A skunked term is a word or phrase whose meaning is changing in such a way that it becomes difficult for people to be sure they are using it correctly.
While terms transition from one meaning to another, they may be used for different and even opposite meanings, which can problematize their use in communications. Skunked terms can also cause issues for software, such as natural language processing (NLP), chatbots, expert systems and machine learning applications, which are reliant on human-generated definitions and may have difficulty with ambiguous terms, especially if the context is not clear.
The lexicographer Bryan A. Garner coined skunked term in the 2008 edition of his guide, Garner’s Modern American Usage. According to Garner, the meaning of a given term may be a point of contention between linguistic purists and liberals for anywhere from ten years to a hundred and should be avoided in formal writing while the disagreement is at its peak:
"A word is most hotly disputed in the middle part of this process: any use of it is likely to distract some readers. The new use seems illiterate to Group 1; the old use seems odd to Group 2. The word has become 'skunked.'"
Common examples of words that are or have been skunk terms include decimate to mean almost destroyed, rather than the original meaning of to kill one in ten and literally to mean figuratively, as in "I literally died" when it is apparent that the speaker is still living. Although linguistic prescriptivists (sticklers, basically) still rail against that use of literally, according to the Oxford English Dictionary, it has been valid for some hundreds of years. Other skunkings involve forms of words such as singular data, criteria and agenda rather than datum, criterium and agendum. Skunked terms can also arise from misused words that catch on, and back formations of existing words.
Language is constantly evolving in the way it is used, and changes in the way words and phrases are used can come about in a number of ways. Verbed nouns, which are very popular in business, are immediately skunked because some people will use them enthusiastically while others deny their validity strenuously. Incentivize, formed from incentive, and Bangalored to mean offshored to Bangalore are a couple of examples.
Skunked terms are common in IT, despite the fact that clarity is essential in technological environments and business writing. The word robot, for example, has always been defined as a machine. However, the term has more recently been used to also refer to software systems with some degree of intelligence and autonomy, such as expert systems, chatbots and virtual assistants. Another evolving situation involves the definition(s) of dark web / deep web. The terms are increasingly used synonymously although their official meaning is distinct.
For technical writers, experts recommend that you avoid skunked terms if possible and otherwise provide enough additional information that the meaning is clear.