Why Words Change Meaning: The Science of Semantic Shift and How Language Evolves Under Our Feet

The word "nice" once meant "foolish" or "ignorant." It comes from the Latin nescius, meaning "not knowing." Over the course of several centuries, it passed through meanings including "timid," "fussy," "precise," "delicate," "pleasant," and finally settled into its current sense of "kind" or "agreeable." At no point did anyone decide to change its meaning. No committee voted to upgrade "nice" from insult to compliment. The change happened gradually, through millions of individual acts of usage in which speakers stretched, adapted, and shifted the word's meaning to serve their communicative needs.

The word "awful" traveled in the opposite direction. It once meant "full of awe"--inspiring reverence and wonder. Used to describe cathedrals, storms, and divine manifestations, it conveyed something that commanded solemn respect. Through progressive association with terrifying rather than merely magnificent sources of awe, it acquired a negative charge and eventually settled into its present meaning of "terrible" or "very bad." Today, "awesome" (which followed a similar path from "awe-inspiring" in the original sense) is undergoing its own transformation, having largely detached from any genuine sense of awe to function as a general-purpose positive intensifier.

These transformations are not exceptional. They are normal. Every word in every language is either currently changing meaning or has changed meaning in the past. The English you speak today is full of words whose current meanings would baffle the speakers who used them a few centuries ago. The English your grandchildren will speak will contain meanings that would baffle you.

Semantic change--the shift in word meanings over time--is one of the most fascinating and most misunderstood aspects of language evolution. It is not corruption. It is not decay. It is not a sign that speakers are getting stupider or lazier. It is the natural, inevitable, and necessary process by which language adapts to the changing needs, experiences, and social realities of its speakers.

"A word is not a crystal, transparent and unchanged; it is the skin of a living thought and may vary greatly in color and content according to the circumstances and time in which it is used." -- Oliver Wendell Holmes Jr.

The Mechanisms of Meaning Change

Semantic change is driven by several interacting forces, each pulling word meanings in different directions. Understanding these mechanisms reveals why certain types of change are predictable and recurring across languages and eras.

Metaphorical Extension: The Most Creative Driver

The most creative and productive driver of meaning change is metaphor--the application of a word from one domain to another based on perceived similarity or structural parallel.

Consider how physical experience gets mapped onto abstract domains:

  • "Grasp" originally meant to seize physically; it now also means to understand intellectually (seizing an idea as one seizes an object)
  • "Bright" originally described light; it now also describes intelligence (smart people "illuminate" understanding)
  • "Navigate" originally described steering a ship; it now describes moving through any complex system (navigating a website, navigating a bureaucracy, navigating a relationship)
  • "Cloud" originally described a visible mass of water vapor; it now describes remote data storage (the metaphor of information floating overhead, accessible from anywhere)
  • "Stream" originally described flowing water; it now describes continuous media delivery
  • "Deep dive" originally described a physical submersion; it now describes thorough analysis

George Lakoff and Mark Johnson's landmark 1980 work Metaphors We Live By demonstrated that metaphor is not merely a literary device but a fundamental structure of human cognition. We understand the abstract by mapping it onto the concrete. Most abstract words in any language can be traced back to concrete physical origins. "Understand" likely derives from "standing under" or "standing among"--physically being in the midst of something. "Comprehend" comes from Latin comprehendere, meaning to grasp or seize physically. "Consider" comes from Latin considerare, which may originally have meant to observe the stars (sidera). "Translate" comes from Latin translatus, meaning to carry across. "Inspire" comes from breathing in (spirare)--ideas entering like breath.

This pattern repeats across virtually every language studied. The abstract vocabulary of emotion, cognition, time, and causation consistently derives from the concrete vocabulary of physical action, spatial position, and bodily experience. Metaphorical extension is how language grows its capacity to describe the abstract world from the concrete world that speakers first learned to name.

Social and Cultural Change: Language as Mirror

When society changes, the meanings of words change to keep pace. The relationship between language and thought is deeply intertwined with shifts in cultural reality. Words must track the changing world they describe.

