Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Probing the reaching–grasping network in humans through multivoxel pattern decoding

Probing the reaching–grasping network in humans through multivoxel pattern decoding Functional magnetic resonance imaging, Introduction: The quest for a putative human homolog of the reaching–grasp- multivoxel pattern decoding, reaching-only ing network identified in monkeys has been the focus of many neuropsycholog- action, visuomotor reach-to-grasp action ical and neuroimaging studies in recent years. These studies have shown that Correspondence the network underlying reaching-only and reach-to-grasp movements includes Maria Grazia Di Bono, Dipartimento di the superior parieto-occipital cortex (SPOC), the anterior part of the human Psicologia Generale, University of Padova, via intraparietal sulcus (hAIP), the ventral and the dorsal portion of the premotor Venezia 8, 35131 Padova, Italy. Tel: +39 049 cortex, and the primary motor cortex (M1). Recent evidence for a wider fron- 8276642; Fax: +39 049 8276600; toparietal network coding for different aspects of reaching-only and reach-to- E-mail: [email protected] grasp actions calls for a more fine-grained assessment of the reaching–grasping and Marco Zorzi, Dipartimento di Psicologia network in humans by exploiting pattern decoding methods (multivoxel pattern Generale, University of Padova, via Venezia analysis—MVPA). Methods: Here, we used MPVA on functional magnetic res- 8, 35131 Padova, Italy. Tel: +39 049 onance imaging (fMRI) data to assess whether regions of the frontoparietal net- 8276618; Fax: +39 049 8276600; work discriminate between reaching-only and reach-to-grasp actions, natural E-mail: [email protected] and constrained grasping, different grasp types, and object sizes. Participants were required to perform either reaching-only movements or two reach-to- Funding Information grasp types (precision or whole hand grasp) upon spherical objects of different This work was supported by a grant from the European Research Council (grant no. sizes. Results: Multivoxel pattern analysis highlighted that, independently from 210922) and the University of Padova the object size, all the selected regions of both hemispheres contribute in coding (Strategic Grant NEURAT) to M. Zorzi. for grasp type, with the exception of SPOC and the right hAIP. Consistent with recent neurophysiological findings on monkeys, there was no evidence for a Received: 4 May 2015; Revised: 27 July clear-cut distinction between a dorsomedial and a dorsolateral pathway that 2015; Accepted: 13 September 2015 would be specialized for reaching-only and reach-to-grasp actions, respectively. Nevertheless, the comparison of decoding accuracy across brain areas Brain and Behavior, 2015; 5(11), e00412, highlighted their different contributions to reaching-only and grasping actions. doi: 10.1002/brb3.412 Conclusions: Altogether, our findings enrich the current knowledge regarding the functional role of key brain areas involved in the cortical control of reach- ing-only and reach-to-grasp actions in humans, by revealing novel fine-grained distinctions among action types within a wide frontoparietal network. activity of single neurons is recorded with techniques Introduction allowing a high level of spatial and temporal resolution. In the domain of motor control great attention has been These studies have identified the main cortical structures given to reaching-only and reach-to-grasp actions, appar- involved in the control of visually guided reach-to-grasp ently simple and straightforward behaviors which are part movements. They are the primary motor cortex (F1), the of our everyday life motor repertoire, and fundamental premotor cortex (area F5), and the anterior part of the for our interaction with the environment. intraparietal sulcus (AIP; Murata et al. 1997, 2000). The A great extent of our knowledge regarding the cortical ability to perform a successful reach-to-grasp action control of reach-to-grasp movements is rooted in neuro- depends primarily on the integrity of F1; indeed, lesions physiological studies on behaving monkeys, in which the of this area in macaques produce a remarkable deficit in ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (1 of 18) This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Multivoxel Pattern Decoding M. G. Di Bono et al. the control of individual fingers, bringing to a loss of monkeys (Cavina-Pratesi et al. 2007; Culham et al. 2006; coordination abilities (Lawrence and Hopkins 1976). Area Kroliczak et al. 2007; Tunik et al. 2007; for reviews see F5, which forms the rostral part of the macaque ventral Castiello 2005; Castiello and Begliomini 2008; Filimon premotor cortex (PMv) and AIP, a small zone lying 2010). Overall, reach-to-grasp fMRI studies converge in within the rostral part of the posterior bank of the intra- considering the anterior part of the human intraparietal parietal sulcus (Matelli et al. 1985; Luppino et al. 1999; sulcus (hAIP), a likely homolog of monkey AIP (Grafton Matelli and Luppino 2001) are directly connected and are et al. 1996; Culham et al. 2003; Frey et al. 2005; Beglio- involved in converting intrinsic object properties (e.g., mini et al. 2007a; Hinkley et al. 2009). The key role of shape, size) into a proper hand conformation for grasping hAIP in the dynamic control of reach-to-grasp move- the object (Jeannerod et al., 1995). ments has also been confirmed in a series of TMS studies In macaques trained to grasp various objects, activity (Glover et al. 2005; Tunik et al. 2005; Rice et al. 2006). of F5 and AIP neurons show not only strong similarities, Tunik et al. (2005) have shown that applying TMS to the but also important differences (Rizzolatti et al. 1988, hAIP induces a delay in grasp adaptation, suggesting that 2002; Taira et al. 1990; Rizzolatti and Arbib 1998). On this area performs a sort of iterative comparison between one side, both F5 and AIP neurons code for reach-to- the incoming sensory information and the motor grasp actions (Murata et al. 1997, 2000). However, AIP command during the ongoing movement. neurons seem to represent the entire action, whereas F5 The quest for the human homolog of macaque F5 has neurons seem to be concerned with a particular segment identified the ventral part of the premotor cortex (PMv) of it (Rizzolatti et al. 1998; Murata et al. 2000). Another as a plausible candidate. However, neuroimaging studies important difference is that visual responses to three-di- investigating brain activity during a reach-to-grasp move- mensional objects are found more frequently in AIP than ment do not provide a coherent picture regarding the in F5 (Murata et al. 2000). This suggests that AIP, involvement of the PMv. Some fMRI studies have although part of a parieto-frontal network dedicated to reported PMv activation during multidigit visually guided hand movements, also contains a population of neurons reach-to-grasp actions (Grol et al. 2007; Cavina-Pratesi that code three-dimensional objects in visual terms. et al. 2010), object manipulation (Binkofski et al. 1999), Building upon this knowledge, Fagg and Arbib (1998) and isometric grasping (Ehrsson et al. 2001), whereas suggest that AIP could store the objects’ sensory proper- other studies found no evidence of PMv involvement dur- ties (Taira et al. 1990; Murata et al. 1997, 2000). These ing visually guided reach-to-grasp action (Culham et al. representations influence the ventral premotor area F5 2006; Begliomini et al. 2007a,b). A possible explanation and also the dorsal premotor area F2, which is involved for this controversial finding, which contrasts with the in visual guidance of the hand (Moll and Kuypers 1977; clear involvement of PMv for reach-to-grasp movements Godschalk et al. 1981; Weinrich and Wise 1982; Passing- in macaques (e.g., Rizzolatti et al. 1988), could be due to ham 1987; Rizzolatti et al. 1988; Raos et al. 2004, 2006). the fact that interspecies differences in the organization of Area F5 plays a primary role in selecting the most appro- the PMv, as well as the development of a motor speech priate type of grip on the basis of the object affordances area in humans, may have changed the location of the provided by AIP, thereby activating a motor representa- human functional homolog of monkey area F5 (Amunts tion of that object. This motor representation is then sup- and Zilles 2001). Moreover, it is worth noting that in the plied to F2, which keeps memory of it and combines it majority of studies, grasping-related activity has been iso- with visual information provided by cortical areas of the lated by subtracting activations obtained during the superior parietal lobe to continuously update the configu- reaching-only from the reach-to-grasp task (Grafton et al. ration and orientation of the hand as it approaches the 1996; Culham et al. 2003; Frey et al. 2005; Begliomini object. The final output is then sent to the F1 for motor et al. 2007a,b). Because in these studies both the reach- execution (for review see Castiello and Begliomini 2008). ing-only and the reach-to-grasp tasks required specific Moreover, the same role of F2 is played by area V6A, motor goals—triggering premotor activity—it might well which is strongly and reciprocally connected with the be that activations within premotor areas could have can- dorsal premotor cortex controlling arm movements, and celed one another when compared (Grafton et al. 1996; elaborates visual information, motion and space, for con- Culham et al. 2003; Frey et al. 2005; Begliomini et al. trolling both reaching-only and reach-to-grasp move- 2007a,b). ments (Galletti et al. 2003; Fattori et al. 2009, 2010). The dorsal part of the premotor cortex (PMd) has been In humans, both functional magnetic resonance imag- suggested as the human correspondent of macaque area ing (fMRI) and transcranial magnetic stimulation (TMS) F2 (Matelli et al. 1991). As demonstrated in macaques studies have demonstrated the existence of localized corti- (Raos et al. 2004), in humans the contribution of PMd to cal reach-to-grasp areas similar to those described in reach-to-grasp action is that of an online monitoring dur- Brain and Behavior, doi: 10.1002/brb3.412 (2 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding ing the execution phase of the action. A study comparing wide frontoparietal network adapted to both object size reach-to-grasp movements with different levels of com- and location. Furthermore, in an electroencephalogram plexity, underlined bilateral PMd involvement in associa- (EEG)/event-related potentials (ERP) study, Tarantino tion with conditions that required higher levels of et al. (2014) showed that the kinematics of reaching-only, accuracy in implementing the action (Begliomini et al. as well as the amplitude and the latency of P300 and 2007b). N400 ERP components in parietal and prefrontal sites, Although the studies reviewed above significantly con- respectively, were modulated by object size, consistent tributed to sketch an overall picture of the neural sub- with physiological findings on nonhuman primates (Fat- strates of reaching-only and reach-to-grasp in humans, a tori et al. 2012). The possibility to shed further light on crucial issue that requires further investigation is how the these issues is offered by a multivariate approach that different areas specifically contribute to the coding of exploits multivoxel pattern analysis (MVPA; e.g., Di Bono grasp type (e.g., precision grasping [PG], whole hand and Zorzi 2008; O’Toole et al. 2007; Pereira et al. 2009; grasping [WHG]) with respect to object size. This knowl- Zorzi et al. 2011). A study by Gallivan et al. (2011) edge is fundamental in order to fully define the paral- showed distinct activity patterns coding different preci- lelism between the monkeys and the human grasping sion grasping actions toward two differently sized objects network. Indeed, Rizzolatti et al. (1988; see also Rizzolatti positioned at two different spatial locations (i.e., the and Luppino 2001) showed that in monkeys, neurons smaller cube on the top of the larger one). The authors within AIP and F5 areas code for grasping actions in rela- claimed that it was possible to decode two different types tion to the type of object to be grasped. More in detail, of grasping, but it was unclear whether this result could F5 neurons seem to be mainly involved in selecting the be related to the object size or to a different direction in most appropriate motor act from a “motor vocabulary.” reaching-only toward the bottom or top object. Gallivan For instance, the act of grasping a raisin (which requires et al. (2011), also showed that voxel pattern activity the opposition of the index finger with the thumb) is within multiple frontoparietal areas during movement encoded by neurons different from those that encode the planning allowed discrimination between reach-to-grasp grasping of an apple (which requires the opposition of and reaching-only actions. More evidence against a clear the thumb with all fingers). distinction between a dorsomedial (e.g., superior parieto- In humans, fMRI studies that directly contrasted PG occipital cortex [SPOC], medial intraparietal area MIP, versus WHG using conventional analysis, revealed activa- and PMd) and a dorsolateral (e.g., hAIP and PMv) path- tion differences between the two grasping actions in con- way, specialized for reaching-only and grasping, respec- tralateral M1 (WGH > PG), bilateral PMv and hAIP tively, was provided by Fabbri et al. (2014). These recent (PG > WHG) (Ehrsson et al. 2000, 2001; Begliomini findings in humans are consistent with the theory of a et al. 2007a). More recent studies have confirmed these dorsomedial visual stream (e.g., V6A) involved in reach- findings, suggesting that grasp types (PG vs. WHG) have to-grasp actions, suggested by Galletti et al. (2003). distinct representations within a wide frontal–parietal net- Indeed, this has been documented by Fattori et al. (2009) work subserving reach-to-grasp movements (Begliomini and more directly by Fattori et al. (2010), who showed et al. 2014). This issue, however, remains controversial evidence of grasping neurons in the medial parieto-occip- given that other studies failed to detect such differences ital cortex