To formally test for serial correlation in your residuals:

Find the box corresponding to the number of X variables in your equation and the number of observations in your data. Choose the row within the box for the significance level ("Prob.") you consider appropriate. That gives you two numbers, a D-L and a D-U. If the Durbin-Watson statistic you got is less than D-L, you have serial correlation. If it is less than D-U, you probably have serial correlation, particularly if one of your X variables is a measure of time.

The reason for the D-L and D-U is that the distribution of the Durbin-Watson statistic depends on the values of the X variables in your data. This is different from the t-test, where the distribution depends only on the number of degrees of freedom. If your X variables are well-behaved, which mainly means that their values are evenly spread out over their range, then the Durbin-Watson statistic's distribution is indicated by the D-U's. If your X variables are not well-behaved, then the Durbin-Watson statistic's distribution might be as low as that indicated by the D-L's. Using the D-L for your critical value makes sure you err on the side of not saying that there is serial correlation.