Highlights d Prefrontal activity in monkeys performing a distance discrimination task is metastable d Duration of metastable states is longer before errors d Latency of state transition is longer in correct difficult trials d States may code for relative distance based on stimulus features or presentation order

Previous studies have established the involvement of prefrontal cortex (PFC) neurons in decision processes in many task contexts. Single neurons and populations of neurons have been found to represent stimuli, actions, and internal deliberations. However, it is much less clear which underlying computations are affected during errors. Neural activity during errors can help to disambiguate confounds and clarify which computations are essential during a specific task. Here, we used a hidden Markov model (HMM) to perform a trial-by-trial analysis of ensembles of simultaneously recorded neurons from the dorsolateral prefrontal (PFdl) cortex of two rhesus monkeys performing a distance discrimination task. The HMM segments the neural activity into sequences of metastable states, allowing to link neural ensemble dynamics with task and behavioral features in the absence of external triggers. We report a precise relationship between the modulation of the metastable dynamics and task features. Specifically, we found that errors were made more often when the metastable dynamics slowed down, while trial difficulty influenced the latency of state transitions at a pivotal point during the trial. Both these phenomena occurred during the decision interval and not following the action, with errors occurring in both easy and difficult trials. Thus, modulations of metastable dynamics reflected a state of internal deliberation rather than actions taken or, in the case of error trials, objective trial difficulty. Our results show that temporal modulations of PFdl activity are key determinants of internal deliberations, providing further support for the emerging role of metastable cortical dynamics in mediating complex cognitive functions and behavior.

No abstract

Brain effective connectivity aims to detect causal interactions between distinct brain units and it is typically studied through the analysis of direct measurements of the neural activity, e.g., magneto/electroencephalography (M/EEG) signals. The literature on methods for causal inference is vast. It includes model-based methods in which a generative model of the data is assumed and model-free methods that directly infer causality from the probability distribution of the underlying stochastic process. Here, we firstly focus on the model-based methods developed from the Granger criterion of causality, which assumes the autoregressive model of the data. Secondly, we introduce a new perspective, that looks at the problem in a way that is typical of the machine learning literature. Then, we formulate the problem of causality detection as a supervised learning task, by proposing a classification-based approach. A classifier is trained to identify causal interactions between time series for the chosen model and by means of a proposed feature space. In this paper, we are interested in comparing this classification-based approach with the standard Geweke measure of causality in the time domain, through simulation study. Thus, we customized our approach to the case of a MAR model and designed a feature space which contains causality measures based on the idea of precedence and predictability in time. Two variations of the supervised method are proposed and compared to a standard Granger causal analysis method. The results of the simulations show that the supervised method outperforms the standard approach, in particular it is more robust to noise. As evidence of the efficacy of the proposed method, we report the details of our submission to the causality detection competition of Biomag2014, where the proposed method reached the 2nd place. Moreover, as empirical application, we applied the supervised approach on a dataset of neural recordings of rats obtaining an important reduction in the false positive rate.

Brain effective connectivity aims to detect causal interactions between distinct brain units and it can be studied through the analysis of magneto/electroencephalography (M/EEG) signals. Methods to evaluate effective connectivity belong to the large body of literature related to detecting causal interactions between multivariate autoregressive (MAR) data, a field of signal processing. Here, we reformulate the problem of causality detection as a supervised learning task and we propose a classification-based approach for it. Our solution takes advantage of the MAR model by generating a labeled data set that contains trials of multivariate signals for each possible configuration of causal interactions. Through the definition of a proper feature space, a classifier is trained to identify the causality structure within each trial. As evidence of the efficacy of the proposed method, we report both the cross-validated results and the details of our submission to the causality detection competition of Biomag2014, where the method reached the 2nd place.

In neurophysiology, nonhuman primates represent an important model for studying the brain. Typically, monkeys are moved from their home cage to an experimental room daily, where they sit in a primate chair and interact with electronic devices. Refining this procedure would make the researchers’ work easier and improve the animals’ welfare. To address this issue, we used home-cage training to train two macaque monkeys in a non-match-to-goal task, where each trial required a switch from the choice made in the previous trial to obtain a reward. The monkeys were tested in two versions of the task, one in which they acted as the agent in every trial and one in which some trials were completed by a “ghost agent”. We evaluated their involvement in terms of their performance and their interaction with the apparatus. Both monkeys were able to maintain a constant involvement in the task with good, stable performance within sessions in both versions of the task. Our study confirms the feasibility of home-cage training and demonstrates that even with challenging tasks, monkeys can complete a large number of trials at a high performance level, which is a prerequisite for electrophysiological studies of monkey behavior.

How the emergent functional connectivity (FC) relates to the underlying anatomy (structural connectivity, SC) is one of the biggest questions of modern neuroscience. At the macro-scale level, no one-to-one correspondence between structural and functional links seems to exist. And we posit that to better understand their coupling, two key aspects should be taken into account: the directionality of the structural connectome and the limitations of describing network functions in terms of FC. Here, we employed an accurate directed SC of the mouse brain obtained by means of viral tracers, and related it with single-subject effective connectivity (EC) matrices computed by applying a recently developed DCM to whole-brain resting-state fMRI data. We analyzed how SC deviates from EC and quantified their couplings by conditioning both on the strongest SC links and EC links. We found that when conditioning on the strongest EC links, the obtained coupling follows the unimodal-transmodal functional hierarchy. Whereas the reverse is not true, as there are strong SC links within high-order cortical areas with no corresponding strong EC links. This mismatch is even more clear across networks. Only the connections within sensory motor networks align both in terms of effective and structural strength.

Our representation of magnitudes such as time, distance, and size is not always veridical because it is affected by multiple biases. From a Bayesian perspective, estimation errors are considered to be the result of an optimization mechanism for the behavior in a noisy environment by integrating previous experience with the incoming sensory information. One influence of the distribution of past stimuli on perceptual decisions is represented by the regression toward the mean, a type of contraction bias. Using a spatial discrimination task with 2 stimuli presented sequentially at different distances from the center, we show that this bias is also present in macaques when comparing the magnitude of 2 distances. We found that the contraction of the first stimulus magnitude toward the center of the distribution accounted for some of the changes in performance, even more so than the effect of difficulty related to the ratio between stimulus magnitudes. At the neural level in the dorsolateral prefrontal cortex, the coding of the decision after the presentation of the second stimulus reflected the effect of the contraction bias on the discriminability of the stimuli at the behavioral level.