V1 to Broca’s Chronicle

This project aimed to provide the first ‘end-to-end’ chronicle of neuronal computation for language production, specifically from entry into the cortex of perceptual information about words via early visual areas (during reading), to exit from the cortex of the speech plan via motor cortex.

I applied the high temporal and physiological resolution of ICE to the whole of the language-associated network as previously identified with fMRI. The patients tested in this project had a broader coverage of electrodes than my previous patients, which made it possible to address aspects of the task from vision to speech.

Using this end-to-end approach, I asked (a) when each region first comes online, from ~V1 to Broca’s area and beyond, (b) if each region can act as multiple computational entities over time (as I discovered is possible within Broca’s area), (c) if activity is strictly feed-forward and sequential, or evidences feedback, and (d) whether dynamics across the network hint at mechanisms for solving the problem of binding the disparate types of information likely computed by each region/entity.

Manuscripts coming soon (fingers crossed!)…

Some preliminary results include several distinct ‘waves’ of neural activity that propagate quickly through the entire cortical system supporting language behavior. Also, the circuits that are traditionally thought to be “early” areas were engaged persistently and much later than expected, and the circuits thought to be “higher” or later areas were engaged very early on within each trial period. The overall time-based model of information flow that is emerging is quite complex and multi-staged.

Note, however, that these dynamics in “activity” are complete independent of dynamics in “connectivity” – see my other results. Magnitude of activity in a given neuronal population of course is independent of the phase-relationships or synchrony across multiple such populations in a large-scale cell assembly – as may constitute the brain’s functional networks.

As a side note, I consider language-related behavior to include all the various activities from low-level vision (or audition) to visual object recognition to the grasp of ‘meaning’ to memory to core ‘linguistic’ processing to social cognition to motor planning and implementation. It seems necessary to me that all these and more are required for an organic language behavior event, e.g., a conversation. (This is why I wish to poll the neural system electrophysiologically, in so many sites.) As such, when considering language behavior so broadly, it seems clear to me that there is no one thing called “language”, any more than there is one thing called “action” or “thinking”. Some of the disparate talents that are confederated and brought together for language behavior will turn out to be more uniquely suited or customized to the language faculty, but it seems inevitable that a.) many will not, and b.) asking whether they are indeed language-exclusive is not the most productive question.

A further note: the framework taken here, namely of polling activity magnitude putatively from perceptual input to motor output implies a progression through time of the level of abstraction of neural processing. While prior research suggests that such a progression is real, my perspective is that this is not by any means the whole story. Importantly, I do not take as an assumption that neural processing of cognition proceeds by a feed-forward trajectory along which an increasingly complex “representation” is assembled. Rather the preferred perspective in this line of research is that many computations take place as network-level events and are a matter of a well-tuned assembly coalescing rather than a representation being amended and passed.