International Journal of Psychophysiology 101 (2016) 33–42

Contents lists available at ScienceDirect

International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho

Age-related differences of neural connectivity during mental rotation Monika Thomas Institute of Physiology and Anatomy, German Sport University Cologne, Am Sportpark Müngersdorf 6, 50933 Cologne, Germany

a r t i c l e

i n f o

Article history: Received 2 October 2015 Received in revised form 15 January 2016 Accepted 20 January 2016 Available online 22 January 2016 Keywords: Aging Mental rotation Graph theory Interhemispheric connections Intrahemispheric connections

a b s t r a c t The purpose of the present study was to investigate age-related effects on functional brain networks during a mental rotation task. At behavioral level age-related cognitive deficits have been shown. Cognitive deficits in older adults are associated with structural decline, especially in frontal and parietal areas and in the corpus callosum. In consequence, functional networks are affected by old age as well. To this end, a graph theoretical approach was taken, which quantifies the global and local efficiency as well as the cost efficiency of frontal and parietal intrahemispheric and interhemispheric networks. Main results indicate that intrahemispheric and interhemispheric networks are differently affected by older age: in the left frontal and the left and right parietal intrahemispheric networks global and local efficiency was reduced, whereas in frontal and parietal interhemispheric networks cost efficiency was decreased. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Mental rotation tasks consider the ability of judging objects, which are presented in different orientations and were firstly introduced by Shepard and Metzler (1971). In this study subjects had to judge whether two objects which were presented are similar. They could show an increasing reaction time with growing angle disparity between both presented objects. Therefore, it has been suggested that during the mental rotation processing the presented stimuli were first mentally rotated into an equal position before a judgment was made. Further studies, using different kinds of objects, obtained similar results (Dalecki et al., 2012; Ionta et al., 2007; Kessler and Thomson, 2010; Thomas et al., 2013). Objects can be classified into three categories: mental rotation of external objects (letters, cubes), of human shapes (hands, body parts) and of complex scenes (landscapes, table scenes). Depending on the object category mental rotation processes were interpreted in different ways: it is thought that external objects are mentally rotated in an allocentric reference frame (object-based transformation), while human hands and scenes activate an egocentric reference frame (perspective taking) (Kessler and Thomson, 2010; Kozhevnikov et al., 2006; Zacks and Michelon, 2005). Several studies with young adults showed that for all three categories reaction times increase with the angular deviation of the object Abbreviations: NR, no rotation; IR, intermediate rotation; FR, full rotation; k, costs; Eglobal, global efficiency; Elocal, local efficiency; Frontalleft, left frontal intrahemispheric network; Frontalright, right frontal intrahemispheric network; Frontalinter, frontal interhemispheric network; Parietalleft, left parietal intrahemispheric network; Parietalright, right parietal intrahemispheric network; Parietalinter, parietal interhemispheric network. E-mail address: [email protected].

http://dx.doi.org/10.1016/j.ijpsycho.2016.01.004 0167-8760/© 2016 Elsevier B.V. All rights reserved.

from the normal upright position (Dalecki et al., 2012; Ionta et al., 2007; Kessler and Thomson, 2010; Shepard and Metzler, 1971; Thomas et al., 2013). Moreover, it has been shown that the direction of deviation, namely clockwise or counter-clockwise, has no impact on the reaction times. Discriminating between a right and a left hand proceeds the fastest when the hand is presented fingertips-up, and slowest when it is presented fingertips-down. However, no differences in judging the left of the right hand could have been shown (Dalecki et al., 2012; Simone et al., 2013; Thomas et al., 2013). Furthermore, it has been shown that mental rotation of internal objects is performed faster and more accurate than of external objects (Amorim and Stucchi, 1997; Dalecki et al., 2012; Keehner et al., 2006; Thomas et al., 2013; Wraga et al., 1999). Behavioral studies investigating mental rotation in older adults found similar results (Gaylord and Marsh, 1975; Saimpont et al., 2009; Simone et al., 2013) and the increase of reaction times with angular deviation was even more pronounced in older participants (Band and Kok, 2000; Dror and Kosslyn, 1994; Gaylord and Marsh, 1975; Saimpont et al., 2009; Simone et al., 2013). Furthermore, even in the normal upright condition older participants had longer reaction times and produced more errors (Band and Kok, 2000; Dror and Kosslyn, 1994; Gaylord and Marsh, 1975; Saimpont et al., 2009). This possibly reflects age deficits for both, the cognitive ability of mentally rotating objects and the ability of judging the objects. Cognitive deficits in older adults are associated with structural decline, which is most pronounced in frontal areas but also is substantial in parietal areas and the corpus callosum (Dennis and Cabeza, 2011). This degeneration entails a reduced efficiency of structural networks and in consequence functional networks are affected by aging as well. More precisely, lower as well as higher functional coupling has been observed with aging and the associated cognitive decline (Antonenko and

34

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

Flöel, 2014; Andrews-Hanna et al., 2007; Geerligs et al., 2014; Wen et al., 2011). One possibility to investigate functional network organization is the graph theoretical approach (Bullmore and Sporns, 2009), where a network is defined by nodes (vertices, e.g. EEG electrodes) and edges (connections between different vertices) (Bullmore and Sporns, 2009; Rubinov and Sporns, 2010). The topography of such brain functional networks is neither completely regular nor completely random — they are so-called “small-world networks” (Achard and Bullmore, 2007; Latora and Marchiori, 2001). Small-world networks are characterized by dense or clustered local connections and only a few long-range connections (Bassett and Bullmore, 2006; Watts and Strogatz, 1998). Such architecture minimizes the path length between nodes, and thus reduces the costs of processing. Achard and Bullmore (2007) investigated the network efficiency and costs in older participants during resting state using fMRI. Efficiency has been defined as a function of the minimum path length between regions. Costs have been defined as the number of connections within the network. In fact, they found a reduced efficiency in frontal as well as temporal and subcortical regions in older adults compared to young adults. Also, Wen et al. (2011) found a decreased efficiency of the corticocortical network in older adults. This decrease of network efficiency may be a sign of neural dedifferentiation, characterizing a less specialized but more diffused pattern of functional connections (Antonenko and Flöel, 2014; Baltes and Lindenberger, 1997; Park and Reuter-Lorenz, 2009). The purpose of the present study was to investigate functional network connectivity of frontal and parietal areas in the elderly compared to young participants during a mental rotation task considering the results of Thomas et al. (2013). Additionally, the focus is on discriminating between different object categories. In this study, we investigated the functional connectivity within and between frontal and parietal areas during mental rotation focusing on young adults using EEG measurement. In contrast to the previously mentioned study, now a graph theoretical approach was used in order to analyze the small-world brain network behavior of older compared to younger adults during a mental rotation task. 2. Methods Ten younger (25.7 ± 1.3 years, 5 male) and ten older volunteers (73 ± 2.5 years, 6 male) participated in this study. The older participants were recruited via a registration list of interest in participating in aging studies. All were free of sensorimotor dysfunctions except for corrected vision, and none of them reported prior experience in mental rotation research. All participants signed an informed consent statement approved by the institutional ethics committee. This study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. The set-up of the experiment was similar to the former study (Thomas et al., 2013). For detailed description please consider this publication.

