Rui leads the Neural & Machine Learning group that leverages AI principles to develop a new generation of computational models of learning in the brain. In particular, the group focuses on understanding how (i) cortical circuits, (ii) neuromodulation and (iii) subcortical regions enable efficient credit assignment in the brain.
Rui did his PhD at the University of Edinburgh (UK) as part of the Institute for Adaptive and Neural Computation where he established a collaboration between theoretical (Mark van Rossum) and experimental groups (P. Jesper Sjöström). During this time Rui was also a visiting PhD student at the University College London (UK) and McGill University (Canada). Next, Rui conducted postdoctoral research in computational neuroscience & machine learning at the University of Oxford (UK) with Tim Vogels where he established collaborations with the groups of Nando de Freitas (Oxford/Google Deepmind) and Nigel Emptage (Oxford). Next, Rui did a short postdoc with the group of Walter Senn (Bern, Switzerland) in collaboration with Yoshua Bengio (MILA, Canada). In 2018 Rui moved to the University of Bristol to start his own group (Neural & Machine Learning group), which in 2023 moved to the University of Oxford.
Scott Waddell studied biochemistry as an undergraduate at the University of Dundee, and researched cancer for his Ph.D. at the University of London. He switched fields and continents as a postdoc in the Department of Brain and Cognitive Sciences at Massachusetts Institute of Technology. After 10 years leading a research group in the Department of Neurobiology at the University of Massachusetts Medical School, he moved to Oxford as a Wellcome Senior Research Fellow in November 2011. He is a member of EMBO and was awarded the 2014 Liliane Bettencourt Prize for the Life Sciences. Scott became a Wellcome Principal Research Fellow in 2016.
Scott’s group has studied neural circuit mechanisms of memory-directed behaviour in the fly since 2001.
Your PhD was in cancer biology. What got you interested in neuroscience?
Cancer research is interesting and very important work but I felt that there was too much fighting over details. I knew I had to find a different field. Throughout my Ph.D. my advisor John Jenkins encouraged me to read different areas of science. John left a photocopy of a paper from Tim Tully on my bench, so I searched the literature to find other labs using genetics to study fly learning and memory. This lead me to Chip Quinn’s pioneering studies. Seeing that single genes impacted this level of behaviour amazed me. Once I read these papers, and others from Martin Heisenberg, Yadin Dudai and Ron Davis, I knew that I had found the field I wanted to be in.
And not just the field; it sounds like you knew where you wanted to be too?
Chip Quinn first taught fruit flies in the early 1970s as a post-doc in Seymour Benzer’s lab at Caltech. After meeting Chip and other researchers at MIT, I realised MIT was where I wanted to work. It was also clear that I would be supported but very independent with Chip, and that is exactly what I wanted. The first time we walked into his lab he remarked, with a high voice, “This could be all yours, including the curtains!”. And so in 1996 I moved out of cancer research and joined the MIT Center for Learning and Memory as a post-doc with Chip. It was a radical shift for sure, and for a couple of years I wondered if the change was too great. But it was something I was genuinely interested in, and I eventually made some exciting progress.
What was your first breakthrough?
I found two neurons in the fly brain that are required for memory. The discovery came from the analysis of amnesiac—one of Chip Quinn’s original memory defective single gene mutant flies. Chip had shown in 1979 that amnesiac flies learn, but forget quickly. Looking for the sites of amnesiac production led me to two large neurons in the brain. I showed, using a genetic switch of synaptic transmission, that if I crippled the function of these neurons the flies forgot, like amnesiac mutants. It seemed likely that the amnesiac gene mutation had in some way impaired the function of these neurons and that memory stability was controlled by these neurons in normal flies. The output of these two neurons can be traced to a brain structure called the mushroom body. The mushroom body had previously been implicated in olfactory memory in the fly, but it still has about 5,000 neurons. So the resolution of memory processing was radically improved and at the same time a philosophical change was taking place in the field. We were starting to alter acutely the function of small groups of neurons as a way of learning how the relevant neural circuits operate.
And this led to other breakthroughs?
