In three experiments, listeners’ abilities to detect changes in randomly generated tonal sequences were determined for sequences or ‘‘patterns’’ ranging in total duration from 62.5 62.5 ms to 2 s. Experiment 1 utilized an adaptive‐tracking procedure, with n, the number of pattern components, as the dependent variable, and included a variety of spectral and temporal discrimination tasks with isochronous patterns. When the to‐be‐detected change was the only variation on a given dimension (e.g., the presence or location of a brief pause), patterns were discriminable when the absolute duration of the changed element, or pause, exceeded a critical value. However, when each pattern component varied on the dimension of the to‐be‐detected change (e.g., frequency), discriminability was strongly related to the number of tones in the pattern, and only weakly to the durations of either the target components or the total pattern. This dependence of discrimination performance on n was also demonstrated with anisochronous patterns in experiment 2. Experiment 3 revealed the same dependence of performance on the number of components per pattern as did experiments 1 and 2, but with Δf/f as the dependent variable, rather than n. The number of pattern components and the proportional duration of the target components, relative to total pattern duration, were confounded in these experiments. Additional research is therefore required to determine whether number or proportional target‐tone duration is the primary determinant of pattern discriminability.