presented. Each stimulus was displayed until participants responded, and was followed by a randomly varying inter-trial interval of 0.5 s to 1 s. For statistical analysis the reaction time (RT) between stimulus onset and key press was averaged across clockwise and counterclockwise orientation angles, because no differences have been detected between these orientations, what was in line with the results of previous work (Dalecki et al., 2012; Thomas et al., 2013). Additionally, RT was averaged across the two respective stimulus shapes of each condition (“G” and “R”, two different hands, gun or knife), and across repetitions. RT data was submitted to an analysis of variance (ANOVA) with the within-factors Condition (LETTER, HAND, SCENE) and Rotation (0°, 30°, 60°, 90°, 120°, 150°, 180°). The outcome was Greenhouse–Geisser adjusted if necessary, and significant effects were scrutinized by Fishers LSD post-hoc tests. Additionally, the error rate (ER) was calculated as the percentage of wrong answers in the entire presented stimuli separately for each group averaged across all conditions and rotations. ER was submitted to an independent t-test. 2.2. EEG EEG signals were recorded by 64 electrodes (10–20 systems) (Jasper, 1958) (see also Thomas et al., 2013). In contrast to the former study 1000 ms segments instead of 800 ms segments relative to the stimulus onset (0 to 1000 ms) of the mental rotation task were used. All segments were subdivided according to the 7 different rotation angles from 0° to 180 °. The signals were transformed in the frequency domain using a Fast Fourier Transformation with a Hanning window function. Coherence was calculated as the quotient from crossspectrum and auto-spectrum implemented in the Brain Vision Analyzer software (Brain Products, Munich, Germany). Selected electrodes located over the frontal and parietal cortices were included in the analyses (see Fig. 1B). Coherence values for each condition (LETTER, HAND, SCENE) and each rotation angle (0°–180°) were measured for each participant in the gamma frequency band (30–45 Hz). Different networks have been defined to differentiate between intrahemispheric and interhemispheric as well as between frontal and parietal connections. The different networks considered the coherence values of the following electrode pairs, which were Fisher-Z transformed: 1. For intra-hemispheric networks:

a. Frontal, left (frontalleft): all possible connections within section I of Fig. 1 b. Frontal, right (frontalright): all possible connection within section II of Fig. 1 c. Parietal, left (parietalleft): all possible connections within section III of Fig. 1 d. Parietal, right (parietalright): all possible connections within section IV of Fig. 1

2.1. Mental rotation 2. For inter-hemispheric networks: As in the former study participants attended in three different mental rotation conditions: condition letter (“G” and “R”), condition hand (left or right hand) and condition scene (images of a person sitting at a round table a weapon lying in front) (Fig. 1A). Participants' task was to judge whether the letter was non-reversed or mirror-reversed, they saw a left or right hand and whether the weapon in the scene was on the left or right hand side of the person. To discriminate between the object presentation participants were instructed to press a key either with the right (‘k’ on the keyboard) or left index finger (‘d’ on the keyboard). In each condition objects were presented in the orientations 0°, ±30°, ± 60°, ± 90°, ±120°, ± 150° and 180°, in which 0° denotes fingers pointing upward, letters normally oriented, and persons turning their back to the observer. In each condition 144 different stimuli were

a. Frontal (frontalinter): all possible connections between Sections 1 and 2 b. Parietal (parietalinter): all possible connections between Sections 3 and 4

As a next step unweighted binary matrices for each participant and each rotation angle separated for each defined network were prepared with a threshold of 0.4. Thus, in total 42 matrices per participant were produced (7 rotation angles × 6 networks). To quantify the network quality structure we used graph theoretical analyses, where a network

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

35

Fig. 1. A. Examples of stimuli of the different conditions: Letter, Hand, Scene. Stimuli presented in the top row require the right button press, stimuli in the bottom row require the left button press. B. Included electrodes: Electrode positions of frontal and parietal networks: I: frontalleft, II: frontalright, III: parietalleft, IV: parietalright, I + II: frontalinter, III + IV: parietalinter.

is defined by having different nodes and edges between them (Bullmore and Sporns, 2009). Thus, the different electrodes represent the nodes while the connections between them represent the edges. Based on the matrices using the brain connectivity toolbox (Rubinov and Sporns, 2010) the following parameters were calculated: 1. Network costs (K): The network costs are calculated as total number of edges within a network (Degree of nodes) divided by the maximum possible number of edges (N2 − N) / 2 (Achard and Bullmore, 2007). 2. Global efficiency (Eglobal): The Global efficiency of a network is inversely related to the minimum path length between each electrode pair. The minimum path length is the number of edges which have to be passed to go from one node to the other (Bullmore and Sporns, 2009). Eglobal is calculated by

Eglobal ¼

X X 1 1 ; NðN−1Þ j∈G i∈G;i≠ j dij

ð1Þ

with N numbers of nodes, i and j representing different nodes and di,j representing the shortest path length defined by X

di; j ¼

auv ;

ð2Þ

auv ∈P

where P is the shortest path length between nodes i and j. a represents the connection status between two nodes (u,v) with au,v = 1 for an existing edge and otherwise au,v = 0. 3. Local efficiency (Elocal): As a measure of local efficiency the clustering coefficient (Watts and Strogatz, 1998) was used. The clustering coefficient is defined as a measure of information transfer in the immediate neighborhood of each node (Achard and Bullmore, 2007; Latora and Marchiori, 2001):

Elocal ¼

1X C; N i i

ð3Þ

with N numbers of nodes. Ci represents the number of edges between the neighbors of node i divided by the maximum possible number of connections ki(ki − 1) / 2. The local efficiencies of all nodes were averaged within a network to calculate the mean local efficiency.

