If I touched Jimmy, he would burst into tears.
Is it true that "it rains in California"? Yes. Is it true that "everybody
always gets gloomy"? No. Therefore, sentence (1) is false.
But of course that line of reasoning doesn't make sense. We have to
understand the phrase
"in California" as taking the entire sentence in its
scope, just as we understand "everbody" as taking in Californians
who experience rain, which is not at all what "everybody"
means in a self-standing sentence.
Sentence (1) is undoubtedly false, but not for
the reason worked out from the truth-table for material implication.
Is it true that "I touched Jimmy"? No. I wouldn't think of it. Is
it true that "he would burst into tears"? I can't answer that, since
a sentence with a conditional modal can't be evaluated on its own.
Assuming, for the same of argument, that the second clause is either true
or false, then sentence (2) is true. But we know, of course, that the
truth of this sentence, as we usually understand it, cannot be determined
in that way.
The point is, of course, that the subtle ways in which we understand the
actual conditional sentences that get used in everyday talk involve
detailed consideration of the actual grammatical form of the sentences
Types of Meanings of Conditional Sentences
Eve Sweetser, in From Etymology to Pragmatics has classified
conditional semantics according to the three domains she speaks of in
that book, the content domain, the epistemic domain, and
the speech act domain. Content-based conditionals are understood
by relating the content of the two clauses to each other. A typical
way in which content conditionals can be understood is for the "P" clause
to identify a situation which causes or automatically results in the
state of affairs signalled by the "Q" clause. This is the case for
Epistemic conditionals are understood as expressions of the reasoning
process. If the state of affairs represented by the "P" clause turns
out to be true, then we are licensed to believe what we are told in the
"Q" clause. Thus:
If you drop it, it will break.
If you say that again, I'll slap you.
If it rains, we'll cancel the picnic.
And speech act conditionals are understood as pre-posing to a speech
act a "P" clause that identifies the situation which got the speaker to
provide the speech act. Thus:
If their lights are on, the Wilsons are home from their vacation.
If the streets are wet, it rained last night.
If she wins, she's been practicing in secret.
We will see, in comparing the verbal forms of conditional sentences,
that some combinations can only have the epistemic interpretation, others
can have either an epistemic or a content interpretation. I have not
explored the formal conditions for being a speech-act conditional.
- If you're hungry, I could find something for you in the fridge.
If you leave before I see you again, have a good time.
If what I said offended you, I apologize.
A major descriptive problem that
grammarians have to face in dealing with
English conditional sentences involves the complex
system of compatibility relations between the two
parts of a conditional sentence. That is, certain
verbal forms occurring in the antecedent clause of a
conditional sentence are compatible only with
certain other verbal forms in the consequent clause.
Some examples of compatible combinations are
Some examples of incompatible (or at least
difficult-to-contextualize) combinations are the
If she opens it, they will escape.
If she opened it, they would escape.
If she had opened it, they would have escaped.
If she opened it, they escaped.
What we need for this set of facts is some set of
general principles according to which these
acceptability judgments, and the accompanying
interpretations, can get explained.
*If she'll open it, they had escaped.
*If she were here, I'll be happy.
*If she opens it, she had misunderstood my message.
The tools we need for stating these
principles include the following:
First, we need to
have a vocabulary for describing the various verbal
forms which enter into the compatibility relations
second, we need to speak of
something I will refer to as "epistemic stance" - the
speaker's stance on the reality of the proposition
expressed in the antecedent clause;
third, we will
need to notice that some sentences give expression
to what we can call the "interlocutors' interest" - the
speaker's view that of the alternatives recognized by
a conditional sentence, one is looked on as matching
the speaker's or the hearer's interest (this will be modified below); and
fourth, we will need to notice features of "polarity" - the difference between
positive polarity and negative polarity.
Describing the selection of verbal forms in English conditional sentences is
by the facts that some of the relevant categories are not identifiable with
particular morphemes or particular
individual grammatical notions, but with
complexes of these. What this means is that we will
have to give different names to forms that have the
same, or almost the same, superficial appearance.
Furthermore, in discussing the categories we need, it
is necessary to keep in mind the difference between
"Time" (which we take as a semantic notion) and
"Tense" (a grammatical notion).
