The purpose of this study was to explain differences in threshold of discomfort (TD) across complex stimuli using acoustic and sound-quality variables as predictors. Two 4-tone complexes and 10 environmental sounds were used as stimuli. The environmental sounds consisted of a baby crying, hairdryer blowing, glass breaking, jet engine propelling, person laughing, motorcycle accelerating, orchestra tuning, telephone ringing, siren blowing, and toilet flushing. One-third octave band (1/3 OB) spectra were obtained for the 12 stimuli, with overall rms amplitude held constant across the stimuli. Nine acoustic quantities describing the high- and low-frequency content, peakiness, and bandwidth of each 1/3 OB spectrum were defined. Twenty adult subjects, 6 men and 14 women, with normal hearing sensitivity participated in the study. TDs were obtained from each subject for all of the stimuli. In addition, subjects rated each of the stimuli for annoyance, harshness, loudness, noisiness, and tinniness on a 10-point scale. These ratings were completed at a level 10 dB below each subjects TD for that stimulus. A hierarchical multiple regression analysis was used to estimate the amount of variance in TDs accounted for by subject differences, defined acoustic properties, and sound-quality ratings. Results indicated that, after controlling for intersubject differences, acoustic and sound-quality factors significantly influenced TD ratings. Increases in high-frequency cutoff and low-frequency cutoff, as well as higher ratings of loudness and annoyance, were associated with higher TDs. Associated with lower TDs were an increase in the frequency of the primary spectral peak, an increase in the number of spectral peaks, an increase in the center frequency, an increase in the area under the frequency response curve 30 dB down from the peak amplitude, an increase in the calculated loudness level re ANSI S3.4-1980, and higher ratings of tinniness.