Poster: Exceptional Imperative Constructions are Semantically Regular

October 25–26
Workshop on Semantic Variation, University of Chicago


In the literature on imperative clauses, there are several “exceptional” constructions that are attributed to a single language at a time. In this work, I focus on two of these constructions: past tense imperatives and marking negative commands with a dummy imperative verb. I show that each construction is regular and semantically transparent when analyzed in a preference-based approach to imperative semantics (following Starr 2012). In this framework, imperatives carry both propositional information (in the verbal domain) and preferential information (in the left periphery). The free combination of elements in the two domains permits these constructions, and their rarity is attributable to syntactic limitations and, in the case of past tense imperatives, pragmatic constraints on their use.

Past Tense Imperatives: Most accounts of imperative semantics require that they describe future actions (e.g. Hamblin 1987), or at least non-past actions (Kaufmann 2011). However, some languages combine past tense morphology with imperative morphology (1–2) and/or word order (2–3). These constructions all share a counterfactual advice interpretation (rendered in English as You should have…).

(1)    Syrian Arabic (Cowell 1964; Palmer 1986)
    kənt          kōl             lamma kənt          fəl-bēt
    you.were eat+IMP when    you.were in.the-house
    “You should have eaten when you were at home.”   

(2)    Estonian (Aikhenvald 2010)
    tulnud             õhtul                 õigel            ajal               koju
    come+PAST=OPT evening+LOC.SG right+LOC.SG time+LOC.SG home
    “You should have come at a proper time in the evening.”   

(3)    Dutch (Mastop 2005)
    Had je       telefoonnummer dan   ook niet aan die   vent gegeven!
    Had your phone-number     then also not  to    that guy  give-PP
    “You shouldn’t have given your phone number to that guy.”   

Past Tense Preferences: I argue that each of the constructions in (1–3) is a straightforward combination of a past tense proposition with the characteristic meaning of imperatives: imposing a preference. Following Starr (2012), this preference is modeled in the preference state of the discourse; analogical to the common ground, it contains the information mutually assumed by the interlocutors.

Take (1), which introduces a preference of the form ⟨p,¬p⟩, where p = the addressee ate when he was at home. This sentence is most felicitous when it is not commonly held that p, just as a non-past imperative is most felicitous when its realization is not certain (Kaufmann 2011). The preference for ¬p is semantically compatible with the counterfactual information, but must be interpreted as advice. This holds even in languages like English, where marking of past tense in an imperative is not allowed (4), but an imperative can combine with a past-tense adverb with similar pragmatic effect (5).

(4)    *Have read the book!

(5)    Read the book yesterday!
    ≈You should have read the book already.

Dummy Imperative Verbs: It has been widely discussed that many languages must mark negative imperatives with an infinitive or subjunctive form (e.g. Zanuttini 1997, Han 2000), but an alternate strategy is to insert a dummy imperative verb (6–8). These elements are taken from verbs that originally meant ‘can’ (6), ‘want’ (7), and ‘stop’ (8), but serve a purely functional role in prohibitions. For example, (6) does not mean “Don’t be able to read newspapers.”

(6)    Serbo-Croatian (Szucsich 2010)
    Nemoj-te            čita-ti       novin-e!
    NEGIMP.2.PL read-INF papers-F.PL.ACC
    “Don’t read newspapers!”

(7)    Latin (Croft 1991)
    nōlī                        amābō verberāre lapidem
    NEG.want.IMP I.pray   beat.INF stone.ACC
    “Don’t beat a stone.”

(8)    Welsh (Willis in press)
    Paid         â       gadael!
    NEG.2S with leave.INF
    “Don’t leave!”

Dummy Verbs are Non-Propositional: In all three cases, the dummy verb surfaces in the left periphery of the clause and either attaches to negation (6–7) or supplants it (8). I argue that this is a reflex of clause typing, which a) contributes the preferential meaning of the imperative and b) enforces agreement with a verbal element. Semantically, since the dummy verb carries no meaning, it cannot be negated. However, the dummy verb also fulfills the syntactic restrictions of imperative clause typing, by absorbing agreement; consequently the content verb is spelled out as an infinitive. The only remaining interpretation is that negation is interpreted within the propositional domain and the dummy verb is interpreted preferentially in the functional domain. The separation of these two domains allows prohibitions, regardless of their surface syntax, to be interpreted in the same compositional manner as all imperative constructions.