All three parameters (K, Eglobal, Elocal) are normalized, as they are calculated in relation to the maximum possible number of nodes and maximum possible efficiencies. Therefore, it is 0 ≤ {K, Eglobal, Elocal} ≤ 1. Due to this normalization cost efficiency defined as Eglobal–K could be calculated (Achard and Bullmore, 2007). Cost efficiency is thought to quantify small-world properties such that high values characterize an economical network with low costs and great efficiency, while low and especially negative values describe an uneconomical network with high costs and low efficiency. Previous analyses of the present data indicated that the results of the rotation angles 30° to 150° did not differ between each other. Therefore, we used the average of these rotation angles for further analyses, such that three rotation values were applied to the statistical analyses: 0° (NR: no rotation), 30° to 150° (IR: interim rotation) and 180° (FR: full rotation). For statistical analyses four-way ANOVAs were calculated for pairs of intrahemispheric networks, with the factors Group (older, young) × Condition (LETTER, HAND, SCENE) × Rotation range (NR, IR, FR) × Network (first ANOVA: frontalright, frontalleft; second ANOVA: parietalright, parietalleft). Further three-way ANOVAs were calculated separately for the frontal interhemispheric (frontalinter) and parietal interhemispheric (parietalinter) networks with the factor Group (older, young) × Condition (LETTER, HAND SCENE) × Rotation range (NR, IR, FR). As dependent variables Eglobal, Elocal and cost efficiency were used. Results were Greenhouse–Geisser adjusted when necessary, and significant effects were scrutinized by Fisher's LSD post-hoc tests. To quantify relationships between the behavioral data (reaction time) and the EEG networks six stepwise multiple linear regression analyses were calculated separately for each rotation range (NR, IR, FR) averaged over the three conditions (LETTER, HAND, SCENE) with the dependent variable reaction time. Within three regression analyses (one for each rotation range) as regressors the EEG parameters of frontal networks were used: Eglobal, Elocal and cost efficiency of frontalright, frontalleft and frontalinter. The regressors of the other three regression analyses (one for each rotation range) were the EEG parameters (Eglobal, Elocal and cost efficiency) of the parietal networks (parietalright, parietalleft and parietalinter). Inclusion criteria for the stepwise multiple regression analyses were F values ≥1. 3. Results 3.1. Behavioral data The results are depicted in Fig. 2. For the young as well as for the older participants reaction times increased with ascending rotation

36

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

Fig. 2. Behavioral data Reaction times [ms] of the mental rotation test separately for the older and the young subject group averaged across subjects separately for each condition and orientation. Error bars represent standard errors.

from 0° to 180° of all three object categories. It further appears that the reaction times for the older participants were higher for all conditions and rotation angles — even for the 0°-rotation. The course of the reaction times from 0° to 180 °and the relationship between conditions was comparable for both participant groups. ANOVA showed significant effects for Group (F (1,18) = 30.13; p b .001; η2 = 0.63), Condition (F (2, 36) = 1.99; p b .001; η2 = 0.4) and Rotation (F (6, 108) = 81.88; p b .001; η2 = 0.82). The LSD post hoc test for the factor Condition showed that SCENE differed significantly from LETTER and HAND. Result of the LSD-Post Hoc test for the factor Rotation showed that the rotation angles 0°–60 ° differed significantly from 90°–180°. Additionally, single rotation angles differed significantly between each other from the 90° until 180°-rotation. Furthermore, ANOVA yielded significant interactions between Rotation ∗ Group (F (6, 108) = 9.95; p b .001; η2 = 0.35), Condition ∗ Rotation (F(12, 216) = 17.41; p b .001; η2 = 0.49) and Condition ∗ Rotation ∗ Group (F(12, 216) = 3.11; p b .001; η2 = 0.15). As the main finding, the LSD Post Hoc Test of Condition ∗ Rotation ∗ Group yielded significantly differing reaction times of each condition and rotation angle between groups. The error rate of older participants (13.7% ± 12.6%) was significantly higher than ER of the young participants (5.3% ± 4.9%) (t (18) = 3.42; p b .01; d = 1.61). 3.2. EEG 3.2.1. Global efficiency (Eglobal) For the frontal intrahemispheric networks Eglobal of frontalleft was reduced in older participants as ANOVA revealed a significant interaction of Network ∗ Group (F (1, 18) = 7.32; p = .014; η2 = 0.29) (Fig. 3a). Results of the LSD Post-Hoc test confirmed the significantly reduced Eglobal of frontalleft compared to frontalright for the older participants (p = .02). For the young participants both networks did not differ between each other. Additionally, ANOVA yielded a significant effect for the factor Rotation range (F (2, 36) = 12.21; p b .001; η2 = 0.4). LSD Post-Hoc test showed that Eglobal was significantly reduced for IR compared to NR (p = .004) and FR (p b .001) (Fig. 3a). All other effects were not significant and are presented in Table 1. For the parietal intrahemispheric networks a significant agerelated reduced Eglobal has been found such that ANOVA yielded a significant effect of the factor Group (F (1, 18) = 4.62; p = .045; η2 = 0.2) and the interaction of Rotation range ∗ Group (F (2, 36) = 5.00; p = .012; η2 = 0.22). The LSD Post-Hoc Test revealed that Eglobal of the

