Ofcom's new broadcast guide bands 183 words by how acceptable audiences find them after 9pm. It puts 'nonce', slang for a child sex offender, in the softest band, beside 'shag' and 'jizz'.
Estimated reading time: 7 minutes
In short. Ofcom published a revised Quick Reference Guide to offensive language on 15 October 2026 [[3]], which for the first time assigns 183 individual words a separate rating for broadcast before and after the 9pm watershed, based on an Ipsos survey of 4,289 adults [[1]]. Words scoring a mean acceptability of 6 or above are banded "More acceptable: Words in this category are generally considered more acceptable by audiences and require less contextual justification" [[1]]. In the "sexual references" category, "Nonce", British slang for a child sex offender, is banded "Less acceptable, requiring clear and strong contextual justification" before the watershed, and "More acceptable, requiring less contextual justification" after it, the same post-watershed band as "Bonk", "Cum" and "Jizz" [[1]]. The guide tells broadcasters its ratings "should not be taken in isolation" and should be read "in the light of the accompanying research report" [[1]]; that 86-page report never discusses the word qualitatively, and its own methodology admits "the research did not seek to validate respondents' interpretations of the words tested" [[2]]. Three documented mistakes: four stars.
Ofcom commissions this research to inform how it enforces Section Two of the Broadcasting Code, the rule requiring "generally accepted standards" to protect the public from "harmful and/or offensive material" [[2]]. The Quick Reference Guide is the practical document broadcasters are pointed to when deciding whether a word needs an apology, a warning, or no caution at all. The 2026 edition is new in one specific way: for the first time, each of the 183 tested words gets its own rating before the watershed and after it, rather than one blended score [[1]]. Among those 183 words is "nonce".
The claims, tested
| The claim | What we found | Verdict |
|---|---|---|
| The guide's ratings reflect "a mean acceptability score... calculated for each word in each broadcast context", with words scoring 6 or above banded "More acceptable... requiring less contextual justification" [[1]], and the renamed categories are designed to show "the contextual acceptability of each word in the specific broadcast setting tested, rather than assigning it a fixed level of offensiveness" [[1]] | "Nonce" sits in the "sexual references" category alongside "Bang", "Bonk", "Cum", "Frigging", "Ho", "Jizz", "MILF", "Prickteaser", "Shag", "Skank", "Slag", "Slapper", "Slut", "Spunk", "Tart" and "Whore". After the watershed, all of them, including "Nonce", carry the identical rating text: "More acceptable, requiring less contextual justification" [[1]]. The seven categories used to sort all 183 words, general swear words, body parts, sexual references, race/nationality/ethnicity, religious references, mental health and physical health conditions, and sexual orientation and gender identity, have no category for words whose ordinary meaning is an accusation of a specific crime rather than crude or sexual language [[2]] | A word whose everyday use accuses someone of sexually abusing a child is scored, categorised and worded identically to generic sexual slang; the guide's own seven-category framework has no bucket to catch the difference |
| The guide instructs broadcasters that "the ratings in this guide should not be taken in isolation. Readers should interpret the ratings in the light of the accompanying research report, which provides further detail on the contextual factors that influence acceptability" [[1]], and that report sets out an extensive qualitative treatment of which categories of language carry risks beyond crudeness, finding that "language relating to race, nationality and ethnicity was among the least acceptable categories tested and some terms were viewed as harmful regardless of broadcast timing" and that "intentional or repeated misgendering and deadnaming" were "generally considered unacceptable" [[2]] | Searching the 86-page report for "nonce" returns exactly one entry: a row of raw percentages in Table 1.6, showing 16% of respondents found it acceptable for broadcast before the watershed and 51% after it [[2]]. There is no qualitative discussion of the word anywhere in the report's Executive Summary, its four-factor framework for context ("what was broadcast", "how was it broadcast", "who was saying it", "why was it being said"), or its sections on protecting specific groups from harm [[2]] | A broadcaster who follows the guide's own instruction and reads the research report for context on "nonce" finds nothing beyond the number that produced its rating |
| The guide's methodology section states plainly: "Respondents were not provided with definitions of the words included in the survey. Familiarity with, and understanding of, the meaning of each word was therefore self-reported and is likely to have varied between respondents... The research did not seek to validate respondents' interpretations of the words tested" [[1]] | That caveat is applied as a blanket note covering all 183 words, with no word-by-word assessment of which terms it matters most for. "Nonce" carries "High familiarity" in the guide's own rating [[1]], meaning most respondents recognised it, but recognising a word is not the same as registering that its ordinary use is an accusation that someone has sexually abused a child, a claim capable of triggering vigilante violence against the wrongly accused, rather than mere crudeness [[1]] | The guide identifies, in its own words, the exact risk that respondents' understanding of a word might not match its real-world weight, then never checks whether that risk applied to the one word on the list where it mattered most |
The mistakes, counted
No category exists for words that accuse someone of a crime (1). Ofcom and Ipsos sorted all 183 tested words into seven categories: general swear words, references to body parts, sexual references, references to race, nationality and ethnicity, religious references, references to mental health and physical health conditions, and references to sexual orientation and gender identity [[2]]. "Nonce" was placed in "sexual references" [[1]]. Its ordinary use in British English is not to describe a sexual act or body part but to accuse someone, often inaccurately and sometimes fatally for the person accused, of being a child sex offender. None of the seven categories, and none of the four contextual questions the qualitative research used to probe acceptability, "what was broadcast", "how was it broadcast", "who was saying it", "why was it being said" [[2]], asks whether a word functions as an accusation of a specific crime. The result is a rating sentence, "More acceptable, requiring less contextual justification" [[1]], identical to the one given to "Bonk" and "Jizz".
