Abstract
In defining physical (i.e. causal dynamic) units to which conscious experience is to be ascribed, integrated information theory (IIT) raises three notable requirements: (1) that a unit to which consciousness is ascribed must be defined, or circumscribed, by some intrinsic aspect or property, where intrinsic implies existing 'for itself' or 'from its point of view'; (2) that the intrinsic aspect that defines the unit to which consciousness is ascribed must be dynamic (i.e. involve causal power) rather than purely structural or kinematic; (3) that this dynamic aspect must involve the integration of elements of information in such a way that semantic values of elements are interdependent, allowing rich meaning. It will be argued that, although these requirements are all well-motivated, IIT falls short on them within a testable natural science framework. Concerns raised about the three requirements all point towards one central problem: that the concept of 'integrated information' does not involve any specific causal relation. Although IIT starts from strictly causal premises, the final analysis appears to invoke two incompatible accounts of the same events, with 'integration' having no causal relational basis. It will also be argued that the way IIT attempts to provide a testable theory of dynamic units to which consciousness is ascribed mustcausal reason inherent in any means of testing, be at best redundant for in the human case and, as a general principle, untenable.