<![CDATA[peter的世界]]> Fri, 01 Aug 2008 06:44:10 CST zh-tw <![CDATA[Peter in Seoul]]>
Dongdamum

Chicken soup is very very very good...

Look spicy, but taste just a little spicy...

7/30 evening, the conference held a reception for all participants.

almost sun set...

so many people...

A Big Show.....


Seoul is an expansive city, for i spent about 30000 won (1000 NT) per day.
And i believe i will spend more in the following days.....

]]> Fri, 01 Aug 2008 06:44:10 CST <![CDATA[桔醬的滋味=媽媽的味道]]> 昨晚看到公視的節目:桔醬的滋味

覺得蠻不錯的,有興趣的可以去找來看.

(或是注意重播的時間)

 

一直覺得公視有不少好節目...

 

"人生有很多種選擇...

沒有是非對錯,只有能否承擔..."

]]> Mon, 30 Jun 2008 05:55:20 CST <![CDATA[君子不器]]> 比喻君子體用兼備，不只一才一藝而已。

出自論語,為政：子曰：君子不器。

朱熹注：器者，各適其用而不能相通。成德之士，體無不具，故用無不周，非特為一才一藝而已。

 

"君子不器",' 當初知道這個成語是件偶然的事,

記得高中時,

本人雖然不是很優秀的學生,但在課堂上卻很會點頭,

於是國文老師便誤以為我很用功....(真不好意思),

要我每天在黑板上寫上一則成語,讓全班同學多認識一些成語.

剛開始,本人覺得這是件苦差,所以隨便弄幾則成語渡日...

一開始,國文老師也不說什麼,到了某天,

他有點嚴厲的說,

"這幾天我寫的成語太過通俗,根本不能讓同學們多學一點"

老實說,當時真的覺得很羞愧...

所以當天回去就查了成語字典,

無意間看到"君子不器",隔天便寫上去.

此後國文老師便沒說什麼了.

(因為我每天不僅讓自己學到新成語,也真的讓同學們學到新成語)

此後我也朝著"不器"的方向前進,盡量讓自己多點才能.

 

過了很久,我心中一直記得這位老師,儘管他對我們班上的同學很感冒...

前幾年,我去面試一份大學兼任講師的職位,

心裡很緊張,到了主任面前,

因為主任是位老教授,很和氣,也沒架子,

心裡才放鬆了點,在面談時,

還算順利,直到最後一刻,那主任說,

"現在的學生/老師都很不成器,

出了問題只會怪別人,不去問自己是否有錯,

以現今的教育來看,很多老師在教書時,

當學生表現差,就只會說學生品質或素質爛,

也不檢討自己是否有改進或改變教學方式,

用更適合學生的方式來教導...

(該主任是國內知名的歷史學者,這段非常有道理的訓話,

在當時我是充耳不聞的,因為太緊張了,

不過日後這段訊後成為我教書的準則.)

最後...主任考我一個成語,那就是"君子不器".

有點跟之前的訓話相關,

幸運的,因為高中的經歷,讓我對"君子不器"印象很深,

所以我就說出了正確的答案.

當然也順利取得教職.

 

最近在趕論文,

而我研究的最後是能夠做到跨領域的結合,

所以我會涉獵很多領域的書籍與論文,

這讓我覺得很充實,所以....

期許每個人都"君子不器"!

]]> Sun, 25 May 2008 11:18:48 CST <![CDATA[走路....有助思考....]]> 最近在忙論文,

不過腦子可不是一直都很清醒,

偶爾也需要發呆...

 

無論如何,發現走路還真是有幫助,

邊走邊想,

雖然回家後可能都忘了,不過這樣的感覺還不錯.

因為在寫論文時,可能會一起想起來...

 

總之,衝衝衝!

希望今年可以畢業.

]]> Wed, 21 May 2008 22:23:04 CST <![CDATA[區分條件句part 2]]> 4.5 Edgington’s Y-shaped analysis

Before introducing Edgington’s analysis, let me shortly introduce Ellis’s in that their analyses are similar.  According to Ellis, a conditional should be explained in terms of our degree of belief in it.  Basically, we can replace the selection function s in Stalnaker’s semantics with a chosen basis for reasoning from the antecedent supposition, which is probably signaled by tense or mood[1].  However, Ellis’s meta-theory is unlike Stalnaker’s at all.  For Ellis, what he offers is not truth condition but acceptability condition, which is represented by probability[2].  Nevertheless, I think Ellis will agree with Davis ’s analysis, if the concept of truth is replaced by the concept of probability and the concept of possible world is replaced by the concept of belief systems[3].  Except this similarity, Ellis claimed that acceptable subjunctives are reflections of their earlier corresponding indicatives.  To explain his claim, let’s take Edgington’s OK pair, ( 6a ) and (6b):

 

( 6a ) If you go in, you will get hurt;

(6b) If you had gone in, you would have gotten hurt.

 

What Ellis claimed is that if (6b) is acceptable, ( 6a ) asserted earlier was acceptable, which accords with our intuition.  Thus, indicatives and subjunctives say the same thing, but concern different times[4].  Bennett precisely summarized Ellis’s and Edgington’s analyses in his book and called them ‘The Correspondence Thesis’ (CT, for short)[5]:

 

CT: For any A and C, if A>C is the right thing to think at a certain time, then at some earlier time A®C was the right thing to think.

 

Here we can treat ‘the right thing to think’ as high probability.  Now, let’s check ( 1a ) and (1b) under CT.

    Before the assassination, it was very unlikely to believe that someone else would kill Kennedy (Y), for he was so respected and supported at that time.  On the contrary, it was very likely to believe that no one would kill him (~Y).  Thus, for people who are familiar with American history around 1963, the probability of Y given X was low, but the probability of ~Y given X was high.  That is to say, ‘If Oswald doesn’t kill Kennedy, no one else will’ was acceptable before the assassination, which is plausible[6].  However, whether conditionals have truth conditions or not, it seems reasonable to think that meaning/content will change with truth conditions or acceptance conditions.  It follows that ( 1a ) and (1b) say different things given one is true/acceptable and the other isn’t.  For Ellis and Edgington, their analyses avoid this obstacle, because, as mentioned earlier, indicatives and subjunctives say the same thing, but concerns different times.  In my opinion, the advantage of using probability instead of truth-value shows up here, for it is convincing that probability or conditional probability will change with time.  Therefore, the difference between indicatives and subjunctives is merely a matter of time.

