ABR measurements in the cat using a forward-masking paradigm. Probe-elicited wave V amplitudes of the auditory brainstem response (ABR) were measured using a forward-masking paradigm. Subjects were anesthetized cats. For individual experiments, probe frequency and intensity were fixed and masker frequencies and intensities were varied. For each masker frequency, the extent to which the probe-elicited wave V amplitude was reduced by the preceding masker was plotted as a function of masker intensity. The rising segments of the masking functions were fitted with straight lines, using a least-squares procedure, to obtain estimates of their slopes. Masking grew most rapidly for masker frequencies below probe frequency, becoming progressively less steep as masker frequency increased. ABR tuning curves were constructed by using the linear fits to define the masker intensity that caused a 50% reduction in probe-elicited wave V amplitude. The shapes of these tuning curves were comparable to whole-nerve action potential (AP) tuning curves obtained under similar stimulus conditions.

Intensity and frequency resolution: masking of absolute identification and fixed and roving discrimination. Auditory intensity and frequency resolution were studied in three paradigms under masking conditions. Absolute identifications of single stimuli (one-interval paradigm) and 2IFC judgments of fixed- and roving-level pairs of stimuli (two-interval paradigm) were obtained from the same experienced observers. Judgments were made under optimal (no mask) conditions, in the presence of a broadband noise mask (simultaneous mask), and when the stimulus(i) to be judged were either preceded (forward mask) or followed (backward mask) by a broadband noise mask. Substantial masking of intensity resolution was found in all mask conditions. Only a simultaneous mask affected frequency resolution. In the no mask condition, performance was best for fixed-level (or frequency) 2IFC discrimination, followed by roving-level (frequency) 2IFC, and finally absolute identification. These differences were maintained under masking for frequency resolution, but not for intensity resolution. The results are discussed in terms of the Braida and Durlach (1988) model of intensity resolution.

Nonsimultaneous maskers can strongly impair performance in an auditory intensity discrimination task. Using methods of molecular psychophysics, we quantified the extent to which (1) a masker-induced impairment of the representation of target intensity (i.e., increase in internal noise) and (2) a systematic influence of the masker intensities on the decision variable contribute to these effects. In a two-interval intensity discrimination procedure, targets were presented in quiet, and combined with forward maskers. The lateralization of the maskers relative to the targets was varied via the int In addition, the finding that backward maskers presented after the target cause the same effect as forward maskers [2] , [4] , [8] is difficult to explain by processes in the auditory periphery and suggests an involvement of more central mechanisms [9] . Several explanations for the effects of non-simultaneous masking have been proposed (an in-depth discussion can be found in [3] ). Zeng et al. [7] suggested that the relatively slow recovery of low spontaneous-rate neurons in the auditory nerve [10] creates a "coding gap" for midlevel standards if an intense forward masker is presented. For example, a forward masker could cause response suppression in auditory nerve neurons (e.g., [19] ), resulting in a lower spike count produced in response to the target, relative to a situation where the target is presented in quiet. At higher processing stages (e.g., primary auditory cortex), enhancement rather than suppression of the neural response by a preceding sound is sometimes observed [20] , although as will be discussed below it is unlikely that the stimulus configuration used in typical experiments on intensity discrimination under nonsimultaneous masking would lead to response enhancement. In the present experiment, methods of "molecular psychophysics" [18] , also known as perceptual weight analysis or behavioral reverse correlation [26] , [27] , [28] , [29] , provided a rich set of behavioral measures that made it possible to test which of the three effects play a role for intensity discrimination under forward masking. In a two-interval intensity discrimination task (see Figure 1 ), we imposed random trial-by-trial variation on the levels of the maskers presented in the first interval ( L M1 ) and in the second observation interval ( L M2 ). As a control condition, the tones were additionally presented monaurally, either to the left or to the right ear. Three ratings were obtained per condition, in random order. Next, the six different masking conditions from the intensity discrimination task were presented. On each trial, the listener heard three trials as in the intensity discrimination task, and first rated the perceived lateralization of the target tones, and then the perceived lateralization of the maskers. Three ratings of target and masker lateralization were obtained per condition, in random order. These DLs were used to select individual intensity increments for the main task. Additionally, detection thresholds in quiet and under forward masking were obtained (in each session: one block per condition, random order). In sessions 9 to 17, the intensity discrimination task with random perturbation of the target levels and a fixed intensity increment was run (six randomly selected conditions per session with the restriction that no condition was presented more than once per session, in random order). At this point, the potential role of persistence of neuronal activation should be considered, which denotes neuronal responses continuing even after the termination of the sound For a conditioner stimulus preceding the target stimulus in time, this enhancement might be caused by persistence of activation, that is, by residual activation caused by the conditioner in the temporal window in which the neural responses to the target are measured. However, persistence can, by definition, only account for effects of forward maskers, not for effects of backward maskers. Therefore, given the very similar effects of forward and backward maskers on intensity discrimination discussed above, it is unlikely that persistence plays an important role for effect A). In contrast, the DL elevation was not correlated with increases in internal noise. Why should it be difficult to selectively attend to the targets, given that at masker-target inter-stimulus intervals (ISIs) of 100 ms or more as in previous experiments the masker and the target presented in each observation interval are clearly perceived as two separate tones [59] ? Oberfeld and Stahn [4] argued that this can be explained by the temporal structure of a two-interval intensity discrimination task. In this situation, according to the concept of object-based attention, it should be difficult to selectively attend to one feature (target intensity) of the auditory object presented in interval 1 while ignoring another feature (masker intensity) of this same object. Two recent studies from our lab supported this hypothesis. In one of the experiments [4] , we compared the DL elevation in the usual two-interval intensity discrimination task under forward masking to a condition where a brief sequence of three forward-maskers was presented in each observation interval. The inter-stimulus intervals between the three maskers were much smaller than the ISI between the target and its temporally most adjacent masker. Therefore, within each observation interval we expected the three forward-maskers to be grouped together on the basis of temporal proximity and thus to be processed as one auditory object, while the target should be perceived as a separate object. This should facilitate selective attention to the targets. Compatible with this hypothesis, the masker-induced DL elevation was significantly smaller in the condition presenting three maskers per interval than in the usual forward-masking condition where only one masker was presented per interval.