General Recognition Theory Research Articles

Face shape and motion are perceptually separable: Support for a revised model of face processing.

A recent model of face processing proposes that face shape and motion are processed in parallel brain pathways. Although tested in neuroimaging, the assumptions of this theory remain relatively untested through controlled psychophysical studies until now. Recruiting undergraduate students over the age of 18, we test this hypothesis using a tight control of stimulus factors, through computerized three-dimensional face models and calibration of dimensional discriminability, and of decisional factors, through a model-based analysis using general recognition theory (GRT). Theoretical links between neural and perceptual forms of independence within GRT allowed us to derive the a priori hypotheses that perceptual separability of shape and motion should hold, while other forms of independence defined within GRT might fail. We found evidence to support both of those predictions.

Sensitivity vs. awareness curve: A novel model-based analysis to uncover the processes underlying nonconscious perception.

In this study, we present a novel model-based analysis of the association between awareness and perceptual processing based on a multidimensional version of signal detection theory (general recognition theory, or GRT). The analysis fits a GRT model to behavioral data and uses the estimated model to construct a sensitivity versus awareness (SvA) curve, representing sensitivity in the discrimination task at each value of relative likelihood of awareness. This approach treats awareness as a continuum rather than a dichotomy, but also provides an objective benchmark for low likelihood of awareness. In two experiments, we assessed nonconscious facial expression recognition using SvA curves in a condition in which faces (fearful vs. neutral) were rendered invisible using continuous flash suppression (CFS) for 500 and 700 milliseconds. We predicted and found nonconscious processing of face emotion, in the form of higher than chance-level sensitivity in the area of low likelihood of awareness.

Don't be a Square: The processing mechanisms characterising the elemental dimensions of width and height.

What are the geometric and information processing characteristics of elementary figures composed of simple physical dimensions? There have been a number of investigations of perception of rectangles, including debate about configurality (e.g., integrality and gestalt properties) as well as the prime perceptual dimensions. Yet, because of ambiguity even in the "right" definition of configurality and an absence of penetrating methodologies, there is still little known concerning the information processing of these patterns. To this end, the present study brings together two separate theory-driven methodologies, general recognition theory (GRT) and systems factorial technology (SFT). The first attacks the problem of dimensional interactions while the latter seeks to uncover process characteristics such as architecture, decisional stopping rules, and workload capacity. The same observers and as much as possible, the same stimuli were used in both approaches. Through our GRT analyses, we found strong evidence for dependencies between the percepts of height and width on both within-stimulus and cross-stimulus bases. Height perception was better with narrow widths and width perception was superior with short heights. In addition, a significant positive within-trial correlation of dimensions was evidenced within squares but not with rectangles. Our SFT initiative uncovered consistent signatures of parallelism paired with super capacity, the latter appearing both through the traditional conditioning on being correct and still present when modest speed accuracy trade-off was accounted for. Thus, the SFT and GRT inferences were quite compatible with a plausible cause of the positive correlations being across-channel facilitatory interactions which led to super capacity processing.

A general recognition theory model for identifying an ideal stimulus

A probabilistic, multidimensional model is described that accounts for sensory and hedonic ratings that are collected from the same experiment. The model combines a general recognition theory model of the sensory ratings with Coombs’ unfolding model of the hedonic ratings. The model uses sensory ratings to build a probabilistic, multidimensional representation of the sensory experiences elicited by exposure to each stimulus, and it also builds a similar representation of the hypothetical ideal stimulus in this same space. It accounts for hedonic ratings by measuring differences between the presented stimulus and the imagined ideal on each rated sensory dimension. Therefore, it provides precise estimates of the sensory qualities of the ideal on all rated sensory dimensions. The model is tested successfully against data from a new experiment.

Effects of shifts in response preferences on characteristics of representation and real-time processing: An application to the Hering illusion.

The multidimensional generalization of signal detection theory known as General Recognition Theory (GRT, Ashby & Townsend, Psychological Review, 93, 154-179 1986) has been used to model and characterize the ways in which changes in encoded perceptual information and the application of decisional operators can produce patterns in behavior that are consistent with notions such as configural processing and representation. In particular, a set of studies (e.g., Cornes et al.,, 2011; Wenger & Ingvalson; 2002, 2003) have shown how decisional influences might exert themselves in situations of configural perception, such that effects such as the Thatcher illusion can be obtained by way of shifts in decisional criteria. The present investigation brought to bear the combined tools of GRT and Systems Factorial Technology (SFT, Little et al.,, 2017; Townsend & Nozawa, 1995) with a classical illusion, to show that it is possible to induce a shift in a decisional criterion by way of varying payoffs and that this shift is accompanied by regular changes in the workload capacity statistic. The combined sets of analyses on the same stimuli reveal orderly effects on the decisional criteria (i.e., the signal detection parameter c), report independence (suggesting perceptual independence), invariant measures of sensitivity (i.e., the signal detection parameter [Formula: see text]),and exhaustive parallel processing accompanied by super capacity. We therefore propose wider use of the combined sets of tools, further exploration of the ability of decisional alterations to affect processing times while leaving accuracy largely unscathed, and reaching out to explore more of the information processing mechanisms of classical illusions.