 

5. Some Remarks

Before presenting my own view (Chapter 2), I would like to point out some flaws in the analyses introduced earlier, for ‘If I have seen further it is by standing on the shoulders of giants’[7].  To reveal their flaws, I will introduce some concepts or distinctions in advance, for these concepts or distinctions are helpful when we want to see their flaws clearer. 

 

5.1 Re-thinking OK phenomena and backward conditionals

People agree that there are some similarity and differences between indicatives and subjunctives, though they may have different explanations for them.  In the following, I will roughly present their similarity and difference, before we step further.

 

(A)  It is a fact that there is some kind of grammatical connection between indicatives and subjunctives.

(B)  It is a fact that people have divergent attitudes toward some OK pairs.

 

Obviously, a theory of conditionals should explain not only these two facts, which concern OK phenomena, but also kinds of conditionals.  The conditionals, pair (1) to (8), mentioned before have something in common; that is, they are forward conditionals or simultaneous ones.  By ‘forward conditionals’ I mean that the antecedent is temporally prior to the consequent, and by ‘simultaneous conditionals’ I mean the antecedent and consequent are synchronous.  Intuitively, except pair (4) and (8) are simultaneous, the rest are forward, for ‘will’ or ‘would’ in the consequent indicate their temporal order.  However, there is another kind, called ‘backward conditionals’, which is opposite to forward conditionals; that is, the antecedent is temporally posterior to the consequent.  For example,

 

(9) If the plane had arrived at 2:00 p.m., it would have left at noon[8].

 

Therefore, one way to detect the flaw of an analysis is to check whether it can give a satisfactory explanation for (A), (B) and backward conditionals, like (9).

 

5.2 Goodman’s insight: ACE framework

In his ‘The Problem of Counterfactual Conditionals’, Goodman pointed out two major problems concerning counterfactual conditionals: first, what is the assumed facts in the antecedent[9] and, second, what is the connected laws between the antecedent and consequent[10].  That is to say, through the antecedent (simply, A), the consequent (simply, C), some assumed facts in A, and the connected laws between A and C, we can evaluate or explain conditionals[11].  As to the laws, I think the context will contain enough information in determining which law we should take in order to explain a conditional.  In most contexts, for example, the default will include logical laws.  In talking of physics, it is natural to include physical laws.  Thus, for argument’s sake, let’s focus on only the facts[12].

    Based on Bennett, I think ‘evidence’ is a better term for the assumed facts, for in explaining conditionals what we appeal to is our evidence, especially in explaining some subjunctives.  Take ( 1a ) and (1b): even though it is a fact that Oswald killed Kennedy in 1963, we can still appeal to some evidence which may result in false beliefs in explaining them.  Some evidence may mislead people into that someone else killed Kennedy in 1963[13], and consequently make people think ( 1a ) as true or acceptable.  Now, it seems that in evaluating some conditionals we can use ACE framework; that is, as Goodman suggested, whether C can be derived from A and E (and the implicit laws)[14].  Except this, according to Bennett, for some conditionals, it is A that can be derived from E and C, and for others E is the explanandum.  The former can be instantiated by backward conditionals and the latter can be illustrated by (10):

 

(10) If those are desert verbena, then this fire is many days old’[15].

 

Let’s consider it further.

    At the beginning, suppose we are in a place where there are cold ashes of what has been an enormous fire, and you say, pointing to some growing plant, ‘if those are desert verbena, then this fire is many days old’.  At first, the best explanation of the coldness ashes should be a recently heavy rain; however, as a plant expert, you notice the plant nearby and recognize it as desert verbena.  As you know, desert verbena will bloom once rain falls on them, and the plants there have not done yet.  So a recently heavy rain is not open to our consideration, and that explains why you say so.

    Some may fail to see the difference and significance among type A, C and E explanation, for, seemingly, there is no difference between explaining C and E in that A (those are desert verbena) and E (desert verbena will bloom once rain falls on them and the plants there have not done yet) can explain C in type E explanation.  In my opinion, due to the complexity of our natural language, I do think that there exist kinds of types of explanations; however, the problem is whether it is significant to make such distinction.  For Bennett, of course, it is significant to see their differences, for he argues that some theory, like Davis ’s, fails to explain type E[16], which we will see later.  Thus, there do exist significant differences among them.

    Let me make some criterion for distinguish them and leave the significance in the following section.  First of all, I think it is context that makes the distinction, for we can explain a conditional within ACE framework (and implicit P).  Second, I assume that type C is the paradigm in the sense that people understand a conditional in the basis of type C.  Now, suppose that there is some difference among type A, C and E, and it follows that we are in a position to discover their differences.  As mentioned earlier, context plays an important role in determining their type, but I think there will be some linguistic device, by which we can see their differences.  To see their differences, we should look at our paradigm, type C, carefully.  In light of Goodman, we can say a conditional ‘if A, C’ is true, iff, C can be derived from A and E, which is shared by type A, C and E.  However, the intended target, the focus, will change with context.  In normal context, what is interested is the consequent C, for we wonder whether, under such and such conditions, C will follow.  In context like (9)[17], its backwardness marked by the temporal order between A and C indicates that A is the intended target.  Therefore, in normal context, the backwardness is what makes A and C different.  As to context like (10), since the best option for C is ruled out, the context involved is abnormal and subjective.  In light of Stalnaker, in normal context, the speaker and audience will share E (the context set), but the intended target may be A or C, for C is supposed, at least on speaker’s view, to be the best option.  Back to (10), if the scenario is normal (the speaker and audience share the same evidence), (10) will not be asserted unless the speaker has additional evidence.  In such context, the audience may puzzle about (10) until E is showed up.  Thus, the absent best option indicates that E is the intended target.

