Approximate estimates of prices, grades, temperatures, sales figures, and other uncertain target amounts (T) can be expressed as uncertainty intervals. In daily life, often only one interval limit (lower or upper) is suggested. Such estimates contain both quantitative and pragmatic (qualitative) information. Limit values function as reference points: lower limits suggest downward comparisons and upper limits suggest upward comparisons, leading, by contrast, to high or low evaluations of T. A series of vignette studies demonstrated inferences people draw from upper and lower limit estimates, partly about the speakers' attitudes and expectations, and partly about the relative magnitude of T. Thus temperatures "above X degrees" and sales "over X million" imply increasing temperatures and growing sales, even with an unknown X. It follows that lower limit predictions will, occasionally, exceed upper limit predictions.