Several attempts to attack tumours in experimental systems have been made using conjugates of chemotherapeutic agents or potent toxins with antibodies (immunotoxins). In vitro studies have been highly successful, showing target specificity of a high order in some cases. However, so far, such conjugates have been inadequate in vivo, probably for two main reasons. First, conventional heteroclonal antibodies are perhaps inappropriate, because purification by biochemical methods leaves a large amount of non-antibody gamma-globulins. The use of monoclonal antibodies may overcome this problem. Second, when whole toxins have been conjugated to antibodies there has been a strong residual nonspecific cytotoxicity due to the binding capacity of a subunit, the B-piece of the toxin. (Diphtheria toxin or ricin consist of two polypeptide subunits. The A-piece is responsible for inhibition of protein synthesis on ribosomes, and the B-piece binds to galactose residues on the cell membrane and facilitates the transmembrane passage of the A-piece.) In the present work the problem of nonspecific binding by the B-piece has been circumvented by using the A-piece only as the toxin component of immunotoxins; these immunotoxins are active both in vitro and in vivo.