Using a probe-recognition technique the signal detection theory parameters d' and Beta were estimated for three types of probe (common surnames, uncommon surnames and synonyms) for material contained in a prose passage. Subjects were presented with the prose passage either in the presence of noise (85dBA) or in quiet (60dBA). In two experiments the effects of noise on auditory and visual presentation of the passage were studied. In both cases the recognition test took place in quiet. Noise decreased values of Beta for rare names and increased Beta for common names in both auditory and visual versions of the task. Noise influenced d' values in the audi tory version only, with d' increasing for common names in loud noise. The results support the view that noise influences performance by disturbing the pigeon-holing mechanism with the qualification that when material may not be recapitulated (as in the auditory presentation in the present study) greater attention may be allocated to easily recognizable material. The findings gi ve li ttle support to theories of noiseinduced deficits in performance based on the masking of inner speech.Theoretical views of the way in which noise influences performance are particularly polarized in regard to the effects on memory. On the one hand there are those who argue that noise influences factors such as the allocation of attention or the adoption of strategies (e. g., Broadbent, 1981), while on the other there are those who posit that noise has its action through the masking of inner speech (e.g., Poul ton, 1979). Problematical for the latter view are studies which (a) find noise improves performance, (b) find two opposite effects of noise, and (c) find the effect extending beyond the period of noise exposure. One study showing all three effects is by Schwartz (1974). The aim of the present small-scale study is to examine further the details of the phenomena to be observed. Schwartz (1974) found that noise influenced the signal detection parameters d' and Beta in a probed-recogni tion test of memory for surnames. Subjects were asked to give confidence judgements to probe items in a transcript of a passage originally