In Koornneef and Van Berkum’s (2006) recent eye-tracking study of implicit-causality effect (Caramazza, Grober, Garvey, & Yates, 1977), mid-sentence delays were observed in the processing of sentences such as ‘David blamed Linda because she(bias-consistent)/he(bias-inconsistent)…’ when the pronoun following because was inconsistent with the bias of the implicit-causality verb. The authors suggested that these delays could be attributed to participants having predicted the occurrence of a bias-consistent pronoun after because. Accordingly, any word other than the bias-consistent pronoun placed after because should cause processing delays. The present investigation set out to test this speculation by presenting participants with sentences of the form ‘David blamed Linda because she(bias-consistent)/he(bias-inconsistent)/there(bias-neutral)…’. Since significant delays were observed in sentences containing a bias-inconsistent pronoun but not in sentences containing there, the results of the present study did not support a word-specific prediction-based account. These implications of these findings are discussed with regards both to the nature of online predictions in discourse processing.