Background: There is overwhelming evidence that anxiety is associated with the tendency to interpret information negatively. The causal relationship between this interpretive bias and anxiety has been examined by modifying interpretive bias and examining effects on anxiety. A crucial assumption is that the effect of the procedure on anxiety is mediated by change in interpretive bias rather than being a direct effect of the procedure. Surprisingly, this had not previously been tested. Aim: The aim is to test whether altered interpretive bias, following Cognitive Bias Modification of Interpretations (CBM-I), affected anxiety. Method: Mediational path analyses were conducted to test the hypothesis that changes in anxiety are due to changes in interpretive bias. A separate experiment was conducted to test which elements of the procedure could be responsible for a direct mood effect. Results: Results from mediation analyses suggested that changes in trait anxiety, after performing CBM-I, were indeed caused by an altered interpretive bias, whilst changes in state anxiety appear to be caused by the procedure itself. The subsequent experiment showed that state anxiety effects could be due to exposure to valenced materials. Conclusions: Changed state anxiety observed after CBM-I is not a valid indicator of a causal relationship. The finding that CBM-I affected interpretive bias, which in turn affected trait anxiety, supports the assumption of a causal relationship between interpretive bias and trait anxiety. This is promising in light of possible clinical implications.