Breast Tissue Density Influences Tumor Malignancy Perception and Decisions in Mammography

High density breast tissue has been found to reduce radiologists’ accuracy in detecting and classifying mammogram abnormalities. The current research examines the perceptual and decisional components that underlie diagnostic classification (independent of detection) in a sample of novices. Mammograms were varied along two dimensions: Breast tissue density (low/high) and the nature of an identified mass (benign/malignant). In two experiments, participants learned to classify images into 4 categories created by factorial combination of these dimensions. In low density tissue, accuracy was higher for benign than for malignant masses. Surprisingly, and in contrast to the mammography literature, accuracy for malignant masses was higher in high-density than in low-density tissue. Cognitive modeling based in general recognition theory (GRT) indicated that low-density/benign category accuracy was largely due to high perceptual discriminability for that category. The high accuracy level for malignant masses in high density tissue was accounted for by decision bound slopes that favored the “malignant” response for items in the high-density malignant category. Because GRT can provide insight into perceptual and decisional determinants of observed classification accuracy, as well as individual-difference level parameters about attention allocation, GRT may be a means to obtain a more detailed understanding of diagnostic classification of complex naturalistic stimuli.

Stereotypical Processing of Emotional Faces: Perceptual and Decisional Components.

People tend to associate anger with male faces and happiness or surprise with female faces. This angry-men-happy-women bias has been ascribed to either top-down (e.g., well-learned stereotypes) or bottom-up (e.g., shared morphological cues) processes. The dissociation between these two theoretical alternatives has proved challenging. The current effort addresses this challenge by harnessing two complementary metatheoretical approaches to dimensional interaction: Garner's logic of inferring informational structure and General Recognition Theory—a multidimensional extension of signal detection theory. Conjoint application of these two rigorous methodologies afforded us to: (a) uncover the internal representations that generate the angry-men-happy-women phenomenon, (b) disentangle varieties of perceptual (bottom-up) and decisional (top-down) sources of interaction, and (c) relate operational and theoretical meanings of dimensional independence. The results show that the dimensional interaction between emotion and gender is generated by varieties of perceptual and decisional biases. These outcomes document the involvement of both bottom-up (e.g., shared morphological structures) and top-down (stereotypes) factors in social perception.

Training away face-type bias: perception and decisions about emotional expression in stereotypically Black faces.

Prior research indicates that stereotypical Black faces (e.g., wide nose,full lips) are perceived negatively relative to non-stereotypical faces (face-type bias). The current study investigated whether stereotypical faces may bias the interpretation of a neutral facial expression to seem threatening. Moreover, could biased responses be trained away with feedback? In two experiments, stimuli (face images) were presented in a speeded identification task that included corrective feedback, and participants indicated whether the face stimuli were stereotypical or not and threatening or not. Stimuli were pre-rated by face-type (stereotypical, non-stereotypical) and expression (neutral, threatening). Computational modeling based on General Recognition Theory indicated that training increased perceptual discriminability between all the faces. By the end of training (in both experiments), discriminability for emotional expression was slightly higher for stereotypical faces. Model parameters (for both experiments) also showed that, early in training, decision boundaries were more biased toward the threatening response for stereotypical faces relative to non-stereotypical faces. The results suggest that decision bias may be malleable with training.

Linking general recognition theory and classification images to study invariance and configurality of visual representations

Linking general recognition theory and classification images to study invariance and configurality of visual representations

Linking the diffusion model and general recognition theory: Circular diffusion with bivariate-normally distributed drift rates

The circular diffusion model is a model of continuous outcome decisions, which are modeled as evidence accumulation by a two-dimensional Wiener diffusion process on the interior of a disk whose bounding circle represents the decision criterion. When there is across-trial variability in the evidence entering the decision process, represented by variability in drift rates, the model predicts that inaccurate responses will be slower than accurate responses, in agreement with, and generalizing, the slow-error property of the one-dimensional diffusion model of two-choice decisions. A natural generalization of the one-dimensional model’s assumption of normally distributed drift rates is provided by general recognition theory, which represents the perceptual effects of a two-dimensional stimulus as a bivariate distribution with possibly correlated components. Analytic predictions are derived for the joint distributions of decision outcomes and decision times for the circular diffusion model with bivariate-normally distributed drift rates with correlated components. The decision time and accuracy predictions of the model are shown to depend on the relationship between the orientation of the ellipses of equal likelihood of the drift-rate distribution and the orientation of the mean drift rate vector. The analytic expressions for the model can be computed efficiently and are well suited to fitting data.