In my own lab we have used similar genetic strategies to show that the mushroom body neuron population is functionally divisible into a group whose activity is required after training for memory consolidation, and a group required for later memory retrieval. This is reminiscent of how memory is consolidated by the hippocampus and cortex in mammals. It seems likely that this ensemble process relies on the two ‘amnesiac’ neurons we previously identified.
We also located small groups of neurons representing a motivational control system for behaviour. Only hungry flies will pursue an odour that they have previously learned signifies the presence of food. My students found that a subset of six dopamine neurons inhibit behaviour in well-fed flies. These dopamine neurons are in turn switched off, releasing memory-guided food-seeking behaviour in hungry flies, by neurons that release the fly analogue of mammalian neuropeptide Y.
More recently my group identified rewarding dopaminergic neurons in the fly. Again, the work uncovers a striking mechanistic similarity between flies and mammals – reward and motivation signalling even uses the same neurotransmitter molecule. However, our work also revealed a few surprises. We found that both the sweet taste and nutrient value properties of sugars are rewarding, but that nutrition is essential to trigger dopaminergic neurons that form persistent long-term memories. We also showed that water and sugar rewards use separate subsets of dopaminergic neurons, suggesting these memories are stored in different places in the fly brain. This seems fairly logical because we know that hunger and thirst can specifically direct food or water-seeking behaviours.
What drives your work today?
I am generally fascinated by how the brain orchestrates intelligent behaviour and we are tackling this at several levels. How are stimuli represented and associated, and how is experience encoded? How are memories consolidated? How are memories appropriately accessed and retrieved to guide behaviour? How are memories updated? How are behaviours instructed? We now know that some of these complex processes can be controlled by manipulating small groups or even single pairs of neurons.
I think an important thing to say is that we are exploring many things at the CNCB that I am confident we can answer.
Is there a single area of research that stands out?
Understanding neural processes of memory is centre stage. Memory is a difficult thing to study, because it is measured by observing a behavioural outcome. However, this immediately introduces a problem that is of particular interest from the point of appetitive memory: has the fly actually forgotten, or is it just choosing not to respond? The dogma is that memory is represented as a specific set of changes in synaptic strength—perhaps within a single circuit. I think we are very close to being able to observe these changes and understand how their use is controlled in an appropriate way. We already know some relevant neural pathways and synaptic connections and we have detailed ideas of how they are regulated.
The main thing we’re doing in Oxford that we weren’t doing in the US is to make physiological recordings of neural events. We are watching and measuring reward signals and trying to understand how they are generated and what they represent. We are also trying to visualise how rewards and motivational states change neural circuit activity.
Such fine experimenting on something as small as a fly’s brain must require sophisticated techniques and sophisticated equipment?
My colleagues at the CNCB are pioneers of optical imaging and we have all the required hardware. That hardware includes machinery like the two-photon microscope—a fluorescence-imaging technique that allows us to see several hundred micrometres into living tissue, which in this work is a very great depth; importantly, more than the thickness of the entire fly brain. Advanced optics becomes even more important as we start to look at molecular processes.
Optogenetic control was pioneered by your colleague Gero Miesenböck and others. What is special about optogenetics?
The beauty of optogenetics—and a related discipline called thermogenetics, in which neurons are activated by heat—is that you can jump into the middle of a circuit. You don’t need to know where the signal has come from, or where it is going. We can identify the neurons that are involved in the expression of some behaviour without knowing how they themselves are controlled; but ultimately, of course, we would like to know what the prior and next step is in the process. And this is where high resolution analysis becomes important and where we hope to make advances.
Is this work unique to the CNCB?
It is clear that, although the most important questions in science are of interest to many, others will not necessarily investigate them in the same way, from the same vantage point, or to the same level of detail. We like to first design a relevant behavioural task for the flies and then identify the neurons and mechanisms that allow the fly to perform it. Surprisingly, some others seem to approach it the other way round – find a few neurons and then try to work out what they do.
Do you have a number of research projects running at the same time?
Yes, many. I like to balance the work going on in the lab, having some relatively simple and apparently straightforward things and some more difficult going on at the same time. The relatively simple projects often don’t work!
