Despite the oil industry's efforts in improving safety, it still presents a high rate of serious accidents, many involving human failure events (HFE), which can be identified, modelled, and quantified through human reliability analysis (HRA). The oil industry commonly analyzes process safety by focusing on technical barriers, and thus it could benefit from HRA. Phoenix methodology is an HRA method that uses a human response model and relates the crew failures modes (CFM) to performance influencing factors (PIFs). Based on Phoenix CFMs and PIFs, two refinery accidents, the BP Texas City (2005) and the Chevron Richmond (2012), are analyzed in this paper. The analysis consists of the construction of the accident timeline; identification of the HFEs and assigning them to appropriate CFMs; and, finally analysis of the PIFs. The analysis helped better understand how the operators responded to an abnormal condition of the process, and why they took the actions they did, investigating the contribution of human error to the accidents. The assessment of the role human error played in these accidents is a major contribution to the understanding of why they happened, and a key information to avoid the same happening again in the future. Moreover, the features and limitations of the application of Phoenix HRA, which was developed based mainly on nuclear power plant operations, to Oil Refinery operation scenarios, are discussed and evaluated. This article provides insights on value of investigating the potential impact of human error in the Petroleum Industry accidents.