hypothetically (:$ ) your questionnaire that you designed is SO bad that the only factors that come out in factor analysis are based on the beginning of the questions?

I have 3 factors

Factor one - all the questions = "it has been easy"

Factor two - all the questions = "I rarely...."

haha - its SO bad. Do you reckon I should pretend that I never made a questionnaire??? and that it doesn't exist. i think sup has forgotten I spent 3 months making it.

If an item loads on to two factor scores what do you do?

Can you choose the one that *seems* to fit best, or do you go with the highest loading, or can you have it in both factors?

I know what factor analysis is about, but I've never actually run one. Are you sure that it has run correctly? It just seems very strange for you to end up with factors that look like that.
I don't think you should pretend that you never made the questionnaire. It's still viable. You can always make the suggestion that an EFA constitutes further work for your questionnaire. I'm at a loss to explain why? Have you tried rotating it? If your questionnaire was so bad, then you'd expect lots of factors or no factors.
What are the 'it has been easy' and 'I rarely' questions trying to measure in general terms (that won't identify you or your research)? Lots of different things from one another?

If I remember what I read about factor analyses correctly, you go for the factor the items loads highest on. However, don't forget that you can rotate the factor analysis until you get the most 'pleasing' factor loadings. It's quite subjective in that respect.

I've tried varimax and direct oblimin rotation.

Orginially the questionnaire is supposed to measure 4 key things.

My first factor = bang on what its supposed to measure

and the latter 3 factors seem to have some patterns, but then one item seems to throw the whole thing off. e.g. 6 items on liking chocolate and one on liking cheese!

My fourth factor consists of 4 items all starting with "it has been easy" but finish with completely different things e.g. "to eat chocolate", "to ride a bike" "to procrastinate"

I'm not exactly sure why I'm doing factor analysis anyway. I thought you had to to make a questionnaire??

All my cronbach's alphas show my items are measuring the right things i.e. the original 4 things I wanted to measure.

Sneaks, it's not a problem. If you've got an item that doesn't seem to belong to any of the factors (which will make up subscales), which you've shown with your EFA, you can delete it as part of item reduction. You've shown that it doesn't actually belong to any of the subscales.

As far as CTT is concerned, using factor analysis is right at the top for developing a questionnaire - kudos to you for doing it. It'll help show you have unidimensional subscales that therefore measure individual constructs. If you've then got Cronbach's alphas of between 0.7 and 0.9 for each of those subscales, all the better. See, told you that questionnaire of yours isn't crap.

ok I've played around with the data and got 3 factors that just about make sense.

However, I originally wanted to measure 4 things - I had 4 subscales.

Now its saying only 3 factors? D'oh!