According to Karl Popper, a respected 20th century philosopher famous among the scientific fraternity, true scientific theory makes predictions that can be empirically tested.

The superhero status of testable predictions has made good sense to me ever since I was exposed to it as an undergraduate back in the Carboniferous. Unless a theory can be tested it falls to the lowly status of opinion where only dubious predictions live; admittedly an overcrowded residence these days.

An idea, supposition or prediction attains the lofty moral position of a scientific theory — a supposition or a system of ideas intended to explain something, especially one based on general principles independent of the thing to be explained — if it can be empirically tested, ideally through manipulations in controlled conditions with heaps of replication.

This much is grasped by most students of science, even the naive ones around when the trees were laid down for coal. It is the basics of the scientific method taught in every good high school.

Unfortunately, this is often as far as it goes. But there is more.

What Popper also realised was that scientists can never prove a theory to be true because the next test might contradict all that preceded it. Observations can only disprove a theory they cannot prove it. Empirical tests can only falsify.

This is way more subtle. Evidence from a controlled experiment might reject the hypothesis the experiment was designed to test but the alternative outcome (where evidence is not sufficient to reject the hypothesis) does not make the alternative (accepting the hypothesis) true. Empirical tests can only disprove, never prove.

Suppose I have a large field that I subdivide into twenty equally sized fields.

Into 10 of these small paddocks, chosen at random, I place five sheep for five days, remove them for 10 days and then put them back in. This rotational grazing goes on for a year. The other 10 paddocks contain no sheep at all.

The hypothesis is that grazing by sheep will decrease the amount of carbon in the soil. So before the sheep are introduced several soil samples are taken from all the small paddocks and tested for their carbon content. More soil samples are taken at the end of the year and their carbon content statistically compared with carbon content in the soil samples taken at the start.

It turns out that after a year the average carbon content from the grazed paddocks averages about 3%, slightly more than it was at the start, a small but statistically significant increase. In the paddocks without sheep, soil carbon also increased too but by no more than would be expected by chance (as determined by the statistical properties of the numbers generated from the soil carbon samples).

The hypothesis – sheep grazing will decrease the amount of soil carbon in the soil – is rejected given the empirical evidence.

The evidence is enough to reject the hypothesis and the temptation is to accept the theory that sheep actually do good things to soil carbon. Only Karl Popper would wriggle a little in his coffin if you made this call because should you do this experiment again, who knows what the outcome would be.

This example is phrased to follow the conventional wisdom. Current theory is that livestock grazing will reduce soil carbon over time as the animals metabolize the primary production and the farmer removes animals or their fleeces to market making for a net loss in soil carbon over ungrazed paddocks.

But if we rephrased the hypothesis as ‘grazing by sheep will increase the amount of carbon in the soil’ and the results of the experiment stay the same, then we accept the hypothesis. Again we are tempted to accept the theory that grazing by sheep is good for soil carbon levels only this time by claiming the results are a proof not a falsification.

Popper gets to wriggle again.

Interesting isn’t it. Even when science is done through determined experiments the outcome is not a given. Conclusions are also dependent on how the empirical test is conceived. This is why theory only gets such a lofty badge when there is repetition of empirical tests sufficient to reduce doubt but even then there is no proof, only falsification.

The sheep grazing example is naive of course and was phrased around hypothesis testing rather than theory. In reality, theory only achieves acceptance after many tests of many specific hypotheses. The process of iteration provides the rigor that allows scientists to rest easily at night without Popperian spectres messing with their dreams.

Only the example is also real.

We are not actually sure of the theory in this case despite the importance of grazing to food production and the reality that soils need as much carbon as possible to maintain that production.

Falsification is very difficult to do in environmental and ecological science, especially where soil is concerned. There is very little in the way of Popperian truth where fields, paddocks and remnant native vegetation is concerned. There have been way too few tests leaving fertile ground for opinion.

However, the risk in leaving issues of food security to opinion should scare the socks off you.