    In short, if C is not the best option, the conditional is type E; if C is the best option, the conditional is either type A or C, which can be tell by the backwardness.

 

5.3 The flaws in Stalnaker’s analysis

Here I would like to address the problem concerning Stalnaker’s Y-shaped analysis only, and leave other objections to Chapter 2 while introducing his theory of conditionals.  Though they are connected, let’s focus on whether his analysis nicely explains two facts, (A) and (B), and backward conditionals.

    First of all, let’s see how Stalnaker explains ( 1a ) and (1b).  Recall his analysis in the following:

 

‘If A, C’ is true in i, if and only if, C is true in s(A, i)

 

The selection function s in ( 1a ) will pick out the base world, i.e. the actual world, itself, for it is possible[18] that Oswald is not the one who killed Kennedy in 1963 in the actual world.  However, it is a fact that Kennedy was killed in 1963, so the best explanation of it is that someone else killed him.  Thus, C is true in s(A, i) and ( 1a ) is true. 

On the other hand, which world will s in (1b) pick out?  First, the subjunctive mood indicates that the antecedent is hypothetical (at least for the speaker), and under this hypothetical situation the speaker and audience may consider something outside the context set, which means s wouldn’t pick out the actual world.  According to Stalnaker, context will contain some information about how to compare possible worlds and pick out the most similar one.  However, it seems that we may have good reason to pick out not only a world where someone else killed Kennedy but a world where no one else killed him.  For one thing, given the situation where Kennedy was so respect and support around 60s, a world where Oswald didn’t kill Kennedy may be a world where no one killed Kennedy.  Therefore, (1b) should be rendered false.  For another thing, given the situation where some people questioned about the Warren Commission Report, a world where Oswald didn’t kill Kennedy is a world where someone else killed Kennedy.  Thus, (1b) should be true.  However, by his fourth restriction on the selection function, given A is possible, either ‘if A, C’ is true or ‘if A, ~C’ is true, but not both.  Now, Stalnaker seems to face a problem.  In my opinion, Stalnaker may reply that it is context that makes the difference; given the same context, the selection function will pick out one and only one the most similar world.  Though I agree with Stalnaker, I think we can bring the influence of tenses into the analysis of conditionals, as Davis[19] and Thomason and Gupta[20] suggested.  Putting everything unexplainable or unclear into the realm of pragmatics is the last option, so let’s try to supplement Stalanker by bringing tense into the whole framework.

 

5.4 The flaws in Davis ’s analysis

With little but important supplement, we can explain (1b) in a satisfactory way.  That is to say, the selection function s should pick out the world where everything is exactly the same with the actual world before Kennedy was killed.  Hence, (1b) should be false. So far so good.  Next, let’s see how to evaluate backward conditionals.

Take (9) first, and it says that ‘If the plane had arrived at 2:00 p.m., it would have left at noon’.  Given that the flight it takes from departing to arriving is about 2 hours, (9) should be rendered true or accepted.  Now, let’s see whether Davis ’s analysis give the right predication.  Which world will s pick out when considering ‘the plane had arrived at 2:00 p.m.’ (and taking the actual world as its base)?  Suppose that the plane arrived at 3:00 p.m. in the actual world, and we fix everything unchanged before 3:00 p.m.  It seems that a world where the plane arrived at 2:00 p.m. is a world where it departed at noon.  Thus, it seems that Davis makes the right predication[21].  Next, let’s see Jackson ’s example[22]:

 

(11) If I have misremembered to the date, the Battle of Hastings was not fought in 1066.

 

According to Jackson , (11) should be rendered accepted[23], for given the date I remembered is incorrect the Battle of Hastings would not happen in that date.  However, on Davis ’s analysis, s may fail to pick out the intended world, for s may pick out a world where I remembered the date as 1067 or 1065 and everything is fixed before the moment of my misremembering.  So it follows that s picks out a world where the Battle of Hastings did happen in 1066 in that everything is fixed unchanged.  Thus, Jackson concluded, Davis fails to make the right predication[24].

    Is Jackson ’s conclusion true?  According to Bennett, though Davis ’s analysis has some problem, Jackson ’s objection is not fatal.  Let’s change Jackson ’s example into the third person version:

 

(12) If Henry has misremembered to the date, the Battle of Hastings was not fought in 1066.

 

Should (12) be accepted?  No, it shouldn’t, for I am certain that the Battle of Hastings was fought in 1066.  Based on my confidence in the date, I may think that Henry’s mistake is to remember the date not as 1066.  Thus, (12) should be unaccepted.  The reason why Jackson ’s first person version is accepted is the result of self-reflection.  As mentioned before, an independent conditional is a conditional such that the truth of its antecedent can ensure the truth of its consequent.  In one’s self-reflection, since he or she is aware of the correct date--otherwise, he or she won’t claim that ‘I have misremembering…’--, and doubts whether the date he or she remembered is correct, the consequent will follow.  Let me use another example to explain[25]:

 

( 13a ) If I have misremembered my birthday, I wouldn’t have been born in 1977.

(13b) If Peter has misremembered his birthday, he wouldn’t have been born in 1977.

 

When I consider ( 13a ), my reflection is that ‘the date (1977) I remembered may be incorrect, and it follows that I was not born in that date.’  The point is that the date in both antecedent and consequent will be the same, and it follows that the antecedent (mistake in remembering the date) will guarantee the consequent (the date is incorrect).  On the other hand, since I know that Peter was born in 1977, my thought of (13b) is that ‘Peter may remember his birthday as 1976, but he was born in 1977’ .  Therefore, ( 13a ) is accepted, but (13b) isn’t.  Thus, Jackson ’s objection fails.