of the macaque monkeys. The abovementioned (e.g., Kuhtz-Buschbeck et al., 2008). results about macaque area V6A suggested SPOC area as Another interesting question that requires further its putative homolog in humans (Pitzalis et al. 2013, investigation is the role of object size in both reaching- 2015; Tosoni et al. 2014). The human homolog of V6A only and reach-to-grasp actions. The visuomotor channel has been also identified as the parieto-occipital junction hypothesis of Jeannerod (1981) states that the grasping by Prado et al. (2005) and as the superior end of the action is composed of grip and transport components, parieto-occipital sulcus (sPOS) by Filimon et al. (2009). which rely on intrinsic (e.g., object size) or extrinsic (e.g., The recent findings on different aspects of reaching- location) object properties. According to this view, object only and reach-to-grasp actions call for a thorough and size and location have to be processed independently in fine-grained assessment of the reaching–grasping network separate visual channels. However, the recent neuroimag- in humans. We exploited pattern decoding methods for ing findings of Monaco et al. (2015) have suggested that, investigating the following key questions: (1) whether in humans, the cortical processing of object size and loca- there are distinct representations for different grasp types tion does not conform to a strict segregation between grip (i.e., PG vs. WHG); (2) whether there are distinct repre- and transport components of the reach-to-grasp action. sentations of object size during reaching-only action; (3) In an fMRI adaptation paradigm, the authors found that whether object size could modulate each grasp type action left aIPS showed adaptation only to object size, whereas a in a congruent/incongruent action setting (e.g., PG ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (3 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. toward a small object and WHG toward a large object an angle of ~30° and they were supported by a foam [congruent] vs. PG toward a large object and WHG wedge permitting direct viewing of the stimulus without toward a small object [incongruent]). Moreover, we mirrors. The apparatus was placed at a natural reaching aimed at: (4) replicating the findings of Gallivan et al. distance (~15 cm) above the participant’s pelvis for (2011) and Fabbri et al. (2014), which provided evidence avoiding further movements of the upper part of the of distinct representations for reaching-only and reach-to- trunk. grasp actions, distributed across a wide frontoparietal net- work; (5) replicating the findings of Monaco et al. (2015) Stimuli and task procedures on the representation of the object size during reach-to- grasp actions. The stimuli consisted of two spherical plastic objects of To address these issues, we reanalyzed the fMRI data of different dimensions (small stimulus: 3 cm diameter; large Begliomini et al. (2007b) using MVPA for investigating stimulus: 6 cm diameter). Participants were requested to the specific contribution of each brain area belonging to perform three different actions toward either the small or the reaching–grasping network in humans. To this end, the large stimulus: (1) grasping the stimulus with a PG; we selected anatomically defined regions of interest (2) grasping the stimulus with a WHG; (3) only reach the (ROIs) within a wide frontoparietal network involved in stimulus (R), by touching it with the hand knuckles, reaching-only and reach-to-grasp action representation maintaining the hand closed like in a fist. Participants (e.g., Gallivan et al. 2011; Fabbri et al. 2014). We then were informed about the type of movement to perform trained a support vector machine (SVM) classifier (see through a sound delivered by pneumatic MR-compatible Pereira et al. 2009, for a tutorial overview) with linear headphones: (1) PG—low tone (duration: 200 msec; fre- kernel on the voxel pattern activity of those ROIs for quency: 1.7 kHz); (2) WHG—high tone (duration: decoding (1) object size in both reach-to-grasp and reach- 200 msec; frequency: 210 Hz); R-double tone (duration: ing-only actions, (2) grasp type, (3) the congruence 70 msec each, staggered by a 60 msec silence period; fre- between grasp type and object size, and (4) the action quency: 445 Hz) and they were instructed to start their type (i.e., reach-to-grasp vs. reaching-only actions). action toward the stimulus only when the sound was delivered. Materials and Methods Experimental design Participants The experiment was conducted by using an event-related Nineteen right-handed participants (12 female; 19– design. inter stimulus interval (ISI) varied from 3 to 8 sec 30 years old) participated in the experiment. All gave writ- with a “long exponential” probability distribution (Hag- ten informed consent before entering in the scanner room. berg et al. 2001). ISIs distribution was fully randomized According to Begliomini et al. (2007b), three participants across trials in each run for each subject. Action toward were not included in the analysis due to the presence of the stimulus (PG, WHG, R) and stimulus dimension head motion. The cut-off used for motion correction tol- (small or large) were manipulated as to create six differ- erance was the size of the voxel (3.3 9 3.3 9 3 mm). In ent conditions (see Fig. 1): (1) “PG toward the small other words, if motion exceeded these measures in transla- object” (PGS); (2) “PG toward a large object” (PGL); (3) tion and/or rotation, the participant was not included in “WHG toward a large object” (WHGL); (4) “WHG the analysis. All participants were right-handed as mea- toward a small object” (WHGS); (5) “reaching-only sured by the Edinburgh Handedness Inventory (Oldfield toward a small object” (RS); (6) “reaching-only toward a 1971). The experimental procedures were approved by the large object” (RL). There were 45 trials for each experi- ethics committee of the University of Padua (see Beglio- mental condition, grouped into mini-blocks of five trials mini et al. 2007b, for all details). belonging to the same condition. Trials were divided in four runs, with a short rest between each run. In the odd runs the object was small, whereas in the even runs the Apparatus object was large. Participants were requested to perform either reaching- only or reach-to-grasp actions toward stimuli presented Imaging parameters by using a metal-free apparatus, which was composed of a table mounted on a plexiglass structure that allowed the Images were acquired with a whole-body 3T scanner (Sie- presentation of real 3D stimuli to participants lying mens Magnetom Trio, TIM system, Siemens, Erlangen, supine in the scanner. Participants had their head tilted at Germany) equipped with a standard Siemens 12 channels Brain and Behavior, doi: 10.1002/brb3.412 (4 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Figure 1. Experimental conditions (adapted from Begliomini et al. 2007b). Participants viewed one of the two stimuli (i.e., a spherical object of two different sizes) and performed three different tasks (i.e., reaching-only and two types of reach-to-grasp actions). The