Use the formula (sample mean difference) � (t)(sem), where t represents the (1 - alpha/2) percentile of a t distribution with n - 1 degrees of freedom, and sem represents the standard error of the paired mean difference = sd / sqrt(n).
There are Pages and JAVA programs on the Web that can perform this calculation (e.g., http://glass.ed.asu.edu/stats/analysis/mci.html).
Such confidence intervals will have a 95% chance of capturing the (true) parametric mean difference.