"Computer" once meant a person who computes--a human mathematician performing calculations for scientific or military purposes. During World War II, hundreds of human computers (many of them women, including the teams documented in Margot Lee Shetterly's Hidden Figures) calculated trajectories and data tables by hand. The invention of electronic computing machines transferred the word to the machine. By the 1970s, the human reference had been almost entirely eclipsed, and today the original meaning is so obscure that it requires historical explanation.

"Gay" meant "lighthearted" or "carefree" through most of its English history. Its association with homosexuality emerged in the mid-twentieth century through gradual adoption within LGBTQ communities and gradual spread into mainstream usage. By the 1970s, the new meaning had become dominant, largely displacing the older sense. Today, the older meaning survives primarily in historical texts and requires contextual disambiguation.

"Broadcast" originally meant to scatter seeds widely--a agricultural term for a specific sowing method. Radio technology adopted the metaphor of "scattering" information widely, and the agricultural sense has been almost entirely replaced. The word now describes a mode of media distribution so fundamental to modern life that its agricultural origin is invisible.

"Quarantine" comes from the Italian quarantina, meaning "forty days"--the period during which ships arriving in Venice during the Black Death were required to anchor offshore. Its meaning has broadened from a specific forty-day isolation to any isolation period for disease prevention. The COVID-19 pandemic revived mass familiarity with the word and introduced related terms like "social distancing" and "bubble" with specialized new meanings.

Technological Innovation: The Vocabulary of the New

New technologies create pressure for new vocabulary, and languages typically respond by extending existing words rather than inventing entirely new ones. The new technology gets named through metaphor: it is like something familiar, so it gets the familiar name with the new application.

Every technological era produces this pattern. The railway era transformed "station," "terminal," "junction," "platform," and "express." The automotive era transformed "drive," "gear," "shift," "park," "accelerate," "brake," and "crash." The computing era has been particularly prolific:

  • "Mouse" (computing device): Extended from the rodent based on physical resemblance--the cable looked like a tail
  • "Web" (the internet): Extended from spider's web based on network structure
  • "Bug" (software error): Possibly extended from an actual moth found in a Mark II computer relay by Grace Hopper's team in 1947
  • "Cookie" (stored browser data): Extended from fortune cookie (small package containing information) or magic cookie (computing term predating the web)
  • "Viral" (content spread): Extended from biological virus based on rapid, self-replicating spread--a metaphor closely tied to the dynamics of how internet culture forms
  • "Troll" (internet provocateur): Extended from trolling (a fishing technique) where the provocateur "fishes" for reactions
  • "Thread" (connected messages): Extended from the thread of a conversation or the thread connecting beads

Each of these extensions began as deliberate metaphor and became, within years or decades, so natural that the metaphorical origin is forgotten. The metaphor dies into the literal, and new speakers learn only the new meaning without awareness of any borrowing from another domain.

Generational Differentiation: Language as Identity Marker

Each generation develops its own vocabulary partly to signal generational identity and distinguish itself from older generations. Slang functions as a tribal marker: using the right terms signals membership; failing to use them (or using them incorrectly) signals outsider status.

This mechanism produces two patterns. First, entirely new slang terms: "groovy" (1960s), "rad" (1980s), "phat" (1990s), "lit" (2010s), "bussin" (2020s)--each generation's term for "excellent" or "impressive" uses different words to mark generational belonging. Second, existing words acquire new generational meanings: "sick" (excellent), "fire" (impressive), "slay" (perform excellently), "mid" (mediocre), "sus" (suspicious).

The generational dimension of semantic change is one of the most visible and most contested forms. Older speakers who encounter familiar words with unfamiliar meanings experience the change as confusion, irritation, or evidence of linguistic decline. This reaction is predictable, universal, and invariably incorrect as a prediction of long-term harm to communication. The words of today that provoke linguistic conservatism will be the standard vocabulary of tomorrow, as reliable as sunrise.