    Although Davis avoids Jackson ’s attack, Bennett still thinks there is some flaw in Davis ’s analysis.  As mentioned in 5.2, type E is the conditionals that C isn’t the best option in explaining.  Suppose that its antecedent is false[26] and what we intend to explain is that desert verbena will bloom once rain falls on them (EV, for short).  Let’s consider (10) and recall the scenario:

 

At the beginning, suppose we are in a place where there are cold ashes of what has been an enormous fire, and you say, pointing to some growing plant, (10) ‘if those are desert verbena, then this fire is many days old’.  At first, the best explanation of the coldness ashes should be a recently heavy rain; however, as a plant expert, you notice the plant nearby and recognize it as desert verbena.  As you know, desert verbena will bloom once rain falls on them, and the plants there have not done yet.  So a recently heavy rain is not open to our consideration, and that explains why you say so.

 

Let me supplement this scenario a little:

 

Suppose Peter knows everything you know and something you don’t; that is, the plant you’re pointing is not desert verbena.  However, Peter believes the fire is many days old, because the locals told him so.  Paul knows everything Peter knows and something he doesn’t; that is, the fire is not many days old.

 

Now, since the antecedent is false, either (~A or C) or (~A or ~C) can equally explain EV.  It follows that whether C is true in s(A, i) or not, EV is explained.  In other words, Peter may reasonably think ‘if A, C’, while Paul equally thinks ‘if A, ~C’.  I think Davis will agree with Paul, because, when we fix everything unchanged before your pointing, C is false in s(A, i)[27].  According to Bennett, Davis is wrong in evaluating indicatives objectively, for when considering type E subjectivity is brought in.  It follows that Peter and Paul are equally reasonably in thinking of (10), and it doesn’t matter whether (10) is objectively false[28].

    Frankly speaking, I don’t see why this is a flaw, for there is a difference between what is true and what is accepted.  It means that given a truth or falsity we, a creature with powerful mental capacity, can accept it or not, especially when we’re uncertain of it.  The point is that Paul, Davis ’s voice, does the job.  For another thing, with the supplement of tense, Stalnaker’s context set can do the rest.  Besides, Bennett’s argument seems to base on two contexts, represented by Peter and Paul, and of course we will have different result.  In my opinion, on the contrary, Davis should win the competition, for he keeps the track on one context.  Nevertheless, I think Davis ’s analysis will have a shortcoming; that is, lack of flexibility in dealing with deterministic past.

    The picture Davis and some possible-worlds theorists (who entertain Y-shaped analysis) draw is deterministic about past, but we should be in a position to talk of past in another way.  Suppose that, though it is a fact that Oswald killed Kennedy, the reason why Oswald did that is his believing that Kennedy would cause nuclear war, which would surely result in very serious damage around the world.  Under a picture of determinists, the possible world mentioned above would have an extremely different past with the actual world, though both contain the event that Oswald killed Kennedy.  In evaluating conditionals, I don’t see why this is impossible in principle, for I have been thinking ‘if human had chosen magic rather than science, what would have happened now?[29]’.  Thus, Davis ’s analysis seems to restrict ourselves within a deterministic past, which, I think, is a defect.

   

5.5 The flaws in Edgington’s analysis

In light of Adams , Edgington gave up truth condition in evaluating conditionals, and attempted to replace it by probability.  However, I think her analysis is bounded, though she claims that her theory can explain more[30].  Let’s recall CT and focus on it, for where there is a flaw in CT there is a flaw in Edgington’s analysis.

 

CT: For any A and C, if A>C is the right thing to think at a certain time, then at some earlier time A®C was the right thing to think

 

First of all, I think Edgington should explain what ‘the right thing to think’ means, for we can treat it as truth-like.  Of course, she has an answer to it; that is, acceptability conditions, which can be represented by probability[31].  Well, is the problem solved?  I don’t think so, for we can understand probability as ‘likely to be true’.  Maybe she can argue the notion of acceptability condition (or ‘probability’ in her sense) is primitive and it is inappropriate to ask for a further explanation.  If so, I wonder how she deals with independent conditionals, which are plainly true.  Unless she wants an ad hoc theory like Bennett’s[32], I think independent conditionals will be a problem for her theory.

    As mentioned in 3.3, IST indicates something opposite to CT.  IST shows that subjunctives are based on indicatives, and that is how we learn indicatives at the beginning.  CT shows that subjunctives seem prior to indicatives.  In my opinion, CT has a serious flaw; that is, there are some subjunctives which don’t have their corresponding indicatives.  For example:

 

(14) If I were you, I would pay the bill.

( 15a ) If I were Superman, I would be able to fly.

(15b) If Superman were I, he wouldn’t be able to fly[33].

 

These examples show some subjunctives don’t have their corresponding indicatives and confirm, though indirectly, IST is plausible[34].  If this is a fact, it follows that the application of CT, if true, is bounded.  In my opinion, this is huge, because CT is supposed to be a general statement about conditionals; it implies that given what we know about subjunctives we will know indicatives.  To see this, let me rephrase CT as following:

 

CT’: For any A and C, if we know the acceptability conditions of A>C at a certain time, then we know the acceptability conditions of A®C at some earlier time.

 

Though CT shows some intimate connection between indicatives and subjunctives, it’s bounded, which is a serious flaw.

 

5.6 A real unified analysis

On the one hand, Y-shaped analysis shows that there is an intimate connection between indicatives and subjunctives, and on the other hand, so-called Direct Argument (DA, for short)[35] shows that there is a close relation between disjunctions and indicatives.  Thus, according to Jackson , a real unified theory is to uniformly explain the relation among them: disjunctions, indicatives and subjunctives[36].

 

5.6.1 Direct Argument

Since we have seen how Y-shaped analysis explains the relation between indicatives and subjunctives, in order to have a unified theory we should look at DA now, for DA suggests there is also a close relation between indicatives and disjunctions.  DA is a kind of inference, from disjunctions to indicatives, and seems plausible.  Basically, we can explain its plausibility by claiming either disjunctions and indicatives have the same meaning (truth conditions) or DA is a different kind of reasoning, which yields its plausibility from context.  The former strategy is endorsed by Grice, Jackson and Lewis[37], whereas the latter is by Stalnaker[38].  Here is DA:

 

‘Either the butler or the gardener did it.  Therefore, if the butler didn’t do it, the gardener did.’[39]

 

This inference seems plausible, given a context where we assume the disjunction is true.  However, as Bennett’s example shows, given a context where the assumption is not accepted, we may disagree with the conditional[40].  For the sake of argument, let’s agree with the plausibility of DA, and consequently a unified theory of conditionals should explain the plausibility of DA.

