Do any languages not have the concept of "words"?

What counts as a word depends on who you ask

Linguists do not agree on a single definition of 'word' that works for every language. Some languages make word boundaries obvious through stress patterns or pauses; others do not. In the polysynthetic languages of northern Australia—Bininj Gun-wok, Ngalakgan, and Murrinhpatha—a single verb can contain many meaningful pieces that English would split into separate words. The question of where one word ends and the next begins is not trivial there [4].

The variation is real even within a single language family. In Murrinhpatha, all stem elements of a verb integrate into a single prosodic word with fixed penultimate stress, though some suffixes sit outside it. In Bininj Gun-wok and Ngalakgan, by contrast, the polysynthetic verb can contain multiple prosodic constituents, potentially anchoring multiple accents, and accent placement varies considerably [4]. So even the prosodic cues that listeners might use to identify 'a word' behave differently across these languages.

When words are hard to pin down, language still works

Written Chinese social media offers a different kind of challenge. The paper on 'Martian Language' describes online Chinese text full of neologisms, creative character placements, and combinations of writing systems. Readers often rely on visual cues, reading aloud, and knowledge of current events to understand a post. Building a fixed lexicon is difficult because inventing new ways to refer to a word or concept is easy and common [1].

This does not mean Chinese speakers lack the concept of words. It means that in some contexts, the word as a stable unit is not the most useful level of analysis—for humans or for machines. The authors argue that new uses of language in social media will challenge machines' abilities to operate with words as the basic unit of understanding, not only in Chinese but potentially in other languages [1].

Words show up across very different language families, but not always the same way

Romance languages have been analyzed using 'wordhood' as one parameter alongside expressibility and coverage when studying phasal polarity adverbials—the equivalents of 'already,' 'still,' 'not yet,' and 'no longer'—across 45 Romance varieties belonging to 34 distinct Romance languages [2]. The fact that wordhood is treated as a measurable parameter suggests that words are a meaningful unit in these languages, even if the analysis reveals diverging phenomena [2].

Pidgin and creole languages raise related questions about function words and their properties. The study of question words in these languages examines their formation and role in linguistic genesis, asking what role the lexicon of the colonial language played [3]. This line of inquiry treats words—including small function words like question markers—as identifiable units whose origins and behavior can be traced.

Words may be an early achievement in language evolution, not a universal given

One philosophical paper argues that the evolutionary origins of words matter for understanding what words are. There is consensus that lexical communication was an early achievement in the phylogeny of our language capacities, and that words became available to our ancestors before the maturation of other components of the grammar. At the beginning of their evolutionary trajectory, words were thus likely to have been significantly different from the vocabulary items of contemporary languages [5].

This suggests that the concept of 'word' is not a fixed universal that every language either has or lacks. Instead, words are a solution to the problem of pairing sound with meaning—one that different languages implement in different ways, and one that may have looked quite different in the deep past [5]. The question 'Do any languages not have words?' may be less useful than asking how each language draws the line between a word and its neighbors.

Sources used in this answer

Related research questions