Identification of objects in an applicable problem domain is the first step in object oriented requirements analysis. Natural language documents are an important source of information for this task, but most currently available tools operate only on formal specifications. Even those that purport to analyze natural language constrain the permissible forms of expression to such an extent that little if any advantage is gained over formal specification languages. Recognition and resolution of ambiguity have proven so difficult that automated interpretation of informal software requirements expressed in unconstrained natural language is not feasible. Heuristic recognition of object expressions in informal specifications, however, does not require logical interpretation based on truth valued semantics Two heuristic principles for object identification are described. The first holds that entities to be modeled as objects in an object oriented software design will likely be those most frequently referred to in the requirements documents. The second proposed heuristic is based on the conceptual similarities between objects in object oriented design and the agent and neutral entities of case grammar. A software object is an abstraction of an entity emphasizing its capacity to undertake actions that affect other entities in the application domain and to be similarly affected by other entities. In case grammar, the agent is the entity that initiates an action denoted by a verb, and the neutral is the entity affected by the action. Recognition of agent and neutral case fillers, then, is tantamount to identification of software objects