On Conditionals

+_--
+
THE JOURNAL OF PHILOSOPHY
VOLUME XCI, NO.
i+.
0
3,
MARCH
4.
1994
- .
..
_.
ON CONDITIONALS
E
ach of the sentences (1) can encode a messageabout a future
sporting fixture:
(la) A Spurs-Liverpoolfinal will be a fitting climax to the season.
(Ib) A Spurs-Liverpoolfinal would be a fitting climax to the season.
(ic) A Spurs-Liverpoolfinal would have been a fitting climax to the
season.
The main thrust of these three messages is of course the same. It
concerns the fittingness of a final between Spurs and Liverpool, as
who should argue both teams play skilful, inventive football. The
three messages share a common core, we might say.
But just as plainly some mysterious force is also at work, making
(Ib) the appropriate formulation before the finalists are determined
and either (la) or (1c) thereafter. We can summarily distinguish
elements of the core from our mystery agent by how they are encoded. From the core arises everything the three sentences share,
while the mystery agent governs the choice between them. Their
different preoccupations will be apparent at once. The choice
between (la), (Ib), and (Ic) is made with an eye to who the real
finalists are going to be, an issue to which the core sentiment is
wholly oblivious.
I shall urge that all this applies also to conditionals. A conditional
has a core; when the conditional is affirmed, the core is sustained by
underlying reasoning; the underlying reasoning is separate from the
preoccupations, whatever exactly they are, of a mystery agent with
control over the form.
In due course, I shall unmask the mystery agent and try to explain
what animates its activities. But with only a few suggestions to offer
0022-362X/94/9103/113-28
(?)1994 The Journal of Philosophy, Inc.
113
114
THE JOURNAL OF PHILOSOPHY
about cores and underlying reasoning, I disclaim any detailed semantic account of conditionals. My principal submission will be the
need to distinguish two semantic themes when explaining them. Predictably, one relates to the core and the other to the mystery agent.
By a conditional I mean a quite particular kind of message. I shall
define one in section IV, although the idea will be clear long before
that, I hope, just from the examples.1 The received analysis of conditionals recognizes three factors: an "antecedent," a "consequent," and some sort of operator. Having disputed it elsewhere,2 I
ignore this analysis here, preferring simply to present my own in its
own terms. In quest of perspicuity, I eventually resort to diagrams.
To minimize technicality, on the other hand, I remain at the level of
examples throughout.
I
At the beginning of September 1940, Britain alone and all but defenseless, conditions greatly favored a German invasion of England,
moon and tide arguing September 7 as optimal. But the invasion
never came: by September 15, the British had shot down many of
the fighters that might have supported an invasion, and we who had
lately maintained (2a) or (2b) found ourselves obliged to abandon
these locutions in favor of (2c):
(2a) If Hitler invadesEngland,Germanywill/can/may win the war.
(2b) If Hitler invaded(was/wereto invade)England,Germanywould/
could/might win the war.
(2c) If Hitler had invaded England, Germany would/could/might
have won the war.
It is not that we were changing our position. The conditions had
changed, not our thinking. If it was good thinking when we said (2a)
or (2b) back on September 1, it was good thinking when we said (2c)
after September 15; for it was the same thinking. It was premised
(perhaps) on England's unpreparedness and the strength of Hitler's
airforce and other such intelligence relating to the early days of
September. This thinking is about Germany's winning the war in the
aftermath of Hitler's invading England, inferring, indeed, a verdict
about this topic. The verdict is expressed by 'will/would', 'can/
could', or 'may/might'.
And here we have the core: a verdict about Germany's winning
the war in the aftermath of Hitler's invading England. It is this core
l In the terms of a taxonomy I reject, my "conditionals" coextend with the
union of "subjunctives" and "future indicatives."
2 E.g., in part 3 of "Antecedents and Consequents," Theoria, LII, 3 (1986):
168-99.