The psychological interaction of spam email features

This study explored distinct perceptual and decisional contributions to spam email mental construal. Participants classified spam emails according to pairings of three stimulus features – presence or absence of awkward prose, abnormal message structure, and implausible premise. We examined dimensional interactions within general recognition theory (GRT; a multidimensional extension of signal detection theory). Classification accuracy was highest for categories containing either two non-normal dimension levels (e.g. awkward prose and implausible premise) or two normal dimension levels (e.g. normal prose and plausible premise). Modelling indicated both perceptual and decisional contributions to classification responding. In most cases, perceptual discriminability was higher along one dimension when stimuli contained a non-normal level of the paired dimension (e.g. prose discriminability was higher with abnormal structure). Similarly, decision criteria along one dimension were biased in favour of the non-normal response when stimuli contained a non-normal level of the paired dimension. Potential applications for training are discussed.Practitioner summary: We applied general recognition theory (i.e. multivariate signal detection theory) to spam email classification at low or high levels of three stimulus dimensions: premise plausibility, prose quality, and email structure. Relevant to training, this approach helped identify perceptual and decisional biases that could be leveraged to individualise training.

Multidimensional signal detection theory: Theoretical issues and empirical applications

Signal detection theory (SDT) provides a powerful set of tools for modeling choice data in a wide variety of domains. Fundamental to SDT is the distinction between noisy evidence and deterministic response selection. Multidimensional signal detection theory (MDSDT, aka general recognition theory) extends the basic concepts of SDT so that they are applicable to more complex tasks and data. Building on the SDT foundation of noisy evidence and deterministic response selection, MDSDT adds multiple ways in which dimensions may (or may not) interact. The additional complexity of multiple dimensions both broadens the range of applications of these tools and presents unique challenges. I will discuss the origins of MDSDT, a number of empirical applications of MDSDT in speech and hearing sciences, and some recent theoretical developments, focusing on model mimicry and identifiability and multilevel extensions of the basic model.

Detecting mean shift integrality using the Hering illusion: initial results using general recognition theory and systems factorial theory

Detecting mean shift integrality using the Hering illusion: initial results using general recognition theory and systems factorial theory

Modeling talker- and listener-based sources of variability in babble-induced consonant confusions.

Speech communication often occurs in the presence of noise. Patterns of perceptual errors induced by background noise are influenced by properties of the listener and of the noise and target speech. The present study introduces a modification of multilevel general recognition theory in which talker- and listener-based variability in confusion patterns are modeled as global or dimension-specific scaling of shared, group-level perceptual distributions. Listener-specific perceptual correlations and response bias are also modeled as random variables. This model is applied to identification-confusion data from 11 listeners' identifications of ten tokens of each of four consonant categories-[t], [d], [s], [z]-produced by 20 talkers in CV syllables and masked by 10-talker babble. The results indicate that dimension-specific scaling for both listeners and talkers provides a good account of confusion patterns. These findings are discussed in relation to other recent research showing substantial listener-, talker-, and token-based sources of variability in noise-masked speech perception.

When a face type is perceived as threatening: Using general recognition theory to understand biased categorization of Afrocentric faces.

Prior research indicates that stereotypical Black faces (e.g., wide nose, full lips: Afrocentric) are often associated with crime and violence. The current study investigated whether stereotypical faces may bias the interpretation of facial expression to seem threatening. Stimuli were prerated by face type (stereotypical, nonstereotypical) and expression (neutral, threatening). Later in a forced-choice task, different participants categorized face stimuli as stereotypical or not and threatening or not. Regardless of prerated expression, stereotypical faces were judged as more threatening than were nonstereotypical faces. These findings were supported using computational models based on general recognition theory (GRT), indicating that decision boundaries were more biased toward the threatening response for stereotypical faces than for nonstereotypical faces. GRT analysis also indicated that perception of face stereotypicality and emotional expression are dependent, both across categories and within individual categories. Higher perceived stereotypicality predicts higher perception of threat, and, conversely, higher ratings of threat predict higher perception of stereotypicality. Implications for racial face-type bias influencing perception and decision-making in a variety of social and professional contexts are discussed.

Converging operations and the role of perceptual and decisional influences on the perception of faces: Neural and behavioral evidence

Theoretical analyses suggest that the regularities indicative of holistic processing can be obtained by combinations of perceptual and decisional factors. Kuefner and colleagues used electrophysiological results to suggest that the composite face effect is driven solely by perceptual factors. Two limitations of their approach are (a) it did not involve behavioral measures of perceptual sensitivity or bias, and (b) it is unclear how the measures used in that study are consistent with other measures of perceptual and decisional processing. Eight observers completed three tasks involving the stimuli used by Kuefner et al.. The first was a direct replication. The second was a complete identification task, associated with the perceptual and decisional distinctions formalized in general recognition theory. The third was an implementation of the Eriksen fianker task, which allows for a pattern of results that have been interpreted in terms of perceptual and decisional influences. While the empirical distinctions used by Kuefner et al. were not consistent with either the EEG data from the other tasks or the established behavioral measures of perceptual sensitivity and decisional bias, the inferences drawn from the EEG and behavioral data from those tasks were consistent with one another, underscoring the importance of converging operations.

