Search

The proposed study is part of a methods pilot that aims to generate sequences of mathematical transforms involved in early sensory and language processing, using the methods from Thwaites et al (2016).

Neural codes are reflected in complex neural activation patterns. Conventional electroencephalography (EEG) decoding analyses summarize activations by averaging/down-sampling signals within the analysis window. This diminishes informative fine-grained patterns. While previous studies have proposed distinct statistical features capable of capturing variability-dependent neural codes, it has been suggested that the brain could use a combination of encoding protocols not reflected in any one mathematical feature alone. To check, we combined 30 features using state-of-the-art supervised and unsupervised feature selection procedures (n = 17). Across three datasets, we compared decoding of visual object category between these 17 sets of combined features, and between combined and individual features. Object category could be robustly decoded using the combined features from all of the 17 algorithms. However, the combination of features, which were equalized in dimension to the individual features, were outperformed across most of the time points by the multiscale feature of Wavelet coefficients. Moreover, the Wavelet coefficients also explained the behavioral performance more accurately than the combined features. These results suggest that a single but multiscale encoding protocol may capture the EEG neural codes better than any combination of protocols. Our findings put new constraints on the models of neural information encoding in EEG.

Brain connectivity analyses have conventionally relied on statistical relationship between one-dimensional summaries of activation in different brain areas. However, summarizing activation patterns within each area to a single dimension ignores the potential statistical dependencies between their multi-dimensional activity patterns. Representational Connectivity Analyses (RCA) is a method that quantifies the relationship between multi-dimensional patterns of activity without reducing the dimensionality of the data. We consider two variants of RCA. In model-free RCA, the goal is to quantify the shared information for two brain regions. In model-based RCA, one tests whether two regions have shared information about a specific aspect of the stimuli/task, as defined by a model. However, this is a new approach and the potential caveats of model-free and model-based RCA are still understudied. We first explain how model-based RCA detects connectivity through the lens of models, and then present three scenarios where model-based and model-free RCA give discrepant results. These conflicting results complicate the interpretation of functional connectivity. We highlight the challenges in three scenarios: complex intermediate models, common patterns across regions, and transformation of representational structure across brain regions. The article is accompanied by scripts (https://osf.io/3nxfa/) that reproduce the results. In each case, we suggest potential ways to mitigate the difficulties caused by inconsistent results. The results of this study shed light on some understudied aspects of RCA, and allow researchers to use the method more effectively.

Work published as: Goddard E, Carlson T, Woolgar A (2021) Spatial and feature-selective attention have distinct, interacting, effects on population-level tuning. Journal of Cognitive Neuroscience (in press Nov 2021). DOI: https://doi.org/10.1162/jocn_a_01796. Please cite this paper if you use the data on this project. This project contains a magnetoencephalography (MEG) data set. Participants (n = 20) made judgements about visual stimuli in different attention conditions. For details of the folder structure and how to use this data see the project wiki. For details of the stimuli and methodology, see the paper above.

Here we share the information including the scripts and data about our papers on representational connectivity analysis. Please cite the below preprint if you use the material shared here: Karimi-Rouzbahani, H., Woolgar, A., Henson, R. and Nili, H., (2021) Caveats and nuances of model-based and model-free representational connectivity analysis. BioRxiv. https://doi.org/10.1101/2021.08.10.455841

This project investigated whether we could use 'fast periodic visual stimulation' (FPVS) and EEG to detect lexical processing in individuals. You can find the project manuscript [here]. We showed rapid streams of visual stimuli to 10 healthy adults, while recording EEG with a 64-channel Biosemi system. The stimulus streams consisted of faces embedded among scrambled faces, or words embedded among pseudowords. A new stimulus appeared every 100 ms (10 Hz), with the embedded stimulus type appearing every 5th stimulus (2 Hz). Following Lochy et al. (2015), we expected that healthy adults would show face- and word-selective responses (ie, a spike at 2 Hz) around parieto-occipital electrode sites. We observed this clearly for faces among scrambled faces. For words embedded among pseudowords, effects were rare and small. The OSF project folder contains presentation scripts (/testing), raw EEG files (/data), and analysis code (/analysis) for the wordFPVS project.

