Many of us trust our students to conduct a good portion of our research. Our students collect the data, collate the data, score the data, and often analyze the data. We remain as vigilant as possible to potential problems – errors can occur at any point along the way. We watch out for outliers, data entry errors, errors in the analyses, and so on, because mistakes happen. Honest mistakes and errors occur in every experiment; we’re used to it. But it rarely occurs to us that the data themselves are not real; our trust in our students overwhelms the possibility that the data may be faked.

Yet, this is what happened to me early on in my career. My student (I’ll call her Jane) was completing her Master’s thesis. I’d recently been inspired to examine the effects of text cohesion with younger readers. Several studies had demonstrated the importance of increasing cohesion for low-knowledge adolescent and adult readers, pointing to the potential importance of cohesion for younger emerging readers. I set out on the quest to investigate this question with Jane. She was very excited about the project and decided to conduct the study for her thesis project. We decided to focus on children in approximately the second grade. We collected second grade books. We chose our passages. We developed the high- and low-cohesion versions. We constructed the comprehension questions and chose the individual difference assessments. Jane completed and defended her thesis proposal; we were all set to run the study.