The CI repressor of bacteriophage lambda is a model for the role of cooperativity in the efficient functioning of genetic switches. Pairs of CI dimers interact to cooperatively occupy adjacent operator sites at O(R) and at O(L). These CI tetramers repress the lytic promoters and activate transcription of the cI gene from P(RM). CI is also able to octamerize, forming a large DNA loop between O(R) and O(L), but the physiological role of this is unclear. Another puzzle is that, although a dimer of CI is able to repress P(RM) by binding to the third operator at O(R), O(R)3, this binding seems too weak to affect CI production in the lysogenic state. Here we show that repression of P(RM) at lysogenic CI concentrations is absolutely dependent on O(L), in this case 3.8 kb away. A mutant defective in this CI negative autoregulation forms a lysogen with elevated CI levels that cannot efficiently switch from lysogeny to lytic development. Our results invalidate previous evidence that Cro binding to O(R)3 is important in prophage induction. We propose the octameric CI:O(R)-O(L) complex increases the affinity of CI for O(R)3 by allowing a CI tetramer to link O(R)3 and the third operator at O(L), O(L)3.