The names of the verbal-form categories we will use are these:
the form which, in the copula, results
in is, am, are and in the non-modal verbs uses
the sibilant suffix to express third-person-singular agreement (walks)
the form which, in the copula, results
in was, were and otherwise, in the
"regular" cases, the simple past-tense
the expression of future meaning with
the modal will followed by the
this form is the same as the past-
tense form, except that, in some
dialects (perhaps especially in the
U.S.) there is a single form for the
this form is the same as the
pluperfect form (had gone, etc.),
except that in colloquial English we
also find a more complex form
(had've gone, etc.), and in colloquial
American English we find a form
identical to what I will call
"conditional perfect": would have
this form is constructed with would or
could plus the unmarked infinitive
(would go, etc).
this form is constructed with would or
could plus the perfect infinitive
(would have gone, etc.)
In general, "perfect aspect" and "progressive
aspect" can coexist with most of these forms and
contribute their own meanings. In other words, in
describing a conditional antecedent, the form "if he
has seen her" will be simply classified as "present"
for present purposes.
In the immediately following discussion we will combine
conditional sentences with sentences having a
temporal subordinate clause.
We can distinguish three sorts of epistemic stance -
positive, neutral, and negative - which will indicate the degree of the
speaker's commitment to the
actuality of the proposition expressed in a
In the case of positive epistemic stance, the
speaker accepts the truth of the proposition
expressed in the subordinate clause: Thus, in "when
Pat opened the door, the dog escaped", the speaker
accepts the idea that Pat did indeed open the door
and asserts that at that time the dog escaped.
In the case of neutral epistemic stance, the
speaker takes no stand on the truth of the
proposition expressed by the subordinate clause.
Thus in, "If Pat left the door open, the dog
undoubtedly escaped", the speaker does not know
whether or not Pat left the door open, but asserts an
unfortunate consequence of such a state of affairs.
And in the case of negative epistemic stance,
the speaker assumes that "P" is not true, where "P" is
a proposition derivable from (and preserving the
polarity of) the form of the antecedent clause. Thus,
in "If Pat had left the door open, the dog would have
escaped," we hear the sentence as revealing the
speaker's belief that Pat did not leave the door open.
In using the words "positive" and "negative"
epistemic stance, rather than, say, "believes true"
and "believes false", I have in mind the fact that we
may be dealing with conceits rather than beliefs.
And in the case of future-time expressions, such as
the difference between "If she invites them, they'll
go" and "If she invited them, they'd go", we will
interpret the latter sentence not as expressing the
speaker's belief that "they" won't get invited, but
that - say - "other things being equal", they're not
likely to get invited.
It seems to me that there are three basic types
of conditional sentences, from the point of view of
Epistemic Stance. I can refer to these as Generic (in
which the speaker accepts the existence of instances
of P but is presenting the "conditional" as a general
principle), Neutral (in which the speaker makes no
commitment about the actuality of P), and Negative
(in which the speaker doubts the actuality of P). The
following tables will show the relationships between
Epistemic Stance, "Time", and Verbal Form. Each
cell in these tables names the form of the verbal
expression that expresses the Epistemic Stance (the
table), the Time (the column), and appearance as
Antecedent or Consequence (the row). Any
conditional sentence can be formed by choosing,
from one of the tables, one cell from the upper
column and one cell from the lower column. (There
are some other constraints, to be noted below.)
Neutral Epistemic Stance
Negative Epistemic Stance
Examples of Neutral-ES and Negative-ES
conditionals, illustrating each formal possibility,
Neutral Epistemic Stance
It should be noticed that there are different pragmatic purposes to
conditional sentences, which we can think of as causative versus inferential.
Those in which the time of the antecedent follows the time of the consequent
are necessarily of the inferential type.
Negative Epistemic Stance
The upper left ("past subjunctive") corner of the
Negative ES diagram has a special status, in that
there is a variety of forms that can express it. The
standard form is identical to the pluperfect: "If I
hadn't opened it." But there is a general colloquial
form "If I hadn't 've opened it" and there is a special
American colloquial form "If I wouldn't have opened
if I hadn't opened it
if I hadn't've opened it
if I wouldn't have opened it
A very important fact to notice about this
collection of alternatives, and their evaluations, is
that it characterizes not only the past Neg-ES forms
of conditional antecedents, but also other contexts
with Neg-ES meanings.
