R

Os

Na

Bb

c

d

De

h

••••

a

ARR1AA

KTMOAAT

oB

0h

Behavioural Brain Research 261 (2014) 328– 335

Contents lists available at ScienceDirect

Behavioural Brain Research

j ourna l h o mepa ge: www.elsev ier .com/ locate /bbr

esearch report

bserved manipulation of novel tools leads to mu rhythmuppression over sensory-motor cortices

orma Naima Rüthera,b,∗, Elliot Clayton Brownb,c, Anne Kleppd, Christian Bellebaume

Institute of Cognitive Neuroscience, Department of Neuropsychology, Faculty of Psychology, Ruhr-University Bochum, Universitaetsstrasse 150, 44801ochum, GermanyInternational Graduate School of Neuroscience, Ruhr-University Bochum, Universitaetsstrasse 150, 44801 Bochum, GermanyMaryland Psychiatric Research Centre, University of Maryland School of Medicine, 55 Wade Avenue, Baltimore, MD 21228, USAInstitute of Clinical Neuroscience and Medical Psychology, Medical Faculty, Heinrich Heine University Düsseldorf, Universitaetsstrasse 1, 40225üsseldorf, GermanyInstitute of Experimental Psychology, Heinrich Heine University Düsseldorf, Universitaetsstrasse 1, 40225 Düsseldorf, Germany

i g h l i g h t s

Observational learning of tool manipulation leads to mu rhythm modulation.Similar modulation is seen for visual exploration of tools.Mu rhythm modulation takes place over sensory-motor cortices.The effect is seen within 200 ms after tool picture presentation.

r t i c l e i n f o

rticle history:eceived 5 November 2013eceived in revised form9 December 2013ccepted 23 December 2013vailable online 3 January 2014

eywords:oolsu rhythmbservation

a b s t r a c t

Tool stimuli can be analyzed based on their affordance, that is, their visual structure hinting at possibleinteraction points. Additionally, familiar tools can initiate the retrieval of stored object–action associa-tions, providing the basis for a meaningful object use. The mu rhythm within the electroencephalographicalpha band is associated with sensory-motor processing and was shown to be modulated during the sightof familiar tool stimuli, suggesting motor cortex activation based on either affordance processing or accessto stored conceptual object–action associations. The current study aimed to investigate the impact of suchassociations, acquired by observation of manipulation, in a training study controlling for inherent objectaffordances and previous individual differences in object-related experience. Participants observed themanipulation of a set of novel tool objects and visually explored a second set of novel tools for which onlyfunctional information was provided. In contrast to non-trained objects, observed objects modulated the

ctionffordanceraining

mu rhythm over left sensory-motor cortex within 200 ms after training. Additionally, both observed andvisually explored objects modulated mu rhythm over right sensory-motor cortex in the same time win-dow to some extent, with the effect being stronger for the latter. This result suggests that motor cortexactivation in visual processing of tools can result from observation of tool manipulation. However, murhythm modulation, albeit with a different and less clear left-lateralized pattern, is also seen when thetools were only made visually familiar and when information was restricted to the tools function.

© 201

∗ Corresponding author at: Institute of Cognitive Neuroscience, Departmentf Neuropsychology, Ruhr-University Bochum, Universitaetsstrasse 150, D-44801ochum, Germany. Tel.: +49 234 32 23174; fax: +49 234 32 14622.

E-mail addresses: naima.ruether@rub.de, norma.ruether@rub.de (N.N. Rüther).

166-4328 © 2014 The Authors. Published by Elsevier B.V. ttp://dx.doi.org/10.1016/j.bbr.2013.12.033

Open access under CC BY-NC-SA licens

4 The Authors. Published by Elsevier B.V.

1. Introduction

Tools represent a special class of object as they are defined by theway one interacts with them, and the purpose they serve. Hence,when seeing manipulable tools, information about potential actioninteraction points, or “affordances” [1], is extracted. Furthermore,familiar tools can initiate the retrieval of stored information about

Open access under CC BY-NC-SA license.

associations between the object, an object-related action and anaction goal, which is based on previous object interaction expe-rience [2]. Several functional neuroimaging studies have foundactivations of a frontoparietal network by the sight of familiar

e.

Brain

ttbpatoc

SraitcFriDecektarrthtohwi

mr(cnbiitaptoptotuoewbusooasvs[

N.N. Rüther et al. / Behavioural

ools, most likely representing the stored motor action representa-ion associated with the tool (for review, see [3–6]). Furthermore,ehavioral studies have shown that the mere sight of tools canrime, or “potentiate”, motor responses to object parts, affording anction through motor facilitation, therefore supporting the idea ofhe automatic extraction of affordances when seeing manipulablebjects. Together, these findings suggest that action informationan be an integral part of object representations.

