Revisiting Bshary & Grutter 2006

In a paper published in Nature in 2006, Redouan Bshary and Alexandra Grutter provided experimental evidence for, both, image scoring by clients and increased cooperation by cleaners in the presence of image-scoring clients in a cleaner fish mutualism, suggesting that the mutualism is a case of indirect reciprocity. Ten years after the paper was published, I asked Redouan Bshary about his motivation to do this study, hicollaboration with Alexandra Grutter and what we have learn since about mechanisms underlying cleaner fish mutualisms.

Citation: Bshary, R., & Grutter, A. S. (2006). Image scoring and cooperation in a cleaner fish mutualism. Nature, 441(7096), 975-978.

Date of interview: 28 November 2016 (via Skype)

Hari Sridhar: I wanted to start by asking you about your motivation for writing this paper. By looking at your publication profile, I came to know that, at this point, you were already about six years into studying this mutualism between cleaner fish and their clients. What was the motivation to do the specific experiments that you present in this paper?

Redouan Bshary: I think it was a mixture. I got aware of the concept of image scoring and indirect reciprocity through the Nowak and Sigmund paper in 1998. I was actually not too convinced by the paper because I thought, if you already make the assumption that there was image scoring (it does not need to evolve) you then get nicer behaviour when you are observed.  That was pretty clear. But it wasn’t clear to me why you should develop image scoring in the first place. Then, by chance, at a conference, I met Claus Wedekind, and he had just done the experiments on humans showing that it works in humans. And he said, yeah, it is actually quite likely that it occurs in animals as well. And because Claus is quite smart, I thought, well, maybe I should think about it, because it is clear that if a cleaner fish has 2000 interactions in a day, quite often, when one interaction takes place there is already a potential client waiting. So, a potential client could observe. I saw that, in my system at least, in principle, there’s the potential for image scoring and that the cleaners adjust to it. And then, I think an intermediate step was that I was very much into biological markets theory from Ronald Noë, who had been my PhD supervisor. In markets, you would predict that if the demand is high, the service quality should go down. And this was actually something that we observed – tried to observe – in nature and quantified. How many clients seek service simultaneously, and how does this influence the service quality? My poor Master’s student, was completely into biological markets, and clearly assumed that service quality should go down, but in the field observation it was the opposite, i.e. if there’s more than one client, service quality goes up. There you have the right ingredient for image scoring and audience effect. And based on these data, we decided, okay, we should do a controlled lab experiment where we kind of manipulate the behaviour of clients.That’s why we use Plexiglas plates rather than real clients so that we have control over it. And the first part is obviously that clients really paid attention to what cleaners are doing. That’s when we used the one-way mirror, and the audience in the middle that sees a cooperative or non-interacting cleaner, and see what they prefer. 

HS: You mentioned that you did a field study with a Master’s student. Was that published?

RB: It’s a book chapter in a book edited by Pete McGregor on communication networks in a 2005 book.

HS: Before I ask you more about this specific study, just stepping back a bit, could you tell me how you got interested in this interaction? Your PhD was on interspecific associations in primates. Why did you make the switch from studying primates to studying cleaner fish?

RB: As a student, I already liked cooperation and game theory best. And then primates were simply the local option that I had to do a PhD. My PhD supervisor was Ronald Noë. He is the one who developed biological market theory, that supply and demand would affect how you split the shares from a cooperative interaction. I really liked his way of thinking. But the funny thing is I had to leave his project to test his ideas, because with primates you could not really work on biological markets. I was really going through literature and trying to find a system where I could look at cooperation and supply and demand, and how it affects how they share. I thought about ants and lycaenids, and I thought about natural gifts in cricket, but then it was actually Hans Fricke, who was also at the Max Planck at the time, who pointed out that maybe cleaner fish would be a good system because they had been studied in the late 50s 60s 70s, before there was proper theory of cooperation and game theory. He said, yeah, there are lots of observations, but the system is not well understood. He gave me some papers to read and I thought that this is potentially really interesting. That’s why I chose to work on cleaners, to look at biological markets.

