In Gibson's theory, an object is defined by its interactions with people, named affordances. Affordances raise from the association of the object's sensory perceptions. Bienenstock Cooper and Munro's (BCM's) cells converge to one of the input patterns with a decentralized and unsupervised learning. We want to extend this mechanism in order to develop a generic incremental modalities association paradigm. We introduce a feedback modulation of BCM's neurons to obtain a spatial auto organization. This feedback will be the reflect of the multi modal constraints.