The Euphemism Treadmill: The Contamination Cycle

Words that refer to taboo, sensitive, or uncomfortable topics undergo a characteristic pattern that the linguist Steven Pinker named the "euphemism treadmill":

  1. A stigmatized concept exists (disability, death, poverty, bodily functions)
  2. A new, neutral or positive term is introduced to replace the negatively loaded existing term
  3. Through association with the stigmatized concept, the new term acquires negative connotations
  4. Another new term is introduced to replace the now-tainted one
  5. The cycle repeats indefinitely

This pattern is visible across domains:

  • Terms for intellectual disability: "idiot" (medical term, originally neutral) > "moron" (medical term, originally neutral) > "retarded" (medical term, eventually pejorative) > "mentally handicapped" > "intellectually disabled"
  • Terms for poverty: "poor" > "needy" > "underprivileged" > "economically disadvantaged"
  • Terms for death: "died" > "passed away" > "passed" > "is no longer with us"
  • Terms for toilet facilities: "privy" > "water closet" > "lavatory" > "bathroom" > "restroom" > "powder room"

The treadmill reveals an important truth: semantic contamination is driven by social attitudes, not by words themselves. No word is intrinsically offensive. Offense is attached to words through their association with stigmatized concepts. Changing the word changes nothing about the stigma; the stigma migrates to the new word. This illustrates how framing through language shapes perception--the same underlying concept can be presented with radically different connotations depending on the word chosen, but only temporarily.

Natural Drift: The Slow Motion of Language

Some semantic change occurs without any obvious social, technological, or generational driver. Words simply drift in meaning through accumulated minor shifts in usage, a kind of linguistic emergence where small, uncoordinated individual choices produce large-scale systematic change over centuries.

"Meat" once meant food in general--"meat and drink" meant food and water. It gradually narrowed to mean specifically animal flesh, perhaps because animal flesh was the most significant food category requiring specific designation. "Deer" once meant any animal (German Tier still has this broader meaning). It narrowed to mean a specific type of animal as English vocabulary expanded to give other animals their own specific names. "Starve" once meant to die from any cause (German sterben still means "to die"). It narrowed to mean specifically dying from hunger, probably because hunger was the most common cause of death for long periods and the word became specialized for that referent.

These narrowing cases show that language, like ecosystems, tends toward speciation over time: general terms narrow into specific ones as vocabulary expands, and new general terms are created through metaphorical extension to cover the newly vacated conceptual space.

The Main Types of Semantic Change

Linguists have identified several recurring patterns in how word meanings shift. These patterns appear across languages and historical periods with remarkable consistency, suggesting they reflect something fundamental about how human minds process meaning.

Broadening (Generalization)

A word's meaning expands from a specific sense to a more general one:

  • "Dog" once referred to a specific breed; now means any domestic canine
  • "Bird" once meant a young bird specifically; now means any bird regardless of age
  • "Salary" comes from Latin salarium (salt money--money given to Roman soldiers to buy salt); now means any regular payment for work
  • "Holiday" comes from "holy day" (a religious observance day); now means any day of rest or vacation regardless of religious significance
  • "Manufacture" comes from Latin manu facere (made by hand); now includes any production regardless of whether hands are involved

Broadening often occurs when a specific referent becomes the dominant or most culturally important example of a category, so the specific term absorbs the general category's meaning.

Narrowing (Specialization)

A word's meaning contracts from a general sense to a more specific one:

  • "Meat": General food > specifically animal flesh
  • "Deer": Any animal > a specific animal family
  • "Wife": Any woman > specifically a married woman
  • "Hound": Any dog > a specific type of hunting dog
  • "Liquor": Any liquid (as in Latin liquor) > specifically alcoholic drink

Narrowing often occurs when vocabulary expands in a domain, allowing the general term to specialize as other terms cover the vacated ground.

Amelioration (Improvement in Meaning)

A word's meaning shifts from negative or neutral to more positive:

  • "Nice": Ignorant/foolish > pleasant/kind
  • "Knight": Boy/servant > honorable warrior/nobleman
  • "Fond": Foolish > affectionate
  • "Pretty": Cunning/tricky > attractive
  • "Terrific": Terrifying > excellent (via "impressively powerful")
  • "Wicked": Evil > (in some dialects) excellent

Pejoration (Deterioration in Meaning)

A word's meaning shifts from positive or neutral to more negative:

