Jump to ContentJump to Main Navigation
Donald DavidsonA Short Introduction$

Kathrin Gl¨uer

Print publication date: 2011

Print ISBN-13: 9780195382976

Published to Oxford Scholarship Online: May 2015

DOI: 10.1093/acprof:osobl/9780195382976.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017

(p.281) Appendix

(p.281) Appendix

Source:
Donald Davidson
Publisher:
Oxford University Press

A T-theory for a Fragment of English1

We shall now consider a very simple fragment of English that we shall call ‘QE’, and show how to construe a T-theory for it. The vocabulary of QE contains the following:

1. Proper names:

‘Elsa’, ‘John’, ‘Mary’, ‘Paul’

2. Predicates:

‘sleeps’, ‘loves’, ‘envies’, ‘helps’

3. Logical particles:

‘and’, ‘or’, ‘it is not the case that’, and ‘if, then’

4. Quantifiers:

‘everyone’, ‘someone’

The grammar is very simple, too. Proper names and quantifiers are both noun phrases (NPs). The category S is that of sentences. The rules are then:

  1. 1. NP + ‘sleeps’ is an S

  2. 2. NP1 + ‘loves’—‘envies’—‘helps’ + NP2 is an S

  3. 3. ‘it is not the case that’ + S is an S

  4. 4. S1 + ‘and’—‘or’—‘if, then’ + S2 is an S

(p.282) By these rules we can e.g. form the following sentences:

  1. (5) Mary sleeps or John loves Mary.

  2. (6) Everyone loves someone.

  3. (7) If Elsa helps Paul, then Paul loves Elsa and John envies Paul.

Even though this fragment is very simple, the examples show that it has ambiguous sentences. (6) can mean that there is some person such that everyone loves that person, or that for each person there is some person, not necessarily the same, that the first person loves. And (7) can mean either that John envies Paul (unconditionally) and if Elsa helps Paul, then Paul loves Elsa, and it can mean that If Elsa helps Paul, then Paul loves Elsa and (under this condition) John envies Paul. These ambiguities matter semantically So we need to disambiguate the sentences before they can be interpreted.

We disambiguate by means of bracketing sentences and indexing the quantifiers for scope relations. Using these devices we get the following readings corresponding to the readings just specified:

  1. (6′) Everyone2 loves someone1.

  2. (6″) Everyone1 loves someone2.

  3. (7′) (If Elsa helps Paul, then Paul loves Elsa) and John envies Paul.

  4. (7″) If Elsa helps Paul, then (Paul loves Elsa and John envies Paul).

In order to apply the Tarskian semantics, the fragment needs to be regimented into the apparatus of quantifiers and variables. Thus we will have

1. Proper names:

‘Elsa’, ‘John’, ‘Mary’, ‘Paul’

2. Predicates:

‘sleeps’, ‘loves’, ‘envies’, ‘helps’

3. Logical connectives:

‘&’, ‘∨’, ‘¬’, ‘→’

4. Quantifiers:

‘∀’, ‘∃’

5. Individual variables:

x1, x2, …

(p.283) The grammar is the usual formation rules for a first-order language. The translation rules are again very simple (we will not spell them out). What we need to take account of is the relative scope of simple sentences containing (different) quantifiers. Corresponding to (6′) and (6″) we have

  1. (6*)x1x2(x2 loves x1)

  2. (6*)x1x2(x1 loves x2)

One must also take care that when a new variable is introduced in the translation, it does not already occur in the scope of the new quantifier.

It should be noted at this point that the first-order language is expressively richer than the QE fragment. For instance, a first-order sentence such as

  1. (8)x1(∃x2(x1 loves x2) → ∃x3(x3 loves x1))

does not have a counterpart in QE. It would need forms such as

  1. (8′) Everyone who loves someone is loved by someone.

  2. (8″) Everyone is such that if he/she loves someone, then he/she is loved by someone.

In (8′) we use a subordinate relative clause, and in (8″) bound pronouns, to express the same proposition, and QE does not contain either of these devices. Adding them to QE would complicate the grammar considerably. It is not needed in order to exemplify the basic features of T-theories. Still, the T-theory that we give below, applied to just the regimented language, does provide an interpretation of (8) (after going through the example derivation below, you could try it as an exercise).

We can now set out the basic clauses for the T-theory for QE. We shall use‘s’ as a meta-language variable for infinite sequences (s = 〈s1, s2, …〉), and ‘a1, ‘a2’, … as meta-language individual variables. The axioms and axiom schemata are the following: (p.284)

1. Singular terms:

⊦ Ref(‘Elsa’) = Elsa

⊦ Ref(‘Mary’) = Mary

⊦ Ref(‘John’) = John

⊦ Ref(‘Paul’) = Paul

2. Predicates:

⊦ ∀s (sat(s,⌜xi sleeps⌝ iff si sleeps)

⊦ ∀s (sat(s, ⌜xi loves xj⌝) iff si loves sj)

⊦ ∀s (sat(s, ⌜xi helps xj⌝) iff si helps sj)

⊦ ∀s (sat(s, ⌜xi envies xj⌝) iff si envies sj)

3. Connectives:

⊦∀s (sat(s, ⌜S1 & S2⌝)iff sat(s,S1)and sat(s,S2))

⊦ ∀s (sat(s, ⌜¬S⌝) iff it is not the case that sat(s,S

⊦∀s(sat(s, ⌜S1S2⌝iff sat(s,S1) orsat(s,S2)))

