We conduct behavioral studies of memory and visual perception in both young and older populations. We have 7 behavioral testing rooms, each with 3D ready Asus monitors capable of refresh rates up to 144 Hz, nVidia shutter glasses for presenting different images to different eyes, and response boxes for the collection of highly accurate response time data. This setup allows us to run simple accuracy experiments, reaction time experiments, psychophysics studies with tight control of display properties, and continual flash suppression experiments in which we present visual images to participants in such a way that they are not aware that anything has been presented.
cMAP Lab Methods
Behavioral Experiments
 
Neuroimaging
We use functional Magnetic Resonance Imaging (fMRI) to examine the neural mechanisms of memory and visual perception. Current studies are investigating the nature of object representations and how they change with experience, as well as the mechanisms underlying cued recall. We use both univariate and multivariate analysis techniques, as well as effective connectivity analyses (DCM). We are also developing experimental and modeling procedures to understand the way in which neural-level responses are mapped onto the more coarse-grained BOLD signal recorded with fMRI, in visual cortex. For our imaging studies we use the new Siemens 3T Skyra at UMass Amherst. We are funded by the NSF for fMRI studies of memory and high-level vision (1554871) and by the NIH for fMRI studies of lower-level vision and modeling the BOLD signal (1RF1MH114277).
Computational Modeling
In the lab, we use computational methods in several ways. We are currently developing a novel technique for modeling the BOLD signal that can characterize properties of the neural-level responses that give rise to voxel tuning functions. We also build neural network models of cognitive function, to explain and predict behavioral data from lesion studies and fMRI data from neuroimaging studies. Finally, we use mathematical modeling techniques to analyze our empirical data, for example using Bayesian hierarchical models, Dynamic Causal Modeling, and other model-fitting and model comparison techniques. Our models and analyses are coded in Matlab and R. All students and postdocs in the lab are issued a high speed laptop or workstation. We also maintain a powerful server for processing neuroimaging data and performing other computationally intensive work.
computational Memory and Perception Lab
© Cowell Lab 2018, University of Massachuetts, Amherst.