The answer is true. The linear regression analysis necessitates all variables to be multivariate normal. This supposition can best be checked with a histogram. Normality can be checked with a goodness of fit test. When the data is not normally distributed a non-linear transformation can answer this issue.