experimental conditions involved either precision grasp (PG), whole hand grasp (WHG), or Reaching-only (R) actions. Participants were instructed about the movement to perform (PG, WHG, and R) with a sound delivered through headphones. According to the size of the object to be grasped, the reach-to-grasp action was defined as congruent (PG toward a small object—PGS; WHG toward a large object—WHGL) or incongruent (PG toward a large object —PGL; WHG toward a small object—WHGS). All actions had to be performed with the right hand. coil. Functional images were acquired with a gradient- rological Institute (http://www.mni.mcgill.ca/) and echo, echo-planar (EPI) T2*-weighted sequence in order distributed with the software SPM. To avoid any circular- to measure blood oxygenation level-dependent (BOLD) ity issue in ROI selection (Kriegeskorte et al. 2009), we contrast throughout the whole brain (47 contiguous axial did not rely on the functional data but selected six ROIs slices acquired with descending interleaved sequence, that were defined on purely anatomical grounds (using 64 9 64 voxels, 3.3 9 3.3 9 3 mm resolution, the SPM Anatomy toolbox; http://www.fil.ion.ucl.ac.uk/ FOV = 210 9 210 mm, flip angle = 90°,TE = 30 msec). spm/ext/#Anatomy). One additional ROI, selected on the Volumes were acquired continuously with a repetition basis of the results of Fabbri et al. (2012), was obtained time (TR) of 3 sec; 117 volumes were collected in each through a spherical image mask using the SPM Sim- single scanning run (5:51 min; four scanning runs in pleROIBuilder toolbox (http://www.fil.ion.ucl.ac.uk/spm/ total). High-resolution T1-weighted images were acquired ext/#SimpleROIBuilder). for each subject (3D MP-RAGE, 176 axial slices, data The seven ROIs were defined as follows: matrix 256 9 256, 1 mm isotropic voxels,  ROI-1: bilateral superior parieto-occipital cortex TR = 1859 msec, TE = 3.14 msec, flip angle = 22°). (SPOC) defined according to the functional study by Fabbri et al. (2012). We extracted a sphere of 8-mm radius, centered on the Talairach coordinates (SPOC Regions of interest LH: 17, 72, 37; SPOC RH: 21, 73, 31). The functional images were preprocessed using the soft-  ROI-2: bilateral superior parietal lobe (SPLap), defined ware package SPM (Wellcome Department of Imaging according to the anatomical study by Scheperjans et al. Neuroscience, University College of London, http:// (2008). We used two different subregions of SPL (la- www.fil.ion.ucl.ac.uk/spm/). For each participant, images beled as SPL 7A and SPL 7P in the Anatomy toolbox) underwent motion correction and unwarping, and each to create this anatomical mask. volume was realigned to the first volume in the series.  ROI-3: bilateral hAIP, defined according to the The mean of all functional images was then co-registered anatomical study by Choi et al. (2006) on the human IPS. We used three different subregions of the anterior to the anatomical scan, previously corrected for intensity inhomogeneity. EPI images were then normalized adopt- IPS (labeled as hIP1, hIP2, and hIP3 in the Anatomy ing the MNI152 template, supplied by the Montreal Neu- toolbox) to create this anatomical mask. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (5 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. ROI-4: bilateral Brodmann area (BA) 1/2/3ab, accord- PG vs. WHG) as input to the classifier. In order to main- ing to the anatomical studies of Geyer et al. (1999, tain sample independence for SVM training and testing, 2000) and Grefkes et al. (2001). for each mini-block (i.e., five trials from the same condi- ROI-5: bilateral primary motor cortex, defined accord- tion), we discarded the first four volumes to capture a ing to Geyer et al. (1996), but selecting only the poste- stable fMRI signal without incorporating any noise from rior part of the primary motor cortex (bilateral BA 4p) trials within the previous mini-block and then created one sample averaging the remaining volume images (e.g., to focus on the hand representation. Pereira et al. 2009). Consequently, the target condition, ROI-6: bilateral premotor area BA 6, defined according relative to each contrast, was coded in a way to have a to the anatomical study of Geyer (2003), roughly vector T {+1, 1} , where i refers to the sample corresponding to the PMd. Specifically, BA 6 is a i i =1,...,N and N is the number of samples relative to both condi- rather large area that includes not only PMd laterally, tions in the classification (e.g., N = 36 in PG vs. WHG but also supplementary motor area (SMA) and classification), in which all the samples corresponding to pre-SMA medially. ROI-7: bilateral BA 44/45, according to Amunts et al. one target condition (e.g., PG) were labeled with +1, (1999), roughly corresponding to the PMv. whereas all the other samples (e.g., WHG) with 1. Cross-validation was used to estimate the test generaliza- To test classifier performance outside our selected net- tion performance. The SVM classifier was trained on the work, we defined one additional control ROI in which no data set using a modified version of leave-one-out cross- BOLD signal was expected and then no consistent classifi- validation. At each step of the cross-validation loop, two cation performance should be possible (see Gallivan et al. samples (one for each condition) were excluded from the 2011, for a similar methodological procedure). Therefore, training set and used to test generalization performance we selected a (8 mm) cubic region outside the skull of (see Zorzi et al. 2011). Classifier accuracy, computed the brain (centroid MNI coordinates: [63, 63, 75]). across the entire cross-validation loop on the test set, was used as statistical measures of binary classification. Preprocessing Statistical analysis on the classifier After ROI extraction, the voxel time series were prepro- performance cessed through a series of commonly used steps: stan- dardization, detrending, and temporal filtering. For each Previous studies (e.g., Chen et al. 2011; Gallivan et al. participant, each of the four runs was processed sepa- 2011) showed that t-test group analysis, with respect to rately. The time series were first standardized in order to nonparametric randomization tests, is a rather conserva- have zero mean and standard deviation 1. Then, linear tive estimate of significant decoding accuracy. Therefore, trends in each time series were removed, and a high-pass we conducted a set of one-tailed t-tests, one for each filter (0.01 Hz) was applied in order to remove low ROI, on the classifier accuracy (against the chance level of frequency drift in the signal. 