Caste and Ecology in the Social Insects

by Edward Wilson and George Oster

1/5/2020

I've probably read only around 5 books per year since I graduated from college. This year I'm resolving to read 10 times as many -- one per week.

Reading slowly has its advantages, though. The books I've read in the last three years have had time to percolate and assimilate into my memory and I can still recite obscure factoids from them. Richard Nixon founded a fraternity at Whittier college called the Orthogonians. In the lead up to the 1953 Iran coup that re-installed the Shah, 4 out of the 5 Tehran newspapers were on the CIA's payroll.

But now that I've finished my first book of 2020, I've had only a few hours to process it in its entirety and I'll already be starting the next one this afternoon. So I'm going to record my thoughts on it here. Maybe I'll do the same for the rest of them.

I picked up Caste and Ecology in the Social Insects by chance at a labyrinthine used bookstore in Ohio. The book is from 1978 and it's possible no one has read it since then (between the pages I found an old airplane ticket I used as a bookmark that had rules explaining the plane's smoking section!). In fact it's possible no one has read any copy of the book since the 70s as it doesn't even make E. O. Wilson's top 30 "Main Works" listed on Wikipedia. However since I am recently in the business of knowing as much as possible about ants I gave this one a read.

The book is filled with math. Differential equations make an appearance in every chapter, as do charts and graphs of every variety. Wilson and Oster set out to mathematically formulate the basis of physical castes in the social insects (mainly ants, but also bees, wasps, and termites). In insects, "castes" refer to groups of individuals of the same species with different physical attributes (most obviously body size) that therefore exhibit different behaviors. For example leafcutter ant colonies consist of tiny "minima" ants that tend the colony's symbiotic fungus, minor and major "workers" that cut and transport the leaves that feed the fungus, comparatively huge "soldiers" that have no task except to fight predators, and the even larger queen who is the sole parent of all the others (the males die shortly after mating).

The worker castes are notable in that they are discrete -- there are no ants whose bodies partially match a major worker and partially match a soldier. So instead of simple variance producing workers along a normally-distributed spectrum of size, some other biological mechanism bucketizes ants into discrete castes with little variance within each caste. They briefly cover what the biological mechanisms might be, but focus more on why having castes at all would be evolutionarily beneficial.

The book attempts to explain the caste differentiation of social insects like the leafcutter ants in mathematical terms by arguing that producing ants of different sizes allows the colony to more optimally propagate its genes to future generations. Their model is to take the set of all tasks that ants perform and plot them in a multi-dimensional space where the location of the task in this space defines the morphology of an ant that optimally performs that task (note that they don't actually plot the tasks of any specific species of ant, but rather just offer this concept theoretically). One could plot a sphere around each task that delineates the boundary beyond which the ant's body shape makes the task impossible to complete. The set of all task-spheres forms a sort of cloud of ant body shapes required to accomplish all tasks posed to the colony. Then they argue that, depending on the distribution of tasks specific to a given species, splitting workers into discrete physical castes more optimally matches this task-cloud than having a simple normal distribution of ant body attributes.

Sounds plausible, right? (Well, maybe. I just covered in four sentences what Wilson and Oster cover in three chapters, so maybe the previous paragraph doesn't make any sense at all...) Regardless, this model falls short in three ways:

Most social insect species don't even have physical castes outside of the queen-worker dichotomy. Edward Wilson (and pretty much every ant biologist I've read) was obsessed with division of labor by caste in ant colonies like leafcutters, but this kind of organization is the exception in social insects, not the rule. Fully 80% of social insect genera have only a single monomorphic worker caste where division of labor is entirely behavioral instead of physical. And only four genera have more than two castes. Instead, most ants expand their optimal task-cloud by coordinating their efforts and working together instead of producing individual ants optimized for the job. The book barely touches on ant coordination and cooperation at all.

The model implies causality where there might not be: they argue that an ant species like the leafcutters evolved multiple castes to better fit the array of tasks at hand, but it could just as well have been that they already had multiple physical castes for other reasons and then later adopted the set of tasks we observe and simply don't perform whole realms of other possible tasks unsuited to their bodies (e.g. breaking open seeds or hunting small arthropods).

Finally, who is to say that evolution would bring about the optimization of anything at all? For any given biological trait, it is impossible to attribute its existence to any particular adaptation that trait enables. For example, it is not true that giraffes have long necks "just so" they can eat the leaves off of tall trees. Male giraffes also use their necks in fights with each other (definitely look up giraffe fights on Youtube, it's wild). But just because giraffes use their necks for those things doesn't mean evolution gave them long necks for those specific purposes. On the other hand evolution didn't act totally randomly as clearly along the line somewhere some zebra-things with longer necks than their siblings produced more-viable offspring. The only real conclusion we can make about the evolution of specific traits is that they are partially the result of direct fitness gains, partially the result of the fitness gains of other traits that must go together because of quirky mechanisms of that species' biology, and partially because those traits randomly occurred in individuals and weren't so bad for them that they couldn't still survive and reproduce.

Wilson and Oster address some shortcomings of their approach in the book, but in my opinion they offer only weak justifications for why their arguments supersede the above pitfalls, mainly centering around the need for more empirical studies. Regardless, reading this was a good exercise for me in quickly digesting dense non-fiction. I also learned some interesting factoids I will share here:

Wilson categorizes ant colonies into three developmental phases: founding, the "ergonomic" growth phase when the colony produces exclusively sterile workers that help grow the colony further, and the reproductive phase when the colony "cashes in" its investment in numerical growth to produce exclusively new reproductive females and males. In one chapter, they show mathematically that this strategy of producing only sterile offspring followed by producing only reproductive offspring is genetically optimal for colonies whose workers cannot survive the winter. They call this a "bang-bang" reproductive strategy. And here I thought the term "bang-bang" was invented by Louis C.K. to refer to eating two full meals one after the other. Also, during the peak of the reproductive stage fully 70% of a colony's individuals could be otherwise-useless young queens and males. The colony has to be so efficient to support so many individuals who aren't contributing!

At least in 1978, it was thought that 13 families of insects independently developed eusociality (that is, form colonies where some members reproduce and some only help raise their siblings but don't reproduce themselves). 12 of these were in the ants, bees, and wasps, and then once in termites. Wasps and bees have practically no instances of discrete physical castes, and ants mostly are monomorphic as well. But termites almost all have very well-delineated roles and castes within their colonies. Maybe the book would have done better focusing on termites instead of ants. Termites rarely forage since they feed on fungus that they tend, and besides the occasional exterminator their primary threat is coordinated attacks by ants, so it makes sense why some individuals might develop specialized traits to fight them (see the termite vs ant battle in the movie Antz).

Lastly, Wilson and Oster noted that along with division of labor ants also have a division of initiative. A small minority of workers are more adventurous and active than their siblings and are the primary scouts and defenders. The rest are only stirred to action after being recruited by their more vigorous nestmates. I had imagined ant colonies as consisting of entirely interchangeable members, but apparently some are more independent than others. This might be an adaptation though since docile ants are more likely to help out their nestmates than be off on their own adventures. This is most striking in the slave driver ants that send one ant in each direction to look for neighboring ant colonies to pillage. When a lone ant returns knowing the location of prey, it sets off a chain reaction of communication and alert that spurs the entire colony to follow it to its target.

This turned out longer than I intended (and took three hours to write!), but will hopefully be a good resource for me to return to down the line in order to refresh myself on this book. I might even write another review next week, especially if this one is received well. I also want to try to improve this webpage since I set it up back in 2013.

Back to Blog