The present controlled interventional study examined the effects of a cognitive intervention (spatial training vs. perceptual control training) and a physical intervention (endurance training vs. non-endurance training) on spatial learning and associated functional brain activations in healthy middle-aged adults. Spatial learning of a virtual maze tended to improve after a spatial training involving viewpoint shifting and path integration. Only the spatial training group showed significant changes in brain activation in the right middle and superior temporal gyrus and medial temporal lobe from pretest to posttest. At posttest, participants of the spatial training group showed lower activation levels than participants in the perceptual training group in a network of brain regions associated with spatial learning such as the hippocampus and parahippocampal gyrus. In contrast, the type of physical intervention did neither increase spatial training gains nor performance and associated brain activity in the virtual maze task. Thus, we did not find support for our hypothesis that cognitive and physical training would result in additive or even supra-additive gains for cognitive functioning.
Based on the findings in animals that both physical exercise and spatial training increased neuronal plasticity particularly in the hippocampus and performance in spatial memory tasks (e.g. ), performance in a spatial learning and navigation paradigm was chosen as the dependent variable in the present intervention study. This task was realized in a virtual maze task and required participants to build up a cognitive map of a virtual town. Only participants who took part in the spatial training, which involved viewpoint shifting and path integration tasks, were able to solve the virtual maze task at posttest and showed reduced brain activations at posttest compared to pretest. To successfully solve the virtual maze task, participants had to integrate the position of landmarks into a cognitive representation of the town and to continuously update their position in the environment by making use of visual motion cues. Thus, participants of the spatial learning group were able to transfer their newly acquired spatial abilities to the test task, i.e. the virtual maze task. This type of transfer has been called “near transfer”  and means that an ability is generalized from the training task, involving different stimuli and tasks, to a more complex test task of the same functional domain. In contrast, participants of the spatial training group did not outperform participants of the perceptual training group in tests of verbal learning and executive functions (‘far’ transfer). This result is in line with previous studies in humans (e.g. ) and animals , and suggests that spatial training induced changes limited to spatial processing systems. Furthermore, our results are in line with training studies in other cognitive domains, suggesting that a transfer from training to test is mainly observed when the training task and the test task depend on overlapping functions associated with similar brain regions . It remains to be determined whether spatial skills acquired in a virtual environment would generalize to the real world as well, although there is evidence that virtually acquired spatial knowledge transfers at least to real-world settings that are similar to the learned virtual environment .
On the neuronal level, the spatial training group, as compared to the perceptual training group, showed significant activation changes from pre- to posttest in the superior and middle temporal gyri and the medial temporal lobe of the right hemisphere.
Lateral temporal cortex has been shown to be involved in declarative memory, both during encoding and retrieval . Moreover, the middle and superior temporal gyri have been discussed as parts of a neural network which encodes spatial relations; especially the right temporal areas seem to be involved in calculating spatial coordinates . Activations in these temporal lobe structures have been shown to correlate with participants’ performance during spatial navigation [40, 58]. It might be speculated that brain activations in lateral temporal regions in the present study were involved in encoding the spatial relationship between buildings in the virtual maze tasks. A reduced activity in these areas at posttest for the spatial learning group might reflect a more efficient encoding of the spatial layout compared to posttest [38, 39].
Furthermore, a comparison of the spatial and perceptual training group at posttest revealed a reduced activation in the spatial training group in a network of brain areas associated with spatial learning, including the hippocampus and parahippocampal gyrus.
Hippocampal and parahippocampal activity have been shown to correlate with navigation abilities , navigation strategies  and increasing practice . Moreover, interindividual differences in navigational expertise have been linked to structural properties of the human hippocampus [81, 82], at least in young adults . Previous data using a similar spatial paradigm as in the present fMRI sessions had shown decreasing hippocampal activity with increasing knowledge of the environment .
Thus, as seen in the present study, the reduced activity in the spatial training group (as compared to the perceptual group) might reflect a better elaborated spatial knowledge.
