Arthrodesis is the current treatment for end-stage hallux rigidus, although there are many reports advocating total joint replacement arthroplasty. The aim of this randomized controlled trial was to evaluate clinical outcomes after first metatarsophalangeal joint (MTPJ) arthrodesis and replacement arthroplasty. Between November, 1998, and January 2001, 63 patients between the ages of 34 and 77 years, with unilateral or bilateral MTPJ arthritis were recruited and randomly selected to have either MTPJ arthrodesis or arthroplasty. Twenty-two patients (38 toes) had arthrodesis and 27 patients (39 toes) had arthroplasty. A single surgeon performed all surgery. The primary outcome measure determining successful surgery was a decrease in pain as measured on a Visual Analogue Scale (VAS). Functional outcome was assessed at 6 months and 1 and 2 years. Cost data were simultaneously collected. At 24 months, pain improved in both groups (p < 0.001), but there were significantly greater improvements after arthrodesis (p = 0.01). All 38 arthrodeses united at a mean dorsiflexion angle of 26 degrees, with few complications. In contrast, in the arthroplasty group, six of the 39 inserted implants had to be removed because of phalangeal component loosening. In the remainder the range of motion gained was poor, and the patients tended to bear weight on the outer border of their foot. The cost ratio was 2:1 in favor of arthrodesis. Outcomes after arthrodesis were better than those after arthroplasty. The results were partially attributable to an unacceptably high incidence of loosening of the phalangeal components, which resulted in removal of the implants. However, even when data from the failures were excluded, arthrodesis was clearly preferred by most patients.