Cluster headache (CH) is diagnosed according to criteria of the International Headache Society (IHS), but, in clinical practice, these criteria seem too restrictive. As part of a nation-wide study, we identified a group of patients who met all criteria minus one (IHS-CH-1), and assessed in which way they differed from CH patients meeting all criteria (IHS-CH). We performed a nation-wide questionnaire study for CH and CH–like syndromes, including questions based on the IHS criteria, and additional features such as restlessness during attacks, nocturnal onset of attacks, circadian rhythmicity of attacks and response to treatment. IHS-CH and IHS-CH-1 patients were compared. Of 1452 responders to two questionnaires, 1163 were IHS-CH and 289 were IHS-CH-1. The majority of the IHS-CH-1 patients were classified as such because their attacks exceeded 3 h (64%, median attack duration: 5 h), or came in a frequency of less than 1 per 2 days (16%). Age at onset was similar between the groups. The male to female ratio was 3.7 : 1 in the IHS-CH group and around 1.6 : 1 in the IHS-CH-1 groups (P < 0.005). Patients with attacks exceeding 3 h less often reported a circadian rhythmicity (IHS-CH-1: 49%, IHS-CH: 64%), episodic periodicity (IHS-CH-1: 65%, IHS-CH: 78%), nocturnal attacks (IHS-CH-1: 67%, IHS-CH: 78%), smoking (IHS-CH-1: 90%, IHS-CH: 80%) and restlessness during attacks (IHS-CH-1: 64%, IHS-CH: 76%) than IHS-CH patients (P < 0.005). Photo- or phono-phobia (IHS-CH-1: 67%, IHS-CH: 54%) and nausea (IHS-CH-1: 38%, IHS-CH: 27%) were more frequently reported by patients who reported to have attacks exceeding 3 h (P < 0.005). Similar proportions reported effect of verapamil on their attacks (IHS-CH-1: 54%, IHS-CH 61%). We conclude that average attack duration exceeding 3 h was frequently the reason for not fulfilling IHS CH criteria. Symptoms often accompanying CH such as restlessness, nocturnal attacks and an episodic attack pattern were relatively frequently present in IHS-CH-1 patients with longer attacks. These patients may therefore be diagnosed with CH. Attack frequency may not be a useful criterion for the diagnosis of CH. The upper limit of 3 h should be increased in future diagnostic criteria.