Hidden truncation models have a long history before Azzalini (1985) popularized and studied the skew-normal (SN) distribution. Birnbaum (1950) studied the distribution and its extensions in the context of educational testing where he showed that the SN distribution can result from linear truncation of multivariate normal random variable. Weinstein (1964), using a convolution of normal and truncated normal random variables derived a distribution similar to SN, expressed implicitly. Roberts (1966) considered the distribution resulting from selecting the maximum or minimum value from suitably standardized measurements taken on a pair of twins. The resulting distribution is also similar to the SN distribution. In the Bayesian context, O’Hagan and Leonard (1976) suggested the use of an extended version of SN distribution as a possible prior for a normal mean. The above early studies showed that simple and common nonlinear operations such as truncation, conditioning and censoring carried out on normal random variables lead to versions of SN random variables.