Are the samples independent? I'll assume that they are.

The samples are small size so the t-distribution should be used. Since and do NOT differ by an order of magnitude, it can be assumed that df =

.

Two tailed therefore require (refer table given at Student's t-distribution - Wikipedia, the free encyclopedia or use the calculator found at http://www.anu.edu.au/nceph/surfstat...e/tables/t.php).

The confidence interval is

Since the degree of freedom is so large you could use the z-distribution without too much error.