| Besides studying multiple topics in visual recognition, I also employ a fairly wide range of techniques to understand the nature of human visual processing. I also work with both adult, child, and infant observers to investigate visual behavior across the lifespan. The diversity of my research program makes it possible to study visual learning and recognition through converging methods at multiple levels of observer expertise. In what follows, I briefly describe the main methodologies I use, and discuss how they contribute to my research goals: Visual Psychophysics/ Behavioral Experimentation While I'm very interested in using high-tech imaging technology and computational modeling to characterize human vision, I remain very interested in studying human vision through simple behavioral methods. In particular, I'm interested in exploring the scope and limitations of invariant visual recognition and "face-specific" visual processing in children and adults (see below). I also often use computational models as tools for creating stimuli that are useful for addressing focused questions regarding natural visual function. This strategy (which could be called "analysis-by-synthesis") is a valuable tool for examining visual representations of complex inputs.  Computational Modeling My earliest work in visual recognition focused on using methods from computer vision and machine learning to explore novel features for recognition. As I continue to explore this issue, computational modeling is a vital tool for maintaining objectivity when suggesting new mechanisms or features that might describe human performance. Formalizing new models to the point where they can be validated computationally is a necessary step towards making a theory concrete, and similarly, new computational results can point the way toward interesting experimental designs for psychophysical study. The interplay of computation and behavioral work has been very fruitful throughout my career thus far.  Computational modeling can suggest new representational strategies for recognition, like the "Dissociated Dipole" operator, depicted above. This operator emerged from some modeling results I carried out early in my graduate career, and has led to important advances in several computer vision systems as well as raised important psychophysical questions. Electrophysiology In my postdoctoral work, I've begun studying infant and adult populations in parallel to understand how face expertise emerges and is refined over time. Being able to compare infant and adult performance rests critically on finding a methodology that translates easily between the two populations. To that end, I've been fortunate to have the opportunity to learn to use event-related potentials (ERPs) as a tool for investigating face perception in infant and adult populations. ERPs are relatively easy to obtain from infants (though not without challenges!) and require no explicit response from the observer. As such, they are invaluable for describing visual processing across the lifespan and provide a window into the neural correlates of visual behavior. In the Nelson Lab, I've also been exploring how to use machine learning techniques on ERP data to carry out "neural decoding" research in infant populations.  At left, some data collected from adult subjects for a collaborative project between myself and Dr. Meg Moulson highlighting typical "Face-sensitive" components of the ERP signal at the "N170." At right, my godson (Max Connor) is all capped and ready to go for an experiment in the Nelson Lab. Max remains my favorite participant of all time. Thanks to Chris and Aurora Connor for their permission to use this image (and read their baby's brainwaves). |