Perpetual January data from a simple global two-level model are used to diagnose the forcing of the model's low-frequency (LF), large-scale flow. The forcing is considered within the context of a hypothetical low-order model that resolves only the slowly varying planetary-scale modes. Thus, the forcing is partitioned into “resolved” components consisting of LF, planetary-scale (i.e., autonomous) processes and “unresolved” components that involve high-frequency (HF) transient eddy and LF synoptic-scale effects. Spatial patterns of the net tendencies of low-pass filtered, planetary-scale barotropic streamfunction and vertically averaged temperature are compared with corresponding patterns of resolved and unresolved forcing during the onset, maintenance, and decay stages of composited persistent anomalies. Growth stages of positive and negative persistent anomalies that are found in both the Atlantic (ATL) and Pacific (PAC) regions are dominated by autonomous processes. For positive anomalies, the net tendencies and resolved forcing patterns resemble large-scale wave trains emanating from the subtropics upstream of the anomaly region. Although only of secondary importance, unresolved forcing contributes positively toward growth just upstream of the ATL anomaly region, but does not contribute substantially in the PAC region. Negative anomaly growth is dominated in both regions by autonomous processes to a greater extent than positive anomaly growth. Unresolved forcing is critical in mechanically maintaining persistent anomalies of both signs, but thermally acts in a destructive sense. Decay stages of all anomalies are dominated by resolved forcing. The extent to which local signature of barotropic and baroclinic energy sources accompany anomaly onset is examined by constructing these terms in the context of a local kinetic energy budget. Superficially, baroclinic (barotropic) conversion tends to be associated with ATL positive (negative) anomaly onset. However, the kinetic energy source patterns are statistically less significant than, and do not typically resemble, the resolved forcing patterns. This is particularly true for the PAC composite anomalies. Clearly, redistribution terms in the local kinetic energy budget are important. But more important, the representativeness of composites in depicting the dynamics of persistent anomalies is called into question. An examination of the slowly varying local resolved and unresolved forcing during the life cycles of individual members of the positive anomaly composite shows considerable variability, suggesting that no single mechanism is solely responsible for anomaly development. For example, cases of anomaly growth characterized primarily by unresolved forcing are noted. Also seen are anomaly events that more closely resemble stationary large-scale waves.