When is a bench not a bench? Semantic transfer and the OED
Lauren Simon holds an MSt in Linguistics, Philology, and Phonetics from University of Oxford. She is currently a venture capital investor, funding purpose-driven startups, but continues to pursue linguistic research topics, primarily in semantics.
Register for our upcoming webinar Semantic transfer and the OED, featuring Lauren as a panelist.
Consider, for a moment, the word bench. Bench made its first appearance in the sense of ‘A long seat, usually of wood or stone, with or without a back,’ in early Old English—that is, between 600 and 950 A.D.; by the time the OED1 entry for bench was published in the third fascicle in 1887, it had 9 senses and 11 subsenses. Today, after a 2017 revision, there are 11 senses of bench and 21 subsenses, the most recent dating to the 1940s. One may ‘approach the bench’ in a courtroom, ‘ride the bench’ in a football match, or ‘bench press’ heavy weights. The word bench can now also indicate the person (or people) who occupy various benches, as in ‘Our team has a deep bench.’
But how and why did bench develop new meanings? Questions like this are at the heart of historical semantics (the study of changes in meaning), yet they have historically been challenging to answer. New senses added to existing words can be hard to spot and even harder to study at scale, since they are identical in form to preexisting ones. How would one search for new senses of bench without having to comb through all the current ones? More importantly, as new senses emerge, how would one know to search for them?
This is where the OED3 revision process comes in; as lexicographers identify and define new senses, they not only bring entries up to date but also produce a wealth of data on how meanings have changed—an invaluable resource for historical semantics researchers. Patterns that emerge from this data can then answer questions like, ‘When do we appropriate an existing word to refer to something novel rather than bestow an entirely new name?’; ‘Do words belonging to one category of meaning (a ‘domain’) tend to evolve toward meanings in certain other categories?’; and ‘Do meanings tend to broaden or narrow over time?’ The answers to questions like these can then be considered in the context of cognitive linguistic theories about how the human brain may interact with language, both in any given moment and through time.
Yet the data embedded in revised OED entries alone is not enough to compile data on changes in meaning. While OED entries include etymological data about how a word—and sometimes a sense or subsense—entered the English language, they only occasionally contain information about how additional senses developed thereafter. So in order to trace patterns of meaning change within English, it is necessary to apply labels to each new sense indicating the change(s) that produced it.
This labeling process was tested on a subset of new senses, specifically those that were added to existing OED entries and refer to a person or people. Examples of these new senses include, ride, n.2, sense 5b (‘He’s my ride’), concession, n., sense 5c (‘Concessions get in for half price’), and report, n., sense 4 (‘The hire will manage a staff of five direct reports’). The below chart shows the frequency of various semantic transfers, that is, of various ways that an original ‘source’ sense’s core content can change to produce the new ‘target’ sense, within that sample.
The majority of the new senses were produced via metaphorization, which exploits similarity between an original source meaning in one semantic category and a new, target meaning in another (forming a ‘metaphor’).
One example of metaphorization within the sample is, genie n., sense 3c, defined as ‘A person likened to a genie in being noted for his or her talent within a certain sphere, or ability to perform difficult or seemingly magical acts.’ The ‘likening’ of such a person to a magical genie (sense 3a) constituted metaphorization and, when repeated enough times, ultimately resulted in the conventionalization of the new sense (3c).
Another 29% of the semantic transfers occurred via metonymization, which instead exploits ’contiguity’ between the target of a new meaning and the original source meaning. (forming a ‘metonymy’) This contiguity can be physical coincidence or overlap, as in, ‘The White House has no comment,’ in which the building metonymically refers to the government situated there. Metonymizations may also draw on partial contiguity, as in well-known part-for-whole patterns like, ‘He has a good ear,’ in which the body part is conceptually contiguous with musicality (‘a good ear’) because of its role in facilitating the ability.
The fact that twice as many of the semantic transfers occurred via metaphorization than metonymization makes sense; the ‘similarity’ mechanism of metaphorization can develop any source meaning into any target, whereas metonymization can link only contiguous categories of meaning. Yet the ‘contiguity’ criterion that restricts metonymization also makes it a rich resource for mapping categories of meaning from a cognitive perspective. In other words, because metonymization can highlight which meanings are ‘next-door’, in a sense, enough of them can help map the landscape of semantic categories as our brains would traverse it. Patterns of metonymizations in the person-referring sense provided enough information to begin mapping of some of these pathways.
Metonymizations in the sample that relied on a contiguity between cause and effect, for instance, almost always also converted from verb to noun somewhere along the way. This makes sense in light of the integral role of actions and events to establishing any cause-effect relationship. One example from the data is punk, n.1 and adj.2, sense B.3b, defined as ‘A coward; a weakling’ which converted from punk, v1., sense 2, ‘To display cowardice; to back out from cowardice; (hence) to withdraw one’s support, to quit’ then metonymized to refer to the person displaying cowardice.
The cause-effect metonymizations in the sample also almost always cast the person as the cause rather than the effect, which is consistent with patterns in countless languages that reflect a tendency to privilege the agency of humans over creatures, objects, and environments. With ride, n.2, for instance, the effect (a ride in a vehicle) is metonymized to refer to the cause (the person giving the ride). Push n.2 in the logging industry slang sense (11) of ‘a foreman; a boss’, was metonymized as the ultimate cause of a push n.2 (sense 1a), ‘an act of exerting vigorous effort for or to achieve some goal’.
Finally, the specific actions or events that appeared as effects in the data were also almost always those that are exclusively associated with humans, i.e. those involving complex emotions, thinking, or self-awareness. Examples include cowardice (punk, n.1 and adj.2), short-temperedness (pop-off, n. and adj.), and complex actions like driving (ride, n.2), reporting (report, n.), rehiring (rehire, n.), exhorting (push, n.2), and carving (scrimshander, n.).
While these are just initial observations, they suggest that certain characteristics of people may influence how likely any given word is to develop a sense referring to a person.
Similar observations emerged from senses that metonymized locations (in time or space) to refer to a person or people located there (‘locative metonymizations’). Most of the senses in the sample that underwent locative metonymizations, even those with definitions describing an individual person, were supported by quotations that only illustrated plural use. The locations that the new senses developed from were also largely ones typically designed to hold multiple people, as with dressing room, n., sense 2b, ‘The members of a sports team collectively’ which metonymized from sense 2a, ‘A changing room at a public venue; spec. a room at a sports ground or performance venue set aside for actors, performers, or competitors to use before and after a performance or game, for preparing, changing clothes, etc.’.
These patterns are not surprising when considered in light of real-life communication needs. While identifying a passenger by their location (as in, ‘Seat 24J has his call light on’) works as a one-off, occasions for identifying an individual by their location relative to others are idiosyncratic and—in the absence of airplane seat numbers—reliant on where the observer is positioned within the scene. This context-dependence makes it less likely that a locative metonymization referring to an individual will become conventionalized and thus be added to the OED. Groups of people, on the other hand, can often be identified by location without much context-dependence.
The opinions and other information contained in the OED blog posts and comments do not necessarily reflect the opinions or positions of Oxford University Press.