One such context is as the complement of the
verb "wish". Wish is the only verb in English which
accepts these forms in its complement. We find
(with the same acceptability judgments):
The verb wish is used not only for expressing
past counterfactual wishes, but also for expressing
present and future wishes. In the case of present-
time wishes, we find the sentential complements of
wish taking the same present-subjunctive form we
found with present Neg-ES antecedents. Thus, in "I
wish you lived closer to Berkeley", the past-tense
form is used to express a wish about a present-time
situation, and in "I wish she were here", the special
form "were" (rather than "was") can be used.
I wish I hadn't said that.
I wish I hadn't've said that.
I wish I wouldn't've said that.
There is one observation that keeps us from
concluding that the complements of wish are simply
identical, in their formal requirements, with Neg-ES
antecedents, and that has to do with the FUTURE
form. The future Neg-ES antecedent form is the
same as the past tense, but in the case of wish, we do
not get "*I wish you introduced me to Louise
tomorrow", but "I wish you would introduce me to
Louise tomorrow." How are we going to account for
the obligatory "would" in this clause?
I propose that clausal complements of wish and
the antecedents of Neg-ES conditonals are indeed
constructed in accordance with the same principles,
but so far we have left out one set of facts. When
such a clause expresses the Interlocutors' Interest
(or that of some other discourse-relevant individual),
the future-time version is formed with the modal
"would". Since "wish" necessarily expresses the
speaker's interests, the construction with "would" is
obligatory in that case.
This means that we should be able to find cases
of "would" in the Neg-ES antecedents of conditional sentences, and that such
clauses should be taken as expressing one or both of
the conversation participants' interests. That is, in fact, what we find.
Consider first a comparison of cases where we
learn from the consequent whether or not the
speaker has a positive interest in the outcome.
Both of these sentences are acceptable. We
can infer from the first one that the speaker wants
the addressee not to have this conversation, and
from the second one that the conversation with the
father is desired. But the grammatical form of the
sentence does not express these judgments. But
now let us look at the same sentences with would:
If you spoke to my father about
we'd get in serious trouble.
If you spoke to my father about
I might get permission to go.
The oddity of the first of these sentences is that the
consequent seems to contradict the assumption
suggested by the verb form in the antecedent,
assuming that the speaker of the sentence does not
?If you would speak to my father
we'd get in serious trouble
If you would speak to my father
I might get permission to go.
Having seen that there is a separate form for
Neg-ES future antecedents revealing participant
interests, we can now ask whether such a possibility
also exists for Neutral-ES sentences. It appears
there is, namely in the form of the modal "will". We
noted earlier that FUTURE Neutral-ES antecedents
use the simple present tense form, instead of the
expected will-future; but we can find "will" in
sentences exhibiting the participants' positive
The questioned sentences in the preceding set are
all odd, since they suggest that the speaker wants it
to rain, or wants the addressee to break a dish.
If the sun'll shine we'll be able to
have our picnic.
?If it'll rain, we'll have to cancel the
If you break another dish, I'll give
you a spanking.
?If you'll break another dish, I'll
give you a spanking.
In earlier work I suggested that the will...will
form of a conditional sentence was dedicated to
"negotiations" or "negotiated offers". supported by
sentences like "If you'll wash the dishes, I'll dry" and
"If it'll make you feel any better, I'll stay another day
or two". But I think now that the explanation of these
forms is more general, and that the "negotiation"
aspect of the interpretation of these sentences is
merely a by-product of the sentences' ability to
express both participants' interests.
There is a generalization to be captured here.
We are now free to say that in future-time
antecedents, the modal will is used, and that this
form has its present-tense form will in the Neutral-
ES case, the past-tense form would in the Negative-
ES case. Hence:
In the cases where the future antecedent expresses the interlocutors'
interests, the form will is used, in each case:
It is well known that the antecedent clauses of
conditional sentences are - or are capable of being -
"negative polarity contexts", but this is only when the
sentence does not express the interlocutors' interest.
Some linguistic forms are generally welcome in only
positive (or "positive interest") sentences, e.g., "a little".
Other expressions, e.g., "any" (in the relevant
meaning), are generally welcome only in sentences
expressing uncertainty or negative interest.