ON CONDITIONALS
115
which the underlying reasoning sustains, reasoning to which it is
irrelevant whether Hitler in fact invaded England. But there, too,
lurks our mystery agent, presiding over the choice between (2a),
(2b), and (2c). And it apparently monitors this irrelevant coincidental with particularity.
II
The mystery agent moves in a mysterious way. Compare the futureconditional interpretations of (2a) and (2b), for instance. There is no
difference in their logical strengths, for if Hitler had invaded England on September 7, the proponents of either would have become
committed to the then burden of 'Germany will/can/may win the
war'. Inferentially, the two are indiscernible. Yet there is a difference. Its existence is easily demonstrated,3 and in any case we can
sense it. But how is it to be described?
A frequent innocent reaction is that (2b) sounds "indefinite,"
certainly less "definite" than (2a)-an intuition that demands its
explanation while failing to enlighten. Otto Jespersen and others
have felt that we say (2a) or (2b) depending on whether we deem an
invasion of England by Hitler more or less probable; but there can
be no such connection, seeing that the following both make perfect
sense:
If Hitler invadesEngland,which I admit is a million-to-oneshot, Germanywill/can/may win the war.
If Hitler invaded (was/were to invade) England, something not at all
improbable,Germanywould/could/might/win the war.
According to F. R. Palmer,4 (2a) "leaves open the possibility" of
Hitler's invading England, while (2b) "indicates that the speaker
thinks it unlikely" that Hitler will invade England (ibid., p. 189). But
what this says of (2a) applies equally to (2b), and what it says of (2b)
is false.
Once it is too late for Hitler to invade England, one stops saying
(2a) or (2b) and says (2c) instead. This is not to acknowledge that
Hitler did not in fact invade England, for (2c) does no such thing: it
makes perfect sense to say
'Those who said (2a) early in September 1940 can now be reported as having
said that Germany would/could/might win the war if Hitler invaded England, but
those who said (2b) then cannot now be reported as having said that Germany
willlcan/may win the war if Hitler invades England. So there must be some difference.
' Mood and Modality (New York: Cambridge, 1986). On probability, see Jesperson, A Modern English Grammar Based on Historical Principles, Part V (London:
Allen and Unwin, 1940), p. 377.
116
THE JOURNAL OF PHILOSOPHY
Whetheror not Hitler did invade England,if he had invaded England
Germanywould/could/might have won the war.
But if not that, then what?
It is not just easy, then, twigging how the mystery agent operates.
But before we turn to that problem we need to know something of
how conditionals are built up out of their bits. There is no hope
otherwise of understanding how they work.
III
Each variant of a sentence (3) can be said about a future German
win, and (3c) can still be said after the war is over:
(3a) Germanywill/can/may win the war.
(3b) Germanywould/could/might win the war.
(3c) Germanywould/could/might have won the war.
I call the messages thus transmitted projective. Again we recognize a
core, this time comprising a verdict about Germany's winning the
war simpliciter. And there again is our enigmatic arbiter of the
form.5 I classify projective messages as judgments because they have
verdicts for immediate informational factors. This distinguishes
them from propositions, such as the natural interpretation of 'Germany won the war'.
There is nothing out in the world to which a verdict corresponds.
In affirming a projective message, the speaker offers a personal
venture of her own, which of its nature calls for justification. This
justification takes the form of underlying reasoning. It is in the nature of this reasoning that it go beyond the facts. It is premised on
facts, certainly, but the latest available facts are present facts, yet the
verdict can be about something happening in the future. The verdict
must be about Germany's winning the war or whatever in an imagined situation. This is patent when the win is future. And it is typically when the win did not actually pan out that (3c) is said about
the past.6
5 Eyeing the work of said mystery agent, we observe that 'would (have)' sometimes prompts the question 'would (have) if what?' The phenomenon awaits explanation, but mine runs too long for inclusion here. The regularity is after all quite
special, holding for neither 'could' nor 'might' and not always for 'would'. On
similar grounds, I leave the observed "indefiniteness" of (2b) unaddressed.
