NeuroImage 98 (2014) 216–224

Contents lists available at ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Imaging tactile imagery: Changes in brain connectivity supportperceptual grounding of mental images in primary sensory cortices

Timo Torsten Schmidt a,b,c,⁎, Dirk Ostwald b,c, Felix Blankenburg a,b,c

a Neurocomputation and Neuroimaging Unit (NNU), Department of Education and Psychology, Freie Universität Berlin, 14195 Berlin, Germanyb Bernstein Center for Computational Neuroscience, 10115 Berlin, Germanyc Max Planck Institute for Human Development, Center for Adaptive Rationality (ARC), 14195 Berlin, Germany

⁎ Corresponding author at: Freie Universität Berlin,Psychology, Neurocomputation and Neuroimaging Unit45, 14195 Berlin, Germany. Fax: +49 30 838 55620.

E-mail address: timo.schmidt@bccn-berlin.de (T.T. Sch

http://dx.doi.org/10.1016/j.neuroimage.2014.05.0141053-8119/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 6 May 2014Available online 13 May 2014

Keywords:Mental imageryVibrotactilePerceptual groundingCore construction systemfMRIPsychophysiological interaction (PPI)Connectivity

Constructingmental representations in the absence of sensory stimulation is a fundamental ability of the humanmind and has been investigated in numerous brain imaging studies. However, it is still unclear how brain areasfacilitating mental construction processes interact with brain regions related to specific sensory representations.In this fMRI study subjects formed mental representations of tactile stimuli either from memory (imagery) orfrom presentation of actual corresponding vibrotactile patterned stimuli. First our analysis addressed the ques-tion of whether tactile imagery recruits primary somatosensory cortex (SI), because the activation of early per-ceptual areas is classically interpreted as perceptual grounding of the mental image. We also tested whether anetwork, referred to as ‘core construction system’, is involved in the generation of mental representations inthe somatosensory domain. In fact, we observed imagery-induced activation of SI. We further found supportfor the notion of amodality independent construction networkwith the retrosplenial cortices and the precuneusas core components, which were supplemented with the left inferior frontal gyrus (IFG). Finally, psychophysio-logical interaction (PPI) analyses revealed robust imagery-modulated changes in the connectivity of theseconstruction related areas, which suggests that they orchestrate the assembly of an abstract mental representa-tion. Interestingly, we found increased coupling between prefrontal cortex (left IFG) and SI during mental imag-ery, indicating the augmentation of an abstract mental representation by reactivating perceptually groundedsensory details.

© 2014 Elsevier Inc. All rights reserved.

Introduction

The cognitive process of forming mental images in the absence ofperceptual stimuli is a fundamental aspect of human cognition. It under-lies future thinking and thereby enables planning and supportsdecision-making. Mental imagery has therefore always attracted cogni-tive neuroscientists (e.g. Shepard and Metzler (1971)) and Kosslyn(1973) and research has broadened to all sensory modalities and con-tent of diverse complexity (Albright, 2012; Hassabis and Maguire,2007, 2009; McNorgan, 2012; Schacter et al., 2012).

Initial neuroimagingworkwas dominated by the question ofwheth-er mental imagery activates early sensory cortices, a phenomenon werefer to as perceptual grounding (Barsalou, 2008; Kosslyn et al., 2001).Overlap in neural activation during imagery and perception strength-ened the view that the nature of mental representations is formedby our perceptual apparatus and experience. Evidence for such earlysensory activation has now accumulated across different modalities,

Department of Education and(NNU), Habelschwerdter Alle

midt).

where activation of early sensory cortices appears dependent on howfine-grained the mental image is (Kosslyn and Thompson, 2003;McNorgan, 2012; Thompson et al., 2009). The imagery literature isdominated bywork on “themind's eye”, while other sensorymodalitiesare underrepresented, such as the somatosensory domain (Newmanet al., 2005; Olivetti Belardinelli et al., 2009; Yoo et al., 2003). To fillthis gap and to facilitate the identification of modality independentprinciples, our study focused on tactile imagery.

The construction process ofmental image generation involvesmem-ory retrieval, forming and maintenance of a mental representation, andfinally a conscious experience similar to perception (Moulton andKosslyn, 2009; Schacter, 2012; Schacter et al., 2012). On the neurallevel, a “core construction network” is activated during the assemblyof mental representations independently of content and modality(Hassabis and Maguire, 2009; Summerfield et al., 2009, 2010). Thefew initial reports on network properties, such as changes in connectiv-ity during imagery, leave the question open of how the constructionnetwork interacts with the brain areas related to perceptual grounding(Chen et al., 2009; Deshpande et al., 2010; Gao et al., 2011; Kasess et al.,2008; Mechelli et al., 2004; Schlegel et al., 2013). It has been suggestedthat themodality independent core network couples to task specific re-gions including sensory areas (Schacter et al., 2012). Here we employed

Fig. 1. Experimental design. A. Three trial types constituted the conditions of the fMRI par-adigm: In the PERCEPTION condition, subjects were presented with a vibrotactile patternstimulus and had to report whether a probe pin was included in the pattern. In theIMAGERY condition, subjects were cued by a color display to imagine a pattern stimulusand perform the same task as in the PERCEPTION condition. The CONTROL conditionwasmatched to the IMAGERY condition in terms of tactile and visual physical stimulation;subjects reported whether or not the vibration of the probe pin was briefly interrupted.B. Left panel: The four vibrotactile stimuli as presented to the left index finger on a 4 × 4pin Braille display. Right panel: Illustration of the vibrotactile pattern presentation consti-tuted of pattern pins (green) and non-vibrating background pins. Subjects had to report ifthe additional probe pin (black) was covered by the pattern. In the CONTROL andIMAGERY conditions, only a probe pin was presented.

217T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

a novel tactile imagery recall task that necessitated to actively form amental representation to test if changes in connectivity can revealnew insights into the construction process of mental representations.

Our paradigmwas inspired by a classic visual imagery task of lettersin a grid (Podgorny and Shepard, 1978). Analogously, we used a 16-pinBraille display to present vibrotactile bar-patterns. Subjects either per-ceived or imagined tactile stimuli and reported whether a probe-pinwas included in a pattern or not. This demanding discrimination taskcompelled subjects to form mental representations of fine-sensory de-tail and allowed the behavioral assessment of their performance.

We hypothesized that this task necessitates perceptual grounding inSI and involves the activation of the core construction network.We thentested for modulations of functional interaction between the construc-tion network and brain areas associated with the content of the mentalrepresentation and provide new evidence for the role of the prefrontalcortex in interacting with early perceptual areas.

Materials and methods

Participants

Fourteen healthy volunteers (age range: 22–30 years; 8 males, oneleft-handed) without any neurological or psychiatric disorder complet-ed the study after giving written informed consent. The studycorresponded to the Human Subject Guidelines of the Declaration ofHelsinki and was approved by the Ethics Committee of the Charité Uni-versity Hospital Berlin.

