This paper constitutes a response to Lust & Mazuka's (1989) defence of the Principal Branching Direction parameter and their critique of O'Grady, Suzuki-Wei & Cho's (1986) experiment, which purported to show that even children learning left-branching languages exhibit a preference for forward patterns of anaphora. Results of new experimental work with children learning Japanese are reported and shown to support the claim that there is a universal preference for forward patterns of anaphora in the early stages of language acquisition.