Neural Dynamics and Cognitive Ontology

Jessica Cohen (University of North Carolina, Chapel Hill)
Recent advances in neuroimaging methods and analysis have led to an expanding body of research that investigates how large-scale brain network organization dynamically adapts to changes in one's environment, including both internal state changes and external stimulation. In this talk, I will provide examples from my research of both externally-driven and internally-driven changes in functional brain network organization that impact cognitive task performance. First, I will discuss findings that whole-brain network organization dynamically adjusts to changing cognitive demands in a task specific manner, and that individuals whose network organization is more flexible perform better (Cohen & D’Esposito, 2016). Next, I will describe new research in which I find that dynamic changes in network organization during a resting state, when there are no changes in external stimuli, can impact subsequent task performance across multiple cognitive tasks. These findings emphasize the importance of probing the dynamics of functional networks when considering how brain network organization relates to successful cognition.

Felipe De Brigard (Duke University)
Although cognitive systems are commonly posited in explanations in psychology and cognitive neuroscience, there is uncertainty as to how to precisely characterize them. Recently, I (De Brigard, 2017) argued against a prominent characterization of cognitive system put forth by Rupert (2009; 2011), on account that it cannot capture the fact that brains exhibit both functional stability and diachronic dynamicity, as manifested, for instance, by changes in brain dynamics as a function of age without concomitant changes in task performance. In this talk I suggest a solution to the challenge of characterizing the notion of cognitive system in a way that allows for both functional stability and diachronic dynamicity. To that end, I build upon recent developments in topology and network analysis to offer a characterization of what may be called “dynamic cognitive systems”. Roughly, my suggestion is that a dynamic cognitive system can be seen as a neural network in which time is parametrized. More specifically, I argue that recent algorithms from temporaldynamic network analyses of task-based neuroimaging data can provide us with models of brain mechanisms that can exhibit stability in performance (i.e., functional stability) while at the same time capture underlying structural changes through time (i.e., diachronic dynamicity). I conclude with the admonition that any attempt to explain cognition mechanistically through neuroscience would require rethinking the notion of cognitive system in dynamic terms.

Bryce Gessell (Duke University)
Many studies on long-term memory formation conceptualize this process as a series of three steps. The first, encoding, is the modification of neural structures in response to incoming information. The second step is consolidation, during which molecular and cellular changes make these earlier modifications more durable. The third step, storage, involves the long-term maintenance of such changes, which preserve a memory trace in a state suitable for reactivation or retrieval. All three steps happen at timescales which can differ by orders of magnitude: encoding may take a single second; consolidation, hours or days or even weeks; storage may continue for many decades. However, it is difficult to find neurobiological correlates whose temporal profiles match those of the steps in longterm memory formation. For example, one form of long-term potentiation (LTP)— a mechanism of memory consolidation and storage—depends on postsynaptic NDMA receptors. But NMDA receptors, as proteins with a short half-life, degrade and recycle frequently, meaning that the temporal stability of consolidation or storage cannot correspond to similarly-stable molecular and cellular structures. Rather, many different neurobiological processes, operating over different timescales, give rise to the three abiding “steps” of long-term memory. In this talk I investigate the extent to which these different neurobiological processes may prompt us to reconsider our analysis of long-term memory formation. In particular, I argue that alternative pathways of early consolidation, such as those enhanced by sleep (Yang et al. 2014; Feld and Born 2017), suggest a more nuanced decomposition. I outline one possible decomposition, paying special attention to how variability at neural substrates across time can culminate in meaningful behavioral and cognitive consequences. I conclude with a self-conscious reflection on the method of my talk, and on the role of molecular and cellular findings in cognitive neuroscience more generally.

Colin Klein (Australian National University)
Biologically realistic models of neural spiking take into account spike timings, yet the relevance of spike timing beyond individual neurons is often unclear. In Izhikevich's (2006) model, spike timing plays a crucial role in allowing for the natural formation of polychronous circuits. These are firing patterns which are timedependent, but do not require synchronous firing of all neurons in an assembly. Further, individual elements can figure in a number of distinct polychronous assemblies, their role determined by the timing of their firing relative to other neurons. I argue that this reflects a distinct organisational principle from traditional notions of pluripotency, redundancy, or re-use, and argue that for the phenomenon to be understood properly requires a shift from a traditional static, capacity-based view of computation to a process-based one.

Sarah Robins (University of Kansas)
The idea that remembering involves a fixed engram, which becomes stable and permanent as a result of consolidation, has been a guiding assumption of the neuroscientific study of memory since its inception. Understanding of and commitment to the engram and consolidation has wavered more recently, as both systems and cellular neuroscientists have shifted to thinking of memory as a continuous and dynamic process. A central theme in 21st century neuroscience of memory is resistance to this traditional picture, with some calling for “the demise of the fixed trace” (Nadel, 2007) and others urging us to reject the “consolidation dogma” (Silva, 2007). One advantage of the traditional model of engrams and consolidation was that it offered clear identity conditions for the engram. The engram was identified functionally as (roughly) whatever neural change occurred as a result of learning and was uniquely implicated in subsequent behavior (Josselyn, Köhler, & Frankland, 2015). While there is a growing consensus that this traditional framework should be rejected, there is no accompanying agreement on how to modify our understanding of the engram. Proposed alternatives involve different ways of conceiving of engrams, not all of which can be true simultaneously. Proponents of reconsolidation suggest that consolidated engrams enter a new phase of vulnerability as a result of retrieval, incorporating new information (Nader & Hardt, 2009). So, on this view, an engram is the same engram over time so long as the same neurons are involved, even if the content/behavior/output produced by those neurons changes. Compare this to the standard consolidation model, whose proponents argue that the engram literally moves over time – from cells in the hippocampus to cells in the neocortex. On this view, the engram is the same engram over time so long as the content/behavior/output remains the same, even if the neurons responsible for producing it change. Which view is correct? The question cannot be resolved empirically. It requires reflection on our conceptual and theoretical commitments around the nature of memory and the role of the engram. In this paper, I apply the resources of contemporary philosophy of neuroscience and philosophy of memory to provide a framework in which the available avenues for theorizing about memory as a dynamic process can be understood.