Jennifer J. Richler’s research while affiliated with Vanderbilt University and other places

There is substantial evidence for individual differences in personality and cognitive abilities, but we lack clear intuitions about individual differences in visual abilities. Previous work on this topic has typically compared performance with only 2 categories, each measured with only 1 task. This approach is insufficient for demonstration of domain-general effects. Most previous work has used familiar object categories, for which experience may vary between participants and categories, thereby reducing correlations that would stem from a common factor. In Study 1, we adopted a latent variable approach to test for the first time whether there is a domain-general object recognition ability, o. We assessed whether shared variance between latent factors representing performance for each of 5 novel object categories could be accounted for by a single higher-order factor. On average, 89% of the variance of lower-order factors denoting performance on novel object categories could be accounted for by a higher-order factor, providing strong evidence for o. Moreover, o also accounted for a moderate proportion of variance in tests of familiar object recognition. In Study 2, we assessed whether the strong association across categories in object recognition is due to third-variable influences. We find that o has weak to moderate associations with a host of cognitive, perceptual, and personality constructs and that a clear majority of the variance in and covariance between performance on different categories is independent of fluid intelligence. This work provides the first demonstration of a reliable, specific, and domain-general object recognition ability, and suggest a rich framework for future work in this area.

The Vanderbilt Holistic Processing Test for faces (VHPT-F) is the first standard test designed to measure individual differences in holistic processing. The test measures failures of selective attention to face parts through congruency effects, an operational definition of holistic processing. However, this conception of holistic processing has been challenged by the suggestion that it may tap into the same selective attention or cognitive control mechanisms that yield congruency effects in Stroop and Flanker paradigms. Here, we report data from 130 subjects on the VHPT-F, several versions of Stroop and Flanker tasks, as well as fluid IQ. Results suggested a small degree of shared variance in Stroop and Flanker congruency effects, which did not relate to congruency effects on the VHPT-F. Variability on the VHPT-F was also not correlated with Fluid IQ. In sum, we find no evidence that holistic face processing as measured by congruency in the VHPT-F is accounted for by domain-general control mechanisms.

In tests of object recognition, individual differences typically correlate modestly but nontrivially across familiar categories (e.g. cars, faces, shoes, birds, mushrooms). In theory, these correlations could reflect either global, non-specific mechanisms, such as general intelligence (IQ), or more specific mechanisms. Here, we introduce two separate methods for effectively capturing category-general performance variation, one that uses novel objects and one that uses familiar objects. In each case, we show that category-general performance variance is unrelated to IQ, thereby implicating more specific mechanisms. The first approach examines three newly developed novel object memory tests (NOMTs). We predicted that NOMTs would exhibit more shared, category-general variance than familiar object memory tests (FOMTs) because novel objects, unlike familiar objects, lack category-specific environmental influences (e.g. exposure to car magazines or botany classes). This prediction held, and remarkably, virtually none of the substantial shared variance among NOMTs was explained by IQ. Also, while NOMTs correlated nontrivially with two FOMTs (faces, cars), these correlations were smaller than among NOMTs and no larger than between the face and car tests themselves, suggesting that the category-general variance captured by NOMTs is specific not only relative to IQ, but also, to some degree, relative to both face and car recognition. The second approach averaged performance across multiple FOMTs, which we predicted would increase category-general variance by averaging out category-specific factors. This prediction held, and as with NOMTs, virtually none of the shared variance among FOMTs was explained by IQ. Overall, these results support the existence of object recognition mechanisms that, though category-general, are specific relative to IQ and substantially separable from face and car recognition. They also add sensitive, well-normed NOMTs to the tools available to study object recognition.

The part-whole paradigm was one of the first measures of holistic processing and it has been used to address several topics in face recognition, including its development, other-race effects, and more recently, whether holistic processing is correlated with face recognition ability. However the task was not designed to measure individual differences and it has produced measurements with low reliability. We created a new holistic processing test designed to measure individual differences based on the part-whole paradigm, the Vanderbilt Part Whole Test (VPWT). Measurements in the part and whole conditions were reliable, but, surprisingly, there was no evidence for reliable individual differences in the part-whole index (how well a person can take advantage of a face part presented within a whole face context compared to the part presented without a whole face) because part and whole conditions were strongly correlated. The same result was obtained in a version of the original part-whole task that was modified to increase its reliability. Controlling for object recognition ability, we found that variance in the whole condition does not predict any additional variance in face recognition over what is already predicted by performance in the part condition.