One potential mechanism for this is described in the Two Actionystems Model [2], which proposes that the action system is sepa-able into two functionally and anatomically distinguishable units,

Structural System and a Functional System. The Structural Systems devoted to perform an online analysis of visual object struc-ures for a rapid interaction and is proposed to be not exclusivelyognitive. During repeated and skilled interaction with objects, aunctional System extracts the characteristics of an action thatemain constant across object-directed interactions, hence provid-ng the basis for the formation of conceptual object representations.uring manual object interaction the extracted functional knowl-dge and object-associated actions thus become integral parts ofonceptual representations, which do, however, comprise differ-nt types of knowledge about an object in different modalities. Aey component for the evolution of human tool use behavior ishe capacity to learn through the observation of others performingn action [7]. This therefore suggests that the formation of mentalepresentations of tools is not only bound to direct interactive expe-ience with the object, but can also result from indirect experiencehrough the observation of others’ interactions with the object. Itas been shown that certain neurons in area F5 of the monkey cor-ex, called “mirror neurons”, discharge during the observation ofthers performing object-related actions [8–10], a property thatas been proposed to represent a direct matching mechanism byhich the observed motor action is mapped onto the observer’s

nternal motor representations [11,12].Electroencephalography (EEG) studies investigating sensory-

otor processing in humans have analyzed the mu rhythm, whichefers to oscillatory brain activity in the alpha and beta band8–12 Hz and ∼20 Hz), in electrodes placed over the sensory-motorortices. Event-related desynchronization (ERD) of the oscillatoryeuronal firing that underlies the mu rhythm can, for example,e found during movement execution (e.g., [13–15]) but also dur-

ng the observation of movements (e.g., [16,17]) and movementmagery (e.g., [18]). These properties of the mu rhythm show thathe motor system can also be engaged in the absence of activection execution, and some researchers argue that this activityartly represents an index of the mirror neuron system [19]. Fur-hermore, findings on the suppression of the mu rhythm duringbject-related in contrast to meaningless actions have been inter-reted as activation of goal-directed motor programs related tohe mirror neuron system [16]. Interestingly, the simple viewingf tool stimuli, even without the demand to interact, can also leado mu rhythm suppression between 140 ms and 300 ms after stim-lus presentation [20], possibly reflecting the automatic access tobject-associated actions. Further support for this comes from earlyvent-related potentials (ERPs) during the processing of tools forhich sources were described in the left postcentral gyrus and

ilateral premotor cortex. However, so far it is not clear if the mod-lation of the mu rhythm during the sight of tool stimuli relates toimple visual affordance processing, or to experience-dependentbject–action associations that can be acquired during observationf object manipulation. As mu suppression has been found both forction execution and for the sight of known objects [13–15,20], it

eems likely that active object experience plays a role in the acti-ation of motor cortex during visual object processing. While muuppression was also seen for the observation of meaningful actions16,17,21], it is not clear what effect the observation of object

Research 261 (2014) 328– 335 329

interactions has on the formation of object–action associations. Themirror neuron hypothesis [8–10] and related simulationist theo-ries [22] posit that direct motor experience may not be necessaryto form such associations, but rather observation of object-relatedmotor actions may be sufficient. However, this is a topic that is stillhotly debated.

A major problem with using familiar tools in experimentsaddressing experience effects is that interindividual differencesin previous knowledge cannot be controlled for. Recently, severaltraining studies have used novel objects to overcome this problem(e.g., [23–26]). In these studies, subjects systematically receiveddifferent types of object-related experience. Even though theobject-related knowledge gained cannot be compared to existingconceptual representations, training studies can lead to importantinsights, providing a model for the mechanisms involved in theemergence of new object representations. One such study foundthat early activation over sensory-motor and occipito-parietalbrain regions during object processing only occurred in the con-dition in which object-related movements had been previouslytrained through pantomimed actions, as compared to training withnon-functional actions that were not relevant to the object beingseen [25]. The authors here concluded that object processing canbe altered with respect to the specific sensory-motor interactionswith objects during knowledge acquisition.

In the current study, novel, tool-like objects ([23] similar to theones used by Weisberg et al. [24]) were used to investigate whetherobserving the manipulation of objects had a differential impact onneural correlates of object–action associations, as reflected in thesensory-motor mu rhythm, when compared to the mere learningabout object function, but without seeing the object being used.Each participant in the present study received three sessions ofobject-related training over 3 days. For one set of objects, par-ticipants observed an experimenter manipulating the invented,unfamiliar tools during training (observation of manipulation train-ing objects, OBS). For another set of invented and unfamiliar toolobjects, participants were informed about the tools’ function butonly visually explored the object (visual training objects, VIS).Finally, a third object set served as an untrained control set (nottrained objects, NO). Before and after training, participants per-formed a visual matching task during which photographs of all toolswere shown and brain activity was measured with EEG.

We expected a post training modulation of ERD of activity inthe sensory-motor mu rhythm depending on the specific trainingexperience with the objects. More specifically, we predicted thatthe processing of objects that had been observed to be manipulated(OBS) would lead to a larger mu rhythm suppression as comparedto objects that had not been seen to be manipulated (VIS and NO).

2. Materials and methods

2.1. Participants

Twenty-one right-handed students (11 women, mean age 25.33,SD = 3.71, range = 20–34) with normal or corrected-to-normalvision participated in the study. No participant had a historyof neurological or psychiatric diseases. All were informed aboutthe testing procedure and signed a written consent. Participantsreceived either course credit or financial reimbursement. The studywas approved by the Ethics Committee of the Medical Faculty atRuhr University Bochum, Germany.

2.2. Overall procedure

Using a children construction toy (K’nexTM), similar to the stud-ies in [23,24], 36 novel objects were constructed. Participants

3 Brain Research 261 (2014) 328– 335

caEwswtdd

2

tttar

taonioivocrFar1

2

oosisswmpaosc

2

obhttsewtta

30 N.N. Rüther et al. / Behavioural

ompleted a visual picture matching task with the objects, beforend after training, during which brain activity was recorded withEG. Training occurred over three sessions on three different daysith some of the invented, unfamiliar objects. The average time

pan between the first and the last training was 8.1 days (SD = 3.7),hereas the mean interval between the last training session and

he post training EEG assessment was 3.2 days (SD = 2.5). The meanuration from the first training to the second EEG session was 10.7ays (SD = 4.4) on average.

