Can the Motivated Emotional Mind model explain phenomenal consciousness beyond GNWT?

Two theories, one set of findings

Global Neuronal Workspace Theory (GNWT) holds that a content becomes consciously accessible when it is selected, maintained, and broadcast widely to systems for memory, decision-making, report, and action, a transition tied to nonlinear 'ignition' of a distributed network and long-range recurrent communication. The Motivated Emotional Mind (MEM) model instead proposes an embodied, semi-hierarchical associative memory in which representational selection, action, recurrent reconstruction of modality-specific fields, interoception, and valence form a single functional cycle. The paper asks three separate questions: whether MEM can reproduce the access functions GNWT explains, whether its mechanisms are genuinely distinct, and whether any existing findings discriminate between them.

The authors are explicit that breadth of scope is not superiority. GNWT has roughly four decades of psychophysical, electrophysiological, neuroimaging, clinical, and computational work behind it; MEM is younger, and some of its distinctive hypotheses have not been tested in studies designed to falsify them. That asymmetry shapes the whole analysis: the paper's conclusion is framed as a hypothesis of greater explanatory potential, not a declaration that MEM has won.

Twenty-five studies rated on sixteen criteria

The corpus contains 25 publications. Twelve are canonical GNWT studies, computational models, or critiques; because that literature says little about interoception, the authors added thirteen studies on bodily signals, affect, heart–brain coupling, respiration, facial expression, and cortical stimulation. The second group was not designed as a test of MEM and, the authors state, cannot be treated as direct confirmation of the theory.

Each publication was rated from 1 to 6 (6 strong support, 1 strong contradiction) against sixteen theory-neutral criteria spanning phenomenal consciousness (P1–P4), access consciousness (A1–A4), system efficacy (S1–S4), and biological plausibility (B1–B4). Empirical studies and meta-analyses received weight 1, hybrid models with biological validation 0.75, computational models 0.5, and theoretical papers or reviews 0.25. Three quantities were then computed per criterion: M, the theory's stated explanatory scope; E, the weighted consistency of the corpus with that criterion; and K = min(M, E), a conservative fit bounded by the weaker component. The authors stress that M is not a measure of truth and E is not a measure of causal strength, and that the numbers form a structured synthesis rather than a meta-analysis.

MEM's advantage sits in scope, not evidence

Mean fit was 5.09 for MEM and 3.00 for GNWT. That gap needs its sampling condition attached: items 13–25 raise the corpus weight given to interoception, affect, and organismic regulation, which favours models that explicitly formalize those variables. The authors read classical GNWT's narrower stated scope in this domain as a difference of theoretical emphasis rather than an empirical challenge.

The criterion-level profile is more informative than the averages. GNWT performs best on threshold dynamics, chronometry, electrophysiology, and reporting conditions. MEM claims broader scope on modality, interoception, learning, and action, but lacks theory-specific evidence. Because K is capped by the weaker of M and E, several criteria where MEM claims a scope of 6 collapse to the same conservative fit as GNWT: P1 (4.65 for both), P3 (4.66), P4 (3.75), A1 (4.04), A3 (4.41), A4 (4.08), S1 (4.39), S3 (3.31), and B1 (4.52). Where GNWT's scope is higher, its K leads: A2 (4.36 versus MEM's 4.36 but with GNWT claiming scope 6), B3 (4.67 versus 4.00), and B4 (4.80 versus 4.00). MEM's clearest scope advantages appear on P2 (4.33 versus 3.00), S4 (4.24 versus 3.00), and B2 (5.04 versus 5.00).

Compatible findings do not separate the theories

The paper walks through canonical GNWT results and asks whether MEM could reproduce each qualitatively. In the attentional blink, the distribution of subjective visibility is more consistent with a discrete transition than a gradual decline; GNWT attributes this to workspace occupancy and failed ignition, while MEM attributes it to a first winning semblion being primed and maintained by winner-take-all competition and recurrent processing, with the second signal failing to establish a stable recurrent loop and secondary perception. Both predict threshold selection.

The same pattern recurs. Late, widespread MEG divergence between seen and unseen stimuli fits ignition, but also fits MEM's distinction between a local feedforward sweep and the later victory of a representation with recurrent reconstruction. Intracranial markers — sustained potentials, high-frequency activity, phase synchronization, long-range connectivity — are the classical ignition signature, yet MEM can also generate them through propagation across linguistic, semantic, and motor associations. The authors' point is that chronometry alone identifies neither the recipients of broadcasting nor the modality-specific target of re-entry; directionality measures and layer-specific interventions would be needed. Compatibility here means the architecture can reproduce a finding qualitatively, not that the result was predicted in advance from independent parameters.

Genuine competition needs a dissociation

The authors separate shared claims from orthogonal ones and from real rivals. Both theories assume distributed processing, competition among representations, a nonlinear transition to a more stable state, recurrence, and downstream influence of the winning content. Other claims are orthogonal: GNWT specifies the conditions and dynamics of global availability, while MEM describes the geometry of learned representations, their receptor grounding, their relations to action programmes, and their affective valuation. These components could coexist in the same brain.

Competition begins only where each theory assigns a constitutive role to a different process. Strong GNWT holds that sufficiently rich ignition and global availability are sufficient for conscious experience; strong MEM holds that modality-specific sensory–interoceptive activation or reconstruction is necessary for phenomenal quality. An experiment can adjudicate only if it independently manipulates global availability and modality-specific re-entry while measuring experience as free as possible from reporting demands. The paper also warns that replacing 'broadcast' with the general term 'propagation' is not enough to replace GNWT: MEM must predict which populations, transmission directions, delays, and thresholds will be observed, and when they will diverge from a global workspace prediction.

Interventions that could falsify MEM

The paper closes its analysis with a proposed experimental programme designed to dissociate sensory re-entry, global availability, report, and interoceptive valuation, and it specifies criteria for falsifying MEM. This follows from the evidential standard the authors set: a direct MEM programme should match GNWT research in preregistration, quantitative temporal and spatial predictions, causal interventions, no-report paradigms, and out-of-sample model comparison.

The authors also flag where MEM is currently underspecified. Report is assigned to secondary perception, recategorization, coupling to linguistic semblions, and activation of a procedural communication programme, but an explicit threshold of reportability and a model of metacognitive confidence are lacking. Action control is rated high for selection and only partial for flexible control, with conflict, response inhibition, and rule switching left underspecified. Phenomenal content is described as receptor-grounded activation and recurrent reconstruction of modality-specific sensory and interoceptive fields leading to secondary perception — a distinctive but as yet unverified claim that requires separating the effect of re-entry on experiential quality from its effects on attention, working memory, report, and task performance.

Underdetermination is the result, not MEM's victory

The defensible reading is that GNWT's findings are substantially compatible with MEM, which argues against the exclusivity of a single interpretation and supports a shared class of recurrent–selective mechanisms. That is not theory-specific evidence for MEM, because post hoc compatibility cannot substitute for prospective prediction. The authors state this limit plainly: reinterpreting GNWT data in MEM terms can demonstrate theoretical underdetermination and the feasibility of an alternative explanation, but it cannot replace a theory-specific test.

MEM's greatest additional potential, in the authors' account, lies in the receptor-grounded structure of semblions, modality-specific re-entry, secondary perception, and the coupling of content with interoception, valence, allostasis, and action programmes. Whether that combination actually explains qualia and phenomenal consciousness better than the classical GNWT core is exactly what the proposed dissociating experiments would have to show. Until then, the paper's own priority is the experiment that separates modality-specific reconstruction from global availability and report — not the mean-fit comparison, which its sampling choice already tilts toward MEM.

Sources used in this answer