Loanword doublets

Why does a language borrow a foreign word for something it already has a perfectly good word for?

English had swine long before pork, and cow long before beef. It had ask before question and question before interrogate. On the face of it this is waste: a second word for a job the first was doing adequately, and then a third. Borrowing to fill a gap makes obvious sense — no English word for sushi, so take sushi. But borrowing where there is no gap seems to buy nothing. So either speakers are behaving oddly, or the pair was never really a pair. Which is it?

Test the premise before accepting the puzzle. Are those pairs actually interchangeable? Try substituting. Swine and pork now name different things — an animal and a meat. Ask and interrogate differ in what they imply about the asker. Frail and fragile both descend from Latin fragilis, and yet a frail old man is not a fragile old man. In every case the pair has separated. The duplication we set out to explain does not appear to exist in the present tense of the language.

Which suggests a better question. Not "why borrow a duplicate?" but "why do we never find one?" Two candidate answers, and they are not the same.

The first is that the split happens afterwards, as a repair. Two forms competing for one meaning is unstable — a listener who hears a speaker choose the odd word assumes something was meant by the choice, and if enough listeners do that, something is meant by it. Linguists call the underlying assumption a principle of contrast: different forms are taken to carry different meanings, so a genuine synonym is under pressure from the moment it arrives. Either it acquires a niche or it dies out.

The second answer is that the pair was never equal, because a word does not arrive by itself; it arrives attached to a situation. Pork did not enter English as an abstract alternative to swine — it came through Norman French speakers in whose company that food was discussed, so it was marked from birth. Interrogate came through Latin texts, carrying the smell of a document. The niche is not found later; it comes in the door with the word.

Both are probably true, and they compound. Now push for a prediction. If the loan enters through a specialised channel while the native word already occupies the everyday default, the borrowed member should end up the marked one — more formal, technical or narrower — more often than chance. English obliges almost embarrassingly, in a three-layer stack that repeats across the vocabulary: ask / question / interrogate, kingly / royal / regal, rise / mount / ascend — native, French, Latin, running from plain to elevated.

So far so tidy. Now let us break it, because the counter-cases are where the mechanism shows. English they, them and their are Norse, and they did not settle into a formal register beside native hie and hem — they destroyed them. Egg is Norse and drove out native ey; Caxton, printing in 1490, complained that a northern merchant asking for eggys could not make himself understood by a woman who knew only eyren. Sky, skin and window are the same story.

Does that refute the account? No — it locates it. The Norse contact was everyday and roughly between equals, speakers living side by side in the Danelaw with no prestige gradient, and the two languages close enough that the words were near-interchangeable to begin with. Where contact is like that, a loan can compete for the unmarked slot and win it outright. Register-splitting is the signature of prestige borrowing — a superposed language of courts, churches and books — not of borrowing as such. So the doublet is the middle outcome, between the loan dying quietly and the loan taking the native word's place.

Three refinements, and one caution.

The animal-versus-meat pattern is real, but the story usually told about it — Saxon peasants tending the beasts, Norman lords eating them — is tidier than the evidence supports. It was popularised by Walter Scott in Ivanhoe, and the split consolidated over several centuries rather than appearing at the Conquest; in French itself, porc and boeuf named the animals perfectly well. Treat it as an illustration of the specialising mechanism, not a documented origin story.

Second, "the loanword is the fancy one" is a tendency, not a rule, and the counter-examples above are not exceptions to be excused — they are diagnostic of a different kind of contact.

Third, doublets need not come from different sources. The sharpest are the same word entering twice by different roads: Latin capitale gives English cattle through Norman French and capital through the scholars. Hostel and hotel, fragile and frail, shirt and skirt work the same way. That is the clean case, because the source is held constant and only the route and era differ.

The caution concerns the principle of contrast itself. It describes what happens to near-synonyms well, but whether it is a genuine cognitive constraint or merely a name for the surviving outcomes is argued — we mostly observe the pairs that differentiated, because the others are no longer in the language. What would refute my picture is a carefully sampled set of doublets in which the borrowed member is as often the plain everyday term as the marked one even under prestige contact, or long-lived pairs that remain freely interchangeable across registers.

• Why languages differ so much in how readily they borrow at all, and what makes one resistant.

Every term the collection defines is gathered in the glossary.