Experimental stimuli and conditions

To test for perceptual grounding and the involvement of theconstruction system in tactile imagery, we designed a vibrotactileimagery task inspired by classical investigation of visual imagery(Podgorny and Shepard, 1978) with three experimental conditions:(1) PERCEPTION, (2) IMAGERY, and (3) CONTROL (Fig. 1A). In short,subjects indicatedwhether a probe stimuluswas covered by a particularvibrotactile pattern or not, where this pattern stimulus was either per-ceived or imagined. In the control condition subjects detected a hardlydetectable gap within the probe stimulus.

Vibrotactile stimuli (1.5 s duration) were delivered to the leftindex finger using a 16-dot piezoelectric Braille display (4 × 4 matrixwith 2.5 mm spacing) controlled by a programmable stimulator(Piezostimulator, QuaeroSys, St. Johann, Germany). Four vibrotactilepatterns were defined consisting of eight pattern pins and eight back-ground pins (Fig. 1B). Pattern pins were driven by a 120 Hz sinusoidalcarrier signal, which was amplitude-modulated by a half sine wave tosmooth stimulus on- and offsets while the background pins wereretracted during stimulus presentation. Prior to the experiment, partic-ipants were trained to associate a set of four color cues with the fourvibrotactile patterns.

The trials in the fMRI paradigm were constituted as follows: (1) Inthe PERCEPTION condition, subjects were presented with one of thefour vibrotactile pattern stimuli. In addition to the pattern pins aprobe pin was applied with 8 Hz oscillatory amplitude modulationaround half a sine wave. On each trial the participant's task was to de-cidewhether or not the probe pin was included in the pattern and indi-cate this by a right-hand button-press response. The mapping betweenyes/no answers and fingers was counterbalanced over subjects. Theprobe was part of the pattern in 50% of the trials. A color cue was pre-sented tomatch the low-level visual stimulation to the IMAGERY condi-tion. As subjects were intensively trained on the cue-patternassociation, the cue wasmeaningless (different color set) to avoid auto-matic imagery upon cue presentation. Thereby the meaningless cueprevented any interaction of imagery and perceptual processes in thePERCEPTION condition. (2) On each trial of the IMAGERY condition, par-ticipants were cued with one of the learned color displays to imagine

the corresponding vibrotactile pattern and as in the PERCEPTION condi-tion to indicate via button press, whether a probe pin was included inthe (mentally represented) pattern or not. (3) In the CONTROL condi-tion, no vibrotactile pattern was presented. Instead, only the probe pinwas presented together with a meaningless color cue. The participant'stask was to indicate whether the probe pin vibration was brieflyinterrupted or not. In 25% of the trials it was paused for half of an 8 Hzsinusoidal oscillation (i.e. 62.5 ms) either 375 ms after stimulus onsetor 375 ms before stimulus offset. Both temporal positions of the inter-ruption were equally probable to ensure the participant's continuing

218 T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

attention throughout the probe pin presentation. In all conditions, trialslasted 3 s comprising a 1.5 s stimulus presentation and a 1.5 s responsewindow.

We employed simple visual stimuli to cue the recall of tactile per-cepts. Visual stimuliwere simple color patches to be used only as a recallcue without any further visual processing demands to foster themodal-ity–specificity of the imagery processes to the tactile domain. To rendervisual strategies unlikely, our taskwas designed to require detailed spa-tial resolution of the tactile percept. Additionally, subjects wereinstructed to mentally reactivate how the percept of the vibrotactilestimulus had felt on their finger. Debriefing of the subjects after the ex-periment did not reveal the use of visual strategies.

fMRI experimental design and data acquisition

fMRI data were acquired in eight runs lasting 6.5 min each. A runcomprised three blocks (PERCEPTION, IMAGERY, CONTROL) eachconsisting of 12 trials. Prior to the blocks, the experimental condition(i.e. task to be performed) was visually indicated to the participantsfor 4.5 s, followed by a 10 s fixation cross until the first trial.Each vibrotactile pattern was presented/imagined equally often (threetimes per block). The order of blocks and of trials within blockswas ran-domized. The inter-trial intervals of the 12 trials within each blockwerescattered between 2 and 9 s in steps of 1 s. During the inter-trial intervalall pins were slightly elevated to provide the location and extent of thetactile display. Participants were instructed to keep their eyes openthroughout the experiment. At the beginning of each run the partici-pants were reminded of the color cue-pattern association by joint pre-sentation of the cue and the tactile pattern, which lasted 6 s. Theassociations (color and pattern) were counterbalanced over partici-pants. Stimulus presentation was controlled using custom MATLABcode (The MathWorks, MA) and the Cogent 2000 Toolbox (developedby John Romaya at the LON at the Wellcome Department of ImagingNeuroscience). Visual cues were presented on a screen that was visiblefrom within the scanner via a mirror system attached to the head coil.

Functional imaging was performed on a 1.5 Tesla Siemens SonataMRI scanner (SiemensMedical Solutions, Erlangen, Germany) equippedwith a standard circular polarized head coil. T2*-weighted images wereacquired with a gradient-echo planar image sequence (TR = 2010 ms,TE = 40 ms, flip angle = 90°). Each EPI volume consisted of 36 axialslices covering the whole brain in an interleaved slice ordering (voxelsize 3 × 3 × 3 mm3, matrix size 64 × 64).

fMRI data analysis

The fMRI data were pre-processed with SPM8 (Wellcome TrustCentre for Neuroimaging, Institute for Neurology, University CollegeLondon, London, UK). To minimize movement-induced image artifactseach data set was realigned to its mean image. Next, masks defininggray and white matter were generated for each subject by segmentingthe mean image of the fMRI scans of each subject using SPM8's defaulttissue probability maps as priors and non-linearly aligned to the groupaverage using DARTEL (Ashburner, 2007; Chu et al., 2011). After esti-mating the inter-subject alignment by matching tissue class images to-gether, the warping parameters were used to transform each subject'sfMRI volumes into MNI space. Finally, the EPI images were re-interpolated to 2 × 2 × 2 mm3 voxel size and spatially smoothed withan 8 mm FWHM Gaussian kernel.

fMRI statistical analysis was performed according to a standard gen-eral linearmodel (GLM) approach using a two-levelmixed-effects anal-ysis in SPM8. At the subject level, regressorswere specified for the eventrelated design as trial onset regressors separately for: PERCEPTION, IM-AGERY, CONTROL and CONTROLinterruptedwhere the last regressor refersto those trials with interrupted presentation of the probe pin. To ac-count for potential confounds of response time differences betweenconditions, we included an additional regressor with the button-press

onset of each trial. Further the movement parameters were modeledas regressors of no interest. After the model was fitted to the experi-mental data contrast images were first evaluated at the single-subjectlevel and then passed to second-level one-sample t-tests. For one sub-ject the data of two runs was excluded as the subject reported to haveconfused the task conditions.