  • "Silly": Blessed/happy (from Old English saelig, related to "soul") > innocent/harmless > foolish
  • "Villain": Farm worker (from Latin villanus, a serf attached to a villa) > wicked person (serfs were associated with violence and low social status)
  • "Notorious": Simply well-known > well-known for bad reasons
  • "Awful": Full of awe/inspiring > terrible
  • "Egregious": Outstanding/remarkable > outstandingly bad
  • "Cunning": Knowledgeable/skilled > sly/manipulative
Change Type Direction Example Old Meaning New Meaning
Broadening Specific > General "Holiday" Religious holy day Any day off
Narrowing General > Specific "Meat" Any food Animal flesh
Amelioration Negative > Positive "Nice" Ignorant Pleasant
Pejoration Positive > Negative "Villain" Farm worker Evil person
Metaphor Concrete > Abstract "Grasp" Seize physically Understand
Metonymy Referential shift "Crown" Physical object Monarchy institution

Metonymy: The Shift by Association

A word's meaning shifts to something associated with it rather than similar to it. Where metaphor relies on similarity, metonymy relies on association, contiguity, or part-whole relationships:

  • "Crown": A physical object > the institution of monarchy (the crown is worn by the monarch, so it comes to represent monarchy)
  • "The press": A printing press > journalism as an institution
  • "Hollywood": A geographic location > the film industry
  • "Washington": A city > the federal government
  • "Tongue": A body part > a language ("mother tongue")
  • "The White House": A building > the executive branch of the U.S. government

How Fast Do Word Meanings Change?

The speed of semantic change varies enormously depending on the word, the social context, and the forces driving the change. Some patterns are consistent.

Rapid Change: Slang and Social Media

Slang changes meaning rapidly because its social function depends on novelty. When a slang term becomes mainstream, it loses its value as an in-group marker and is replaced. The members of the originating group abandon the term precisely because outsiders have adopted it.

Social media has created an additional acceleration mechanism. A new term can spread from an originating community of thousands to a general population of millions within days rather than years. "Ratio'd" (when replies to a tweet outnumber likes, implying the content is widely criticized), "based" (expressing authentic, unapologetic views regardless of popularity), "cope" (dismissing others' views as psychological coping mechanisms), and "NPC" (non-player character, used to describe people behaving as if running a script) all acquired their internet-specific meanings within years and in some cases within months of widespread adoption.

Moderate Change: Standard Vocabulary Over Decades

Most standard vocabulary changes meaning over decades to centuries. "Awesome" took several decades to shift from "inspiring genuine awe" to "really good." "Terrible" took centuries to complete its journey from "inspiring terror" to simply "very bad." "Terrific" traveled in the opposite direction over a similar timeframe.

Slow Change: Core Vocabulary Over Millennia

Core vocabulary--basic words for body parts, family relationships, small numbers, and fundamental natural phenomena--changes meaning very slowly. Words for "mother," "water," "sun," "eye," and small numbers have remained relatively stable across millennia in many language families. Comparative linguists can trace these words through thousands of years of change and find remarkable continuity at the core vocabulary level even as periphery vocabulary transforms dramatically.

The linguist Swadesh developed a "basic vocabulary" list of 100-200 items specifically to enable this kind of long-range comparison because basic vocabulary is the most stable. The stability of core vocabulary is why we can recognize that English "father," German Vater, Latin pater, Greek pater, and Sanskrit pita all descend from a common ancestor word, even though the languages themselves have diverged dramatically over five thousand years.

The Accelerators of Modern Semantic Change

Several factors are accelerating semantic change in the modern era beyond historical norms.

Social media enables new usages to spread to millions of speakers in days rather than years or decades. The community dynamics that once governed semantic change (new meanings spreading through face-to-face contact networks) now operate at the speed of the internet. Twitter/X, TikTok, Reddit, and other platforms create conditions for rapid semantic diffusion that fundamentally reshaped how internet culture forms.

Global communication exposes speakers to more varied usages from more diverse communities simultaneously. English speakers in the UK, Australia, India, Nigeria, and the United States now interact constantly, creating a complex ecology of semantic variation in which multiple meanings coexist and compete.

Reduced prescriptive authority allows innovation to spread more freely. Fewer people consult dictionaries as arbiters of correct usage; dictionary makers themselves have largely abandoned prescriptive stances in favor of descriptive ones, acknowledging that their role is to document living language rather than control it.

Rapid cultural change creates constant demand for new meanings to describe new experiences. The digital economy, changing social norms around gender and sexuality, climate change, and rapid technological development all require vocabulary that existing words must stretch to cover.

