Mutual dependencies between objects arise frequently in programs, and programmers must typically resort to manually filling “initialization holes” to help construct the corresponding object graphs, i.e. null values and/or explicitly mutable locations. This report describes a “base-line” proposal for a generalized form of value recursion in an ML-like language called initialization graphs, where value recursion is given the simplistic semantics of a graph of lazy computations whose nodes are sequentially forced, with uses of recursive values checked for initialization-soundness at runtime. We then develop examples using this mechanism to show how problematic the issue of value recursion is for ML-like languages, and in particular how sophisticated reactive objects cannot be defined in the language without using initialization holes, and how this forces ML programmers to break abstraction boundaries. At the same time we show how OO languages rely extensively on null pointers during initialization. We propose that a general, semi-safe mechanism allows value recursion to be used in conjunction with existing sophisicated abstract APIs such GUI libraries, and allows freshly defined APIs to be both abstract and yet not require clients to use explicit initialization holes. We propose that the initialization mechanism permits more programs to be expressed in the mutation-free fragment of ML, though we do not formally prove this result.