Task- and age-effects in functional neuroimaging studies of working memory – a quantitative meta-analysis

Introduction: Impairment of working memory is found in patients suffering from a variety of neuro-psychiatric disorders such as dementia, Parkinson's disease or schizophrenia. The neurobiological basis of this function and its socio-economically highly relevant impairment, however, are still largely unclear. While there is already a large number of functional imaging studies investigating the neural correlates of the working memory¹, their results are often nonuniform and partly contradictory. To allow more generalisable inference about the involved brain regions as well as their age dependence we here report a coordinate-based meta-analysis on working memory studies.

Methods: This quantitative assessment included 47 functional MRI and PET studies examining different aspects of working memory in a total of 762 healthy subjects. These studies reported a total of 159 individual experiments that tested the neural correlates of a large variety of different paradigms such as the n-back, Sternberg or delayed match to sample tasks. The convergence of the coordinates reported for significant activation maxima in the different experiments was assessed by means of the „Activation likelihood estimation“ (ALE)² algorithm. This procedure accommodates the spatial uncertainty of the reported coordinates and tests for location in the brain where the convergence across results from different experiments was higher than expected by chance, i.e. given random spatial distribution of foci.

Results: In a conjunction of verbal and non-verbal working-memory tasks we found joint bilateral significant convergence of activations in Broca's region (BA 44 and 45), the intraparietal sulcus (IPS) and the SMA as well as the right prefrontal cortex. In the direct contrast we observed signfiicantly higher convergence of activation foci for non-verbal tasks in the IPS, the SMA and the right frontal lobe. Verbal tasks, on the other hand, were more strongly associated with activation of Broca's region and the lateral temporal lobe on the left hemisphere. Significant correlations with age were found in three regions. Older subjects were significantly less likely to activate the SMA and the right IPS. In contrast, they showed increased probabilities for activation of left Broca's region.

Discussion: This quantitative meta-analysis of the results from 159 neuroimaging experiments, demonstrated a reliable activation of a widespread fronto-parietal network by working memory tasks. However, we also found evidence for task specific variations within this system. Whereas verbal working memory tasks were associated with activity in the classical left-hemispheric language areas, non-verbal working memory tasks seem to recruit a dorsal fronto-parietal network primarily associated with spatial processing and motor planning. This dissociation could explain the differenzial impairment of working memory components in patients but also the sometimes divergent results from neuropsychological assessments. Moreover, we found significant age-related effects on activation probabilities indicating, that younger subjects may be more efficient in recruiting dorsal fronto-parietal networks whereas older subjects seem to be more likely to activate Broca's region. The latter may be ascribed to a compensatory recruitment of the „phonological loop“ (according to Braddly) due to reduced automated processing.

References: [1] Nebel et al Cognitive Brain Research 25 (2005). [2] Eickhoff et al Hum. Brain Mapp. (2009.)