.3. Novel object stimuli

Each novel object served one of six possible functions relatedo a specific manipulation (separation, pushing, tagging, crushing,ransporting and moving other small objects, e.g., paper rolls orable tennis balls). Every object had at least one moveable part or

handle that could be used for the specific type of manipulationelated to object function.

A separate group of subjects (N = 33) evaluated each object inerms of singularity, visual complexity and possible real-objectssociated affordances (how much the object resembles a knownbject and which affordances it raises) with a structured question-aire. Singularity was assessed by asking how outstanding an object

s compared to the other objects (on a 7-point scale) and what otherbject was most similar to a given object and how strong the sim-larity was (again on a 7-point scale). Participants also rated theisual complexity (“How complex is the object visually?”) of thebjects on a scale from 1 (low visual complexity) to 7 (high visualomplexity). Additionally, participants rated how much the objectesembled a real object and, if yes, how strong this association was.inally, participants were asked to write down if they could think of

specific function the depicted object might have. Based on theseatings, the 36 objects were divided into three object sets including2 objects each matched with respect to the assessed parameters.

.4. Training sessions

The training sessions lasted 80 min each, and served to inducebject representations in the participants. Only two of the threebject sets were used for training, whereas the third object seterved as a control condition (NO), and were therefore includedn the visual matching task. Training for the two trained objectets differed to induce qualitatively different types of object repre-entations. One set of objects (VIS) was part of a “visual training”hile the second object set (OBS) was part of an “observation ofanipulation training”. Hence, after three training sessions, each

articipant had completed three “visual trainings” with 12 objectsnd three “observation of manipulation trainings” with 12 differentbjects. Each object appeared once in each of the three training ses-ion. Assignment of the object sets to the training conditions wasounterbalanced and randomized across participants.

.4.1. Observation of manipulation trainingIn the “observation of manipulation training”, the execution

f object manipulation was demonstrated for each of the objectsy an experimenter. No participant executed the manipulationim-/herself. Each object was placed on a table in front of the par-icipants while naming the specific function of the object. In thisraining, however, the experimenter then explained the discreteteps of manipulation by a standardized verbal description whileach step of manipulation was demonstrated simultaneously,

hich illustrated the intended function of the object. Subsequently,

he object was manipulated by the experimenter for 90 s whilehe participants were instructed to carefully watch. To ensure thatll participants were attending to the manipulation, they were

Fig. 1. Example trials of the visual matching task.

instructed to silently count how often each object was manipulatedwithin the 90 s.

2.4.2. Visual trainingFor the objects with “visual training”, each was placed on the

table in front of participants, one at a time, and the specific func-tion of the object was named, but without any indication of how theobject could be manipulated. Then participants had 90 s to visuallyexplore and simultaneously describe the object. As for the “obser-vation of manipulation training” object set, it was prohibited totouch the objects. In the cases where participants were not ableto fill the required 90 s with verbal description, the experimenterprompted a more detailed depiction of the object by asking specificquestions about the object (e.g., “how many of these yellow partsdoes the object have”).

2.5. The matching task

Before and after training, participants performed an identi-cal visual matching task (similar to [23,24]) with pictures of theinvented tool-like objects, to investigate training-induced changesof object processing (see Fig. 1). The matching task was presentedon a computer screen, and each trial started with an exclamationmark, with a varying duration of 600–1600 ms, which was followedby a fixation cross presented for 500 ms. A photograph of one of theobjects was then shown for 1000 ms, followed by another fixationcross with varying duration between 500 and 700 ms. Afterwards,a second photograph appeared showing an object from a differentperspective for 1000 ms. Participants were then required to makea response within 1500 ms (left or right button press) to determinewhether the two photographs were of the same object or not.

The task consisted of 3 blocks containing 72 trials each (216 trialsin total). On each trial, two photographs were shown, amounting to432 object pictures in total (12 pictures per object). Half of the trialsshowed the same object on both photographs. For non-matchingphotographs, both objects were from the same training group.To account for possible repetition effects, each object was pho-tographed from four different perspectives (see also [23]). Hence,each individual picture was shown three times, with the picturesbeing randomly assigned to either the first or second presentationper trial. The software “Presentation” was used to control stimu-lus timing and response recording (version 13.0, NeurobehavioralSystems, Albany, CA, USA).

After the post training EEG measurement, participants receiveda paper–pencil questionnaire including colored pictures depictingeach of the 36 objects from one perspective. They were then asked

to indicate, if the depicted object was part of the observation orvisual exploration training condition (OBS, VIS) or if the depictedobject was not part of the training (NO). Participants indicatedthis by marking the respective training condition (“observation of

Brain

mtta

2

iIGoaGwFCta1Gf

2

hi

saAbicbbnbotcitsattdsods1t1ctluatlta

p = .001) which shows that participants were overall more accurateafter object training in the post compared to the pre training ses-sion. A significant main effect of STIMULUS TYPE (F(2,40) = 4.733;p = .014) indicates differences in accuracy between object types,

Table 1Mean RT data (ms) and accuracy (%) for the performance in the visual matching task(standard deviation in brackets).

