Making linguistic theory is like specifying a programming language...

Why can some NPs take scope where others cannot? In particular, why is it that indefinite NPs can often take wider scope than other NPs can in the same context (Fodor and Sag 1982)? For example, a professor in (1) can take scope over every paper, as shown in (1b). If a is changed to every as in (2), then that reading becomes unavailable, as shown in (2b). (1) I read every paper that a professor in my department recommended.... a. ∀x. � paper(x) ∧ ∃y. professor(y) ∧ recommend(y, x) � ⇒ read(me, x)... Yet I skipped this paper because no professor recommended it. b. ∃y. professor(y) ∧ ∀x. � paper(x) ∧ recommend(y, x) � ⇒ read(me, x)... I understood her research interests very well when I finally finished. (2) I read every paper that every professor in my department recommended.... a. ∀x. � paper(x) ∧ ∀y. professor(y) ⇒ recommend(y, x) � ⇒ read(me, x)... It only took a week because the professors ’ interests are so diverse. b. *∀y. professor(y) ⇒ ∀x. � paper(x) ∧ recommend(y, x) � ⇒ read(me, x)... Yet I skipped this paper because no professor recommended it. It is widely noted that quantifier scope in English is bounded by relative clauses, if-clauses, and perhaps tensed clauses in general. It is also well known that indefinite determiners such as a, some, and a certain are exempt from this restriction. We can try to explain this discrepancy in two ways. On one hand, maybe indefinite NPs take scope through a mechanism separate from (and more liberal than) that for “genuine quantifiers”. The analysis