50%) to obtain group statistics regarding the discrimina- tion between the two conditions included in each classifi- cation. We used false discovery rate (FDR) for correcting Classifier analysis for multiple comparisons. Furthermore, for each classifi- We used SVM with linear kernel (the C parameter was cation we assessed the possible differences between ROIs fixed to 1, which is the default value) as multivoxel pat- and hemispheric asymmetries by performing an ANOVA tern classifier. We performed six classifications: (1) Object on the classifier accuracy using ROI (SPOC, SPLap, hAIP, size in Reach-to-grasp (i.e., PGS + WHGS vs. BA 1/2/3ab, BA 44/45, BA 6, BA 4p) and hemisphere (left PGL + WHGL); (2) Object size in Reaching-only (i.e., RS vs. right) as factors. Finally, to assess the sensitivity of vs. RL); (3) Grasp type (i.e., PGS + PGL vs. each ROI for each classification, we performed a repeated WHGS + WHGL); (4) Congruence between grasp type measure (RM) ANOVA on the classifier accuracy, using and object size (i.e., PGS + WHGL [Congruent] vs. classification as a within-subject factor. PGL + WHGS [Incongruent]); (5) PG vs. Reaching-only (i.e., PGS + PGL vs. RS + RL); (6) WHG vs. Reaching- Results only (i.e., WHGS + WHGL vs. RS + RL). For each partic- ipant, we trained a linear classifier on the voxels within In this section we report, for each classification (i.e., Object each selected ROI, separately for each hemisphere. We size in reach-to-grasp action, Object size in reaching-only used only the fMRI volumes corresponding to the experi- action, Grasp Type, Congruence, PG vs. Reaching-only, mental conditions for each classification (e.g., grasp type: WHG vs. Reaching-only) the results obtained by training Brain and Behavior, doi: 10.1002/brb3.412 (6 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding linear SVM classifiers on each selected ROI, separately for bilateral SPOC, right hAIP, and the control ROI (see the left and the right hemisphere. For each ROI, the results Table 1; Fig. 2, panel B), revealing an hemispheric asym- are expressed in terms of classification performance on the metry for the hAIP. test set. To investigate possible interhemispheric asymmetries for the ROIs, we performed RM-ANOVA on the classifier accuracy using ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA Object size in reach-to-grasp action 44/45, BA 4p, and BA 6) and hemisphere (left vs. right) Independently from the grasp type, it was not possible to as within-subject factors. The analysis revealed a main discriminate between grasping a small and large object effect of ROI (F(6, 90) = 4.03, P = 0.001, g = 0.21) and from all the selected ROIs in both hemispheres, Control hemisphere (F(1, 15) = 8.12, P = 0.012, g = 0.35). The ROI included (mean accuracy = 0.47  0.02 SEM, all two-way interaction was not significant (F = 1.49). ts < 0.59). Decoding accuracy was higher when decoding from the left (M = 0.59  0.02 SEM) than from the right (M = 0.57  0.02 SEM) hemisphere. Paired t-tests (FDR Object size in reaching-only action corrected, corrected a = 0.007) showed higher decoding It was not possible to discriminate between reaching-only accuracy in the somatosensory cortex (BA 1/2/3 ab) a small and large object from all the left and right selected (M = 0.63  0.02 SEM) with respect to SPOC (M = ROIs, Control ROI included (mean accu- 0.54  0.02 SEM, t(15) = 3.73, P = 0.002), hAIP (M = racy = 0.53  0.03 SEM, all ts < 2.3). 0.54  0.02 SEM, t(15) = 4.78, P < 0.001), and the selected motor areas (BA 4p) (M = 0.56  0.02 SEM, t (15) = 4.2, P = 0.001) (see Fig. 3, panel A). No further Grasp type significant results were observed. Results for grasp type classification are summarized in Table 1. Congruence The classifier analyses showed that it was possible to linearly decode the type of grasp from the voxel pattern From none of the left and right selected ROIs (Control activity of all the selected ROIs with the exception of ROI included), it was possible to discriminate between congruent and incongruent conditions (mean accu- racy = 0.48  0.02 SEM, all ts < 0.78). Table 1. Grasp type classification. Results obtained by training linear SVM classifiers on each selected ROI, separately for the left and the Precision grasping versus reaching right hemisphere. For each ROI, the results are expressed in terms of classification performance on the test set (M  1 SEM) and the t Results for reach-to-grasp using PG versus reaching-only statistics for assessing classification significance. classification are summarized in Table 2. The classifier analyses showed that, independently from ROI Left hemisphere Right hemisphere the object size, it was possible to linearly discriminate SPOC .52  .03 .55  .03 between PG and Reaching-only from the voxel pattern t(15) = 0.75, ns t(15) = 1.49, ns activity of all the selected ROIs with the exception of the SPLap .61  .02 .54  .03 control ROI (see Table 2; Fig. 2, panel C). t(15) = 4.55, P < .001 t(15) = 3.55, P < .01 hAIP .57  .03 .51  .02 To investigate possible interhemispheric asymmetries, t(15) = 2.32, P = .017 t(15) = .48, ns we performed an RM-ANOVA on the classifier accuracy BA 1/2/3ab .67  .02 .59  .02 using ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA 44/45, t(15) = 7.31, P < .0001 t(15) = 3.92, P < .0001 BA 4p, and BA 6) and hemisphere (left vs. right) as BA 4p .56  .02 .56  .02 within-subject factors. The analysis revealed a main effect t = 2.4, P = .015 t(15) = 2.54, P = .015 of ROI (F(6, 90) = 9.38, P = 0.001, g = 0.39) and hemi- BA 6 .6  .2 .56  .03 sphere (F(1, 15) = 7.71, P = 0.014, g = 0.34). The two- t(15) = 2.95, P < .001 t(15) = 2.14, P = .025 p BA 44/45 .58  .03 .57  .03 way interaction was not significant (F = 1.31). Indepen- t(15) = 2.99, P < .005 t(15) = 2.42, P= .015 dently from the selected ROI, classifier accuracy was higher Control ROI .5  .02, when decoding from the left (M = 0.67  0.01) than from t = .17, ns the right (M = 0.63  0.01) hemisphere. Paired t-tests (FDR corrected, corrected a = 0.031) showed that decod- SVM, support vector machine; ROI, regions of interest; SPOC, superior ing accuracy from BA 1/2/3ap (M = 0.73  0.02 SEM) parieto-occipital cortex; SPLap, superior parietal lobe; BA, Brodmann area; hAIP, anterior part of the human intraparietal sulcus. and BA 6 (M = 0.72  0.02 SEM) was higher with ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (7 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. (A) (B) (C) (D) Figure 2. (A) Regions of interest (ROIs) used in the multivariate classifier analyses, transparently superimposed on top, lateral and mesial view of a standard template using BrainNet Viewer (http://www.nitrc.org/projects/bnv/) (Xia et al. 2013). ROI-1 (yellow) includes SPOC areas (Fabbri et al. 2012). ROI-2 (violet) includes SPLap areas (Scheperjans et al. 2008). ROI-3 (red) includes three subregions in the hAIP (Choi et al. 2006). ROI-4 (pink) includes BA 1/2/3ab (Geyer et al. 1999, 2000; Grefkes et al. 2001). ROI-5 (blue) includes the posterior part of the BA 4 (Geyer et al. 1996). ROI-6 (green) includes BA 6 (Geyer 2003). ROI-7 (orange) includes BA 44/45 (Amunts et al. 1999). (B) Mean linear SVM classification accuracy for grasp type decoding as a function of the involved ROIs in the left (L) and right (R) hemisphere. (C) Mean linear SVM classification performance for discriminating (independently from the object size) between PG and Reaching-only conditions as a function of the involved ROIs in each hemisphere. (D) Mean linear SVM classification performance for discriminating (independently from the object size) between WHG and Reaching-only conditions as a function of the involved ROIs in each hemisphere. Error bars indicate one standard error of the mean. Asterisks assess statistical significance with one-tailed t tests across subjects with respect to 50% (significance levels: *P < .05; **P < .01; ***P < .001 ). respect to all the other ROIs (all ts ≥ 4.5, all Ps < 0.015). 