Abstract A basic question concerning microtubule assembly is the polarity of growth, namely, whether subunits can add to either end of a growing microtubule or whether growth proceeds by subunit addition to only one end. To approach this question in an in vitro system, experiments were carried out on the addition of microtubule subunits to isolated flagellar axonemes. Flagella were detached from Chlamydomonas by brief treatment with non-ionic detergent, isolated by differential centrifugation, and incubated with crude high-speed extracts of porcine brain tissue or with purified tubulin (obtained by repetitive temperature-dependent assembly and disassembly). Electron microscopy of negatively stained samples showed as many as 11 long microtubules added at one end of more than 90% of the axonemes. Colchicine (100 μ m ), CaCl 2 (2.5 m m ), and low temperature (0 °C) both prevented and reversed microtubule assembly but had no effect on axonemal length. In crude extracts microtubules formed on both members of the axonemal central pair but on only the A-tubule of the outer doublets. Flagellar fragments, produced by mechanical shearing, were also incubated with microtubule subunit. Single tubules formed at only one end of outer doublet fragments; the appearance of single tubules on one or both members of central pair fragments was predominantly unidirectional. Structural analysis of frayed axonemes and the asymmetry of side-arm attachments permitted the absolute polarity of the axonemal fragments to be determined and revealed that assembly proceeded by addition of subunits to the distal ends of the axonemal microtubules. Using purified brain tubulin, a limited extent of proximal addition and growth on the B-tubule also occurred. The extent of proximal addition increased with increasing protein concentration and temperature. We conclude that the microtubules of flagella have an intrinsic polarity reflected in their side-arm attachments and in their directionality of growth.