Based on contemporary theories of learning, this article takes a critical view of the concept of learning objects. Learning objects decompose content into granular pieces of information that can be stored, retrieved, and reused in instruction. Current conceptualizations of learning objects support traditional, objectivist forms of instruction. While there are no implicit restraints on the concept of learning objects in terms of their complexity, interactivity, and cognitive functionality, the current industry standards cannot describe the rich interactions necessary for meaningful learning, such as problem solving. Also, the metadata used to describe them provides very little useful information for the instructional designer deciding how to use learning objects. We argue for richer, multidimensional conceptions of learning objects, including information objects, conversation objects, learning objects, thinking objects, knowledge objects, and activity objects. Additionally far richer metadata are needed to describe learning functions, purpose, and outcomes.