6 A verdict does not have to be about imagined things. For instance, every
variant of
have won the war.
Germany will/can't/may/would/could/might
has an interpretation whose verdict is about Germany actually winning the war in
the past. In fact, English encodes just two kinds of judgment, and this is the other
kind. It will be noted that (3c) is ambiguous between them.
Strict regularities relate these two kinds ofjudgment. Say 'My washing will/may
ON CONDITIONALS
117
How is this imagined situation arrived at? In my submission, it is
arrived at by seizing upon certain facts of an earlier date D, and
imagining a future to unfold in which those facts are allowed to
flourish and prosper, in what for them is a normal course of events. It
is as if the speaker accepted the history of the world just up until D,
and carried on from there in a fantasy in which certain arrangements, actual at D, were preserved. Illustration will eventuate in
sections VIII and IX.
The enigma remains of the mystery agent. But with the reflection
that English must be a system for putting ideas into words comes the
realization of what part of the answer must be. The forms (a), (b),
and (c) can only be registering tense. After all, the choice between
(3a), (3b), and (3c) exactly mirrors the choice between 'invades',
'invaded', and 'had invaded'. And the latter trio are well-known for
registering tense t, present, past, or pastpast.7 And the obvious inference, that the former choice is governed by t in exactly the same
way, is amply confirmed by independent evidence of many kinds.
So the mystery agent transpires to be none other than the tense.
And now the enigma, relocated, is how it achieves its effects. How,
simply by registering A-series location, do we contrive the differences we perceive, however dimly, between (a), (b), and (c)?
We must never confuse the tense of a message with the time y that
it is intuitively about. (3)'s verdict is about Germany's winning the
war, so here y is the time of the German win, future or past. y
belongs with the core. t is the time registered by the form, an utterly
different conception. Indeed, inspection will discover that projective messages always have y later than t ... which means that core
and mystery agent cannot be entirely unconnected.
IV
Now, quite evidently, the sentences (2) are generated from their
conditional interpretations by an elaboration of the encoding program that generates the corresponding sentences (3) from their projective interpretations. The program that generates (2) incorporates
an extra episode, responsible for adding a clause to the predicate of
(3). It is called an "adverbial clause of condition," or conditional
be dry soon' now and you expect to be able to say 'My washing will/may be dry
now' soon-although
if something goes wrong you may end by saying 'My washin would/might have been dry soon/now' under a projective interpretation.
'Had invaded' can occur as the past of 'has invaded'. But it makes no sense to
say 'Hitler has invaded England last month', and in
Hitler had invaded England the previous month.
'had invaded' registers that t is past with respect to something already past.
118
THE JOURNAL OF PHILOSOPHY
clause for short, and it consists of 'if' followed by a subject and a
predicate. The effect, predictably, is to complicate an otherwise
simpler projective message, and specifically to complicate what the
verdict is about. For a conditional interpretation of (2), the verdict is
about Germany's winning the war in a situation at y imagined to
result from Hitler's invading England in an earlier situation at x
imagined as developing out of facts of a yet earlier date D.
As well as the time y, conditionals also have a time x of the ifcondition's satisfaction; and encoded in the clause's predicate we
find another tense t'. It is characteristic of a conditional clause that it
have x later than t'. The relation between x and y, meanwhile, is
forged in the underlying reasoning, and can vary greatly. For two
events, as in (2), x usually precedes y; but not always: 'If I kill myself,
I'll kill Grannie first'.8
Both tenses in a conditional are chosen, I submit, to locate the
same point in the B-series-usually t' = t, as research will discover.
But even when, as sometimes happens, t' diverges from t, each is
chosen so as, in its own way, to locate the same date.9
I can now honor a promise, defining a conditional as any message
8 For the sake of concision,
I stick to events throughout this essay; but an
adequate semantics would cover states as well. In hard cash, the difference is that
there are present states but no present events. Thus, a conditional with two states
can have x and y both present: think of
If Her Majesty was/were here now, she would be furious.
