Training was adaptive according to a pre-specified algorithm with three levels of difficulty; as subjects became more proficient,task difficulty increased. Subjects were scanned using fMRI before and after training. During fMRI assessment, the letter-memory cri-terion task and two transfer tasks were administered: a numerica ln-back task and a Stroop task. A number of other transfer tasks were given outside the scanner pre and post training, tapping episodic memory, speed, verbal fluency, and reasoning (Dahlin et al., 2008a).Fig. 1A depicts gains in letter memory performance during scanning in younger and older adults, expressed as effect sizes. Fig. 1B por-trays gain trajectories for the two age groups across the five-week training period for the same letter-memory task assessed outside the scanner. As can be seen, both groups improved greatly from the intervention, although performance of the old at the end of the training the training period did not surpass that of the young in the early phases of training. Further, Fig. 1A shows a clear performance incrementon an untrained 3-back task (that also requires updating) in the young, but not in the old. By contrast, there was no transfer to the Stroop task, which taxes inhibition rather than updating (Miyakeet al., 2000). In addition, with the exception of n-back, no transfer was observed to the tasks assessed outside the scanner. This pat-tern indicates that the intervention effect was process-specific; the training did not affect executive functions or cognition in general,but specifically influenced WM updating.