DW interpretation to reject the Null Hypothesis

If the DW test stat indicated serial correlation, how can Smith then reject the Null Hypothesis? Wouldn’t the model have to be re-fit (correct the serial correlation) to then allow Smith to perhaps reject the Null?
The question did not give the DW test stat number.

The null hypothesis is that there is no serial correlation.