At a glance
Five papers from 1992 to 2023 reshaped how language scientists describe reflexive constructions, from Reinhart and Reuland's Reflexivity to a 28-language open-access survey.
Five research papers, published between 1992 and 2023, carry most of the load in how language scientists describe reflexive constructions — the part of grammar that lets a clause point back at itself. The order below is an editorial judgment on two tests: how much later work took the paper's questions as given, and how much of the world's languages it could account for.
The list starts with a 1993 reformulation of binding conditions in Linguistic Inquiry and closes with an 867-page open-access volume describing reflexives in 28 languages on six continents. Between them sits a 63-page paper arguing that the whole business can be explained by counting how often pronouns refer back.
What the five share is a change of object. Reflexives stopped being a projection of English and became a countable empirical phenomenon: the likelihood that a pronoun in a given slot refers back to the subject shapes which forms a language bothers to build.
- Reinhart and Reuland, "Reflexivity," Linguistic Inquiry 24(4), 1993
- Cole, Hermon and Yanti, "Grammar of Binding in the Languages of the World: Innate or Learned?" Cognition 141, 2015
- Janic, Puddu and Haspelmath (eds.), Reflexive Constructions in the World's Languages, Language Science Press, 2023
- Haspelmath, "A Frequentist Explanation of Some Universals of Reflexive Marking," Linguistic Discovery 6(1), 2008
- Pollard and Sag, "Anaphors in English and the Scope of Binding Theory," Linguistic Inquiry 23(2), 1992
What counts as the phenomenon, and what counts as evidence
A reflexive construction is a clause in which two participants turn out to be the same entity: Mary saw herself. The dedicated form carrying that meaning is a reflexivizer, a term Katarzyna Janic and Martin Haspelmath fixed for any specialized form that expresses coreference within a clause, whether the form is a pronoun or a marker fused into the verb.
Not every language has one. In the 28-language sample behind the 2023 volume, only two lacked a dedicated reflexivizer: Kazym Khanty, a Uralic variety spoken in northwestern Siberia, and Zenzontepec Chatino. Both still separate coreference from disjoint reference, leaning on context and information structure instead of a grammaticalized marker.
Two more terms recur and cause most of the confusion. A logophor is an anaphoric form whose antecedent is fixed by whose point of view a sentence reports rather than by syntax. Verbs also split into introverted and extroverted classes — 'wash' against 'kick' — and languages often mark coreference differently depending on which class is involved.
Elicitation has limits that show up in the published data. Chini, a Lower Sepik-Ramu language of Papua New Guinea, has no natural way to say 'give something to oneself'; speakers found the frame absurd. In Chini, too, verbs such as 'hate' or 'kill' are not used reflexively in ordinary practice. A missing form in a grammar is not always a grammatical gap.
1. Reinhart and Reuland (1993): locality moves from noun phrases to predicates
Tanya Reinhart and Eric Reuland's Reflexivity, in Linguistic Inquiry 24(4), is the paper most often cited as the alternative to Chomsky's binding theory. It stopped stating conditions on anaphors and pronouns and stated two conditions on predicates instead: a reflexive-marked syntactic predicate is reflexive, and a reflexive semantic predicate is reflexive-marked.
In practice a SELF anaphor does the marking, and a personal pronoun cannot do that job. John likes him fails because both arguments belong to the same predicate and only the reflexive can link them. John said Ann likes him passes because the pronoun and its antecedent sit in different predicates.
The reformulation left a residue the authors did not hide. A reflexive with no coargument to bind it, inside a picture noun phrase or certain prepositional phrases, falls outside the conditions and behaves as a logophor, licensed by point of view. That residue became a research programme in its own right: exempt anaphora.
2. Cole, Hermon and Yanti (2015): an argument rather than a description
Most papers on reflexives describe a language. This one used two Malay varieties to argue about innateness, and the argument has run ever since.
Peter Cole, Gabriella Hermon and Yanti compared Traditional Jambi Malay, where pronouns and reflexives are not distinct classes, with Jambi City Malay, where the distinction behaves close to a hard rule. They called the pair a live snapshot of grammaticalization: what is a preference in one variety is a requirement in the other. The paper is indexed on PubMed.
Their conclusion, in Cognition 141, was that binding systems are learned and that universal grammar does not solve the poverty of the stimulus on its own. Data from signed and creole languages were offered as corroboration.
