histon, we have a problem (six, in fact)
the extraordinary remote 18th International Linguistics Olympiad
so this past week, the IOL finally returned but not as we remembered it from before.
the 18th edition of the IOL ran from 19-24 July 2021, hosted in the coastal town of Ventspils, Latvia. 216 contestants in 54 teams from 34 countries participated remotely in their home countries, sitting the 6-hour individual contest within a span of 14 hours and the 4-hour team contest within a span of 10 hours, the entirety of which was invigilated in-person and over video — a logistical task so completely monstrous that it highlights just how much had to go right for this olympiad even to take place in a fair environment. there were so many people who worked tirelessly to pull this feat off, a statement which is always true of IOL, but was particularly evident given the current circumstances.
many other things changed because they had to — remote social events were organised by the local organisers, but IOL contestants have always had an initiative to take matters into their own hands when it comes to socialising. a slack server set up for socialising quickly gave way to at least two further platforms, including the traditional FB messenger group chat and a discord server (the highlight of which was the #memes channel that filled the void left behind by the dormancy of the IOL memes page).
the IOL jury, responsible for grading the papers, also had to adapt significantly to the new situation. used to working in a single room for days on end with set breaks for food and sleep, we had the unenviable task of adapting to a 24-hour grading cycle, with our members now spread across the world from California to Japan. we also had significantly more extra work to do than usual: video invigilation was necessarily done by jury members, and in many countries the on-site invigilators were also part of the jury. in the end, with the help of one gruelling all-nighter that left a good few jury members rather sleep deprived, we were able to grade the papers in time without sacrificing thoroughness or fairness, which is remarkable.
in some places, this was helped by the grouping of jury members in specific locations — some travelled to the host city, Ventspils, itself, while those members based in the UK (including myself) congregated in the village of Histon, just outside of Cambridge.
staying in histon was actually really lovely. we were right next to the village green and a stone’s throw away from a small high street. our entire stay there i kept my phone and laptop in Eastern European Summer Time, which was a slightly surreal experience, even if i could check the wall to see the “real” time as well. and when it came to grading, i cannot begin to describe how useful and grounding it was to have other people around who could help me out, e.g. with reading a phrase in Russian or simply keeping me sane through that very long night on thursday. even though England had declared its “freedom day”, we still took occasional lateral flow tests to ensure we were being responsible, one of very few reminders of the world outside of the IOL bubble.
the problems
at this point i think i’ve spoken enough about the organisational side of things now (and probably in a fairly dull way) so now let’s get to something a bit more fun — the problems themselves! the people working in histon actually contributed 4 of the 6 problems that ended up on the individual paper, with the other individual problem and the team problem written by André Nikulin, who was instead based in Brazil.
part of the delay of publishing this newsletter has been waiting for this year’s problems and results to be uploaded to the IOL website (the other part of the delay being that i was exhausted), so that whoever reads this can see exactly what i’m talking about.
the first problem, on the numeral system of Ekari, followed in the footsteps of Alexey’s 2018 problem on Arapesh numerals by being incredibly difficult without strategies like requiring the students to match up correspondences or involving mostly monolingual equations. i was somewhat surprised by the low score, i have to admit, if only because i know that numeral systems were one of the few things i was able to prepare in advance for every IOL and Ekari’s base is one i very much knew off the top of my head as a contestant. the beauty of the problem is of course that knowing the base is only half of it, and this one has plenty more tricks up its sleeves to keep you entangled.
the second problem, on semantic compounds in Zuni, was the hardest problem of the lot, but this was definitely the easiest version we managed to make. the combination of semantic matches and morphological marking was a difficult balance to strike and i think we managed it in the end, even if the scores don’t quite match up. definitely worth a try — it’s one of my favourite problems that i’ve had a hand in writing, and there were so many wonderful compounds Ellie and i had to cull to make it solvable, including abo’yanne ‘sky’ (literally stone/glass + hat) and mo’k’yayanne ‘eyeball’ (lit. a spherical collectivity of water). once the final version was confirmed, i even put together a list of all the words we had included in various versions of the problem and there were 94 of them! in another universe, that could perhaps have been the team problem.
the third problem, on the classifier and demonstrative system of Kilivila, was the easiest problem on the paper, but with an average of 9.41, it’s also by no means trivial. in the end, only two people achieved a full score of 20 out of 20, with many people missing a couple of criteria here not necessarily through lack of understanding but presumably from forgetting to include them in their explanations. this was the problem that literally kept us up at night — given that almost all contestants submitted something for it, and most submitted at least 3 pages, and that the list of criteria we were checking for was large, it’s no surprise that this took a long time to get through. it was clearly worth it in the end, though: the contestants chose it as their favourite problem of the set! i think there’s a lot here to interest people, partly because there’s just so much going on in it anyway. my favourite aspect of it, as a problem author, was being able to include an element of irregularity or ambiguity into the world of the linguistic olympiad, which can often treat language as this regular, logical machine. even the shape of the sentences were artificial — the word order for sentences or within noun phrases is not fixed, but in this problem it made more sense to pretend that they are. hopefully as we move forward, these kinds of pretences will become less and less necessary, as we (hopefully) move towards naturalistic language data.
the fourth problem, on the language game/cant of Agbirigba, was a lot of fun, especially given my interests in vowel harmony. this also has the fun inclusion of some exceptionality, which appeals to me on the same basis as the irregularity in the Kilivila problem, but also involves some bona fide phonology, something that i often find is difficult to find problemable phenomena in that don’t require vast prior knowledge.
the fifth problem, on the verbal morphology of Rikbaktsa, ended up being in the middle of the pack in terms of difficulty, which really speaks to just how difficult this year’s paper was, given that the parsing involved here is quite tough. it is, of course, meticulously written as morphology problems go, and rounds off the set in a very satisfying way. it is noteworthy that this feels like one of the most linguistic problem sets in a while — from Zuni to the end, there’s something interesting to say linguistically about all four problems (about number in nouns, classifier systems/irregularity, feature spreading and phonologically conditioned suppletive allomorphy). Ekari is the odd one out here, not quite melding the numeral and the linguistic to the same degree as Arapesh, but then it’s fitting that it is found first on the set.
i admittedly don’t have much to say about the team problem on Garífuna, Lokono and Kari’ña, if only because i only tested part of it a long time ago and don’t remember much of it. in a normal year, i would have also graded it in part, but thanks to the all-encompassing nature of Kilivila grading, i wasn’t involved at any stage.
this was one of the hardest problem sets we have ever set, if not the hardest. the highest individual score went to Roman Shabanov, from the Netherlands (which felt rather poetic given my recent move there) with a score of 78.3. this is drastically different from 2019, where the winning individual score was 98.7. our ideal problem set is found somewhere in the middle, i think, but given the recent tendency towards higher and higher scores, it was quite nice in a way to see that the race between the increasingly astonishing abilities of the contestants and the problem committee’s imagination hasn’t been lost just yet.
the inner workings of the IOL problem committee and jury have always been somewhat of a black box to the outside world. all that secrecy might lend itself to worrying that we have something negative to hide, when this could ultimately not be further from the truth. this year in particular has proved to me what i already knew: that the people who make the IOL run each year are some of the most talented, dedicated and passionate people i will ever have the good fortune to know and work with. they are people who consistently go far beyond the call of duty to ensure that everyone who participates in the competition each year is assessed fairly and awarded for their achievements and, most importantly, comes out of the IOL having enjoyed themselves and learned more about languages and cultures around the world. i for one cannot wait to do my part, as chair of the problem committee for the 19th IOL (!!!), to continue to make the magic happen.