I would predict that girls will remember the poem better than boys, and that they do so regardless of color. [Editor: this is because past experiments have shown that, on average, girls have better verbal skills than boys.]
I suggest that when you test this prediction in your project, do not just use any arbitrary color. Give each of your participating subjects a chance to pick their favorite color, from among those that you can produce. This feature of the design may enhance the chances of your finding differences between black and white versus color, and differences between girls and boys. In any case, you guard against a possible criticism that your color treatment did not work because you used unpleasant colors. Also, if you want to make the experiment more ambitious, use at least two different gender-neutral poems.