Decision making, lily pads, and the impact of testosterone

Testosterone has long been the scapegoat for many bad choices. But is the hormone, responsible for aggression, really so influential in decision-making? That’s the question that researchers sought to answer in a new study that focuses on the relationship between testosterone and cognitive abilities. The paper, “Single Dose Testosterone Administration Impairs Cognitive Reflection in Men,” was authored by Wharton marketing professor Gideon Nave, Amos Nadler of the Ivey School of Business, David Zava of ZRT Laboratories and Colin Camerer of Caltech. It was recently published in the journal Psychological Science. Nave and Nadler spoke with Knowledge@Wharton about the research and its implications.

An edited transcript of the conversation follows.

Knowledge@Wharton: Tell us about what you studied and how you studied it.

Amos Nadler: This study looks at testosterone, which is a sex hormone in the body. We wanted to know how this particular sex hormone would affect the way that people, specifically men, make somewhat complicated decisions. Testosterone has been part of mammalian physiology for millions of years, and the role of testosterone has been to essentially ensure survival. It’s been functional in social hierarchies and aggression and basic reproductive functions. We wanted to know how testosterone would affect people’s ability to think through decision-making situations. You could imagine possibly the hormone that is associated with, say, reproduction and fighting possibly would not be conducive to making better decisions. That is what we tested.

We said we’re going to use a large sample because we want to see what the data would look like with strong confidence about the results. We took about 250 young men from Southern California and did a double-blind study, so nobody knew whether they were getting a testosterone placebo, and we looked at the results in terms of their ability to answer three questions.

Knowledge@Wharton: What were the key takeaways of the paper?

Gideon Nave: There is a specific test called the cognitive reflection test (CRT), which is designed to put men in the situation where there is an intuitive answer that jumps to one’s mind, but it is wrong. You need to overcome it and just calm down a little bit and think it through and get to the correct answer. We found that testosterone is making men rely on their wrong, instinctive answer that jumps to their head. This is the takeaway: In the specific case of the CRT, we know that it is bad for you. But in real life many times your intuitions are right, so I wouldn’t make a strong case whether this is good or bad. Obviously, in the case of the CRT, it is bad. But as we know, there are many visual illusions that one can get in the lab. We have people making all sorts of mistakes in laboratory experiments, but when we go out to the real world we don’t see people stumbling up on things. So, it really depends on the context whether the effect of testosterone is going to be bad or good. Again, in the case of the CRT, it was quite bad.

Knowledge@Wharton: Could you give us an example of one of those test questions?

Nadler: OK, I’m going to test you on the spot. See if you can do this. Imagine a pond of lily pads that are growing and doubling every day. On day 48, the pond is full. On what day is the pond half full of lily pads?

Knowledge@Wharton: I want to say it would be on day 24, but I’m sure that’s wrong or you wouldn’t be asking this question.

Nadler: You’re reflecting already on the possibility, but what do you want to say?

Knowledge@Wharton: I want to say that the pond would be half full halfway through the time, which 48 divided by 2, which is 24. But I’ll bet I’m wrong.

Nadler: You’re correct that you’re wrong, but I’m going to push you harder. What is the right answer?

Knowledge@Wharton: I don’t know. I’m going to give up. I want to know.

Nadler: So the first impulse is exactly what most people do. You were able to catch yourself, but what we find is that there is what’s called the “intuitive incorrect answer” — which would be that half of 48 is 24, and that must be the answer. That’s the answer that most people jump to. But if you think about it, that isn’t correct because on day 48 it’s full.

Let’s try backwards induction. We said the lily pads doubled every day. If it’s full on day 48 and you go backwards, how full is it on day 47?

Knowledge@Wharton: Oh, it would be half full.

