Band 3, the anion exchanger of human erythrocytes, contains up to 14 transmembrane (TM) segments and has a single endogenous site of N-glycosylation at Asn642 in extracellular (EC) loop 4. The requirements for N-glycosylation of EC loops and the topology of this polytopic membrane protein were determined by scanning N-glycosylation mutagenesis and cell-free translation in a reticulocyte lysate supplemented with microsomal membranes. The endogenous and novel acceptor sites located near the middle of the 35 residue EC loop 4 were efficiently N-glycosylated; however, no N-glycosylation occurred at sites located within sharply defined regions close to the adjacent TM segments. Acceptor sites located in the center of EC loop 3, which contains 25 residues, were poorly N-glycosylated. Expansion of this loop with a 4-residue insert containing an acceptor site increased N-glycosylation. Acceptor sites located in short (<10 residues) loops (putative EC loops 1, 2, 6, and 7) were not N-glycosylated; however, insertion of EC loop 4 into EC loops 1, 2, or 7, but not 6, resulted in efficient N-glycosylation. Acceptor sites in putative intracellular (IC) loop 5 exhibited a similar pattern of N-glycosylation as EC loop 4, indicating a lumenal disposition during biosynthesis. To be efficiently N-glycosylated, EC loops in polytopic membrane proteins must be larger than 25 residues in size, with acceptor sites located greater than 12 residues away from the preceding TM segment and greater than 14 residues away from the following TM segment. Application of this requirement allowed a significant refinement of the topology of Band 3 including a more accurate mapping of the ends of TM segments. The strict distance dependence for N-glycosylation of loops suggests that TM segments in polytopic membrane proteins are held quite precisely within the translocation machinery during the N-glycosylation process.