A strain of Escherichia coli constructed by Shapiro has a segment of Mu bacteriophage DNA inserted between the araC and lacZ genes. Excision events that produce an in-frame fusion of lacZ to araB result in a cell (here designated Ara-Lac+) that can grow on lactose if arabinose is present as an inducer. Whether or not these excision events occur in the absence of selection for the Ara-Lac+ phenotype has figured prominently in the debate of the phenomenon known as 'directed' or 'adaptive' mutation. In an attempt to settle the issue, we have used classic fluctuation tests to show that cells capable of producing a clone of descendants that are phenotypically Ara-Lac+ do, indeed, arise in stationary phase cultures kept starving in depleted minimal medium. We found that Ara-Lac+ progenitors arise rapidly under these conditions, in contrast to the delayed appearance of Ara-Lac+ mutants when cells are incubated on lactose-arabinose minimal plates. Similar results are reported in the accompanying paper by Maenhaut-Michel and Shapiro, who used indirect selection to isolate Ara-Lac+ cells in the absence of selection. However, their sequencing data have introduced a new unexpected complication to the interpretation of all such experiments, and it is no longer clear exactly when the fusions arise.