Modal expressions as a general rule are highly context sensitive, relying for their interpretation even more than most expressions on the conversational background. (See Kratzer 1981, 1991 for a formal treatment of conversational background and modal words.) Sometimes, the conversational background against which a modal word is interpreted is quite general; for example, the background invoked by the adjunct in 1 is the set of laws of the state of Maine, and the matrix sentence is held to be true in situations in which those laws apply.
1. According to Maine law, dogs may run free.
In other cases, the conversational background is much more situation-specific; in 2, for example, the background brought to mind by the adjunct is information about what happened on a particular day in the parking lot.
2. In view of what happened last week in the parking lot, residents may take steps to hire a security guard.
2 is ambiguous: on its epistemic reading, it means that, according to what the speaker knows (including that something particular and relevant happened in the parking lot the previous week), it is possible that the residents will take steps to hire a security guard. On its deontic reading, 2 means that the residents have the right to do this (because of whatever it was that happened in the parking lot.) The central claim I make in this paper is this: epistemic interpretations arise against situation-specific conversational backgrounds whereas root (deontic, ability and dispositional) interpretations often arise against conversational backgrounds that describe a range of situations (as in the most obvious reading of 1). Several observations support this claim. (i) Wide-scope adverbs of quantification, durative temporal adverbs and indefinite locative adverbs all show a bias in favor of root modals. (Iatridou 1990 pointed out the adverbs of quantification facts.) See 3-5.
3. Usually, students may use the copy machine on the first floor. 4. While employed by Unisys, Bob must wear a suit. 5. At high altitudes, Brenda can run a six minute mile.
(ii) The available temporal interpretations for the sentence under the scope of an epistemic modal coincide with the temporal interpretations generally available for the present tense. Specifically, they get either 'now' interpretations for stative sentences and habitual interpretations for non-stative sentences, whereas root modal sentences tend to be temporally underdetermined. See 6-7.
6. Bill must be on time. epistemic reading: It must be the case that Bill is on time (now). deontic reading: It is required that Bill be on time (on whatever occasions the conversational background makes relevant.) 7. Monks may spend time alone. epistemic reading: It is possible that monks (habitually) spend time alone. deontic reading: It is allowed that monks spend time alone/Monks are allowed to spend time alone.
(iii) Type/token NPs in subject position generally get token (or, referential) interpretations in epistemic modal sentences, but are generally ambiguous in root modal sentences. See 8.
8. The District Attorney may be a Canadian.
epistemic reading: It is possible that the person who is
the DA in this context is in fact a Canadian.
deontic reading (a): It is permissible that the DA in this
area (whoever it is) be a Canadian.
deontic reading (b): The person who is the DA in this
context has the right to be (or, become) a Canadian.
I relate these three sets of observations to the central claim that epistemic conversational backgrounds are indexical to a particular situation. Formally, the difference between epistemic and root interpretations is that in epistemic cases, the modal quantifies over worlds, but in the root cases it quantifies over world-situation pairs. I compare this analysis to Iatridou's 1990 analysis of the (related) contrast between epistemic (or, subjective) modality and metaphysical (or, objective) modality, which rests on the claim that epistemic but not metaphysical modals come supplied with a time variable, and show why the present account is preferable.
References
Iatridou, S. 1990 "The Past, the Possible and the Evident", in _LI_ 21.123-129.
Kratzer, A. 1981 "The Notional Category of Modality" in _Words, Worlds and Contexts_, H.J. Eikmeyer and H. Reiser, eds. (Berlin: Walter de Gruyter).
___________ 1991 "Modality" in A. von Stechow and D. Wunderlich, eds. _Semantics: An International Handbook of Contemporary Research_ (Berlin/NY: Walter de Gruyter).