In this context, what P = 0.04 (i.e., 4%) means is that if the null hypothesis is true and if you perform the study a large number of times and in exactly the same manner, drawing random samples from the population on each occasion, then, on 4% of occasions, you would get the same or greater difference between groups ...