NO VIS OBS

RTs (ms)Before learning 518 (117) 518 (104) 521 (116)

N.N. Rüther et al. / Behavioural

anipulation”, “visual exploration” or “not part of the training”)hat were listed next to the object pictures. This procedure was usedo ensure that subjects were able to distinguish between the objectsnd remembered their occurrence during training (see [24]).

.6. EEG recording

EEG assessment during the performance of the visual match-ng task took place in the psychophysiology laboratory of thenstitute of Cognitive Neuroscience at Ruhr University Bochum,ermany. EEG was recorded at 30 scalp sites, with a sampling ratef 500 Hz, using a Brain Amp “Professional Powerpack” amplifiernd the “Brain Vision Recorder” software (Brain Products, Munich,ermany). Using an elastic nylon electrode cap, Ag–Cl electrodesere arranged according to the International 10–20 System (F7,

3, Fz, F4, F8; FT7, FC3, FCz, FC4, FT8; T7, C3, Cz, C4, T8; TP7, CP3,Pz, CP4, TP8; P7, P3, Pz, P4, P8; PO7, PO3, POz, PO4, PO8). Mas-oid electrodes served as references and one ground electrode wasttached to recording site FPz. All impedances were kept below0 k�. “Brain Vision Analyzer 2” software (Brain Products, Munich,ermany) and MATLAB (Mathworks, Natick, MA, USA) were used

or off-line EEG data analysis.

.7. Analysis of ERD/ERS

Electrodes of interest for ERD/ERS analysis were at sites C3 (leftemispheric) and C4 (right hemispheric), since ERD at these sites

s related to activation of primary sensory-motor cortex [15,27].A Butterworth Zero Phase Filter (low cutoff 0.5305, time con-

tant 0.300 s with 12 dB/oct; high cutoff 40 Hz, 12 dB/oct) waspplied to the EEG raw data. Then, an Independent Componentnalysis (ICA) was performed for each participant to remove eyelink artifacts from the raw data [28]. An ICA converts the EEG data

nto a matrix containing spatially fixed and temporally independentomponents where the number of EEG channels matches the num-er of components. Eye blink related distortions are characterizedy symmetric, frontally pronounced positivity. The eligible compo-ent showing these characteristics was removed from the data sety performing an ICA back transform. Subsequent visual inspectionf the data ensured that ocular artifacts were largely removed byhis approach. By careful visual inspection, trials including mus-ular, movement-related or other artifacts on the electrodes ofnterest (C3, C4) were then removed from the data set. Data werehen segmented into different conditions according to the objectets (i.e. OBS, VIS and NO). Only the first object picture per trial wasnalyzed, because we were mainly interested in modulations dueo different training conditions while processing of the second pho-ograph might have been confounded by the task demands, e.g., theecision process, and by whether the two presented pictures repre-ented matches or non-matches. A 1000 ms epoch before stimulusnset was taken as the reference interval for ERS/ERD analysis,uring which the exclamation mark and fixation cross were pre-ented. The main epoch used for analysis for the first picture was000 ms in duration, and thus covered exactly the presentationime of the picture. The alpha frequency bands of 8–10 Hz and0–12 Hz were analyzed, which both reflect sensory-motor pro-essing in frontoparietal cortices [29], but can be differentiated byheir somatotopical specificity for foot and hand movements. Theower mu rhythm does not show differences in mu rhythm mod-lation for finger or hand movements over motoric hand or footreas, hence representing a more widespread, non-specific activa-

ion. On the contrary, the upper mu rhythm is characterized by aarger modulation for hand movements compared to the modula-ion during the observation foot movements over the motoric handreas [14].

Research 261 (2014) 328– 335 331

To compute ERD/ERS, the data were bandpass filtered andamplitude samples were squared. The power samples across alltrials were then averaged. ERD was calculated as power decreasewhereas ERS was calculated as power increase in relation to thereference interval [30]. The formula for ERD/ERS calculation was[(power of frequency band − power of reference interval)/power ofreference interval × 100] [15]. After computation of ERD/ERS, thetraces were smoothed using a moving average with a time windowof 126 ms. Topographic maps were created using spherical splineinterpolation in “Brain Vision Analyzer 2” software (Brain Products,Munich, Germany).

2.8. Statistical analysis

Statistical analyses were computed with IBM® SPSS® Statistics20.0.0. Mean reaction times and accuracy in the visual match-ing task were analyzed with a 2 × 3 repeated-measures analysisof variance (ANOVA) comprising the factors TIME (pre, post) andSTIMULUS TYPE (OBS, VIS, NO). Reaction times were averagedacross match and mismatch trials and across correct and wrongresponses. As for reaction time data, accuracy was analyzed acrossmatch and mismatch trials.

ERD/ERS analysis was conducted separately for epochs of 200 mswithin the 1000 ms following stimulus presentation, yielding 5epochs of 200 ms each trial (see Fig. 3). A repeated-measuresANOVA comprising the factors TIME (pre, post), STIMULUS TYPE(OBS, VIS, NO) and ELECTRODE (C3, C4) was then conducted for each200 ms epoch within the specific frequency bands of 8–10 Hz and10–12 Hz. Results for all main effects and interactions including atleast one of the factors TIME or STIMULUS TYPE will be reported. Ofparticular interest were those interaction effects including both fac-tors TIME and STIMULUS TYPE. Paired t-tests were used to resolveinteractions. Greenhouse–Geisser corrected results and degrees offreedom are reported in case of significant violations of sphericityas measured by the Mauchly’s test [31].

