?

Is it wise to use means in the contingency table for calculating chi squared, or should I just use totals?

I have done a biology experiment where I tested the feeding habits of two tanks of snails. The tanks were the same, but the idea was to calculate a mean from them to make the data more representative. If I just look at the means there is clearly a pattern, but when I do the chi squared test on them, it comes out as not significant.

Please enlighten me. I can't find this anywhere else.