Engineering neural tissue is one of the most challenging goals of tissue engineering. Neural tissue is highly complex and possesses an organized three-dimensional (3D) distribution that is essential for tissue function. An optimal scaffold for tissue engineering has to provide this distribution until the cells are able to activate their normal functions and develop neural connections with the host tissue. Different strategies such as gene therapy and cell transplantation particularly in retinal tissue have been tested, but so far they have only induced retinal degeneration in animals. The objective of this work was to study neural cell assembly as a function of scaffold features and surface chemistry for application in retinal tissue engineering using microfabricated patterns with a well-defined geometry. Because retinal neurons are known to be arranged in hexagonal arrays, hexagonal scaffolds of poly(DL-lactide-co-glycolide) acid were fabricated using a pressure-assisted microsyringe (PAM) system. The behavior of a model cell, neuroblastoma originating from human retina (SH-SY5Y), was analyzed after seeding on the scaffolds, measuring cell density as a function of line width and length of the scaffold to identify the optimal hexagonal geometry. We also evaluated the influence of scaffold on cell metabolism using the methyl thiazolyl tetrazolium assay and on neurite extension. As far as two-dimensional scaffolds are concerned, the results show that although metabolic activity per cell remains constant, hexagons with sides of 500 microm and line widths of 20 +/- 5 microm are optimum for neural cell adhesion in terms of cell density. On 3D scaffolds, cell metabolism is about three times higher than controls, and the optimum number of layers in the scaffold is three or four.