If Her Majesty had been here now, she would have been furious.
Often the underlying reasoning in such a case has x and y identical, whereas the
account in this essay assumes that x < y, and is therefore insufficiently general. In
partial reparation, I propose that the verdict in these latest examples is about Her
Majesty's being furious in a situation at y which (a) is imagined to develop out of
facts of an earlier D, and (b) includes the imagined presence of Her Majesty at y.
9 Very often when t and t' diverge, the conditional combines a present state with
a past event:
If I didn't have this thing firmly in hand I wouldn't have come in here, I
assure you (Jerome Weidman).
A speaker who said 'If I hadn't had this thing firmly in hand . . .' would thereby
permit a past x: in effect, I had this thing firmly in hand when I came in here. But
by making t' simply past, Weidman's speaker makes x present: I still have this thing
firmly in hand. Evidently, this present state has endured since before the past y.
As I remarked, the relation between x and y can vary greatly. But my present
point is just that there is no contradiction in locating the same date as both past
and pastpast.
Sentences like
If Hitler invades England, Germany might win the war.
are frequently found under conditional interpretations. I concede that this fact is
a problem if t and t' are to locate the same D, but lack the space to respond. The
phenomenon is not found with 'will/would'.
ON CONDITIONALS
119
encoded by the encoding program which generates the sentences (2)
from their by now familiar "conditional" interpretations.
v
In a conditional, then, an otherwise simpler projective message is
complicated by having the satisfaction of an if-condition intrude at
some time x in the imagined development that elicits the situation at
y from facts of D. But there are two ways of reasoning about the
if-condition's satisfaction. Dominant reasoning attends just to the
influence of the if-condition's satisfaction upon the situation at y.
Recessive reasoning concedes that influence, perhaps even takes account of it, but it also attends to how the if-condition might come or
have come to be satisfied. Some conditionals demand recessive reasoning: 'If Grannie escapes, it will be through the sewer'; 'If we fail,
it will be because . . .'. But, broadly, both possibilities are there, a
point illustrated by the following experiment of Alan Gibbard.
Sly Pete is cheating and can see his opponent's cards. Knowing
this, and that Sly Pete is desperate to win, but not what is in either
player's hand, how do English speakers react to (4a)?
(4a) If Sly Pete calls, he will win.
Gibbard'0 tried it, and found that some assented and some did not.
The same verdict holds for Sly Pete's winning in the aftermath of
calling; different reactions. But of course: reason recessively that if
Sly Pete calls it will be because he holds a winning hand, and you will
assent; reason dominantly that who will win depends on the unknown hands, and you will withhold assent.
Either way, dominant or recessive reasoning, the verdict is about
Sly Pete's winning in the aftermath of his calling in a situation at x
descended from facts of an earlier D-no difference there. But
there are essential differences in the underlying reasoning. For instance, the satisfaction of the if-condition is thought of as happening
gratuitously in dominant reasoning and as arising out of given realities of D in recessive. Here, then, is another dimension altogether to
the logic of conditionals.
VI
My theory of tense for projectives, including conditionals, is that t
locates D. The core, it will be remembered, is a verdict-about, let
us say, Germany's winning the war in the wake of Hitler's invading
10
He says his "informal polls" on the point "have proved inconclusive": see his
"Two Recent Theories of Conditionals," in Ifs, W. L. Harper et alia, eds. (Dordrecht: Reidel, 1981), p. 228. Regarding the topic of this section, compare part I
of David Lewis, "Counterfactual Dependence and Time's Arrow," Nou2s, XIII
(1979): 455-76.
120
THE JOURNAL OF PHILOSOPHY
England. But how are we to think of the situation as arising, to
which the underlying reasoning is applied? In what circumstances is
Hitler to be thought of as invading England? I propose that the
situation is thought of as arising in a normal course of events developing out of historical facts of the date D. And t locates D. The tense
of a conditional is an A-series location of the date whose facts, when
thus projected, deliver the intended circumstances for the application of the core reasoning.
