The relationship between succession and tropical forest diversity has been much debated. A fundamental disagreement hinges on whether high local species richness is a transient successional property, albeit one that can be maintained by disturbance, or is rather a property of stable late successional communities. This paper addresses this controversy employing a series of long-term permanent sample plot data spanning seven decades. W.J. Eggeling studied the vegetation of Budongo Forest, Uganda during the 1930s and 1940s. He described a series of ten plots (1.4 and 1.86 ha) as a successional progression of forest types in which tree species numbers show a unimodal rise-and-fall over time - a pattern best known from Connell's illustration of his intermediate disturbance hypothesis. Tree communities in five of the original plots have been intermittently re-assessed over the subsequent decades. One data-series provides observations spanning 54-years from one intact ‘undisturbed' old-growth forest plot. The remaining four plots were assessed before and after controlled disturbances (tree poisoning) executed in the late 1950s and early 1960s, and the resulting data-series span c. 20 years of pre-disturbance and c.35 years of post-disturbance changes. The unimodal pattern of species-richness in the original comparative plot-series is paralled by a similar rise-and-fall in stem-densities, but rarefaction confirms that the unimodal pattern in richness also holds for fixed stem-counts. The proportion of species occurring in both large and small stem-size-classes increases across the series. As richness declines in later succession, low abundance species occur predominantly in larger stem-sizes. All time-series show a rise in species richness ranging from 12 to 177 % ha¯¹ (over 50-60 years). Each of the disturbed plots ultimately reaches greater richness than was recorded anywhere in Eggeling's original series. Contrary to expectation a small rise was also recorded in the undisturbed late successional plot (c.42 species >= 10 cm diameter ha¯¹ , rising to c.47). The lowest species density observed in the study is a 1940s record of c. 10 species >= 10 cm diameter ha¯¹ in monodominant Cynometra [Caesalpinoidae] forest and the highest record is c. 61 ha¯¹ recorded in 1992, in the youngest vegetation type monitored. These observations indicate both the volatile nature of tree-richness patterns and the limitations of simple models as aids to interpretation when confronted with real patterns of long-term change.