A central question for cognitive neuroscience is how feature-combinations that give rise to episodic/source memories are encoded in the brain. Although there is much evidence that the hippocampus (HIP) is involved in feature binding, and some evidence that other brain regions are as well, there is relatively little evidence about the nature of the resulting representations in different brain regions. We used multivoxel pattern analysis (MVPA) to investigate how feature combinations might be represented, contrasting two possibilities, feature-based versus holistic. Participants viewed stimuli that were composed of three source features - a person (face or body), a scene (indoor or outdoor), and an object (bike or luggage) - which were combined to make eight unique stimulus identities. We reasoned that regions that can classify the eight identities (a multiclass classification) but not the individual features (a binary classification) likely have a holistic representation of each identity. In contrast, regions that can classify the eight identities and can classify each feature are likely to contain feature-based representations of these identities. To further probe the extent of feature-based or holistic classification in each region, we developed and validated a novel approach that directly compares binary and multiclass classification. We found clear evidence for holistic representation in the parahippocampal cortex (PHC), consistent with theories that posit that pattern-separation-like binding mechanisms are not unique to the HIP. Further clarifying the mechanisms of feature binding should benefit from systematic comparisons of multi-feature representations and whether they vary with task, type of stimulus, and/or experience.

Achieving a mechanistic explanation of brain function requires understanding causal relationships among regions. A relatively new technique to assess effective connectivity in fMRI data is Dynamic Causal Modeling (DCM). As DCM is more frequently used, it becomes increasingly important to further validate the technique and understand its limitations. With DCM, Bayesian Model Selection (BMS) is used to select the most likely causal model. We conducted simulations to test the degree to which BMS is robust to two types of challenges when applied to DCMs, those inherent to data ( Category 1 ) and those inherent to model space ( Category 2 ). Category 1 challenges tested properties of the data (low signal-to-noise, different response magnitudes and shapes across regions) that could either blur the distinction between models or potentially bias model selection. These challenges are impossible or difficult to measure and control in real data, so investigating their effect upon BMS through simulation is critical. Category 2 challenges tested properties of model space that create subsets of confusable models. Our results suggest that given data that conform to the prior assumptions of DCM, BMS is robust to challenges from Category 1 . However, in the face of Category 2 challenges (when a more homogenous model space was tested) the false positive rate rose above an acceptable level. We show that such errors are neither trivial nor easily avoided with existing approaches. However, we argue that it is possible to detect Category 2 challenges, and avoid inappropriate interpretations by conducting simulations prior to applying DCM.DCM: Dynamic Causal ModelingBMS: Bayesian Model SelectionfMRI: functional magnetic resonance imagingBOLD: blood oxygen level dependentFMC: Family Model ComparisonHRF: hemodynamic response functionROI: region of interestSNR: signal to noise ratioR1: region 1R2: region 2U1: input 1

Episodic memory is characterized by remembering events as unique combinations of features. Even when some features of events overlap, we are later often able to discriminate among them. Here we ask whether hippocampally mediated reactivation of an earlier event when a similar one occurs supports subsequent memory that two similar but not identical events occurred (mnemonic discrimination). In two experiments, participants viewed objects (Experiment 1) or scenes (Experiment 2) during functional MRI (fMRI). After scanning, participants had to remember whether repeated items had been identical or similar. In Experiment 2, representational similarity between the 1st and 2nd presentation predicted participants' ability to remember that the presentations were different, suggesting that the first item was reactivated while viewing the second. A similar but weaker result was found in Experiment 1 that did not survive correction for multiple comparisons. Furthermore, both experiments yielded evidence that the hippocampus was involved in reactivation; hippocampal pattern similarity (and, in Experiment 2, hippocampal activity during the 2nd presentation) correlated with pattern similarity in several regions of visual cortex. These results provide the first fMRI evidence that hippocampally mediated reactivation contributes to the later memory that two similar, but different events occurred. © 2016 Wiley Periodicals, Inc.

Decades of research have demonstrated that a region of the right fusiform gyrus (FG) and right posterior superior temporal sulcus (pSTS) responds preferentially to static faces and biological motion, respectively. Despite this view, both regions activate in response to both stimulus categories and to a range of other stimuli, such as goal-directed actions, suggesting that these regions respond to characteristics of animate agents more generally. Here we propose a neural model for animacy detection composed of processing streams that are initially differentially sensitive to cues signaling animacy, but that ultimately act in concert to support reasoning about animate agents. We use dynamic causal modeling, a measure of effective connectivity, to demonstrate that the directional flow of information between the FG and pSTS is initially dependent on the characteristics of the animate agent presented, a key prediction of our proposed network for animacy detection.