Humans have the remarkable ability to adapt to environmental pressures through learning – a process that is accompanied by heterogeneous changes in brain structure and function. Parsing the neurophysiological mechanisms of learning could provide a better understanding of how the brain is able to fine-tune its organization and dynamics to accurately and efficiently accomplish day-to-day tasks. Importantly, such alterations continue to impact future behaviors well beyond the critical period over which learning occurs, and in a manner that is unique to each individual \cite{Painter2014,Tobler2005}. These and related empirical observations have resulted in a shift in investigational paradigms that seek to understand general mechanisms of learning and the ensuing variability in these processes that drive inter-individual differences in behavior \cite{Squire1992}.
Many factors may contribute to inter-individual variability in learning, including the mechanics of the particular skill or information that is being learned, and the strategies by which the same phenomenon, concept, or principle can be apprehended \cite{KANFER1990221,marton1976qualitative}. It is intuitively plausible that different individuals could be predisposed to engage in different learning strategies based on the nature of their central nervous system \cite{bassett2017network}. Unique genetic and environment factors present in the early stages of development can give rise to different wiring patterns in the brain \cite{di2014unraveling}, in turn learning to different patterns of neural activity in response to stimuli \cite{Atun-Einy2012,harrison2011learning,wigfield2000expectancy}. A particularly parsimonious language in which to describe and characterize such patterns is that of network science \cite{bullmore2009complex,bassett2017networka}, where regions of the brain are represented as network nodes whose activity can vary over time \cite{murphy2016explicitly}, and connections between regions are represented as network edges whose strength can also vary over time \cite{gu2017functional}. While some organizational principles of brain network organization and dynamics appear to be conserved across individuals \cite{betzel2016multi}, others – including measures of activity \cite{prat2011individual,prat2007individual} and connectivity \cite{KANFER1990221} – vary appreciably \cite{Finn2015Functional}. It is as yet unknown to what degree these shared versus unique features of brain network structure and dynamics explain the processes of learning and the resultant behavior.
From a computational perspective, identifying tools that can be used to compare shared versus unique features of brain network structure and dynamics in a statistically principled fashion are relatively sparse. Arguably one of the most elegant potential approaches is intersubject correlation (ISC), which has historically been used to quantify the extent to which neuronal processes are shared across individuals \cite{hasson2004intersubject}. ISC estimates the similarity of brain activity across subjects by using the temporal signature of neuronal activity in a particular brain region as a predictor of the neuronal activity in the corresponding brain region of another individual. Thus, ISC is a “model-free” approach that is thought to reflect the degree of inter-subject synchrony in emotional or mental states, which is not simply explained by common evoked responses to the tasks \cite{nummenmaa2012emotions,simony2016dynamic}. While ISC can effectively capture the shared neuronal response of individual brain regions, it does not measure the degree to which the functional connections between pairs of brain regions are commonly modulated across individuals. Recent work demonstrating that topological features of functional brain networks can predict the capacity of an individual to learn new skills \cite{bassett2015learning,mattar2017network,schoenbaum2000changes}, motivates the development of an approach that can quantify the inter-subject similarity of functional connectivity. Such techniques to estimate the similarity of functional brain organization of human cortex across subjects during learning in the context of both functional activity and functional connectivity constitute important tools to probe the neural basis of individual differences during learning.
In this study, we utilize ISC to measure the response of brain activity shared across individuals learning the value of novel objects over the course of 12 training sessions taking place on four days. In addition to the ISC, we introduce the intersubject functional connectivity (ISFC), which measures inter-regional functional correlations across subjects. The ISFC is estimated by calculating the Pearson correlation coefficient between the pattern of functional connectivity of a single subject and the average pattern of functional connectivity across all remaining subjects. We hypothesized that subject-general patterns of activity and connectivity would be located in motor cortex (consistent with the shared finger movement to press the response box) \cite{bassett2011dynamic,mattar2017network}, visual cortex (consistent with the shared processing of the visual stimuli of the novel objects), and other areas previously associated with the learning of value, including lateral occipital cortex \cite{persichetti2015value}. Based on prior work showing that learning modulates functional connectivity on an individual basis \cite{baeg2007learning,fatima2016dynamic,welberg2009learning}, we hypothesized that increase in intrinsic functional connectivity shared across the subjects over a period of learning modulated by the coherence of stimulus induced activity. Such findings would demonstrate the utility of intersubject analyses (ISC and ISFC) in the context of functional activity and functional connectivity could provide clear neurophysiological correlates and the dynamic of intrinsic functional connectvity and task-dependent activity in the human brain shared across the subjects over a period of time during learning.