Carlson (1983, 2006) argues that functional elements (to be dubbed f-elements) often present mismatches in form and interpretation of the kind lexical elements do not. Carlson points out that there is a learning problem if the learner is supposed to figure out functional meanings from what he/she hears. His solution to the problem is that f-elements themselves are meaningless; functional meanings are carried by features or phonetically null operators that appear on the phrases over which they scope, and their effects percolate down to heads in order to receive expression, in one way or another.
Results in the past decades converge impressively with Carlson’s vision. Consider, for example, the assumption of null operators in Ladusaw 1992 and Zeijlstra 2004 for negative concord, in Beghelli & Stowell 1997 for distributivity, in Kratzer 2005 for existential and other quantifiers, in Kusumoto 2005 for past tense, in Katzir 2011 for poly(-in)definiteness, and so on. There are also variations on that theme. The exhaustifiers EXH in Fox 2007 and O in Chierchia 2013 have syntactic status but no matching f-elements. Still further; the binding combinator “z” in Jacobson 1992 and the abstract negation in Chierchia 2013:235 do not seem to have syntactic status. The disembodied meet and join operations in Szabolcsi 2015 do not have syntactic status; they kick in either by default or to satisfy the presuppositions of overt MO or KA particles. Thus, for the logical scaffolding to be unpronounced or to be even disembodied seems to be the norm. If that is so, careful considerations of compositionality / learnability are called for. (SLIDES)