Identification of the second of two targets (T2) is impaired when presented shortly after the first (T1). This "attentional blink" (AB) is thought to arise from a delay in T2 processing during which T2 is vulnerable to masking. Conventional studies have measured T2 accuracy which is constrained by the 100% ceiling. We avoided this problem by using a dynamic threshold-tracking procedure that is inherently free from ceiling constraints. In two experiments we examined how AB magnitude is affected by three masking-related factors: (a) presence/absence of T2 mask, (b) T2-mask stimulus onset asynchrony (SOA), and (c) level of T2 impoverishment (signal-to-noise ratio [SNR]). In Experiment 1, overall accuracy decreased with T2-mask SOA. The magnitude of the AB, however, was invariant with SOA and with mask presence/absence. Experiment 2 further showed that the AB was invariant with T2 SNR. The relationship among mask presence/absence, SOA, and T2 SNR and the AB is encompassed in a qualitative model. (Contains 5 figures and 1 footnote.)