Several leading theories of consciousness, including Global Workspace Theory, claim that a central function of consciousness is to permit the integration of information from different functional regions, such as different sense modalities: call such theories of consciousness “integration theories”. A presupposition of integration theories is that unconscious processing cannot achieve this kind of integration, as witnessed by such claims as “unconscious input processing is limited to sensory regions”, and “consciousness is needed to integrate multiple sensory inputs” (Baars 2002, p 47-48). I report on work by our team at the Universities of Sussex and Wisconsin that set out to determine whether associative learning can take place without conscious perception of the stimuli and whether this can be achieved where stimuli to be associated are presented in different modalities. Employing a new paradigm, it was demonstrated over four experiments that pairs of stimuli presented subliminally were associated, as evidenced by the ability of one stimulus to prime classification of the other. Experiment 1 presented stimuli auditorily, Experiment 2 visually, and Experiment 3 presented one of the paired words auditorily and the other word visually. Experiment 4 generalised this paradigm to non-linguistic stimuli (auditory and visual). All four experiments found the same significant inhibitory priming effect with concordant test pairs associated with slower classification judgements. Thus it appears that, contrary to integration theories, unconscious inputs from distinct sensory modalities can be integrated.
After reviewing the background and the above experimental findings, I will assess the impact of the findings on integration theories. In particular, I will consider how integrationist theories might be modified to accommodate these findings. For example, one might abandon the idea that consciousness is required for integration tout court, and instead claim that it is required for integrated information to be deployed in a flexible way, e.g. that spans longer time scales, can potentially manifest itself in a wide range of effects depending on the needs of the subject (e.g., not just reaction time), or is generally under the executive control of the subject.
Ron Chrisley received a Bachelors of Science in Symbolic Systems from Stanford University in 1987, with honours and distinction. He was an AI research assistant at Stanford, NASA, and Xerox PARC, and investigated neural networks for speech recognition as a Fulbright Scholar at the Helsinki University of Technology and at ATR Laboratories in Japan. In 1997 he received a DPhil in Philosophy from the University of Oxford, and in 1992 he took up a lectureship in Philosophy in the School of Cognitive and Computing Sciences at the University of Sussex. From 2001-2003 he was a Leverhulme Research Fellow in Artificial Intelligence at the School of Computer Science at the University of Birmingham. Since 2003 he has been the director of the Centre for Research in Cognitive Science (COGS) at the University of Sussex, where he is also on the faculty of the Sackler Centre for Consciousness Science.