Pure alexia (PA) is characterised by abnormally strong length effects in word reading times. It is often thought to result from damage to visual processing. This visual damage was also found to cause impaired object recognition performance (e.g., Roberts et al., 2012), suggesting a general visual deficit. Many computational models of reading have successfully simulated different forms of acquired dyslexia (e.g., Coltheart et al., 2001; Plaut et al., 1996). However, an adequate computational account of pure alexia has yet to be produced. We developed a large-scale and complete connectionist model with asymmetric hemisphere processing to support both word and object recognition. When damage was applied to the left visual processing layer in the model, the model produced abnormal length effects and impaired object recognition similar to those seen in PA patients. The results provide evidence to support the view of a common visual processing in visual word recognition.