Automatically understanding the rhetorical roles of sentences in a legal case\njudgement is an important problem to solve, since it can help in several\ndownstream tasks like summarization of legal judgments, legal search, and so\non. The task is challenging since legal case documents are usually not\nwell-structured, and these rhetorical roles may be subjective (as evident from\nvariation of opinions between legal experts). In this paper, we address this\ntask for judgments from the Supreme Court of India. We label sentences in 50\ndocuments using multiple human annotators, and perform an extensive analysis of\nthe human-assigned labels. We also attempt automatic identification of the\nrhetorical roles of sentences. While prior approaches towards this task used\nConditional Random Fields over manually handcrafted features, we explore the\nuse of deep neural models which do not require hand-crafting of features.\nExperiments show that neural models perform much better in this task than\nbaseline methods which use handcrafted features.\n