Robots and autonomous systems are being developed at an ever-increasing rate. Autonomous systems are already prolific in many households around the world, and their adoption is only expected to increase over the coming decades. Even so, many of the systems that are deployed today are still prone to small operational errors such as struggling to navigate complex environments. While the argument over how acceptable these kinds of errors are is still ongoing, these systems are in fact being deployed throughout society and small errors have the potential to gradually erode the trust in them. One way of reducing this erosion of trust in robots is to have the robot provide a spoken explanation for why the error happened. However, speech is not always a given in robots and it is currently unknown how just possessing the ability to speak impacts the impression of a robot. To shed some light on this question, we present data from two online human-robot interaction experiments. We had 227 participants view videos of a humanoid robot exhibiting faulty or non-faulty behaviours while either remaining mute or speaking. The participants evaluated their perception of the robot's trustworthiness, likeability, animacy, and perceived intelligence. While a non-faulty robot achieved the highest trust, a faulty robot that could speak managed to almost completely mitigate any degradation of trust. We theorize that having the ability to speak increases the perceived intelligence and capability of the robot, which in turn increases trust. It is also possible that speaking causes the robot to appear more like a sentient or living being, causing people to be more lenient when evaluating it.