I am doing a study to wild pollinators and seed production at the moment, and I want to know whether different species have a preference towards different parts of the flower head. I have noted for each species how many times it was present on which part of the round flower head (lower part, middle part or upper part). I have performed a Chi-square test (predicted values are 0.3333 * the total number of observed individuals).For most species I have found that the observed values significantly differ from the expected values (I do not know if I say this correct in this way, but there is a significant difference). However, I want to know which groups differ significantly from each other (i.e. bottom vs. middle, bottom vs. top and middle vs. top for each species).

I found somewhere that you can perform a Chi-square test like this:

e.g. bottom vs. non-bottom (i.e. middle and top together) and as expected values: for bottom 0.3333 * total observed individuals of that species and for non-bottom 0.6666 * total observed individuals of that species. And then you have to apply a Benforroni correction (thus 3 comparisons --> alpha = 0.05 / 3 = 0.017).

Is this correct or are there better ways to do this?

Thank you in advance for your answer!