With various reaction time paradigms, panic patients have been shown to have selective attention for threatening sensations. However, almost all of these paradigms use words describing sensations and not the threatening sensations themselves. To increase the ecological validity, in the current study selective attention for heartbeat information was measured with a paradigm derived from the dot probe detection task but using 'real' heartbeat information instead of words. The results indeed showed selective attention for ECG information in panic patients. However, an accelerated ECG did not attract the attention of panic patients more than a slower ECG. Against expectation, both panic patients and normal controls shifted their attention away from an accelerated ECG. Possible explanations are explored.