We also sought to identify the network that underpinsthe mental representation of vibrotactile patterns, shared byPERCEPTION and IMAGERY. We tested the respective contrasts tothe CONTROL condition against the conjunction null hypothesis(Friston et al., 1999; Nichols et al., 2005). Specifically, theCONTROL condition trials were divided equally and arbitrarilyinto two distinct sets and modeled as separate regressors. This sep-aration allowed for independent computation of the conjunctioncontrasts (PERCEPTION NCONTROL1 and IMAGERY N CONTROL2),which required separate trials for independent baselines. At thesecond level this modification resulted in a 2 × 2 within-subjectANOVA, with factors condition (PERCEPTION, IMAGERY) and con-trol (CONTROL1, CONTROL2), which was assessed using a flexiblefactorial within-subject GLM including a between-subject factor.

To test for temporal aspects of BOLD responses, we computed peri-stimulus time histograms using the rfxplot second-level analysis tool-box (Gläscher, 2009), where time-courses were only adjusted forblock and nuisance effects.

Connectivity analysis

Task-dependent connectivity modulations were assessed using thepsychophysiological interaction (PPI) approach as implemented inSPM8 (Friston, 2011; Friston et al., 1997; Gitelman et al., 2003). To iden-tify how core regions of the construction system interactwith perceptu-al areas during imagery, we defined four seed regions-of-interest withinthe core construction network based on our GLM analysis. For eachseed region the first eigenvariate of the BOLD time-series was extractedfrom voxels within a 4 mm radius sphere around the subject-specificpeak voxel in the contrast IMAGERY N PERCEPTION, individuallythresholded: left retrosplenial cortex (RSC) (x = −14.0 ± 2.9, y =−60.1 ± 5.7, z = 21.3 ± 5.6), right RSC (x = 15.3 ± 3.7, y =−58.4 ± 4.5, z = 20.7 ± 6.5), precuneus (x = 2.0 ± 2.7, y =−58.9 ± 3.5, z = 47.9 ± 6.0), and left IFG (x = −41.7 ± 3.0, y =22.4 ± 5.3, z = 24.7 ± 5.1). The deconvolved time-series were mul-tiplied by the psychological variables (IMAGERY N PERCEPTION con-trast, expressed as onset weights [1 −1]), and reconvolved with theHRF to obtain the PPI interaction-terms (Gitelman et al., 2003). Onthe single-subject level our multi-run PPI design contained sets ofthree regressors per run: interaction-term, time-series and psycho-logical factor supplemented with a run constant. PPIs were estimat-ed and the interaction term parameter estimates were forwardedto second-level t-tests.

All reported coordinates correspond toMNI space. The SPManatomytoolboxwas used to establish cytoarchitectonical referenceswhere pos-sible (Eickhoff et al., 2007, 2008). All activations are reported at p b 0.05,corrected for false discovery rate (FDR) at the cluster level, except if stat-ed otherwise.

Results

Behavioral data

To test that the three task conditions were similarly demanding,we compared performance levels expressed as percentage of cor-rect responses: PERCEPTION: 76.5 ± 11.8% (SD), IMAGERY: 73.7 ±9.3%, CONTROL: 76.5 ± 12.3 (one-way ANOVA between conditionsF(2,26) = 0.56, p = 0.577). This corresponds to a hit-rate forPERCEPTION of 61.3 ± 15.9%, for IMAGERY of 58.8 ± 12.4%, and forCONTROL of 46.1 ± 30.2% and a correct rejection-rate for PERCEPTION

219T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

of 84.1± 9.3%, for IMAGERY of 78.7± 8.2%, and for CONTROL of 91.9±5.4%, resulting in mean d′ for PERCEPTION of 1.58 ± 0.70, for IMAGERYof 1.29 ± 0.59, and for CONTROL of 1.69 ± 1.59 and mean bias C forPERCEPTION of −0.12 ± 29.3, for IMAGERY of −0.04 ± 16, and forCONTROL of 0.13± 0.44. A repeated-measures ANOVA revealed no sig-nificant differences between the three conditions in d′ (F(2,26) =1.407, p =.263) or in C (F(2,26) = 2.947, p = 0.70). Median responsetimes (RTs), assessed as delay after onset of the response display,were compared in a one-way ANOVA (F(2,26) = 4.88, p = 0.016)and post-hoc paired t-tests revealed significantly elongated RTs inIMAGERY (666 ± 106 ms) as expected (Formisano et al., 2002; Kosslyn,1976) when compared to PERCEPTION (578 ± 130 ms, t(13) = 4.68,p b 0.01) but not to CONTROL (595 ± 139 ms, t(13) = 1.99, p = 0.06).For one subject the RTs of two runs were lost due to computer failure.

fMRI data

The contrast PERCEPTION N CONTROL was evaluated to identifytask-related network activation of processing a patterned vibrotactilestimulus when perceiving it. As expected, this contrast revealedincreased activation in a tactile processing network comprising

Fig. 2. Results of the GLM analysis. A. The contrast PERCEPTION N CONTROL shows a netwIMAGERY N CONTROL contrast was computed to test for perceptual grounding of the mental rvation related to the construction and mental pattern representation. C. The conjunction analystation of the tactile pattern stimulus and performance of the probe allocation task. Activationevidence that the construction of a mental representation leads to similar neuronal proIMAGERY N PERCEPTION contrast reflects activation related to the construction of the mentastimulus time histograms for activations related to the construction system and perceptual gIMAGERY conditions. Activation in the CONTROL condition is substantially smaller despite theprecuneus and retrosplenial cortices respond only during the IMAGERY condition emphasizing

contralateral primary and secondary somatosensory cortex, intra-parietal sulcus (IPS) and supplementary motor area (SMA; see Fig. 2Aand Table 1A; Blankenburg et al., 2006; Bodegard et al., 2001; Burtonand Sinclair, 2000; Francis et al., 2000; Moore et al., 2013; Wackeret al., 2011). The contrast also revealed activation in right lateral occip-ital complex (LOC) matching reports on visuo-haptic object-relatedclusters termed lateral occipital tactile–visual area (LOtv; Amedi et al.,2001, 2002, 2007).

Perceptual grounding and stimulus representation

To test for perceptual grounding of mental representations, we per-formed two analyses. First, we contrasted the IMAGERY and CONTROLconditions, where only the first required a mental representation of atactile stimulus while both conditionswerematched regarding physicalstimulation.We found activation in IPS, superior parietal lobe (SPL), in-ferior parietal lobe (IPL), the frontal eye fields (FEF), inferior frontalgyrus (IFG) and bilateral LOtv. Crucially, this contrast also revealed acti-vation of the right SI, supporting the concept that the mental image isgrounded in perceptual areas. As the employed IMAGERY condition in-volves the recall of a previously learned stimulus, memory retrieval

ork of brain areas activated by the perceptual processing of the tactile stimuli. B. Theepresentation formed during imagery. We found activation of SI, supplemented with acti-is shows activations shared by IMAGERY and PERCEPTION related to themental represen-of SI (p b 0.001 uncorrected) and lateral occipital tactile–visual area (LOtv) give furthercessing as during perception, which is referred to as perceptual grounding. D. Thel image. All activations are shown at p b 0.05 FDR corrected on the cluster level. E. Peri-rounding (±SEM). The evoked responses in SI appear similar for the PERCEPTION andfact that the physically applied stimulus is identical to the IMAGERY condition. Left IFG,their role for assembling the mental representation from memory.

