Aim/purpose-The purpose of the research is to verify the Capital Asset Pricing Model (CAPM) in the Polish capital market based on a conventional and downside risk approach. Design/methodology/approach-The author in this study, using individual securities and portfolios, compares the unconditional risk-return relationships with the conditional risk, estimated in up and down market using realised returns in cross-sectional regressions. Except for a beta coefficient, the CAPM is tested with co-skewness as a higher order co-moment and downside betas as a risk measure in a downside approach. Findings-The unconditional regressions give evidence of existing risk premium associated with co-skewness and downside beta, and confirmed the validity of the downside CAPM. The author, based on conditional relations, found that risk-return relations depend on the state of the stock market. The average premium for systematic risk in term of beta coefficient is significantly positive in up market periods and significantly negative in down market periods. The use of conditional models did not explicitly confirm the suitability of co-skewness in asset pricing. Research implications/limitations-The main implications include the fact that the conventional beta coefficient is an appropriate risk measure when we consider using it separately for up and down market. A valuable extension of this research would be a benchmarking analysis to compare results for the Polish capital market against other emerging and developed markets. Originality/value/contribution-The author in this paper proposes an alternative approach to testing risk-return relationships based on the CAPM in comparison to commonly used tests founded upon joint estimations of these relationships in periods of both positive and negative market excess return. The noteworthy contribution of this study is an application of the downside beta coefficient and the co-skewness coefficient in crosssectional regressions.