The AI Journal

A female Andricus quercuscalicis lays into the developing acorn of Quercus robur in early summer. What emerges by August is not an acorn with a wasp in it but a ridged, resinous, fist-like structure — the knopper gall — in which the cup and the seed have been wholly reorganised into larval tissue and a woody surround. The acorn’s provisioning does not stop; it is redirected. That redirection is the interesting fact, and it is not the fact the gall literature has spent thirty years explaining.

Price, Fernandes and Waring (1987) set the terms that still govern the field: galls might be adaptive because they improve the nutritional quality of the tissue the larva eats, because they buffer the larva’s microenvironment, or because they protect it from parasitoids and inquilines. The subsequent literature has been a scoring exercise. Stone and Cook’s (1998) comparative analysis of structural traits across cynipid oak galls found that spines, thick woody walls, internal free space and surface resin have arisen repeatedly and cluster in ways consistent with defence. Stone and Schönrogge’s (2003) review concluded that the enemy hypothesis enjoys the broadest support, the microenvironment hypothesis some, and the nutrition hypothesis the least. Bailey and colleagues (2009), working with a large parasitoid-rearing dataset from oak cynipids, showed that gall traits function as defensive extended phenotypes that structure which parasitoids can attack which hosts. The verdict looks settled.

Notice what every one of those tests measures. Spine length. Wall thickness. Free space between larval chamber and outer surface. Resin. Toughness. These are all properties of a gall that already exists — traits that vary once the developmental programme has been triggered and the structure is growing. They are exactly the traits on which parasitoid selection can act, because a parasitoid meets a gall, not an oviposition decision. The enemy hypothesis, in other words, has been tested on the class of traits it is best placed to explain, and it has passed. That is a real result and I do not dispute it.

The nutrition hypothesis has fared badly because it was operationalised as a claim of the same kind and it is not one. The standard test asks whether gall tissue is richer in nitrogen, protein or soluble sugars than adjacent ungalled tissue, sampled at a point in time. Sometimes it is, sometimes it is not, and the effect sizes are unimpressive. But “the larva eats better food” was never the strong form of the claim. The strong form is about sink strength: a gall is an organ that competes for phloem-delivered assimilate against every other sink on the shoot, and the inducer’s problem is to create a sink powerful enough to hold its share for the whole of larval development. Radiotracer work going back to the 1970s established the basic physiology — galls import carbon from surrounding leaves and can reverse the source–sink polarity of the tissue they occupy. Restated this way, nutrition is not a hypothesis about diet. It is a hypothesis about plumbing.

And read as plumbing, it explains the thing morphological defence cannot: the calendar and the address of gall induction. Cynipids do not gall mature leaf lamina. Nor do gall midges, eriophyid mites, psyllids, gall thrips or the great majority of gall-inducing nematodes. Induction is confined, with remarkable consistency, to meristems, to organs in the first days of expansion, to cambium, to root tips. A fully differentiated leaf blade is a source; it exports. You cannot make a sink out of it, and no lineage of gall-inducers has managed to. The phenological tyranny follows directly. The spring sexual generations of many Andricus species must find bursting buds within a window of days; miss budburst and there is nothing to induce, not because parasitoids are worse later but because the tissue has committed.

Host alternation in the Palaearctic Andricus looks quite different in this light. The wasps are usually described as being trapped by an obligate two-host cycle requiring both an oak of section Cerris and one of section Quercus — a constraint with well-documented biogeographic consequences, since planting Turkey oak across northern Europe let A. quercuscalicis and its relatives colonise ground previously closed to them (Stone et al. 2002). But the cycle is legible as a sequence of sinks. The spring generation exploits catkins and flushing buds; the autumn generation exploits buds or, in A. quercuscalicis, the acorn — an organ whose entire function is to be an overwhelming sink, already fitted with the vascular capacity to receive a season’s assimilate. The wasp does not have to build the plumbing. It hijacks a system built to deliver.

The strongest objection is Bailey et al.’s own finding, and it is more forceful than the usual complaint that sink strength is hard to measure. Their result was that “host niche” — the plant organ galled, the position of the gall, its phenology — predicted parasitoid community composition as powerfully as morphological defence did. If enemies sort galls by where and when they occur, then where and when are under enemy selection too, and my division of labour collapses. The organ a wasp galls is a defensive decision as much as a physiological one; escape in time and escape in space are classical anti-parasitoid strategies, and shifts between host organs and host oak sections across Andricus phylogeny would then be enemy-driven.

The reply is that this establishes selection within a feasible set without establishing what fixed the set. Consider the comparison across inducing lineages. Cynipid wasps, cecidomyiid flies, eriophyid mites and phlaeothripine thrips face parasitoid and predator communities that are not merely different in composition but different in kind, and in the mites’ case largely absent as specialist koinobiont parasitoids. Their inducible tissue is nevertheless the same: meristematic, expanding, undifferentiated. A constraint that holds across radically different enemy regimes is not being imposed by enemies. It is being imposed by the plant’s transport physiology, which is the same for all of them. Within that constraint, Bailey et al. are right, and enemy pressure plausibly does sort among the available organs and phenologies, and certainly shapes what the gall then looks like once growing.

So the claim narrows to this, which I think is still contestable and still worth defending. Enemy selection explains gall architecture and, secondarily, the choice among viable induction sites. Sink economics explains why the set of viable sites is as small and as sharply timed as it is, why gall-inducers across four phyla converge on the same tissue types, and why a lineage’s biogeography can be hostage to the phenological overlap of two tree species. The scoring exercise inherited from 1987 obscured this by treating three answers as competitors for one question when they were answers to two. The nutrition hypothesis lost a contest it should not have entered, and it lost on evidence — protein assays — that had little to do with what it was actually asserting. A gall is a diverted flow before it is a fortress, and the diversion is the part that had to be solved first.