In most environments, the visual system is confronted with many relevant objects simultaneously. That is especially true during reading. However, behavioral data demonstrate that a serial bottleneck prevents recognition of more than one word at a time. We used functional magnetic resonance imaging (fMRI) to investigate how parallel spatial channels of visual processing converge into a serial bottleneck for word recognition. Observers viewed pairs of words presented simultaneously. We found that retinotopic cortex processed the two words in parallel spatial channels, one in each contralateral hemisphere. Responses were higher for attended words than ignored words, but were not reduced when attention was divided, even though behavioral performance suffered greatly. We then analyzed two word-selective regions along the occipito-temporal sulcus (OTS) of both hemispheres (i.e., sub-regions of visual word form area, VWFA). Unlike retinotopic cortex, each word-selective region responded to words on both sides of fixation. Nonetheless, a single region in the left hemisphere (VWFA-1 in posterior OTS) contained spatial channels for both hemifields that were independently modulated by selective attention. Thus, the left posterior VWFA supports parallel processing of multiple words. In contrast, a more anterior word-selective region in the left hemisphere (VWFA-2 in mid-OTS) showed limited spatial and attentional selectivity, consistent with activity of a single channel. Therefore, the visual system can process two words in parallel up to a late stage in the ventral stream. The transition from two parallel channels to a single channel in more anterior regions is consistent with the observed bottleneck in behavior.