In many practical situations, we have probability distributions which are close to normal but skewed. Several families of distributions were proposed to describe such phenomena. The most widely used is skew-normal distribution, whose probability density (pdf) is equal to the product of the pdf of a normal distribution and a cumulative distribution function (cdf) of another normal distribution. Out of other possible generalizations of normal distributions, the skew-normal ones were selected because of their computational efficiency, and not because they represent any real-life phenomena. Interestingly, it turns out that these distributions do represent a real-life phenomena: namely, in a natural simple model of scientists' strength, this strength is skew-normally distributed. We also describe what happens if we consider more complex models of scientists' strength.