Combined marking of reflexivity: Typology and grammaticalization

This paper investigates a phenomenon which has been identified in several unrelated languages, but of which no comprehensive survey has been presented so far: the combination of two or more separate grammatical markers to express reflexivity. Using a convenience sample of 50 genetically and areally diverse languages, we will present a typological overview of the construction-types under examination and discuss their grammaticalization as full-fledged reflexive markers.
Combined-marking reflexives may involve three types of grammatical elements:

(i) a personal, logophoric or reflexive pronoun;
(ii) a middle(-like) marker, usually with valency-reducing functions;
(iii) an intensifier of the type of English reflexive pronouns, in appositional or adverbial use (e.g. I myself said that/I said that myself).

The three possible two-marker combinations are illustrated below:

(1) Pronoun + middle(-like) marker: Paicî (Austronesian)
é pi-tâmâgööri ê wë Maria
3.SG PREF-know 3.SG ART.PERS Maria
‘Maria knows herself.’
(Moyse-Faurie 2008: 147)

(2) Pronoun + intensifier: Tsakhur (Nakh-Daghestanian)
Rasul-e: wuʒͮ-e: wuʒͮ get-u.
Rasul-ERG INTS-ERG LOG.ABS beat-PFV
‘Rasul beat himself.’
(Lyutikova 2000: 229)

(3) Middle(-like) marker + intensifier: Mezquital Otomí (Otomanguean)
Ra Šuwa bi-n-hyó-sɛ
DET Juan 3.PAST-REFL/MID-kill-INTS.AO
‘Šuwa killed himself.’
(Gast & Siemund 2006: 368)

Without the additional marker (the middle prefix in (1), the intensifier in (2)-(3)), all three sentences would be ambiguous: (1)-(2) could be interpreted as transitive clauses with non-coreferential arguments; (3) as an intransitive clause with passive or impersonal meaning.
We hypothesize that combined-marking reflexives arise in a process of grammaticalization consisting of four stages:

1. Each marker performs its original function: the semantics of the construction is always compositional and never merely reflexive (e.g. Italian Dio si è creato da sé, ‘God created himself by himself’, where the intensifier da sé emphasizes the agentivity of the subject-referent: the clause cannot simply be rendered as ‘God created himself’).
2. The combination of markers starts conventionalizing as a reflexive construction, but a compositional interpretation is still possible in context. In Khaling (Sino-Tibetan), for instance, reflexivity can be expressed by combining the multifunctional middle marker -si with the intensifier -tāːp; this, however, does not necessarily entail a reflexive interpretation (Jacques et al. 2016: 40). Thus, in context, -si is still interpretable as an antipassive, autobenefactive or impersonal marker, with -tāːp retaining its intensifying meaning.
3. A compositional interpretation is no longer possible: the combination has grammaticalized as a full-fledged reflexive construction, but is not strictly obligatory (e.g. in (3) a non-reflexive interpretation is excluded, but a reflexive one is still possible, in context, even without the intensifier sɛ̌hɛ́/-sɛ).
4. The combination has become obligatory for the expression of reflexivity, e.g. Mam (Mayan), where the referential reflexive iib’ cannot occur without the antipassive suffix -n (England 1983: 187).

Further sub-distinctions exist within all three classes of elements: different types of pronouns, different functions of middle markers and different uses of intensifiers (adnominal vs. adverbial, agent- vs. nonagent-oriented). Since all these subtypes can occur alongside markers of the other two classes, the paper will also investigate whether all possible co-occurrence patterns are equally likely to grammaticalize as combined-marking reflexives.