Many mammalian and avian complex carbohydrates (glycoproteins and glycolipids) have highly branched oligosaccharides. Although the function of complex carbohydrates is not known, there is evidence to suggest that oligosaccharide branching may be an important factor in the process by which cells recognize one another and their environment. Asparagine-linked (N-glycosyl) oligosaccharides can be subdivided into at least 12 classes according to their branching patterns. It is presently believed that these classes all stem from a common precursor oligosaccharide containing three D-glucose, nine D-mannose, and two N-acetyl-D-glucosamine residues. This precursor is incorporated into the protein backbone in the rough endoplasmic reticulum and is then processed within the endoplasmic reticulum and Golgi apparatus by a series of highly specific glycosidases and glycosyltransferases to yield the various classes of N-glycosyl oligosaccharides. The branches that occur in N-glycosyl oligosaccharides are usually initiated by the incorporation of a N-acetylglucosamine (GlcNAc) residue. Our laboratory has studied four of the N-acetylglucosaminyltransferases (GlcNAc-transferases) involved in this initiation process. We have defined various factors which determine the synthetic pathway. There are at least three types of control that are commonly found. (i) Tissues differ in the relative activities of the different glycosyltransferases and glycosidases and, therefore, competition between two or more enzymes for a common intermediate often determines the synthetic route. (ii) The incorporation of a key glycosyl residue into an oligosaccharide may convert a nonsubstrate to a substrate for either a glycosyltransferase or a glycosidase. (iii) Conversely, the incorporation of a key residue may convert a substrate into a nonsubstrate. Other controls are undoubtedly operative during glycoprotein synthesis: e.g., the effect of the polypeptide sequence on transferase specificity, the distribution of transferases along the endomembrane system, and compartmentation and the availability of substrates and cofactors. These factors have not been studied in our laboratory. However, the oligosaccharides made by the hen oviduct correlate quite well with the control factors elucidated by our approach; other tissues are presently under investigation. Recent studies on the three-dimensional structures of N-glycosyl oligosaccharides have enabled us to explain certain features of glycosyltransferase substrate specificity on the basis of steric factors.