When an improbable or an (un)desirable topic has been introduced, one may reply by diminishing the intensity of the modality : the event is not so improbable afterwards, or it is not so much trouble, or it is not so nice (of course this possibility does not exist in the paradoxical mode). There are several ways offered to the interlocutors to achieve this.
[ex_sailing]
context: discussion about a French skipper who competed in the America's cup challenge, with the support of many people in France.
A1- Alors Marc Pajot s'est fait écraser encore.
B1- Deux minutes, c'est pas écrasé!
A2- Deux minutes quarante secondes, si.
C1- Oui, mais enfin, tu sais, hein, il a quand même tenu le coup jusqu'au dernier moment, hein, lui.
A3- Quel coup?
C2- Hé ben enfin, il a pas sombré, il est pas tombé, il est arrivé!
A4- Tu veux dire qu'il est arrivé, il est arrivé en demi-finale
C3- en demi-finale, écoute!
A1- So Marc Pajot was beaten badly once more.
B1- Two minutes, you can't say he was beaten badly
A2- Two minutes forty seconds, yes he was.
C1- Yes, but you know, he was in the running till the last moment, wasn't he?
A3- What kind of running?
C2- Now, he didn't sink, he didn't fall, he finished!
A4- You mean, he reached the semi-finals
C3- Yes, the semi-finals, it isn't that bad!
B1 is a banalization. It does not deny the fact that M. Pajot was beaten, nor the undesirability of this fact, but the intensity of the latter. In the improbable mode, things are a bit more complex, but nevertheless very interesting because they are in total agreement with what formal probabilities predict [Dessalles 1993]. A first strategy consists in mentioning additional facts that increase the a posteriori probability according to Bayes rule.
[ex_thirst]
context: A and B are speaking about D, their great child. D (one year old) seems to remember them after a separation of several months. B claims not to be surprised : the boy laughed when hearing their voice at the phone. A few seconds later C notices that D swallows a big quantity of water.
A1- Apparemment, il nous avait pas oubliés.
B1- Non. Il nous a pas oubliés quand ... il riait aux éclats quand il entendait notre voix.
[pause]
C1- Hé ben, il avait soif!
B2- Oui, il avait soif. Je m'en suis douté, qu'il avait soif!
A1- It seems that he didn't forget us.
B1- No. He didn't forget us when ... he laughed when he heard our voice.
[pause]
C1- Gee, he was thirsty!
B2- Yes, he was. I suspected that he was thirsty!
The first topic is about a child (D) that A and B see after a long separation. D is so young that it was a priori improbable that he could remember them. A1 brings information, in the Shannon sense. B1 does not invalidate the reasoning leading to IMPR in any way. We may represent the logical context this way :
[ young( D ) & long_separation( D, A_and_B ) & not forgot( D, A_and_B ) ] => IMPR
The additional knowledge underlying B1 may be represented by :
laughed_when_hearing( D, A_and_B ) => not forgot( D, A_and_B )
So B1 performs no invalidation. On the contrary, it seems to confirm A1 ! This excerpt is quite remarkable, because B shows exactly the same behavior with C one second later, on the next topic. C is impressed by the quantity of water D is swallowing, and thus brings information by focusing on an improbable fact. B2 seems once again to be a confirmation.
Many authors consider only cooperative aspects of conversation, in which information and confirmation play important roles. However, at the logical level at which we look at this excerpt, we can see that B's utterances, in both cases, do not merely acknowledge the interlocutors' statements. It is very important to see that B1 and B2 do much more : they aim at diminishing the originality of A1 and C1 respectively. What A and C assert is not so much the event they noticed as its a priori improbability. B twice changes this probability. After B1, for instance, Prob(not forgot) has to be replaced by Prob(not forgot | laughed-when-hearing), which is much greater (it is actually equal to 1 if we consider that [laughed => not forgot]). In other words, with the knowledge that D laughed when hearing A and B, it is much less improbable that he remembers them. Again with B2, B simply indicates that the probability of the little boy being thirsty was for her not so low since she suspected he was.
We can thus anticipate several ways for a speaker Y to perform a banalization by raising the a priori probability P(Ev0) of a given event Ev0 reported by X as improbable. Y can simply indicate that her/his own estimation is above X's, as was the case in B2 :
PrY( Ev0 ) >> PrX( Ev0 )
Y can also reveal that (s)he knows additional facts F1, ... Fn so that:
Pr ( Ev0 | F1, ... Fn ) >> Pr ( Ev0 )
as was the case in B1. But there is yet other possibility, as we show with the next excerpt.
[ex_loud] (from [Tannen 1984:101])
A1- Speaking of which they had the Loud Family. Remember the Loud Family? On Saturday Night Live? [TV program]
B1- What was the Loud Family?
A2- Dju hear about that? THEY TALK LIKE THIS.
B2- I know lots of people in New York who talk like that.
In this case B increases the empirical frequency of the fact which has been presented as improbable by A. Using his own sample (people he knows), the fact is no longer unique or even rare. This strategy may give rise to so-called story rounds [Tannen 1984] that everyone has experienced : after a first story telling some improbable event, the next speaker tells another story about an analogous event. The more analogous, the more efficient it is as a way to reduce the information brought by the first story (both events belong to the same universe of events). However, the second story leads most of the time to a topic change, since listeners forget about the first story, and the new topic is then considered as an improbable fact in itself. Another similar story will come as banalization, and so on. Notice that the similarity between stories is a requirement of banalization.
Interlocutors may use the full range of what formal probabilities predict in order to perform a banalization in the improbable mode. For instance, in order to banalize a dated fact, it is better to mention a similar fact that is recent, since an old fact may imply that the frequency is still low (this is consistent with what a Poisson law suggests).
There is a last way of making a relevant reply, which is by nature limited to the (un)desirable case.