Studies of lexico-semantic processing (i.e. the processing of the meaning of stimuli in their context) provide insights into how the brain recognises and integrates visual and linguistic stimuli to make sense of the world around us. Electroencephalography (EEG), with its fine temporal resolution, is ideal to study the temporal unfolding of semantic processing in response to various stimuli. In particular, the N400 event-related potential (ERP) reflects the difficulty of integrating the meaning of a stimulus with its semantic context. Traditionally, stimuli that are incongruent with the semantic context elicit more negative N400 ERPs relative to congruent stimuli. The N400 has been extensively studied in groups of adults, children, and special population, using a variety of auditory (spoken words, sentences) and visual (images, written words) stimuli. However, much less is known how robust these effects are at the individual level. This has important implications, both theoretically regarding semantic integration, and for the clinical evaluation of semantic processing in populations that may have hidden cognitive abilities, such as people with non-verbal autism or disorders of consciousness. Recent studies suggest a large heterogeneity of N400 responses across individuals. Notably, these studies primarily used spoken word stimuli, which are inherently extended in time. The recognition point varies between spoken items and across people which might contribute to difficulty in detecting time-locked effects. In the current study, we bypassed this issue by using words to establish a semantic context and prediction, but testing N400 responses to matching and mismatching pictures. We predicted that this would allow us to detect N400-like responses more reliably than reported for spoken words and sought to quantify individual-detection rates for this paradigm in healthy adults. We plan to examine the EEG response to identical pictures when they are preceded by either matched or mismatched written words, using both multivariate and univariate analyses at individual and group levels. We will also evaluate the quality of data recorded by the consumer-grade EEG system EPOC+, by comparing it with simultaneous recording from a research-grade Brain Products actiCAP/actiCHampSystem.

Computer-aided detection (CAD) systems generally use a binary cue to indicate a target in an image to an observer (Cunningham, Drew, & Wolfe, 2017). This means that the CAD system provides a marker around (or an arrow pointing towards) any location in an image where the strength of the signal is above a certain threshold and does not provide a marker for those locations where the signal is below the threshold. CAD systems are not completely accurate. In target-present trials, the CAD system may correctly mark the target only (hit), however it may also commit three errors: It may correctly mark the target and incorrectly mark one or more distractors (hit / false alarm), fail to mark the target and incorrectly mark one or more distractors (miss / false alarm), or fail to mark the target and correctly not mark any distractors (miss). In target-absent trials, the CAD may correctly not mark any distractors (correct rejection), or incorrectly mark one or more distractors (false alarm). CAD systems that use binary cues have been shown to lead to increased detection (by the observer) of targets that are marked by the CAD, but they also lead to decreased detection of targets that are missed by the CAD, especially when a distractor has also been cued (Drew, Guthrie, & Reback, 2020; see also Russell & Kunar, 2012 for the exogenous cueing of low prevalence targets). When the CAD system does not provide any cue in a target-present trial (miss), the increased misses of these targets could be due to an overreliance of the observer on the information that the CAD conveys (the overreliance hypothesis; Kunar Watson, Taylor-Phillips, & Wolska, 2017). When the CAD system fails to cue a target and also cues a distractor (miss / false alarm) the increased misses of these targets could be due to an overreliance on the CAD and/or because the exogenous CAD cue(s) capture the observer’s attention to specific locations in the image, drawing it away from non-cued locations that may contain a target. Here, we will test the overreliance hypothesis by removing the localised attentional capture aspect of the CAD cues. We will use non-spatial CAD cues to convey the CAD information to the observer through a coloured border around the image when a target is detected by the CAD system. These cues will provide CAD information to the observer without capturing attention to specific locations. These cues differ from spatial cues in the errors that the system can make. When a target is present and the system provides a cue, it can no longer make the hit / false alarm or miss / false alarm errors. Participants will complete a letter search task under two conditions: with non-spatial CAD cues and without. We predict that participants will perform better (than without CAD) when the non-spatial CAD correctly cues a target (hit), demonstrating that participants are using the CAD information. We also predict that participants will perform worse (than without CAD) when the non-spatial CAD fails to cue a target (miss), thus providing support for the overreliance hypothesis. If participants perform better on CAD hit trials, but no worse on CAD miss trials, both relative to without CAD, this suggests that attentional capture may be responsible for some of the errors in spatial CAD.

Attention can be deployed in different ways: When searching for a taxi in New York City, we can decide where to attend (e.g., to the street) and what to attend to (e.g., yellow cars). Although we use the same word to describe both processes, nonhuman primate data suggest that these produce distinct effects on neural tuning. This has been challenging to assess in humans, but here we used an opportunity afforded by multivariate decoding of MEG data. We found that attending to an object at a particular location and attending to a particular object feature produced effects that interacted multiplicatively. The two types of attention induced distinct patterns of enhancement in occipital cortex, with feature-selective attention producing relatively more enhancement of small feature differences and spatial attention producing relatively larger effects for larger feature differences. An information flow analysis further showed that stimulus representations in occipital cortex were Granger-caused by coding in frontal cortices earlier in time and that the timing of this feedback matched the onset of attention effects. The data suggest that spatial and feature-selective attention rely on distinct neural mechanisms that arise from frontal-occipital information exchange, interacting multiplicatively to selectively enhance task-relevant information.

Bernstein Conference 2021 abstract. http://bernstein-conference.de