HS: How did you pick Labroides dimidiatus as your study species? This is the species you’ve been working on right from the beginning.

RB: Yeah, so it was already the best-studied species. I knew a lot about their behaviour, and also the Red Sea is the nearest coral reefs we have from Europe. In the Red Sea, there’s a research station in Egypt. I went and had a look, and that’s where I found exactly the kind of habitat that I wanted – you have these patch reefs surrounded by sand, where you can then distinguish between species that stay at the patch, and so are residents, and species that switch between patches, and so are visitors with access to several cleaning stations. That’s exactly what I was looking for. So I had everything in Egypt, only four hours flight away. Australia was 30 hours of flight time!

HS: But didn’t you also work in Australia, in the Great Barrier Reef, for this study?

RB: That’s true. The field observations are in Egypt. But if you want to do lab experiments, a research station in Australia is much more convenient than the one in Egypt. So that’s why I did all my lab experiments, pretty much, in Australia, and pretty much all field observations in Egypt.

HS: When did you start working in Australia?

RB: The first time I went to the Red Sea is 96. And then, the research started in 97. And then in 98 I went for the first time to Australia. This is because of Lexa Grutter, my co-author on the Nature paper. She studied the same cleaner species before me in Australia. She started in 92 or so, maybe 93. But she was more interested in ecological questions, including the importance of ecto-parasites on fish. So she knew how to breed parasites in the lab. I thought this is really cool, and that we should do stuff together. And so, since 1998, I’ve collaborated with Lexa.

HS: What did each of you bring to this particular study?

RB: It was clear that this was my line of research. It was about cooperation theory and testing game theoretic models. The experimental design was from me and I did the data collection. She helped with research permits, catching the fish, discussing stuff with me and obviously the writing. She’s the native english speaker who can polish it to a level that these top journals like!

HS: Give us a sense of what the work involved, i.e. what was your daily routine when you were doing the work for this study?

RB: You catch the fish and then you actually have two very nice weeks where you just have to feed them and they get used to being in captivity. Then you train them to eat off plexiglass plates. That’s actually very straightforward; it takes two days for them to start eating. You train them that there are different types of plates and discrete food items on it. For the experiments on image scoring – we are talking about the bystander clients – you have four tanks, so you can test four fish on the same day. You start at 7 am, you switch on the light and make the setup. Then at 8 am you start with the light conditions. There’s the one way mirror, and the bystander can see the two cleaners on both sides. At 8.15 am, you put the models in  -one with food;  one without – and start the camera, and then you go out. You come back at 8.30 am and collect it. And then you prepare the next tank and that you start at 9 am, and then at 10 and then at 11. So you do your four experiments – 15 minutes each. And then you can rearrange, because now the clients and cleaners are done with. You remove everything and you put in the new fish. And then you have the afternoon and the next day to watch videos. Then you start the next cycle and the next cycle. I think we had 16 clients in total (4 rounds of 4). This is to look at what the clients are doing. For the second experiment with the cleaners:  they are already trained to eat off plates and know that there are discrete items. On a typical day you would do 20 trials of plates per cleaner where you either have one plate or two plates. And this means that you do it from 8 am- 12 pm and 1 pm–5 pm. You go around with your plates and let them eat and score what they’re doing, and then you put new food items on the plate and you go to the next cleaner and test them. You just go around, all day pretty much, to get your data.

HS: Do you remember the year and months when you did these experiments?

RB: I think the client experiment was 2002-2003, because we repeated it twice to be sure about what we are seeing. The cleaners experiment : I’m not sure whether it’s 2002. What’s in the paper?

HS: I’m looking for it but I don’t think it’s mentioned . Can you tell us a little about the field station in Australia where you did these experiments?

RB: This is Lizard Island research station. It’s an amazing place. They have four houses for the scientists to stay. Each house has, at a time, between four and 10 places. And then they have a central building where you have a library and directors and the lab facilities plus aquarium space. They have four aquarium rooms and plenty of benches outside. If you just needed holding aquaria, the station would provide it. We wanted to have a nice shiny new aquarium so that we can film what’s happening inside. You can buy your own aquarium, store it there forever and simply use it whenever you want. So, whenever I go back so I still have my aquaria there.