The dexterity of this formulation will not make me the envy of my
friends. Perhaps I can do better pictorially.
VII
I shall represent time horizontally in the customary way, with left
earlier than right. The firm horizontal line depicts the deposits of
fact that history has so far vouchsafed us. This line gives out at 0,
the present moment. I make a habit of writing A-series times above
the level of the history line and B-series times below it. Also below it
belong identifiers of historical facts. A broken line, meanwhile, represents a normal course of events. When Hitler's invading England
is stipulated by 'if, this is acknowledged by brackets around 'Hitler
INVADE' where it intrudes upon the normal course of events. The
question mark before 'Germany WIN' is a reminder that Germany's
winning is merely what the verdict is about. The verdict itself is given
over on the right: ventured on the basis of these imagined temporal
proceedings, it plays no part in them. For the sake of expedition, I
shall ignore the cases for 'can/could' and 'may/might', commending
them to the reader's investigation. The verdict for 'will/would', simplest verdict of all, is epitomized as 'YES'.11 And now we are ready to
experiment.
1 The lexemes 'will/would', 'can/could', 'may/might', and six others are called
modals, and their meanings, presumed unique, are called modalities. No adequate
account exists of the English modalities. Nevertheless, an adequate theory of
conditionals requires one, for projective messages use modalities for their verdicts.
But one thing seems obvious: that 'will/would"s modality is a degree simpler
than any of the others. Where they all intrude their own strictures of one kind or
another, it intrudes ... nothing. The overall effect when it is used as a verdict is
therefore simply that of endorsement, which is what I am representing as 'yes'perhaps unwisely, since the essential point is that 'will/would' is the null modal.
And if it be asked what the difference is between a null modal and no modal at all,
I respond: listen to the difference between 'My washing will be dry now' and 'My
washing is dry now'.
It is doubtless for this nullity that 'will' is paraded as marker of a "future
tense," a conception according to which 'Germany will win the war' merely says
about the future what 'Germany won the war' says about the past. I urge its
untenability in part I of "Against the Indicative," Australasian Journal of Philosophy, LXXII, 1 (March 1994): 17-26.
121
ON CONDITIONALS
VIII
It is September 1, and we are maintaining (2a) when the news comes
through: Hitler assassinated. (2a) becomes something we no longer
hold, and we adjust by switching to (2c)-still about a future y, of
course. One's first thought is that the new locution concedes that
Hitler is not now going to invade England; but no, for we happily
couple (2c) with
(5a) If Hitler invades England,it will be the first time a country has
ever been invadedby a dead commander.
But here, perhaps, is a clue. It is a dead Hitler who is imagined to
invade England in (5a) and a live one in (2c). A live one in (2a), too,
back when we were saying it, although if (2a) were said now, a dead
one. And speaking now, we can say that Hitler is at present dead but
was alive at a time past with respect to the past assassination.
Figure 1 depicts (2a) said before Hitler's assassination, and figure
2 depicts (2c) and (5a) both said after it:
(2a)
x
y
(Hitler
?Germany
INVADE)
---_-_-WIN
-I-V----
Hte
-.
[YES]
alive
Sept1
Sept7
Say1942
Figure 1
0
x
(Hitler
INVADE)
(2c)
(5a)
(5a)
S_
|
_________________________
y
?Germany
WIN
(Hitler ?first
time
INVADE)
-
Y
[YES]
Hitler
alive
-
Bang!
Hitlerdead
Sept7
Sept1
Say1942
Figure 2
Hitler invades England alive or dead in the imagined enactment
according as he is actually alive or dead at t, we observe, with each
normal course of events preserving a status quo. Or as I put it in
section III, certain facts of D are allowed to "flourish and prosper."
122
THE JOURNAL OF PHILOSOPHY
In some sense, the speaker must have been relying on having
Hitler remain unassassinated when she affirmed (2a), else why the
switch to (2c)? But not as a premise of her underlying reasoning.
