As I recall, the t-statistic for significance of a Pearson's r is t = [r*sqrt(N-2)]/sqrt(1-r^2). For N = 168 you have df = 166 and hence a critical t-value of 1.974358 (found using R command qt(.975,166) -- we use .975 for a two-sided test, but you can substitute a 1 - a/2 in place of the .975 for any value of alpha). Then simply substitute N and t into the equation and solve for r; I got something in the neighborhood of r = +/- .15.