Akaike discusses the role of the likelihood function in Bayesian statistics, emphasizing its importance as the primary source of objectivity. He argues that the choice of the prior distribution should be based on the expected behavior of the likelihood function. The paper highlights the practical utility of Bayesian methods through examples, including seasonal adjustment of time series. It critiques the subjective interpretation of Bayesian procedures, pointing out conceptual difficulties and advocating for a more objective approach based on likelihood. The likelihood function is seen as essential for defining Bayesian procedures and for evaluating the performance of prior distributions. The paper also presents a general Bayesian model for linear problems, demonstrating its application in various statistical estimation techniques. Numerical examples illustrate the effectiveness of Bayesian methods, including polynomial fitting and seasonal adjustment, showing how they can handle complex models with many parameters. The discussion emphasizes the importance of using the likelihood function to guide the choice of priors and to evaluate the performance of Bayesian models. The paper concludes that Bayesian methods, while rooted in subjective probability, can be practically useful when guided by the objective measures provided by the likelihood function.Akaike discusses the role of the likelihood function in Bayesian statistics, emphasizing its importance as the primary source of objectivity. He argues that the choice of the prior distribution should be based on the expected behavior of the likelihood function. The paper highlights the practical utility of Bayesian methods through examples, including seasonal adjustment of time series. It critiques the subjective interpretation of Bayesian procedures, pointing out conceptual difficulties and advocating for a more objective approach based on likelihood. The likelihood function is seen as essential for defining Bayesian procedures and for evaluating the performance of prior distributions. The paper also presents a general Bayesian model for linear problems, demonstrating its application in various statistical estimation techniques. Numerical examples illustrate the effectiveness of Bayesian methods, including polynomial fitting and seasonal adjustment, showing how they can handle complex models with many parameters. The discussion emphasizes the importance of using the likelihood function to guide the choice of priors and to evaluate the performance of Bayesian models. The paper concludes that Bayesian methods, while rooted in subjective probability, can be practically useful when guided by the objective measures provided by the likelihood function.