Controlling for performance capacity confounds in neuroimaging studies of conscious awareness

Abstract

Studying the neural correlates of conscious awareness depends on a reliable comparison between activations associated with awareness and unawareness. One particularly difficult confound to remove is task performance capacity, i.e. the difference in performance between the conditions of interest. While ideally task performance capacity should be matched across different conditions, this is difficult to achieve experimentally. However, differences in performance could theoretically be corrected for mathematically. One such proposal is found in a recent paper by Lamy, Salti and Bar-Haim [Lamy D, Salti M, Bar-Haim Y. Neural correlates of subjective awareness and unconscious processing: an ERP study. J Cognitive Neurosci 2009,21:1435-46], who put forward a corrective method for an electroencephalography experiment. We argue that their analysis is essentially grounded in a version of High Threshold Theory, which has been shown to be inferior in general to Signal Detection Theory. We show through a series of computer simulations that their correction method only partially removes the influence of performance capacity, which can yield misleading results. We present a mathematical correction method based on Signal Detection Theory that is theoretically capable of removing performance capacity confounds. We discuss the limitations of mathematically correcting for performance capacity confounds in imaging studies and its impact for theories about consciousness.

Initial-state-dependent, robust, transient neural dynamics encode conscious visual perception

“…One concern in interpreting the above results is that seen and unseen trials differ in aspects other than subjective perception of the stimulus, such as performance in the orientation discrimination task. To control for objective performance, we re-conducted the analysis using correct trials only, which yielded nearly identical results ( Fig 5B ), suggesting that the decoding results were not driven by the categorical distinction between correct and incorrect trials (but see [ 34 ]). In addition, to ensure that the significant decoding result in the pre-stimulus period was not confounded by the (<5 Hz) low-pass filter’s spreading of post-stimulus activity into the pre-stimulus period, we repeated the above analysis using a moving-average window applied to full-band data.…”

Initial-state-dependent, robust, transient neural dynamics encode conscious visual perception

“…To control for objective performance, we re-conducted the analysis using correct trials only, which yielded nearly identical results ( Fig. 7B), suggesting that the decoding results were not driven by the categorical distinction between correct and incorrect trials (but see [33]). In addition, to ensure that the significant decoding result in the pre-stimulus period was not confounded by the (<5 Hz) low-pass filter's spreading of post-stimulus activity into the prestimulus period, we repeated the above analysis using full-band data.…”

Hard criteria for empirical theories of consciousness

“…Even though problems with this contrastive approach have been highlighted recently (Aru et al, 2012;Balsdon & Clifford, 2018;Kleiner & Hoel, 2020;Lau, 2008;Peters & Lau, 2015), all that is needed is the existence of conscious and unconscious states, and methods to contrast them (but see Salti et al, 2019, for an argument that the conscious vs. unconscious dichotomy is misguided). For visual masking and other paradigms, specific procedures have been established to differentiate between conscious and unconscious perception (Morales et al, 2015;Overgaard et al, 2010;Schmidt & Vorberg, 2006).…”