3. Results

3.1. Behavioral data

3.1.1. Performance in the matching taskMean accuracy and mean reaction times (in ms) during the

matching task before and after training are depicted in Table 1. AnANOVA with the factors STIMULUS TYPE (OBS, VIS, NO) and TIME(pre, post) for reaction time data revealed a main effect of TIME(F(1,20) = 15.052; p = .001), showing that participants were faster torespond in the visual matching task after object training. No furthersignificant effects for reaction time data were found (all p > .360).

The respective ANOVA for accuracy data with the same fac-tors also yielded a significant main effect of TIME (F(1,20) = 16.615;

After learning 480 (135) 467 (125) 472 (128)

Accuracy (%)Before learning 92.53 (4.27) 93.52 (4.10) 93.52 (4.42)After learning 94.91 (3.18) 97.42 (2.25) 96.76 (3.12)

3 Brain

alafF(

3

r(et

3

3

pa

owtplSCotaemf(ceb

aHidVcpniff6gtpm

lEbTKTcrss

32 N.N. Rüther et al. / Behavioural

nd post hoc t-tests revealed that participants performed generallyess accurately for NO in comparison to VIS (t(20) = −3.34; p = .003)nd compared to OBS (t(20) = −2.41; p = .026). No significant dif-erence in the accuracy between OBS and VIS was found (p = .324).urthermore, no significant interaction of both factors was foundp = .420).

.1.2. Assignments of objects to the training conditionAfter the second EEG session was completed, participants cor-

ectly recognized on average 10.95 OBS (SD = 1.16), 10.81 VISSD = 2.25) and 11.14 NO (SD = 1.01) (maximum score is 12 forach object set). No difference between object sets with respecto assignment performance was found (p = .701).

.2. ERD/ERS analysis

.2.1. Analysis of the lower mu frequency band of 8–10 HzFig. 2 displays the grand average ERD/ERS traces for pre and

ost training sessions at electrodes C3 (A) and C4 (B) during theccomplishment of the visual matching task for OBS, VIS and NO.

In the frequency band of 8–10 Hz, none of the main effectsr two-way interactions reached significance for any of the timeindows (all p > .05). However, a significant three-way interac-

ion was found in the time window of 0–200 ms (F(2,40) = 3.460; = .041; all p > .05 for the other time windows). For the reso-ution of this interaction, one-factorial ANOVAs with the factorTIMULUS TYPE were conducted, separately for the electrodes3 and C4 and for pre and post training acquisitions. This res-lution revealed no pre training differences between stimulusypes at C3 and C4 electrodes (both p > .400). At electrode C3,

significant linear trend for differences between stimulus typesmerged in the post training data (F(1,20) = 4.735; p = .042). Maxi-al ERD was seen for OBS (mean = −19.83% ERD/ERS; SD = 20.95),

ollowed by VIS (mean = −13.48% ERD/ERS; SD = 20.57) and NOmean = −9.45% ERD/ERS; SD = 21.37). When single conditions wereompared directly by means of post hoc t-tests, a significant differ-nce was found only between OBS and NO (t(20) = 2.18; p = .042),ut not between OBS and VIS or VIS and NO (both p > .189).

A general post training effect of STIMULUS TYPE was foundt C4 (F(2,40) = 3.350; p = .045), with a different overall pattern.ere, there was no clear relationship between the type of train-

ng and the ERD. Instead, ERD for VIS and OBS were, at leastescriptively, comparable. When compared statistically, ERD forIS (mean = −17.50% ERD/ERS; SD = 19.29) was significantly largerompared to NO (mean = −7.29% ERD/ERS; SD = 18.80; t(20) = 2.76;

= .012). For OBS (mean = −13.95% ERD/ERS; SD = 17.76) ERD wasumerically larger than for NO, but this difference was not signif-

cant (p = .116). Likewise, VIS and OBS did not differ significantlyrom each other (p = .413). No significant effects of interest wereound in the remaining time windows (200–400 ms, 400–600 ms,00–800 ms and 800–1000 ms) at 8–10 Hz (all p > .326). Topo-raphic maps of ERD activity for the first 200 ms pre and postraining are displayed in Fig. 3, showing that OBS led to a large ERDronounced at scalp regions over the left-central, primary sensory-otor cortex.To exclude the possibility that there was a link between the

ower accuracy for NO in the post training EEG session and theRD effect for the different object types, a correlation analysisetween accuracy values and post training ERD was performed.he accuracy data were not normally distributed (defined by theolmogorov–Smirnov test for normality distribution; all p < .05).he post hoc bivariate Spearman analysis revealed no significant

orrelations between ERD at electrode C3 with post training accu-acy in the matching task (all p > .339). Additionally, there was noignificant correlation between the magnitude of ERD and timepan in which the three training sessions took place (time interval

Research 261 (2014) 328– 335

between the first and last training session), or between ERD and thetime interval between the first day of training and the post train-ing EEG assessment, or between ERD and the time span betweenthe last training session and the post training EEG assessment (allp > .105).

3.2.2. Analysis of the upper mu frequency band of 10–12 HzNo significant effects of interests were found for all time win-

dows in the frequency band of 10–12 Hz (all p > .05).

