Fake accents are also great as examples of the power of suggestion. The easiest trick is to just find a few words that suggest the place in question according to stereotypes, learn to do them well, and just sprinkle them in liberally. So if you needed to suggest Irishness, you could just learn Irish-sounding versions :
and just use them in some combination.
'I love Guinness with me 'taters, 'specially in County Cork. But not with the fookin' English'.
If you need to actually give a randomly chosen dialogue in a foreign accent, it's considerably harder, since you can't just pick your own words. The chance of being able to convince people depends greatly on their own familiarity with the accent. The hardest is to convince native speakers, since they'll know immediately what sounds wrong. The gold standard for all this is of course Hugh Laurie - Americans who watched House are constantly surprised to find out that his normally speaking voice is strongly English. This is the real Hugh Laurie voice. You can here his House accent here and here.
My fake American accent is marginal at best. By which I mean, it's pretty good by the standard of most people's fake accents, but put me next to a native-speaking American and you can clearly tell where my flubs and weird vowel sounds are. C.f. Hugh Laurie, my American friends generally find it painful to listen to. So if the test is 'If you suspect it might be fake, can you quickly find evidence to confirm this hypothesis?', then I flunk it by a mile.
But most of the time, this isn't actually the test. The real test is 'If you didn't know in advance that it was fake, is it bad enough to raise in your mind the possibility that it might be an impersonation?'. It turns out that this is a much easier standard to beat, because most of the time people aren't on the lookout for someone using a fake accent.
Being a man of science, I decided to try this in the wild. For the first 40 minutes of meeting new Americans, I'd use my fake American accent, then switch to Australian. I'd then ask the person if they suspected that it was fake. Based on a pretty big sample, the percentage who suspected it was fake was between about 5 and 10%. And this is for an accent so bad that people who know me find it gratingly unpleasant to listen to. But people who don't know me just interpret the mistakes as being some sort of regional variation - the slightly Australian 'r' sounds were forgiven as being some sort of East coast/Boston twang.
It's really an example of the curse of knowledge - people who know some information are typically very bad at putting themselves in the position of someone who didn't know the information. If you know my accent is fake, you suspect that everyone will be able to tell that it's fake. But it doesn't work that way.
The other funny observation on this came from my friend SH, who watched one of my recent attempts. He said that my body language became somewhat forced. It was like, he said, watching me trying to perform a difficult calculation. I'd totally believe it - some significant part of your brain is devoted to making the words come out in a different way, and this is actually pretty hard work.
Convincing them that you're not weird after you switch accents, however, is considerably harder. Nobody said science was easy.