LIT-063: The indeterminacy of word segmentation and the nature of morphology and syntax

Martin Haspelmath (2011), Folia Linguistica 45(1), 31–80 (2011). Reprinted in Folia Linguistica 51(s1000), 2017, as DOI 10.1515/flin-2017-1005, which is the identifier the survey gave — DOI-10.1515/flin.2011.002

Key takeaways

• None of the ten standard criteria for a cross-linguistic “word” (pauses, free occurrence, mobility, uninterruptibility, non-selectivity, non-coordinatability, anaphoric islandhood, nonextractability, morphophonological idiosyncrasy, deviations from biuniqueness) is necessary and sufficient, and no combination reproduces orthographic practice. So linguistics has no principled basis for a universal morphology–syntax division, and morphosyntax should be treated as one domain.

Seeded from the abstract and a skim, before the reading in NOTE-073. What follows is what the work says about itself.

The morphology–syntax distinction is usually taken for granted, but it depends on a cross-linguistically valid notion of the morphosyntactic word. Haspelmath argues no good criteria for such a notion exist. He examines ten criteria and shows none is necessary and sufficient alone, and no combination yields a definition matching linguists’ spelling conventions. “Word” can be defined per language or as a fuzzy concept, but the fuzzy version only helps if the affix–word–phrase continuum shows clustering, and there is no systematic evidence of that. So there is currently no good basis for splitting morphosyntax into morphology and syntax.

Standing in the record

Filed from the survey of 2026-09-25 of work the anthology set aside as out of scope (tier B): 480 seconds of active reading over 8 sessions in the papers-feed tracker. It was filed Deferred, unread. NOTE-073 is the close reading of 2026-09-25, and it placed the work: Active — This is the standard statement of the negative case about the cross-linguistic word, and both LIT-071 and LIT-031 cite it for exactly that; read it with the author’s own later constructive follow-up, “Defining the word” (WORD 69(3), 2023, not read here).

When seeded, before the reading in NOTE-073, the entry gave a priority for a deeper reading: high — 480 s over 8 sessions, it anchors the linguistics cluster, and it is the one work here with a direct tokenization link. The skim covers the argument, but the criterion-by-criterion evidence in §3 is what a filing would cite.

What the seeding asked a deeper reading to check (NOTE-073 is that reading):

• It bears directly on tokenization. If “word” has no cross-linguistic definition, whitespace or word-level segmentation is an orthographic convention, not a linguistic unit, which supports subword and byte-level approaches (H-Net and similar). The triage asks for the “when in doubt, anthology” test here, since the anthology’s representation-and-encoding topic may be a better home than a new linguistics tag.
• It is the typological foundation that c62 (Krauska & Lau, who cite the 2017 reprint) and c63 (Bruening) build on.
• A deeper read should check §6’s clustering argument, since that is the empirical claim a corpus or statistical study could actually test.

Access when seeded: Crossref for both DOIs. The 2017 record (Folia Linguistica 51, issue s1000, dated 2017-01-28, author name stored as “Haspelmath Martin”) has the same title and abstract as the 2011 original (45(1), dated 2011-01 with no day), so I treat it as a reprint and date the work by its 2011 first appearance (YYYY-01-01 because no day is given). De Gruyter PDFs returned an empty 202 challenge. I read the author’s deposited 2011 PDF on Zenodo (record 225844, 50 pp., keywords “word, morphology, syntax”): abstract, all section heads, §4, §7, §8, §9 and §10 (conclusion). Filing decision for the caller: keep the survey’s DOI (2017 reprint) or switch to the 2011 original.