The correlation coefficient decreased because we added an ``untypical'' subject, with a low ACHSCOR and a high INTEL score. Adding the point on the scatterplot, it is clear that we will need to make far greater compromises in fitting a line to the new data than the original one. 
 MTB >  insert 0 1 'ACHSCOR' 'INTEL'
 DATA> 20 30
 DATA> end
       1 ROWS READ
 MTB >  corr 'ACHSCOR' 'INTEL'

 Correlation of ACHSCOR and INTEL = 0.287