Over a 21-month period, 16 patients with possible scaphoid fractures, where conventional radiographs were essentially normal on the day of injury and 10 days later, were examined by coronal short tau inversion recovery (STIR) and T1 weighted sequences on a 0.2 Tesla scanner. The magnetic resonance (MR) examinations were done in between electively booked patients and management was altered in 50% of patients. It was expedient to do MR as radio-isotope facilities were not available at the hospital. In four patients, the examination was entirely normal. In another four patients, intercarpal fluid collections only were demonstrated (two generalized and two localized). In these patients a plaster cast was not re-applied. In two patients, bony injuries of the distal radius were demonstrated instead, with the scaphoid being normal. In the remaining six patients, variable appearances were shown, ranging from an actual fracture line demonstrated on both STIR and T1-weighted images with the remainder of the scaphoid being hyperintense on STIR and hypointense on T1, to non-visualization of the actual fracture line, which appeared to be represented by a much more hyperintense band through the waist of an otherwise hyperintense scaphoid on the STIR sequence. In all these patients a plaster cast was re-applied empirically, including those patients with a definite fracture and those patients with apparent bone contusion only. Although the T1-weighted images were anatomically superior, their addition did not alter management and it is suggested that coronal STIR images should suffice to demonstrate occult fractures of the scaphoid.