45 (M = 0.63  0.03 SEM, t = 1.6) (see Fig. 3, panel B). Furthermore, higher accuracy was observed when decod- No further significant results were observed. ing from SPLap (M = 0.65  0.02 SEM) with respect to hAIP (M = 0.6  0.02 SEM, t(15) = 5.63, P < 0.0001). Whole hand grasping versus reaching Finally, decoding accuracy from SPOC areas was lower than those obtained from all of the other ROIs (all Results for reach-to-grasp using WHG versus Reaching- ts ≤ 2.8) with the exception of hAIP (t = 0.84) and BA 44/ only classification are summarized in Table 3. Brain and Behavior, doi: 10.1002/brb3.412 (8 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding (A) (B) (C) Figure 3. Results of the RM-ANOVA on the decoding accuracy. (A) Grasp type: independently from the hemisphere, decoding from somatosensory areas was significantly more accurate than from SPOC, hAIP, and BA 4p. (B) PG versus Reaching-only: independently from the hemisphere, decoding from somatosensory areas and BA 6 was significantly more accurate than from SPOC and hAIP. Moreover, decoding accuracy from voxel pattern activity of BA 6 was significantly higher than from SPLap. (C) WHG versus Reaching-only: independently from the hemisphere, decoding from somatosensory areas was significantly more accurate than from all the other ROIs. In contrast, decoding accuracy from SPOC areas was significantly lower than that from all the other ROIs. Moreover, decoding from BA 6 was significantly more accurate than from hAIP and BA 44/45. For all the three classifications, independently from the selected ROI, the decoding accuracy was significantly higher in the left (contralateral) hemisphere than in the right (ipsilateral) hemisphere (see the bottom part of each panel). Error bars indicate one standard error of the mean across subjects. Asterisks assess statistical significance levels, as reported in the Result section. The classifier analyses showed that, independently form accuracy from SPOC (M = 0.6  0.02 SEM) areas was the object size, it was possible to linearly discriminate lower than that from all of the other ROIs (all Ps < 0.002) between the WHG and the Reaching-only conditions except for BA 44/45 (M = 0.67  0.03 SEM, t = 2.27). from the voxel pattern activity of all the selected ROIs Moreover, decoding from BA 4p (M = 0.7  0.02 SEM) with the exception of the control ROI (see Table 3; Fig. 2, was more accurate that from hAIP (M = 0.66  0.02 SEM, panel D). t(15) = 2.52, P = 0.023) (see Fig. 3, panel C). No further For investigating possible interhemispheric asymmetries significant results were observed. for the selected ROIs, we performed an RM-ANOVA on the classifier accuracy using ROI (SPOC, SPLap, hAIP, BA Classification comparison 1/2/3ab, BA 44/45, BA 4p, and BA 6) and hemisphere (left vs. right) as within-subject factors. The analysis revealed a As a final step we compared the decoding accuracies main effect of ROI (F(6, 90) = 14.66, P < 0.001, among the three possible classifications (i.e., Grasp type, g = 0.49) and hemisphere (F(1, 15) = 10.96, P = 0.005, PG vs. Reaching-only, and WHG vs. Reaching-only). We g = 0.42). The two-way interaction was not significant computed a RM-ANOVA on the classifier accuracy using (F = 0.82). Independently from the selected ROI, classifier Classification (three levels) as within-subject factor, sepa- accuracy was higher when decoding from the left rately for each ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA (M = 0.72  0.01) than from the right (M = 0.68  0.01) 44/45, BA 4p, and BA 6). The analysis revealed for all the hemisphere. Paired t-tests (FDR corrected, corrected ROIs, except for SPOC areas (F = 2.44, P = 0.1), a main a = 0.033) showed higher decoding accuracy from BA1/2/ effect of Classification (all Fs ≥ 6.58, all Ps < 0.004, all 3ap (M = 0.79  0.02 SEM) with respect to all of the other g ≥ 0.31). A significant linear contrast (all Fs ≥ 13.14, all ROIs (all ts > 3.81, all Ps < 0.002) except for the BA 6 Ps < 0.002, all g ≥ 0.47) for all the ROIs, suggests that (M = 0.77  0.02 SEM, t = 1.004). Moreover, also decod- the decoding accuracies linearly increased from the Grasp ing from BA 6 was more accurate than from all of the other type toward PG versus Reaching-only and WGH versus ROIs (all ts ≥ 2.4, all Ps ≤ 0.029). In contrast, decoding Reaching-only classifications. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (9 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Table 2. PG versus reaching classification. Results obtained by train- Discussion ing linear SVM classifiers on each selected ROI, separately for the left and the right hemisphere. For each ROI, the results are expressed in Here, we exploited the potential of MVPA for better char- terms of classification performance on the test set (M  1 SEM) and acterizing the specific contribution of brain areas belong- the t statistics for assessing classification significance. ing to the reaching–grasping network in humans. We ROI Left hemisphere Right hemisphere performed MVPA on the activation patterns detected within ROIs of a wide frontoparietal network, for investi- SPOC .58  .03 .57  .03 gating three main aspects characterizing reaching-only and t(15) = 2.72, P = .016 t(15) = 1.96, P = .034 reach-to-grasp actions: the role of object size, the grasp SPLap .67  .03 .64  .02 t(15) = 6.65, P < .001 t(15) = 9.34, P < .001 type, and the congruence between the grasp type and the hAIP .63  .03 .56  .02 object size. In addition, to better define a possible differen- t(15) = 4.41, P = .001 t(15) = 2.26, P = .039 tial contribution of grasping-related areas, we also directly BA 1/2/3ab .75  .02 .71  .02 compared reach-to-grasp and reaching-only actions. t(15) = 12.9, P < .0001 t(15) = 1.27, P < .0001 Results showed no critical role of object size in per- BA 4p .69  .03 .62  .03 forming both reach-to-grasp and reaching-only actions. It t(15) = 7.39, P < .0001 t(15) = 4.68, P < .0001 was possible, however, to discriminate between grasp BA 6 .69  .02 .62  .03 t(15) = 8.22, P < .0001 t(15) = 4.68, P = .019 types (PG and WHG) regardless of the object size from BA 44/45 .62  .03 .63  .03 activation patterns within all the selected ROIs, with the t(15) = 4.68, P = .019 t(15) = 4.02, P < .0001 exception of bilateral SPOC and right hAIP. No effects Control ROI .52  .03, were found concerning the congruence between grasp t = .85, ns type and object size. Distinctions between reach-to-grasp SVM, support vector machine; PG, precision grasping; ROI, regions of (PG and WHG separately) and reaching-only actions interest; SPOC, superior parieto-occipital cortex; SPLap, superior pari- emerged from all the selected ROIs. Overall, decoding etal lobe; BA, Brodmann area; hAIP, anterior part of the human intra- accuracy was higher in distinguishing reach-to-grasp from parietal sulcus. reaching-only than in distinguishing PG from WHG