The only way to validate such observations is to check if they have predictive power. I'm sure the author checked that to her satisfaction. You are free to do the same.