During lab studies of text entry methods it is typical to observer very few errors in participants' typing - users tend to type very carefully in labs. This is a problem when investigating methods to support error awareness or correction as support mechanisms are not tested. We designed a novel evaluation method based around injection of errors into the users' typing stream and report two user studies on the effectiveness of this technique. Injection allowed us to observe a larger number of instances and more diverse types of error correction behaviour than would normally be possible in a single study, without having a significant impact on key input behaviour characteristics. Qualitative feedback from both studies suggests that our injection algorithm was successful in creating errors that appeared realistic to participants. The use of error injection shows promise for the investigation of error correction behaviour in text entry studies.