Helen Anne Curry and Sabina Leonelli’s chapter in Agricultural Science as International Development: Historical Perspectives on the CGIAR Era, entitled “Crop Descriptors and the Forging of ‘System-Wide’ Research in CGIAR,” advances a provocative argument: IBPGR’s famous, even iconic, standardized descriptor lists — the shared controlled vocabularies used to record traits like plant height, fruit shape and disease resistance in many genebanks around the world — were never just a neutral book-keeping exercise. By 1997, the authors note, something like 80% of genebank curators were using them. Those that didn’t risked being marginalized and excluded from other activities.
The implication is that what seemed like rather “boring” infrastructure quietly decided which traits got noticed, valued and studied, and thus whose priorities counted in the process of conservation and use of plant genetic resources. But not just conservation and use of plant genetic resources. No, descriptors in effect helped drive CGIAR’s whole research agenda for 30 years:
We argue that the project of producing descriptors both defined and embodied CGIAR institutional identity and objectives as these evolved from the 1970s to the 2000s.
I have some sympathy with some of this, not gonna lie. Standards are never innocent or neutral, and it’s a useful corrective to point out that coding conventions, qualitative state labels and data formats can carry as much institutional power as more seemingly consequential, not to say exciting, technical interventions.
But I think the authors overstate their case. Here’s why.
First, this is actually mostly a genebank story, not a user story. The 80% adoption figure is about curators describing accessions; not breeders, researchers or indeed farmers actually drawing on that data downstream. Compliance at the genebank end tells us the descriptor system succeeded as a curation standard. Cool. But it tells us hardly anything about whether it shaped research agendas out in the world, where the more interesting and complex power dynamics actually play out.
And it’s also more of an IBPGR story than a CGIAR story. The descriptor-list project was driven by IBPGR (and its successors IPGRI/Bioversity), the CGIAR center dedicated to plant genetic resources, not by the commodity-breeding centers like CIMMYT or IRRI, whose own research and breeding agendas the chapter wants the descriptors to somehow explain. Treating a program run by one specialized center, albeit with inputs from other centers, as evidence of a system-wide CGIAR phenomenon is over-reaching.
In any case, few people actually use the lists unchanged. In practice, the published descriptor lists tended to function more as a starting template than a straitjacket. Genebanks still routinely add, drop or modify descriptors. Breeding programs often have their own. If actual practical use diverges from the nominal standard as often as it seems to, the claim that descriptors forged a unified CGIAR research agenda needs more evidence. A lot more evidence.
Finally, the aggregation infrastructure barely exists, alas. If descriptors were really operating as a powerful, SPECTRE-like mechanism of control, you’d expect a robust ecosystem of databases pulling together and cross-referencing that data at scale. So where is it? In reality, apart from passport data (the basic what/where/when information about an accession) there are strikingly few databases that aggregate and share descriptor data across genebanks. A standard that isn’t widely aggregated or queried in practice is a much weaker lever on research priorities than the chapter’s framing suggests.
None of this undoes the chapter’s central insight — that standardization is a form of power, and it’s always useful to know who’s wielding power and how. But I’d recommend scaling back the empirical claim. To me, descriptors don’t look like a means to surreptitiously run global agricultural research; they look like a partial, un-enforceable, unevenly-followed standard that genebanks voluntarily adopted because it seemed like a good idea, without it much affecting what happens next.
Which is not to say that it wouldn’t be such a bad thing if descriptors were in fact more deeply and widely adopted, rigorously enforced and consistently used. That would make genebanks better run and more useful, I think. But it still wouldn’t help them, or CGIAR, take over the world, Blofeld-like, cat in lap.