• Three experiments manipulating visual feedback while typing single words. • Error detection moderately impacted by visual feedback removal. • Error correction relying heavily on visual feedback. • Positional information helpful, but letter identity essential for corrections. • Application of signal detection theory to multichannel language monitoring. This study examined the role of external information in monitoring language production. In a typing-to-dictation task, participants were deprived of all or part of visual feedback. Data were analyzed using signal detection theory (SDT) applied to a multi-component monitoring framework. Results showed that removing the visual information affected the correction of typing errors more than their conscious detection (Exps 1, 2). Reinstating partial visual information (positional information) increased correction rates but not to the level of full visual information, independently of the probability of error detection (Exps 2, 3). Analysis of SDT parameters showed that while manipulating visual information affected the informativeness of the signal for both correction and conscious detection of errors, participants treated this change differently in the two tasks. We discuss the implications of the results, and more generally, the utility of SDT, for theories of monitoring and control in language production.