4. Discussion

The present study aimed at elucidating the impact of observa-tional learning of manipulation on the processing of pictures ofinvented, previously unfamiliar tools. Within three training ses-sions, participants observed one set of the invented tools beingmanipulated (OBS) and visually explored a second set for whichthey were only informed about the tools’ functions but not the con-crete manipulation (VIS). A third set served as an untrained control(NO) and was not part of the training. In sum, our results demon-strated that the different types of training did have a differentialmodulatory effect on sensory-motor cortex activation when seeingthe tools after training. More specifically, on scalp regions over leftsensory-motor cortex this modulation was characterized by a lineartrend of mu ERD depending on the type of object-experience (NO,VIS and OBS, respectively), whereby seeing non-trained objectsinduced the least ERD, and those for which manipulation wasobserved induced the greatest activity. Over right sensory-motorcortex, significant mu rhythm modulation by training was alsofound, with significantly larger ERD for VIS in comparison to NO.

It is now generally accepted that conceptual representationsinvolve different types of knowledge related to specific modalities[32]. For tools, associations with manipulation and function playa particularly important role. The role of experience in the emer-gence of new representations is the matter of an ongoing debate.The present study addressed this issue by systematically varyingthe type of object-related experience. Although it is not possibleto induce highly elaborated object representations in just threetraining sessions, the design with previously unfamiliar, manipula-ble objects [23,24] allowed insights into an important mechanisminvolved in the emergence of tool representations, that is, thelearning of object–action associations. The current study is, to ourknowledge, the first to provide evidence for an effect of the type ofobject-related experience on object-processing by motor-relatedbrain structures, as reflected by the mu rhythm. The finding of murhythm modulation for tools which were observed during manipu-lation is consistent with studies on the processing of familiar tools,which also showed early mu rhythm suppression [20]. For the OBSof the present study, the observation of the concrete steps of manip-ulation could have led to the formation of action representationsand motor plans that were associated with the objects. There is evi-dence that the mu rhythm reflects general motor-related activity:during action imitation and observation, mu rhythm modulationcorrelates with BOLD responses in a rather broad cortical networkconsisting of inferior parietal lobe, premotor and frontal cortex, aswell as medial frontal cortex, thalamus, temporal lobe and cerebel-lum [19].

A possible mechanism underlying the effect of manipulationobservation on the subsequent processing of tools might relateto the human mirror neuron system (hMNS). It has been shownthat certain neurons in area F5 of the monkey cortex, called

“mirror neurons”, discharge during the observation of othersperforming object-related actions [8–10]. This property has beenproposed to represent a direct matching mechanism by whichthe observed motor action is mapped onto the observer’s internal

N.N. Rüther et al. / Behavioural Brain Research 261 (2014) 328– 335 333

F ft senO = not

mDstt

ig. 2. Grand average training-specific ERD traces measured on scalp regions over leBS = observation of manipulation training objects, VIS = visual training objects, NO

otor representations, consequently activating them [11,12].

uring the observation of manipulation training in the present

tudy, mirror neurons might have been activated, finally leadingo the activation of the motor system elicited by the mere sight ofhe tools objects. Even though some studies suggest that the mu

sory-motor cortex (A) and right sensory-motor cortex (B), before and after training.trained objects.

rhythm is related to activation of the hMNS ([16] for review, see

[29]), the link of the present findings to the hMNS is speculative.A further possible explanation of the mu modulation could relateto an involvement of the canonical neuron system that was shownin the monkey cortex to respond to graspable objects per se

334 N.N. Rüther et al. / Behavioural Brain

Fig. 3. Topographic maps of grand average training-specific ERD. Displayed is theactivity within 0–200 ms before training (A) and after training (B) for OBS, VIS andNO. The rectangle marks the analyzed electrodes on scalp regions over left andrV

aidsm

lrFitsw(huar(paartrpaw

paiVwtatpafimh

special? Brain Res Cogn Brain Res 2005;22(3):457–69.

ight sensory-motor cortices. OBS = observation of manipulation training objects,IS = visual training objects, NO = not trained objects.

nd during self-executed object-directed grasping [33]. A partialnvolvement of the canonical neuron system seems likely, but noirect conclusions can be drawn from the current design. Futuretudies should investigate in more detail where the mu rhythmodulation evoked by tool stimuli originates from.Concerning the timing, the early effect within 200 ms is in

ine with the effect reported by Proverbio [20], who found muhythm desynchronization for familiar tools within 140–175 ms.urthermore, the left lateralization of the specific effect for OBSs in line with previous studies investigating conceptual represen-ations induced via active manipulation experience [23,24], withtudies investigating processing of familiar tools [4,20,34,35] andith some studies on the observation of object-directed grasping

e.g., [36]). For example, a recent ERP study showed a similar left-emispheric asymmetry for the visual processing of tools: seeingnimanual and bimanual tools activated the left premotor cortex,s was indicated by source analysis [35]. There is also evidence thatepresentations of motor action planning are largely left lateralizedfor review, see [37,38]). Additionally, the left-lateralization of theresumably generated action representations and motor plans is inccordance with the Function system of the “Two Action System”pproach [2] that is proposed to calculate and store conceptualepresentations comprising features of actions, creating associa-ive action-object links. Therefore, the effect for OBS found at scalpegions over left sensory-motor cortices in the present study sup-orts the interpretation of an activation of action representationsnd motor plans during the sight of objects, for which manipulationas learned by observation.