Nadler: Right, it doubles. The correct answer is 47. When we look at the data for the testosterone group, more of them gave the answer of 24, the intuitive incorrect answer. And the group that received the placebo gel, more of them gave the correct answer. That’s a good example of the cognitive reflection task, where you have the intuitive incorrect answer that you want to jump to but then catch yourself, just like you did. Then you think it through and you’re able to get the correct answer. That’s what the paper is about — that with these three seemingly simple questions, the testosterone treatment reduced their ability to catch that they’re getting the wrong answer. It took them a little bit longer to get the right answer.

Knowledge@Wharton: Were there any conclusions that surprised you?

Nave: One of them is that not only they got it wrong and they got it wrong exactly in the intuitive sense, but we also paid people so there was a clear incentive to get it right. Despite being paid, the people who received testosterone were less likely to get it correct.

We also did a control test that examined the capacity to do these calculations, or how fast they can do addition and subtraction. We found no influence. So, it’s really not about the capacity to do these mathematical calculations, not about motivation or engagement. It’s really about this reflection, this stopping for one second before you give the answer, checking yourself and doing this very little mathematical work to get it right. It was quite surprising that it is very selective to this concept of reflection.

Nadler: I would have to agree that the selectivity was a surprise. We thought that if it affects the mathematical abilities, that it would also affect their cognitive reflection. But we found that it was selective. If you could compare the groups that got testosterone and the groups that got a placebo, there’s really no difference between them with math. You might say testosterone affects mathematical abilities, but we can show that the distributions are identical. They were also given money for getting the right answers on both tasks, so it’s not as if they got a financial reward for the math and no financial reward for the actual cognitive test.

It was surprising how selectivity affected cognition, which really tells us a lot about cognition and how complex it can be. It’s not the ability to add five two-digit numbers, it’s the ability to say, “This may not be the correct answer.” There isn’t an intuitive incorrect answer with the mathematics, and I think probably the difference why we get the separation there is that people don’t generally walk around with an intuitive response to something like that. I found that very fascinating. I think there’s also room for more research to look at how cognition functions.

Knowledge@Wharton: How does this research apply in real life, both for the person who is making the decision and the person impacted by the decision?

Nadler: I would say there are two components. One is looking at the testosterone. All people have testosterone, but testosterone varies, and I think that’s an important part to keep in mind here. What this experiment did was elevate levels of testosterone for one group while it left the other group exactly where they were with the placebo. What we’re really trying to understand is how elevated levels of testosterone might affect your decision-making process. There are situations in life where you would experience elevated testosterone levels, and that’s really what we’re trying to get at. It’s called the activational properties of testosterone. It’s the short-term properties and how it changes behavior in this very short time frame, from minutes to hours, but not in terms of weeks or years.

People experience different states of testosterone, which we simulated. For example, there’s sexual activity that increases testosterone. Testosterone levels rise in preparation for a challenge, which is called the Challenge Hypothesis. There’s also the putative Winner-Loser Hypothesis. If you win something large, the body is supposed to produce a lot more testosterone. It’s shown very clearly in animal literature, which is something we are also looking at in the human literature. The whole point was to see how our own biological states, the endogenous hormones that we release, how those are going to affect the decisions and the way that we think when we’re in those different states. That was the first one, the hormonal aspect. The second part is really about how this paradigm mimics other real-world situations.

Nave: Obviously, what we care about is real-world situations, and there is no consensus on whether this paradigm is exactly mimicking it. We have to keep in mind that we used a drug that is very widely prescribed, and there is some evidence that even some clinics in Wall Street are prescribing it to people working on the trading floor. In this sense, it is quite economically valid, and people who are taking testosterone should be wary that when they do so it not only affects their physiology, it may also have a strong impact on their psychology…. I think this is one very clear implication.

One thing I would be worried of is that we have seen an explosion of stories about this research, and many of them have been quite inaccurate. One main takeaway I see is that not every gender difference between men and women is caused by testosterone. It’s not even necessarily biological. So, I would not jump to a conclusion about saying anything about gender with respect to this research.

