Headlinese

Why can a headline drop half of its grammar and still be read without difficulty?

A headline says Minister Resigns Over Leaked Report. Written out properly it would be The minister has resigned over a leaked report, and nobody reading the short version noticed anything missing. Strip a sentence like that in ordinary prose and it sounds like a telegram or a fever dream; strip it above a news story and it reads as normal.

The obvious conclusion is that the removed words — the, a, has — were doing nothing. That deserves a hard look, because if they were genuinely inert, headlines ought never to go wrong. They go wrong constantly, and always in the same way.

Consider Squad Helps Dog Bite Victim, which is about a victim of a dog bite and reads as a squad assisting with the biting. Or British Left Waffles on Falkland Islands, in which a noun phrase and a verb phrase have swapped places entirely. Or the one that gave the whole genre its nickname: Violinist Linked to JAL Crash Blossoms, a 2009 headline about a musician whose career flourished after her father died in a plane crash, and which a copy-editing forum seized on as the name for the type — the crash blossom.

Look at what fails in each. Not meaning: nobody misunderstands what dog, bite or victim denote. What fails is parsing — which word is the verb, which noun group belongs to which. And the words that would have settled it are precisely the ones deleted: an article marking where a noun phrase starts, a form of be marking a verb as a verb, a relative pronoun marking a clause as subordinate. So function words are not empty. They are the scaffolding of the parse.

Then why can they usually go? Because they are the most predictable things in the language. English has a handful of articles, a handful of auxiliaries, one copula. At each of those slots the reader faces two or three options, heavily constrained by the surroundings, and restoring the right one is nearly automatic. Content words are the opposite: at ___ Resigns almost any human noun could appear. The principle is unremarkable once stated — you can delete what the receiver can reconstruct — and it is also why the frequent words are the short ones, both facts falling out of the same predictability.

But note what this does not say. Predictable does not mean harmless. Some function words are both easy to guess and load-bearing for the parse, and dropping those is exactly what breeds crash blossoms. The copula is the worst offender, since without it a word that is ambiguous between noun and verb has nothing to disambiguate it.

There is a second thing going on, and missing it makes headlinese look purely subtractive. Headlines have their own positive grammar, learned like any other. Present tense reports a past event: Queen Opens Bridge happened yesterday. A bare infinitive marks the future: PM to Resign. A comma stands in for and, a colon for said, as for simultaneity. None of these are English's ordinary rules. So a reader of headlines is not decompressing a sentence — they are reading a small separate register, and the evidence that it is learned is that children and second-language readers find headlines disproportionately hard.

Why does the register exist at all? The cause was physical: a headline had to fit a fixed column width in metal type, and the count was unforgiving. Redundancy did not create headlinese; it permitted it once the width demanded it. Which is why the style loosens where the constraint has — online headlines, written for search and sharing rather than for a column measure, drift back toward full sentences.

Two refinements tie the reasoning together.

First, "half the grammar" is not deleted at random. What goes is a specific and fairly stable set — articles, forms of be, auxiliaries, relative pronouns, sometimes of and possessives — and what stays is everything whose identity the reader could not have guessed. The selection is governed by predictability, not by some intuition about which words carry meaning.

Second, this is not the same trick as compressing a file. Lossless compression exploits statistical redundancy in the data itself and guarantees exact recovery. Headlinese throws information away outright and relies on the reader to regenerate it from a shared model of English and of news. That is why it has a failure mode at all, and why the failures cluster where the shared model is thinnest: unfamiliar names, unfamiliar subject matter, and words that could be either noun or verb.

• Why noun pile-ups such as School Funding Row Talks Collapse are readable at four words and impossible at six.
• How search and social sharing changed headline grammar once column width stopped constraining it.

Every term the collection defines is gathered in the glossary.