What’s the atmosphere like in the lab?
The lab is very interactive, collaborative and social. I encourage people to work on something that is a shared interest so that they are personally invested and motivated enough to get on with it. This work is not a 9 to 5 job for me at all. It’s something I can’t put to one side. I think about experiments in the middle of the night and I think about them in the middle of conversations with people when I should be listening.
Do you think about the potential benefits of your work?
I’m curiosity driven. I genuinely believe that what we’re doing will have some benefit for human health down the road, but it’s not necessarily what drives me on a daily basis. I think history has told us that some of the most spectacular advances in our biological understanding have been happened upon by open-minded, curiosity-driven researchers getting a result they don’t understand and re-interpreting it.
What are your goals for next few years at the CNCB?
Our hope is that we can identify and understand the operation of small neural circuits that support memory and direct educated behaviours. If we can do this, we will not just be understanding the fly’s mind, but intelligence in all animals. Given the remarkable conservation of genes, I am willing to bet that several fundamental principles of neural circuit operation will be similarly conserved. In the fly I think we can realistically investigate some of these processes down to the level of handfuls of, and in some cases single, neurons. This cellular resolution allows us to understand what the signals are, how they are generated by molecules and where they are conveyed to, and how these signals affect the function of the circuit. Having that level of understanding of the neural mechanisms of behavioural control would be quite profound.
It is sometimes said that the problem of consciousness is the big problem for the 21st century. Would you agree?
I think every field – including learning and memory—has terms that people have trouble with and argue about. It’s fascinating to think about how you could define consciousness with an experiment, but with an animal that can’t communicate anything to you except through its actions, it is difficult. That said, I don’t think consciousness is limited to humans. Even fruit flies appear to attend selectively to sensory stimuli and can form memories of prior experience. Focusing on certain things at the expense of others and remembering are parts of consciousness; if you didn’t have these functions, you wouldn’t be aware of who you are, where you are and in control of what you are doing; assuming we are in control! So I’m working on processes that are relevant to consciousness, but it’s not something I feel I need to define in the brain of the fruit fly. Honestly, I’d get frustrated by the lack of progress.
Stephen Goodwin
Biography
Born in Banbridge, Northern Ireland, Stephen Goodwin grew up in Belfast and attended Methodist College Belfast grammar school. He studied genetics as an undergraduate at the University of Glasgow and researched Drosophila learning and memory for his Ph.D. After a postdoctoral stint at Brandeis University, he spent 10 years leading a research group at the University of Glasgow before moving to Oxford in 2009.
How did you get into biology, and how did you come to be working in this area?
My father was a mechanical engineer by training but an enthusiastic biologist, and in the late 70s he studied for an Open University degree in biology. He was a believer in broadening your mind in a way that isn’t connected to your working life and getting a sense of achieving something new and amazing. He adored the way different disciplines could cross-pollinate new and exciting ideas, and preached this doctrine throughout his life. I think I got my initial curiosity for working things out from my father — he spent much of his career involved in design engineering. He was always amazed by evolutionary design in the natural world—this resonated with his own views on incremental design in engineering. At school I actually had an aptitude for languages, biology and art; I really disliked mathematics and physics. I started University with no clear view of my future, but fortunately I fell in love with my genetics course from my second year onwards. I have very fond memories of listening to one of my lecturers, Richard Wilson, wax lyrical about a fly mutant called fruity (now known as fruitless). After University I started a PhD working with yeast cell-cycle mutants, but after six months I packed it in. I didn’t realise then but I needed to work on something with legs and a body. An accidental phone call with one of my lecturers in Glasgow, Kim Kaiser, and I found myself working on fruit fly learning and memory in his lab. You can see where this is going: in fly learning and memory, all roads lead to the godfather of neurogenetics Seymour Benzer, and Benzer guided me to one of his F1 progeny, Jeff Hall, and ultimately courtship behaviour. The rest, as they say, is history, or rather, neurogenetics.
What are your research questions?
