Working memory (WM) functioning declines in old age. Due to its impact on many higher-order cognitive functions, investigating whether training can modify WM performance has recently been of great interest. We examined the relationship between behavioral performance and neural activity following five weeks of intensive WM training in 23 healthy older adults (M=63.7 years). 12 participants received adaptive training (i.e. individually adjusted task difficulty to bring individuals to their performance maximum), whereas the others served as active controls (i.e. fixed low-level practice). Brain activity was measured before and after training, using fMRI, while subjects performed a WM task under two difficulty conditions. Although there were no training-related changes in WM during scanning, neocortical brain activity decreased post training and these decreases were larger in the adaptive training group than in the controls under high WM load. This pattern suggests intervention-related increases in neural efficiency. Further, there were disproportionate gains in the adaptive training group in trained as well as in non-trained (i.e. attention, episodic memory) tasks assessed outside the scanner, indicating the efficacy of the training regimen. Critically, the degree of training-related changes in brain activity (i.e. neocortical decreases and subcortical increases) was related to the maximum gain score achieved during the intervention period. This relationship suggests that the decreased activity, but also specific activity increases, observed were functionally relevant.