“Always” and “Only”:Why not all Focus Sensitive Operators are Alike.David Beaver ([email protected]) and Brady Clark([email protected])Stanford UniversityAbstract. We discuss focus sensitivity in English, the phenomenon whereby in-terpretation of some expressions is affected by placement of intonational focus. Weconcentrate in particular on the interpretation of “always” and “only”, both ofwhich are interpreted as universal quantifiers, and both of which are focus sensitive.Using both naturally occurring and constructed data we explore the interaction ofthese operators with negative polarity items, with presupposition, with prosodicallyreduced elements and with syntactic extraction. On the basis of this data we showthat while “only” lexically encodes a dependency on the placement of focus, “always”does not. Rather the focus sensitivity of “always” results from its dependency oncontext, and from the fact that focus also reflects what is given in the context. Weaccount for this split using an analysis couched in event semantics.Keywords: focus,semantics,pragmatics1. Introduction1.1. Focus sensitivity of “always” and “only”An expression is focus sensitive if its interpretation is dependent on theplacement of focus. In English, focus is typically marked by a pitch riseon a stressed syllable. Can a single mechanism explain all cases of focussensitivity? Some, e.g., Rooth (1992), have explicitly stated that theirgoal is to find a single all encompassing theory of focus sensitivity, andof focus more generally. In this paper, we will present data showing thatfocus sensitivity is not a uniform phenomenon. Focus sensitivity of someoperators is derived by a grammatical mechanism. We call these oper-ators focus functional. Focus sensitivity of other operators results froma pragmatic mechanism. At least some non-focus functional operatorsmanifest focus sensitivity entirely as a side effect being anaphoric.There is a wide range of constructions in English that are known toexhibit some form of focus sensitivity. These include questions, coun-terfactual conditionals, discourse connectives, adverbs of quantification,quantificational determiners, frequency adverbs, generics, emotive fac-tives and attitude verbs, negation, superlatives, modals, and a heteroge-c 2002 The Authors. Draft - comments welcome. .distribute.tex; 30/01/2002; 18:45; p.12neous set of other operators (“only”, “even”, “also”, “again”).1To ourknowledge, no case has ever been made for dividing these focus sensitiveexpressions into two separate classes. We will provide varied evidencefor such a bifurcation, although leaving for further work a thoroughsurvey of which expressions belong in which class. We concentrate onan in depth analysis of two constructions rather than a broad surveyof the entire class.We will examine the interpretation of “always” and “only”. It shouldbe noted that “always” and “only” appear to have similar meanings.They are both standardly analyzed as universals, and they are bothstandardly taken to be focus sensitive. What we will show is thatsentences involving “always” and “only” gain their interpretation in(dramatically) different ways.The focus sensitivity of “always” and “only”, and their universalcharacter, is illustrated by the following example paradigms. In eachgroup, the (a) and (b) examples have similar interpretations available.These interpretations are approximated by (c) and (d). In (c), “always”and “only” quantify over individuals.2In (d), “always” and “only”quantify over events.3In Section 5, we return to an events analysisalong the lines of the formulae in (1d) and (2d).(1) Focus sensitivity of “always” and “only”: direct objectfocus.a. Sandy always feeds [Fido]FNutrapup.b. Sandy only feeds [Fido]FNutrapup.c. ∀x feed(sandy, x, nutrapup) → x = fido“Everything Sandy feeds Nutrapup to is Fido”d. ∀e (feeding(e) ∧ agent(e) = sandy ∧theme(e) = nutrapup) → goal(e) = fido1See Rooth (1996a) and Hajiˆcov´a et al. (1998) for examples of most of theseconstructions.2The analyses in (c) are similar to a structured meaning semantics interpretationof the (a) and (b) examples; see Krifka (1992) and von Stechow (1989). Note thatsentences with “only” like (1b) are often argued to include as part of their basicmeaning the proposition that the sentence minus “only” is true, i.e. Sandy feeds FidoNutrapup. Various authors have claimed that this proposition is not an entailment.Horn (1996) has claimed that this an inference arises from a combination of theuniversal meaning as given in (1c) and the standard universal presupposition thatthe domain of quantification is non-empty, here that Sandy fed something Nutrapup.In this case, the inference from (1b) to Sandy having fed Fido Nutrapup has a similarstatus to the implication from “Everyone likes Sandy” to “Someone likes Sandy”. Wewill assume a position like Horn’s on this issue without providing further evidence.3Event semantic analyses of focus sensitivity have been proposed in Bonomi andCasalegno (1993) and Herburger (2000).distribute.tex; 30/01/2002; 18:45; p.23(2) Focus sensitivity of “always” and “only”: indirect ob-ject focus.a. Sandy always feeds Fido [Nutrapup]F.b. Sandy only feeds Fido [Nutrapup]F.c. ∀x feed(sandy, fido, x) → x = nutrapup“Everything Sandy feeds to Fido is Nutrapup”d. ∀e (feeding(e) ∧ agent(e) = sandy ∧goal(e) = fido) → theme(e) = nutrapupExamples like those above show why semanticists have tended toassume that “always” and “only” can be analyzed similarly. Typically,researchers refer to a single mechanism, so called association with focus,to explain the meaning difference between (1a) and (2a), and to explainthe difference between (1b) and (2b).1.2. Theories of focus interpretationWhile all theories agree that focus sensitive operators like “always”and “only” interact with focus in linguistic contexts like that exem-plified by (1) and (2), there is controversy about how grammaticizedthe relationship between “always”, “only” and their associated focusis (Partee 1999:215ff): do the lexical entries of “always” and “only”stipulate association with a focused constituent in their syntactic scopeor is the interaction illustrated by (1) and (2) not an absolute lexicalrequirement of the constructions, and perhaps optional in certain con-texts?