Compare the following,
In the former case, I invite you to come closer,
and propose a reason why you should be interested
in doing so. In the latter case, I discourage you from
coming closer, and I propose a reason for you to
want to do otherwise.
If you come a little closer, you'll
be able to see better.
If you come any closer, I'll call the
If we were to examine the compatibility
problems for antecedent and consequent verbal
forms in English conditional sentences, mentioned
at the beginning of this section, we will find that the
ones which are possible are those that "fall out from"
the combined principles governing tenses,
epistemic stance, and interlocutor interests, and that
the ones which are impossible cannot be derived
from the patterns that such principles create.
What are the things that need to be explained in a theory of English
The P clause can appear before or after the Q clause.
The Q clause can begin with "then" only when the P clause comes first.
The relation between the P-clause and its relevant Q-clause appears to be one
of long-distance dependencies, which makes it similar to the LI construction.
But there's a contradiction here, since the "sealing" (or "island" phenomena
that we observe in other instances of LI (in particular in respect to the so-
called "WH-island")) doe not seem to hold in the case of conditional
There are compatibility relations in the form of the verbal expressions in the
two clauses, such that the form of the verbal expression in each clause
depends on whether
the clause is P or Q
the clause is of tense past, present, or future
or perhaps simply past vs. nonpast, in the negative cases?
the epistemic stance is neutral or negative
the event signalled by P is of positive interest to somebody
In some cases the structural feature which is the formal reflection of some
such combination of conditions is morphologically complex, and not in a way
that lends itself to being understood as a single constituent. This is
so, because we have regarded VPs with complex auxiliaries as right-branching
structures, where each successive auxiliary is in construction with the entire
remainder of the VP. That is, we are not free, willy nilly, to regard "would
have" as a single unit, since the structure we have imposed on such
sequences is, say, "[would [have [eaten it]]]" and not "[would have] [eaten it]"
(One wonders what this has to do with some observations of Joyce Tang
Boyland (whose office is across the corridor from our classroom) about
modal+have sequences functioning as single auxiliaries, in such colloquial
sentences as "What would have you said?". See J. T. Boyland, "A
corpus study of would + have + past-participle in English", to appear
in Nigel Vincent, ed., Historical Linguistics, 1995.)
Negative epistemic stance has to be understood in a special way in case the
intended time is 'future'. When it is present or past, the meaning is
essentially that of counterfactuality; but when it is future it appears to
be a matter of other-things-being-equal likelihood. ("If she asks me, I'll
aaccept" versus "If she asked me, I'd accept".)
The person whose "positive interest" is in question for the positive polarity
sentences is one or both interlocutors in a typical case, but it might also be the
"individual" we've been referring to as prag, that is, the person from
whose point of view something is being expressed, which by default
will of course be the speaker or addressee.
In our informal discussion of the two parts of a conditional sentence we have
been free to use words like "antecedent" vs. "consequent", or even
"subordinate clause" versus "main clause", but we need to do better than that
in representing the construction that will be able to take care of all of this. We
will in fact need some way of showing that the P clause has an "anchoring"
function with respect to the Q clause, since it provides the conditions under
which the interpretation of the Q clause is based.
This might need to be integrated into our proposals for the architecture of
our semantic representations, where we once proposed distinguishing
"settings" from "scenes". The "settings" part included temporal settings,
including tense and aspect, but also temporal, locational, and hypothetical
There appear to be some features shared by temporal and conditional adjuncts.
One of these is the use of present-tense for future meaning in the subordinate
clause. Another is the possibility of a sort of "correlative" phrasing, with
"then", when the subordinate clause precedes the main clause. Thus we have
When he starts treating me like a friend, then I'll start showing him some
If he does this, then I'll do that.
Shall we arrange for the conditional construction to iterate? I would prefer
to think of sentences like the following as instances of language play, and not
as evidence of some natural process in English: "If we had some ham,
we could make some ham and eggs, if we had some eggs."
All of this means, of course, that when we write up the construction
description, it will need to contain a number of primitive semantic/pragmatic
notions that require interpretations that have not yet been completely pinned
down. These will include our notions of anchoring, epistemic stance, prag's
positive interest, and the notion of prag itself.
Watch This Space