Abstract. I compare the burst detection sensitivity of CGRO's BATSE, Swift's BAT, the GLAST Burst Monitor (GBM) and EXIST as a function of a burst's spectrum and duration. A detector's overall burst sensitivity depends on its energy sensitivity and set of accumulations times ∆t; these two factors shape the detected burst population. For example, relative to BATSE, the BAT's softer energy band decreases the detection rate of short, hard bursts, while the BAT's longer accumulation times increase the detection rate of long, soft bursts. Consequently, Swift is detecting long, low fluence bursts (2-3× fainter than BATSE).What is the relative sensitivity of different detectors for detecting gamma-ray bursts, and how should this sensitivity be compared? How do these differences shape the observed burst populations, which must be taken into account in determining the underlying burst distribution? Here I compare BATSE's Large Area Detectors on CGRO (the past), the Burst Alert Telescope (BAT) [1] on Swift (the present), and the GLAST Burst Monitor (GBM) and EXIST (the future). BATSE and the GBM are/were sets of NaI(Tl) detectors while the BAT and EXIST are/will be CZT coded mask detectors. The energy range of NaI(Tl) detectors is ∼20-1000 keV while for CZT it is ∼10-150 keV. I apply a semi-analytic methodology using simplified models of the trigger systems of the different detectors.Most instruments detect bursts using either rate triggers or image triggers. A rate trigger determines whether the increase in the number of counts in a time bin ∆t and energy band ∆E over the expected number of background counts is statistically significant. An image trigger determines whether the image formed from the counts in the time bin ∆t and energy band ∆E contains a new point source. Usually an image trigger is preceded by a rate trigger that starts the imaging process[2]; the rate trigger is set to permit many false positives that are eliminated by the image trigger. If the number of burst counts is S and the number of non-burst counts is B, then the rate trigger significance σ r (for BATSE and the GBM) and the image trigger significance σ i for ∆t and ∆E are σ r = S √ B andwhere f c accounts for the finite size of the detector pixels. For Swift f c ∼0.7[3], which explains why for a given burst the rate trigger significance is greater than the imaging significance [4]. The BAT uses a more complex rate trigger than shown above. For directions other than the burst position the counts S from the burst contribute to the