Page 402 - Encyclopedia of Philosophy of Language

Formal Semantics
A 'statement' consists of the utterance of an indica- tivesentence0inanutterancesituation:u:$.Whereas sentences have meanings, a statement u:4>is assigned an 'interpretation' [u:0] = {.y|w[<£]s}. The interpret- ation of the statement u: I am stroking Jackie is the collection of situations s such that «STRO- KE,/,fl,£;l»es, where l,a,b are as above. This col- lection contains possible situations, but also impossible (incoherent) ones. Notice that for simple positive sentences </»like 'I am stroking Jackie,' [u:</>] is'persistent':ifse[u:<f>]and s^s', thens'e[u:#].Truth is a property of statements, not of sentences. A state- mentu:<l>is'true'inA/iffthereisanssuchthat seM0 and se[u:</>].
In keeping with the underlying philosophy, there are two notions of logical consequence for statements. If 4>and f are statements, then "Fis a 'strong conse- quence' of <l>iff [<i>] is a subcollection of [ f ] , and f is a 'weak consequence' of 4>iff f is true in every M where <I>is true. The notion of strong consequence is relevant for the logic of information containment: if someone knows that <P, and 'Pis a strong consequence of <P, then (s)he knows that f . The notion of weak consequenceismoretraditional.If<Pistrue,and fis a weak consequence of <P, then f must be true as
prove the first (and weaker) claim, one has to show that[John]hadhiseyesopenandfunctioning,and that an event of a certain sort was taking place before him. Toprovethestrongerclaim,oneneedstoprove something about what he recognized and what thoughts were going through his mind' (S&A: 179). See- reports like (1) are considered paradigmatic for a semantics of the attitudes. One reason for this is that they report, in a sense, the most realistic attitude; what is seen is real. Another reason is that they have the least controversial semantic properties, like:
(a) Veridicality: if a sees 0, then 0*, where 0* is the tensed version of <f>. Jackie sees Molly scratch. So, Molly scratches. (<t>must be a sentence giv- ing rise to a persistent interpretation. It does not follow that if Alice is seeing no one walk on the road, then no one is walking on the road.);
(b) ConjunctionDistribution:ifasees0and^,then a sees <f>and a sees ty;
(c) DisjunctionDistribution:ifasees<f>orij/,thena sees ^ or a sees \j/.
Properties like these should shed light on the proper treatment of see-reports. Standard possible world semantics assumes that the semantic value of embed- ded clauses is the set of possible worlds in which they are true. This leads to the problem that if a sees <f>and
well. Note, however, that the weak consequences of a
statementdonotcompletelycoincidewiththeconse- ^islogicallyequivalentto\J/,thenasees$.Hence,
quences in classical logics.
The relational theory of meaning, in which u and 5
are the same type of thing, easily explains the phenom- enon of 'inverse information.' An utterance can con- vey information about a described situation, but also about the utterance situation. Suppose someone next door says 'I am Jon.' This utterance describes situ- ations in which the speaker of the utterance is called Jon. However, for someone who knows Jon, the utter- ance may convey information about the utterance situation: he is speaking there! Note that this infor- mation is not part of the linguistic interpretation of the utterance.
From the outset, situation semantics has been con- cerned with attitude reports, utterances of sentences containing verbs like 'see (that),' 'know that,' 'believe that,' which are used to report perception and cognition. The seminal paper 'Scenes and Other Situ- ations' (1981, reprinted in Barwise 1989) studies the semantics of sentences reporting visual perceptions. Syntactically, one can distinguish at least two kinds of those reports. In (1), the verb 'see' has an untensed sentence (or 'naked infinitive') as its complement; in (2), it selects a //zar-complementizer followed by a tensed sentence:
Semantically, 'epistemically neutral' 'see' can be contrasted with 'epistemically positive' 'see that': 'To
John saw Jackie bite Molly entails John saw Jackie bite Molly and Tully scratch or not scratch, since Jackie bite Molly and Jackie bite Molly and Tully scratch or not scratch denote the same set of possible worlds.
In the special case of verbs like 'see,' this assumption has even worse implications. Using veridicality, con- junction distribution, and disjunction distribution, one can derive 'omnipercipience' ('you've seen one, you've seen them all'): if for some $, a sees 0, then
for all true ^, it must hold that a sees if/.
Situation semantics solves those problems by
assigning clauses a more fine-grained semantic value: situations. See denotes a primitive relation, SEE, between a location, an individual (the agent), and a situation. The infon <SEE, /, a, e; 1 > corresponds to the fact that a sees e at /, and see-reports are interpreted in the following way: ufJohn is seeing Jackie bite Molly\s iff <SEE,/,a,j';l>es and <Brr£,/,ft,c; l>es', where / = u(is), I temporally overlaps with the location of u, a=u(John), b = u(Jackie), and c=u(Molly).
It is easy to see that this analysis validates con- junction and disjunction distribution. However, ver- idicality is not yet guaranteed. If (SEE,/,a,5';1>es, and s is actual, nothing follows about s being actual or factual; s' is just a possible (or even impossible) situation doing duty as a constituent of a fact. To get veridicality, a constraint is needed: (INVOLVES,5,S';
1>, with 5={<SEE,/,a,s; !>} and S"=s (where 5 is a situation indeterminate).
The effect of this constraint is that for every actual 5', there is a factual s> such that for all locations /,
380
John saw Jackie bite Molly.
(1) John saw that Jackie bit Molly. (2)