Thomas J. Palmeri’s research while affiliated with Vanderbilt University and other places

Evidence accumulation is a computational framework that accounts for behavior as well as the dynamics of individual neurons involved in decision making. Linking these two levels of description reveals a scaling paradox: How do choices and response times (RT) explained by models assuming single accumulators arise from a large ensemble of idiosyncratic accumulator neurons? We created a simulation model that makes decisions by aggregating across ensembles of accumulators, thereby instantiating the essential structure of neural ensembles that make decisions. Across different levels of simulated choice difficulty and speed-accuracy emphasis, choice proportions and RT distributions simulated by the ensembles are invariant to ensemble size and the accumulated evidence at RT is invariant across RT when the accumulators are at least moderately correlated in either baseline evidence or rates of accumulation and when RT is not governed by the most extreme accumulators. To explore the relationship between the low-level ensemble accumulators and high-level cognitive models, we fit simulated ensemble behavior with a standard LBA model. The standard LBA model generally recovered the core accumulator parameters (particularly drift rates and residual time) of individual ensemble accumulators with high accuracy, with variability parameters of the standard LBA modulating as a function of various ensemble parameters. Ensembles of accumulators also provide an alternative conception of speed-accuracy tradeoff without relying on varying thresholds of individual accumulators, instead by adjusting how ensembles of accumulators are aggregated or by how accumulators are correlated within ensembles. These results clarify relationships between neural and computational accounts of decision making.

Object representations from convolutional neural network (CNN) models of computer vision (LeCun, Bengio, & Hinton, 2015) were used to drive a cognitive model of decision making, the linear ballistic accumulator (LBA) model (Brown & Heathcote, 2008), to predict errors and response times (RTs) in a novel object recognition task in humans. CNNs have become very successful at visual tasks like classifying objects in real-world images (e.g., He, Zhang, Ren, & Sun, 2015; Krizhevsky, Sutskever, & Hinton, 2012). We asked whether object representations learned by CNNs previously trained on a large corpus of natural images could be used to predict performance recognizing novel objects the network has never been trained on; we used novel Greebles, Ziggerins, and Sheinbugs that have been used in a number of previous object recognition studies. We specifically investigated whether a model combining high-level CNN representations of these novel objects could be used to drive an LBA model of decision making to account for errors and RTs in a same-different matching task (from Richler et al., 2019). Combining linearly transformed CNN object representations with the LBA provided reasonable accounts of performance not only on average, but at the individual-participant level and the item level as well. We frame the findings in the context of growing interest in using CNN models to understand visual object representations and the promise of using CNN representations to extend cognitive models to explain more complex aspects of human behavior. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Objective This research was designed to examine the contribution of self-reported experience and cue utilization to diagnostic accuracy in the context of radiology. Background Within radiology, it is unclear how task-related experience contributes to the acquisition of associations between features with events in memory, or cues, and how they contribute to diagnostic performance. Method Data were collected from 18 trainees and 41 radiologists. The participants completed a radiology edition of the established cue utilization assessment tool EXPERTise 2.0, which provides a measure of cue utilization based on performance on a number of domain-specific tasks. The participants also completed a separate image interpretation task as an independent measure of diagnostic performance. Results Consistent with previous research, a k-means cluster analysis using the data from EXPERTise 2.0 delineated two groups, the pattern of centroids of which reflected higher and lower cue utilization. Controlling for years of experience, participants with higher cue utilization were more accurate on the image interpretation task compared to participants who demonstrated relatively lower cue utilization ( p = .01). Conclusion This study provides support for the role of cue utilization in assessments of radiology images among qualified radiologists. Importantly, it also demonstrates that cue utilization and self-reported years of experience as a radiologist make independent contributions to performance on the radiological diagnostic task. Application Task-related experience, including training, needs to be structured to ensure that learners have the opportunity to acquire feature–event relationships and internalize these associations in the form of cues in memory.

We investigated whether a task requiring concurrent perceptual decision-making and response control can be performed concurrently, whether evidence accumulation and response control are accomplished by the same neurons, and whether perceptual decision-making and countermanding can be unified computationally. Based on neural recordings in a prefrontal area of macaque monkeys, we present behavioral, neural, and computational results demonstrating that perceptual decision-making of varying difficulty can be countermanded efficiently, that single prefrontal neurons instantiate both evidence accumulation and response control, and that an interactive race between stochastic GO evidence accumulators for each alternative and a distinct STOP accumulator fits countermanding choice behavior and replicates neural trajectories. Thus, perceptual decision-making and response control, previously regarded as distinct mechanisms, are actually aspects of a common neuro-computational mechanism supporting flexible behavior.

Visual art is pervasive in modern society. From advertising to fine arts galleries, the visual arts play a visible role in how we view and understand the world. In this review, we consider research that speaks to whether our experiences with art can change the way we see. Numerous studies speak, often indirectly, to this question—addressing whether artists see things differently from nonartists. Specifically, we reviewed literature that investigates the ways artistic ability and artistic training interact with visual abilities from the perspective that artists can be described as experts in visual media. Some work suggests that those who identify as artists or undergo artistic training perform better than nonartists on measures of low-level vision, high-level perception, and visual cognition, and show differences in brain activity while engaged in perceptual or artistic tasks. Other studies do not support these conclusions, however, and report no differences between artists and nonartists. We conclude that experimentally designed and well controlled training studies are necessary to elucidate whether artistic training shapes the brain and its perceptual and cognitive processes or whether budding artists gravitate toward the visual arts because of existing visual abilities.

The target article, “Robust Modeling in Cognitive Science,” proposes a number of recommended practices in computational modeling in response to the growing “crisis of confidence” facing many scientific disciplines, including psychology and neuroscience. Those of us who do modeling, write about modeling, teach modeling, and mentor modelers worry deeply about best practices and any new suggestions for making modeling more transparent, trusted, and robust are welcome. Many of the recommendations seem uncontroversial. My commentary focuses on forms of preregistration and postregistration, which constitute three of the four key ideas highlighted as take-home recommendations at the conclusion of the target article. I have chosen to consider these recommendations by reflecting on my own past experiences developing new models and modeling approaches.

One of the more principled methods of performing model selection is via Bayes factors. However, calculating Bayes factors requires marginal likelihoods, which are integrals over the entire parameter space, making estimation of Bayes factors for models with more than a few parameters a significant computational challenge. Here, we provide a tutorial review of two Monte Carlo techniques rarely used in psychology that efficiently compute marginal likelihoods: thermodynamic integration (Friel & Pettitt, 2008; Lartillot & Philippe, 2006) and steppingstone sampling (Xie, Lewis, Fan, Kuo, & Chen, 2011). The methods are general and can be easily implemented in existing MCMC code; we provide both the details for implementation and associated R code for the interested reader. While Bayesian toolkits implementing standard statistical analyses (e.g., JASP Team, 2017; Morey & Rouder, 2015) often compute Bayes factors for the researcher, those using Bayesian approaches to evaluate cognitive models are usually left to compute Bayes factors for themselves. Here, we provide examples of the methods by computing marginal likelihoods for a moderately complex model of choice response time, the Linear Ballistic Accumulator model (Brown & Heathcote, 2008), and compare them to findings of Evans and Brown (2017), who used a brute force technique. We then present a derivation of TI and SS within a hierarchical framework, provide results of a model recovery case study using hierarchical models, and show an application to empirical data. A companion R package is available at the Open Science Framework: https://osf.io/jpnb4.