We are studying how the fundamental properties of ‘maleness’ and ‘femaleness’ are encoded in the circuitry and chemistry of the brain and how these internal states combine with sensory stimuli to elicit sex-specific behaviours. We use Drosophila courtship behaviour to study how sex-specific neural circuitry and behaviours are established during development by the action of complex networks of genes. Our studies focus on two pivotal transcription factors of the sex-determination hierarchy, fruitless (fru) and doublesex (dsx) that act together to specify and configure both the anatomy and physiology of sex-specific neural circuitry.
Why are these research questions important?
Understanding how the central nervous system translates biologically relevant stimuli into behaviour is one of biology’s biggest mysteries. Courtship behaviour in Drosophila has proven to be an outstanding model for understanding this complex problem.
What’s the best piece of advice you’ve been given?
Most of my life I have been terrible about asking for advice, and like many people rarely take it. Nonetheless, four pieces of sagely advice stand out for me, and I have tried to live by these ‘truths’.
‘Truth’ number 1: during my aimless teenage years I had been dithering between a career in law or science (in reality I wanted to be David Bowie). My solicitous grandfather, Fred Haugh, told me: “Stephen, my lad, this is a very simple dilemma, choose science as it’s a ticket to ride, you will have the opportunity to work in different countries, and meet people from interesting cultures; law will keep you stuck at home”. ‘Truth’ number 2: on leaving Jeff Hall’s lab, it was traditional for the lab to throw a big party for the leaver. This involved a lot of booze, greasy food, and gift giving (a wonderful combination). The lab gave me a Red Sox baseball jersey with my name emblazoned on the back, and a box set of Frank Sinatra rare recordings. I was really surprised when Jeff handed me a CD, “Never Mind the Bollocks, Here’s the Sex Pistols”. I was never a big fan of punk music (The Clash being the exception). “This should be your mantra, live by it and don’t forget it”, Jeff pronounced. Later the penny dropped (it always drops slowly for me): what Jeff was telling me was don’t worry about the unimportant stuff; just worry about getting the science right. ‘Truth’ number 3: did I mention that my wife is the real brains of our operation? She gets me to do things that I don’t like doing. ‘Truth’ number 4: finally, notwithstanding the influence that Kim and Jeff had on me as a scientist, I believe that your peers are very influential. Being intrinsically lazy I have always liked to surround myself with bright people, I think it ups your game.
Which question about your work annoys you the most — and why?
“Is there a human gene for that?” Do I have to say why? Well, the fly has already taught us a great deal about the molecular structure of our nervous system and now it will teach us how this structure functions. Because of the vast array of genetic tools available for the fly, and its rich behavioral repertoire, we can identify the cellular components of neural circuits, map function in these circuits and define causal relationships between neural activity and behaviour.
How has the field of Drosophila behavioural genetics changed since you were a PhD student?
When I started my PhD, only the freaks and geeks (myself included) were doing fly behavioural genetics. At that time, neurodevelopmental biology would dwarf the other sessions at meetings; the behavioural talks would be scheduled for the last day, when everyone was either too hungover to concentrate, or checking out to make their flight home. Fast-forward 20 years and things have changed dramatically. The behavioural genetics field has worked extremely hard to understand the organisation and function of the brain, and the challenge now is to understand exactly how the brain’s neural circuits carry out the information processing that directly underlies behaviour. At fly meetings, neural circuits and behaviour are now de rigueur. This change in fashion has led inevitably to increased competition. No one denies that competition is a good thing, but it also leads to an atmosphere of ‘fear and loathing’. The field is no longer a ‘petulant adolescent’, it has lost some of its immaturity, impulsiveness and sensitivity, it’s getting harder to have the time to be curious about something, work it up, and rejoice in the haphazard and explorative nature of it. As scientists we are privileged and fortunate to be working in this field, it’s a gift, and with this comes responsibility, not entitlement. We need to set an example and be better role models for the next generation of fly researchers.
So which do you prefer, scientific ‘hero’ or ‘anti-hero’?
I have always fallen for the scientific anti-hero. Apart from their obvious scientific achievements, they are often loners, misfits, mavericks in their thinking and behavior. As in all good movies, you need a scientific anti-hero to serve as a counter for the too good to be true super-hero protagonist. Integrity, honesty, and the ability to say what needs to be said, even if it clears the room, are all hallmarks of my scientific anti-hero, Jeff C. Hall.