older participants was significantly reduced for NR (p = .038) and IR (p = .016) compared to young participants. Further, ANOVA yielded a significant effect of Rotation range (F (2, 36) = 28.64; p b .001; η2 = 0.61) and a significant interaction of Rotation range ∗ Network (F (2, 36) = 6.26 p = .004; η2 = 0.26). Relevant results of the LSD Post-Hoc Tests showed that Eglobal of all Rotation ranges differed significantly for parietalleft with IR b NR, FR and NR b FR (all p ≤ .001). For parietalright Eglobal of IR was significantly lower than NR and FR (all p b .01). Additionally, Eglobal of FR was significantly lower for parietalright compared to parietalleft (p b .001) (Fig. 3b). All other effects were not significant (Table 1). For the frontal interhemispheric network and the parietal interhemispheric network no significant Group differences has been observed. However, for both networks, ANOVAs revealed a significant effect of the factor Rotation range (frontalinter: F (2, 36) = 13.4; p b .001; η2 = 0.43; parietalinter: F (1.21, 21.91) = 10.39; p = .003; η2 = 0.37). For frontalinter the results of the LSD Post-Hoc Test showed that Eglobal of all Rotation ranges differed significantly with IR b NR, FR and NR b FR (all p b .05) (Fig. 3c). All other effects were not significant (Table 1). For parietalinter the LSD Post-Hoc Test showed that Eglobal of FR was significantly greater than of NR (p = .015) and IR (p b .001) (Fig. 3d). Additionally, ANOVA of parietalinter showed a significant three-way interaction of Condition ∗ Rotation range ∗ Group (F (2.03, 36.59) = 4.25; p = .021; η2 = 0.19). However, the LSD Post-Hoc Test did not show any relevant age-related differences. All other effects were not significant (Table 1). 3.2.2. Local efficiency (Elocal) As for Eglobal a significant age-related reduced Elocal of frontalleft has been shown. Thus, ANOVA revealed a significant interaction between Network ∗ Group (F(1,18) = 4.58; p = .046; η2 = 0.27). Results of the LSD Post-Hoc test showed a significant reduced Elocal of frontalleft of the older participants compared to frontalleft of the young participants (p = .028) (Fig. 4a). Additionally, ANOVA yielded a significant effect for the factor Rotation range (F (2, 36) = 9.51; p b .001; η2 = 0.4). LSD Post-Hoc Test showed that all Rotation ranges differed significantly with IR b NR, FR and NR b FR (all p b .05) (Fig. 4a). All other effects were not significant (Table 1). For the parietal intrahemispheric networks no significant agerelated effects have been found. But ANOVA revealed a significant effect for the factor Condition (F (2, 32) = 5.07; p = .011; η2 = 0.22). LSD Post-Hoc Test indicated that Elocal for the condition SCENE was significantly lower than for the condition LETTER (p = .004) and HAND

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

37

Fig. 3. Global efficiency Eglobal averaged across subjects and conditions separately for each rotation range (NR, IR and FR) with error bars representing standard errors of: A frontalright and frontalleft, B parietalright and parietalleft, C frontalinter and D parietalinter.

(p = .028) (inset of Fig. 4b). Further, results showed a significant effect for the factor Rotation range (F (2, 36) = 28.59; p b .001; η2 = 0.61) and the interaction between Rotation range ∗ Network (F (2, 36) = 4.04; p b .05; η2 = 0.18). Relevant results of the LSD Post-Hoc Test showed that Elocal of all Rotation ranges differed significantly for parietalleft with IR b NR, FR and NR b FR (all p b .01). For parietalright Elocal of IR was significantly lower than of NR (p = .000) and of FR (p b .001). Additionally, Elocal of NR and IR was significantly lower for parietalleft compared to parietalright (p b .001) (Fig. 4b). All other effects were not significant (Table 1). For the frontal and parietal interhemispheric networks no significant Group effects have been detected. But for both networks the factor Rotation range became significant (ANOVA frontalinter: F (2, 36) = 10.46; p b .001; η2 = 0.37; ANOVA parietalinter: F (2, 36) = 16.55; p b .001; η2 = 0.48). LSD Post-Hoc Tests showed that in both networks Elocal differed significantly between all Rotation ranges with IR b NR, FR and NR b FR (all p b .05) (Fig. 4c, d). Additionally, for parietalinter ANOVA yielded a significant effect for the factor Condition (F (2, 26) = 3.85; p = .03; η2 = 0.18) indicating a significant lower Elocal for the condition SCENE compared to the condition LETTER (LSD Post-Hoc Test: p = .011) (inset of Fig. 4d). All other effects were not significant (Table 1).

3.2.3. Cost efficiency The results of the parameter cost efficiency are depicted in Fig. 5. For the frontal intrahemispheric networks and for the parietal intrahemispheric networks the ANOVAs yielded no significant main effects, such that no group differences or network differences could have been detected (Table 1). For the parietal intrahemispheric networks a significant four-way interaction between Condition*Rotation range*Network*Group has been detected (F (4, 72) = 3.26; p = .02;

η2 = 0.16). However, the LSD Post-Hoc Test showed no relevant group differences. Within frontalinter and parietalinter older participants had a reduced cost efficiency compared to young participants independent of the condition or the Rotation range such that ANOVA revealed a significant effect of the factor Group (frontalinter: F (1, 18) = 6.531; p = .02; η2 = 0.27; parietalinter: F(1,18) = 5.56; p = .03; η2 = 0.24) (Fig. 5c, d). Additionally, a significant effect of the factor Rotation range was found for frontalinter (F (2, 36) = 6.44; p = .004; η2 = 0.26). The LSD Post-Hoc tests showed a significantly lower cost efficiency for IR compared to NR (p = 0.003) and FR (p = .003). All other effects were not significant (Table 1). 3.3. Stepwise multiple regression 3.3.1. Multiple regressions with the dependent variable reaction time of NR The analysis comprising the parameters of the frontal networks included four regressors: Eglobal of frontalleft and frontalinter, cost efficiency of frontalinter and frontalright. Only the first three reached statistical significance (Table 2) resulting in a significant total model (R2 = .62, F (4, 15) = 6.05, p = .004; f2 = 1.61). Results of the analysis comprising the parameters of the parietal networks could not indicate any correlation between parietal parameters and reaction time. The total model was not statistically significant (R2 = .29, F (2, 17) = 3.43, p = .056) where two regressors were included with none reaching statistical significance: Eglobal and cost efficiency of parietalright. 3.3.2. Multiple regressions with the dependent variable reaction time of IR The analysis comprising the parameters of the frontal networks resulted in the inclusion of five regressors: the same four regressors as

38

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

Table 1 Results of ANOVAS with dependent variables Eglobal, Elocal, cost efficiency separately for the frontal intrahemispheric networks (frontalleft, frontalright), the parietal intrahemispheric networks (parietalleft, parietalright), the frontal interhemispheric network (frontalinter) and the parietal interhemispheric network (parietalinter). P-levels are defined as following: *** p ≤ .001, ** p ≤ .01, * p ≤ .05 and n.s. p N .05. When necessary, Greenhouse–Geisser adjusted values are quoted.

