processing configural face information and integrating configural and parts information into a coherent whole (holistic processing). Gender recognition provides a further opportunity to more fully examine this issue as this ability may be intact in DPs and it has been shown to depend on processing configural information and holistic processing in neurotypical individuals. In the present study we first determined that, indeed, gender discrimination performance was similar in DPs and controls. Second, we found that inversion and scrambling (which we propose measures holistic processing and sensitivity to configural information, respectively) produced comparable deficits in DPs and controls, suggesting that both groups use holistic processing and configural information to recognize gender. This indicates that holistic processing and using configural face information are not general impairments in DP and may be more specific to face identity.     

A reach-to-touch investigation on the nature of reading in the Stroop task

In a Stroop task, participants can be presented with a color name printed in color and need to classify the print color while ignoring the word. The Stroop effect is typically calculated as the difference in mean response time (RT) between congruent (e.g., the word RED printed in red) and incongruent (GREEN in red) trials. Delta plots compare not just mean performance, but the entire RT distributions of congruent and incongruent conditions. However, both mean RT and delta plots have some limitations. Arm-reaching trajectories allow a more continuous measure for assessing the time course of the Stroop effect. We compared arm movements to congruent and incongruent stimuli in a standard Stroop task and a control task that encourages processing of each and every word. The Stroop effect emerged over time in the control task, but not in the standard Stroop, suggesting words may be processed differently in the two tasks.     

Testing the assumptions of exponential,additive reaction time models

Two assumptions commonly made by choice reaction time (RT) models are (1) that certain experimental tasks can be found that cause an extra processing stage to be inserted into the cognitive process and (2) that the duration of one or more processing stages is random with an exponential distribution. Few rigorous tests of either assumption exist. This paper reviews existing tests and presents several new results that can be used to test these assumptions. First, in the case in which the duration of an inserted stage is exponentially distributed, it is shown that the observable RT density functions must always intersect at the mode of the density requiring the extra processing stage. Second, when only the first assumption (Assumption 1) is made, it is shown that the cumulative RT distribution functions and, in many cases, the hazard functions must be ordered. Finally, when only Assumption 2 is assumed, it is shown that, under fairly weak conditions, the taft of the RT density function must be exponential. The first two results are applied to data from a memory scanning experiment, and the two assumptions are found to receive tentative support     

CDF-XL: computing cumulative distribution functions of reaction time data in Excel

In experimental psychology, central tendencies of reaction time (RT) distributions are used to compare different experimental conditions. This emphasis on the central tendency ignores additional information that may be derived from the RT distribution itself. One method for analysing RT distributions is to construct cumulative distribution frequency plots (CDFs; Ratcliff, Psychological Bulletin 86:446–461, 1979). However, this method is difficult to implement in widely available software, severely restricting its use. In this report, we present an Excel-based program, CDF-XL, for constructing and analysing CDFs, with the aim of making such techniques more readily accessible to researchers, including students (CDF-XL can be downloaded free of charge from the Psychonomic Society’s online archive). CDF-XL functions as an Excel workbook and starts from the raw experimental data, organised into three columns (Subject, Condition, and RT) on an Input Data worksheet (a point-and-click utility is provided for achieving this format from a broader data set). No further preprocessing or sorting of the data is required. With one click of a button, CDF-XL will generate two forms of cumulative analysis: (1) “standard” CDFs, based on percentiles of participant RT distributions (by condition), and (2) a related analysis employing the participant means of rank-ordered RT bins. Both analyses involve partitioning the data in similar ways, but the first uses a “median”-type measure at the participant level, while the latter uses the mean. The results are presented in three formats: (i) by participants, suitable for entry into further statistical analysis; (ii) grand means by condition; and (iii) completed CDF plots in Excel charts.     

The Ornstein-Uhlenbeck model for decision time in cognitive tasks: An example of control of nonlinear network dynamics

The Ornstein-Uhlenbeck (OU) model for human decision-making has been successfully applied to account for response accuracy and response time (RT) data in recent two-choice decision models. A variant of the OU model is shown to arise from the response dynamics of a nonlinear network consisting of randomly connected neural processing units. When feedback control of the network is effected by the stimulus onset, the average network response is an autocorrelated random signal satisfying the stochastic differential equation for the OU process. An alternative, more general, stimulus detection procedure is proposed which involves the use of an adaptive Kalman filter process to track any temporal change in autoregressive parameters. The predicted decision time distributions suggest that both the OU and the Kalman filter processes can serve as alternative models for RT data in experimental tasks. Received: 15 October 1998 / Accepted: 27 May 1999     

