Source
Bruening (2018), LIT-071 (Language 94(1), 1–42), read in full in NOTE-065, with the abstract and opening of Müller’s reply (2018, Language 94(1):e54–e66).
What was actually shown
The target. The lexicalist hypothesis in one sense only: a separate, encapsulated word component that feeds syntax one way (p. 1). Not “lexicalist” in the sense of rich lexical entries, of which the paper has “absolutely nothing to say” (p. 2). Words are granted for the argument (fn. 2).
Wrong, on three counts.
- Phrases feed word formation (§2). Phrasal compounds (a don’t-you-dare! look), zero-derivation from phrases (know-it-all, Bonnie and Clyded) and overt derivation (do-it-yourselfer, unputdownable) are productive, obey phrasal syntax inside, and stay open to anaphora ((12)). The “quotation” defence fails, because the encapsulation it gives is strict cyclicity, which holds of every stem ((15)–(18)).
- Phrasal operations reach sub-word units (§3). Word-part ellipsis (pre- and post-revolutionary) respects morpheme boundaries ((24)–(29)) and is the process seen in right-node raising. Focus targets morphemes (a-moral, not im-moral).
- The claimed differences in principle fail (§4). Locality (self-), adjunct scope (Venda causative, Japanese -sugiro), headedness (enrage, two years ago) and idiosyncrasy (multiword expressions at least as numerous as words, fn. 17) do not separate words from phrases.
Superfluous (§5). A formal X0, “a segment of a syntactic node that includes nothing but a single bundle of morphosyntactic features” ((62), p. 25), makes complex heads and phrase-plus-null-N0 both X0s. Then: extraction targets only XPs, which rules out word parts and also whole words ((55)–(58)); apparent X0 coordination is phrasal (the but did not test); inbound anaphoric islands follow from pro-forms’ lack of content; the no-phrase constraint is affix-specific selection, with complement preservation tracking it ((89)–(95)). The resulting typology “makes no reference to the notion of a word” (p. 29). Parsimony then favours one component, “assuming that they are equivalent in their empirical coverage” (p. 1).
What could have come out otherwise. Phrasal input to word formation could have been confined to listed or quotative cases, and word-part ellipsis could have ignored morpheme boundaries. In English, neither happened.
What this does not say
- That the case is made beyond English. Other languages are cited from secondary literature, often in a sentence. Müller’s reply presses exactly this, and the paper expects “there may be languages where extraction can target an X0” (p. 26).
- That every sub-argument stands. The German resultative nominalization is “not decisive” (p. 12), derived nouns’ structure “requires more research” (p. 31), the analysis of -ness may have “a better analysis” (p. 35), and fn. 9 leaves non-quotative phrasal compounds open. Müller disputes X0 extraction, phrasal coordination and the resultatives.
- That every lexicalist framework is refuted. “Superfluous” is shown case by case for Bresnan and Mchombo, Williams, Newmeyer and Chomsky 1970. The author says the criticisms “may not apply to certain other approaches” (p. 2).
- That one component is simpler overall. The X0/XP replacement needs null heads, segment-based adjunction and extraction blind to X0s. That it is cheaper than a lexical component is argued, not measured.
- That there is no cross-linguistic word. That is Haspelmath’s claim (THEORY-477). Bruening sets it aside, so this account survives if Haspelmath is wrong, and the reverse.
- That lexical items are not form–meaning–syntax triads. Bruening declines that sense of lexicalism. The second assumption Krauska and Lau reject (THEORY-588) gets no support here.
Connections
- Haspelmath (THEORY-477) reaches the same verdict on lexical integrity from typology, and calls it partly circular (his §8). The two use the same diagnostics: word-part coordination, Georgian inbound anaphora and extraction.
- Krauska and Lau (THEORY-588) take this account’s conclusion, one component for words and phrases, as a premise for a claim about production models.
- NOTE-065 notes that ANTH-THEORY-022 (a phrase in a word slot, from ANTH-LIT-410) describes the construction of §2.1. Bruening is background for “productive, not listed” there, not evidence for its claim about frequency.
Note of 2026-10-29: what holds it back
None of the three tests outside English that the promote_when names is in the record: whether an X0 is ever extracted (Müller’s German cases), whether any head coordination is irreducibly X0-level under the substitution test, and whether phrase-selecting affixes preserve complements where X0-selecting ones do not. A narrowing to the English evidence of §§2–3 would free nothing, because THEORY-588 presupposes the one-component conclusion, which would go to a kind-1 entry carrying this promote_when.