I see the following points about VVPAT:

• the winner of the election is decided in the first count (probably the only count) which is based on electronic votes. Infact VVPATs are counted in the second count (recount or audit), but this rarely happens thus VVPATs will most likely not be used or counted. You are able to view the Paper Audit Trail, and make sure it is correct. You have no way of knowing what your electronic ballot says. You can feel relatively certain that if there is a hand recount, your vote will be counted properly. But since hand recounts are very rare, when you look at that piece of paper, you are not actually verifying your vote. There is still nothing to verify that your actual vote was correct.
• VVPAT recounts shouldn't occur only when elections are very close.
  Infact where deliberate fraud does take place, the magnitude of the fraud may not be small. And also the magnitude of accidental errors may not necessarily be small. Thus fraud and errors can produce very different results. Unfortunately many people, and state laws, only want recounts to be conducted when elections are very close. It seems that people are willing to do recounts in the case of small accidental errors but not to detect fraud or large errors!
• it is not possible to make a statistical "recount" of VVPATs by manually counting a small percentage of them and seeing if the result is more or less the same as the electronic one. Infact, as candidates of USA-2000 election well remember, elections can be very close and so a precise count of all the VVPATs could be necessary. Some legislations require a little 1% recount to validate electronic results!

Thus VVPAT can't be used to verify electronic electoral results unless they are all counted. But if we really print and count VVPAT for each casted vote then we simply run a paper election which ballots are printed by machines instead of being hand written by electors!

• we double the efforts of each election, which is now made of an electronic one and a paper one
• we greately increase the election cost (try to imagine how it costs to buy and maintain a PC in each voting boot, plus the software, plus the network apparatus and lines, plus the high-tech skill involved, ... and compare it with the cost of ballot papers and pencils!)
• we know from the beginning that the official result will always be the one coming from the counting of VVPATs. Infact we use them to confirm electronic results, thus in case of discepancies they surely win. Thus, what for do we also run an electronic election?

• all the electors should verify their own vote simultaneously at the same time in which a (proven error-and-fraud-free) tally is executed to produce the final result. If the counting would not occur in the same moment while ALL the votes are being verified, we could not have the prove votes being properly tallied up: it's a joke to program any computer to show to an elector his true vote and then not taking it in account during the count.

• votes would be no more anonymous since voters could be tracked (otherwise we couldn't know who can verify them).
  We would miss the anonymity requirement due to the possibility to link a vote with its voter. It is not enough to say that the "key" to make such link might be only available to the elector himself. In paper ballots such key doesn't exist at all! Anyway, even if we could find a safe way to allow voters to verify the vote stored in their behalf ...
• there is no way to know if a claim of error would be honest
  We know electors can't be given (for their own sake!) any "receipt" stating how they voted, and thus there is no way for them to prove, if it is the case, the vote stored in their behalf is not the one they really cast.
  Even if it would exist an algorithm allowing the verification of the recorded votes without breaking their anonymity, it should be used with great care. In fact it would in any case show the electors how their vote has been recorded and thus it would be much like as they were given a receipt of the cast vote. But vote "receipt" can't be used!
• we must have options in case such verification fails
  Any verification process must have at least two options to be taken upon its result: if the prototype of the new car is properly working we start producing it on large scale, but if such verification is negative we don't.
  But what could we do if somebody claims his vote to be erroneously recorded? There would be no prove pro or against such claim; should we change the result of elections because of it? Electors might even change their mind, what should we do if half a million people request to change their vote? Should we allow it? Will the pro tempore winner agree?
  Electronic voting's verification is quite a strange process: regardless how it goes, the election's results are in any case confirmed! We didn't notice it but we entered a logical loop:
  1. due to the large interests moved by elections (we talk about ruling people and nations), we want electoral results to be verifiable.
  2. we want to use electronic vote and find out the only way to verify the final result is to verify each electors' vote.
  3. discrepancies between recorded votes and what electors claim they cast cannot be proved, thus we can't modify the result nor cancel an election simply because (few?) people claim vote have been falsely accounted for.
  4. not permitting any changes we act like if results were error-free and fraud-free and thus we could blindly accept them. This is an obvious contrast to the starting point 1)
  If anybody can honestly say elections results' don't need any verification, please go to e-vote otherwise go back to point 1)