The Vanderbilt Holistic Face Processing Test (VHPT-F) is a new measure of holistic face processing with better psychometric properties relative to prior measures developed for group studies (Richler et al., 2014). In fields where psychologists study individual differences, validation studies are commonplace and the concurrent validity of a new measure is established by comparing it to an older measure with established validity. We follow this approach and test whether the VHPT-F measures the same construct as the composite task, which is group-based measure at the center of the large literature on holistic face processing. In Experiment 1, we found a significant correlation between holistic processing measured in the VHPT-F and the composite task. Although this correlation was small, it was comparable to the correlation between holistic processing measured in the composite task with the same faces, but different target parts (top or bottom), which represents a reasonable upper limit for correlations between the composite task and another measure of holistic processing. These results confirm the validity of the VHPT-F by demonstrating shared variance with another measure of holistic processing based on the same operational definition. These results were replicated in Experiment 2, but only when the demographic profile of our sample matched that of Experiment 1.

Several studies using reliable measures of individual differences in object recognition with several object categories find that car recognition ability shares little to no variance with other object recognition abilities. In fact, car recognition is at least as independent from the recognition of other non-face categories as face recognition. Consequently, using cars to represent non-face objects may be problematic, especially when only one non-face object category is used. While the relative independence of car recognition ability has been replicated in several datasets, this result has only been obtained using learning measures that repeat stimuli across trials. This learning component common to every test in these datasets may moderate correlations between tests. For example, it may be that greater experience with cars leads to faster learning over trials within this category, leading to a dissociation with other categories that are learned more slowly. To test if the independence of car recognition generalizes outside of learning tasks, we created the Vanderbilt Car Memory Test (VCMT), which does not repeat stimuli across trials. The VCMT is modeled after the Vanderbilt Face Memory Test (VFMT). Each trial begins with 2 car images presented for 4 seconds, followed by a 3-AFC. We honed the test through 3 iterations to produce good internal reliability (approximately .85) with an average accuracy of approximately 60% (SD = 12%). This test, and similar tests with other non-face categories, will be useful to evaluate accounts of why car recognition ability is independent from the recognition of most other object categories. Meeting abstract presented at VSS 2016

Face perception is a critical and complex cognitive operation, and it poses unique the cognitive demands. This chapter addresses the question of whether face perception can be viewed as a cognitive phenotype. Evidence from neurophysiological and neuropsychological studies are summarized that reflect specialization of parts of the visual system for face processing, for example, face-tuned neurons in the superior temporal sulcus of non-human primates, and cortical regions associated with prosopagnosia in humans. Data from cognitive neuroscience (especially functional imaging studies) are presented, illustrating that many distinct brain regions show that activity in response to faces though the lateral fusiform gyrus or “fusiform face area” (FFA) shows a particularly robust response. Current interpretations of FFA activity and how it may be functionally parsed out from the activity of the occipital fusiform area and anterior temporal lobe are then laid out, and this raises the tricky question of what differs between the perception faces and non-face objects as may be expressed along these neural centers. Our discussion on face perception as a potential cognitive phenotype suggests that the domain of operation of the face perception neural mechanism is not all or none. We point instead to a more general purpose visual learning system that happens to be critical in face perception and possibly most fully realized in face perception. The processes of face perception are especially illustrative of difficulties that can be inherent in establishing evidence for a domain-specific cognitive phenotype.

The Cambridge Face Memory Test (CFMT, Duchaine & Nakayama, 2006) is a measure of face recognition ability. It was designed to prevent use of salient diagnostic features, problematic in older tests, and is expected to promote holistic processing, a hallmark strategy of face recognition. However, the actual strategy used on the CFMT has not been directly tested. Attempts to correlate CFMT performance with holistic processing have produced conflicting results, but holistic processing measurements are often unreliable. A recent high-powered study with the first test designed for reliable individual differences in the measurement of holistic processing (the Vanderbilt Holistic Face Processing Test, VHPT-F; Richler et al., 2014, reliability ~.6) found no relationship between holistic processing and the CFMT. Unlike previous measures of holistic processing, faces do not repeat across trials in the VHPT-F, so we hypothesize that the correlation between the CFMT and prior holistic processing measures may stem from stimulus repetition. Because stimuli repeat in the CFMT (6 target faces are repeatedly tested), it is possible that the face learning ability measured by the CFMT is most relevant when only a handful of faces are discriminated across trials, reducing the relevant dimensions and promoting a part-based strategy. We created a new test of face matching ability similar to the CFMT but without face repetition. On each trial, 2 faces are studied for 4s, followed by a 3-AFC. We collected data from 100 subjects for two separate forms of the VFMT1.0 (each 48 trials) in an online sample. It revealed good performance on catch trials, a good spread of item difficulties, average performance of 60% (SD = 20%), and reliability of .7 for both versions. In future work, we will use item analyses to refine the test, and explore correlations with CFMT, the Vanderbilt Expertise Test with objects, and the VHPT-F. Meeting abstract presented at VSS 2015.