Table 1Anatomical location and MNI coordinates of GLM analysis. p b 0.05 corrected for FDR onthe cluster-level. FEF: Frontal eye field, IFG: Inferior frontal gyrus, IPL: Inferior parietallobe, IPS: Inferior parietal sulcus, LOtv: Lateral occipital tactile–visual area, medFS: Medialfrontal sulcus, Parahipp: Parahippocampal region, PFC: Prefrontal cortex, Prec: Precuneus,RSC: Retrosplenial cortex, SI: Primary somatosensory cortex, SII: Secondary somatosenso-ry cortex, SMA: Supplementary motor area, SPL: Superior parietal lobe, VI: Primary visualcortex.

Peak MNI coordinates

Cluster size Anatomical region X Y Z z-Score peak

A. PERCEPTION N CONTROL1597 Right SPL 28 −66 50 5.41

Right IPS 34 −42 44 4.57Right SI 56 −24 40 4.48

738 Left IPS/SI −58 −24 36 5.34−40 −40 48 4.83

244 Left IPL −26 −72 32 4.52292 Right LOtv 58 −60 −10 4.5175 Right precentral gyrus 60 6 30 3.7860 Left FEF −30 −4 50 3.77

B. IMAGERY N CONTROL4707 Right IPL 46 −70 28 5.77

Right SPL 20 −64 56 5.53Left IPL −32 −74 42 5.38

88 Right LOtv 56 −48 −12 4.92524 Left IFG −44 6 30 4.56319 Right FEF 26 10 50 4.55418 Left FEF −26 −6 50 4.5153 Right Parahipp 34 −38 −14 4.43134 Right IPS 46 −34 48 4.42316 Right RSC 18 −62 22 4.38170 Left LOtv −58 −62 −10 4.2653 Right SI 56 −18 42 4.0896 VI −6 −84 −4 3.87

C. (IMAGERY N CONTROL) & (PERCEPTION N CONTROL)415 Right SPL 28 −60 54 5.03277 Right IPL 38 −70 26 4.90382 Left IPS −44 −36 44 4.6692 Right SPL 36 −48 58 4.27184 Right IPS 36 −34 42 4.23139 Right LOtv 58 −56 −10 4.1757 Right SI (uncor.) 56 −22 40 4.17117 Left IPL −28 −80 34 4.14

D. IMAGERY N PERCEPTION400 Right IPL 48 −70 32 5.16231 Left IFG −42 22 26 5.031158 Left RSC −18 −58 22 4.96

Right SPL 10 −66 56 4.88Prec 4 −58 50 4.30

739 Right RSC 18 −54 18 4.8979 Left IPS −34 −54 40 4.41113 Left FEF −28 0 58 4.25438 Left IPL −38 −84 28 4.2168 VI −2 −84 2 4.09132 Right FEF 24 14 44 4.03137 Prec −6 −74 54 3.9957 medFS −16 26 38 3.9580 Left PFC −44 44 −4 3.95

220 T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

related activation was expected. Correspondingly, we found the rightparahippocampal region activated. Fig. 2B and Table 1B summarizethe corresponding activation pattern.

Second, we sought to identify brain regions related to the mentalstimulus representation by means of a conjunction analysis. Becausefor both the PERCEPTION and IMAGERY conditions amental representa-tion of the tactile pattern had to be formed, we computed the conjunc-tion of their contrasts to the CONTROL condition and tested against theconjunction null hypothesis (Nichols et al., 2005). As presented inFig. 2C and Table 1C we found further support for perceptual groundingby means of SI activation (p b 0.001 uncorrected due to strong a prioriassumptions). Additionally we found activation in bilateral IPS and IPL,and right SPL and LOtv.

Engagement of the core construction network in tactile imagery

We next tested whether tactile imagery involves the recruitment ofthe construction system. We contrasted the IMAGERY against thePERCEPTION condition as in both conditions, subjects performed thesame task, i.e. processing a bar-pattern stimulus and allocating a probepin within this representation. In the PERCEPTION condition the stimu-lus was perceived; in the IMAGERY condition its representationwas de-rived from memory. This contrast therefore reflects the constructionrelated network without task-specific or representation-specific activa-tions. In accordance with the description of the core constructionsystem, this contrast revealed increased activations of bilateralretrosplenial cortices (RSC), the precuneus and also the left IFG(Hassabis and Maguire, 2009; McNorgan, 2012). We further observedactivation of bilateral FEF, SPL, IPL and primary visual cortex (VI) andprefrontal cortex (PFC) as shown in Fig. 2D and Table 1D.

BOLD time-courses

To characterize the temporal dynamics of recorded BOLD responseswe computed peri-stimulus time histograms (BOLD time-courses) aspresented in Fig. 2E for right SI, left IFG, precuneus and bilateral RSC.Evoked responses follow the hemodynamic response function, whereresponses in SI for IMAGERY and PERCEPTION appear similar in theirdynamics and amplitude while the CONTROL condition elicited weakerresponses.

Interaction between the core construction and the content representationsystem

To identify the dynamic interaction of the construction system withthe network related to the mental pattern representation, we imple-mented psychophysiological interaction (PPI) analyses. Seed regionsfor the PPI analyses were defined within the construction networkand PPIs calculated for the contrast IMAGERY N PERCEPTION. Resultsare summarized in Fig. 3 and Table 2.

Increases in connectivity from the RSC and precuneus to a distribut-ed network of areas indicate their prominent function of orchestratingthememory retrieval and (re)construction processes duringmental im-agery. Connectivity to bilateral IPS as representation-related areas wasmodulated from the precuneus and right RSC and to the right SI frombilateral RSCs and precuneus. In contrast, the analysis for the left IFG re-vealed a selective and substantial increase in coupling to the right SI.This connection had spatial specificity onto the BA1 sub-region of SIwhich was also activated during imagery (compare Figs. 2B and C).This finding indicates that the coupling of IFG and SI is crucial for aug-menting an abstract mental image with perceptually grounded detailswhen mentally reactivating perceptual processes.

