Sometimes, science does not go the way you plan.

That is usually framed as failure: the experiment did not work, the results were inconclusive, the story was unclear. But science can also advance when we go back to old data and ask a different question. That is exactly what happened with a study of Field Scabious, Knautia arvensis, a familiar wildflower of meadows and grasslands.

Back in 2001, I carried out an experiment to test the function of the showy outer florets around the edge of the flower head. These enlarged “ray florets” look as though they ought to be important in attracting pollinators. The idea was simple: trim them in different ways and see whether pollinating insects responded – you can see the results of that trimming in the photo above.

The answer, at the time, was baffling. Bumblebees, hoverflies and butterflies (all of which are effective pollinators of this plant) did not seem to care very much. Visitation rates by the insects hardly changed, and neither did seed set. But seed weight did change, as did the amount of sugar being produced by the trimmed flower heads. At the time I could not make sense of this in relation to the question I had asked, so the data were archived and left alone.

But not forgotten.

Fast forward 25 years, and I was chatting about this data set with colleagues when I was at the Kunming Institute of Botany in China. That got me reading some more recent work about florivory -the damage done to flowers by animals – and I realised that this old experiment might make more sense if viewed from another angle. Perhaps I had not really been testing floral attraction at all. Perhaps, without intending to, I had carried out a simulated florivory experiment.

Seen in that light, the results became much more interesting.

Cutting the ray florets did not stop pollinators from visiting. Nor did it reduce the number of seeds produced. But it did change the plant’s internal economics. Nectar quality declined in the most heavily cut flowers, and the seeds that those flower heads produced were consistently lighter. In other words, the flowers still functioned well enough to get pollinated, but the plant appeared to invest less in rewards for pollinators and less in each of its offspring.

That matters because lighter seeds may have poorer prospects later in life, even if the plant initially appears to reproduce successfully. The damage did not cause total reproductive failure. Instead, it produced a subtler effect: hidden costs that only become visible when you look beyond simple seed counts.

Newly contexualised in this way, we wrote up this work, being completely honest about the history of the study, submitted it to the Journal of Pollination Ecology, where it received very positive reviews. It’s now been published and you can download a copy by following the link in the reference:

Ollerton, J., Xu, X. & Ren, Z.-X. (2026) Misconceived experiments may yield valuable insights: simulated florivory has unpredictable consequences for plant reproduction in Knautia arvensis (Caprifoliaceae). Journal of Pollination Ecology (in press)

I like this study because it tells two stories at once. One is ecological: florivory may not always reduce pollinator visits, but it can still alter plant reproduction in potentially important ways. The other is about how science actually works. Not every good paper begins with a perfect hypothesis and a clean result. Sometimes the value lies in returning to an awkward, neglected dataset and realising that it was trying to tell you something different all along.

Science does not always go the way you plan. But occasionally that is when it becomes most revealing.