 

5.6.2 Statement and Judgment

Except describing the world, as a statement, we can use language to do some speech acts, such as promise, command, judgment, and so on[41].  For example, when someone says ‘There are some beers in refrigerator,’ he can use this statement to invite his thirsty guest to have some.  Obviously, the speaker does something by saying that statement, not to mention those with speech-act markers, such as ‘promise’.  Thus, statements can play dual roles in communication; describing the world and doing some speech act.

    There are kinds of speech-act markers and my focus is mainly on statements, so I won’t spend too much space on discussing them.  However, I would like to discuss the functions of ‘will’ and ‘would’ in statements.  I think that ‘will’ and ‘would’ are like speech-act markers, for they suggest the speaker’s intention.  For example, when Peter told to Paul that ‘I will bring you a bike’, it seems that Peter intends to do something in the future.  If he really brings a bike to Paul, the statement truly describes the world; if not, falsely.  On the other hand, Peter’s intention suggests something like promise, though he doesn’t express it explicitly.  ‘Would’ has similar function, when it is merely a past tense of ‘will’.  When it functions as a subjunctive marker, it looks like making a judgment.  For example, when someone says ‘if I won the Pools, I would donate part of it to people in need,’ he is making a judgment that given a situation obtained he will do such and such.  Now, let’s back to conditionals.  How about conditionals?  Can they be used to make both a statement and judgment?  In my opinion, the answer is positive, for many people agree that we can talk of the acceptability conditions of conditionals.  Anyway, let me postpone this to Chapter 2.

 

5.6.3 An open alternative

Based on Jackson , I think disjunctions and indicatives have the same truth conditions, and I also agree with Y-shaped analyses that indicatives and subjunctives have similar acceptability conditions.  Now, I think a real unified theory is possible, only if either subjunctives and indicatives have the same truth conditions or disjunctions and indicatives have similar acceptability conditions.  For many people, a real unified theory is impossible, for the options are not open.  Nevertheless, I will argue for the former option and try to develop a real unified theory of conditionals in the end.

Thu, 08 May 2008 07:33:21 CST <![CDATA[區分條件句]]>

Classifying Conditionals

    ‘If’ is a loaded word, by which various sentences are formed, and consequently makes philosophers and logicians want to know its semantics.  To avoid unnecessary complexity, we need to classify the data concerning ‘if’ before offering an analysis.  Though the analysis of ‘if’ will connect with the way we classify conditionals, it is legitimate to separate these two issues, for we can’t analyze anything unless we know roughly what and where our target is.  Therefore, we have to find out what kinds of sentences are conditionals and whether there is only one kind of ‘if’ or not, and then focus on our target. 

    In short, this chapter is mainly devoted to the problem concerning OK phenomena.  To explain OK pairs, there are many kinds of classifications of conditionals in literature, but unfortunately it seems all of them fail to offer a satisfactory explanation.  Nevertheless, the main purpose of my dissertation is to offer a better analysis of conditionals, so I won’t spend too much space on the issue of OK pairs.

There are five sections in this chapter: ‘Conditionals’, ‘Dependent Conditionals’, ‘OK Phenomena’, ‘Kinds of Classifications’, and ‘Some Remarks’.  In the first section, I will roughly point out what kinds of sentences are conditionals, and then distinguish dependent/contingent conditionals, which are our target, from independent/ necessary ones in the second section.  After knowing our first target, I will discuss a debate about OK pairs, which show that there might be two semantically different kinds of ‘ifs’, in the third section.  In the literature, there are kinds of classifications, though unsatisfactory, some of which will be introduced in the forth section.  Finally I will present some remarks on this issue.

 

 

1. Conditionals: What kinds of sentences are conditionals?

There are kinds of sentences in natural language, and some of them play a crucial role in our daily reasoning, which interest logicians.  Intuitively, conditionals are those containing ‘if’, though the difficulty to give a precise definition of conditionals is far from imagination. 

Let me first list several examples[1], which are not so-called conditionals, though these examples contain ‘if’:

 

In the days that followed, Nigel kicked himself for not untying the bunch of flowers and looking to see if there was a card inside.

In contrast to the outside, the area was softly carpeted, softly lit, as if illness and death had to be cushioned away, made to look as if they didn’t exist.

And she also reminded herself that even if it had included a kiss it didn’t mean a thing -- especially as it had come from a man who was living in the outback to get away from women.

Licence Revoked was originally going to be the title for this last Bond film, and if only they had.

 

The ‘if’ in the first sentence can be replaced by ‘whether’, and obviously does not function as a connective.  As to those in the second, third and forth, they will occur with another word and together form an idiom, and some don’t function as a connective.  Still, there are other kinds of sentence containing ‘if’, but we won’t see them as conditionals.  The reason why they are not conditionals is that: first, ‘if’ is not part of an idiom and, second, they don’t have a ternary structure O(A, C), where ‘O’ represents ‘if’ functioning as a connective, and ‘A’ and ‘C’ stand for the antecedent and the consequent respectively.  However, sometimes even though their structure is ternary, we won’t think them as conditionals.  For example, ‘She was always home before midnight: if she missed the bus, then she would walk’, in which ‘if’ can be replaced by ‘whenever’[2].  Therefore, the criterion above doesn’t constitute a sufficient condition for a sentence to be a conditional.  Unfortunately, it seems that we can’t offer a precise criterion for distinguishing conditionals from other kinds of sentences, because of the complexity of our natural language[3].  Nevertheless, it doesn’t follow that we can’t go any further, for we can stick to some paradigms and start our analysis, which is a legitimate strategy.  Thus, let’s suppose we know, though roughly, what kinds of sentences are conditionals, and step into a further analysis[4].

