The classical experiment on masking of pure tones by white noise by Hawkins and Stevens [J. Acoust. Soc. Am. 22, 6-13 (1950)] has left two general notions about broadband masking that are still being echoed in present-day textbooks on hearing. One is that broadband masking is linear (i.e., if the noise power level increases, the signal threshold level increases by exactly the same amount). The other is the existence of a nonmonotonic function of frequency, the critical ratio (CR) which, unlike the critical band (CB) or equivalent rectangular bandwidth (ERB) functions, shows a distinct upturn below 300 Hz. Exploring the possibility that Hawkins and Stevens' measurements may have been adversely affected by acoustic leakage at the headphones at low frequencies, the experiment was repeated with insert earphones. The resulting critical ratio function shows a monotonic dependence on frequency down to 90 Hz, similar to CB and ERB functions. Moreover, the CR function appears to closely approximate the ERB function when the dependence of each filter's detection efficiency on masker energy uncertainty is incorporated. Masking appears to be linear below 1000 Hz, but thresholds, expressed as S/N0 ratios, were found to increase systematically with increasing masker intensity at higher frequencies.