Rather, having Hitler remain alive was understood as part of what,
without necessarily realizing, she meant. The underlying reasoning
was never intended to apply without a live Hitler.
It would seem that when a conditional is affirmed about a future
y, the underlying reasoning is bound to insist upon countless such
preservations. And when one of them fails in reality, we single it out
and imagine a rerun of history in which it succeeds, just as was
initially expected. Since the rerun must start from earlier than the
already past failure, we naturally select the (c) form. Thus we cleave
to our underlying reasoning, which presumes a live Hitler or whatever, but at the cost of applying it only to a rerun.
Ix
Said in time of peace, (6a) postulates a sudden war and (6b) does
not:
(6a) If war is declared tomorrow,I shall enlist.
(6b) If war were (to be) declared tomorrow,I would enlist.
(6c) If war had been declared tomorrow,I would have enlisted.
These are matters of observation. The first is easily explained: a war
tomorrow in circumstances developing naturally out of today's
peace is certainly going to be a sudden war. Regarding the second, it
seems that the speaker who chooses the (b) form thereby allows
room for a course of events beginning in the past in which war is
declared tomorrow unsuddenly-indeed, in no particular way. Reflection discerns that (6b) avoids (6a)'s imputation of suddenness
while putting nothing in its place, and the hypothesis suggests itself
that the past t of the (b) form is chosen to favor no particular matter
of fact:
tomorrow
x
y
(DECLARE ?Speaker
war)
ENLIST
(6b)
~~
~~~x
Y
?Speaker
ST _
war). . . . ENLI
I
(DECLARE
(6a)
?2
[YES]
Nothreatit
of war
Figure 3
[YES]
123
ON CONDITIONALS
The circumstances that hold at (6b)'s t are not given as different
from those obtaining at 0. If they were different, there would be a
past time at which they changed, making them by now pastpast.
Locating D as simply past is quite a different thing, a matter just of
surrendering control over the way the attendant circumstances are
to develop.
Sometimes immanent war is averted at the eleventh hour, and the
relieved speaker will switch to (6c), in which war is declared tomorrow in circumstances arising from how things were in the awful days
before the thankfully past eleventh hour:
Q
tomorrow
x
y
(DECLARE ?Speaker
war)
ENLIST
--- - - - - - - - - - - - -- - - - - - - - - - - - - - - - -
(6c)
Threat/
of war
/-
T
11thhour
[YES]
NothreatA
of war
Figure
4
x
Whether she asserts (2a) or (2b), the speaker's underlying reasoning,
the same for them both, is premised on the latest military intelligence. This is obvious: it would be grotesque to suppose that the
sponsor of (2b) might be obliged to make do with information that
was out of date. But how is this reliance on present facts allowable,
when (2b)'s fantasy starts in the past?'2
My answer is that the present facts merit their place in the fantasy
on the understanding that these things would be the same anyway.
They would certainly be the same whether Hitler invaded England
or not, but that does not amount to much, seeing x comes after the
bothersome facts. Rather, the facts of D that sufficed to produce
them in reality can be relied upon to produce them in the fantasy as
well. When the military facts of 0 show up at 0 in the fantasy for
(2b), it is because they are thought of as descended in exactly the
same way from the same facts of D down either path. An important
division is emerging. The core reasoning makes essential use of
12 Cf. David Lewis on "factual premises": see Conditionals (Cambridge: Blackwell, 1973), p. 68.
124
THE JOURNAL OF PHILOSOPHY
things that "would be the same," while the mystery agent is off
arranging circumstances for the core reasoning to be unleashed on.