⊦ ∀s (sat(s, ⌜S1S2⌝) iff: if sat(s, S1 ), then sat(s, S2))

4. Quantifiers:

⊦ ∀s (sat(s, ⌜∃xiPxi⌝) iff for some ai (sat(s[ai/si], ⌜Pxi⌝ )))

⊦ ∀s (sat(s, ⌜∀xiPxi⌝) iff for every ai (sat(s[ai/si], ⌜Pxi ⌝)))

where ⌜xi⌝ etc. is schematic, with instances like ‘x13’,

and where s[b/si] is the sequence differing from s at most by having b in its i:th position.2

5. Terms in formulas:

⊦ sat(s,⌜ Pt⌝) iff sat(s[Ref (t)/si], Pxi), provided ⌜ xi⌝ does not already occur in ⌜P⌝.

We will also need rules of inference in order to derive the right theorems. But we don’t want the full power of frst-order logic, since that would have the consequence, for instance, that if ⊦ A iff B is a theorem, so is ⊦ A iff (B or (p and not p)). But the latter would not be interpretive if the former is. Therefore, we limit ourselves to a few more specialized rules of inference. They provide one alternative for capturing what Davidson had in mind when speaking of “canonical derivations” (see above, p. 65). (p.285)

Rule 1:

If ⊦ a = b and ⊦ X(a), then ⊦ X(b)

Rule 2:

If ⊦ ∀s(A iff B) and ⊦ ∀s(X(A)), then ⊦ ∀s(X(B))

Rule 3:

If ⊦ ∀s(A(s) iff B) and ‘s’ does not occur in B, then ⊦ ∀s(A(s)) iff B

Rule 4:

If ⊦ A iff B and ⊦ B iff C, then ⊦ A iff C

Rule 5:

⊦ True (s) iff True (s′), if s′ is the regimentation of s

Rule 6:

⊦ for all s, a, i (s[a/si]i = a)

In Rules 1 and 2 ‘X’ is schematic for the sentential context. We can now define truth:

Rule 7:

⊦ True S iff for all s (sat(s, S))

To get a bit of a feeling for how this works, let’s run through a derivation. Below, for the sake of readability, we shall use italics instead of quote marks for expressions that are mentioned.

  1. (9)

    i)

    ⊦ True(If someone sleeps, then Paul sleeps) iff

    True(∃x1(x1 sleeps) → Paul sleeps)

    (rule 5)

    ii)

    ⊦ True(∃x1(x1 sleeps) → Paul sleeps) iff

    for all s(sat(s,∃x1(x 1 sleeps) → Paul sleeps))

    (rule 7)

    iii)

    ⊦ For all s (sat(s, ∃x1(x1 sleeps) → Paul sleeps) iff

    if sat(s, ∃x1 (x1 sleeps)), then sat(s,Paulsleeps))

    (Connectives)

    iv)

    ⊦ For all s (sat(s, ∃x1 (x1 sleeps)) iff

    forsome a1(sat(s[a1/s1],x1 sleeps)))

    (quantifiers)

    v)

    ⊦ For alls (forsome a1(sat(s[a1/s1],x1 sleeps)) iff

    for some a1(s[a1/s1]1 sleeps))

    (predicates, rule 2)

    vi)

    ⊦ For all s (forsome a1(s[a1/s1]1 sleeps)) iff

    for some a1(a1 sleeps))

    (rule 6, rule 1)

    vii)

    ⊦ For all s (sat (s, Paul sleeps) iff

    sat(s[Ref (Paul)/s2], x2 sleeps))

    (terms in formulas)

    viii)

    ⊦ For all s (sat(s[Ref(Paul)/s2], x2 sleeps) iff

    s[Ref(Paul)/s2]2 sleeps)

    (predicates)

    ix)

    ⊦ For all s (s[Ref(Paul) /s2]2 sleeps iff

    Ref(Paul) sleeps)

    (rule 6, rule 1 )

    x)

    ⊦ For all s (Ref(Paul) sleeps iff

    Paul sleeps)

    (singular terms, rule 1)

    xi)

    ⊦ For all s (if sat(s, ∃x1 (x1 sleeps)), then sat(s, Paul sleeps) iff

    if for some a1 (a1 sleeps), then Paul sleeps)

    ((9iv)-(9x), rule 2)

    xii)

    ⊦ For all s (sat(s, ∃x1 (x1 sleeps) → Paul sleeps) iff

    if for some a1 (a1 sleeps), then Paul sleeps)

    ((9iii), (9xi), rule 2)

    xiii)

    ⊦ For all s (sat(s, ∃x1 (x1 sleeps) → Paul sleeps)) iff

    if for some a1 (a1 sleeps), then Paul sleeps

    ((9xii), rule 3)

    xiv)

    ⊦ True(∃x1 (x1 sleeps) → Paul sleeps) iff

    if for some a1 (a1 sleeps), then Paul sleeps

    ((9ii), (9xiii), rule 4)

    xv)

    ⊦ True(If someone sleeps, then Paul sleeps) iff

    if for some a1 (a1 sleeps), then Paul sleeps.

    ((9i), (9xiv), rule 4).

    (p.286)

Notes:

(1.) Special thanks to Peter Pagin for helping me make this appendix.

(2.) The method of quantifying over individual meta-language variables applied in the quantifer clauses is suggested in Wiggins 1980, 325.