Gero Miesenböck
Biography
Gero Miesenböck studied medicine at the University of Innsbruck in his native Austria and did postdoctoral research at Memorial Sloan-Kettering Cancer Center in New York. He was on the faculty of Memorial Sloan-Kettering Cancer Center and Yale University before coming to Oxford in 2007. Gero is the founding director of the CNCB.
Gero has invented many of the optogenetic techniques used for visualizing and controlling nerve cells with light. He has also been a pioneer in the use of flies to study neural circuits.
As the current Waynflete Professor of Physiology do you feel that you are part of a tradition?
When I arrived in Oxford, my lab was housed in the Sherrington Building, named after perhaps the most famous Waynflete Professor of Physiology. In 1937, Charles Sherrington described the brain as ‘an enchanted loom where millions of flashing shuttles weave a dissolving pattern, always a meaningful pattern though never an abiding one; a shifting harmony of subpatterns.’ Despite the Victorian prose, Sherrington’s vision of flashes of light signalling the activity of nerve cells contains a modern idea: that the brain could reveal its inner workings optically. Many of the technical aspects of my work revolve around this idea.
It’s interesting that, as an avowed reductionist, you should start with the poetic.
Good poetry is also reductionist: it makes complex thought or sentiment elemental. I in fact wanted to become a writer for a very long time. My father was a classicist, and I was tempted to follow him into literature. But he was frustrated with his chosen profession, its subjective standards and its limited potential for discovery. He persuaded me, in strong terms, to choose science as the more exciting path. I started reading more and more science when I was at school and I became gripped by it. But I was gripped from an abstract point of view; I liked ideas and the way they were expressed and tested. I wasn’t tinkering around with a chemistry set.
Your father was clearly an early mentor. Were there others?
In 1989, I spent three months at the University of Umeå, near the Arctic Circle, to learn an experimental technique I needed for my dissertation. Unwisely, I had chosen winter for my visit. In the near-complete darkness of northern Sweden, our circadian clocks started to run free. People worked the strangest hours because it didn’t matter when you went to bed or got up: it was nearly always dark. The only event that brought everyone together was our journal club—a weekly discussion of the scientific literature that is a fixture in many labs. There I came across papers that blew me away with their boldness and the beauty of their ideas. The papers were the work of James Rothman. I made up my mind right there that I wanted a postdoctoral position with Rothman.
So how did you get yourself out of Austria and over to America?
I wrote to Rothman a year or two later when I was finishing medical school. It didn’t take long until a rejection letter arrived, in which my surname was spectacularly misspelled. Admittedly it is an unusual name, but 4 out of the 10 letters were incorrect. Come on! But I persisted. I visited Felix Wieland, a collaborator of Rothman’s in Heidelberg, and he must have put in a word for me, because a few days later I got a phone call at midnight, and it was Rothman saying that he might have made a mistake. In 1992 I moved to the USA for what I thought would be two or three years and ended up staying for 15. Rothman took a risk on a complete unknown. I try to remember this nowadays when I get applications that seem to come out of nowhere, but may in fact be the result of years of reflection and preparation.
What was your first project in the Rothman lab?
I initially worked on protein sorting in cells. But soon my work concentrated on what has become a recurring theme: the development of biological tools to illuminate biological problems. The first such tool used a light-emitting protein to make communication between nerve cells visible. To get the system to work, I had to isolate the DNA that encodes the protein from a particular species of shrimp found along the south coast of Japan. But getting my hands on these shrimp proved difficult. Late every evening in New York, I asked a Japanese colleague to make phone calls to universities, aquariums and natural history museums in Japan, but no luck. Finally, I just rang the Marine Biological Laboratory in Woods Hole, Massachusetts, and asked to speak to their bioluminescence expert, fully expecting to be told to get lost. But incredibly, the phone operator knew how to field my request. I was asked to hold on for a moment and then a voice came on the line. It was Osamu Shimomura. By chance I found myself talking to the one man who could help me.
