Decision Theory Journal Club: The failure of rationality when foraging

When an animal forages for food, it leaves it current location for what it hopes is a better locale.  We like to believe that this foraging decision is made when the animal expects to get more food if it leaves than if it stays.  Simple and obvious, right?  Unfortunately for our intuition, this doesn’t seem to be the case.  When I was at the Foraging workshop at Cosyne, one of the main themes of the day seemed to be that animals aren’t optimal foragers and they don’t look like our idealized rational economic agents, either.

Let’s consider an animal that is forced to choose between different sources of food.  Life isn’t easy, so it takes some time to gather the food from each source and after gathering it must leave the area before its allowed to gather food there again.  The way this particular experiment happens to be set up, if our animal was an Optimal Forager that was trying to get as much food as possible, it would apportion its time solely on the easiest food source, ignoring the rest.  Anything else is a waste of time and energy.  And yet – and yet! – the animal doesn’t do this; it still will go gather food at one location, move to the next and gather food there, and move on again.  Look at how often (pWait) an animal will wait at a food source versus how often they should wait.  It doesn’t exactly look rational to me.

Rats aren't rational economic actors

Okay, so maybe Optimal Foraging isn’t the way to go; maybe an animal doesn’t apportion its time based on maximizing food rate but uses some other economic criterion.  Maybe it matches its investment and reward so that the more reward it gets from a food source, the more time it spends there?  Data says nope.  Maybe it is a temporal discounter and prefers rewards that happen sooner over rewards that happen later?  Nope again.

Let’s consider another option: maybe the animals get attached to the different options, or just like sampling different opportunities.  What if we introduce a parameter into these different theories that represents an unwillingness to reject potential food options?  This helps a little with the temporal discounting model but where it really shines is when added to the Optimal Foraging model.   But foragers aren’t just always opposed to rejecting a food option.  Rather, as the environment becomes more resource-rich, they are more willing to reject a food option when the environment is resource poor (low opportunity cost of time) than when it is resource rich (high opportunity cost of time).

Rats don't like to give up an opportunity

But does this tell us what we think it does?  It seems like much of the effect is explained by the length of time the animal must wait for the longest food source; the length of the other options don’t seem to affect whether an animal will reject those options.  This gives us another possible option: the task was just too easy for the animals.  I would like to see whether the usefulness of the discounting model changes as everything gets harder and the environment becomes sufficiently ‘resource poor’, and also how animals behave when the basic optimal foraging model predicts something other than “always go to the easiest option”.


Wikenheiser, A., Stephens, D., & Redish, A. (2013). Subjective costs drive overly patient foraging strategies in rats on an intertemporal foraging task Proceedings of the National Academy of Sciences, 110 (20), 8308-8313 DOI: 10.1073/pnas.1220738110

6 thoughts on “Decision Theory Journal Club: The failure of rationality when foraging

  1. That was a very interesting entry.
    Based on the references I must assume that you are only referring to behaviours in rats so now I’m left wondering about other species, more specifically primates.
    I did some research on temporal discounting in great apes and it’s a topic I’m very interested about. Do you have some insights about that?

    • You’re right, this is all in rats, I should have made that more clear. One thing I didn’t mention in this post is that not only didn’t the temporal discounting seem to do a good job of explaining the behavior, but the little explanation the models could give came when temporal discounting was near 0. In other words, it doesn’t do a great job of explaining, but the more you discount the worse of a job that it does.

      The next paper from this journal club that I’m going to post on (next week?) is basically the latest result in economics from an experiment with humans. They basically find that, when you control for motivation etc as much as you can, humans don’t really seem to temporally discount (for money) either. They do, though, seem to for work; in other words, money now might as well be the same as money later (to these relatively well-off undergraduates) but work later is MUCH preferable to work now (again, to these lazy undergraduates ;).

  2. Whenever I see these sort of “deviations from rationality” results, I read them more as informative on how the experimental conditions don’t accurately reflect typical naturalistic settings for the animal. I guess this comes from my interest in things like rule rationality vs. act rationality in econ, subjective rationality for interactions of evolution and learning, and Zollman & Skyrms view of meta-games in EGT. I would really like to see a naturalistic study that shows large deviations from good foraging strategies, that would be fun to try and explain!

    • These are valid points. However, to me, the big surprise with the Wikenhiser and Redish paper is not the deviation from optimal foraging, but the failure of (suboptimal) discounting models to explain behavior. Discounting models generally provide a highly robust explanation for behaviors obtained in a large set of laboratory circumstances, and the Wikenhiser results show a new situation in which they don’t hold true.

      Anyway, I totally agree with you it’d be nice to see a naturalistic study of foraging behavior but these kinds of studies are a gigantic pain in the ass and hugely expensive.

      • This is why I use C. elegans…cheap and easy 😉 I’m looking into a foraging assay right now to see how ‘optimal’ they are; they will, of course, be suboptimal since there are tons of strains that have different foraging strategies.

        The problem I run into for discounting is that I’m not sure how an animal knows the foraging landscape. In a dynamic (and LARGE) world, how do you estimate the potential reward? I have some ideas but if there is a literature on this please let me know

Leave a Reply to adithyaentertainment Cancel reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s