Frontal intrahemispheric

Parietal intrahemispheric

Frontalinter

Parietalinter

Effect

Eglobal

Elocal

Cost efficiency

Group Condition Condition ∗ Group Rotation Rotation ∗ Group Network Network ∗ Group Condition ∗ Rotation Condition ∗ Rotation ∗ Group Condition ∗ Network Condition ∗ Network ∗ Group Rotation ∗ Network Rotation ∗ Network ∗ Group Condition ∗ Rotation ∗ Network Condition ∗ Rotation ∗ Network ∗ Group Group Condition Condition ∗ Group Rotation Rotation ∗ Group Network Network ∗ Group Condition ∗ Rotation Condition ∗ Rotation ∗ Group Condition ∗ Network Condition ∗ Network ∗ Group Rotation ∗ Network Rotation ∗ Network ∗ Group Condition ∗ Rotation ∗ Network Condition ∗ Rotation ∗ Network ∗ Group Group Condition Condition ∗ Group Rotation Rotation ∗ Group Condition ∗ Rotation Condition ∗ Rotation ∗ Group Group Condition Condition ∗ Group Rotation Rotation ∗ Group Condition ∗ Rotation Condition ∗ Rotation ∗ Group

F(1,18) = 0.84n.s. F(2,36) = 1.42n.s. F(2,36) = 2.19n.s. F(2,36) = 12.21*** F(2,36) = 3.02n.s. F(1,18) = 0.80n.s. F(1,18) = 7.32* F(4,72) = 0.66n.s. F(4,72) = 0.64n.s. F(2,36) = 0.04n.s. F(2,36) = 0.81n.s. F(2,36) = 0.17n.s. F(2,36) = 1.41n.s. F(4,72) = 1.61n.s. F(4,72) = 2.28n.s. F(1,18) = 3.14n.s. F(2,36) = 2.04n.s. F(2,36) = 1.23n.s. F(2,36) = 17.58*** F(2,36) = 2.06n.s. F(1,18) = 1.16n.s. F(1,18) = 0.89n.s. F(4,72) = 1.22n.s. F(4,72) = 0.43n.s. F(2,36) = 0.20n.s. F(2,36) = 0.00n.s. F(2,36) = 6.32** F(2,36) = 3.09n.s. F(4,72) = 1.97n.s. F(4,72) = 0.88n.s. F(1,18) = 2.08n.s. F(2,36) = 1.36n.s. F(2,36) = 2.78n.s. F(2,36) = 13.40*** F(2,36) = 1.04n.s. F(4,72) = 1.22n.s. F(4,72) = 1.29n.s. F(1,18) = 2.25n.s. F(2,36) = 0.31n.s. F(2,36) = 2.09n.s. F(2,36) = 10.39*** F(2,36) = 1.43n.s. F(4,72) = 2.36n.s. F(4,72) = 4.25n.s.

F(1,18) = 2.13n.s. F(2,36) = 1.68n.s. F(2,36) = 0.94n.s. F(2,36) = 9.51*** F(2,36) = 2.69n.s. F(1,18) = 0.54n.s. F(1,18) = 4.58* F(4,72) = 0.45n.s. F(4,72) = 0.48n.s. F(2,36) = 0.19n.s. F(2,36) = 1.98n.s. F(2,36) = 0.47n.s. F(2,36) = 2.16n.s. F(4,72) = 0.42n.s. F(4,72) = 0.14n.s. F(1,18) = 1.03n.s. F(2,36) = 5.07* F(2,36) = 0.24n.s. F(2,36) = 28.59*** F(2,36) = 3.04n.s. F(1,18) = 0.79n.s. F(1,18) = 0.45n.s. F(4,72) = 1.25n.s. F(4,72) = 1.13n.s. F(2,36) = 0.35n.s. F(2,36) = 3.25n.s. F(2,36) = 4.04* F(2,36) = 0.68n.s. F(4,72) = 0.54n.s. F(4,72) = 0.73n.s. F(1,18) = 0.92n.s. F(2,36) = 2.33n.s. F(2,36) = 1.86n.s. F(2,36) = 10.46*** F(2,36) = 1.88n.s. F(4,72) = 0.44n.s. F(4,72) = 0.20n.s. F(1,18) = 0.39n.s. F(2,36) = 3.88* F(2,36) = 0.26n.s. F(2,36) = 16.55*** F(2,36) = 1.86n.s. F(4,72) = 1.32n.s. F(4,72) = 0.39n.s.

F(1,18) = 0.86n.s. F(2,36) = 0.26n.s. F(2,36) = 1.06n.s. F(2,36) = 0.38n.s. F(2,36) = 0.06n.s. F(1,18) = 0.22n.s. F(1,18) = 1.54n.s. F(4,72) = 0.48n.s. F(4,72) = 1.23n.s. F(2,36) = 1.83n.s. F(2,36) = 0.92n.s. F(2,36) = 0.01n.s. F(2,36) = 2.01n.s. F(4,72) = 1.32n.s. F(4,72) = 1.81n.s. F(1,18) = 3.15n.s. F(2,36) = 2.14n.s. F(2,36) = 3.17n.s. F(2,36) = 0.01n.s. F(2,36) = 0.62n.s. F(1,18) = 4.13n.s. F(1,18) = 0.14n.s. F(4,72) = 0.13n.s. F(4,72) = 0.13n.s. F(2,36) = 1.17n.s. F(2,36) = 1.19n.s. F(2,36) = 0.05n.s. F(2,36) = 0.04n.s. F(4,72) = 0.56n.s. F(4,72) = 3.26* F(1,18) = 6.53* F(2,36) = 1.99n.s. F(2,36) = 1.30n.s. F(2,36) = 6.44** F(2,36) = 0.93n.s. F(4,72) = 1.01n.s. F(4,72) = 0.27n.s. F(1,18) = 5.56* F(1.36,24.59) = 0.80n.s. F(1.36,24.59) = 1.61n.s. F(1.37,24.76) = 1.89n.s. F(1.37,24.76) = 2.21n.s. F(2.32,41.86) = 2.76n.s. F(2.32,41.86) = 1.45n.s.

for NR and Elocal of frontalright with the same regressors reaching statistical significance as for NR (Eglobal of frontalleft and frontalinter, cost efficiency of frontalinter) (Table 2). The total model was statistically significant (R2 = .62, F (5, 14) = 4.57, p = .011; f2 = 1.61). The analysis comprising the parameters of the parietal networks resulted in a significant total model (R2 = .33, F (2, 17) = 4.14, p = .034; f2 = 0.49). The same two regressors as for NR were included but none reached statistical significance (Table 2). 3.3.3. Multiple regressions with the dependent variable reaction time of FR Results of the analysis comprising the parameters of the frontal networks resulted in a significant total model (R2 = .68, F (4, 15) = 7.92, p = .001; f2 = 2.13). Four regressors were included with the first three reaching statistical significance (Table 2): Elocal of frontalleft, cost efficiency of frontalleft and frontalinter and Eglobal of frontalinter. The total model of the analysis comprising the parameters of the parietal networks did not reach statistical significance (R2 = .48, F (5, 14) = 2.62, p = .071; η2 = 0.92). Nevertheless, two of the five included regressors were statistically significant: Eglobal and cost efficiency of parietalleft. Eglobal, Elocal and cost efficiency of parietalright were not significant regressors.