Discussion

This article reports on an fMRI experiment during which subjectswere required to form a mental representation of vibrotactile patternstimuli. First, our report complements the imagery literature by pre-senting a stringently controlled study in the tactile domain. We foundclear evidence for perceptual grounding of themental image in primarysomatosensory cortex, supplementing reports from other modalities(Kosslyn and Thompson, 2003; McNorgan, 2012). Secondly, our find-ings support the notion that tactile imagery involves the brain's con-struction system, which can be distinguished from a network relatedto the mental content representation (Hassabis and Maguire, 2007,2009). Finally our connectivity analysis revealed increased coupling be-tween core elements of the construction network and content relatedbrain areas during imagery including themodulation of connectivity be-tween prefrontal cortex (left IFG) and right SI.

Fig. 3. Psychophysiological interaction (PPI) analysis. Increases in functional connectivityfor IMAGERY N PERCEPTION for seed regions placed within the construction system:A. left retrosplenial cortex (RSC), B. right RSC, C. precuneus, D. left inferior frontal gyrus(IFG). All p b 0.05 corrected for FRD on the cluster level. E. Summary of connectivity in-creases between the reported seed regions within the construction network and areas re-lated to the mental representation as identified in the conjunction analysis (Fig. 2C).Bilateral RSCs and the precuneus show increased connectivity to a distributed networkof brain areas, which emphasizes their role in orchestrating retrieval and construction ofa mental representation. In contrast, the left IFG shows topologically specific coupling toBA1 hand region of SI, which indicates a mechanism to augment a mental representationwith sensory details.

Table 2Anatomical location andMNI coordinates of PPI analysis results. p b 0.05 corrected for FDRon the cluster-level. FEF: Frontal eye field, IFG: Inferior frontal gyrus, Ins: Insula, IPS: Infe-rior parietal sulcus, SI: Primary somatosensory cortex, SII: Secondary somatosensory cor-tex, SPL: Superior parietal lobe, SMA: Supplementary motor area.

Peak MNI coordinates

Cluster size Anatomical region X Y Z z-Score peak

A. left RSC154 Left SII −56 −18 28 4.49111 Right SII/SI 66 −22 32 4.33511 Left SI −58 −20 44 4.32146 Right Ins 38 14 8 3.85

B. right RSC1532 Left SII −52 −28 26 4.79

Left IPS/SI −54 −32 48 4.17581 SMA −8 14 44 4.60291 Left FEF −24 −6 68 4.57471 Left IFG −54 6 24 4.46352 Right SI/SII 58 −12 36 4.37457 Right IPS 28 −34 42 4.34182 Right IFG 50 10 20 4.30132 Right Ins 34 22 0 4.11180 Left Ins −24 20 4 4.08120 Left IPL −28 −82 40 3.97169 Right SPL 14 −66 56 3.76

C. Precuneus124 Right IPS 32 −34 42 4.96135 Right SI 58 −12 30 4.36157 Right Ins 36 36 −4 4.33361 Left IPS −40 −28 34 4.05169 SMA 8 8 46 3.94

D. left IFG185 Right SI 66 −32 38 4.30

66 −18 40 4.20

221T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

To our knowledge, and as reviewed byMcNorgan (2012), there haveonly been three fMRI reports on imagery in the somatosensory domain.The studies of Newman et al. (2005) and Olivetti Belardinelli et al.(2009) investigated rather abstract aspects of somatosensory imageryby letting subjects compare haptic/form properties or by cueing the

imagery of active touch. Yoo et al. (2003) reported the only study com-paring tactile imagery to perception. Specifically, the authors usedbrushing stimuli in a block design and reported an initial indication ofperceptual grounding. Our study substantially extends this work asour design allowed for investigating the mental construction processand required a mental representation of high granularity. Furthermore,the current study advances the study by Yoo et al. by providing a behav-ioral report for successful imagery and control conditions, which enableto differentiate between construction and representation-related acti-vations, as well as the study of their interaction by means of PPIanalyses.

Perceptual grounding of the mental representation

Previous work has focused on the activation of primary sensory cor-tices during mental imagery as an activation overlap in early sensoryareas is interpreted as grounding of the mental representation inperception-specific neural processes (McNorgan, 2012). Such activationis seen to support the fact that mental representations are rather‘depictive’ in their nature than propositional (Kosslyn and Thompson,2003). Inclusion of a control condition allowed for testing whether SIis recruited during imagery by contrasting IMAGERY versus CONTROL.In addition, the computation of a conjunction analysis enabled us toidentify the spatial overlap in the neural representations for imageryand perception, as the content to bementally representedwasmatchedbetween PERCEPTION and IMAGERY. In fact, we found activation of SI inboth analyses and this activation topologicallymatched the SI activationduring perception within the hand region of BA1. Considering the tem-poral aspects of the SI responses, we found indistinguishable dynamicsand amplitude during IMAGERY and PERCEPTION while observingsmaller responses in the CONTROL condition. This is particularlyinteresting as a recent visual imagery study by Albers et al. (2013)found temporally overlapping BOLD responses in early visual cortices

222 T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

when working memory content was derived from either imagery orperception. Taken together, our findings indicate bottom-up (percep-tion) and top-down (imagery) induced activations in early sensory cor-tices to follow similar BOLD dynamics.

Additionally,we observed activation in a sub-region of the lateral oc-cipital complex (referred to as LOtv; Amedi et al., 2001, 2002), whichhas been related to modality independent processing of object form(Amedi et al., 2007; Tal and Amedi, 2009) and also during visual imag-ery (Hassabis et al., 2007; Lacey et al., 2009). Lacey et al. (2010) ad-dressed the controversy of whether haptic shape perception dependson visual imagery by investigating LOC activity and connectivity modu-lations by the familiarity of objects (Deshpande et al., 2010). Their fMRIstudy supports that shape representation in LOC can be assessed byeither top-down (visual imagery) or bottom-up (haptic perception).This is in agreement with our results indicating LOtv as part of the con-tent representing network shared by imagery and perception.

The brain's construction network in imagery

Imagery is not themere representation of mental content but ratheran active construction process. The notion of ‘construction’ has been de-veloped in the study of episodic memory and future thinking wherecomplex mental assemblies are composed (Buckner and Carroll, 2007;Hassabis andMaguire, 2009; Schacter et al., 2012). It appears that a sim-ilar brain network supports episodic memory processing (Hassabis andMaguire, 2009; Hassabis et al., 2007; Summerfield et al., 2009), simula-tion of future events (Schacter et al., 2012), scene construction, perspec-tive taking (Buckner and Carroll, 2007), or spatial imagery (Thompsonet al., 2009). The common network is referred to as a ‘core’ constructionsystem (Hassabis andMaguire, 2007, 2009) and comprises medial tem-poral lobe (MTL) regions (hippocampus and parahippocampal gyrus),retrosplenial and posterior parietal cortices, ventromedial prefrontalcortex and the precuneus (Hassabis and Maguire, 2009; Mullally andMaguire, in press; Schacter et al., 2012).

