We show that smoothed versions of the high-resolution calibration curve should be used when ¹⁴C ages are calibrated with large (> ∼30 ¹⁴C yr) measurement errors (represented by standard deviation δ_m) or are mixtures of elements of variable age (natural sample error with standard deviation δ_n). The degree of smoothing should agree with the standard deviation of total sample error, δ_t, the square root of the quadratic sum of δ_m and δ_n. However, in most cases, δ_t is not well known, especially due to difficulties in quantifying δ_n. We present an inverse method that gives a measure of mean δ_t, for different materials that are widely used in (conventional) ¹⁴C dating. Calculations with large (>100) data sets of wood, charcoal, ombrotrophic peat and minerotrophic peat/gyttja samples indicate that δ_t of such materials is generally much larger than previously assumed, mainly because of large values of δ_n. This means that particularly in organic deposits, strongly smoothed calibration curves should be used where medium-term ¹⁴C variations (wiggles) are completely straightened. This has especially major consequences for calibrating ¹⁴C histograms for natural ¹⁴C variations. We conclude that ¹⁴C histograms consisting of samples of organic deposits do not require correction for medium-term ¹⁴C variations and that uncalibrated ¹⁴C histograms need not be as suspect as is usually believed.