XI
Actually, Hitler was never assassinated. When it came to invading
England, what really happened was that he simply missed his opportunity. After September 15, it was too late, and we who a fortnight
since had maintained (2a) or (2b) found ourselves obliged to switch
to (2c). First, why had we to abandon (2a) and (2b)? Because a
premise of our underlying reasoning had been shot from under us
on September 15: things that "would be the same" were the same
no longer. Secondly, why could we still maintain (2c)? Because its
fantasy revises history. Contrary to fact, it has Hitler forestalling the
reverse of September 15 by invading England on September 7, permitting exactly the underlying reasoning that sustained (2a) or (2b)
back on September 1. Figure 5 shows all three sentences (2) said at
the same 0 later than September 15:
x
(Hitler
INVADE)
9
\
|
*
?Germany
WIN
~~~Conditions
favor
Germany
Germany
x
(Hitler
INVADE)
(2
'b)
*
x
I
(Hitler
INVADE)
.
? ConditionsfavorGermany
*1g
*7
y
?Germany
WIN
[YES]
disfavor
Conditions
Germany
British
airvictory
?
Y
?Germany
WiN
W
[YES]
-
disfavor
Conditions
Germany
'
I
(2a)
0Y
15 September
Say1942
Figure 5
The italics in figure 5 identify new elements in our diagrams:
matters of fact that figure in the underlying reasoning. Under the
history line we have them as facts. But they show up in the fantasy as
well, on the ground that, in their regard, things in the fantasy would
have fallen out exactly as they actually did.
XII
There is another motive for rerunning history besides wanting to
revise it, namely, to retrace it: an i]Lcondition will be imagined satisfied at a present or past x in an effort to find out whether it actually
is or was:
125
ON CONDITIONALS
If the fugitive had fallen from the roof, he would have left an indentation in the rose garden; but as you see, the rose garden is undisturbed ...
If the fugitive had fallen from the roof, he would have left just such an
indentation in the rose garden as we now observe . . .
0
yI
?Indentation
x
(C)
(Fugitive
FALL)
in9arden
[YES]
Evidence
Fugitive
fell?
Figure 6
XIII
Perhaps this is the moment to explain how the mystery agent
achieves its effects in our opening example. I propose that the tense
moves to ensure that Spurs and Liverpool are the legal finalists
when the match is played, having defeated all previous opponents or
whatever is required. Indeed, a proponent of (1) would likely claim
that having Spurs and Liverpool as legal finalists was part of what
she meant when she spoke of a Spurs-Liverpool final, or that that
was understood.
Once Spurs and Liverpool have each won a place in the final, the
fantasy can project that present fact so that the match is played
between legitimate finalists, and we say (la):
y
?Fitting
0
(la)
climax [YES]
SandLbothA
legalfinalists
Figure 7
As a matter of observation, a speaker who says (1a), given that it is
not yet determined who the finalists are to be, thereby registers her
conviction that Spurs and Liverpool will both win their ways to the
final. Why is this? Because if Spurs and Liverpool, though not yet
established as finalists, are both to be legitimate finalists in circumstances arising from how things are now, this requires that they both
become legitimate finalists in those same circumstances, justifying
'will'. But before it is established whether Spurs and Liverpool are
to be the finalists, the uncontentious utterance is (lb), which per-
126
THE JOURNAL OF PHILOSOPHY
mits both teams to find their unhurried ways to the final from whatever beginnings:
0
y
?Fitting
(1b)
-----------------------------------
climax
_[YES]
So
so good
farA
Figure 8
Finally, once Spurs have been put out of contention, the fantasy
has to start from a pastpast stage of the competition when Spurs
were still in contention, and we say (1c):
0
y
- - - - - ?- - - - - . -
(1c)
SofarA
so good
7
?Fitting
climax [YES]
SandLnot ""
Spurs bothlegalfinalists
out
Figure 9
There are swags of examples like the Spurs-Liverpool one. Take
the following, for example, all said about a future y.
(7a) This play will infuriate Her Majesty.
(7b) This play would infuriate Her Majesty.
(7c) This play would have infuriated Her Majesty.
Doubtless the core reasoning is premised on aspects of the play and
propensities of Her Majesty. But the verdict is understood to be
about Her Majesty's being infuriated by the play upon encountering
it, and it is Her Majesty's confrontation with the play that originates
from t.
