The impossibility of voice mismatches in sluicing has proven to be a fruitful empirical domain for the formulation of the identity condition on ellipsis. English data like (1) show that a verb in the active voice cannot serve as the antecedent for a verb in the passive voice inside a sluice. This has led researchers to propose that some syntactic identity is required for ellipsis (Merchant 2013).

(1) Active—Passive mismatch

   *Someone hugged Robin, but we don’t know who by.

However, certain languages do allow for non-identical voice specification between the verb in the antecedent and the ellipsis site. Chamorro (Austronesian) permits antipassive-active voice mismatches (Chung 2013). The possibility of this type of mismatch leads Chung to propose a hybrid identity condition which employs both semantic and (limited) syntactic identity, including (2):

(2) Chung (2013)’s Case condition on ellipsis

  “If the interrogative phrase is a DP, it must be Case-licensed 
   in the ellipsis site by a head identical to the corresponding 
   head in the antecedent clause.”

In this talk, I present novel data regarding a specific type of mismatch in Kaqchikel (Mayan) sluicing. In Kaqchikel, an active verb can serve as the antecedent for a verb in the Agent Focus voice inside a sluice and vice-versa. Agent Focus is used when A’-extracting the subject of a transitive clause:

(3) Active—Agent Focus mismatch

   Jun winaq  x-Ø-tzaki-j     la  che… 
   one person CPL-B3S-cut-ACT DET tree

   Man w-etama-n    ta  achike winaq   
   NEG A1S-know-PRF NEG which  person

   <x-Ø-tzaki-n      la  che >. 
   CPL-B3S-cut-A.FOC DET tree
   ‘A person cut that tree... I don’t know which person.’

I argue that the particular agreement configurations in active versus Agent Focus clauses in Kaqchikel provide evidence against Chung (2013)’s Case identity condition. I show that in the Kaqchikel example in (3), the wh-remnant in the second clause does not enter into an agreement relation with the same head as the corresponding indefinite in the antecedent, violating the condition in (2). I go on to sketch possible analytical pathways to account for this novel data without recourse to an identity condition akin to Chung’s.