4. Discussion The purpose of the study was to determine age effects on electro cortical synchronization patterns during mental rotation of three different object categories and different rotations. Behavioral data showed that older participants had higher reaction times independent of the condition or rotation angle. Even for the 0°-rotation older participants needed longer reaction times compared to young participants, indicating age deficits in the choice reaction time performance which were also found in other studies (Fozard et al., 1994; Hultsch et al., 2002). In fact, older participants showed a comparable pattern of the relation between the conditions and rotation angles compared to the young participants. Overall, the reaction times of the older adults were arranged at an averaged offset of 574 ms ± 214 ms with a stronger increase of reaction times with increasing angle disparity, such that age deficits of mental rotation were most pronounced for the 180°- rotation and as well for the LETTER condition. These results are in line with previous studies which observed higher reaction times for older participants compared to young participants for different kinds of stimuli (Band and Kok, 2000; Dror and Kosslyn, 1994; Gaylord and Marsh, 1975; Saimpont et al., 2009).

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

39

Fig. 4. Local efficiency Elocal averaged across subjects and conditions separately for each rotation range (NR, IR and FR) with error bars representing standard errors of: A frontalright and frontalleft, B parietalright and parietalleft, C frontalinter and D parietalinter. The insets represent Elocal averaged across both subjects groups and all rotation ranges separately for each condition with error bars representing standard errors: inset in B averaged across parietalleft and parietalright and inset in D of parietalinter.

4.1. Age-related EEG results EEG results of the present study could show that during a mental rotation task the economical network properties were affected by older age. More precisely, Eglobal and Elocal have been found to be affected in frontalleft. Additionally, Eglobal of parietal intrahemispheric networks was affected by age during no rotation (NR) and intermediate rotation (IR). Cost efficiency was reduced in the interhemispheric networks. Remarkably, reduced efficiency has been observed during both, the no rotation and rotation situation. Due to that, age-related deficits should be rather interpreted as dependent to cognitive processes relating to choice reaction time performance than to mental rotation processes. This finding is in line with the behavioral results as reaction times were found to be generally prolonged in older adults independent of the rotation situation. The age-related reduction of Eglobal and Elocal of frontalleft indicates a worse network performance compared to young participants. More precisely, the reduced Eglobal reflects a reduced capacity for parallel information processing between nodes using multiple series of edges (Achard and Bullmore, 2007). The reduced Elocal refers to a degraded ability in maintaining information processing when the index node is eliminated, described as fault tolerance (Achard and Bullmore, 2007; Latora and Marchiori, 2001). For the interhemispheric networks no age-related effects of Eglobal and Elocal but age-related reduced cost efficiency has been detected. These results support the idea that the capacity for parallel information processing between hemispheres within frontal and parietal areas was preserved through higher costs in older adults. In contrast, this was not true for the intrahemispheric networks, where results reflect that a lower efficiency within these networks was not accompanied by higher

costs. This interpretation is based on the different meanings of Eglobal and Elocal compared to cost efficiency. As mentioned in the introduction, economic or so called small-world networks are defined as networks with high efficiency and low costs, with costs estimated as the number of edges of a network in relation to the maximum possible number of edges (Achard and Bullmore, 2007; Bassett and Bullmore, 2006). Hence, small-world properties require at least positive cost efficiency. In other words, high cost efficiency reflects economic network properties. For example, about 37% of Eglobal for about 10% of costs has been detected in case of a functional brain network (Achard and Bullmore, 2007). In summary, the age related effects on Eglobal, Elocal as well as on cost efficiency can be understood to be in line with structural decline of frontal and parietal areas and the corpus callosum (Dennis and Cabeza, 2011) and may be explained by neural dedifferentiation meaning a less specialized and more diffuse pattern of functional connections (Antonenko and Flöel, 2014; Baltes and Lindenberger, 1997; Park and Reuter-Lorenz, 2009). Nevertheless, it should be remarked that an age-related reduced skull conductivity could influence coherence (Wendel et al., 2010). That means, that possibly coherence changes while brain dynamics have not changed. But if this is true for the present data it would have been expected that age-related coherence changes become obvious in both, the intrahemispheric and interhemispheric networks. But in fact this was not the case. 4.2. Relationship of cognitive behavior and function brain network processing The results of the multiple linear regression showed that Eglobal, Elocal and cost efficiency of frontal networks have an impact on the reaction

40

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

Fig. 5. Cost efficiency Cost efficiency averaged across subjects and conditions separately for each rotation range (NR, IR and FR) with error bars representing standard errors of: A frontalright and frontalleft, B parietalright and parietalleft, C frontalinter and D parietalinter.

Table 2 Results of stepwise multiple regression analyses with the dependent variable reaction time separately for each rotation range (NR, IR, Fr). As regressors the EEG parameter (Eglobal, Elocal, cost efficiency) were used with separate analyses for the frontal and parietal networks.

Rotation range NR

Frontal networks

Parietal networks

Rotation range IR

Frontal networks

Parietal networks

Rotation range FR

Frontal networks

Parietal networks

Regressors

b

t

p-Value

Eglobal frontalleft Cost efficiency frontalinter Eglobal frontalinter Cost efficiency frontalright R2 = .61, F(4, 15) = 6.05, p = .004 Cost efficiency parietalright Eglobal parietalright R2 = .28, F(2, 17) = 3.43, p = .061 Eglobal frontalleft Cost efficiency frontalinter Eglobal frontalinter Cost efficiency frontalright Elocal frontalright R2 = .62, F(5, 14) = 4.57, p = .011 Cost efficiency parietalright Cost efficiency parietalinter R2 = .32, F(2, 17) = 4.13, p = .034 Elocal frontalleft Cost efficiency frontalinter Cost efficiency frontalleft Eglobal frontalinter R2 = .67, F(4, 15) = 7.92, p = .001 Cost efficiency parietalright Cost efficiency parietalleft Eglobal parietalleft Eglobal parietalright Elocal parietalleft

