Suppose one measures two variables for each member of a group of individuals, and that the correlation coefficient of the variables is positive (negative). If the value of the first variable for that individual is above average, the value of the second variable for that individual is likely to be above (below) average, but by fewer standard deviations than the first variable is. That is, the second observation is likely to be closer to the mean in standard units. For example, suppose one measures the heights of fathers and sons. Each individual is a (father, son) pair; the two variables measured are the height of the father and the height of the son. These two variables will tend to have a positive correlation coefficient: fathers who are taller than average tend to have sons who are taller than average. Consider a (father, son) pair chosen at random from this group. Suppose the father's height is 3SD above the average of all the fathers' heights. (The SD is the standard deviation of the fathers' heights.) Then the son's height is also likely to be above the average of the sons' heights, but by fewer than 3SD (here the SD is the standard deviation of the sons' heights).