Trust in what you think. I agree with you. Wittgenstein said much that was important but Philosophy has continued on without him in the analytic sense. He wished to highlight uses of language and that must be borne in mind here. Of course daydreams are not necessarily lucid or banal. Suspect all limiting results unless they're proven, and even then be careful. Wittgenstein also said that: 

"Imponderable evidence includes subtleties of glance, of gesture, of tone. I may recognize a genuine loving look, distinguish it from a pretended one (and here there can, of course, be a 'ponderable' confirmation of my judgment). But I may be quite incapable of describing the difference. And this not because the languages I know have no words for it. For why not introduce new words?—If I were a very talented painter I might conceivably represent the genuine and the simulated glance in pictures."  Phil. Investigations  (Elizabeth Anscombe)

I think W. was also trying to say that we felt it but it wasn't always confirmable statistically. However shortly after W. died, Paul Ekman did some very interesting work on human facial expressions, counter to Mead's beliefs against universals of this type existing across cultures. In fact one uses such knowledge all our lives without necessarily realizing it explicitly. So W. was apparently wrong.

Hi Martin,
Why do you say Wittgenstein was wrong? It seems to me your example of Paul Ekman proves him to be correct. Wittgenstein does say one can distinguish the looks, only that we cannot describe the difference. Somewhat like what he says about a game, that we all know what a game is, we can give examples, but not a definition.

Interesting phrase Mia. I also wouldn't agree with it as it seem Wittgenstein phrase is based on a presumed separation of consciousnesses and unconsciousness when no such real separation can be done with regards anything we do.

Dear Mike,
Because I recall reading W. many years ago where he stated that these imponderable forms could be privately agreed upon but there wasn't any proof as far as a statistical test went, i.e. not empirically verifiable in sense that a large number of people would agree. It was left at that and that it might become ponderable later is rather interesting and I wondered under what circumstances Anscombe added it in? Recently I went to look for it because of your post but had no luck. With Ekman we can certainly describe the difference because, for example, based on his work, computer algorithms were developed for Homeland Security to use to ID terrorist behaviour from live video also security personnel can be trained in it's use. Might need a Wittgenstein scholar for this one perhaps?