Eric Reuland answered in Cognition in 2017, arguing that the same facts follow from a small set of universal principles plus the vocabulary each language happens to have. Part of his case rested on whether the Jambi pronoun dio? is specified for number. Cole, Hermon and Yanti re-ran acceptability judgments with eight native speakers, including ethnic Chinese residents of Jambi City, and rejected that analysis. The exchange is the clearest record of where the disagreement actually sits.
3. Janic, Puddu and Haspelmath (2023): the cross-linguistic baseline
Reflexive Constructions in the World's Languages is the first edited volume devoted to the topic, and the whole book is free to download from Language Science Press. Twenty-eight chapters, each based on original fieldwork, use a single questionnaire and a shared vocabulary, which is what makes the results comparable at all.
The closing comparative chapter delivers the numbers. Most languages have one or two reflexivizers; three is rare, recorded for Jóola Fóoñi, Kuuk Thaayorre and Kakataibo. Nominal reflexivizers outnumber verbal argument markers and voice markers. Body-part terms are the main historical source for reflexive nouns, though the authors record an unexpected one in Walman, where the source expression means 'base of a tree, cause, reason'.
Two results cut against expectation. No reflexive voice marker in the sample encoded coreference and nothing else; every one also covered middle functions. And a long-standing generalization held up: if a language has a reflexivizer, it uses it in the third person singular, from where it may spread to other persons.
The volume is also candid about its own gaps. Its editors note that reflexive constructions in Georgian are unusual in permitting a reflexive pronoun in subject function, and that the questionnaire has not yet covered oblique and long-distance constructions in comparable detail.
4. Haspelmath (2008): why the forms look the way they do
Before the 2023 volume there was this paper, in Linguistic Discovery 6(1), which set out universals the volume later tested. Its complaint was methodological: the binding literature had produced language-specific generalizations rather than cross-linguistic claims anyone could check. The full text sits in an open archive.
The explanation on offer was frequency. Where coreference with the subject is the likely reading, a language can get away with lighter marking; where it's unlikely, speakers need a dedicated and heavier form. Haspelmath argues that this asymmetry, position by position, motivates the observed inventory of reflexive forms.
The supporting cases are chosen to break English habits. Loniu, an Oceanic language of Papua New Guinea, has no reflexive pronouns and leaves coreference with an object pronoun vague. Hausa permits partial coreference, where a reflexive covers the subject and others, which English forbids. Haspelmath notes that only Mira Ariel's work had taken a similar usage-based line, developed independently of his.
Photo by Mick Haupt on Unsplash
5. Pollard and Sag (1992): the same answer from a different framework
Carl Pollard and Ivan Sag reached the coargument idea in Linguistic Inquiry 23(2) at the same time as Reinhart and Reuland, working from head-driven phrase structure grammar rather than generative syntax. Their domain for proper reflexives and reciprocals was the coargument domain, and they were explicit that not all English reflexives answer to the same conditions.
That convergence is why the paper still gets cited. When two incompatible formalisms land on the same empirical generalization within a year of each other, the generalization is unlikely to be an artefact of either one. Pollard and Sag's later book-length treatment consolidated the exempt category that Reinhart and Reuland had also been forced to acknowledge.
It ranks fifth here on one measure only: the 1993 formulation travelled further into subsequent syntax, and citation habits followed it.
What is still being argued
The live question is where the pronoun/reflexive split comes from. Haspelmath, and Cole with Hermon and Yanti, put it in use: forms grammaticalize out of frequency and function. Reuland puts it in a narrow set of principles plus each language's vocabulary, a position he restated in a 2017 reply to the Jambi work.
Stephen Levinson had made a related case earlier, arguing in Journal of Linguistics in 1987 that Gricean pragmatics could carry much of the explanatory load, with Guugu Yimidhirr and English as his examples. His 1991 follow-up added a diachronic account that brought long-distance reflexives and logophoricity into the same frame.
Book-length attempts to settle it exist. Eric Reuland's Anaphora and Language Design and Johan Rooryck and Guido vanden Wyngaerd's Dissolving Binding Theory both appeared in 2011, and they pull opposite ways: one rebuilding binding on Agree and chains, the other arguing the theory should be dissolved into other parts of grammar.
Data will decide more of this than argument will. The 2023 questionnaire is the first instrument that makes descriptions from different fieldwork traditions comparable, and whether the next decade of papers uses it — or returns to bespoke examples in single languages — is the test worth watching.
- linguistics-research
- linguistics-research
- reflexive-constructions
- reflexive-constructions
- binding-theory

