Human immunodeficiency virus type 1 (HIV-1) reverse transcriptase catalyzes DNA synthesis from RNA and DNA templates by a sequential mechanism. This enzyme is neither processive nor distributive but has a rather intermediate behavior; at any template position, there is a certain probability that the replica strand will be extended, which we define as extensibility. The extensibility depends on the substrate concentration, i.e. on the concentration of the cognate (and to a smaller extent of the noncognate) deoxynucleoside triphosphates, in a typical Michaelis-Menten mode. The extensibility varies from position to position in a sequence-dependent manner, being particularly low at certain sites, accordingly called pause sites. The rate and fidelity of successive incorporation of nucleotides were measured and then compared with numerical integrations of the pertinent rate equations, which were composed to describe a suitable reaction mechanism and parameterized starting starting with rate constants reported in the literature. We found that agreement between stimulation and experiment requires two-step binding of enzyme to the template-primer. In an initial second-order step, an "outer" binary complex is rapidly formed; this is followed by a slower conformational change into an "inner" complex. During multiple rounds of nucleotide incorporation, the complex remains in the inner form; the rate-determining step for enzyme release is the reversion from the inner to the outer complex, with a standard rate constant of 0.2s-1. This rate constant may be significantly increased at pause sites. In order to match the experimental results, the standard rate constants had to be modified for pause sites. At low concentrations or in the absence of the cognate nucleotide, the site-specific misinsertion frequency, a function of the nucleotide pool is bias and of the efficiency to discriminate against a noncognate nucleotide, can be determined from the dependence of extensibility on concentration of cognate and noncognate substrates. The error frequency was found to be somewhat smaller than the misinsertion frequency, because mismatches are extended less efficiently than matched pairs.