−1.086 −1.408 1.438 0.407

−3.58 −3.05 −2.60 1.91

p = .002 p = .007 p = .02 p = 0.074

−0.38 −0.29

−1.85 −1.41

p = 0.081 p = 0.177

−1.012 −1.583 1.731 0.497 −0.378

−2.79 −3.17 2.71 2.02 −1.74

p = .014 p = .006 p = .016 p = .062 p = .103

−0.373 −0.301

2.02 −1.74

p = .109 p = .189

−0.917 −0.742 0.457 0.557

−3.71 −3.51 2.45 2.03

p = .002 p = .003 p = .027 p = .061

−0.151 −1.048 1.556 −0.555 −0.795

−0.63 −2.55 2.18 −1.92 −1.37

p = .536 p = .022 p = .046 p = .074 p = .191

R2 = .48, F(5, 14) = 2.62, p = .071

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

times for all three rotation ranges. Conspicuously, while the results of no rotation and intermediate rotation were comparable, the results of full rotation were different as also an impact of parietal networks has been detected. This reflects the special role of the full rotation situation, which is discussed more precisely in a further paragraph. Reaction times of no rotation (NR) and intermediate rotation (IR) are influenced by Eglobal of frontalleft and cost efficiency and Eglobal of frontalinter. The first two mentioned parameters are also reduced in older adults. Therefore, these effects can be interpreted as an influencing factor on age-related prolonged reaction times. This implicates that a less economic information processing results in a longer processing time. Therefore, it is consistent, that the relationship between both parameters and the reaction time is negative, such that a prolonged reaction time is accompanied by reduced Eglobal of frontalleft and cost efficiency of frontalinter. In contrast, it is somehow surprising that the relationship between the reaction times and Eglobal of frontalinter is positive and thus reflecting an increased efficiency with prolonged reaction times. Here, it has to be mentioned that the effect of Eglobal of frontalinter is independent of age effects. Necessarily, we have to remember that Eglobal is associated with the involvement of multiple series of edges ensuring short path length and a high capacity for parallel information processing. Considering this aspect the relationship between prolonged reaction times and increased efficiency seems coherent. For the full rotation situation (FR) a negative impact of Elocal of frontalleft as well as cost efficiency of frontalinter on the reaction times has been detected. As these parameters are additionally affected by age, this can be interpreted as an age-related effect on the reaction times. This is not feasible for the cost efficiency of frontalleft which was found to have a positive impact on the reaction times of full rotation. This relationship is difficult to interpret. As the ANOVA of cost efficiency of frontal intrahemispheric networks revealed no significant results, this relationship may be an incidental finding. Furthermore, the reaction time of FR has been found to be positive affected by Eglobal of parietalleft, while this parameter was not affected by age. However, the influence of Eglobal of parietalleft indicates the above mentioned special role of the full rotation situation compared to NR and IR. This was strengthened by the results of ANOVA inducing a significantly greater Eglobal of parietalleft compared to NR and IR. As a further result, cost efficiency of parietalleft and parietalright were found to have a negative impact on the reaction times of FR. As for the frontal intrahemispheric networks the ANOVA of cost efficiency of parietal intrahemispheric networks revealed no significant results. Therefore, the impact of these parameters may be again interpreted as incidental findings. 4.3. Effects related to condition and rotation Differences between the rotations were found in every network for Eglobal and Elocal and for the cost efficiency only in frontalinter. The results of Eglobal and Elocal reflected a U-shape of the three rotation ranges with reduced efficiency for the intermediate rotation (IR) compared to no rotation (NR) and full rotation (FR). This result is in line with the findings of (Thomas et al., 2013), indicating a special role for the FR compared to the IR. In contrast, cost efficiency did not differ between the rotation ranges except of frontalinter. For all other networks, the efficiency as well as the costs were reduced during IR such that cost efficiency remained stable. Therefore, one could argue that the way of information processing is different between IR and FR but not less economic. In fact, the mental rotation of the FR, describing a two-axes inversion, seems to involve distinct task-specific processes compared with mental rotation of smaller rotations (Dalecki et al., 2012; Franklin et al., 1992; Shelton and McNamara, 2001; Thomas et al., 2013). Differences between the conditions are only reflected for Elocal of all parietal networks with a lower Elocal for the condition SCENE compared to the other conditions. In fact, the difference is in line with the

41

behavioral data, where as well SCENE differed significantly from LETTER and HAND. This result is contradictory to the idea that SCENE and HAND as internal objects were processed by egocentric-perspective taking while LETTER as external object is processed by object-based transformation (Kessler and Thomson, 2010; Kozhevnikov et al., 2006; Zacks and Michelon, 2005). Therefore, the present result does not reflect the different processes. Considering the behavioral data, comparable results has been detected as reaction times of SCENE differed significantly from LETTER and HAND. Taken together, evidence suggests that these differences may reflect a higher cognitive demand for the conditions LETTER and HAND rather than differences in the explained processes. The fact that the conditions differed also in case of no rotation, confirms this assumption. In conclusion, the results of the present study confirm the findings of Thomas et al. (2013) concerning differences between intermediate rotation and full rotation extending these findings to older participants. Beyond that, age-related effects on economic brain functional network processing during a mental rotation task were investigated. The most relevant age-related finding to emerge from this study is a reduction of Eglobal and Elocal in intrahemispheric networks of frontal and parietal areas, while cost efficiency is degraded in interhemispheric networks. Finally, it should be noted, that coherence analyses of EEG data do not represent precise brain sources, but rather the communication between the activity measured by the different sensors. Therefore, the interpretation of the results in relation to the brain sources should be treated with care. To solve this problem of interpretation further studies should be designed by using both the EEG measurement as real-time measurement together with imaging methods to investigate the real brain sources. Acknowledgments Thanks are due to L. Geisen who has written the software for realtime experimental control, O. Bock and A. Simon for proof reading the article and S. Saalmann for data collection. Responsibility for the contents rests with the author. References Achard, S., Bullmore, E., 2007. Efficiency and cost of economical brain functional networks. PLoS Comput. Biol. 3 (2), e17. http://dx.doi.org/10.1371/journal.pcbi.0030017. Amorim, M.A., Stucchi, N., 1997. Viewer- and object-centered mental explorations of an imagined environment are not equivalent. Brain research. Cogn. Brain Res. 5 (3), 229–239. Andrews-Hanna, J.R., Snyder, A.Z., Vincent, J.L., Lustig, C., Head, D., Raichle, M.E., Buckner, R.L., 2007. Disruption of large-scale brain systems in advanced aging. Neuron 56 (5), 924–935. http://dx.doi.org/10.1016/j.neuron.2007.10.038. Antonenko, D., Flöel, A., 2014. Healthy aging by staying selectively connected: a mini-review. Gerontology 60 (1), 3–9. http://dx.doi.org/10.1159/000354376. Baltes, P.B., Lindenberger, U., 1997. Emergence of a powerful connection between sensory and cognitive functions across the adult life span: a new window to the study of cognitive aging? Psychol. Aging 12 (1), 12–21. Band, G.P., Kok, A., 2000. Age effects on response monitoring in a mental-rotation task. Biol. Psychol. 51 (2–3), 201–221. http://dx.doi.org/10.1016/S0301-0511(99)00038-1. Bassett, D.S., Bullmore, E., 2006. Small-world brain networks. Neuroscientist 12 (6), 512–523. http://dx.doi.org/10.1177/1073858406293182. Bullmore, E., Sporns, O., 2009. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat. Rev. Neurosci. 10 (3), 186–198. http://dx. doi.org/10.1038/nrn2575. Dalecki, M., Hoffmann, U., Bock, O., 2012. Mental rotation of letters, body parts and complex scenes: separate or common mechanisms? Hum. Mov. Sci. 31 (5), 1151–1160. http://dx.doi.org/10.1016/j.humov.2011.12.001. Simone, L.d., Tomasino, B., Marusic, N., Eleopra, R., Rumiati, R.I., 2013. The effects of healthy aging on mental imagery as revealed by egocentric and allocentric mental spatial transformations. Acta Psychol. 143 (1), 146–156. http://dx.doi.org/10.1016/j. actpsy.2013.02.014. Dennis, N.A., Cabeza, R., 2011. Age-related dedifferentiation of learning systems: an fMRI study of implicit and explicit learning. Neurobiol. Aging 32 (12), 2318, e17. http://dx. doi.org/10.1016/j.neurobiolaging.2010.04.004. Dror, I.E., Kosslyn, S.M., 1994. Mental imagery and aging. Psychol. Aging 9 (1), 90–102. Fozard, J.L., Vercruyssen, M., Reynolds, S.L., Hancock, P.A., Quilter, R.E., 1994. Age differences and changes in reaction time: the Baltimore longitudinal study of aging. J. Gerontol. 49 (4), P179. http://dx.doi.org/10.1093/geronj/49.4.P179. Franklin, N., Tversky, B., Coon, V., 1992. Switching points of view in spatial mental models. Mem. Cogn. 20 (5), 507–518.

