In a recent article in this Review, Robert Fogel and Stanley Engerman (hereafter, F-E) defended their previous assertion that pre-Civil War southern agriculture was more efficient than its northern counterpart. The new study rebuked a number of critiques of their prior works (1971, 1974) and concluded that incorporation of suggested revisions actually widened the estimated differential in total factor productivity to 39 percent. The study also reiterated their finding that total factor productivity increased with the size of slaveholding. Using the Parker-Gallman cotton South sample for 1860, F-E estimated that farms with over fifty slaves enjoyed a level of efficiency over one-third higher than farms with no slaves. i The objective of this note is to demonstrate that F-E have misinterpreted the latter finding. They viewed the positive association between size of slaveholding and efficiency as an indicator of the proficiency of slave labor. In addition, they argued that this association indicated the existence of scale economies in production but that these economies only existed beyond some minimum slaveholding where the regimented gang labor system could be used. It is our contention that in making these assertions, F-E combined and confused two concepts-the number of slaves and scale. The latter term is generally defined in terms of output or the general level of inputs, while the former only provides an imperfect index of one input, labor. The two concepts are not necessarily consistent measures of size and it is therefore necessary to distinguish between them in the type of analysis that F-E pursued. The preferred distinction between the concepts implies that there are actually two issues to appraise. The first-the F-E hypothesisis whether larger slaveholdings resulted in higher productivity and the second is whether scale (as correctly measured by output) caused higher efficiency. Obviously, the two issues are not mutually exclusive, but it is our contention that F-E's failure to separate the issues led them to draw invalid inferences concerning the relationship between slavery and agricultural efficiency. In the analysis that follows we have employed the same data as F-E; namely, the Parker-Gallman sample for the cotton South in 1860. From this sample of 5,228 farms, we have excluded the same 929 farms as F-E so that we are dealing with exactly the same subsample. Further, we have defined land, labor, capital, and output using the measures put forward by F-E; in other words, the data in this paper are identical to those of F-E.