The observation that this network highly overlaps with the defaultmode network (Raichle et al., 2001) fostered the discussion on thefunctional specificity of its components. It has been suggested thatconstruction-related regions flexibly couple with specific brain regionsdependent on the imagery demands, such as perspective taking(Schacter et al., 2012). In line with this hypothesis our study focusedon the imagery of sensory content to test for interactions ofconstruction-related and grounding-related brain regions. The employedtask compelled subjects to reactivate a sensory experience and therebydiffers from classic mental construction paradigms during which sub-jects reconstruct and explore scenes in episodic memory or future simu-lations. They overlap in recalling self-related memory and activelygenerate a vivid mental representation. Kosslyn (2005) also emphasizesthe allocation of attentionwithin amental image as intrinsic to the imag-ination process. While visual imagery requires attentional shifting in amental visuospatial representation, our tactile task necessitated allocat-ing attention within (peri)-personal space towards the fingertip. Thesetask differences need to be consideredwhendiscussing the functional in-terpretation of the core construction network structures.

The MTL regions are well known for their involvement in tasks suchas spatial navigation and episodic memory. Schacter and Addis (2009)reviewed the potential dissociation of the contributions of the hippo-campus and parahippocampal cortex to construction processes, sug-gesting that the latter allows access to contextual associations that arethen combined with further details by the hippocampus (see alsoAddis and Schacter, 2012). In line with this, Mullally and Maguire(in press) suggest that the hippocampus mainly contributes to spatialcoherence by binding together diverse temporal and spatial aspects.The cued recall imagery task that was employed here did not posemajor binding demands and intense training minimized subjects' recallefforts. The contrast IMAGERY N CONTROL revealed activation in theparahippocampal region and thereby provides further evidence for

the modality independent involvement of MTL for mental imagery. Ad-ditionally, our results suggest that simple sensory representations canbe constructed independently of the hippocampus.

The precuneus and the RSC are known to be highly connected to var-ious brain sites relating to functional involvement in memory and con-struction processes. Together with the posterior cingulate cortex (BA23and 31), the RSC (BA29 and 30) is described as part of the parieto-medial temporal pathway of visual information processing (Kravitzet al., 2011). Although no unifying description of RSC's functional rolehas yet been given, its activity has been related to perspective taking,shifting between viewpoints and reference frames (reviewed by Vannet al. (2009)).

A recentmeta-analysis onmental imagery in differentmodalities re-ports modality-independent activation of the left IFG (McNorgan,2012). While the IFG is not described as a core component of the con-struction network, it is part of the lateral prefrontal cortex (lPFC) andis known for its role in working memory (WM), including tactile WM(Auksztulewicz et al., 2011; Preuschhof et al., 2006; Spitzer et al.,2010, 2014). Conceptually an involvement of WM in imagery is plausi-ble because the mental representation has to be maintained for con-scious access and deliberate content processing (Baars and Franklin,2003; Likova, 2012). Imagery can even be regarded as the active/dynamic component of WM in line with the view of construction-related areas to update/manipulate WM content (Albers et al., 2013;Likova, 2012; Tong, 2013). Evidence for information maintenance inlPFC during WM was recently given by Lee et al. (2013). The authorsused a decoding approach to distinguish where different types of infor-mation are coded during a WM task. Their results indicated that infor-mation on rather abstract, non-perceptual properties appear to bemaintained in lPFC, which is in line with reports on the prefrontal cod-ing of analog stimulus properties during tactile WM (Spitzer et al.,2013). In our imagery study, such an abstract aspect could be the con-ceptual information ofwhat should be imagined, such as the associationof the color cue and the corresponding tactile percept. However, theIFG's role for an abstract representation of stimulus properties or stimu-lus identity remains speculative and encourages future investigations,especially when considering it's coupling to SI as revealed by our con-nectivity analysis.

While our paradigmwas not designed to dissociate supramodal im-agery processes from those specific for the tactile modality, the identi-fied imagery network strikingly overlaps with the well described coreconstruction network (Hassabis and Maguire, 2009). Also, Daselaaret al. (2010) and Zvyagintsev et al. (2013) specifically aimed to identifysupramodal imagery networks when contrasting visual with auditoryimagery and gave support for a network highly overlapping with thedefault mode network. Conceptually it appears difficult to depictconstruction processes as truly supramodal or amodal, as similarlydiscussed for mental representations to be either abstract/symbolic ormodality-specific (perceptually grounded) or both. While our subjectsreceived visual cues, the construction processwas directed to the tactiledomain as subjects were necessitated to reactivate fine sensory details.

Together our results support the notion that parahippocampal cor-tex, retrosplenial cortices and precuneus contribute, independent ofmodality, to the initiation of mental content inWM. This content is aug-mented with perceptually grounded details via coupling to sensoryareas, where SI might constitute as a dynamic sensory information‘buffer’ (Albers et al., 2013; Kosslyn, 2005; Likova, 2012).

The interaction of construction and content networks

To date, few studies have investigated changes in connectivity forimagery (Gao et al., 2011; Mechelli et al., 2004; Schlegel et al., 2013;Szameitat et al., 2012). We used PPI analyses to identify connectivitychanges related to the construction of the mental representation.More specifically, we were interested in how the constructionsystem interacts with areas related to the content-representation and

223T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

perceptual grounding and therefore defined seed regions within theconstruction network. Firstly, during imagery the precuneus and bilat-eral RSCs showed increased connectivity to a distributed network relat-ed to theprocessing of tactile stimuli (cf. Figs. 3A, B, C and Table 2A, B, C).This emphasizes their functional role for memory retrieval and the ac-tive reconstruction of the mental image. Supplemented with the IFG,they appear as the core of the construction network to assemble amen-tal representation by reactivation of early perceptual areas.

Interestingly, we found a substantial and selective increase in cou-pling of left IFG with right SI, i.e. the same BA1 sub-region activated byimagery. This interaction of IFGwith SImay support perceptual ground-ing by augmenting an abstract mental representation with sensory de-tails that are grounded in primary sensory cortices. These results arethe first to indicate a direct involvement of prefrontal cortex interactionwith primary sensory areas, related to the construction process of amental image.

Conclusion

We presented an imagery task that stringently controlled for thecontent that had to bementally represented.We found this mental rep-resentation to be perceptually grounded, as SI activation was shared byimagery and stimulus-driven perception. Furthermore, we found thecore construction system to be involved in the active construction ofthe mental stimulus representation and the involvement of prefrontalcortex, specifically the left IFG. Most interestingly our analysis revealedincreased connectivity between prefrontal cortex and SI during imag-ery, which is the first indication of a direct mechanism that augmentsan abstract mental representation with perceptually grounded sensorydetails.

Acknowledgments

TTS is supported by the Research Training Group GRK 1589/1“Sensory Computation in Neural Systems” by the German ResearchFoundation (DFG). FB was supported by a grant from the GermanFederal Ministry of Education and Research (BMBF).

References

Addis, D.R., Schacter, D.L., 2012. The hippocampus and imagining the future: where dowestand? Front. Hum. Neurosci. 5, 173.

