Neglect patients, when asked to bisect a horizontal line, typically show large rightward errors with long lines and a decreased error with medium length lines. With very short lines the bisection bias reverses from the right to left side of the line physical centre (the so-called crossover effect). It is commonly pointed out that such a leftward bias is difficult to explain by traditional theories of neglect. Several accounts propose two distinct mechanisms, one that works for short lines and one that works for long. In the present study we demonstrated that the crossover effect can be explained by means of a unitary mechanism that derives from the space anisometry hypothesis. This hypothesis postulates that in neglect patients representational space is progressively 'relaxed' contralesionally and progressively 'compressed' ipsilesionally. In a series of five experiments, we investigated the crossover effect in 26 right-brain damaged patients: 17 with neglect without hemianopia, 4 with neglect and hemianopia and 6 without neglect or hemianopia. Patients were to bisect or extend lines of objectively and subjectively different lengths. The modulation of subjective length was created by an Oppel-Kundt illusion that is thought to resemble the distortion of representational space that occurs with neglect. All groups, except for the patients with neglect and hemianopia, were prone to the illusion. The rightward bias was reduced when the illusion induced a perceptual distortion opposite to that thought to underlie neglect. Importantly, the strength of the illusion decreased with reducing the physical line length and reversed with very short lines. These results argue for a simple and unitary explanation of the crossover effect in spatial neglect within the framework of the space anisometry hypothesis.