The Joke That Would Not Travel

ANo branch of translation admits defeat as often as the translation of jokes. A sentence carrying information can be carried across; a sentence carrying a pun carries its own language with it, and the language cannot be moved. In 1978 the Belgian critic Hubert Vandelle described comic translation as "the discipline's confession", the place where the pretence of equivalence is dropped in public. Two responses have been available ever since. The translator may render what the original says and let the joke die on the page, sometimes with a footnote to explain what has been lost; or the translator may write a new joke, unrelated in wording, which is meant to do to the second audience what the first joke did to the first. The second option, known as functional substitution, is now standard practice in subtitling and dubbing alike. Whether it is translation at all is a question the field has been arguing over for four decades, and the argument has lately acquired data.

BRenata Molnar, who directs the subtitling archive at the Skalborg School of Applied Linguistics, set out to establish what translators actually do rather than what the manuals instruct them to do. Between 2014 and 2021 her group annotated 1,240 comic passages taken from 18 feature films released in four languages, classifying every translation as literal rendering, substitution or omission. Wordplay was substituted in 71 per cent of cases and cut outright in 12; humour resting on situation rather than on words was rendered literally in 84 per cent. Molnar draws a deliberately modest conclusion from this. Substitution, she argues, is not a licence seized by bold translators but a routine forced by the clock: a subtitle must fit two lines and roughly six seconds, and an explanation does not fit inside either. Her critics, she remarks drily, have never had to fit an explanation into six seconds.

CThat defence invites a test, and Molnar's group has run one. In 2022 they showed 320 viewers a comedy in a language none of them spoke, subtitled either literally or with substituted jokes, and recorded audible laughter in the room. Substituted subtitles produced laughter at 2.4 times the rate of literal ones, and the viewers who saw them rated the characters as warmer. The effect, however, held only for jokes built on the sound or the shape of words; where the humour came from a character's situation the two versions were indistinguishable, and in three scenes the literal version drew the louder response. Molnar publishes the exception as prominently as the finding. What her figures establish, she insists, is not that substitution is the better method but that it repairs one specific kind of damage, the damage done when a joke depends on a language the audience has no access to.

DJulian Petrakis, a scholar of comparative literature at the Institute of Narrative Media in Bergen, does not dispute the laughter counts; he disputes what laughter is evidence of. A joke, he argues, is never merely a device for producing amusement. It is also a piece of characterisation: it tells an audience who is speaking, what that person finds funny, and what they are prepared to say out loud. Replace it and the amusement may survive while the character quietly changes into somebody else. To demonstrate this he asked 240 bilingual viewers to describe the personality of a single character after watching one version or the other; those who had seen the substituted subtitles used the word "crude" more than twice as often. Petrakis concedes the obvious limitation without being pressed: his viewers knew both languages and may have been reading as critics rather than as audiences, and he examined one character in one film.

EThe dispute has consequences well beyond the seminar room. Since 2021 a large streaming service has required its subtitling contractors to log every substituted joke in a file delivered with the episode, a rule drafted after complaints from rights holders rather than from viewers. Petrakis regards this logging requirement as the most useful development of the past decade, because it creates a record that can still be inspected long after the translator has moved on. Molnar, whose laughter figures are the ones the platforms themselves most often quote, is here the more sceptical of the two: a log, she points out, records that a substitution was made without recording what the alternative would have cost, and a translator who leaves the pun untouched and unfunny logs nothing at all. Both object to the industry habit of advertising subtitles as accurate without ever saying accurate to what.

FWhat has emerged is less an answer than a division of the material. Where a joke's machinery is formal — a rhyme, a mispronunciation, a surname that happens to mean something — the loss under literal rendering can be described precisely, and a substitution can be judged by whether it rebuilds the same machinery. Where the joke rests instead on who is speaking, the loss is harder to name and far easier to cause, and no quantity of laughter measured in a viewing room will detect it. Whether comic writing is drifting towards the formal or away from it, neither researcher will say. What both refuse is the framing that has organised the public argument, namely a choice between a faithful translation and a funny one, when what a joke says and what a joke does were never separate to begin with.