The initiator of coliphage lambda DNA replication, lambda O protein, may be detected among other 35S-labeled phage and bacterial proteins by a method based on immunoprecipitation. This method makes it possible to study lambda O proteolytic degradation in lambda plasmid-harboring or lambda phage-infected cells; it avoids ultraviolet (u.v.)-irradiation of bacteria, used for depression of host protein synthesis, prior to lambda phage infection. We confirm the rapid decay of lambda O protein (half-time of 80 s), but we demonstrate the existence of a stable lambda O fraction. In the standard five minute pulse-chase experiments, 20% of synthesized lambda O is stable. The extension of the [35S]methionine pulse, possible in lambda plasmid-harboring cells, leads to a linear increase of this fraction, as if a part of the synthesized lambda O was constantly made resistant to proteolysis. Less than 5% of lambda O protein synthesized during one minute is transformed into a stable form. We presume that the stable lambda O is identical with lambda O present in the normal replication complex and thus protected from proteases. We cannot find any stable lambda O in Escherichia coli recA+ cells that were irradiated with u.v. light prior to lambda phage infection, but their recA- counterparts behave normally, suggesting that recA function interferes in the assembly of a normal replication complex in u.v.-irradiated bacteria. The stable lambda O found in lambda plasmid-harboring, amino acid-starved relA cells is responsible for the lambda O-dependent lambda plasmid replication that occurs in this system in the absence of lambda O synthesis. The existence of stable lambda O raises doubt concerning its role as the limiting initiator protein in the control of replication. Another significance of lambda O rapid degradation is proposed.