A label, a person, a token
In current English an identifier is whatever picks one thing out of a crowd. It can be a person whose job is to establish who someone is. It can be a number, name, or code that points uniquely at a record. It can be a guidebook that tells a mushroom from its lookalikes. In computing it is the formal name of a variable, function, process, or user, and in databases it often names a primary key. Everyday speech stretches the same idea to barcodes, serial numbers, and the strings that stand in for people in a file once the real names have been stripped out.
The origin of the word identifier is later and more local than that range of uses suggests. The noun is not an old inherited form. It is an agent noun built in English from the verb identify plus the native suffix -er, first recorded in 1870 with the sense of a thing that identifies.
From French into English
The verb arrived more than two centuries earlier. In the 1640s English took identify from French identifier. Accounts of the last step do not match word for word. One standard etymology traces the French verb to identité, the French noun for sameness. Another points to Medieval Latin identificare. A third analysis splits the same Latin pieces as identicus plus facere, to make. The disagreement is about the last step of formation, not about the raw materials. French and Medieval Latin both built a verb meaning, in effect, to make the same, or to treat as the same, on a learned noun of sameness.
English identity itself is a little older than the verb: around 1600, from French identité, already in French by the fourteenth century, from Medieval Latin identitas. An earlier English spelling, idemptitie, shows up in the 1560s from a Medieval Latin form idemptitas, with an extra p that later dropped out of the standard word. Identical, from Medieval Latin identicus, is of the same generation, first as a term in logic in the 1610s.
So identifier sits at the end of a chain: a nineteenth-century English agent noun on a seventeenth-century French loan, itself a learned formation over Late Latin identitas.
What the Latin is made of
Late Latin identitas meant sameness. It is an abstract in -tas from classical Latin idem, the same. Idem is a reinforced demonstrative: id, that or it, plus a suffix of the same family as English speakers still meet in the citation formula idem, used to avoid repeating an author’s name. Behind id lies a Proto-Indo-European pronominal stem reconstructed as *i-, the same distant relative that shows up in words for that one over there in several Indo-European languages.
The combining stem ident- is slightly awkward. Idem did not offer a clean stem on which to hang -tas the way unitas hangs on unus. Accounts differ on how the extra t got there. Some connect ident- with identidem, again and again, itself from idem et idem. Others see analogy with abstract nouns such as entitas. The Late Latin noun has also been described as a calque of Greek tautotes, sameness. None of those proposals is a settled proof; what is clear is that ident- became the stem for the whole family: identitas, identicus, and identificare.
The second half of identify is the verb-forming element English writes as -fy. It comes through French -fier from Latin -ficare, a combining form of facere, to make, from the Proto-Indo-European root *dhe-, to set or put. Read morpheme by morpheme, identify is to make the same. Identifier is then the agent: the person or thing that does that making. The gloss is almost too neat for modern ears, because we no longer hear identify first as a logical operation of equating. In the 1640s that was exactly the point.
Related words keep the same Latin spine. French identité, Spanish identidad, Italian identità, and German Identität sit beside the agent nouns identificateur, identificador, identificatore, and Identifikator. English identification, from French identification and recorded from the 1640s as the act of treating one thing as another, belongs to the same cluster.
How the senses moved
Identify did not enter English meaning to name a stranger in a lineup. Its first recorded sense was to regard as the same: to treat A as identical with B. That is a philosopher’s and logician’s verb. By 1769 a second sense is in use: to determine the identity of someone or something, to recognize or prove it to be the same as a known person or object. Around 1780 a reflexive and associative sense appears, to make oneself one with a group or cause. Only in 1886 is the verb attested as serving as a means of identification, the inanimate sense in which a number or a mark does the identifying.
The noun identifier, dated 1870, belongs with that middle history. It names the agent of determining identity, at first typically a thing rather than a person. The personal reading is still live: a racetrack official who checks tattoos, colours, and markings is an identifier in the plain agent sense. Field guides that help a reader name a bird or a mineral are identifiers of another kind. The computing sense is a later specialization of the same agent noun: a token that refers uniquely to a variable, a process, an HTML element, or a row in a table. Abbreviations such as ID, from identification and recorded from the mid-1950s, travel with that technical use but are a shortening of a different noun.
The oldest English sense of the verb, treating two things as one, has not vanished. It survives in identify with, in the psychological vocabulary of identification recorded from the nineteenth century, and in the later idiom of identifying as a member of a group. Those uses share the Latin idea of sameness. What changed is the object: not two propositions in a logic book, but a person and a community, or a person and a name they claim.
The same, made usable
A useful residual of the history is how close the programming jargon still sits to the Latin. When a compiler treats every occurrence of the same identifier as the same object, it is doing, mechanically, what Medieval identificare was coined to do: make this instance the same as that one. The 1870 agent noun was already waiting for a world that needed millions of such sameness-markers. The word did not have to be reinvented. It only had to be pointed at code.