Research
Interests
I have always been fascinated by how we process the complex world around us. Specifically, I am interested in how we process the world using vision. The eyes - our sensory organ - take in both relevant and irrelevant information around us, and the brain - our perceptual organ - instantaneously and seemingly effortlessly transforms this into a relevant experience. How does the eye function as a sensory tool to allow such a rich perceptual experience? Most of our daily visual experiences deal with continuous object recognition. As we walk through a room, we look at and see a chair, a desk, a computer, a sticky note, all in succession. We immediately identify these objects, understand how they relate, and decide to act according to this information. I’d like to understand how the brain utilizes visual information to generate perception in noisy, real-life situations. I am interested in exploring the interplay between bottom-up and top-down attention (semantics, saliency, and goals), the role of eye movements in prioritizing visual information in perception, and what processes or factors determines what visual information is given meaning (what do we visually experience?). Together, these interest areas can inform us of how the brain generates a visual percept, and how this is updated to reflect a change in information, or how past information is relied on to generate a stable view.
Methods
I work with human participants using behavioral psychophysics experiments. In the past I have employed fMRI and eyetracking to obtain physical measures from the brain of the behavioral activity I observe.
Projects
Serial dependence during slow change blindness
My current goal is to investigate serial dependence as a mechanism driving slow change blindness. When individuals view a slowly changing scene, they are often blind to large changes which occur very slowly. Why does this happen? Are people stuck in the earliest perception of the changing object? Do they keep up with the current representation but in turn override earlier information? Or perhaps their current perception lags behind reality due to a reliance on earlier information. This latter theory would align with serial dependence. In general, I am interested in how a slowly changing object is perceived at any given moment during the slow change. The first step in this investigation is to design a method to evaluate if the perception of an item has changed. To do so, I use maximum likelihood difference scaling (MLDS; Maloney & Yang, 2003) to evaluate perceived differences between color stimuli. I then show how the difference scaling shifts when a preceding stimulus alters color perception. Up next is to use this method to evaluate changes in perception while viewing a slowly changing stimulus!
Slow change blindness
Slow change blindness is a not-often-studied and therefore not-well-understood phenomenon where observers fail to notice otherwise obvious changes in an image because they occur very slowly. This is a surprising failure because the item has remained in sight the entire time, and even if observers do not notice the slow change as it occurs, one would expect that at some later instance they would notice that an object had changed - yet this is not the case. With Jan Brascamp at MSU and colleagues from NYU, I developed a stimulus generation method and corresponding slow change stimulus set that reliably induces slow change blindness. This unlocks the potential to study slow change blindness in a systematic way! You can read about this stimulus set and download some example for yourself here!
Neural correlates of consciousness
Another project I am working on looks at the neural correlates of consciousness. We use the classic change blindness phenomenon with retro and post cues to train participants to access their iconic memory. Iconic memory is a very short lived (1000ms) visual memory storage that contains a rich representation of what was just looked at. Information in iconic memory can be consciously experienced even after the stimulus has disappeared by quickly attending to it. Iconic memory is thought to be an example of phenomenal awareness: information that you are not presently conscious of (access consciousness), but that you can be conscious of as soon as you attend to it. Here, we ask questions including: Is recurrent processing sufficient for consciousness? Where are the neural correlates of consciousness in the brain?
Pupillometry
They say the eyes are the windows to the soul, but we in pupillometry know that the pupils are the windows to the brain! Pupillometry is the measure of pupil size in response to a stimulus. Many people are familiar with the pupil light reflex (PLR), in which the pupil constricts (gets smaller in diameter) in response to increased light. There is also an established pupil response to cognitive tasks. When completing a variety of mental tasks (such as discriminating between line orientations, doing math, or recalling items), the pupil dilates (gets larger in diameter). There are a few brain areas that are correlated with such responses, including the locus coeruleus (LC) and the superior colliculus (SC). With this project, we were interested in whether different cognitive tasks evoke different response shapes, and what pupil response to other light driven (but not the general PLR) stimuli look like. Do similarities in pupil response shapes reflect similarities in mechanisms? Do differences in pupil response shapes reflect differences in mechanisms? Ultimately, can we draw conclusions about brain activity based on changes in pupil size?
Object recognition
As an undergraduate, I helped on a project studying object recognition in noise (Gaussian and Fourier noise) of humans and convolutional neural networks (CNNs) to identify object features that aid recognition. The paper is published here.
My honors thesis project was on the role of background information on object processing in humans and CNNs. Using Adobe Photoshop, I created greyscale composite images of a foreground object superimposed on a background scene. I varied visual clutter, the amount of texture, pattern, or excess background information in an image, as well as semantic congruency, whether the foreground object and background content made sense to occur together. I compared human and network performance across categories. I wondered: How much does background information play a role in object recognition? Do convolutional neural networks (CNNs) maintain human level performance when background information is manipulated?