Grammaticalization — the historical process by which lexical items develop into grammatical ones, as English 'be going to' developed out of the motion verb 'go' — exhibits a robust unidirectional tendency in semantics: concrete, specific meanings evolve into more schematic, abstract ones. Jäger & Rosenbach (2008) proposed that this unidirectionality is driven by asymmetric priming, a cognitive association in which a source meaning strongly and facilitatively primes its grammaticalized counterpart but not the reverse. The first experimental test (Hilpert & Correia Saavedra 2018, comprehension only) found the opposite pattern — an inhibitory effect — and left open whether the predicted facilitative asymmetry obtains in language production and in naturalistic usage, the regime where priming-driven change must operate. The open problem is to determine, empirically, whether facilitative lexical-to-grammatical asymmetric priming exists in production paradigms and naturalistic production data and can drive unidirectional semantic change, or whether converging evidence eliminates the hypothesis in favor of alternative mechanisms.

Background

Grammaticalization is the diachronic process whereby lexical items and constructions acquire grammatical functions, as when the English motion construction 'be going to' developed into a future marker. A central empirical generalization of grammaticalization theory is the unidirectionality of semantic change: meanings tend to evolve from concrete and specific toward more schematic and abstract (a development often described as semantic bleaching), while the reverse development, grammatical item becoming lexical, is rare. What has been missing is a cognitive mechanism explaining why the change runs predominantly one way.

Priming is the psycholinguistic phenomenon whereby recent exposure to a form or concept changes the processing of a related form or concept, typically speeding it (facilitative priming). Asymmetric priming denotes a directional pattern of such associations: one idea strongly evokes another (the word 'paddle' strongly evokes 'water') while the reverse evocation is much weaker ('water' only weakly evokes 'paddle'). Jäger & Rosenbach (2008) proposed that asymmetric priming is the 'missing link' explaining unidirectional language change: in a grammaticalizing pair, the lexical source meaning A facilitatively primes the grammaticalizing meaning B far more than B primes A, and repeated asymmetric priming in usage accumulates into diachronic unidirectionality. Crucially, their proposal rests on the uniformitarian assumption that cognitive mechanisms measurable in present-day speakers also operated in the (mostly prehistoric) communities where attested grammaticalization took place. They further argued that priming effects are most consequential in language production, since production choices determine what is transmitted across usage events; the psycholinguistic literature on structural priming (the tendency to repeat previously used syntactic structures) indeed shows priming to be strongest and most persistent in production.

Hilpert & Correia Saavedra (2018) provided the first experimental test using a maze task, a self-paced reading variant in which participants build a sentence word by word, choosing between two alternatives at each point. Sentences contained both a lexical and a grammatical use of the same word (e.g., 'kept' with its lexical sense 'retain' and its grammatical aspectual use). Contrary to the asymmetric priming hypothesis, they observed a negative (inhibitory) priming effect: the grammaticalized variant was processed more slowly after its lexical counterpart, a pattern they attributed to horror aequi, the well-documented tendency of language users to avoid immediate repetition of identical or near-identical forms and constructions. They tested comprehension only, and flagged production-to-production priming, corpus evidence from naturalistic production, and the limits of the uniformitarian assumption as open questions. A subsequent corpus study by the same authors (2020), operationalizing the hypothesis's predictions in token-based semantic vector spaces (distributional vector representations of individual word tokens in corpus data) for the word 'use', found no support. Kapron-King, Kirby, Trousdale & Smith (2025) found no asymmetry in individual preferences for extending artificial words between lexical (body-part) and grammatical (prepositional) meanings, and showed computationally that a larger lexical than grammatical inventory could alone produce unidirectionality. Johansson, Nykiel & Salte (2026) independently confirmed the inhibitory pattern for the grammaticalized intensifier 'way', attributing it to the processing cost of switching between declarative memory-based and procedural (rule-based) processing. Meanwhile, Baumann & Sommerer (2018) showed in a mathematical population-dynamics model that asymmetric priming could in principle generate reductionist, unidirectional change — but also diversification and stable coexistence of variants — so the mechanism remains theoretically viable without empirical demonstration. No study to date has tested the predicted facilitative lexical-to-grammatical priming in language production, the regime that both the hypothesis's proponents and its first experimental critics identified as decisive.

Problem Statement

Does facilitative asymmetric priming between lexical items and their grammaticalized counterparts (e.g., 'go' priming 'be going to', but not the reverse) actually obtain in human language use — in particular in language production, where priming is strongest and where usage-driven change must operate — and can it be established as the cognitive mechanism that drives the unidirectionality of semantic change in grammaticalization?

The question concerns whether the directional prediction of Jäger & Rosenbach's (2008) asymmetric priming hypothesis — a facilitative lexical-to-grammatical priming asymmetry — holds on empirical evidence in production, the regime its critics identified as decisive, and whether that asymmetry can account for the unidirectionality of semantic change. The source flags that attested grammaticalization occurred under social conditions no longer observable and that past cognition is inaccessible, which bounds what any present-day study can establish about the original change.

References

  1. Hilpert, M., & Correia Saavedra, D. (2018). The unidirectionality of semantic changes in grammaticalization: An experimental approach to the asymmetric priming hypothesis. English Language and Linguistics, 22(3), 357–380. DOI: 10.1017/s1360674316000496. https://doi.org/10.1017/s1360674316000496
  2. Jäger, G., & Rosenbach, A. (2008). Priming and unidirectional language change. Theoretical Linguistics, 34(2), 85–113. DOI: 10.1515/thli.2008.008. https://doi.org/10.1515/thli.2008.008
  3. Hilpert, M., & Correia Saavedra, D. (2020). Using token-based semantic vector spaces for corpus-linguistic analyses: From practical applications to tests of theoretical claims. Corpus Linguistics and Linguistic Theory, 16(2), 393–424. DOI: 10.1515/cllt-2017-0009. https://doi.org/10.1515/cllt-2017-0009 (published online 26 September 2017)
  4. Kapron-King, A., Kirby, S., Trousdale, G., & Smith, K. (2025). Grammatical unidirectionality is not reflected in individual preferences when performing artificial semantic extension. Language and Cognition, 17, e60. DOI: 10.1017/langcog.2025.10018. https://doi.org/10.1017/langcog.2025.10018
  5. Johansson, C., Nykiel, J., & Salte, A. (2026). Way better that way. Linguistics, 64(4), 869–900. DOI: 10.1515/ling-2024-0194. https://doi.org/10.1515/ling-2024-0194
  6. Baumann, A., & Sommerer, L. (2018). Linguistic diversification as a long-term effect of asymmetric priming. Language Dynamics and Change, 8(2), 253–296. DOI: 10.1163/22105832-00802002. https://doi.org/10.1163/22105832-00802002
  7. Rosemeyer, M., & Grossman, E. (2021). Why don't grammaticalization pathways always recur? Corpus Linguistics and Linguistic Theory, 17(3), 653–681. DOI: 10.1515/cllt-2020-0053. https://doi.org/10.1515/cllt-2020-0053