Does a changing climate mean an increase in conflict and civil unrest around the world?
Some of the world's military authorities believe it might - elements in the US armed forces hierarchy, for example, see climate change as a security issue for just that reason.
And just last month the UN Security Council held its second major discussion on climate-driven conflicts, concluding that "the possible adverse effects of climate change could, in the long-run, aggravate certain existing threats to international peace and security".
But it's a hugely difficult question to gauge.
If modelling some of the physical parameters of climate change accurately - rainfall, to take an obvious example - is hard, extrapolating to factors that could affect unrest such as water scarcity is obviously a much bigger challenge.
And the equation becomes even more complex when you add in human responses.
For example, rapid development of crops able to tolerate drought would materially change the outlook for millions of people already living on the edge of an adequate food supply.
So if modelling the future is hugely tricky here, is the past a better guide?
A couple of years ago, researchers published an analysis concluding that high temperature has been one of the factors driving conflict in Africa.
Last year, another group published a counter-analysis that said no, it hasn't been. When you looked deeper into the roots of those African conflicts, they said, and when you looked at different definitions of local climatic factors and of the term "conflict", temperature appeared not to be an issue.
The biggest factor changing the climate on a year-to-year basis is the El Nino Southern Oscillation (ENSO).
The warm El Nino phase brings rain to the western coast of South America and hot, dry conditions further east. Africa is visited by both heavy rains and drought.
According to the Columbia research, El Nino also brings conflict.
Looking at the years from 1950, the researchers found that El Nino doubles the risk of civil conflict across a swathe of 90 countries affected by the changed climate.
In countries not seriously affected, there was no change in the conflict rate.
Many of the conflicts in the register were African. But taking these out of the mix, the link still held.
In principle this could tell us something about a climate-constrained future, for a number of reasons.
Firstly, El Nino events can raise the global average temperature by a couple of tenths of a degree Celsius, and are therefore sometimes regarded as "test-beds" for what every year will look like soon.
Secondly, some computer models indicate that El Nino conditions will become more common in future.
But whether this latest work tells us anything about conflict under human-induced climate change is a moot point.
The researchers themselves don't make such a claim, though they do describe it as "the first demonstration that the stability of modern societies relates strongly to the global climate".
As they note, changes driven by ENSO are fairly rapid, from one year to the next - very different from the underlying upwards march of temperatures (and concomitant changes in other things) driven by greenhouse gas concentrations.
So the capacity of people to adapt may be different.
It's also worth chucking into the mix that the global average temperature variation between even the strongest El Nino event and the next ordinary year is a fraction of a degree Celsius - certainly less than the rise of about 0.6C that we've seen since the beginning of the study period, in 1950.
As they make clear themselves, the researchers also haven't any idea what the mechanism is behind the link they've found.
This is something that troubles Halvard Buhaug, the Norwegian researcher behind the 2010 paper that found no link between high temperatures and African conflict.
"This doesn't provide any explanation of a mechanism for the connection between the ENSO cycles and conflicts, and I'm quite puzzled about what it could be," he told me.
"The effect is immediate, within a matter of months from the El Nino start until the conflict is observed. But it doesn't work through local temperature or rainfall changes."
Looking at the whole field of climate-conflict studies, he said, a key was to understand the reality of what might link short- or long-term climatic change to outbreaks of conflict.
When the UN Security Council last debated climate change, a number of developing countries objected.
They said it was the wrong forum, and that the issue was not fundamentally one of security but of development.
The Columbia research backs them up to some extent.
Man-made climate change may indeed be a "threat multiplier" for social stress, as many have argued; but economic and social development clearly emerges as a threat-diminisher.
The examplar is Australia, one of the countries whose climate has been most badly affected by ENSO variations in recent decades, but which did not register a single civil conflict in the half-century research period.
Follow Richard on Twitter