Basic semantics for quantification in English

Jeff Speaks
phil 43916
September 16, 2014

1 Models and assignments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Interpretation of sentences containing ‘every’ . . . . . . . . . . . . . . . . . . . 2
3 ‘A’ and ‘the’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1 Models and assignments

Our language now contains trees like

NP1 S

Det Nc e1 VP

every cat Vi

is cute

What we need to know now is how to interpret trees like this: what does it take for them
to be true or false?

Two answer this question we need to introduce two new concepts: the concept of a model
and the concept of an assignment.

A model M is a pair of two things: a valuation function V and a domain (or universe) U of
discourse. These ideas are already familiar, in somewhat different form, from our simpler
language. Recall that we needed to talk not just about, for example, Jis boringK, but also
Jis boringKv – where the latter is JboringK relative to some circumstance of evaluation. A
valuation function V1 is a function from the expressions of our language to their semantic
value in a situation v1 .

A domain U1 is just the list of things that exist in v1 . The semantic values assigned to
expressions by V1 must be built up from elements of U1 . In the examples we’ve been
discussing, the domain was {Pavarotti, James Bond, Sophia Loren}.

1
So we can talk about the semantic value of “is boring” relative to a model in much the
way we before talked about Jis boringKv . This is a first step in making more precise what
we meant by ‘relative to a state of the world’ before.

An assignment g is a function from traces to elements of the domain. Recall that we

have infinitely many traces e1 , e2 , . . . in our language. One assignment, given the above
domain of discourse, might be

[e1 → James Bond

e2 → Sophia Loren
en → Pavarotti ] for any n > 2.

Another might be

[e1 → James Bond

e2 , 33 → Pavarotti
en → Sophia Loren ] for any n > 3.

A very simple one is

[en → James Bond ] for any n

All that’s required is that the assignment g be a function from the variables to members
of the domain.

Rather than talking about semantic values relative to v, we can now talk about the
semantic value of an expression relative to a model and an assignment (written as, e.g.,
Jis boringKM1 ,g1 ). This notion is defined (for the lexicon described above) as follows:

For a model M1 = hU1 , V1 i, if A is an expression which is not a trace, then JAKM1 ,g1 =
V1 (A); and if A is a trace, then JAKM1 ,g1 = g1 (A).

What does every valuation function return when given ‘Pavarotti’ as argument?

What is g 2 (e4 )?

2 Interpretation of sentences containing ‘every’

Let’s return to the sentence above:

2
 S

NP1 S

Det Nc e1 VP

every cat Vi

is cute

What does it take for this sentence to be true or false? We know that JcatKM,g will be a
set of things – the set of cats in the relevant model. So our question boils down to the
question of how we should understand the contribution of ‘every’ to the truth conditions
of sentences in which it occurs.

The basic idea is that we start with an assignment g of values to all of the traces, and we
then consider every assignment function which agrees with g on every trace other than e1 .
If our sentence is true in M with respect to every such assignment, then our quantified
sentence is true relative to M and g.

We use

g1 [u/e1 ]

to mean

the assignment function which differs from g 1 only in assigning u as the value
of e1

Using this notation, we can state the lexical entry for ‘every’ as follows:

[u/ei ]
J[[every β]i S]KM,g =1 iff for all u ∈ U , if u ∈ JβKM,g , then JSKM,g =1

What does this say?

Can you use this rule to derive truth conditions for ‘Every cat is cute’ ?

If this sentence is true relative to one assignment g 1 , can it be false relative to another
assignment g 2 (assuming that we hold fixed the model)?

3 ‘A’ and ‘the’

Sentences involving ‘a’ and ‘the’ don’t differ in their structure from sentences containing
‘every’; so what we need to be able to interpret these sentences are just the lexical entries
for ‘a’ and ‘the.’ They are as follows:

3
 [u/ei ]
J[[a β]i S]KM,g =1 iff for some u ∈ U , u ∈ JβKM,g and JSKM,g =1

[u/ei ]
J[[the β]i S]KM,g =1 iff for some u ∈ U , JβKM,g ={u} and JSKM,g =1

What does the clause for ‘the’ say?

What does this lexical entry tell us about the truth conditions of ‘The cat is cute’ ? Is
this surprising?

Not all quantifier phrases are as simple as ‘every cat.’ In particular, some quantifier
phrases contain others, as in ‘every cat next to a dog.’ Do our semantic rules tell us how
to handle quantifier phrases of this sort? What would happen if you tried to apply the
rules above to this phrase?