Why Semantic Change Cannot Be Prevented

No--prevention is impossible. This finding is one of the most robust in linguistics, supported by centuries of attempted and failed resistance.

Language exists in speakers' minds, not in dictionaries. Dictionaries record usage; they do not control it. When speakers change how they use a word, the dictionary eventually follows. The Oxford English Dictionary's function is descriptive, not prescriptive--it documents what speakers do, not what they should do.

Semantic change is decentralized. No individual or institution controls how hundreds of millions of speakers use words. Change emerges from the aggregate of countless individual usage decisions, none of which requires permission or approval.

Change serves real functions. Speakers change word meanings because doing so serves their communicative, social, and identity needs. Preventing change would require preventing speakers from adapting language to their actual requirements.

The Academie Francaise has been attempting to prevent semantic change in French since 1635--nearly four centuries. French has continued to change at roughly the same rate as other languages. The Academy's authority is real within formal institutional contexts, but it has no mechanism for controlling the language of daily speech, which is where semantic change actually occurs.

Every generation of language commentators has included voices declaring that current semantic changes represent unique and catastrophic linguistic deterioration. Jonathan Swift in the early 18th century deplored what he considered the degradation of English. Victorian grammarians worried that American English was corrupting the language. 20th-century prescriptivists objected to "contact" as a verb, "hopefully" as a sentence adverb, and "they" as a singular pronoun. All of these changes became standard, unremarkable features of English. Every such commentator has been proven wrong by subsequent generations who communicated just as effectively with the changed meanings.

Why Older Speakers Resist New Meanings

The resistance of older speakers to semantic change is one of the most predictable social dynamics in language history. It deserves explanation because it is so universal and so invariably unsuccessful.

Cognitive Investment in Existing Meanings

Speakers have invested decades in learning word meanings, and encountering familiar words with unfamiliar meanings creates cognitive disruption. When a word no longer does what the speaker expects, the speaker experiences friction that feels wrong even when the communication succeeds.

"Literally" functioning as an intensifier ("I literally died laughing") is not just linguistically irritating to those who learned the older meaning--it is cognitively disorienting, because the word no longer provides the disambiguation signal the speaker relied on. The new meaning is not inherently worse; it is simply different, and difference from the familiar is experienced as deficit.

Identity and Language

Language is tied to identity. The way you speak marks you as belonging to a particular generation, region, class, and social group. When younger speakers change the language, older speakers may experience this as an erosion of the linguistic world they inhabit--not just a change in vocabulary but a change in the culture that the vocabulary represents. The complaint about linguistic change is often a complaint about cultural change expressed through the available linguistic vocabulary of objection.

The Cycle of Generational Resistance

The pattern is remarkably consistent across recorded history:

  1. Young people develop new usages that serve their communicative and social needs
  2. Older people notice the change and object, often with moral vocabulary ("corruption," "degradation," "laziness")
  3. The new usages spread despite objections because the communicative community that uses them is larger and more active than the community trying to resist
  4. The young people grow old and begin objecting to the next generation's innovations
  5. The cycle repeats

Shakespeare's contemporaries objected to his linguistic innovations--he introduced or popularized hundreds of words and grammatical constructions that were considered neologisms or errors. Today those same innovations are considered the apex of English literary achievement. The evaluative criteria do not change; the application of those criteria follows generational patterns that have nothing to do with linguistic quality.

Does Semantic Change Harm Communication?

In the short term, semantic change can create temporary confusion between speakers who use different meanings. A teenager saying "That's sick!" (meaning excellent) to a grandparent who hears "sick" as negative has created a brief miscommunication. A professional using "literally" for emphasis while a colleague interprets it as factual assertion has introduced potential ambiguity. These confusions are real but temporary.

The period of confusion is the transitional phase during which old and new meanings coexist and speakers must attend to context to determine which meaning is operative. This transitional phase is uncomfortable, but it resolves as one meaning or the other achieves dominance--or as both coexist as separate senses that speakers distinguish contextually without difficulty.

The Long-Term Picture

Over the long term, semantic change does not harm communication. It maintains communication by ensuring that language remains adapted to the communicative needs of its current speakers. A language that could not change meanings would be a language that could not describe new technologies, new social arrangements, new experiences, or new ideas. It would be a language that was functionally dead--preserved in its form but incapable of serving the needs of living speakers.

