In my experiments I test whether an additive to the soil does good things to the plants. We'll call it RGS (Really Good Stuff) for now. When I add RGS, I also add (because it's part of the chemical) something we'll call ES (Extra Stuff). To verify that the observations are due to RGS and not ES, I have a control to which I add only ES. This is a negative control.
A while ago, I analyzed my plant material and noticed that the plants that had only gotten ES, tested positive for RGS. I had already seen that the observations weren't as expected. ES plants were behaving too much like RGS plants.
The next experiment had the same problem. As I was writing this up for the past few weeks, I had a real problem explaining my data, because the ES plants were doing weird things. I did some tests last night, and more detailed ones this morning. And the verdict is: someone must have mixed up the labels on the container and put a form of RGS in the container labeled ES. All the while I though I was putting in ES as a negative control, I was simply adding another RGS treatment.
When I started noticing these problems with my plants, I added another negative control to my experiments, to which I didn't add anything at all (no ES or RGS). Those experiments are still salvageable, because at least I have a negative control. The ones where ES was my only negative control are useless.
I now have to re-do the entire statistical analysis and go back over all my data, trying to figure out when exactly this problem occurred. On the bright side: my data makes much more sense now. ES plants were behaving like RGS plants, because that's exactly what they were. It's easier to explain, but I really didn't need to spend more time analyzing data, thank you.
I'll spend the rest of the day working on statistics, and making new graphs and tables for the chapter in question, and sadly, tossing out a bunch of experiments.
This is exactly what the negative control is supposed to do, though. Science did do it's job here, and let me know that something was up. At least I figured it out now, and not after the paper was published. I would have had a hard time explaining my data.
I also have to notify all the other people in the lab that might have used ES as a negative control. They're likely to run into problems too. If I ever get my hands on the person that mislabeled those containers....