Common response processing and psychological refractoriness

This experiment examined whether structural considerations, that is the functional closeness of limb control centres (Kinsbourne and Hicks 1978), have a role to play in common processing models of psychological refractoriness. A typical double stimulation paradigm was used (Kantowitz 1974). On experimental trials, subjects responded with the right hand (RT1) to an auditory signal (0.80 probability). After a varied interstimulus interval (ISI) of 100, 200, 300 or 700 msec, one of three lights corresponding to one of the other limbs provided the signal for the second response (RT2). Control trials for RT1 and RT2 involved both stimuli, but subjects were required to make only one response. Results indicated that experimental RT1 was lengthened considerably over control values. While typical double stimulation effects were found at short ISIs, that is RT2 was lengthened, RT2 at long ISIs was actually faster than double stimulation control values. Although the RT results were not in a pattern that would be predicted by functional distance (Kinsbourne and Hicks 1978), within subject correlations of RT1 and RT2 by limb condition suggest that structural considerations may have a role to play. While within subject correlations were positive at all levels of ISI, they decreased as the interval between the two signals increased. The implications of these results for common processing models are discussed.     

Hierarchical Bayes Models for Response Time Data

Human response time (RT) data are widely used in experimental psychology to evaluate theories of mental processing. Typically, the data constitute the times taken by a subject to react to a succession of stimuli under varying experimental conditions. Because of the sequential nature of the experiments there are trends (due to learning, fatigue, fluctuations in attentional state, etc.) and serial dependencies in the data. The data also exhibit extreme observations that can be attributed to lapses, intrusions from outside the experiment, and errors occurring during the experiment. Any adequate analysis should account for these features and quantify them accurately. Recognizing that Bayesian hierarchical models are an excellent modeling tool, we focus on the elaboration of a realistic likelihood for the data and on a careful assessment of the quality of fit that it provides. We judge quality of fit in terms of the predictive performance of the model. We demonstrate how simple Bayesian hierarchical models can be built for several RT sequences, differentiating between subject-specific and condition-specific effects.     

Retrieval processes in recall

A reaction time (RT) paradigm was developed to study retrieval processes in paired associate (PA) recall. Prior to the experimental session, subjects learned lists of PAs (varying in length from three to 24 pairs); during the experimental session, subjects' RT to say the response word from the onset of a visually presented stimulus word was measured. The implications of several classes of retrieval models were discussed in the context of this paradigm. The shape of the RT-list length function, practice, and sequential effects were all of interest in distinguishing among models. Four experiments were reported which were designed to (1) establish the baseline effects in this paradigm, (2) determine which of these effects should be attributed to the retrieval stage of processing, and (3) investigate the effect of semantic memory in this task. Results suggest that simple serial scanning models are inadequate to handle the data from this task. Strength, direct-access, or parallel processing models, on the other hand, seem to capture the qualitative effects present in our experiments. When a strength model was formalized and fit to the data from Experiment I, it was found that a two-trace version gave good quantitative fits while a one-trace version did not, suggesting that both short- and long-term memory independently contribute in this task.     

Inferential bridging relations reveal distinct neural mechanisms: evidence from event-related brain potentials

This study investigates the online comprehension of Determiner Phrases (DPs) as a function of the given-new distinction in two-sentence texts in German and further focuses on DPs whose interpretation depends on inferential information (so-called 'bridging relations'). Previous reaction time studies report an advantage of given over new information. In the present study, this difference is reflected in distinct neural mechanisms: event-related potentials reveal that previously introduced (i.e., given) DPs elicit a reduced N400, while new DPs show an enhanced N400 followed by a P600. Crucially, inferentially bridged DPs, which are hypothesized to share properties with new and given information, first pattern with given DPs (showing an attenuated N400) and then with new DPs (showing an enhanced P600). The data demonstrate that salience relations between DPs and prior context ease DP integration and that additional cost arises from the establishment of independent reference. They further reveal that processing cost associated with the interpretation of bridged DPs results from the anaphoric complexity of introducing an independent referent.     