For VIS, information about the function of the objects wasrovided verbally during the training, possibly leading to associ-tions between the object and functional goal-related semanticnformation. The activation of the motor cortex in response toIS after training corroborates recent findings of Cross et al. [39]ho showed that both observation of knot tying and linguistic

raining can induce object representations. However, they foundctivation for linguistically induced representations about how toie knots in the parietal cortex, and not, as it is suggested by theattern found in the current study, in more anterior regions. Thectivation observed in the present study might be triggered by the

ormation of predictive models of motor plans that were generatedn response to the semantic information. Studies have shown that

u frequency can be modulated by reading sentences relating toand actions [40] and that manual action training can improve

Research 261 (2014) 328– 335

linguistic understanding of semantically related sentences [41],therefore demonstrating a close relation between the semanticand the motor system and between different types of access tothe semantic system. The predictive action plans induced by VISappear, however, to differ from the motor plans elicited by OBS, asis indicated by the different topographies of OBS and VIS trainingeffects.

The effect for OBS and VIS measured on scalp regions oversensory-motor cortices was restricted to the 8–10 Hz frequencyband. Contrary to our finding, the effect found by the Proverbio[20] study was only seen in the 10–12 Hz frequency band. It is con-ceivable that the processing of familiar and previously unfamiliartools recruits, at least to some extent, different mechanisms. A pos-sible explanation is provided by a study of Pfurtscheller et al. [14]who argue for a dissociation between the lower (8–10 Hz) and theupper mu rhythm (10–12 Hz). Both frequency bands show the typ-ical desynchronization before movement initiation but the lowermu rhythm shows no somatotopic specificity, i.e. it is desynchro-nized by either finger or foot movements, whereas the location ofthe upper frequency is different for finger versus foot movementsand also evoked by movement imagination [18]. The authors sug-gest that the somatotopically unspecific effect within the lowermu rhythm reflects an unspecific activation of the motor cor-tex independent of the actual execution of specific movements,maybe representing a presetting mechanism of motor neurons thatare part of the activated representation areas. Hence, the effectobserved by Proverbio [20] could rely on the activation of specificmotor plans that were evoked during the sight of the familiar toolstimuli, whereas the objects of the present study elicited a moreunspecific activation of sensory-motor cortex.

In sum, the present study showed that the perception of unfa-miliar tools that were previously observed to be manipulated,and were hence associated with concrete actions, led to mu sup-pression on scalp regions over left sensory-motor cortex withinthe first 200 ms after stimulus presentation. This effect suggestssensory-motor cortex activation as a result of indirect manipula-tion experience with an object, independent of object affordances,possibly related to the stored representations of object–action asso-ciations. Additionally, the perception of objects that were onlyvisually explored, and for which information about the specificfunction was provided verbally, also led to mu rhythm suppres-sion. For those objects, semantic information might have led to theformation of predictive motor plans that are related to the providedgeneral tool-associated function.

Acknowledgment

The authors thank the German Research Foundation for support-ing this work (Deutsche Forschungsgemeinschaft, DFG; SFB 874/TPB6 to C.B.).

References

[1] Gibson JJ. The theory of affordances. In: Shaw R, Bransford J, editors. Perceiving,acting and knowing: toward an ecological psychology. Erlbaum: Hillsdale, NJ;1977. p. 67–82.

[2] Buxbaum LJ, Kalenine S. Action knowledge, visuomotor activation, and embod-iment in the two action systems. Ann N Y Acad Sci 2010;1191:201–18.

[3] Lewis JW. Cortical networks related to human use of tools. Neuroscientist2006;12(3):211–31.

[4] Chao LL, Martin A. Representation of manipulable man-made objects in thedorsal stream. Neuroimage 2000;12(4):478–84.

[5] Creem-Regehr SH, Lee JN. Neural representations of graspable objects: are tools

[6] Martin A, Wiggs CL, Ungerleider LG, Haxby JV. Neural correlates of category-specific knowledge. Nature 1996;379(6566):649–52.

[7] van Schaik CP, Deaner RO, Merrill MY. The conditions for tool use in primates:implications for the evolution of material culture. J Hum Evol 1999;36:719–41.

Brain

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

N.N. Rüther et al. / Behavioural

[8] Dipellegrino G, Fadiga L, Fogassi L, Gallese V, Rizzolatti G. Understanding motorevents – a neurophysiological study. Exp Brain Res 1992;91(1):176–80.

[9] Gallese V, Fadiga L, Fogassi L, Rizzolatti G. Action recognition in the premotorcortex. Brain 1996;119:593–609.

10] Rizzolatti G, Fadiga L, Gallese V, Fogassi L. Premotor cortex and the recognitionof motor actions. Cogn Brain Res 1996;3(2):131–41.

11] Iacoboni M, Woods RP, Brass M, Bekkering H, Mazziotta JC, Rizzolatti G. Corticalmechanisms of human imitation. Science 1999;286(5449):2526–8.

12] Rizzolatti G, Fadiga L, Fogassi L, Gallese V. Resonance behaviors and mirrorneurons. Arch Ital Biol 1999;137(2–3):85–100.

13] Chatrian GE, Petersen MC, Lazarte JA. The blocking of the rolandic wicketrhythm and some central changes related to movement. ElectroencephalogrClin Neurophysiol 1958;10(4):771–2.

14] Pfurtscheller G, Neuper C, Krausz G. Functional dissociation of lower and upperfrequency mu rhythms in relation to voluntary limb movement. Clin Neuro-physiol 2000;111(10):1873–9.