We lived on the island. Obviously, you pay bench fees – a fixed amount per day – and there are big fridges and a barge coming every two weeks. You order, with a list, food from a supermarket, and they ship it to you – they pack it for you and ship it. You get your food once every fortnight so you have to make sure that you order the right amount, and don’t forget anything! And then, you have your little boat and tanks for diving, with which you simply go out, catch fish when you need it. Then you set up your tanks and put your fish in.There’s a flow through system so that there’s always high water quality. I don’t know what else to say. The directors and staff members are just amazing. I mean, they do everything to help you with your research. It’s an amazing place to do research.

HS: Do you continue to work in this place?

RB: Yes, I do. I go now every year,  and my group goes there to do research.

HS: Did you have help while doing these experiments or were you doing everything on your own?

RB: Well, as I said, Lexa and I caught the fish together, but then I took care of them and did all the experiments myself.

HS: Did you design the experimental setup you used?

RB: Yes. We already used the Plexiglas paradigm in our Biology Letters paper in 2005 and our Ecology Letters paper in 2002.

HS: Does the experimental setup you used still exist at the station?

RB: We constantly make new plates, but there might be some still lying around.They’re probably getting pretty old now. Half of them are probably thrown away. But some of the plates and the separations, are still out there.

HS:How did you pick Scolopsis bilineatus to use as the client species?

RB: It is very common, and relatively easily to catch and keep in the lab. So, it was just a convenient species, and Lexa’s idea, actually.

HS: For the experiments where you presented a model instead of a real fish, you used a Chaetodon lunulatus as the model. How did you pick this species?

RB: I think this was pretty random. It was just a nice picture on the web. And we wanted a different species. I think this was pretty much it. We could have chosen many other species. I think we could have chosen any from 20 to 50 other species as well.

HS: For this experiment, why did you want to use a different species and not the client species you were testing?

RB: Just in case that scolopsis might be more critical if we use conspecifics and decide that this is not a real species. Because we used a laminated picture. We basically considered that if we take a different species, they (the scolopsis) care less about the fact that it’s not alive.

HS: For the second set of experiments, i.e. the feeding experiments of the cleaner fish,you say that, after previous experiments on partner control mechanisms in cleaning mutualism, you replaced client mucus and parasites with plates. Could you tell us why you didn’t use real fish in these experiments?

RB: There are kind of two reasons. We did a follow up study in 2011 where we have real clients because we got some comments from colleagues who said it would be nice to have it also with real fish. The problem with real fish is that you don’t see the parasites. So, basically, you don’t see what the cleaner fish is doing on them (do they cooperate by eating a parasite or cheat by eating mucus). Whereas if you have a plexiglass plate, and you have these two types of food items, the prawn and the flakes, then you see exactly what they’re eating. And here is the second important thing – which is that you can really decide on conditionality, i.e. what are the decision rules. Because with clients, it’s not that each time the cleaner fish takes a nibble, a mouth contact, they remove a parasite. Quite often they remove dead mucosa. So you don’t see it, but the client doesn’t mind at all. So, basically, with the clients, you never see what the cleaners are doing, i.e. whether it’s cheating or not. There’s only the correlate of the client jolting that you can use. But also you cannot introduce the conditionality where you say, well, if the cleaner eats a preferred item, I remove the client and the bystander.With plates you can have full control over the conditionality. And that’s obviously the big problem with any studies on cooperation – how do you introduce conditionality in there?

HS: I would like to go over the Acknowledgements, to get a sense of how you knew these people and how they helped.

RB: Okay.

HS: The first name is W. Wickler.

RB: Wolfgang Wickler. He was my big boss at Seewiesen, at the Max Planck Institute, when I was doing my PhD, where I also stayed for my post-doc. I started working on cleaner fish while I was still at Seewiesen. He promoted my career a lot. And that’s why I put him in. Also, the year before, he was also on Lizard Island and we discussed about the stuff.