Analytic processing of distance

How does a human observer extract from the distance between two frontal points the component corresponding to an axis of a rectangular reference frame? To find out we had participants classify pairs of small circles, varying on the horizontal and vertical axes of a computer screen, in terms of the horizontal distance between them. A response signal controlled response time. The error rate depended on the irrelevant vertical as well as the relevant horizontal distance between the test circles with the relevant distance effect being larger than the irrelevant distance effect. The results implied that the horizontal distance between the test circles was imperfectly extracted from the overall distance between them. The results supported an account, derived from the Exemplar Based Random Walk model (Nosofsky & Palmieri, 1997), under which distance classification is based on the overall distance between the test circles, with relevant distance being extracted from overall distance to the extent that the relevant and irrelevant axes are differentially weighted so as to reduce the contribution of irrelevant distance to overall distance. The results did not support an account, derived from the General Recognition Theory (Ashby & Maddox, 1994), under which distance classification is based on the relevant distance between the test circles, with the irrelevant distance effect arising because a test circle's perceived location on the relevant axis depends on its location on the irrelevant axis, and with relevant distance being extracted from overall distance to the extent that this dependency is absent.

A Model-Based Approach to the Wisdom of the Crowd in Category Learning.

We apply the "wisdom of the crowd" idea to human category learning, using a simple approach that combines people's categorization decisions by taking the majority decision. We first show that the aggregated crowd category learning behavior found by this method performs well, learning categories more quickly than most or all individuals for 28 previously collected datasets. We then extend the approach so that it does not require people to categorize every stimulus. We do this using a model-based method that predicts the categorization behavior people would produce for new stimuli, based on their behavior with observed stimuli, and uses the majority of these predicted decisions. We demonstrate and evaluate the model-based approach in two case studies. In the first, we use the general recognition theory decision-bound model of categorization (Ashby & Townsend, ) to infer each person's decision boundary for two categories of perceptual stimuli, and we use these inferences to make aggregated predictions about new stimuli. In the second, we use the generalized context model exemplar model of categorization (Nosofsky, ) to infer each person's selective attention for face stimuli, and we use these inferences to make aggregated predictions about withheld stimuli. In both case studies, we show that our method successfully predicts the category of unobserved stimuli, and we emphasize that the aggregated crowd decisions arise from psychologically interpretable processes and parameters. We conclude by discussing extensions and potential real-world applications of the approach.

Feature binding in visual short term memory: A General Recognition Theory analysis.

Creating and maintaining accurate bindings of elementary features (e.g., color and shape) in visual short-term memory (VSTM) is fundamental for veridical perception. How are low-level features bound in memory? The present work harnessed a multivariate model of perception - the General Recognition Theory (GRT) - to unravel the internal representations underlying feature binding in VSTM. On each trial, preview and target colored shapes were presented in succession, appearing in either repeated or altered spatial locations. Participants gave two same/different responses: one with respect to color and one with respect to shape. Converging GRT analyses on the accuracy confusion matrices provided substantial evidence for binding in the form of violations of perceptual independence at the level of the individual stimulus, such that positive correlations were obtained when both features repeated or alternated together, while negative correlations were obtained when one feature repeated and the other alternated. This "cloverleaf" GRT pattern of binding was similar whether the spatial location of the preview and target repeated or altered. The current results are consistent with: (a) the discrete memory "slots" model of VSTM, and (b) the notion that spatial location is not necessary for the formation of "object files." The GRT approach presented here offers a viable quantitative model for testing various questions regarding feature binding in VSTM.

Testing Separability and Independence of Perceptual Dimensions with General Recognition Theory: A Tutorial and New R Package (grtools).

Determining whether perceptual properties are processed independently is an important goal in perceptual science, and tools to test independence should be widely available to experimental researchers. The best analytical tools to test for perceptual independence are provided by General Recognition Theory (GRT), a multidimensional extension of signal detection theory. Unfortunately, there is currently a lack of software implementing GRT analyses that is ready-to-use by experimental psychologists and neuroscientists with little training in computational modeling. This paper presents grtools, an R package developed with the explicit aim of providing experimentalists with the ability to perform full GRT analyses using only a couple of command lines. We describe the software and provide a practical tutorial on how to perform each of the analyses available in grtools. We also provide advice to researchers on best practices for experimental design and interpretation of results when applying GRT and grtools