Albers, A.M., Kok, P., Toni, I., Dijkerman, H.C., de Lange, F., 2013. Shared representationsfor working memory and mental imagery in early visual cortex. Curr. Biol. 23,1427–1431.

Albright, T.D., 2012. On the perception of probable things: neural substrates of associativememory, imagery, and perception. Neuron 74, 227–245.

Amedi, A., Malach, R., Hendler, T., Peled, S., Zohary, E., 2001. Visuo-haptic object-relatedactivation in the ventral visual pathway. Nat. Neurosci. 4, 324–330.

Amedi, A., Jacobson, G., Hendler, T., Malach, R., Zohary, E., 2002. Convergence of visual andtactile shape processing in the human lateral occipital complex. Cereb. Cortex 12,1202–1212.

Amedi, A., Stern,W.M., Camprodon, J.A., Bermpohl, F., Merabet, L., Rotman, S., Hemond, C.,Meijer, R., Pascual-Leone, A., 2007. Shape conveyed by visual-to-auditory sensorysubstitution activates the lateral occipital complex. Nat. Neurosci. 10, 687–689.

Ashburner, J., 2007. A fast diffeomorphic image registration algorithm. Neuroimage 38,95–113.

Auksztulewicz, R., Spitzer, B., Goltz, D., Blankenburg, F., 2011. Impairing somatosensoryworking memory using rTMS. Eur. J. Neurosci. 34, 839–844.

Baars, B.J., Franklin, S., 2003. How conscious experience and working memory interact.Trends Cogn. Sci. 7 (4), 166–172.

Barsalou, L.W., 2008. Grounded cognition. Annu. Rev. Psychol. 59, 617–645.Blankenburg, F., Ruff, C.C., Deichmann, R., Rees, G., Driver, J., 2006. The cutaneous rabbit

illusion affects human primary sensory cortex somatotopically. PLoS Biol. 4, e69.Bodegard, A., Geyer, S., Grefkes, C., Zilles, K., Roland, P.E., 2001. Hierarchical processing of

tactile shape in the human brain. Neuron 31, 317–328.Buckner, R.L., Carroll, D.C., 2007. Self-projection and the brain. Trends Cogn. Sci. 11, 49–57.Burton, H., Sinclair, R.J., 2000. Attending to and remembering tactile stimuli: a review of

brain imaging data and single-neuron responses. J. Clin. Neurophysiol. 17, 575–591.Chen, H., Yang, Q., Liao, W., Gong, Q., Shen, S., 2009. Evaluation of the effective connectiv-

ity of supplementary motor areas during motor imagery using Granger causalitymapping. Neuroimage 47, 1844–1853.

Chu, C., Mourao-Miranda, J., Chiu, Y.C., Kriegeskorte, N., Tan, G., Ashburner, J., 2011. Utiliz-ing temporal information in fMRI decoding: classifier using kernel regressionmethods. Neuroimage 58, 560–571.

Daselaar, S.M., Porat, Y., Huijbers, W., Pennartz, C.M., 2010. Modality-specific andmodality-independent components of the human imagery system. Neuroimage 52(2), 677–685.

Deshpande, G., Hu, X., Lacey, S., Stilla, R., Sathian, K., 2010. Object familiaritymodulates effective connectivity during haptic shape perception. Neuroimage 49(3), 1991–2000.

Eickhoff, S.B., Paus, T., Caspers, S., Grosbras, M.H., Evans, A.C., Zilles, K., Amunts, K., 2007.Assignment of functional activations to probabilistic cytoarchitectonic areas revisited.Neuroimage 36, 511–521.

Eickhoff, S.B., Grefkes, C., Fink, G.R., Zilles, K., 2008. Functional lateralization of face, hand,and trunk representation in anatomically defined human somatosensory areas.Cereb. Cortex 18, 2820–2830.

Formisano, E., Linden, D.E., Di Salle, F., Trojano, L., Esposito, F., Sack, A.T., Grossi, D., Zanella,F.E., Goebel, R., 2002. Tracking the mind's image in the brain I: time-resolved fMRIduring visuospatial mental imagery. Neuron 35, 185–194.

Francis, S.T., Kelly, E.F., Bowtell, R., Dunseath, W.J., Folger, S.E., McGlone, F., 2000. fMRI ofthe responses to vibratory stimulation of digit tips. Neuroimage 11 (3), 188–202.

Friston, K.J., 2011. Functional and effective connectivity: a review. Brain Connect. 1,13–36.

Friston, K.J., Buechel, C., Fink, G.R., Morris, J., Rolls, E., Dolan, R.J., 1997. Psychophysiologicaland modulatory interactions in neuroimaging. Neuroimage 6, 218–229.

Friston, K.J., Holmes, A.P., Price, C.J., Buechel, C., Worsley, K.J., 1999. Multisubject fMRIstudies and conjunction analyses. Neuroimage 10, 385–396.

Gao, Q., Duan, X., Chen, H., 2011. Evaluation of effective connectivity of motor areas dur-ing motor imagery and execution using conditional Granger causality. Neuroimage54, 1280–1288.

Gitelman, D.R., Penny, W.D., Ashburner, J., Friston, K.J., 2003. Modeling regional and psy-chophysiologic interactions in fMRI: the importance of hemodynamic deconvolution.Neuroimage 19, 200–207.

Gläscher, J., 2009. Visualization of group inference data in functional neuroimaging.Neuroinformatics 7, 73–82.

Hassabis, D., Maguire, E.A., 2007. Deconstructing episodic memory with construction.Trends Cogn. Sci. 11, 299–306.

Hassabis, D., Maguire, E.A., 2009. The construction system of the brain. Philos. Trans. R.Soc. Lond. B Biol. Sci. 364, 1263–1271.

Hassabis, D., Kumaran, D., Maguire, E.A., 2007. Using imagination to understand the neu-ral basis of episodic memory. J. Neurosci. 27, 14365–14374.

Kasess, C.H., Windischberger, C., Cunnington, R., Lanzenberger, R., Pezawas, L., Moser, E.,2008. The suppressive influence of SMA on M1 in motor imagery revealed by fMRIand dynamic causal modeling. Neuroimage 40, 828–837.

Kosslyn, S.M., 1973. Scanning visual images: some structural implications. Percept.Psychophys. 14, 90–94.

Kosslyn, S.M., 1976. Can imagery be distinguished from other forms of internal represen-tation? Evidence from studies of information retrieval times. Mem. Cogn. 4, 291–297.

Kosslyn, S.M., 2005. Mental images and the brain. Cogn. Neuropsychol. 22 (3/4), 333–347.Kosslyn, S.M., Thompson, W.L., 2003. When is early visual cortex activated during visual

mental imagery? Psychol. Bull. 129, 723–746.Kosslyn, S.M., Ganis, G., Thompson, W.L., 2001. Neural foundations of imagery. Nat. Rev.