HS: R. Bergmüller

RB: Yeah, Ralph Bergmüller, Karen Cheney, Phil Munday and M. Blows. Basically, we sent the paper to those people and asked for comments. They all commented on previous versions.

HS: Were all these people your colleagues?

RB: Ralph Bergmüller was a post-doc of mine at the time. Karen Cheney was a colleague of Lexa and Phil Munday was also working a lot on Lizard Island. Blows  is actually a colleague of Alexa; I never met him personally.

HS: Do you remember how long it took you to write up this paper and when and where you did most of the writing?

RB: I wrote it in Neuchatel. For me, it was quite a transition. The data were definitely in by 2003. But then, I started in Liverpool as a lecturer, so I had to spend a lot of time preparing lectures. I was only there for one year. In 2004, I came to Neuchatel. And then, I was again busy preparing even more lectures. That’s probably why it took two years to have the version ready for submission.

HS: Tell us a little about the writing process itself. Did you and Alexandra Grutter meet often to discuss the writing or was it done mostly over email?

RB: It was done remotely, back and forth. We might have met on Lizard Island in 2005. Typically, we met once a year, on Lizard Island, at the time. I’m sure we discussed stuff then as well. But it was mostly remote.

HS: Would you say this paper had a relatively smooth ride through peer review? I’m guessing Nature was the first place you submitted it to?

RB: Yeah. We got, obviously, one round of comments, but it was already pretty clear that, as long as we take them seriously, it would get in.

HS: So, not substantially different from the first draft?

RB: No.

HS: How was the paper received when it was published? Did it attract a lot of attention and was it controversial?

RB: No, I don’t think so. I never heard that it was controversial. At least, as far as I know, people were pretty excited about finding such a thing in a fish. It was covered in the media. There was a News & Views in Nature, New Scientist picked it up, and so did Science News, which is kind of the American equivalent.

HS: Did this paper have any kind of direct impact on your career?

RB: Well, it certainly helped me get a more senior status within the Swiss scientific landscape. In the next grant application, I got quite a bit more money. It certainly helped.

HS: Did this paper influence your subsequent research trajectory?

RB: Yeah. I was always testing game theory and cooperation. So, it fit perfectly, in terms of papers that I published before and afterwards.

HS: So, it wasn’t a turning point in terms of your research interests?

RB: No.

HS: This paper also been cited a lot. Do you have a sense of what it gets cited for?

RB: Not so much. I think it’s mainly for the one special case of sophisticated fish behaviour; or the first special case. It’s also cited quite a bit in the social sciences, where they simply decided that this is kind of the pet citation!

HS: What does it get cited for in the social sciences?

RB: In the context of cooperation and reputation, especially the latter.

HS: Today,10 years after the paper was  published, I wanted to ask you to reflect on the main conclusions of the paper. I’ll just read a couple of lines that capture the main conclusions:“Our results show that both requirements for simple indirect reciprocity—image scoring by clients and an increased level of cooperation by cleaners in the presence of image-scoring clients— exist in a cleaning mutualism. Our experimental results confirm interpretations based on field observations.”

Would you say that still holds true more or less?

RB: Well, we actually found “stupid cleaners”, which don’t care about their reputation. And we’re trying to find out, now, the ecological conditions that promote or hinder “reputation management”. So, we still work on reputation; it’s clear that it’s out there. We have reproduced the results in different ways, repeatedly.  But now we have found that, under certain conditions, cleaners don’t have to care about their reputation, and we try to understand this better now – what promotes it. But otherwise, it’s certainly still true.What I personally like about the paper is that we anchored cooperation theory in a much broader context. Because it is true that, within the cooperation literature, there’s hardly anything out there about reputation management, But the idea that interactions takes place in a communication network, and that there are  audience effects, i.e. that you adjust to being observed or not, in a competitive context, it’s everywhere. So, cooperation is just one side of the coin. And, really, we shouldn’t be too surprised about both image scoring and audience effects because those effects are really very widespread in nature. I think it’s important to ground the concepts a little bit. We shouldn’t make too much fuss about it and say it’s cognitively demanding and things.There’s a clearly developed theory out there, and it fits. That’s what I was very keen on: to put it in the general framework of communication networks.

