Works in Progress / Guest Speaker
including Ben Harvey, Utrecht University
WIN Wednesday
Wednesday, 18 October 2023, 12pm to 1pm
Hybrid via Teams and in the Cowey Room, WIN Annexe
Join via TeamsWorks in Progress (WIP)
Control-dependent working memory codes for protection versus manipulation
Recht & Manohar
Information held in working Memory (WM) is under our active control: we use it to guide actions and achieve goals. For example, we can prioritise, update, manipulate, or protect the information. (Miller et al., 2018; Nobre and Stokes, 2019). New findings suggest that we use a dynamic coding scheme to achieve this, a code that varies with task-specific and response-dependent factors (Kikumoto et al. 2022; 2023).
When information is protected from interference, it may be shifted into an orthogonal neural coding space, which is unaffected by other concurrent processing. And when information is mentally manipulated, some neural coding dimensions may remain stable, while others are updated.
Using MEG, we will investigate how two types of WM control – namely the manipulation of WM content and the protection from interference – impact WM codes. We will compare three conditions presented in blocks: a baseline condition, during which a visual stimulus must be maintained for a given duration and then compared to a reference line; a manipulation condition, where the same stimulus must be rotated before the comparison to reference; and an interference condition, where an irrelevant numerosity task is intercalated between stimulus and reference. Throughout the retention period, a neutral visual stimulus (or ‘ping’) will be shown at various moments to detect activity-silent stimulus traces. We will train classifiers to decode the stimulus within and across conditions.
We hypothesise that after manipulation and interference, the “ping” will reveal that coding axes of the remembered stimulus will be misaligned with the coding in the baseline condition. Moreover, if the same kinds of WM code are used for stability and manipulation, we may also find that coding after interference aligns well with the coding after manipulation. We also predict that interference and manipulation directly alter the code, when comparing pre to post, within conditions. Such observations would indicate the presence of an activity-quiet WM code that adjusts in response to task demands, but may be common to protecting and manipulating information WM during the retention period.
Guest Lecture
Human quantity processing, from sensory to cognitive systems
Ben Harvey, Utrecht University Department of Experimental Psychology.
Human perception of physical quantities like numerosity and event timing supports many cognitive functions, from foraging to mathematical and scientific thought. Investigation of the underlying neural processes has focussed on quantity-tuned neural responses, which respond maximally to different numerosities or timings in different neural populations. Here I will describe 7T fMRI studies in which we show quantity-tuned neural populations in many areas of the human brain and for many quantities. These are spatially organised to map the quantity across the cortical surface. I will then discuss recent studies in which we reveal how these quantity-tuned responses are derived from responses in sensory systems, how they are then transformed between different brain areas, and how responses to different quantities are related. I will first show how straightforward analyses of visual images can determine their numerosity with little effect of item size or spacing, and that this analysis predicts responses to numerosity in early visual cortex and neural network models more closely than numerosity itself. Once quantity-tuned responses have been derived, I will show how responses to visual numerosity and haptic numerosity are related. Finally, I will describe a set of brain areas whose responses depend on working memory load near those showing quantity-tuned responses. I propose that this grouping allows a linking of physical quantity representations and working memory, supporting higher cognitive functions. Together, these results give an integrated overview of quantity-tuned neural response, their derivation in sensory systems, and their extension into higher-level cognitive processing.