 

2. Dependent Conditionals: First Target

After being able to collect most uncontroversial instances of conditionals, we can firstly consider whether conditionals have truth-value, for most people think sentences in indicative mood have truth-value.  For example,

 

The earth is flat.

Two plus two equals four.

Some people believe that money can buy everything.

 

Though they are concerned with different objects[5], we think they have truth-value, for they describe the world either correctly or not.  If they correctly describe the external world, we think they are true; if not, false.  Since conditionals are one kind of sentence, they should have truth-value.  Some may argue that conditionals are truth-valueless, for it is unclear whether conditionals that have false antecedents have truth-value.  However, let’s postpone this issue to Chapter 2, where I will introduce Edgington’s arguments against the view that conditionals have truth-value.  In my opinion, conditionals do have truth-value, if we want to have a unified theory of conditionals.  Before we go further, let’s firstly separate two kinds of conditionals: dependent/ contingent and independent/necessary conditionals[6].  For example:

 

If Eric is at home, he watches TV.

If Eric is at home, he is at home.

 

Intuitively, the truth of the second conditional doesn’t rely on any contingent facts, whereas the first does.  Let’s consider the second conditionals further.  Prima facie it has a ternary structure O(A, C), where ‘A’ represents ‘Eric is at home’ and ‘C’ stands for ‘he/Eric is at home’, though some may argue that we should not treat ‘if’ as one unit but with the clause[7].  No matter which structure correctly reflects conditionals, I think conceptually the ternary structure is convincing, though it is debatable whether it is in practice real, for it may turn out that in effect we can’t separate ‘if’ from the clause.  Hence, it seems legitimate to threat conditionals as ternary in structure.  Back to our example, what makes it true?

    In general, here we endorse so-called Correspondence Theory of Truth[8], which claims that a sentence a is true in the actual world, if and only if, a describes a fact.  Based on its ternary structure, what makes its components true may shed some light on what makes it true, though some may argue that ‘if’ isn’t truth functional. 

What makes ‘Eric is at home’ true?  Obviously, the situation that Eric is at home will make ‘Eric is at home’ true, and similarly the situation that Eric watches TV will make ‘Eric watches TV’ true.  Let’s consider the first conditional first.  I think most people will agree that, given Eric is at home and watches TV, the first conditional will be true, or at least acceptable.  If this intuition is common to most people, the second conditional will be true, for its antecedent is the same with its consequent.  It follows that, given the truth of its antecedent, its consequent will be true, and consequently the conditional in question will be true, or at least acceptable.  The difference between the first and the second conditional is the dependence of contingent facts.  The nature of the second conditional is that the truth of its antecedent with some law will guarantee its consequent from being false[9].  We can find lots of examples with this very nature in the following:

 

If neither Kate nor Daria answered the phone, then the phone was not answered by Kate.

If the river were to rise another two feet, it would be two feet higher than it is now.

 

According to their nature, let’s call the second kind independent conditionals, and name those similar to the first kind dependent conditionals.  Since independent conditionals will be true no matter what, investigating them won’t shed any light on the nature of conditionals.  Thus, our focus will be dependent conditionals, which is our first target.

 

3. OK Phenomena: What is the intuition behind OK pairs?

3.1 OK pair

After finding out our first target, let’s go further.  Among dependent conditionals, there seems to be an important difference between so-called OK pairs (Edgington calls these pairs “OK cases”[10]).  Take Adams ’ example[11]:

 

( 1a ) If Oswald did not kill Kennedy, someone else did;

(1b) If Oswald had not killed Kennedy, someone else would have.

 

People who are familiar with American history around 1960s agree with ( 1a ), but dissent from (1b).  From the divergent attitudes toward ( 1a ) and (1b), our intuition suggests that there must be some difference between ( 1a ) and (1b).  In order to see the difference clearer, I think we should ask what the nature of OK pairs is.  Suppose we can collect all OK pairs into a set, called OK set.  Can we offer a definition of it by abstraction[12]?

 

3.2 Indicatives and Subjunctives

Initially, I think that there are two things we should consider further: a grammatical distinction and divergent attitudes.  As to the former, it’s marked by (i) ‘would’ in the consequent and (ii) the back-shifting of the verbs in both the antecedent and the consequent.  Thus, ( 1a ) is called ‘indicative conditional’ and (1b) is called ‘subjunctive conditional’.  For those who take English as their first language, this grammatical distinction is taken for granted.  However, for people who take English as their second language, it is quite difficult to learn it well[13].  In what follows is an explanation of subjunctive:

 

the forms.  If she were coming, she would be here by now…. These sentences all contain verbs in the subjunctive mood, which is used chiefly to express the speaker’s attitude about the likelihood or factuality of a given situation. If the verbs were in the indicative mood, we would expect she was coming in the first sentence….

if clauses—the traditional rules.  According to traditional rules, you use the subjunctive to describe an occurrence that you have presupposed to be contrary to fact: if I were ten years younger, if America were still a British Colony. The verb in the main clause of these sentences must then contain the verb would or (less frequently) should: If I were ten years younger, I would consider entering the marathon. If America were still a British colony, we would all be drinking tea in the afternoon. When the situation described by the if clause is not presupposed to be false, however, that clause must contain an indicative verb. The form of verb in the main clause will depend on your intended meaning: If Hamlet was really written by Marlowe, as many have argued, then we have underestimated Marlowe’s genius. If Kevin was out all day, then it makes sense that he couldn’t answer the phone.

