In this theory-building chapter we explore neural mechanisms potentially fulfilling key requirements for conscious representations. First we focus on the brain's capacity to align and bind elements of percepts such as features in different (sub)modalities. Continuing with the question how we come to experience sensory modalities as qualitatively different from one another, it is argued that no central readout mechanism, attaching verbal “labels” to sensory input, is needed to solve the problem. Instead, the correlative structure in multimodal networks may provide a basis for qualitatively different experiences. This proposal leads to another question: if we assume that individual object features are coded by neuronal firing rate, how should this multimodal network structure be functioning to code such correlative relationships in addition? Dual coding of feature presence by firing rate, and feature relationships by firing phase, offers a promising scheme to solve this problem neurophysiologically. Similarly, cognitive interpretation-the act of interpreting a sensory input as signifying a semantic category-is suggested to depend on iterative interactions between low-level sensory and high-level mnemonic systems. Combining sensory and cognitive interpretation, global percepts may achieve temporary stability by dynamic attractors in coupled multisensory and mnemonic systems.