Every language in active use is simultaneously ancient in its deepest structures and brand new in its latest innovations. English retains grammatical structures and core vocabulary that derive from Proto-Germanic, spoken approximately 3,000 years ago. It simultaneously coins new words and meanings every day in response to contemporary communicative needs. This combination of stability and change is not a defect. It is the mechanism by which human language has remained humanity's most powerful tool for communication and thought for over a hundred thousand years.

"Language is never innocent." -- Roland Barthes

The words we inherit carry the accumulated history of the communities that used them. The words we innovate carry the emerging needs of the communities we are creating. Both are necessary. Neither is corrupting the other. Semantic change is the mechanism through which a living language remains alive--changing constantly so that it can continue to give people the words they need to describe the world they actually inhabit.

Documented Case Studies in Rapid Semantic Change: From "Spam" to "Woke"

Several recent semantic changes have been documented with unusual precision because they occurred in the age of large digital text corpora, allowing linguists to track the spread and transformation of meanings with statistical rigor that was impossible for earlier changes.

The word "spam"--meaning unwanted electronic mass communication--provides one of the most clearly documented cases of rapid semantic extension. The word derives from a 1970 Monty Python sketch in which the canned meat product "Spam" was repeated relentlessly in a cafe's menu, overwhelming other choices. The connection to unwanted email was made in the early 1990s by users of Usenet discussion boards who used "spam" to describe mass unsolicited postings. Internet historian Brad Templeton documented this transition in a 2003 retrospective, tracing the first documented use of "spam" in its electronic communication sense to a 1993 Usenet post by Joel Furr describing a mass posting incident. Within three years, the term had spread from Usenet to email and had been adopted in journalistic coverage of internet culture. By 2004, Merriam-Webster added the electronic meaning as the primary definition. The transition from technical community slang to everyday vocabulary--crossing the chasm from internet users to the general population--was tracked by Google Books Ngram Viewer data showing exponential growth in published usage from 1995 to 2005. The case is notable for the clarity with which the initial metaphorical extension can be identified and the speed with which it became universal.

The semantic trajectory of "woke" illustrates how a word can be simultaneously adopted and transformed by different communities in ways that produce politically charged ambiguity. The word's origins in African American vernacular, meaning "alert to racial injustice," have been traced by linguist David Bowie and others to documented usages in Black communities from at least the 1930s. Playwright Barry Beckham used it in his 1971 play Garvey Lives!, and Lead Belly's 1938 recording "Scottsboro Boys" includes the phrase "I advise everybody to be a little careful when they go along through there--stay woke." By 2014, following Erykah Badu's 2008 song "Master Teacher" and the Black Lives Matter movement, "stay woke" entered mainstream awareness. Corpus linguist Anna Solin at the University of Helsinki tracked the word's frequency in the Corpus of Contemporary American English between 2014 and 2022 and found a sustained increase in usage alongside a marked shift in syntactic context: from appearing primarily in progressive and African American media with positive connotations, to appearing increasingly in conservative media with negative connotations. By 2020, the same word carried almost opposite evaluative valences depending on the speaker's political community--a rare documented case of simultaneous amelioration in one community and pejoration in another.

The word "literally" offers the most-studied case of ongoing semantic broadening producing prescriptivist objection. Linguist Mark Peters at Grammar Girl documented in a 2013 analysis that "literally" used as an intensifier (rather than as a factual claim about non-figurative meaning) appears in published text as far back as 1769, in usage by the Irish novelist Maria Edgeworth. The usage was widespread in 19th-century literature, including in works by Charles Dickens, Charlotte Bronte, and F. Scott Fitzgerald. Despite this long history, the usage generated significant public complaint beginning in the 1990s and reaching peak media coverage around 2013 when several major outlets published articles decrying the "misuse." Merriam-Webster and the Oxford English Dictionary both added the intensifier meaning to their definitions in 2013, documenting usage while generating further controversy. The linguistically interesting aspect is that "literally" is following a well-documented path of intensifier semantic bleaching: words that begin as factual intensifiers (really, very, awfully, terribly) gradually lose their specific content and function as general emphasis markers, just as "really" (meaning in reality) and "very" (meaning truly, from Latin verus) did in earlier centuries.