HS: As a thought experiment, if you were to redo these experiments today, would you change anything?

RB: We made the corresponding experiment with real fish, so that now we also kind of know what happens if a bystander observes real interaction. But otherwise, yeah, the, the basic principles of using plexiglass plates and deciding how they behave, is a paradigm that we still use in very much the same way.

HS: Have the sites around Lizard Island from where you got the fish changed a lot in the last 10 years or have they remained more-or-less the same?

RB: No, there’s quite a lot of change. There have been hurricanes that have destroyed some areas. After the cyclones in 2014 and 2015, there was an El Niño in 2016, leading to warm water and coral bleaching. This caused a decline of cleaner densities by 80%. So, there’s been quite some destruction in the last three years on lizard island.

HS: Has it affected the populations of the fish that you study?

RB: Well, we have to see what comes out of this, whether this has any impact on their social behaviour. We still have to find out. We collected data but we haven’t examined the results yet [updated in 2020: published as Triki et al. 2018 in Global Change Biology and Triki & Bshary 2019 in PeerJ: cleaner population down to 20%, fish populations down to 40%].

HS: In the last line of the paper, you say “It will be a challenge to find more complex forms of indirect reciprocity and image scoring—in particular the purely altruistic system that humans are capable of playing—in other animals”.

In the last 10 years, have there been other examples documented of this kind of behaviour? Do we now know of even more complex forms of indirect reciprocity and image scoring in other animals?

RB: No, we’re still waiting for this. I wouldn’t call it altruistic anymore, by the way. That was a terminology that I gave up because Stu West and colleagues, claim that altruism should really only be reserved to the Hamiltonian sense; not to the Trivers’s sense. Fair enough. But, basically, only in humans, so far, is it true that every single act is an investment; there’s no other example out there.

HS: And in terms of behaviour similar to the fish – have there been examples of that?

RB: Umm, let me just think a little bit. I still think there’s hardly anything out there where you have. I should check whether there’s something on rats by the Taborsky group, because they’re definitely studying it, and they’re finding effects. I’m not sure whether that’s published yet. I know his work very well because we are colleagues nearby; he is in Bern. So, I’m very aware of what he’s doing. But there’s very little else like this.

HS: Do you have a sense of why that is? Is it just that the behaviour itself is likely to be very rare or that it hasn’t attracted enough research attention?

RB: I think, in the end, this I’m still pretty much the only one who makes 100% living out of studying cooperation and  game theory and animals. There’s much more on cooperative breeding, obviously. It’s just that it’s not the biggest field to study – game theory and cooperation in animals.And you need the game theory approach if you really want to test these models, rather than putting it more into ecological context, or looking at mechanisms like the physiology underlying social behaviour and stuff. I think that’s probably one of the reasons why there isn’t that much out there.

HS: Have you ever read this paper after it was published?

RB: No. I come back to the methods once in a while, but not the whole paper.

HS: Would you count this as one of your favourites among all the papers you published?

RB: Yeah, definitely. In some ways it was ideal: you have a theory, you have a system where it works, you test it, you have field observations that make very clear predictions, and then you simply confirm the prediction. In this sense, it was obviously a very straightforward paper – theory driven experiments and clear results. It was very nice.

HS: What would you say to a student who’s about to read this paper today, 10 years after it was published? What should he or she takeaway from it? Would you add any caveats to keep in mind while read this paper?

RB: I think the paper itself is very straightforward. I would give the caveat that life is never that simple, so think about extensions and where to take it. I mean, just the fact that we found cleaners was pure accident, but we found cleaners that don’t care about their reputation. What does that mean? Basically, if you put game theory more into an ecological context, you realize that sometimes you find it and sometimes you don’t and there are reasons for this. So, I think, generally, game theory is great but it lacks the ecology. It’s the scientist – the empiricist – who has to add ecology. It’s like with Hamilton’s formula. I mean, it’s such a beautiful formula, but to understand b & c you have to know your species really well!