A recent paper claims that a newly proposed method classifies EEG data recorded from subjects viewing ImageNet stimuli better than two prior methods. However, the analysis used to support that claim is based on confounded data. We repeat the analysis on a large new dataset that is free from that confound. Training and testing on aggregated supertrials derived by summing trials demonstrates that the two prior methods achieve statistically significant above-chance accuracy while the newly proposed method does not.

Neuroimaging experiments in general, and EEG experiments in particular, must take care to avoid confounds. A recent TPAMI paper uses data that suffers from a serious previously reported confound. We demonstrate that their new model and analysis methods do not remedy this confound, and therefore that their claims of high accuracy and neuroscience relevance are invalid.

New results suggest strong limits to the feasibility of object classification from human brain activity evoked by image stimuli, as measured through EEG. Considerable prior work suffers from a confound between the stimulus class and the time since the start of the experiment. A prior attempt to avoid this confound using randomized trials was unable to achieve results above chance in a statistically significant fashion when the data sets were of the same size as the original experiments. Here, we attempt object classification from EEG using an array of methods that are representative of the state-of-the-art, with a far larger (20×) dataset of randomized EEG trials, 1,000 stimulus presentations of each of forty classes, all from a single subject. To our knowledge, this is the largest such EEG data-collection effort from a single subject and is at the bounds of feasibility. We obtain classification accuracy that is marginally above chance and above chance in a statistically significant fashion, and further assess how accuracy depends on the classifier used, the amount of training data used, and the number of classes. Reaching the limits of data collection with only marginally above-chance performance suggests that the prevailing literature substantially exaggerates the feasibility of object classification from EEG.

A recent paper [31] claims to classify brain processing evoked in subjects watching ImageNet stimuli as measured with EEG and to employ a representation derived from this processing to construct a novel object classifier. That paper, together with a series of subsequent papers [11, 18, 20, 24, 25, 30, 34], claims to achieve successful results on a wide variety of computer-vision tasks, including object classification, transfer learning, and generation of images depicting human perception and thought using brain-derived representations measured through EEG. Our novel experiments and analyses demonstrate that their results crucially depend on the block design that they employ, where all stimuli of a given class are presented together, and fail with a rapid-event design, where stimuli of different classes are randomly intermixed. The block design leads to classification of arbitrary brain states based on block-level temporal correlations that are known to exist in all EEG data, rather than stimulus-related activity. Because every trial in their test sets comes from the same block as many trials in the corresponding training sets, their block design thus leads to classifying arbitrary temporal artifacts of the data instead of stimulus-related activity. This invalidates all subsequent analyses performed on this data in multiple published papers and calls into question all of the reported results. We further show that a novel object classifier constructed with a random codebook performs as well as or better than a novel object classifier constructed with the representation extracted from EEG data, suggesting that the performance of their classifier constructed with a representation extracted from EEG data does not benefit from the brain-derived representation. Together, our results illustrate the far-reaching implications of the temporal autocorrelations that exist in all neuroimaging data for classification experiments. Further, our results calibrate the underlying difficulty of the tasks involved and caution against overly optimistic, but incorrect, claims to the contrary.

Significant scientific and translational questions remain in auditory neuroscience surrounding the neural correlates of perception. Relating perceptual and neural data collected from humans can be useful; however, human-based neural data are typically limited to evoked far-field responses, which lack anatomical and physiological specificity. Laboratory-controlled preclinical animal models offer the advantage of comparing single-unit and evoked responses from the same animals. This ability provides opportunities to develop invaluable insight into proper interpretations of evoked responses, which benefits both basic-science studies of neural mechanisms and translational applications, e.g., diagnostic development. However, these comparisons have been limited by a disconnect between the types of spectrotemporal analyses used with single-unit spike trains and evoked responses, which results because these response types are fundamentally different (point-process versus continuous-valued signals) even though the responses themselves are related. Here, we describe a unifying framework to study temporal coding of complex sounds that allows spike-train and evoked-response data to be analyzed and compared using the same advanced signal-processing techniques. The framework uses a set of peristimulus-time histograms computed from single-unit spike trains in response to polarity-alternating stimuli to allow advanced spectral analyses of both slow (envelope) and rapid (temporal fine structure) response components. Demonstrated benefits include: (1) novel spectrally specific temporal-coding measures that are less confounded by distortions due to hair-cell transduction, synaptic rectification, and neural stochasticity compared to previous metrics, e.g., the correlogram peak-height, (2) spectrally specific analyses of spike-train modulation coding (magnitude and phase), which can be directly compared to modern perceptually based models of speech intelligibility (e.g., that depend on modulation filter banks), and (3) superior spectral resolution in analyzing the neural representation of nonstationary sounds, such as speech and music. This unifying framework significantly expands the potential of preclinical animal models to advance our understanding of the physiological correlates of perceptual deficits in real-world listening following sensorineural hearing loss.