The Interactive Account of ventral occipitotemporal contributions to reading By Cathy J. Price and Joseph T. Devlin Danika Teverovsky – BCS 265 – April 2nd, 2015

Why I’m interested I’m a junior BCS major on the perception track Multi-modal processing Visuospatial integration, visuo-auditory integration Dyslexia – a family friend’s experience Organization of information

Some buzz words – the ever-helpful glossary Bottom-up sensory information Top-down predictions Generative models Predictive coding Prediction error We’ll get to all of these.

Ventral occipitotemporal cortex (vot) – Anatomy Figure 1: visual word recognition in the vOT a) Anatomy and its relation to activation for visual word recognition (red-yellow) shown on the ventral surface of an inflated left hemisphere Figure 1 a)

VOT – Receptive fields Figure 1 b) rotated and c) Figure 1 b) “Examples of simple shape stimuli that are important for recognizing both visual words and objects. Neurons within V2 respond to these types of simple shapes and project to V4, where the cells have more complex receptive fields that respond to combinations of these shapes within a retinotopic reference frame. These in turn project to vOT neurons that have receptive fields with multidimensional tuning functions, where simple shape elements are combined nonlinearly in an object-centered reference frame. Thus, unlike earlier visual areas, it is difficult – if not impossible – to find the optimal stimulus driving a cell using a simple line drawing” c) “complex, object-centred receptive field for a vOT neuron. On the left are three ‘J’s of different sizes in different retinal positions. Within early retinotopic areas, each J would be encoded by non-overlapping sets of neurons. By contrast, the receptive field illustrated on the right by a three by three grid of panels provides amore compact, stable object-centred representation. Here, curvature and orientation are plotted recursively within each receptive field region such that it will respond strongly to any combination of a vertical straight line at the top right and a concave-up curved horizontal line at the bottom. Although it is tempting to call this a ‘J detector’, this would be incorrect – the receptive field responds equally well to the handle of an umbrella or trunk of an elephant but does not respond to the letter j written in script.” I’m taking BCS 223 Vision and the Eye right now, and we are talking about retinotopic mapping and visual fields. How it relates to reading, though, _________ Figure 1 b) rotated and c)

Bottom-up & top-down – review Bottom-up sensory information: external information arrives at the senses and projects to primary sensory cortices. These drive secondary, tertiary and higher order association cortices via forward connections arising primarily from superficial (layer II and III) pyramidal neurons. Within the ventral occipitotemporal cortex (vOT), the primary source of bottom-up information is visual, presumably from areas” Top-down predictions: “the automatic input a region receives from areas above it in the anatomical hierarchy. These connections attempt to predict the bottom-up inputs based on the context and active features. Important sources of top-down input to vOT are (deep) pyramidal cells in cortical areas that contribute to representing the sound, meaning and actions associated with a given stimulus.”

Generative models and predictive coding Probabilistic modeling Adjust the weight of information from top-down sources versus bottom-up to better predicitons Machines built with models like these – exhibit interactivity Ability to functionally describe the flow of information in a domain-general model

Interactive account Premise: perception requires interaction between sensory and higher order cortices This is done via forward and backward connections vOT an interface for this information

Interactive account Figure 2 a)

Interactive account Figure 2 b)

Dyslexia and learning to read Developmental dyslexia – reading development falls behind compared to other academic development Achieved reading skill often limited, and reading is slow Affects ~10-20% of the population Dyslexics show abnormally low vOT activation and reduced functional connectivity between the vOT and other language areas

This model explains a lot Activation increases for pseudowords (ex. biat) relative to: consonant letter strings (psuedowords more word- like, therefore engage top-down predictions) Rtghlkd words (because larger prediction error for pseudowords) Boat Word selectivity

Pros and cons of visual-word-form specialization This paper (Price & Devlin) and the Dehaene & Cohen paper argue about the vOT: Price & Devlin -> multi-functional Dehaene & Cohen -> specialized for visual words Arguing specific function gets messy quite quickly For example, what can a hand do? Is it functionally specialized to tie your shoes? Type on a keyboard? Use a writing utensil? Hands can grasp, that is more broad. But it is multi-functional

Outstanding questions Are there anatomical sources of top-down information? Are there anatomical pathways linking the vOT to the higher order cortex? Are there pathways that bypass the vOT (and link visual and auditory cortices)? Left versus right vOT contributions? Are reading-specific neurons of the vOT able to be physiologically differentiated (as opposed to the domain-general approach)? How does anatomical damage affect function? What does this mean for dyslexia?

Future Discussion questions Is the vOT specialized for visual-word-form? Price & Devlin say no…Dehaene & Cohen say yes…what do we think? Is one view more advantageous than the other? US Economy/Wall St analogy – US economy happens on Wall St; is Wall St is only for financial institutions?

Thank you!