The extended time-based resource-sharing (TBRS) model suggested a working memory architecture in which an executive
loop and a phonological loop could both support the maintenance of verbal information. The consequence of such a
framework is that phonological effects known to impact the maintenance of verbal information, like the word length effect
(WLE), should depend on the use of the phonological loop, but should disappear under the maintenance by the executive
loop. In two previous studies, introducing concurrent articulation in complex span tasks barely affected WLE, contradicting
the prediction from the TBRS model. The present study re-evaluated the WLE in a complex span task while controlling for
time parameters and the amount of concurrent articulation. Specifically, we used a computer-paced span task in which
participants remembered lists of either short or long words while concurrently either articulating or making a location
judgment. Whereas the WLE appeared when participants remained silent, concurrent articulation eliminated the effect.
Introducing a concurrent attention demand reduced recall, but did not affect WLE, and did not interact with concurrent
articulation. These results support the existence of two systems of maintenance for verbal information.