Remember, just because the modal verb would appears in the main clause, this doesn’t mean that the verb in the if clause must be in the subjunctive if the content of that clause is not presupposed to be false: If I was (not were) to accept their offer—which I’m still considering—I would have to start the new job on May 2. He would always call her from the office if he was (not were) going to be late for dinner…[14]

 

It seems that we can have two characteristics from the explanation above: (iii) the content of if-clause (the antecedent) being presupposed to be false and (iv) containing ‘would’ in the main clause (the consequent), which is the same with (i) mentioned earlier.  So far it seems that we can distinguish subjunctives from indicatives by (i), (ii), and (iii).  Therefore, given a conditional, if it satisfies (i), (ii) and (iii), it belongs to subjunctives; if not, indicatives[15].  In the first section, we have assumed that we can pick out conditionals from other kinds of sentences, so we can assume our domain contains and only contains conditionals. More specially, we take our domain as a set of dependent conditionals, which we can pick out in the second section.  And from (i), (ii) and (iii) we can have following sets: (w(x): x contains ‘would’ in the consequent; b(x): x contains back-shifting verbs in both the antecedent and the consequent; f(x): x presupposes the falsity of its antecedent)

 

Subjunctives: {x½w(x) and b(x) and f(x)}

Indicatives: {x½~w(x) or ~b(x) or ~f(x)}

 

3.3 The Indicative-Subjunctive Transformation

Can we define OK set?  Unfortunately, not yet, for we still need one assumption; that is, for every indicative conditional there is a corresponding subjunctive conditional[16].  This assumption is plausible, for given an indicative conditional we can follow some transformation rules to yield a subjunctive conditional, but not vice versa.  According to Gauker, we can have so-called “The Indicative-Subjunctive Transformation” (IST, for short): (Gauker, 2005: 231)

 

IST (two-step version)[17]

First step: Add the auxiliary “will” to the main verb of the consequent and modify the rest of the verb to preserve temporal reference.

Second step: Backshift the main verb in both the antecedent and the consequent.

 

It seems that IST catches the difference between indicatives and subjunctives; that is, feature (i) and (ii).  However, it seems that (iii) is missing, for IST tells us nothing about speakers’ belief.  Let me explain it further.

Usually, back-shifting the main verb in the antecedent is a marker, by which addressee can reasonably assume the speaker disbelieves in the antecedent.  Maybe it’s too strong, for sometimes the speaker does believe in the antecedent, for example[18]:

 

(2) If he had taken arsenic, he would have shown just these symptoms.

(3) They would be doing that, if a bus were coming.

 

(2) and (3) show that the speakers have strong evidence to believe in the antecedent, even though they use subjunctive mood.  So we need to modify (iii) into (iii’), which says that the content of if-clause (the antecedent) being presupposed (by speakers) to be hypothetical[19].  Thus, (ii) in a sense contains some element of (iii’), which can be revealed in conversation[20].  Now, let’s see some pairs formed by using IST: (Obviously, we can see how (1b) is derived from ( 1a ) by using IST)

 

( 4a ) If the butler didn’t do it, the gardener did.

 

By the first step of IST, we have:

 

(4b) If the butler didn’t do it, the gardener will have done.

 

By the second, we have:

 

( 4c ) If the butler hadn’t done it, the gardener would have done.

 

Thus, we pair ( 4a ) and ( 4c ) as a member of OK pairs.  Some may argue that we don’t know whether the speaker takes the antecedent as hypothetical, how can we pair them?  Take the following example: (A professor says to his student)

 

( 5a ) If you have any question, my assistant will help you[21].

 

Suppose the student does have questions, it seems that the professor won’t say (5b):

 

(5b) If you had any question, my assistant would help you.

 

I agree we can’t be sure whether the speaker thinks the antecedent as hypothetical, but what I want to collect is not an ‘actual’ OK pair but a possible OK pair.  As mentioned before, context plays a role in determining whether an indicative actually has its corresponding subjunctive.  However, the example above shows another flaw in IST; that is, sometimes ‘will’ is already contained in the antecedent.  Thus, we need to modify the first step of IST as following:

 

First step: Add the auxiliary “will” to the main verb of the consequent, if “will” is not contained in the consequent, and modify the rest of the verb to preserve temporal reference

 

So far we almost reach our goal, but there is still one flaw in the second step of IST.  For example[22]: (the speaker intends to warn his/her audience against going into a house, where the ceiling collapses after ( 6a ) being asserted and before (6b) being asserted.)

 

( 6a ) If you go in, you will get hurt;

(6b) If you had gone in, you would have gotten hurt.

 

First of all, none doubts Edgington’s example is an OK pair, when she used it to claim that the distinction between indicatives and subjunctives is a matter of time, probably marked by tense of verb.  However, if we apply IST to ( 6a ), we will have ( 6c ):

 

( 6c ) If you went in, you would get hurt.

 

Thus, if we agree ( 6a ) and (6b) belong to OK set, we need to modify IST as following:

 

Second step: Backshift, relative to assertion time, the main verb in both the antecedent and the consequent.

 

So far it seems that we can define OK set as following: (Let our domain be dependent conditionals)

 

{½xÎ(~w(x) or ~b(x) or ~h(x)[23]) and yÎ(using IST on {x½~w(x) or ~b(x) or ~h (x)})}

 

Now, we have explained the grammatical distinction between our initial pair, ( 1a ) and (1b), but how about the divergent attitudes toward them?  This problem will be answered in the next section, but before we step into next section one thing should be noticed: we may not have such divergent attitudes toward OK pairs.  As mentioned earlier, some pairs in OK set may not be actualized.  Similarly, our attitudes toward them may not be divergent, for it is context that makes them divergent, which is deserved to be explained further.

 

4. Kinds of Classifications: Are OK phenomena explained?

There are kinds of classifications in literature, but I will just introduce some of them, for I think others are wrong.  In general, there are two ways in dealing with OK phenomena; that is, Y-shaped or not[24].  The former takes ‘if’ in indicatives as the same with the one in subjunctives, whereas the latter takes them as different.  In this section, I will first explain the difference between these two first and then the reasons why we should prefer a Y-shaped analysis of conditionals.  In the end, I will introduce three different Y-shaped analyses and one thing should be kept in mind: these analyses are supposed to explain the divergent attitudes toward some OK pairs.

 

4.1 One ‘if’ or two ‘ifs’?

As mentioned before, we treat conditionals as ternary in structure.  Hence a conditional can be analyzed as O(A, C) in structure, where ‘O’ represents the connective, ‘A’ and ‘C’ represent the antecedent and the consequent respectively.  Therefore, if there is any difference between indicatives and subjunctives, it must lurk in either ‘O’ or ‘A/C’.

