I am interested in cross-linguistic variation in case assignment to single-NP elliptical answers – for example “What did they see?”; “A goat”. By “case” I mean a distinct form of a word selected to indicate grammatical role in the sentence, thus excluding use of prepositions and postpositions (so, “with an axe” is not an instrumental case form in English). German and Insular Scandinavian, Slavic (widely if not universally), Finnic, Saami, Sanskrit, Korean, Classical Arabic and (on weak phonological grounds) Quechua have case. Chinese, Lushootseed, Sundanese and Colloquial Arabic do not. I take the existence of allomorphy to be sufficient grounds for calling something “case” as opposed to being a prepositional / postpositional phrase. If there's NP-internal case agreement, even better.
The situation in English is murky because only pronouns are marked for case, though any NP can be marked with -s. Pronouns are nominative (I, he, her, we, they) when they are subjects (“I/*Me went to the store”) though in a subject conjunction, it is optional – “Tom and I/me went to the store”). Otherwise (and when not possessive), the accusative form is used. In eliptical answers, the accusative is used, except when a genitive would be called for in which case you use the “absolute” possessive form or -s if the answer is an NP. So,
Who took the book? Me/*I; Him/*He Who did the dog chase? Me/*I; Her/*She Whose car did they buy? Mine/*Me/*I/*My; Yours/*Your/*You; Hers/*Her/She; Bob’s/*Bob
In North Saami, if you ask a question about a subject the (elliptical) answer is in the nominative; if it’s about an object, it should be in the accusative, and so on. On the other hand, in Southern Nilotic, nominative case is possible only when an overt verb precedes the subject, so elliptical answers use the non-nominative form.
With this background: in what languages is case assigned to elliptical NP answers in the same way as it is to NPs integrated into a full S; in what case languages are elliptical answers given some default case form; and in what languages do you get both patterns, depending on some other factor?
[Addendum] The S. Nilotic claim is based on my notes on Kipsigis. The citation form of "pig" is ingúrwêt; "He killed a pig" is kapár ingúrwêt and "A pig killed him" is kapár ingúrwet -- case is marked by tone change. Either subject or object can be moved left, and an extracted subject NP does not have nominative tone. So using citation / accusative (names) Mʊ́sa, Chʊ̂ma, "Musa saw Juma" is kakɛ́r Mʊsá[s] Chʊ̂ma ~ kakɛ́r Chʊ̂ma Mʊsá[s], and extracted, you have Mʊ́sa kɔ́kakɛ́r Chʊ̂ma (or Chʊ̂ma kɔ́kakɛ́r Mʊsá[s]) which focuses the extracted NP. So, movement messes with case anyhow. What I know is that the guy rejected ingúrwet as a possible form, and I did ask "Even as an answer to 'who killed him'?". But I didn't pursue that rigorously. Creider & Creider 1989 (A grammar of Nandi) may fill the gap.