In this paper, I will argue that if-clauses can be focused and that they associate with focus sensitive operators. To account for differences in the interpretation of focused conditionals, I will propose that there are two focus operators in English corresponding to the tunes H*LL% and L+H*LH% of Pierrehumbert (1980), with the semantic content of a universal/ exhaustive and existential/partial operator respectively .
It has been claimed that if-clauses, unlike when-clauses, do not undergo association with focus (Rooth 1985, Roberts 1994, von Fintel 1994, Heim quoted in Iatridou 1992). Rooth (1985), discusses the following sentences:
(1) If/When John is in the shower, he always shaves.=(3a)
(2) a. John always shaves when he is in the shower.=(3a)or(3b)
b. John always shaves if he is in the shower. (3a) only
(3) a. always (John is in the shower)(John shaves)
b. always (John shaves) (John is in the shower)
First, he observes that S-initial if/when-clauses can only
serve to restrict a focus-sensitive operator, so that (1) has the
reading in (3a). Secondly, he claims that there is an assymmetry
in the behavior of S-final if/when clauses. Sentences with a
when-clause as in (2a) are ambiguous between the reading given in
(3a) and that in (3b) corresponding to whether the main clause or
the when-clause is focused, respectively. On the other hand,
sentences with S-final if-clauses are not ambiguous: (2b) only
has the reading in (3a). Rooth's tentative suggestion
essentially involves the idea that if-clauses, being topics
(Haiman 1978), cannot be focused.
von Fintel and Roberts have already pointed out that if- clauses can be focused. First, even on the assumption that if- clauses are topics, it is not clear that they cannot be focused, given the existence of "contrastive topics". Secondly, from a phonological and discourse-functional perspective, it is clear that if-clauses CAN be focused as shown in (4):
(4) A: What would motivate John to shave?
B: John always shaves, [if his MOTHER is coming to visit]F
In (4), the if-clause carries the nuclear accent of the
utterance and provides the new information requested in the
question - both properties associated with focus. von Fintel and
Roberts nevertheless agree with Rooth that differences in focus
structure in conditionals do not have truth conditional effects.
This claim is clearly false when one examines conditionals with
the prototypical focus-sensitive operator only:
(5) a. John only apologizes, [if/when he makes a MISTAKE].
b. John only [APOLOGIZES] if/when he makes a mistake.
(5a) and (5b) are truth-conditionally distinct. Consider a situation in which John apologizes for being rude to someone (assuming that being rude is not considered 'making a mistake'). Then, (5a) is false while (5b) may be true. Whatever the correct analysis of association with focus turns out to be (pragmatic, contextually determined; or semantic, gramatically determined), it is clear that it should apply equally to the S-final if- clauses and when-clauses in the examples above, accounting for the observed truth-conditional differences. However, not all if-clauses can associate with focus. The if-clause in a so-called relevance conditional as in (6a) or in a factual conditional as in (6b) does not associate with only.
(6) a.There's only a beer in the fridge, if you're thirsty.
b.She should only do some of the work, if she's so sick.
The reason is that the if-clauses in these examples are
outside the scope of the operator in the main clause. Iatridou
(1991) argues that the if-clause in these conditionals is
attached higher in the tree (attached to I-bar or IP/CP in the
terminology of GB). Furthermore, even though they may carry a
nuclear accent, and hence be phonologically focused, they must
form a separate intonational phrase from the main clause. In
other words, they constitute backgrounded/ nonrestrictive/
appositive information and therefore cannot provide the answer to
a question.
Assuming then that (hypothetical) if-clauses do associate
with focus, the question is what readings these utterances have
and how these readings are derived from an account of association
with focus. My claim is that focus on the main clause of a
conditional, has a reading equivalent to one where the if-clause
restricts an (implicit or overt operator), while focus on the if-
clause, in the context of a question, produces a reading
equivalent to a biconditional. Focus on the if-clause in (2b)
then produces a reading equivalent to:
(6) (always (John is in the shower)(John shaves))and
(always (John shaves)(John is in the shower))
Analogously, if B's claim in (4) is true, then in any
situation where John is shaving, I can conclude that his mother
is coming to visit. There is an important qualification however:
the biconditional reading is crucially dependent on a falling
intonational pattern - transcribed in Pierrehumbert notation as
H*LL%. It is also possible to focus the if-clause using a
different contour, namely a fall-rise, or L+H*LH%. With this
pattern we no longer get a biconditional reading, obtaining
instead the reading indicated by Rooth. Although it has been
claimed that the L+H*LH% contour is associated with themes/topics
(most notably by Steedman), I will propose that there are two
focus operators, one with universal force (H*LL%) and the other
existential (L+H*LH%) - the focus operator is then not simply a
presuppositional operator. I will then show how the two operators
can account for the interpretation of conditionals with focused
if-clauses. The analysis will also be shown to extend to simple
declaratives, accounting for cases that have been problematic for
the general correlation between nuclear stress/ answer-new
information/focus, such as Rooth's:
(7) Carl likes herring.
L+H* LH\%
(7) does not entail that Carl is the only relevant person who
likes herring, but merely that he is one such person. What we do
give up is the idea that answers or focus must be exhaustive.
Selected References
Rooth, Mats 1985. Association with Focus. Ph.D Dissertation. University of Massachusetts. Amherst, Massachusetts.
von Fintel, Kai 1994. Restrictions on Quantifier Domains. Ph.D. Dissertation. University of Massachusetts. Amherst, Massachusetts.
Roberts, Craige 1994. If and when: The Semantics of Conditional and Temporal Subordinating Conjunctions. Ms. The Ohio State University. Columbus, Ohio.
Iatridou, Sabine 1991. Topics in Conditionals. Ph.D. Dissertation. MIT. Cambridge, Massachusetts.