This reminds me of a story I heard in statistics involving a claim by a woman that she could tell the difference between the difference between adding milk to tea vs adding tea to milk. A statistician overheard this boost and decided to design an experiment to determine whether her claim was accurate. If you use just two cups, one with milk added and one with tea added, well, even I have a 50% chance of choosing correctly just by picking one at random. It doesn't really tell us anything if she picks out the one with milk added.
It is possible to use statistics to estimate the probability that the observed results could be explained by someone just randomly choosing cups of tea. If that probability is low*, there is a statistically significant chance that there is something else to explain the results (i.e. that the tester can taster can tell the difference). Ideally, the scientist will design experiments before hand so that they know what significance level they can expect at the conclusion of the experiment. Using 2 cups of tea gives us a 50% probability that random chance explains the solution, so is not very meaningful. It turns out that an 8 cup experiment, 4 with milk added and 4 with tea added, makes it much more difficult to correctly identify the 4 cups with milk added if you just select them at random. Indeed, there's only a 1.4% that someone will correctly select the 4 cups. If the lady is successful, there is a high probability that she can actually discern the difference. If she got even one incorrect, however, that claim would not be statistically sound (there's an almost 1 in 4 chance of getting three of four cups correct if you just randomly select them--not terribly impressive).
Your pasta question is almost exactly the same as the question involving tea: can a person tell the difference between food A and food B. Only instead of whether tea is added to milk or milk is added to tea, you examine US pasta versus Italian pasta. The tea experiment required the drinker to correctly pick which ones had milk added. If you are only interested in whether your guests can tell the difference in the, but not necessarily whether they can correctly identify which is US and which is Italian, a similar experiment would only achieve a 2*1.4%=2.8% significance level. Still pretty good.
Of course, preparing 8 dishes per guest may be a bit of work. If you do 6 dishes, with 3 American and 3 Italian, you'd be right at 5% significance if a guest correctly chooses the pastas. Using 4 dishes would give you a 1/6 (16.7%) significance, and 3 dishes would just give a 1/3 (33.3%) significance. Up to you to determine how much confidence you want, and how much work you want to put into it ;).
Note that these numbers are for a single individual testing the dishes. Presuming you have a number of guests, the analysis gets more complicated, since you're adding in an additional variable (each individual is different). Generally, having more guests would help with you confidence that there is a difference in the pastas...if they are all able to pass whatever test you set up. If only a handful are, it's much harder to draw conclusions.
* The choice of the target probability that random chance can explain the observations (the null hypothesis) is arbitrary. A lower probability means you have more confidence in your observations, but requires more work. In many scientific fields, a level of 5% is considered "statistically significant", but there is push-back that a much smaller level should be used, as alluded to by @doneal24 in the comments.
