In distributed graph algorithms, a key computational resource is the communication radius of the algorithm, i.e., its locality. The class LD captures the distributed languages that can be decided by a local algorithm; its nondeterministic analog is the class NLD, which captures the distributed languages that can be decided by a local algorithm with local advice. Inspired by the polynomial hierarchy in complexity theory, this has been further extended into a hierarchy of local decision, where each node runs an alternating Turing machine. However, in prior work, the computational efficiency of each network node as nodes where allowed to run Turing machines for unbounded number of steps. This results in some undesirable and unanticipated properties: for example, the class NLD includes some Turing-undecidable languages.