TALKS
2025-2026
The group addresses questions, such as what’s the difference between a posed and a spontaneous expression? how fast can we perceive a face or an emotional expression?, what strategies do radiologists employ to detect breast cancer and is this skill trainable? how do clinical conditions, such as depression, autism, affect face recognition? To address these questions, researchers in the collaborative employ a variety of empirical techniques involving psychophysics, cognitive experiments, eye tracking, neural imaging (fMRI, EEG), and computer modeling.
Catch up on the 2025-26 DMC talk season here!
______________________________________________________________________________________
September 10th, 2025
Speaker: Oscar Solis (Evans Lab)
Title: Neural Correlates of Visual Complexity and Image Memorability
Abstract: Humans have a remarkable ability to remember thousands of images (Standing, 1973). Not only does the semantic category depicted in the image contribute to this ability but also the idiosyncratic details that make exemplars of the same category distinct from each other (Evans & Baddeley, 2018). Previous work has shown that images that are perceived as more complex tend to be more memorable (Kyle-Davidson et al., 2025), suggesting that visual complexity could contribute to the detail component of visual long-term memory. Here I present work done up to date in our investigation of the neural correlates of this relationship using the Natural Scenes Dataset (Allen et al., 2022). This dataset consists of high-resolution functional magnetic resonance imaging (fMRI) scans from 8 participants who viewed thousands of photorealistic images during a continuous recognition memory task. We have run whole-brain correlational analyses between single-trial betas and metrics derived from these images: complexity and memorability scores generated from human data (still to come) as well as predictions made by a range of computational models. We hypothesize that activity in medial temporal lobe and ventral visual areas will correlate with memorability scores (Bainbridge et al., 2017). We also predict that activity in early visual cortex and higher visual areas such as parahippocampal place area will be correlated with complexity scores (Zhou et al., 2023). By inspecting the overlap in correlation maps between these two features, we will identify the possible neural substrates that underlie the behavioural relationship between complexity and memorability.
Link to Zoom recording of Oscar's Talk
___________________________________________________________________________________
October 8th, 2025
Speaker: Quoc Vuong
Title: A hierarchical model for affect judgments of natural images: Exploring the role of local, global and category visual features
Abstract: Natural images capture real-world scenarios, portraying people and objects situated in a scene. Importantly, observers can experience emotional responses and show corresponding neural and physiological changes when looking at images. These responses can be characterised along an arousal (activation) and valence (pleasantness) dimension, often referred to as the circumflex model. Images can thus serve as strong affective stimuli that moderate arousal and valence levels in observers. The human visual system has a hierarchical structure that is very effective for many visual tasks including affect judgments. Here I present a (preliminary) deep neural network and some pilot data to explore: (1) what visual features related to visual categories (objects, faces and scenes) can predict human arousal/valence ratings; and (2) how “global” valence ratings of full images relate to “local” valence ratings of image regions. Finally, I discuss the need to extend these models to other naturalistic stimuli such sounds and videos.
Link to Zoom recording of Quoc's Talk
___________________________________________________________________________________
November 12th, 2025
Speaker: Ivette Colón (Rogers Lab)
Title: Neural representations of faces are widely distributed and highly individualized
Abstract: As many in the DMC know, there is a historical assumption that our brains are a modular system– where discrete areas handle particular aspects of processing, with perhaps the most famous example being the FFA (fusiform face area) for face processing. While there is no doubt that certain areas are involved in certain processes, there is increasing evidence that there may be more information in other parts of the brain that contribute to those same processes. In this talk, I will present a (very) recently completed study in which we gather neural representations for a set of highly-controlled faces, places, and objects from 20 people in a slow-event related fMRI design. Then, using a multivariate decoding technique called Iterated LASSO, we find— separately for each participant, and across two scans per person— anatomically distributed patterns of voxels that carry reliable signals about whether a stimulus is a face, place, or object. Finally, we used these signal-carrying voxels as candidate stimulation areas in a transcranial magnetic stimulation (TMS) task, which allows us to examine whether brain areas outside of canonical face processing areas causally contribute to participants’ face processing. I will show that a) information about faces is encoded throughout the cortex, b) where exactly is variable across individuals, c) these differences are stable over time, and d) stimulation of these areas shows similar behavioral patterns as stimulation of established face processing areas. Together these results suggest that neural representations of faces may be more widely distributed and individualized than previously thought.
Link to Zoom recording of Ivette's Talk
___________________________________________________________________________________
December 10th, 2025
Speaker: Heida Maria Sigurdardottir
Title: What does your mind’s eye actually see? Introducing a visually anchored measure of imagery
Abstract: We often talk about having a “vivid imagination” or a “blind mind’s eye,” but we rarely ask what those experiences actually look like. Most imagery research still relies on verbal questionnaires such as the Vividness of Visual Imagery (VVIQ), where people rate vague prompts (“a friend’s face,” “a shop front”) on a 1-5 vividness scale. That makes it hard to know whether a “2” for one person resembles a “2” for anyone else, and whether we are really measuring how imagery appears rather than how people talk about it.