Since classical logic being developed, many people think that truth-functional account gives indicatives a good explanation in mathematics.  However, when logical empiricists analyzed some dispositional concepts, they realized that there might be another kind of conditionals in scientific realm.  Thus, this historical event indicates that there might be two different ‘ifs’, for truth-functional account fails to explain subjunctives[25].  Then, Adams ’s OK cases reinforce the effect of that historical event.  In any case, some people, like Lewis, think that the difference between indicatives and subjunctives is due to the ambiguous connective ‘if’[26].  Let’s rephrase ( 1a ) and (1b) in Lewis’s term[27]:

 

( 7a ) If it is the case that Oswald did not kill Kennedy, it is the case that someone else did;

(7b) If it were the case that Oswald did not kill Kennedy, it would be the case that someone else did.

 

From ( 7a ) and (7b), Lewis thinks that we can see the reason why we may have divergent attitudes toward pairs like ( 1a ) and (1b) is that the connectives embedded in them are so different.  Due to this very difference, we cannot have a unified theory of conditionals, and it is an accidental fact if there is any similarity between them[28].  On the other hand, people like Stalnaker, Ellis, Davis, and Edgington think the difference between indicatives and subjunctives is not due to the connective ‘if’ but the antecedent and the consequent, and thus ( 7a ) and (7b) are wrong re-formulations of ( 1a ) and (1b)[29].  Since their analysis of conditionals begins with what is common to indicatives and subjunctives and then forks in order to explain the difference between them, we can call their analysis ‘Y-shaped’[30].  In my opinion, Y-shaped analysis of conditionals should be preferred[31], for I agree with Ellis that a unified theory of conditionals, which offers some kind of Y-shaped analysis, meets two principles in theory evaluation.

 

4.2 Why Y-shaped analysis should be preferred

According to Ellis, there are two principles in theory evaluation; that is, the principle of comprehensiveness and the principle of evidential support[32].  The first says that in face of pervasive phenomena, which should not be treated as accidental, we expect to be able to give a comprehensive account of them.  The second says that a good theory must be sustained by the evidence throughout its domain, though it may alter some of our intuitions within that domain.  Now, suppose there are only two theories available, on Ellis’s view we should prefer the one that meets the two principles mentioned above.  As we have seen, there are two kinds of theories handling OK phenomena; attributing OK phenomena to the connective or to the connected sentences.  When facing these, we find not only some difference but also some similarity between indicatives and subjunctives.  By the first principle, we should not treat the similarity or difference between indicatives and subjunctives as accidental.  Furthermore, a unified theory is expected, for it seems that those who attribute OK phenomena to the connective thought the similarity between indicatives and subjunctives as accidental[33].  On the other, by the second principle, we should prefer an evidential support theory.  There are some facts that should not be ignored: (i) we use the same word ‘if’ in making both indicatives and subjunctives; (ii) we learn how to use or form a indicative first and then know how to use or form a subjunctive by changing the tense or mood; (iii) the plausible translation into indirect speech shows that the ‘same-saying’ relationship can hold between indicatives and subjunctives.  Therefore, a unified theory meets the two principles and should be preferred. 

    So far we have seen that a unified theory of conditionals should be preferred, but it’s not over yet, for there are still kinds of Y-shaped analyses.  Next I will introduce Stalnaker’s, Davis ’s, and Edgington’s Y-shaped analysis of conditionals.

 

4.3 Stalnaker’s Y-shaped analysis

According to Stalnaker, a conditional ‘if A, C’ is true in i, iff, C is true in s(A, i), where i is the base world (the actual world) and s is a function from antecedent and i to a world[34].  Intuitively, the function s with two arguments should not pick out an arbitrary world, so s has to meet four semantic conditions[35].  First, for all antecedents and base worlds α, A must be true in s(A, α); that is, all independent conditionals, such as ‘If snow is white, snow is white’, will be true, which accords with our intuition.  Second, For all antecedents A and base worlds α, s(A, α) = λ only if there is no possible world with respect to α in which A is true, where λ represents the absurd world[36]; that is, when A is possible, either ‘if A, C’ or ‘if A, ~C’ is true, but not both, which also accords with our intuition.  Third, for all base worlds α and all antecedents A, if A is true in α, then s(A, α) = α; that is, given any base world, the most similar world will be itself, given A is true in it, which is a reasonable requirement.  Fourth, for all base worlds α and all antecedents B and B’, if B is true in s(B’, α) and B’ is true in s(B, α), then s(B, α) = s(B’, α); that is, based on the same consideration, there will be one and only one possible world that is the most similar to the base world, which is also a reasonable requirement.  However, given these four constraints, we won’t have one unique selection function s, for what counts the most similar depends on something else.  Let’s consider this further.

    Given any conditional, its antecedent can be true in the actual world or not.  If true, according to Stalnaker’s semantics for conditionals, s should pick out the actual world (the third condition); if not, what counts the most similar depends on context.  Naturally, speakers will take something for granted in conversation.  And we can call those assumed by speakers and audiences ‘common knowledge’[37], which can be represented by a set of possible worlds, called ‘the context set’[38].  For example, when considering ‘if Peter’s laptop had not broken, he would not have bought a new one’ in a context where both the speaker and his/her audience believe that Peter always wants to have another laptop, the selection function s may pick out a world where Peter’s laptop were not broken and he bought a new one.  However, while considering a different context, s may pick out a world where Peter’s laptop were not broken and he didn’t buy a new one, if both the speaker and his/her audience believe that Peter is frugal of money.  In different contexts, what counts the most similar will change.  Here we can see the difference between indicatives and subjunctives is not semantic, for Stalnaker semantics applies to both.  But, when we consider subjunctives, the subjunctive mood indicates that the selected world may be outside the context set.  Thus, the difference lurks in pragmatics[39].

 

4.4 Davis ’s Y-shaped analysis

For Davis , based on Stalnaker’s semantics, we can have a better analysis of conditionals.  By considering Adams ’s example, he concluded Stalnaker’s semantics (‘The Stalnaker Principle’ called by him) is true of indicatives[40]