The proponent of (7a) is understood to expect Her Majesty to see
the play. But of course: (7a)'s verdict is about her being infuriated
by seeing the play in circumstances arising from how things are now,
which requires her to see the play in circumstances arising from how
things are now. Still, it is important to appreciate in what expecta-
127
ON CONDITIONALS
tion of something consists, namely, in seeing it as arising out how
things are now-not in believing some proposition about the future.
XIV
On September 1, 1940 (to refresh the reader's memory) we were
asserting some sentence (2) under a conditional interpretation, with
September 7 in mind for x. Selecting just two from among the
countless different developments that might have awaited us at that
0, I shall describe how we would have reacted in each, and urge that
my account explains that reaction. For the aim of this penultimate
section is to show my account of conditionals at work.
First, Germany's military strength, overwhelming on September
1, might for some reason have abated, and so rapidly that by September 6, invading England on the morrow would have seemed the
height of folly. This development would have obliged us to recant.
We were wrong back on September 1, because our then underlying
reasoning relied on an expectation that was unfulfilled, namely, of
Germany's continued military superiority. From an 0 on September
6, we would reject every variant of (2). It is futile selecting (2c) in the
hope of starting the fantasy from a pastpast D before Germany's
military decline, because Germany's military decline is one of the
things that would have happened anyway, just as Germany's then
superiority was back on September 1. Not for the first time, a premise of earlier underlying reasoning has been shot from under us.
0
(2bC)
(2b,c)
_
__ __ __ __
Conditions
favorable
*
(2a)--
x
(Hitler
__________________________INVADE)
Conditions
favorable
y
(Hitler ?Germany
_ _. ,_ __INVADE) __WIN _"YS
)
Conditions
unfavorable
_
Conditions
unfavorable
1
September
6
?Germany
[Y S
- - - - WIN
- [Y
7
Figure 10
Secondly, Hitler might have been assassinated, say late on September 1, and Germany's military strength so rapidly have abated
that by September 6 a German invasion of England on the morrow
was strongly contra-indicated. We are restricted to the (c) form anyway, because we want a live Hitler in the fantasy. But do we still
maintain (2c) on September 6? The answer, a matter of observation,
128
THE JOURNAL OF PHILOSOPHY
is "yes" if we think that the German decline resulted from Hitler's
assassination and "no" if we think it would have happened anyway:
x
(Hitler
INV,DE)
(2c)
[YES]
~~0
I
Conditions
favorable
y
?Germany
WN _
Bang
Conditions
unfavorable
Hitler
alive
1
6
7
Figure 11
In my design, it depends whether we think of the adverse conditions
of September 6 as originating at bang or at t. If only at bang, then
the German military decline is local: in the fantasy, Hitler invades
England in the same conditions as when we were wont to say (2a) or
(2b) before the assassination. But if the adverse conditions on September 6 are causal descendants of just facts of D, they will show up
in the fantasy as well.
xv
I have distinguished two semantic themes in conditionals. Sustaining
the core is the underlying reasoning, premised on things that
"would be the same." Fixing the tense are the circumstances in
which the underlying reasoning is applied. Their connection is that
the verdict is about something happening at y in imagined circumstances got by projecting facts of t in a normal course of events.
I have said little about the core and less about its underlying
reasoning. For expounding the core, my sorest necessities are meanings for the lexemes 'will/would', 'can/could', and 'may/might', together with an account of how these meanings are used as verdicts.
Regarding the underlying reasoning, I principally contend that it is
meant (not to establish a proposition but) to justify a verdict. I
repent having elsewhere referred to it as the underlying argument,
because 'argument' suggests premises and a conclusion that is like
them, whereas the thinking behind the affirmation of a projective
message moves from factual premises to a verdict, something else
completely. With any account of the underlying reasoning waiting
upon the sorely needed account of verdicts, I am a long way from a
"semantics" for conditionals.
V. H. DUDMAN
Macquarie University