Another thing we have to keep in mind is that this was indeed one of the largest studies, maybe the largest study, conducted [on this subject] to date. We see effects that are very consistent, and we think that statistically the results are very strong. But we always have to keep in mind that this is one study, and we are hoping that we are going to replicate it in a much larger sample. We are actually working on this now. I think that we need to all stay with our feet on the ground and make sure that we are going to get more information very soon.

Knowledge@Wharton: You mentioned that people shouldn’t think it completely explains differences between men and women, but are there other misperceptions that this research dispels?

“One main takeaway I see is that not every gender difference between men and women is caused by testosterone. It’s not even necessarily biological.”–Gideon Nave

Nave: In many ways, this research actually confirms some of the misperceptions that people have. Once I gave a talk about this, and a woman came to me and thanked me for proving that all men are idiots. I think the main takeaway as a misperception is that we need to think of what is the function of testosterone and why we see this. Just as I gave this example of visual illusion, many times it is good to go with your gut, especially in situations when responding slowly is especially costly. Competition is one of them. Maybe there are situations when letting your instinct guide your behavior is going to get you a better solution. I wouldn’t jump to conclusions that testosterone makes men stupid, not even men who are taking testosterone. It’s just something that we need to be aware of and know that this drug will influence people in this way, and this hormone can have this sometimes unpredictable influence on our behavior and our decision-making.

Nadler: I certainly agree that this is a study. It is a large sample size, but there is interesting concordance with another study that I ran recently. As Gideon mentioned, there is quite a bit of this being prescribed to Wall Street traders, people working in the financial sector that are moving substantial amounts of money. A recent study that I did looked at the influence of the exact same drug on price bubbles in experimental markets, and it found that the exact same drug that we use and the same one that is being prescribed actually causes prices to go up dramatically higher than their fundamental value and then to crash. The bubbles were larger and they lasted longer.

We could see that these two studies actually fit together, that the traders weren’t looking at the fundamental value of the stock that they were trading. There are some interesting considerations to think about in terms of the real-world impact. It’s not just the state that people are in that we are trying to simulate, it’s that people are taking the drug to artificially raise their levels, and it’s changing the way that they think. Now, that’s not something the FDA thinks about. They’re looking at the efficacy of the drug. But what we’re finding, sort of by accident, is that these hormones do influence other areas of cognition, and that could have macro-economic consequences that need to be looked at as well.

Knowledge@Wharton: The large sample size sets this research apart from other similar research being conducted. Are there other aspects that set it apart?

Nadler: This field is still really in its infancy, and some of the papers that were built upon [included] fairly small samples. Ours, coming in at 243, is dramatically larger than the next study that I can think of offhand that has been published in journals. What we’re trying to do with this is both validate previous findings, and if it’s not there, if we cannot replicate, we need to better understand what the real effect is. So, I think it’s important to not put too much weight on the first study and to be able to replicate with kind of an open mind, and be transparent with other researchers about what the actual effect is. I know that this is something that Gideon has worked on with Colin Camerer, who is a co-author on this paper as well as David Zava from ZRT Labs.

Nave: Another quite outstanding part of this project is that we really monitored people’s physiological state. We didn’t only induce testosterone, but we also measured many other hormones, including stress hormones, so we could really control for the people’s hormonal state during the time that they took the test. One other result that we confirmed is that there was also a link between the stress hormone cortisol and scoring low in this test, suggesting that cortisol is also related to this tendency to give a fast intuitive response that in this case is wrong.

Knowledge@Wharton: What is next for this research?

Nave: First of all, we are aware of the fact that one study is promising, but we are going to keep pursuing this and replicate the effect in larger samples. One advantage of the CRT is that it is only three questions, so it is very easy to repeat it when we do any other task.

Other questions that this research leaves us with include: What is the actual mechanism? Does it make us feel more confident about our intuition? Or maybe it makes us less suspicious overall, or maybe it makes us want to respond faster? There are many possible underlying psychological mechanisms that we need to know better. This is just the first step towards understanding them. But once we have established this effect, we need to really go and find the root cause, and this is what we are planning to do.