CBB talk series

Speaker

Title

Abstract

Thurs, Jan 30

Rotem Botvinik-Nezer, Dartmouth College

Variability in the analysis of a single fMRI dataset by many teams

The "replication crisis" in many scientific fields has raised concerns regarding the reliability of published results. One reason for the high rate of false positive results is the large number of "researcher degrees of freedom", where the process of data analysis can be performed in multiple ways. This is specifically apparent in neuroimaging, where there is a thriving "garden of forking analysis paths". In the Neuroimaging Analysis Replication and Prediction Study (NARPS: https://www.biorxiv.org/content/10.1101/843193v1.full), we tested the variability fMRI results across  analysis pipelines that are used in practice in research laboratories. Seventy analysis teams independently analyzed the same fMRI dataset to test the same ex-ante hypotheses. Overall, our findings show that analytic flexibility substantially affect reported results. In this talk, I will present the background for this unique project, highlight the main findings and discuss implications and potential solutions.

Thurs, Feb 13

Tommy Botch, Dartmouth College

How colorful is visual experience? Evidence from gaze-contingent virtual reality

Color ignites visual experience, imbuing the world with meaning, emotion, and richness. Intuitively, it feels that we are immersed in a colorful world that extends to the farthest limits of our periphery. In this talk, I will present a series of studies in which we show that this impression is surprisingly inaccurate. We used gaze-contingent rendering in immersive virtual reality (VR) to reveal the limits of color awareness during active, real-world visual experience by systematically altering the visual environments such that only the parts of the scene they were looking at were presented in color and the rest of the scene (i.e., the visual periphery) was entirely desaturated. 

In Study 1, we found that observers were often completely unaware of these drastic alterations to their visual world. In the most extreme case, almost a third of observers failed to notice when less than 5% of the visual display was presented in color. In Study 2, we measured color detection thresholds using a staircase procedure while a new set of observers explicitly attended to the periphery. Still, we found that observers were unaware when a large portion of their field of view was desaturated. In a Study 3, we confirm that individual differences in detection thresholds are test-retest reliable within participants..

In brief, our results provide the first measurements of color awareness during active, naturalistic viewing conditions and show that our intuitive sense of a rich, colorful visual world is largely incorrect. I’ll conclude by briefly describing several applications of this approach to the study of psychiatric conditions such as autism.

Thurs, Feb 20

Giancarlo La Camera, Stony Brook University

Cortical computations via metastable activity

Metastable brain dynamics are characterized by abrupt, jump-like modulations so that the neural activity in single trials appears to unfold as a sequence of discrete, quasi-stationary ‘states’. Metastable activity occurs both in response to an external stimulus and during ongoing, self-generated activity. These spontaneous metastable states are increasingly found to subserve internal representations that are not locked to external triggers, including states of deliberations, attention and expectation. Focusing on metastability allows us to shift our perspective on neural coding from traditional concepts based on trial-averaging to models based on dynamic ensemble representations. In this talk, I will present some recently established links between metastable dynamics, expectation and decision making, together with a theoretical proposal of how these links may emerge in networks of spiking neurons. 

Thurs, Feb 27

Mohsen Rakhshan, Dartmouth College

Neural substrate underlying computations of volatility and adaptive learning in a complex environment

In value-based decision making, integration of reward outcomes over time is essential. When an unpredicted reward outcome occurs, a decision-maker must be able to distinguish between the uncertainty due to the stochastic nature of events (variability) and the uncertainty due to the changes in the environment (volatility). Although variability should not lead to behavioral adjustments, volatility should adjust the learning. While volatility can be specific to a stimulus (or action), it is currently unclear how humans perceive the level of volatility of the stimulus (action) in a complex environment where multiple stimuli (actions) with different levels of volatility are interacting. Here, we developed a novel concurrent probabilistic reversal learning task to study how learning from reward feedback is influenced by the interaction of volatility of different stimuli and their associated actions. Furthermore, we propose that a network of reward-dependent metaplasticity can provide a plausible mechanism for both integration of reward under uncertainty and estimation of uncertainty in the complex environment. 

Bogdan Petre, Dartmouth College

Evoked pain intensity representation is distributed throughout the brain 

Information is coded in brain activity at different scales: locally, distributed across regions and networks, and globally. For pain, the scale of representation is controversial, and quantitative characterizations of spatial information distribution are lacking. Although generally believed to be an integrated cognitive and sensory phenomenon implicating diverse brain systems, both local and global representations are invoked to explain pain physiology. In this person-level meta-analysis (or mega-analysis) of data from 289 participants across 10 studies, we use model comparison combined with multivariate predictive models to investigate the spatial scale and location of acute pain representation. We compare models based on (a) a single most pain-predictive region, identified in a data driven manner; (b) a single best large-scale cortical resting-state network; (c) selected cortical-subcortical systems related to evoked pain in prior literature (‘multi-system models’, including Neurosynth.org); and (d) a model spanning the full brain. We estimate the accuracy of pain intensity predictions using cross validation (7 studies) and subsequently validate in three independent holdout studies. All spatial scales convey information about pain intensity, but distributed, multi-system models better characterize pain representations than individual regions or networks. Full brain models showed no predictive advantage over multi-system models with feature selection guided by previous literature. These findings suggest that the representation of evoked pain experience is distributed across multiple cortical and subcortical systems. They also provide a blueprint for identifying the spatial scale of information in other domains.