On the dissociation between reaction time and response duration as a function of lexical and sublexical reading: An examination of phonetic decoding and computational models

Neurobiological models of reading account for two ways in which orthography is converted to phonology: (1) familiar words, particularly those with exceptional spelling-sound mappings (e.g., shoe) access their whole-word lexical representations in the ventral visual stream, and (2) orthographically unfamiliar words, particularly those with regular spelling-sound mappings (i.e., pseudohomophones [PHs], which are orthographically novel but sound like real words; e.g., shue) are phonetically decoded via sublexical processing in the dorsal visual stream. The present study used a naming task in order to compare naming reaction time (RT) and response duration (RD) of exception and regular words to their PH counterparts. We replicated our earlier findings with words, and extended them to PH phonetic decoding by showing a similar effect on RT and RD of matched PHs. Given that the shorter RDs for exception words can be attributed to the benefit of whole-word processing in the orthographic word system, and the longer RTs for exception words to the conflict with phonetic decoding, our PH results demonstrate that phonetic decoding also involves top-down feedback from phonological lexical representations (e.g., activated by shue) to the orthographic representations of the corresponding correct word (e.g., shoe). Two computational models were tested for their ability to account for these effects: the DRC and the CDP+. The CDP+ fared best as it was capable of simulating both the regularity and stimulus type effect on RT for both word and PH identification, although not their over-additive interaction. Our results demonstrate that both lexical reading and phonetic decoding elicit a regularity dissociation between RT and RD that provides important constraints to all models of reading, and that phonetic decoding results in top-down feedback that bolsters the orthographic lexical reading process.     

Changes in alertness and processing capacity in a serial learning task

Second-to-second changes in processing capacity were examined using a concurrent task paradigm involving serial learning and simple reaction time (RT) in two experiments. Experiment 1 demonstrated the feasibility of combining probe RT with a slow-paced serial learning task including perceptually isolated items. Experiment 2 employed the same concurrent tasks to compare cardiac interbeat interval (IBI, the reciprocal of heart rate) with probe RT. Second-to-second changes during item processing were found consistently for both measures. However, these changes appeared to be more related to observing and response requirements than to the specific cognitive processes required by the tasks. Correctly anticipated items were associated with (l)RT changes, suggesting heightened allocation of capacity, and (2)a cardiac IBI response of episodic cardiac deceleration imposed upon the task-induced cardiac speeding. The probe RT and cardiac IBI measures showed reasonable convergence in assessing the allocation of processing capacity. The concept of the accessibility of allocated capacity was introduced in considering instances of the divergence of probe RT and cardiac IBI.     

Working memory and attention in pediatric brain tumor patients treated with and without radiation therapy

Children are at risk for cognitive difficulties following the diagnosis and treatment of a brain tumor. Longitudinal studies have consistently demonstrated declines on measures of intellectual functioning, and recently it has been proposed that specific neurocognitive processes underlie these changes, including working memory, processing speed, and attention. However, a fine-grained examination of the affected neurocognitive processes is required to inform intervention efforts. Radiation therapy (RT) impacts white matter integrity, likely affecting those cognitive processes supported by distributed neural networks. This study examined working memory and attention in children during the early delayed stages of recovery following surgical resection and RT. The participants included 27 children diagnosed with pediatric brain tumor, treated with (n = 12) or without (n = 15) RT, who completed experimental and standardized measures of working memory and attention (n-back and digit span tasks). Children treated with radiation performed less well than those who did not receive radiation on the n-back measure, though performance at the 0-back level was considerably poorer than would be expected for both groups, perhaps suggesting difficulties with more basic processes such as vigilance. Along these lines, marginal differences were noted on digit span forward. The findings are discussed with respect to models of attention and working memory, and the interplay between the two.     

An integrated perspective on the relation between response speed and intelligence

Research in the field of mental chronometry and individual differences has revealed several robust regularities (Jensen, 2006). These include right-skewed response time (RT) distributions, the worst performance rule, correlations with general intelligence (g) that are more pronounced for RT standard deviations (RTSD) than they are for RT means (RTm), an almost perfect linear relation between individual differences in RTSD and RTm, linear Brinley plots, and stronger correlations between g and inspection time (IT) than between g and RTm. Here we show how all these regularities are manifestations of a single underlying relationship, when viewed through the lens of Ratcliff’s diffusion model ( [Ratcliff, 1978] and [Ratcliff et al., 2008] ). The single underlying relationship is between individual differences in general intelligence and individual differences in “drift rate”, which is just the speed of information processing in Ratcliff’s model. We also test and confirm a strong prediction of the diffusion model, namely that the worst performance rule generalizes to phenomena outside of the field of intelligence. Our approach provides an integrative perspective on intelligence findings.