42

M. Thomas / International Journal of Psychophysiology 101 (2016) 33–42

Gaylord, S.A., Marsh, G.R., 1975. Age differences in the speed of a spatial cognitive process. J. Gerontol. 30 (6), 674–678. Geerligs, L., Maurits, N.M., Renken, R.J., Lorist, M.M., 2014. Reduced specificity of functional connectivity in the aging brain during task performance. Hum. Brain Mapp. 35 (1), 319–330. http://dx.doi.org/10.1002/hbm.22175. Hultsch, D.F., MacDonald, S.W.S., Dixon, R.A., 2002. Variability in reaction time performance of younger and older adults. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 57 (2), P101. http://dx.doi.org/10.1093/geronb/57.2.P101. Ionta, S., Fourkas, A.D., Fiorio, M., Aglioti, S.M., 2007. The influence of hands posture on mental rotation of hands and feet. Exp. Brain Res. 183 (1), 1–7. http://dx.doi.org/10. 1007/s00221-007-1020-2. Jasper, H.H., 1958. Report of the committee on methods of clinical examination in electroencephalography. Electroencephalogr. Clin. Neurophysiol. 10, 370–375. Keehner, M., Guerin, S.A., Miller, M.B., Turk, D.J., Hegarty, M., 2006. Modulation of neural activity by angle of rotation during imagined spatial transformations. NeuroImage 33 (1), 391–398. http://dx.doi.org/10.1016/j.neuroimage.2006.06.043. Kessler, K., Thomson, L.A., 2010. The embodied nature of spatial perspective taking: embodied transformation versus sensorimotor interference. Cognition 114 (1), 72–88. http://dx.doi.org/10.1016/j.cognition.2009.08.015. Kozhevnikov, M., Motes, M., Rasch, B., Blajenkova, O., 2006. Perspective-taking vs. mental rotation transformations and how they predict spatial navigation performance. NeuroImage (20), 397–417. Latora, V., Marchiori, M., 2001. Efficient behavior of small-world networks. Phys. Rev. Lett. 87 (19), 198701. Park, D.C., Reuter-Lorenz, P., 2009. The adaptive brain: aging and neurocognitive scaffolding. Annu. Rev. Psychol. 60 (1), 173–196. http://dx.doi.org/10.1146/annurev.psych. 59.103006.093656.

Rubinov, M., Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. NeuroImage 52 (3), 1059–1069. http://dx.doi.org/10.1016/j.neuroimage. 2009.10.003. Saimpont, A., Pozzo, T., Papaxanthis, C., 2009. Aging affects the mental rotation of left and right hands. PLoS One 4 (8), e6714. http://dx.doi.org/10.1371/journal.pone.0006714. Shelton, A.L., McNamara, T.P., 2001. Systems of spatial reference in human memory. Cogn. Psychol. 43 (4), 274–310. http://dx.doi.org/10.1006/cogp.2001.0758. Shepard, R.N., Metzler, J., 1971. Mental rotation of three-dimensional objects. Science 171 (3972), 701–703. Thomas, M., Dalecki, M., Abeln, V., 2013. EEG coherence during mental rotation of letters, hands and scenes. Int. J. Psychophysiol. 89 (1), 128–135. http://dx.doi.org/10.1016/j. ijpsycho.2013.06.014. Watts, D.J., Strogatz, S.H., 1998. Collective dynamics of 'small-world' networks. Nature 393 (6684), 440–442. http://dx.doi.org/10.1038/30918. Wen, W., Zhu, W., He, Y., Kochan, N.A., Reppermund, S., Slavin, M.J., Brodaty, H., Crawford, J., Xia, A., Sachdev, P., 2011. Discrete neuroanatomical networks are associated with specific cognitive abilities in old age. J. Neurosci. 31 (4), 1204–1212. http://dx.doi. org/10.1523/JNEUROSCI.4085-10.2011. Wendel, K., Väisänen, J., Seemann, G., Hyttinen, J., Malmivuo, J., 2010. The influence of age and skull conductivity on surface and subdermal bipolar EEG leads. Comput. Intell. Neurosci. 2010, 397272. http://dx.doi.org/10.1155/2010/397272. Wraga, M., Creem, S.H., Proffitt, D.R., 1999. The influence of spatial reference frames on imagined object- and viewer rotations. Acta Psychol. 102 (2–3), 247–264. Zacks, J.M., Michelon, P., 2005. Transformations of visuospatial images. Behav. Cogn. Neurosci. Rev. 4 (2), 96–118. http://dx.doi.org/10.1177/1534582305281085.