Search

A study shows the transience of early visual representations (while the stimulus is still on) and the persistence of higher representations (outlasting the stimulus) as various categorical distinctions emerge at staggered latencies. Rather than slavishly following the stimulus, representations interact through recurrent signals to infer what's there.

Peters and Kriegeskorte review the behavioural and neural-network-modelling literature on object-based visual representations. They call for new tasks that will bridge research in cognitive sciences and engineering in this domain.

Developing a better understanding of neural codes should enable the links between stimuli, brain activity and behaviour to become clearer. In this Perspective, Kriegeskorte and Wei examine neural tuning and representational geometry — complementary approaches used to understand neural codes — and the relationship between them.

An EEG, fMRI, and behavioural investigation of five object dimensions related to animacy provides insights into how distinct dimensions (and animacy) are represented in the brain and behaviour.

With the advances in neural language models, the question arises if some models align better with human processing than others. Golan et al. identify sentences that language models disagree about and use them to compare the shortcomings of different language models.

Are vocal emotions perceived as categories (for example, anger) or as dimensions (for example, an intense negative emotion)? Here, using functional magnetic resonance imaging and magneto-encephalography, Giordano et al. find that categories prevail in early (less than 200 ms) neural representations, whereas dimensions emerge later (240 ms and more than 500 ms).

Previous research shows how taste types are represented across regions of the brain in non-human animals. Here, the authors examine how four basic tastes are represented in the human brain, showing evidence of the human gustatory cortex in the insula.

Do artificial neural networks, like brains, exhibit individual differences? Using tools from systems neuroscience, this study reveals substantial variability in network-internal representations, calling into question the neuroscientific practice of using single networks as models of brain function.

The authors found human neuroimaging evidence that entire valence spectrum is represented as a collective pattern in regional neural activity, with sensory-specific signals in the ventral temporal and anterior insular cortices and abstract codes in the orbitofrontal cortices. In this way, the subjective quality of affect can be objectively quantified across stimuli, modalities and people.

Forgetting can at times serve an adaptive purpose. Here the authors develop a method for dynamically tracking neocortical activity patterns related to the retrieval of individual episodic memories. They show that remembering gradually enhances relevant memories but also suppresses the cortical traces of interfering memories, causing adaptive forgetting.

Responding to widespread concerns about reproducibility, the Organization for Human Brain Mapping created a working group to identify best practices in data analysis, results reporting and data sharing to promote open and reproducible research in neuroimaging. We describe the challenges of open research and the barriers the field faces.

The authors review recent work at the intersection of cognitive science, computational neuroscience and artificial intelligence that develops and tests computational models mimicking neural and cognitive function during a wide range of tasks.

This perspective illustrates some of the problems involved in analyzing the complex data yielded by systems neuroscience techniques, such as brain imaging and electrophysiology. Specifically, when test statistics are not independent of the selection criteria, common analyses can produce spurious results. The authors suggest ways to avoid such errors.

A deep network is best understood in terms of components used to design it—objective functions, architecture and learning rules—rather than unit-by-unit computation. Richards et al. argue that this inspires fruitful approaches to systems neuroscience.

Artificial neural networks are being widely used to model behavioural and neural data. In this Perspective article, Doerig et al. present neuroconnectionism as a Lakatosian research programme using artificial neural networks as a computational language for expressing falsifiable theories and hypotheses about the brain computations underlying cognition.