Neurosci. 2, 635–642.Kravitz, D.J., Saleem, K.S., Baker, C.I., Mishkin, M., 2011. A new neural framework for

visuospatial processing. Nat. Rev. Neurosci. 12, 217–230.Lacey, S., Tal, N., Amedi, A., Sathian, K., 2009. A putative model of multisensory object rep-

resentation. Brain Topogr. 21, 269–274.Lacey, S., Flueckiger, P., Stilla, R., Lava, M., Sathian, K., 2010. Object familiarity modulates

the relationship between visual object imagery and haptic shape perception.Neuroimage 49 (3), 1977–1990.

Lee, S.H., Kravitz, D.J., Baker, C.I., 2013. Goal-dependent dissociation of visual and prefron-tal cortices during working memory. Nat. Neurosci. 16 (8), 997–999.

Likova, L.T., 2012. A cross-modal perspective on the relationship between imagery andworking memory. Front. Psychol. 3, 561.

McNorgan, C., 2012. A meta-analytic review of multisensory imagery identifies the neuralcorrelates of modality-specific and modality-general imagery. Front. Hum. Neurosci.6, 285.

Mechelli, A., Price, C.J., Friston, K.J., Ishai, A., 2004. Where bottom-up meets top-down: neuronal interactions during perception and imagery. Cereb. Cortex 14,1256–1265.

Moore, C.I., Crosier, E., Greve, D.N., Savoy, R., Merzenich, M.M., Dale, A.M., 2013. Neocorti-cal correlates of vibrotactile detection in humans. J. Cogn. Neurosci. 25, 49–61.

Moulton, Kosslyn, 2009. Imaging predictions: mental imagery as mental emulation.Philos. Trans. R. Soc. Lond. B Biol. Sci. 364, 1273–1280.

Mullally, S.L., Maguire, E.A., 2014. Memory, imagination, and predicting the future: a com-mon brain mechanism? Neuroscientist (in press, epub ahead of print).

Newman, S.D., Klatzky, R.L., Lederman, S.J., Just, M.A., 2005. Imagining material versusgeometric properties of objects: an fMRI study. Brain Res. Cogn. Brain Res. 23,235–246.

Nichols, T., Brett, M., Andersson, J., Wager, T., Poline, J.B., 2005. Valid conjunction inferencewith the minimum statistic. Neuroimage 25 (3), 653–660.

Olivetti Belardinelli, M., Palmiero, M., Sestieri, C., Nardo, D., Di Matteo, R., Londei, A.,D'Ausilio, A., Ferretti, A., Del Gratta, C., Romani, G.L., 2009. An fMRI investigation onimage generation in different sensory modalities: the influence of vividness. ActaPsychol. (Amst) 132, 190–200.

Podgorny, P., Shepard, R.N., 1978. Functional representations common to visual percep-tion and imagination. J. Exp. Psychol. Hum. Percept. Perform. 4, 21–35.

224 T.T. Schmidt et al. / NeuroImage 98 (2014) 216–224

Preuschhof, C., Heekeren, H.R., Taskin, B., Schubert, T., Villringer, A., 2006. Neural correlates ofvibrotactile working memory in the human brain. J. Neurosci. 26, 13231–13239.

Raichle, M.E., MacLeod, A.M., Snyder, A.Z., Powers, W.J., Gusnard, D.A., Shulman, G.L.,2001. A default mode of brain function. Proc. Natl. Acad. Sci. U. S. A. 98, 676–682.

Schacter, D.L., 2012. Adaptive constructive processes and the future of memory. Am.Psychol. 67, 603–613.

Schacter, D.L., Addis, D.R., 2009. On the nature of medial temporal lobe contributions tothe constructive simulation of future events. Philos. Trans. R. Soc. Lond. B Biol. Sci.364 (1521), 1245–1253.

Schacter, D.L., Addis, D.R., Hassabis, D., Martin, V.C., Spreng, R.N., Szpunar, K.K., 2012. Thefuture of memory: remembering, imagining, and the brain. Neuron 76, 677–694.

Schlegel, A., Kohler, P.J., Sergey, V., Fogelson, V., Prescott, A., Dedeepya, K., Tse, P.E., 2013.Network structure and dynamics of themental workspace. Proc. Natl. Acad. Sci. U. S. A.110 (40), 16277–16287.

Shepard, R.N., Metzler, J., 1971. Mental rotation of three-dimensional objects. Science 171,701–703.

Spitzer, B., Wacker, E., Blankenburg, F., 2010. Oscillatory correlates of vibrotactile frequen-cy processing in human working memory. J. Neurosci. 30, 4496–4502.

Spitzer, B., Gloel, M., Schmidt, T.T., Blankenburg, F., 2013. Working memory coding of an-alog stimulus properties in the human prefrontal cortex. Cereb. Cortex (Epub aheadof print).

Spitzer, B., Goltz, D., Auksztulewicz, R., Blankenburg, F., 2014. Maintenance and manipu-lation of somatosensory information in ventrolateral prefrontal cortex. Hum. BrainMapp. 35 (5), 2412–2423.

Summerfield, J.J., Hassabis, D., Maguire, E.A., 2009. Cortical midline involvement in auto-biographical memory. Neuroimage 44, 1188–1200.

Summerfield, J.J., Hassabis, D., Maguire, E.A., 2010. Differential engagement of brain re-gions within a ‘core’ network during scene construction. Neuropsychologia 48,1501–1509.

Szameitat, A.J., McNamara, A., Shen, S., Sterr, A., 2012. Neural activation and functionalconnectivity during motor imagery of bimanual everyday actions. PLoS One 7,e38506.

Tal, N., Amedi, A., 2009. Multisensory visual–tactile object related network in humans:insights gained using a novel crossmodal adaptation approach. Exp. Brain Res. 198,165–182.

Thompson,W.L., Slotnick, S.D., Burrage, M.S., Kosslyn, S.M., 2009. Two forms of spatial im-agery: neuroimaging evidence. Psychol. Sci. 20, 1245–1253.

Tong, F., 2013. Imagery and visual working memory: one and the same? Trends Cogn. Sci.17 (10), 189–490.

Vann, S.D., Aggleton, J.P., Maguire, E.A., 2009. What does the retrosplenial cortex do? Nat.Rev. Neurosci. 10, 792–802.

Wacker, E., Spitzer, B., Lützkendorf, R., Bernarding, J., Blankenburg, F., 2011. Tactile motionand pattern processing assessed with high-field FMRI. PLoS One 6, e24860.

Yoo, S.S., Freeman, D.K., McCarthy III, J.J., Jolesz, F.A., 2003. Neural substrates of tactile im-agery: a functional MRI study. Neuroreport 14, 581–585.

Zvyagintsev, M., Clemens, B., Chechko, N., Mathiak, K.A., Sack, A.T., Mathiak, K., 2013.Brain networks underlying mental imagery of auditory and visual information. Eur.J. Neurosci. 37, 1421–1434.