15] Babiloni C, Carducci F, Cincotti F, Rossini PM, Neuper C, Pfurtscheller G,et al. Human movement-related potentials vs desynchronization of EEG alpharhythm: a high-resolution EEG study. Neuroimage 1999;10(6):658–65.

16] Muthukumaraswamy SD, Johnson BW, McNair NA. Mu rhythm modulationduring observation of an object-directed grasp. Brain Res Cogn Brain Res2004;19(2):195–201.

17] Perry A, Bentin S. Mirror activity in the human brain while observing handmovements: a comparison between EEG desynchronization in the mu-rangeand previous fMRI results. Brain Res 2009;1282:126–32.

18] Neuper C, Scherer R, Wriessnegger S, Pfurtscheller G. Motor imagery and actionobservation: modulation of sensorimotor brain rhythms during mental controlof a brain–computer interface. Clin Neurophysiol 2009;120(2):239–47.

19] Braadbaart L, Williams JHG, Waiter GD. Do mirror neuron areas mediate murhythm suppression during imitation and action observation? Int J Psychophys-iol 2013;89(1):99–105.

20] Proverbio AM. Tool perception suppresses 10–12 Hz mu rhythm of EEG overthe somatosensory area. Biol Psychol 2012;91(1):1–7.

21] Muthukumaraswamy SD, Johnson BW. Primary motor cortex activation duringaction observation revealed by wavelet analysis of the EEG. Clin Neurophysiol2004;115(8):1760–6.

22] Gallese V, Goldman A. Mirror neurons and the simulation theory of mind-reading. Trends Cogn Sci 1998;2(12):493–501.

23] Bellebaum C, Tettamanti M, Marchetta E, Della RP, Rizzo G, Daum I, et al.Neural representations of unfamiliar objects are modulated by sensorimotorexperience. Cortex 2013;49(4):1110–25.

24] Weisberg J, Turennout M, Martin A. A neural system for learning about objectfunction. Cereb Cortex 2007;17(3):513–21.

25] Kiefer M, Sim EJ, Liebich S, Hauk O, Tanaka J. Experience-dependent plasticityof conceptual representations in human sensory-motor areas. J Cogn Neurosci2007;19(3):525–42.

[

Research 261 (2014) 328– 335 335

26] Creem-Regehr SH, Dilda V, Vicchrilli AE, Federer F, Lee JN. The influence ofcomplex action knowledge on representations of novel graspable objects: evi-dence from functional magnetic resonance imaging. J Int Neuropsychol Soc2007;13(6):1009–20.

27] Kumar S, Riddoch MJ, Humphreys G. Mu rhythm desynchronization revealsmotoric influences of hand action on object recognition. Front Hum Neurosci2013;7:66.

28] Lee TW, Girolami M, Sejnowski TJ. Independent component analysis usingan extended Infomax algorithm for mixed subGaussian and superGaussiansources. Neural Comput 1999;11(2):417–41.

29] Pineda JA. The functional significance of mu rhythms: translating seeing andhearing into doing. Brain Res Brain Res Rev 2005;50(1):57–68.

30] Pfurtscheller G, Aranibar A. Evaluation of event-related desynchronization(ERD) preceding and following voluntary self-paced movement. Electroen-cephalogr Clin Neurophysiol 1979;46(2):138–46.

31] Greenhouse SW, Geisser S. On methods in the analysis of profile data. Psy-chometrika 1959;24(2):95–112.

32] Kiefer M, Pulvermuller F. Conceptual representations in mind and brain:theoretical developments, current evidence and future directions. Cortex2012;48(7):805–25.

33] Murata A, Fadiga L, Fogassi L, Gallese V, Raos V, Rizzolatti G. Object represen-tation in the ventral premotor cortex (area F5) of the monkey. J Neurophysiol1997;78(4):2226–30.

34] Proverbio AM, Adorni R, D’Aniello GE. 250 ms to code for action affordanceduring observation of manipulable objects. Neuropsychologia 2011;49(9):2711–7.

35] Proverbio AM, Azzari R, Adorni R. Is there a left hemispheric asymmetry fortool affordance processing? Neuropsychologia 2013;51(13):2690–701.

36] Grafton ST, Arbib MA, Fadiga L, Rizzolatti G. Localization of grasp representa-tions in humans by positron emission tomography. 2. Observation comparedwith imagination. Exp Brain Res 1996;112(1):103–11.

37] Haaland KY, Harrington DL. Hemispheric asymmetry of movement. Curr OpinNeurobiol 1996;6(6):796–800.

38] Majdandzic J, Grol MJ, van Schie HT, Verhagen L, Toni I, Bekkering H. The roleof immediate and final goals in action planning: an fMRI study. Neuroimage2007;37(2):589–98.

39] Cross ES, Cohen NR, Hamilton AFD, Ramsey R, Wolford G, Grafton ST. Physicalexperience leads to enhanced object perception in parietal cortex: insights fromknot tying. Neuropsychologia 2012;50(14):3207–17.

40] Alemanno F, Houdayer E, Cursi M, Velikova S, Tettamanti M, Comi G, et al.Action-related semantic content and negation polarity modulate motor areas

during sentence reading: an event-related desynchronization study. Brain Res2012;1484:39–49.

41] Locatelli M, Gatti R, Tettamanti M. Training of manual actions improves lan-guage understanding of semantically related action sentences. Front Psychol2012;3:547.