The effectiveness of two types of tonal cues for reducing frequency uncertainty was studied in a tonal detection-in-noise task. Signals varied at random from trial to trial over the range 750–3000 Hz. The three conditions included: (1) maximum uncertainty in which there were no cues; (2) minimal uncertainty in which “iconic cues” were identical to the signal to be detected; and (3) partial uncertainty in which “relative cues” were set to 2/3 of the signal frequency, i.e., at the musical 5th. Results show that relative cues and iconic cues were both effective in reducing uncertainty compared to the no-cue condition, but that performance with relative cues was poorer than with iconic cues by 1.4 dB. In addition, a modified probe-signal method was used to estimate the widths of the subjective listening bands. Application of a model of the auditory filter [R. Patterson and B. C. J. Moore, Frequency Selectivity in Hearing, edited by B. C. J. Moore (Academic, New York, 1986)] to these data showed that the subjective listening bands used with iconic cues were similar in width to typical measures of the critical band but that the bands used with relative cues were wider by a factor of roughly 1.6.