A lot of science is scut work. While I do ponder my findings, develop new hypotheses, and publish papers that might (slightly) change the way that others view the natural world, these activities are, in fact, only the most glamorous ones in the scientific profession. I spend far more time worrying that I am mistaken about a result. You see, scientists are nit-pickers who know from personal experience the difficulty of proving something is true. Scientists are skeptical of their own findings as well as those of their colleagues. We are always on the lookout for false assumptions, biased data samples, misleading correlations, and experimental results that are artifacts (false outcomes) of our methods. Thus, we spend a good deal of time poking and prodding our data, turning it this way and that, and making certain it is bullet-proof, before we try to publish or present it to colleagues. Scientists run many tedious and seemingly repetitive statistical tests aimed at testing a single hypothesis and ruling out alternative explanations for patterns we find in our data. If a flaw in our reasoning, an untested assumption, or a problem in experimental design weakens or invalidates our findings, we want to discover it ourselves in the solitude of our office — not have a listener at a talk or a reviewer of a grant proposal do so.

Scientists have a much higher threshold for accepting statements as fact than does the public at large. Indeed, flawed and misleading conclusions — which would bring harsh criticism to a scientist uttering them — are rampant in public discourse. The biased sample problem leads to many misleading conclusions. Following a political debate, TV commentators always tell us who won, based upon a sample of viewers. Unless one is very careful to control the political makeup of an audience, however, the outcome of such a poll is certain to be biased. Naturally, Fox News and CNN have audiences that differ greatly in their political leanings; in addition, people who watch debates on television or in person represent a biased sample of voters, not the population at large. Finally, viewers are more likely to see the candidate they favor as the winner of a debate, so favoritism towards one candidate will make that candidate more likely to be seen as the winner, even if their performance was worse that their opponent’s. That is, if 65% of all voters favor Ms. Sims over Mr. Peach ahead of the debate, and 55% of all voters say afterwards that Ms. Sims won the debate, Mr. Peach almost certainly performed better, because he beat his poll numbers.

I face the biased sample problem constantly in my analysis of loon behavior. For example, we have observed that loons shifting from a first to a second breeding territory tend to move a very short distance, often settling to breed on a new lake right next to their old one. It is tempting to surmise that loons that move between territories cover only a short distance in order to take advantage of their knowledge of the local area and ease their transition to the new breeding space. This sounds plausible but ignores the fact that shifters are not a random cross-section of the population. Instead, these loons are almost all old individuals with low fighting ability that have been evicted from first territories. Moreover, the new territories they shift into are not average breeding territories but new, untested ones with limited nesting habitat that seldom yield offspring. So old, worn out loons do not seem to be carefully choosing to settle in a new breeding space that they know well; rather, they are desperately setting up a new territory near their original one — and in a place that no other loon wishes to use — because it is not worthwhile trying to compete for a proven territory anywhere else.

At the moment, I have turned a critical eye towards black flies and nest abandonment. I have “known” for decades that black flies cause high rates of nest abandonment in certain years, as they did in 2017. But it is one thing to know something is true, and quite another to convince other scientists of what you know. So I have gone back to field records from 1994 to 2017 and tallied occasions when field observers reported severe infestations of black flies on loons or around their nests during the early nesting period. Then I looked at the correlation between reports of severe black flies and rates of nest abandonment across years. The result, as shown in the figure above, is unsurprising. In years when black flies were reported to be abundant, nest abandonments were very common.  (By the way, that data point in the upper right corner is 2014.)

While I was certainly not on pins and needles during this latest analysis of black flies, it is a crucial piece of the puzzle. Lacking any direct data showing that black flies caused loons to abandon nests, my best evidence to support this conclusion was that cool springs lead to a high rate of nest abandonment. The strong correlation pictured above now implies a direct causal connection between flies and abandonments.

I breathed a sigh of relief at this finding. I am not sure how I would have responded if I had found that years of severe black flies were NOT correlated with rates of nest abandonment. Yet I cannot rest. I can imagine a scientific reviewer complaining that the correlation might have resulted from observer bias. For example, once an observer starts to notice that black fly population is high early in the year and possibly related to nest abandonment, he or she might be more likely to report severe black fly infestations on subsequent days. Such behavior by field observers might explain the correlation I found, at least in part.

In short, I am still uneasy about my analysis of black fly impacts on loon nesting. I am looking for additional statistical analyses that could help convince a skeptical audience of the link between flies and abandonments. That is what life is like for a scientist.