Theories of consciousness can be categorized as either first-order or higher-order. Higher-order (HO) theories originally emerged in philosophy, and have now differentiated to include a number of different variants that each make distinct predictions. First order (FO) theories propose that consciousness is tied to the properties or strength of a FO perceptual representation (e.g. in global workspace theory, consciousness is determined by whether a FO representation is globally broadcast). In contrast, HO theories propose that consciousness depends on a FO state being in some way monitored or meta-represented by a HO representation. However, unlike the active research fields pursuing empirical tests of FO theories, HO theories have received relatively little direct empirical investigation in the neurosciences. They are also often lumped into a single “higher-order theory” when in fact, there is just as much differentiation and disagreement in the higher-order camp as there is in the first-order camp. A project led by Stephen Fleming at University College London, co-directed by Axel Cleeremans at Université Libre de Bruxelles aims to harness the power of adversarial collaboration to test between distinct predictions of higher-order theories. It will focus on two “axes” of disagreement between HOT variants.

One axis examines to what degree HO representations are rich or sparse (note here rich vs. sparse refers to the neurocomputational architecture supporting phenomenal experience, not whether or not perceptual experience is itself rich or sparse). Sparse HOTs propose that phenomenal experience is jointly determined by FO and HO states, with HO states playing a lean role in tracking the precision, intensity or reliability of FO states. However, even within sparse theories there is disagreement as to the nature of the HO code – for instance, whether it distinguishes between perception and imagination. In contrast, rich HO theories propose that phenomenal experience is fully determined by HO states, such that HO representations are just as rich and detailed as perceptual experience itself.

A second (related) axis of disagreement is on whether HO representations can “misrepresent” their FO targets, and in what way. For instance, a subject may be conscious of the color green even if their first-order perceptual system is signaling strong evidence for red. The possibility of misrepresentation is central to non-relational HOTs: if there are mismatches between FO and HO states, conscious experience changes in tandem with the HO state. Strong misrepresentation is precluded under relational theories under which the FO state supplies the mental content which is monitored by a HO index.

The two axes of disagreement will be tested in distinct experiments with a number of convergent methodologies, including psychophysics, neuroimaging, and hypnotic suggestion. The project team follow the high standards set by previous adversarial collaborations in collecting large samples to allow novel findings to be replicated on a hold-out dataset using identical procedures, reducing the risk of obtaining false positives in high-dimensional brain and behavioral data. All of the team's experimental procedures, datasets and analysis tools developed as part of this project will be made freely available to other researchers and adhere to gold standards in open science. This feature of the project will facilitate future discovery science aimed at further evaluating higher-order (and other) theories of consciousness.