Trends in Cognitive Sciences
ForumA Novel Framework for Unconscious Processing
Section snippets
Behavioural Tests of Unconscious Processing
Over the past decade or so, there has been a palpable shift towards the use of subjective reports to identify states of visual unawareness [1]. Trials in which observers report no awareness of a stimulus are used to infer the properties of unconscious processing mechanisms across multiple cognitive domains, including visual perception [2], learning, and memory [3].
Relying on subjective measures of awareness to understand unconscious information processing remains the subject of ongoing strong
A Brain-Based Framework
Unconscious events associated with null perceptual or mnemonic sensitivity may only produce weak effects on the guidance of behaviour. Importantly, demonstrating the absence of behavioural effects from unconscious stimuli may be taken as evidence against the existence of unconscious processing. Crucially, in the framework we propose here this need not be the case. We contend that the representation of unconsciously processed items can be isolated through fine-grained analyses of brain activity
Exploiting Machine Learning, Pattern Analyses, and Computational Models
Figure 2 illustrates some of the approaches currently used in our lab to uncover the brain representation of unconsciously processed information. These include the use of transfer learning based on models pretrained using conscious items that are re-used to characterise unconscious knowledge. The potential of transfer learning lies in the re-use of pretrained models from different experiments/domains/task/stimuli to address the common and distinct brain representational spaces across different
Unconscious Processing across Domains
A key issue is the extent to which unconscious information processing in this framework can be isolated across multiple cognitive domains of perception, learning and memory, language, and emotion. For instance, implicit learning studies involve acquiring knowledge about complex spatiotemporal regularities for visible items. Concluding that the knowledge is unconscious has often been based on subjective awareness tests that do not meet stringent criteria (Figure 1; see [4]). One solution would
Caveats
Our framework does not fully address the extent to which unconscious information processing influences behaviour. We think our proposal is a special case in which a reductionist neuroscience approach is needed to circumvent the long-held concerns with existing approaches to unconscious information processing. The framework, however, could be developed to test whether or not and when an unconscious representation is associated with behaviour, when a particular brain pattern is present or absent,
Acknowledgements
D.S. acknowledges support from the Spanish Ministry of Economy and Competitiveness, through the ‘Severo Ochoa’ Programme for Centres/Units of Excellence in R & D (SEV-2015-490) and project grants PSI2016-76443-P from MINECO and PI-2017-25 from the Basque Government.
References (12)
- et al.
The anatomy of non-conscious recognition memory
Trends Neurosci.
(2016) Decoding the meaning of unconsciously processed words using fMRI-based MVPA
NeuroImage
(2019)No-report paradigms: Extracting the true neural correlates of consciousness
Trends Cognit. Sci.
(2015)Brain mechanisms underlying the brief maintenance of seen and unseen sensory information
Neuron
(2016)- et al.
Beyond functional connectivity: Investigating networks of multivariate representations
Trends Cognit. Sci.
(2018) Learning of goal-relevant and -irrelevant complex visual sequences in human v1
NeuroImage
(2018)