The Science of Semantic Change: What Computational Linguistics Has Revealed

Computational analysis of massive text corpora has transformed the scientific study of semantic change from historical detective work into a quantitative science, enabling researchers to measure rates of change, identify the social networks through which new meanings spread, and test theories about what makes some words change faster than others.

William Hamilton, Jure Leskovec, and Dan Jurafsky at Stanford published a landmark 2016 paper in ACL (Association for Computational Linguistics) titled "Diachronic Word Embeddings Reveal Statistical Laws of Semantic Change." Using word embedding models trained on historical corpora spanning 1800 to 2000, they documented two laws of semantic change: the "law of conformity" (words used more frequently change meaning more slowly--high-frequency words like "be," "have," and "go" are the most stable) and the "law of innovation" (words with many closely related synonyms change meaning faster than words with few synonyms, because the synonyms absorb communicative pressure, allowing the original word more freedom to shift). These statistical laws confirm intuitions that historical linguists had proposed but could not previously quantify. The finding that frequency predicts stability has direct implications for predicting which words in current use are likely to shift in meaning over the next century: low-frequency technical or specialist terms are the most vulnerable to rapid semantic change, while high-frequency core vocabulary is the most stable.

Researchers at the University of Vermont led by Peter Sheridan Dodds developed a "hedonometer" approach to measuring semantic change through large-scale sentiment analysis, published in PLOS ONE in 2015. By tracking the average emotional valence (positive or negative) of words used in 10% of all English-language Twitter messages from 2009 to 2015, they documented that the vocabulary of Twitter--and by extension, the semantic environment of social media communication--skews systematically positive: words used on Twitter are substantially more positive on average than words appearing in general published text, news sources, or Google Books historical corpora. This finding was initially counterintuitive (critics expected social media to be more negative) but consistent with a well-established phenomenon in linguistics: people communicate positive information at higher rates than negative because positive experience is more shareable and more prosocial. The study's relevance to semantic change is that social media platforms systematically select for positive-valence language usage, which over time may contribute to amelioration pressure on words that become associated with sharing culture. This represents a platform-specific environmental pressure on semantic change that has no historical precedent.

Linguist Jack Grieve at the University of Birmingham and colleagues published research in 2018 in Nature Human Behaviour tracking the geographic spread of new lexical items (new words or new meanings) across the United States using geocoded Twitter data. They documented that new terms consistently emerge first in specific urban centers (New York, Los Angeles, Atlanta, and Chicago accounted for the majority of documented first uses in their corpus) before spreading through social network connections to other cities and eventually to rural areas. The spread pattern followed a social network diffusion model more closely than a geographic proximity model: cities with dense social media connections to the originating city adopted new terms faster than geographically closer cities with fewer social network ties. The research provides the most precise empirical documentation ever achieved for the social network mechanisms that linguists had theorized about semantic change diffusion for over a century. For socially significant new meanings--particularly slang and politically charged terminology--the diffusion process that previously took decades now takes weeks or months because social media connections allow simultaneous exposure of distant nodes in the network.

References

Tags

linguistics semantic change language evolution etymology word meaning

Frequently Asked Questions

Why do word meanings change?

Through metaphor, social change, technology, cultural shifts, generational differences, and natural linguistic drift over time.

What are types of semantic change?

Broadening (wider meaning), narrowing (more specific), amelioration (more positive), pejoration (more negative), and metaphorical extension.

How fast do word meanings change?

Varies—some words stable for centuries; others shift within years, especially slang. Social media accelerates change for some terms.

What's an example of amelioration?

'Nice' once meant 'foolish' or 'ignorant,' improved to mean pleasant. 'Knight' evolved from 'servant' to honorable warrior.

What's an example of pejoration?

'Silly' once meant 'blessed' or 'happy,' declined to mean foolish. 'Villain' was neutral 'farm worker,' became evil person.

Can semantic change be prevented?

No—natural process impossible to stop. Prescriptivists try but fail. Language evolves regardless of resistance.

Why do older generations resist new meanings?

Investment in existing meanings, identity tied to language use, and perception that change represents decline rather than evolution.

Does semantic change harm communication?

Temporarily—creates confusion between generations or groups. But language adapts, and new meanings become standard over time.

Share this article