actions. In both cases the left (controlateral) hemisphere played a prominent role in terms of decoding accuracy. Table 3. WHG versus reaching classification. Results obtained by Object size in reach-to-grasp action training linear SVM classifiers on each selected ROI, separately for the left and the right hemisphere. For each ROI, the results are expressed The evidence that object size did not play a relevant role in terms of classification performance on the test set (M  1 SEM) in reach-to-grasp action is consistent with the findings of and the t statistics for assessing classification significance. the reference study by Begliomini et al. (2007b), where the GLM did not reveal a modulation of the BOLD activ- ROI Left hemisphere Right hemisphere ity induced by object size. This allows us to discard the SPOC .63  .02 .57  .03 hypothesis that object size may account for the differen- t(15) = 5.99, P < .001 t(15) = 2.64, P < .005 tial activations within key areas concerned with visuomo- SPLap .7  .02 .7  .02 tor reach-to-grasp actions. This is, however, in contrast t(15) = 1.32, P < .001 t(15) = 13.47, P < .001 with a very recent finding of Monaco et al. (2015), where hAIP .69  .03 .63  .02 t(15) = 6.42, P < .001 t(15) = 5.45, P < .001 the authors used fMRI adaptation for investigating BA 1/2/3ab .76  .02 .73  .03 whether object size and location play a significant role in t(15) = 1.27, P < .0001 t = 7.76, P < .001 reach-to-grasp actions. Specifically, left hAIP showed BA 4p .73  .03 .67  .03 adaptation effect only to object size, whereas left SPOC, t = 7.76, P < .001 t(15) = 6.004, P < .001 primary somatosensory and motor areas (S1/M1), PMd BA 6 .79  .03 .75  .03 and SMA were sensitive to both object size and location. t(15) = 11.55, P < .001 t(15) = 2.07, P < .001 This discrepancy could be ascribed to several factors. BA 44/45 .68  .03 .66  .03 t(15) = 7.09, P < .001 t(15) = 5.86, P< .001 First, the paradigm of Monaco and colleagues was specifi- Control ROI .52  .03, cally conceived to highlight adaptation phenomena. t = .77 Indeed, the systematic variation intrinsic properties (e.g., object size) of the stimulus is crucial for adaptation SVM, support vector machine; ROI, regions of interest; SPOC, superior mechanisms. In contrast, in the study of Begliomini et al. parieto-occipital cortex; SPLap, superior parietal lobe; BA, Brodmann (2007b) this aspect was manipulated in a different way area; WHG, whole hand grasping; hAIP, anterior part of the human intraparietal sulcus. (i.e., object size was kept constant within each run). Cru- Brain and Behavior, doi: 10.1002/brb3.412 (10 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding cially, participants were informed about the size of the within the human hAIP (Begliomini et al. 2007a,b), object to be grasped at the beginning of each run (i.e., whether the human hAIP contains neural populations small object in the odd runs, and large object in the even selectively involved in the coding of different grasping ones). schemata remained to be clarified. Here we demonstrate that only left hAIP can discriminate between grasp types. This result is in agreement with the study by Gallivan Object size in reaching-only action et al. (2011), the first using a decoding method for dis- The fact that no critical role of object size emerged for criminating between different types of precision grasping the reaching-only action is in contrast with recent find- (toward a small vs. a large object stacked in a top and ings by Tarantino et al. (2014). The authors registered bottom location, respectively). However, Gallivan et al. kinematic and evoked related potentials while participants (2011) did not include the right hAIP in their decoding were asked to reach-only for differently sized objects. analysis, neglecting a possible role of the ipsilateral hemi- Results showed that the kinematics of reaching-only sphere in coding for different grasp types. Their ROI action, as well as the amplitude and the latency of P300 selection procedure was relying on the results of the GLM and N400 ERP components in parietal and prefrontal group random effects voxelwise analysis. Despite the fact sites, respectively, were modulated by object size, consis- that they avoided the “double dipping” problem tent with physiological findings on nonhuman primates (Kriegeskorte et al. 2009) by performing this analysis on a (Fattori et al. 2012). The discrepancy between these and different data set that was not used for decoding analysis, our findings could rely on the better temporal resolution this ROI selection procedure suffers from the limitations provided by ERPs with respect to fMRI, and might sug- of the GLM and it implies discarding all regions that do gest that object size, or more precisely the level of accu- not show significant effects at the level of single voxel racy of the movement determined by it, could modulate analysis. However, in our study, MVPA showed that no reaching-only actions on a temporal, rather than a spatial grasp type discrimination is possible from right hAIP. basis. Critically, we did not find any involvement of SPOC areas in distinguishing PG from WHG. Evidence of the involvement of left SPOC in discriminating between two Grasp type different precision grasping comes from the study of Gal- Concerning the grasp type, here we showed that discrimi- livan et al. (2011). Because in that study also object posi- nation between PG and WHG is possible from several tion was manipulated, it is unclear whether this result is areas of our selected network. In particular, decoding due either to the object size or to a different direction in accuracy was higher within the left (contralateral) rather reaching toward the bottom or top cube. The spatial than the right (ipsilateral) hemisphere. Conventional uni- aspect is crucial since it has been demonstrated that variate analyses performed by Begliomini et al. (2007b) SPOC activity is strictly related to the transport compo- revealed only an effect of grasp type (i.e., nent of the reach-to-grasp action (Cavina-Pratesi et al. [PGS + PGL] > [WHGS + WHGL]) in the left hAIP. 2010). In contrast, our results are consistent with the This discrepancy could be ascribed to the differences study of Fabbri et al. (2014), in which the comparison between univariate and multivariate analysis and to the between PG and WHG actions toward a spherical object fact that the findings by Begliomini et al. (2007b) were of constant size did not reveal any grasp type selectivity obtained by means of a subtraction procedure (reach-to- for the left SPOC. However, Fabbri et al. (2014) focused grasp—reaching-only) which is conventionally adopted by their attention only on the left hemisphere, discarding studies focusing on visuomotor transformation compo- possible results within the right hemisphere, whereas here nents underlying grasping (Culham et al. 2003, 2006). we show that the lack of grasp type selectivity character- Here we confirmed the involvement of left AIP in coding izes both contralateral and ipsilateral SPOC. differences between the two types of grasp, also at the The contribution of the SPLap in