OK, thank you for clarifying - I understand what you're after now.

Unfortunately, even if you sampled the entire population, there's no guarantee that you will find a significant relationship, so with correlation/regression type studies, it's difficult to predict the necessary sample size.

If the relationship is in fact strong and reliable, then you won't need very many - but if it's relatively weak, you'll need a big sample size (but then again, what you've shown, with a big sample size is that there is a weak relationship - so a lot of work for very little payoff).

In other words, you kind of need to know what the correlation coefficient is (at least a rough estimate) before you can determine the minimum sample size required to detect such a correlation.

So, my answer, although not what you wanted to hear, is that you need to sample "as many as you can" within your allotted budget and time, and then see how strong or weak the correlation is. Of the references I've been able to locate on the web, they say anywhere between 30 and 60 should be enough in most cases, but more is better.

My question is: Why would there be a relationship (correlation) between male and female physical attributes such as size/age/weight - wouldn't there be a more-or-less random pairing of mates? Why, in nature, would there be a predictable relationship here? Just curious....