Eighty women undergoing multimodality treatment for large (>4cm) or locally advanced (T3, T4, Tx, N2), breast cancers participated in a randomised controlled trial (RCT) to evaluate the immuno-modulatory effects of relaxation training and guided imagery. Patients underwent chemotherapy followed by surgery, radiotherapy, and hormone therapy. Those in the intervention group were taught relaxation and guided imagery. Patients kept diaries of the frequency of relaxation practice and imagery vividness. On 10 occasions during the 37 weeks following the diagnosis, blood was taken for immunological assays CD phenotyping: T cell subsets (helper, cytotoxic), natural killer (NK) and lymphokine activated killer (LAK) cells, B lymphocytes and monocytes; cytotoxicity: NK and LAK cell activities; cytokines interleukin 1 beta (1beta), 2, 4 and 6 and tumour necrosis factor alpha. Significant between-group differences were found in the number of CD25+ (activated T cells) and CD56+ (LAK cell) subsets. The number of CD3+ (mature) T cells was significantly higher following chemotherapy and radiotherapy, in patients randomised to relaxation and guided imagery. Using a median split, women who rated their imagery ratings highly had elevated levels of NK cell activity at the end of chemotherapy and at follow-up. Significant correlations were obtained between imagery ratings and baseline corrected values for NK and LAK cell activity, and IL1beta. Relaxation frequency correlated with the number of CD4+ (T helper) cells, the CD4+:8+ (helper:cytotoxic) ratio, and IL1beta levels. Relaxation training and guided imagery beneficially altered putative anti-cancer host defences during and after multimodality therapy. Such changes, to the best of our knowledge, have not been previously documented in a RCT.