For once I'm almost in agreement with you. In "the field", so to speak, instances where people think they know telepathically when someone they know is about to call is very likely to be due to a combination of selective memory, as you say, and implicit knowledge about calling habits IMO. I think there may be an underlying ESP effect but its going to masked out by such things. The only way to find out if there is an underlying ESP effect is to do the experiments.
So we are also discussing valid experimental methodologies for telephone telepathy. You seem to think that people claim to simple know who is calling them in advance, irrespective of whether they know the person or not. I disagree. People claim to know specific people who are about to call them, which means we should be using known callers in the experiment. By all means, propose a design that uses 100 unknown callers and 4 known. As long as you have a means to work out a hit rate against chance and have the controls in place then its a valid experiment. However, an equally valid experiment can be devised just using known callers.
The blindingly obvious is right before you eyes if you but look. I'll spell it out for you, but this is for the benefit of the slow lurkers.
These claimants DO claim that they know who is calling. What they FAIL TO TAKE INTO ACCOUNT is that they are likely to be performing no better than chance because they fail to understand their "testing environment". In effect, they are counting the hits and not the misses, in a biased testing process.
Furthermore, if anyone uses a testing methodology that is basically just a guessing-game, it is quite conceivable that you CAN score significantly better than chance
in a few tests in a series. Even tossing heads-and-tails randomly is bound to produce long runs of heads or tails at some point.
Now put these two points together, please.
We have a bunch of innocent folks who would not know an unbiased test if they tripped over it.
We have a testing environment that is implicitly biased heavily in favour of them making correct selections.
We have them counting their hits and not the misses.
We have them being astounded that they then think they have ESP or psi or whatever.
We have them making public claims on that basis.
We have the usual psi crowd going all gooey that this is the real deal, but not doing any actual research of evaluation.
We then have Sheldrake weighing in on the subject and implying there may be something in it by conducting equally skewed testing.
Finally we have these basic flaws in his work pointed out that not even the greenest researcher would have overlooked. In fact, so glaringly wrong that it seems all the more incredible that Sheldrake really did make such a boob. Instead, one starts to get the impression that it was deliberate, in a blatant attempt to publish as a means of retaining tenure, rather than publish to expand the sum of human knowledge.
This is not the first time such glaring mistakes, and possible paper-milling, have been made by psi researchers, incidentally.
I am forever reminded of the story of The Emperor's New Clothes when I read these sorts of reports, incidentally...
I would share the general skepticism that indeed that a guessing-game is what it would be. However I am open to accepting proof to the contrary.
I think the latter is inevitable. 