Understanding patterns of species diversity and its scale effects is a central goal of ecology and a fundamental to the management and reservation of biodiversity. The relationship between patterns of species diversity and scale is also significant for understanding formation and maintenance mechanisms of species diversity. Species abundance and richness are the two most basic diversity variables, which represent two categories of variables that have a different spatial property. The former is additive when aggregated across spatial scale while the latter is non-additive. In other words, their usefulness in measuring biodiversity depends in large part on their response to scale effects. Therefore, it is essential to investigate the spatial variation of species abundance and richness and their differences across spatial scales in different forest types. The karst forest in Maolan National Natural Reserve of Guizhou Province, southwestern China is a rare, original forest remnant in the mid-subtropics of the world. In order to investigate the changes in the spatial patterns of abundance and richness of species across multiple scales in karst forest, two 1hm² forest plots in the core zone of Maolan Reserve were established. All woody species with the diameter at breast height (DBH) ≥1cm were mapped, identified to species and measured. Two 1hm² plots were divided into a grid system using five quadrat sizes: 5 m×5 m (400 quadrats), 10 m×10 m (100 quadrats), 20 m×20 m (25 quadrats), 25 m×25 m (16 quadrats) and 50 m×50 m (4 quadrats), and then the species abundance and richness in each quadrat were counted for each grain size. Spearman rank correlation coefficients were calculated to assess the relationship between abundance and richness at different grain sizes. In the meantime, spatial variance and coefficient of variation (the standard deviation divided by the mean, CV) of abundance and richness were also calculated. And then, the variance and CV were plotted against the grain size in order to examine the effect of spatial scale on the spatial variability in abundance and richness. The results showed that: 1) species abundance and richness were dependent on spatial scales in mixed evergreen-deciduous broad-leaved karst forests. 2) Species abundance showed a linear relationship with scale due to the additive nature, variance of the species abundance linearly increased with scales, whereas the coefficient of variation decreased. 3) Variance of the species richness was unimodal with increasing scale, and was maximum at the 25 m×25 m scale, the coefficient of variation of species richness decreased with increasing scale. In all, with respect to additive variables, it is relatively easy to extrapolate them from one spatial scale to another spatial scale, as they and the spatial scale usually form a linear relationship. In contrast, non-additive variables are difficult to extrapolate across spatial scales, because they often respond nonlinearly to spatial scale changes. All these indicated that large-scale species abundance can be estimated though extrapolation from smaller scales, whereas richness can not be estimated. That meant high species richness observed at a particular spatial scale may became low at another spatial scale in mixed evergreen-deciduous broad-leaved karst forests. This was because the fine-scale species richness could not be simply added when scaling up. This inconsistency made it quite difficult to extrapolate species richness from one spatial scale to another spatial scale. Our results suggested that it is necessary to take spatial scale into account when making plans for the diversity management and conservation.