This study examined the effect of high and low anxiety provoking instructions in subjects submitted to a Hyperventilation Provocation Test (HVPT). Subjects were 43 out-patients referred to our clinic for a diagnostic examination of Hyperventilation Syndrome (HVS). Results showed that anxiety levels were affected by the instruction manipulation, but the magnitude of this effect was less than expected and the instruction manipulation had no effect on intensity and type of reproduced symptoms, nor on symptom recognition. Subjects who met Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1 987) criteria for Panic Disorder (PD) were not more responsive to the instruction manipulation than non-PD patients. It is argued that the small effect of the manipulation is probably not due to the solidity of the HVPT but to the pervasiveness of pretest cognitions and expectations. In line with this, the report of HVS symptoms appeared highly related to psychological trait measures like anxiety, fear of bodily sensations, and a general tendency to report somatic symptoms.