Background. Phase III trials of rehabilitation of paresis after stroke have proven the effectiveness of intensive and extended task practice, but they have also shown that many patients do not qualify, because of severity of impairment, and that many of those who are treated are left with clinically significant deficits. Objective. To test the value of 2 potential adjuvants to normal learning processes engaged in constraint-induced movement therapy (CIMT): greater distribution of treatment over time and the coadministration of d-cycloserine, a competitive agonist at the glycine site of the N-methyl-D-aspartate glutamate receptor. Methods. A prospective randomized single-blind parallel-group trial of more versus less condensed therapy (2 vs 10 weeks) and d-cycloserine (50 mg) each treatment day versus placebo (in a 2 x 2 design), as potential adjuvants to 60 hours of CIMT. Results. Twenty-four participants entered the study, and 22 completed it and were assessed at the completion of treatment and 3 months later. Neither greater distribution of treatment nor treatment with d-cycloserine significantly augmented retention of gains achieved with CIMT. Conclusions. Greater distribution of practice and treatment with d-cycloserine do not appear to augment retention of gains achieved with CIMT. However, concentration of CIMT over 2 weeks ("massed practice") appears to confer no advantage either.