Our aim was to compare the clinical features of panic disorder (PD) patients sensitive to hyperventilation or breath-holding methods of inducing panic attacks. Eighty-five PD patients were submitted to both a hyperventilation challenge test and a breath-holding test. They were asked to hyperventilate (30 breaths/min) for 4 min and a week later to hold their breath for as long as possible, four times with a 2-min interval. Anxiety scales were applied before and after the tests. We selected the patients who responded with a panic attack to just one of the tests, i.e., those who had a panic attack after hyperventilating (HPA, N = 24, 16 females, 8 males, mean age ± SD = 38.5 ± 12.7 years) and those who had a panic attack after breath holding (BHPA, N = 20, 11 females, 9 males, mean age ± SD = 42.1 ± 10.6 years). Both groups had similar (chi² = 1.28, d.f. = 1, P = 0.672) respiratory symptoms (fear of dying, chest/pain disconfort, shortness of breath, paresthesias, and feelings of choking) during a panic attack. The criteria of Briggs et al. [British Journal of Psychiatry, 1993; 163: 201-209] for respiratory PD subtype were fulfilled by 18 (75.0%) HPA patients and by 14 (70.0%) BHPA patients. The HPA group had a later onset of the disease compared to BHPA patients (37.9 ± 11.0 vs 21.3 ± 12.9 years old, Mann-Whitney, P < 0.001), and had a higher family prevalence of PD (70.8 vs 25.0%, chi² = 19.65, d.f. = 1, P = 0.041). Our data suggest that these two groups - HPA and BHPA patients - may be specific subtypes of PD.