Interdisciplinary research starts by questioning basic, ingrained ideas. This should remain a guiding principle.
A picture quiz about stereotypes got things started with a bit of fun at last month's lecture about interdisciplinary research at the London International Development Centre (LIDC), UK.  It was the speaker Peter Mollinga's way of grabbing our attention and making the first of many insightful points.
For the quiz, Mollinga, professor of development studies at SOAS (the University of London's School of Oriental and African Studies), put the faces of two unfamiliar men and one woman on the screen and took a poll: who did we think was the natural scientist, who was the social scientist and who was the economist?
We duly raised hands and, as an audience, turned out to be predictable in our stereotyping. In fact, our responses matched those of many others, in different parts of the world, collected by Mollinga over many years.
He then admitted that he had no idea who the people he had shown us actually were - so there was no real match to our collective notions of who belonged where. And so his first point was made: to do interdisciplinary research, it's essential to question basic, ingrained ideas.
What followed was at times just as entertaining, at times familiar, often philosophical - and consistently thoughtful. Yet it left me with a troubling thought: could the struggle to wrestle this idea of interdisciplinarity into academic practice be a symptom of the very problem it aims to solve?
There are various explanations for why disciplines were created in the first place, Mollinga told the audience. They can be seen as the scientific equivalent of division of labour in industry - a way to increase productivity - or as a consequence of rising student numbers.
It's not difficult to appreciate the benefits of specialisation - in-depth knowledge, for example, or gaining an advantage when job hunting. In his introductory remarks, LIDC director Jeff Waage pointed out that plenty of progress can be made by working with individual disciplines. But there are shortcomings, he said, and one clear example of that is the Millennium Development Goals.
Although it's a lot easier to go about understanding the world by breaking it down into manageable parts, a reductionist approach to science falls short because the whole is more than the sum of its parts, Mollinga pointed out. Disciplines "need to be saved from themselves" - on their own, they can't understand and address socially relevant, complex problems.
This echoes a parallel discussion, in other circles, about complexity science's role in development. Commenting on a recent book on this topic, Andrée Carter, the director of the UK Collaborative on Development Sciences, has written about the need for science to adapt, by drawing knowledge from a range of disciplines and stakeholders, so it can better meet the challenges within international development.
Interdisciplinarity can be broadly defined as both working across sectors, and across disciplines within sectors. Often, Mollinga said, what happens in practice is that the elements of a project are disciplinary, and then they're brought together in an 'integrated model'. "[But] what defines interdisciplinarity is forcing ourselves to have a shared question [from the start]," he stressed. It can take five years or more to undertake this kind of joint problem framing.
Time, and consequently, money, aren't major obstacles, he added - the reason why truly interdisciplinary projects are uncommon has more to do with resistance within academic circles.
This comes down to different languages, paradigms, institutional incentives and power dynamics over who might be included or excluded in an interdisciplinary group.
SciDev.Net's practical guide on how to communicate across disciplines offers tips on avoiding some of these issues. But change needs to start early - with education.
Writing about the need to support multidisciplinary education, Daya Reddy, president of the Academy of Science of South Africa, says this should be about carefully restructuring learning and institutions themselves.
One of the most fascinating parts of Mollinga's talk came with his definition of the word discipline. It derives from disciple, or follower, he said; but it is also about being disciplined and systematic.
He moved on to conjure up vivid images of academic tribes and territories borrowed from the "must-read" book by Tony Becher and Paul Trowler.  People are taught to identify with disciplines, he remarked, and questioning that is difficult.
Education and communication tips aside, Mollinga suggested another way forward: adopting a 'boundary work' framework to find ways - concepts, tools and processes - to productively negotiate disciplinary boundaries.
Ecosystem services or vulnerability are examples of concepts that many disciplines can use, he said, even if they mean something a little different in different scientific circles.
It was at this point that I remembered an earlier part of the talk. We heard that disciplines aren't pre-existing or static - many are created every year, and there are lots of hybrids. A discipline is what we decide to call it, Mollinga said; there is nothing natural about it.
And so I wondered whether the study of interdisciplinarity would come to be seen as a discipline in itself. Could it end up creating new boundaries, if the concepts we try to pin down end up becoming solid structures themselves?
If the answer is yes, then it's an uncomfortable conclusion. Perhaps it is possible to, instead, accept a notion of flexible boundaries, where researchers creatively negotiate shared languages and territories with each project and its circumstances.
After the event, during an interview with my colleague Cristina Gallardo, Mollinga described interdisciplinarity as a post-colonial problem: scientists from 'developed' countries once preached disciplinarity, and now they preach interdisciplinarity. It could even be argued, he said, that the hands-on way in which countries with limited resources approach problem-solving is closer to interdisciplinarity.
There may be a lot to learn by looking more closely at persisting paradigms and divisions that go beyond academic disciplines.
Opinion & Special Features Editor, SciDev.Net
 Interdisciplinary research requires joint problem framing - argues Prof. Peter Mollinga of SOAS at LIDC's inaugural lecture on interdisciplinarity and international development (LIDC, 15 January 2014)
 Tony Becher and Paul Trowler Academic tribes and territories (Open University Press, 2001)