Wiring up the nervous system occurs as a self-organizing process during animal development. It has long been proposed that directional growth of axons towards their targets is achieved by gradients of guiding molecules and the conceptual framework of gradient guidance was introduced more than a decade ago. Novel experimental results now allow the formulation of models incorporating more mechanistic detail. We first summarize some crucial in vitro and in vivo results concerning the development of the chick retinotectal projection. We then review two recent theoretical models based on these findings (the models of Nakamoto and colleagues, and of Honda). Neither model considers the latest observation that putative guidance ligands, in addition to their tectal expression, are expressed in a similar pattern on the retina and that a disturbance of this expression affects topography These findings suggest that retinal axons might grow into the tectum until they have reached a ligand concentration matching that of their site of origin. We call this the imprint-matching concept of retinotectal guidance. As a framework for pinpointing logical difficulties of the mechanistic description of the guidance process and to stimulate further experiments we finally suggest two extended versions of Honda's model implementing imprint matching, which we call 'the variable set-point' and 'the gradient-sensitive adaptation' model. Strengths and weaknesses of both mechanisms are discussed.