Participants in the present sample (middle-aged men and women), however, did not reach a performance in the virtual maze task that was as high as the performance of the participants of Wolbers and Büchel . We think this is mainly due to the age difference of the participants. Moreover, Wolbers and Büchel had recruited only male adults while both men and women were recruited for the present study. Although Wolbers and Büchel  provided evidence for reduced hippocampal activity along with superior spatial performance, it has to be noted that there are other studies which demonstrated increased hippocampal activity in participants with more successful spatial navigation skills (e.g. [84, 85]). Moreover, age-related decline in spatial navigation performance has been associated with reduced hippocampal and parahippocampal activation [85, 86]. Thus, future studies are necessary to disentangle more precisely how age, navigation performance and strategies affect functional brain activation patterns in the medial temporal lobe.
Reduced activations in the spatial training group, as compared to the perceptual group, were seen in the frontal cortex as well. The medial frontal gyrus has been associated with spatial short-term memory . Moreover, the prefrontal cortex seems to play an important role for ‘higher cognitive processes’ in general and has been linked to the difficulty of a task . A reduced activity in prefrontal areas subsequent to practice is a well known correlate of learning and has been interpreted as an increase in neural efficiency .
Although the spatial training group tended to improve in the virtual maze task, and showed significant changes in associated brain activations from pre- to posttest, the aerobic endurance training group did not gain more than the non-endurance control group. This held true for the learning improvements across training sessions as well. The result differs from reports of previous studies in older adults. They reported enhanced cognitive performance after an aerobic endurance training as compared to a non-endurance training (e.g. [17, 19]). The sample in the present study was younger (40 to 55 years) than the samples of most previous studies (mostly older than 65 years). It has been discussed that beneficial effects of physical fitness on cognition increase with increasing age. In a meta-analysis, Colcombe and Kramer  provided evidence that participants between 66 and 70 years of age benefit more from exercise interventions than adults between 55 and 65 years of age. It might be speculated that beneficial effects of aerobic exercise on spatial learning are not yet observable in middle-aged adults, and may become evident only when these functions have already been subject of a larger degree of age-related decline. Although brain activations were not modulated by physical exercise on the group level, individual cardiovascular fitness and training induced fitness gains were positively correlated with brain activations during the virtual maze task for participants who took part in an additional spatial training (reported in ). The present results, together with our previous report, suggest that a spatial training has an immediate effect on neural networks associate with spatial learning (at least in middle-aged adults), whereas the influence of an endurance training might cause only subliminal effects on neurocognitive functioning, which are detected only in older age after age-related decline has further progressed. Indeed, prospective studies provided evidence that physical activity in midlife correlated with gray matter volume 20 years later  and participants that reported regular physical activity in midlife had a reduced risk of dementia at follow-up .
Another reason for the lack of a significant difference between the endurance and non-endurance training group might be that the stretching and coordination exercise in the present study promoted functions supporting spatial learning. Indeed Ruscheweyh et al.  demonstrated that physical activity has beneficial effects on memory functions independent of exercising intensity. Moreover, Voelcker-Rehage et al.  reported higher performance in older adults in a visual search task and in an executive task after a coordination training compared to a control group. FMRI data revealed changes in frontal and parietal areas, both after a cardiovascular training and a coordination training, suggesting that besides cardiovascular trainings, other types of exercise might have beneficial effects on cognition as well .
The sample size in the present study was rather small. On the behavioral level, effects of the spatial training showed only trends in the predicted direction. Moreover, null effects such as the non-significant interaction of physical and cognitive training must be interpreted with caution. The present results, together with our previous report of cardiovascular fitness modulating brain activations during a virtual spatial learning task , however, provide first empirical support in humans that the combination of a cognitive training with a physical intervention might promote neuronal plasticity in midlife. More studies with larger sample sizes should further address the question of additive or supraadditive effects of cognitive and physical trainings on cognitive functions in humans. Another limiting factor of the present study might be the rather overall low performance of participants in the maze task. This was unexpected given the results of previous studies with this paradigm [53, 66] and, as discussed above, is most likely due to age differences between study samples. Spatial memory has been shown to decrease as early as in midlife . Nevertheless, middle-aged adults still outperformed participants older than 55 years of age . Future studies should use a test task better adapted for the age group under investigation.