The control of immunoglobulin (Ig) isotype production has a wide range of implications, ranging from the regulation of gene expression in eucaryotic cells to clinical applications such as allergy and "facilitation." In spite of the rapidly accumulated knowledge (1-3) on the structure and organization of Ig C-genes and flanking "switch regions," there is at present no general agreement on the rules determining the selective expression of one or another isotype in specific antibody responses. A model based on the order of structural C-genes along the direction of transcription (4) and on the repeated finding that all C-genes 5' to those being transcribed in secretory cells are deleted from the expressed chromosome (5-8) has recently received much attention. This model (9, 10) assumes no specific control of switch recombination, rather, it postulates randomness in these genetic events and, consequently, dictates, that cells in a clonal progeny will successively delete C-genes and express those located most 3' in the chromosome. Because, in accordance with current molecular models (1) and as shown by direct evidence (11), switch recombination requires DNA replication, it follows that such a model predicts a direct correlation between the number of mitoses undergone by a competent cell and the number of switch recombination events that have taken place in its DNA. This model explains, for example, the high frequency of IgA-secreting cells in some anatomic sites or biological situations by postulating a long process of replication in these cells before being induced to terminal maturation (10).We have now directly tested these assumptions by making use of a serial transfer system, originally described by M611er (12) and later used to analyze clonal stability of V-gene expression (13). It is particularly important to point out that, in this transfer system, the progenies of immunocompetent B cells are analyzed for their whole life span because the experiment is continued until clonal senescence limits further proliferation. It appeared that the analysis of the isotypes produced by antibodysecreting cells along their whole life-span would provide direct indications as to the validity of that random recombination hypothesis. The results obtained do not support that model.