The report the guide tells broadcasters to consult says nothing about this word (2). The Quick Reference Guide is explicit that its ratings are a starting point, not a final answer: "the ratings in this guide should not be taken in isolation. Readers should interpret the ratings in the light of the accompanying research report" [[1]]. That 86-page report gives sustained, specific, qualitative attention to language touching race, religion, disability and gender identity, stating for example that "language relating to race, nationality and ethnicity was among the least accepted categories tested and some terms were viewed as harmful regardless of broadcast timing" [[2]]. "Nonce" gets no equivalent treatment. It appears once, in Table 1.6, as two percentages: 16% acceptable before the watershed, 51% after it [[2]]. A broadcaster who does exactly what the guide asks, reads the report for context, finds none.
The guide names its own blind spot, then does not check whether this word falls inside it (3). The methodology section states that respondents were given no definitions of the 183 words and that "the research did not seek to validate respondents' interpretations of the words tested" [[1]]. That is a candid admission that a respondent's acceptability score for a word might not track the word's real meaning. It is not a hypothetical risk for "nonce": the word's ordinary use is an accusation of a specific, serious crime, and a respondent who registered only "offensive slang I've heard before" without weighing that accusatory force would produce exactly the kind of score the guide's own caveat warns about. The guide records "High familiarity" for the word [[1]] but makes no attempt to establish whether that familiarity extended to its real-world consequences.
Credit where due
The guide is unusually candid about its own limits. It states outright that respondents were not given definitions and that their interpretations were never validated, rather than presenting its ratings as settled fact [[1]]. It explains why the "moderately acceptable" band was deliberately narrowed ("mean acceptability scores tended to cluster around the midpoint of the ten-point scale... a broader range would have resulted in a disproportionately large number of terms being assigned to this category") [[1]], a transparency about its own statistical choices that most guidance documents skip. The underlying report's treatment of language touching race, religion, disability and gender identity is extensive and careful, showing the research was capable of giving a category special qualitative attention when it judged the stakes warranted it [[2]]. And the dual pre/post-watershed ratings are a genuine improvement in granularity over the single blended scores the guide replaces, giving broadcasters a sharper tool for the other 182 words on the list.
Verdict
Four stars, from three documented mistakes. Ofcom's new guide is more detailed than its predecessor and more candid about its own methodology than most comparable documents. But the categorisation that sorts 183 words into seven buckets has no way to separate crude language from a word whose ordinary use accuses someone of a specific, serious crime, so "nonce" ends up rated identically to "bonk" and "jizz". The report broadcasters are told to consult for context says nothing about it. And the guide's own admitted blind spot, that nobody checked what respondents actually understood by the words they rated, was never tested against the one word on the list where the gap between "offensive slang" and "accusation of child sexual abuse" matters most.
The star score counts three documented mistakes. First, the guide's seven-category framework has no category for words whose ordinary meaning accuses someone of a specific crime, so "Nonce" is scored, categorised and worded identically to generic sexual slang such as "Bonk" and "Jizz". Second, the 86-page research report the guide instructs readers to consult for context contains no qualitative discussion of "Nonce" at all, only two percentages in a data table, despite giving extensive qualitative treatment to language touching race, religion, disability and gender identity. Third, the guide's own methodology admits respondents' interpretations of the 183 words were never validated, a risk it does not check against this specific word. Three falls in the 1 to 3 band: four stars; the bands are on the ratings page. This piece makes no finding about the conduct of any named individual at Ofcom or Ipsos; the guide's category design and the report's contents are documented facts, not any one official's decision. Checked directly against the Quick Reference Guide and the full research report, both fetched and read in full 16 October 2026.
Sources
- Quick Reference Guide to offensive language, Ipsos UK for Ofcom, October 2026
- Public attitudes towards offensive language on television, radio and video-on-demand, Ipsos UK for Ofcom, October 2026
- Context and intent shape audience views on offensive language, Ofcom, published 15 October 2026