Osamu Shimomura won the Nobel Prize in 2008 for his discovery of the green fluorescent protein (GFP), an essential reagent in biology. GFP glows greenly when exposed to blue light. It was first purified and studied by Shimomura in the 60s and 70s.
I explained the shrimp situation to him and he sent me a sample he had collected and sun-dried in Japan in 1944, along with a stack of reprints. One of these reprints had a stunning cover photograph: a manuscript page reporting the structure of the shrimp’s light-emitting chemical. The page was illuminated by the chemical reaction described on it—a little like Escher’s famous lithograph of hands drawing themselves.
Although Shimomura’s shrimp proved completely useless, they are one of my prized possessions. The light-generating system is still well enough preserved that you can see a beautiful, eerie blue glow when you crush a few shrimp and add water. The genetic material, of course, was degraded, but Shimomura directed me to a fresh supply of shrimp, from which I managed to isolate a piece of DNA encoding the glowing enzyme. It was this enzyme that produced the first images of synaptic transmission: Sherrington’s weaving shuttles. A later incarnation of the same principle, termed synapto-pHluorin, drew directly on Shimomura’s discovery of GFP. It used a pH-sensitive GFP mutant to provide an optical report of communication between neurons.
What was your first project in your own lab?
The initial aim was to image information flow in a neural circuit with synapto-pHluorin. But then very quickly, in the summer of 1999—it was one of those moments where I can even remember the time and the date and the room I was in—I had the idea of using light not only to observe but also to control. That then quickly became another focus of the lab.
When was that, and where were you?
It was the late afternoon of June 12, 1999, a Saturday. We were living on Union Square in Manhattan at the time. I had taken my daughter on the Staten Island Ferry across New York Harbor, come back home, and stretched out on the bed, ready to return to a book I was absorbed in, Independence Day by Richard Ford. As I was reaching for the book, drifting from the real world into Ford’s fictional New Jersey, there was the idea.
Where did the remote control idea come from?
I guess I had the advantage of being a newcomer to neurobiology. I was not too weighed down by received wisdom, maybe not too weighed down by neuroscience knowledge in general. But I had worked in a leading cell biology lab. I had seen that to establish causality and dissect a complex mechanism it’s essential to be able to control it. In neuroscience, I felt there was still way too much observation and not enough intervention. So I thought, wouldn’t it be wonderful if the two ingredients that I had relied on in my work with synapto-pHluorin—genetics and optics—could be combined again but this time for the opposite way of communicating with the neuron.
At first we were completely alone doing this work. Now, of course, many people have adopted and also improved the approach.
With optogenetic remote control you can say that neuron X is important for behaviour Y. What other questions can you address?
There is a whole range of questions we can ask when we have the ability to control specific groups of neurons. Optogenetics allows us to make non-invasive and physiological connections to brain tissue. We can use these techniques to work out the wiring diagrams of neural circuits. We can apply spatiotemporal patterns of input activity and measure what kinds of signals a target cell or a group of target cells is looking for. It means that rather than just finding the anatomical connections between neurons we can deduce the input/output characteristics of a circuit. A still higher level of analysis is to see what exact features of activity patterns are relevant for perception, action, cognition, memory and so forth.
Have you always been attracted to technological solutions and inventing new methods?
It was not a deliberate choice. I wanted to do certain types of experiment, and I couldn’t make any progress until I developed what was necessary. I wanted to work on networks of neurons and it was clear that there needed to be a new way of interacting with them.
Also, there is beauty in these development efforts in their own right. It’s really a pleasure to see something through from the conceptual stage, to trying to get it work, and then actually using it. Seeing the first images of a fruit fly smelling an odour as revealed by synapto-pHluorin and looking back over the entire arc that began with the engineering of the GFP mutant, or seeing the first remote-controlled fly take flight at the flash of a laser beam, I have to say these were very satisfying moments.
Why did you choose to work on the fly?