In this talk, I introduce the Visual Imagery Visually Anchored Scale (VIVAS), a new tool that lets participants reconstruct their mental images instead of just rating them. On each trial, people briefly see an object (faces, animals, buildings, food, manmade objects, or novel objects), then imagine it, and adjust three dimensions (opacity, color saturation, and sharpness) until an on-screen image matches what they see in their mind’s eye. I will argue that tools like VIVAS can move imagery research beyond vague vividness ratings toward concrete, appearance-based measurement that is more sensitive to both dimensional and content-specific differences in visual imagination.
Link to Zoom recording of Heida's Talk
___________________________________________________________________________________
January 14th, 2026
Speaker: Géraldine Jeckeln (O'Toole Lab)
Title: Examining perceptual confidence for own-race and other-race face-identity matching decisions
Abstract: People identify faces of their own race more accurately than other-race faces, a phenomenon known as the other-race effect (ORE). Confidence reliably indicates face-identity matching accuracy for same-race faces, but it is not known whether this relationship holds for other-race faces. We examined the confidence-accuracy relationship for same- and other-race faces in Caucasian and African American participants. To eliminate response bias, we measured confidence using a confidence forced-choice task (Mamassian, 2016), rather than with confidence ratings. Participants completed two consecutive face-identity matching trials and selected the trial on which they felt more confident (confidence-chosen trial). The results showed an ORE for face-identity matching accuracy, but not for the confidence-accuracy relationship. We also examined the extent to which the relative difficulty of paired trials related to confidence decision consistency. Specifically, on paired trials with large differences in difficulty, people should consistently judge the easier trial as the confidence-chosen trial; when trials are more evenly matched in difficulty, confidence judgments will be less consistent. For African American participants, the difference in difficulty between the paired trials predicted confidence for both same and other-race decisions. For Caucasian participants, the difference in difficulty between the paired trials predicted confidence for same-, but not other-race, face-identity matching decisions. Overall, confidence predicted accuracy in both same- and other-race face-identity matching, with difficulty consistently guiding confidence for same-race faces. These findings enhance our understanding of how confidence decisions are formed and guide cross-race face-identity matching decisions.
Link to Zoom recording of Gerie's Talk
___________________________________________________________________________________
February 11th, 2026
Speaker: Jim Tanaka
Title: Mind reading and the mental maps of expertise: Using machine learning to uncover category structure in experts and novices.
Abstract: As cognitive scientists, we aim to understand how the human mind represents the world by integrating behavioral, physiological, and neural measures such as reaction times, eye movements, and fMRI. In this talk, I will describe a machine-learning approach called PsiZ (Roads & Mozer, 2019) that uses similarity judgments to infer an observer’s category structure as a psychological embedding. We apply this technique to study recognition of stimuli from a variety of domains, including warbler birds, rocks and Japanese kanji characters. Our work examines expertise at both the group and individual levels, as well as how category representations change as observers develop from novices to experts through laboratory training, classroom learning and real-world experience.
Link to Zoom recording of Jim's Talk
___________________________________________________________________________________
March 11th, 2026
Speaker: Deyan Mitev (Koldewyn Lab)
Title: Against the male average: Hierarchical Drift Diffusion Modelling of Sex Judgments
Abstract: Ambiguous faces are more likely to be categorised as male, particularly under conditions of uncertainty. Additionally, recent visual search studies have shown an asymmetry in the representation of sex such that females are found more readily than males. Such findings have been taken to support the male default hypothesis, which proposes that the prior or default representation of a face is closer to male than female. We tested this hypothesis using evidence accumulation modelling. Participants (N = 40) categorised faces as male or female. Stimuli consisted of morph continua generated from 20 pairs of real human faces. We fitted a Hierarchical Drift Diffusion Model to examine whether sex categorisation biases emerged from a starting point bias (reflecting prior expectations) and/or differences in drift rate (reflecting evidence accumulation efficiency). The male default hypothesis predicts a starting point bias toward the male decision boundary and faster evidence accumulation for male faces. We found a significant difference in drift rate, with steeper drift toward the male boundary compared to the female boundary. This indicates faster accumulation of perceptual evidence supporting male decisions. However, there was no starting point bias toward either decision boundary. These findings do not support the prediction that male faces constitute the default prior representation. Instead, they suggest that male categorisation advantages emerge due to differences in perceptual evidence accumulation rather than pre-decisional bias. This contributes to our understanding of the androcentric bias as a function of perceptual decision-making rather than baseline priors.
Link to Zoom recording of Deyan's Talk
___________________________________________________________________________________
June 10th, 2026
Speaker: Oruc Lab
Title: TBA
Abstract: TBA
Link to Zoom recording of Talk: Coming soon!
___________________________________________________________________________________
July 8th, 2026
Speaker: Greene Lab
Title: TBA
Abstract: TBA
Link to Zoom recording of Talk: Coming soon!
__________________________________________________________