I also collected small amounts of seed for 29 species, and measured length and width either under x40 or x100 magnification. Generally I counted at least 10 seeds to give a reliable estimate, but in some cases only one or two. The length; width ratio was not a good discriminator because long seeds were generally thinner and short seeds fatter, as if the volume was roughly the same (I need to look at this further). The average seed length does though offer potential to discriminate, and again interestingly G, conopsea is a long way from A. pyramidalis.
Also the genera are broadly grouped by seed length but this is accentuated when the seed length is plotted against the pod length : width ratio. Clear clusters for different genera can be seen.
This chart seems to provide a way of discriminating between genera - dactylorhiza from gymnadenia say, and then the seed length and pod length : ratio might hint at species identification, but it is early days. Too late this year to do any more work so a project for next year confined probably to those genera with several species such as the marsh and fragrant orchids and the helleborines, so that is the summer dealt with. I also need to look at the literature - there is probably a whole lexicon on this and I need to compare and contrast, so that is winter dealt with also.