The nematode C. elegans has just 302 neurons and a correspondingly, shall we say, basic behavioural repertoire. Rodents are too complex. If you look at a mouse brain under a microscope you instantly realise that you are seeing only a small part of a much bigger structure. In contrast, if you look at a fly brain under a microscope you get the impression that you are seeing something self-contained. Most of the relevant parts are there, visible at once. The scale of the biological structure and that of our analytical methods, which operate at the resolution of individual cells, matches. Then there are the added benefits: a century of genetic work has been done on the fly, and its behaviour is rich.
But it is easy to underestimate the complexity of even the most simple-seeming organisms. In his late years Francis Crick, co-discoverer of the structure of DNA, became interested in consciousness. Crick once said to Seymour Benzer that he didn’t think the fly was very interesting from the point of view of consciousness, to which Benzer responded—I don’t recall the exact words, but it went something like ‘Francis, don’t underestimate flies; they can do more than you can do: For example, can you fly away and land upside down on the ceiling?’
What are your favourite papers?
There’s of course the one discussed in that fateful journal club in Umeå: ‘The rate of bulk flow from the endoplasmic reticulum to the cell surface’ by Felix Wieland, Michael Gleason, Tito Serafini, and Jim Rothman (Cell 1987: 50, 289-300).
Another eye-opener was ‘The statistical nature of the acetycholine potential and its molecular components’ by Bernard Katz and Ricardo Miledi (J. Physiol. 1972: 224, 665-699). This paper is a fine example of how much you can learn from your data if you have the courage to guess the underlying mechanism and the mathematical chops to formalise your guess. Chuck Stevens’s work is also an inspiration in this regard.
And which of your own papers are you particularly fond of?
Well, there’s the two that laid the foundations of optogenetic control: ‘Selective photostimulation of genetically chARGed neurons’ with Boris Zemelman, Georgia Lee and Minna Ng (Neuron 2002: 33, 15-22), and ‘Remote control of behavior through genetically targeted photostimulation of neurons’ with Susana Lima (Cell 2005: 121, 141-152).
Perhaps because I love literature, I tend to agonize over my own writing. I’m rather slow, managing at most 500 words a day even if I do nothing but write. But occasionally there is a paper that virtually writes itself. The most recent example is the fairy tale of Sandman: ‘Operation of a homeostatic sleep switch’ with Diogo Pimentel and Jeff Donlea (Nature 2016: 536, 333-337).
You take a top-down, reductionist approach in your work. Why?
Our goal is to understand the cellular basis of behaviour. There are two ways of approaching the problem: from the bottom up, or from the top down. If you start from the bottom up, by studying individual cells and their interactions, you quickly run into a problem: there is a horizon of predictability beyond which you cannot see. Even if you understand each individual component and each pairwise interaction in great detail, put just a few of these components together and you’ll discover that you can’t make any predictions at all about the behaviour of the resulting system. The problem is not particular to neuroscience or biology. A famous example is the three-body problem in celestial mechanics. Poincaré showed that the motion of a system of orbiting masses governed by precise Newtonian laws gives rise to deterministic chaos if there are more than two masses involved. So if someone tells me they are going to embark on a massive project in which they will analyse every neuron and every connection in the brain and then model the whole thing in a computer and find out how it works, well good luck but I won’t be closing down my lab just yet.
So what’s the alternative?
Given this horizon of predictability from the bottom up, the rational approach would seem to start from the top down. Find an interesting behaviour that taxes a particular circuit, and then take the system apart. Much of our work is predicated on the belief that brains do not employ an endless variety of circuits but rather a limited set. You need circuits that can compare signals, apply thresholds, or integrate information. You need oscillators to keep time, you need buffers that can hold the intermediates of your computations, you need memory you can write to and read from, and so on. If you understand any one of these circuits in any behavioural context, chances are you have learned something general.
What brought you to Oxford?
After New York I went to Yale. But after a couple of years I was approached by Oxford, and my first thought was, No way! I had spent a month in England as a boy, sent to Bournemouth for elocution lessons. As you can hear, they were not successful. I was bored and homesick and vowed I would never visit this country again. I kept to my promise for 30 years, but my wife persuaded me that I should at least consider the job. And surprisingly, I find that I love it here.