The dilution effect occurs when the introduction of non-diagnostic information lessens the impact of diagnostic information despite having no relevance to the hypothesis in question. While the effect has been reproduced in several studies, the psychological basis of the effect remains unclear. Some believe it to be conversational while others believe it to be cognitive and social. The paper tests the conversational basis of the effect by removing the influence of the experimenter from the design. Rather, the studies make use of a legal setting with witness testimonies. The studies replicate the dilution effect, which suggests that the basis of the results in the original studies is not conversational. The results suggest that the credibility of the source strongly influences whether or not the effect occurs. If reliable sources provide the non-diagnostic information, the effect lessens. Conversely, if unreliable sources provide the non-diagnostic information, we observe a stronger dilution effect.