Assays of immune function (recall skin tests to microbial antigens; total circulating lymphocytes, T-cells, B-cells; lymphocyte blastogenesis with PHA, Con A, and pokeweed mitogens; and serum immunoglobulins IgA, IgM, IgG) were obtained in 408 patients with unresectable gastrointestinal carcinoma. The overall patient population, in comparison to normal controls, was characterized by reduced response to recall skin tests, reduced total lymphocyte and T-cell counts, reduced lymphocyte blastogenesis assays, increased B-cell counts and increased IgA and IgM. Significant immunosuppression was associated with prior radiation or chemotherapy, and with impaired patient performance status. There was no apparent correlation between extent of clinically evident malignant disease and immune function within this patient population. No assay of immune function matched the prognostic value of the more readily available and less expensive determinations of performance status, serum alkaline phosphatase, or SGOT. Only reactivity to recall skin tests had a significant correlation to patient survival independent of performance status. Among patients with little or no disability, only intensity of skin test reactivity correlated significantly with survival; and among those with greater disability, there was correlation only with proportion of skin tests positive. The combination of candida and streptokinase antigens provided the best recall skin test survival correlation. Adding a third, fourth, or fifth antigen did not add to prognostic value. From an overall standpoint, the immune determinants which we studied do not appear to provide useful additions to the evaluation of the patient with unresectable gastrointestinal cancer.