In recent experiments on the Kerr effect (electric birefringence) in polyelectrolyte solutions without salt, an anomaly was reported. For a certain intermediate concentration range, the sign of the induced birefringence is opposite to that found at either low or high concentration. An attempt is made to rationalize this data in terms of a locally disc-like clustering of anisometric, polarizable effective particles. If this picture is correct, the birefringence anomaly may help distinguish between competing theoretical models of semidilute polyelectrolyte structure