The noise criterion (NC) curves were developed both as a method for evaluating existing noise situations and as a means for specifying design goals for noise control. They have also influenced the N ratings used abroad. It has been demonstrated that if a background noise whose spectrum conforms to an NC‐curve shape is deliberately generated, it does not sound to a listener as a pleasant or neutral noise, but is both “hissy” and “rumbly.” A recent tendency in consulting practice, therefore, has been to specify noise levels that are lower than NC curves at both low and high frequencies. Also, the original NC curves were derived using Stevens's Mk. I method of calculating loudness level from noise levels measured with the old octave bands, the lowest of which was the “below 75 Hz” band. A reexamination of the noise criterion curves has been undertaken using Stevens's Mk. VI method and the new octave bands. The result is a new set of curves, called “preferred noise criterion (PNC) curves,” having levels in the bands below 125 Hz and above 1000 Hz that are lower than those of the 1957 NC curves by 2–5 dB. The new set also takes into account the engineering realities of achieving the specified noise levels with conventional air‐handling equipment. The results of a number of recent noise‐control projects in building are presented for comparison.