The effect of repeating features in a short-term memory task was tested in three experiments. Experiments 1 and 2 used a recognition paradigm. Participants encoded four serially presented objects and then decided whether a probe matched one of them with regard to all three features. In the control condition no feature was repeated; in the experimental condition features were repeated in two memory objects. Experiment 3 used a cued recall paradigm with the same list design. After list presentation one feature was used as a cue uniquely indicating one of the memory objects. Participants recalled the remaining two features of the probed object. Feature overwriting as one component of the interference model of Oberauer and Kliegl (2006) predicts worse performance in the experimental compared to the control condition. Results of all three experiments did not support this hypothesis. Recognition performances in Experiments 1 and 2 were not impaired by repeating features. Recall performance in Experiment 3 was better for repeated features, contrary to the predictions of feature overwriting. Predictions from feature overwriting for the shape of serial position curves were also not confirmed.