False Negative

I ran the data through the analysis program five times before I shot him a text and asked him to come to my lab.

He arrived, claiming he didn't have much time, until I told him what the data seemed to be saying. He asked how many times I'd checked it, then watched as I checked it a sixth time.

We called in a couple more colleagues and we all manually checked the raw data, and none of us could spot a problem with it.

So, naturally, we junked it and started fresh, getting data that was clearly more accurate.