The lawyer's office was on the third floor in Achrafieh and the lift was out, which everyone treated as weather. I had come about my grandmother's flat, or more precisely about a document concerning the flat, and I had rehearsed on the plane from Manchester. Understanding him was never the problem. I understood every word, including the ones about the encumbrance. Then it was my turn, and I heard myself produce, in a clean unaccented Beirut Arabic, this sentence: my mother says it is fine but we want to see the paper first.
That is a nine year old's sentence. It has a borrowed authority in it, and a want, and not one clause capable of holding a condition. What I needed to say was that we would accept the arrangement provided the charge was discharged before signature, and I had all of that in English, and about a third of it in a French I will not inflict on anyone, and in Arabic I had my mother says it is fine.
He was kind about it. That was the bad part. He switched to English and he did it smoothly, the way you take a heavy bag off somebody without making a thing of it.
I flew back and started reading about language attrition, on the assumption that there was a literature about losing a language and that I was in it.
I am not in it. The term is narrower than the feeling.
Monika Schmid, whose Language Attrition (Cambridge University Press, 2011) is the standard reference, draws the boundary in a footnote in one of her survey articles: attrition covers "migrants who left the country in which their L1 was spoken and in which they had acquired it after puberty", and it "should be distinguished from incomplete acquisition, that is, the restructuring of linguistic knowledge by the acquisition of another language earlier in life". Elsewhere she puts a number on the line. Quantitative studies find the actual degree of change "surprisingly low", she writes, "as long as the speaker did not migrate before the age of 12". Below that line things look different: "among younger speakers, it appears that knowledge of the native language can indeed be lost dramatically or even totally, while post-puberty migrants often perform within the native range on a variety of tasks."
I left Beirut at four, Accra at eleven, and my mother's kitchen at eighteen, and not one of those is the number the research is measuring. It measures leaving the country where the language is spoken, and I did that before I could read it. So by the field's own definitions I am not an attriter. I am the other thing. I dislike the word incomplete and my dislike is not an argument.
The wider finding is stranger and more useful. Adults essentially do not lose their first language. Schmid reports that speakers with the longest migration you could fit inside an adult life, five to six decades, show remarkably few signs of deterioration, and concludes that "a completely acquired native language system can, in fact, be extremely stable and impervious to 'atrophy' or deterioration in any structural sense". Studies hunting dramatic vocabulary loss keep failing to find it. What they find is more disfluency in free speech, slightly slower naming of low-frequency words, and an effect Schmid reads as competition rather than damage: "it is not so much that attriters 'forget' their L1, but that the competing L2 has become too strong to be entirely inhibited."
So my problem was never a leak. Nothing drained out. There was less in there to begin with, and the missing part has a name.
Arabic is diglossic. The colloquial variety comes from the house and Modern Standard Arabic comes from school, and for a child raised outside the Arabic-speaking world there is no school, so there is no Standard. Albirini, Benmamoun and Saadah tested Egyptian and Palestinian heritage speakers in the United States and found that they "display significant competence in their heritage colloquial varieties" while there are "gaps in that knowledge", along with "significant transfer from English, their dominant language".1 That is a good description of me and it stops short of the interesting part, which is that the gap is not distributed randomly. It is exactly the register in which Arab adults conduct adult business.
I did not lose the language of contracts and condolences and complaints to a utility company. It was never delivered. My Arabic came out of a kitchen in Osu where four people who all spoke English were being affectionate at each other in Lebanese, and a kitchen is not a register, it is a room.
Schmid has a line that made me put the paper down. Formal use of a language, at work, under pressure, in front of monolinguals, is what keeps it in shape. Family use is not: "Use of the native language within the family or with friends in the country of migration, that is, informally and with other speakers who are also bilingual, does not place the same constraints against mixing the languages on the speaker, and is therefore less conducive to language maintenance." The only Arabic I ever had was the kind that does not maintain Arabic. And the vicious circle, which Schmid credits to Grosjean and Py, is that you then take the output of other bilinguals as evidence of what the language is, and everyone drifts together, holding hands.
She also notes that attriters do worse in formal test conditions than in free speech, and reads it carefully: the difference "may to some extent be linked more to issues of confidence in the speaker's own proficiency than to an actual erosion of knowledge". What varies between individuals, she argues, is "not so much a matter of underlying knowledge but of executive control". I have read that sentence about nine times. In the lawyer's office I was not searching for words. I was watching myself being about to fail, in real time, at speed, and the watching is what took the words.
A note on the thing everyone brings up. Hartshorne, Tenenbaum and Pinker put roughly 670,000 English speakers through a grammar test and reported in Cognition in 2018 that grammar-learning ability holds up until about 17.4 years and declines from there, much later than the old story. A 2022 reanalysis in Language Learning, by van der Slik and colleagues, took the same data apart and found a continuous decline fits better than a cliff for monolinguals, bilinguals and early immersion learners, with the sharp break surviving only in the later and non-immersion groups. Both papers are about acquiring a second language, not about what happens to a first one, and I ran the two together for about a fortnight before I noticed. There is a real age effect. There is no cliff anybody fell off with a bang.
Twice a week now I sit at the kitchen table in Levenshulme with a tutor in Beirut on a laptop and we go through Modern Standard Arabic out of a textbook written for fourteen year olds. I can read a newspaper column slowly and aloud. I cannot argue in it. There is no scene at the end of this where I ring the lawyer back and finish the sentence.
The thing I resent is smaller than the loss and harder to say out loud. When I speak Arabic well, people tell me my Arabic is beautiful, and they mean it warmly, and what they are hearing is a child's fluent Arabic, which is the most charming register in any language. Being told you are charming in a language is how you learn you are not being taken seriously in it.
I have stopped correcting them.
Albirini, Benmamoun and Saadah, "Grammatical features of Egyptian and Palestinian Arabic heritage speakers' oral production", Studies in Second Language Acquisition 33:2 (2011), 273 to 303. They also counted codeswitching in the narratives, which is the part of the paper I have quoted at people most often. ↩