We discuss point sources foregrounds for the MAP experiment .
We consider several possible strategies for removing them and we assess how the statistics of the CMB signal are affected by the residual sources .
Assuming a power law distribution for the point sources , we propose a method aimed to determine the slope of the distribution from the analysis of the moments of the observed maps .
The same method allows for a determination of the underlying CMB variance .
We conclude that the best strategy for point sources finding is the simultaneous thresholding of the filtered map at all frequencies , with a relatively low threshold .
With this strategy , we expect to find 70 ( 95 ) % of the sources above 3 ( 4 ) \sigma .
Assuming the most conservative case for point sources detection , the recovered slope of the point sources distribution is 2.55 \pm 0.15 , for a fiducial n = 2.5 value .
The recovered CMB pus noise map variance is within 0.2 \% from the real one , with a standard deviation of 0.31 \% , while Cosmic variance contributes 2.2 \% to the same CMB plus noise map .