traditional argument for [[universal grammar]] follows from [[poverty of the stimulus]]:
![[poverty of the stimulus#^argument]]
people have issue with most of these points.
summarized in [[2003TomaselloBiologicalCulturalOntogenetic]] and summarized again here:
| | [[generative grammar]] | [[usage based linguistic]]s |
| ------------------------------------------------------------ | --------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------- |
| the nature of [[linguistic competence]] is | primarily [[form]]al and [[abstract]] | primarily [[discourse\|communicative]] / [[symbol]]ic |
| speakers use [[grammar rule]]s that | are purely [[form]]al [[abstract]] [[human language grammar universal]]s | vary in [[abstract]]ion and depend on [[pragmatic]] factors eg [[intention]] |
| child [[language acquisition\|acquires language]] by | collecting [[dataset]] and running [[hypothesis test]]s against [[grammar rule]]s | [[theory of mind\|intention reading]], [[imitation learning]], and other [[human cognitive skill]]s |
| [[primary linguistic data]] takes the form of | a [[dataset]] of [[utterance]]s | [[intention]]al [[discourse\|communication]] |
| the [[poverty of the stimulus]] argument | provides evidence for [[universal grammar]] | misinterprets data and has many plausible explanations besides [[universal grammar]] |
| [[human language]] specific [[grammar rule]]s are learned by | [[falsify]]ing [[principles and parameters\|linguistic parameter]]s ("linking") | [[pragmatic]] processes, [[inductive bias]]es, [[metaphor\|analogy]], and [[statistical learning]] |
| [[linguistic innate]] knowledge takes the form of | [[universal grammar]] | [[human cognitive skill]]s |
| [[represent\|latent representation]]s of [[grammar rule]]s | are [[time invariant]] and [[abstract]] ([[continuity assumption]]) | transform from item-based [[archetype]]s into general [[abstract]]ions |
| [[idiom]]s and "special constructions" | must be [[human memory\|memorized]] apart from core grammar | are a core part of [[language acquisition]] |
^criticism
plus what does "[[linguistic competence|mental grammar]]" even *mean*?
is it a [[belief state]]? learned [[represent]]ation?
somehow [[encoding]] into [[synapse|synaptic]] weights?
also "linking problem" still unresolved:
eg to set "Head" parameter ([[head]] final vs initial):
child needs to identify [[lexico|phrase]] heads.
how is that even possible?
not like speakers are snapping their fingers after saying each phrase head.
and parameter-setting would be expected to be more discrete;
typical generativist response is [[linguistic performance]] factors
(ad hoc and un[[falsify]]able for now)
[[Occams razor]] seems to be in favour of [[usage based linguistic]]s:
[[universal grammar]] seems to need so much adjustment to accommodate [[human language]]s beyond [[Europe]] (ie [[Eurocentric]])
![[2025-07-16 Kempner lunch and learn linguistics talk#sources]]