discriminating preci- level of voxel patterns. Moreover, MVPA revealed that sion versus whole hand grasp actions is consistent with other ROIs were involved in grasp type coding (all but the findings of Fabbri et al. (2014). Specifically, SPLa bilateral SPOC and right hAIP), because activity modula- broadly corresponds to monkey ventral intraparietal area tion within the voxel patterns related to the two condi- (VIP; Mars et al. 2011) and has been reported to be sen- tions were linearly separable within each ROI. sitive to the spatial congruency between visual and tactile Evidence that neurons within hAIP can selectively code information (Duhamel et al. 1998). for different grasp types comes from neurophysiological The involvement of bilateral BA 1/2/3ab in coding the studies (Murata et al. 2000). Although there is evidence grasp type could be explained by the sensitivity to differ- for different levels of activity depending on type of grasp ent somatosensory feedback provided by the two grasping ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (11 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. actions (i.e., PG and WHG). This peculiarity is confirmed or grasped (Ehrsson et al. 2000, 2001) objects. Bilateral by the results obtained for the bilateral dorsal premotor involvement of PMv during grasping movements has been (BA 6) and motor (BA 4p) cortices: somatosensory infor- also observed in TMS studies (Davare et al. 2006, 2008) mation from the hand should be integrated with motor revealing that lesioning either the left or the right PMv commands from frontal motor areas specifying the type modifies fingertip positioning, which is a prerequisite to of movement necessary to achieve the goal of grasping grasp an object properly (Sartori et al. 2011). (Gardner et al. 2007). Previous neurophysiological data report grasp type On the basis of neurophysiological and neuroimaging specificity within M1. M1 neurons active during WHG are studies, the role of the PMd for distal forelimb move- silent during PG (e.g., Muir and Lemon 1983). Although ments is becoming increasingly established (Raos et al. in humans different levels of activity in M1 for PG and 2004; Begliomini et al. 2007b). Here we extend this litera- WHG (Ehrsson et al. 2001; Begliomini et al. 2007a) have ture by demonstrating that within the left BA 6 different been reported, different spatial distributions of activity patterns of activity associated to different grasp types are associated with different grasping schemata had yet to be evident. This is in agreement with neurophysiological demonstrated. Here we showed that the bilateral BA 4p findings showing that F2 and F5 share similar functional significantly discriminated among grasping schemata. properties and act in concert for the control of grasping Since the movement is performed with the right hand, (Raos et al. 2004, 2006). In particular, F5 would be one might have expected this functional property to be mainly devoted to grasp selection, while F2 would moni- evident solely in left BA 4p. In general, however, the con- tor hand shaping during the ongoing movement, assuring tribution of the ipsilateral hemisphere could be hidden movement accuracy. Therefore, it might well be that the when using traditional GLM analysis, since in this case the discrimination ability shown here by left BA 6 indicates a research question is based on searching where in the brain differential hand shape monitoring depending on grasp there is a significant greater BOLD activity for an experi- types. Grasp type classification was also possible within mental condition with respect to a second one. This the right BA 6: as demonstrated by previous findings, this assumption could discard the involvement of brain areas result could be explained in terms of learning new motor where the experimental manipulations produce an effect sequences or by high requirements in terms of precision at the level of activation patterns rather than at the level and coordination, independently from the hand used of single voxel activity. In contrast, MVPA is intended to (Davare et al. 2006; Begliomini et al. 2008). In this uncover whether and to what extent a brain area is coding regard, PG requires high precision in positioning the two differential voxel pattern representations for two experi- fingers on the opposite sides of the object, whereas WHG mental conditions. We found that also the ipsilateral requires coordination among phalanxes of all fingers. hemisphere has a role in representing different grasp types, Therefore, it is conceivable that the right BA 6 acts in but the decoding accuracy was significantly higher in left concert with the left BA 6, in order to fulfill the accuracy than in the right BA4p. Recent findings show that admin- and coordination requirements intrinsic to the considered istering rTMS (repetitive TMS) on ipsilateral M1 affects types of grasp (Begliomini et al. 2007b). the timing of muscle recruitment, resulting in a loss of Neurophysiological data suggest a key role for PMv in coordination during hand movement (Davare et al. 2007). selecting the most appropriate motor configuration on This phenomenon potentially occurs on the basis of recip- the basis of 3D analysis provided by AIP (Fagg and Arbib rocal connections between cortices via the corpus callosum 1998). In this respect, human neuroimaging findings have (Boroojerdi et al. 1996; Di Lazzaro et al. 1999). provided mixed results. Whereas isometric grasping tasks Overall, we showed, for the first time, that grasp type detected PMv activity (Ehrsson et al. 2001), visually could be decoded from a wide frontoparietal network in guided tasks did not (Culham et al. 2006; Begliomini both hemispheres, with the left (controlateral) hemisphere et al. 2007a,b). Therefore, it was unclear whether the playing a more informative role with respect to the right human PMv really holds a function of “motor vocabu- (ipsilateral) one. However, since participants were able to lary” similarly to macaque F5. Our results extend this lit- see their own movements, results about grasp type could erature by showing that in humans bilateral BA44/45 be also interpreted as different representations mediated exhibits a differential activation pattern in association by the vision of a different movement. with different grasp types and supports the parallelism between macaque and humans in grasp type selectivity at Congruence the level of premotor cortices (Murata et al. 1997; Carpa- neto et al. 2011). Furthermore, several functional imaging Despite the fact that in the study by Begliomini et al. studies have shown activation in both the left and right (2007b) the contrast between natural and constrained PMv when subjects manipulated (Binkofski et al. 1999) reach-to-grasp actions (i.e., our Congruence classification) Brain and Behavior, doi: 10.1002/brb3.412 (12 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding revealed a greater activation within few voxels belonging right hAIP in visuomotor reaching-only rather than to bilateral PMd and left M1, this was not the case with grasping action representation. MVPA. Thus, a difference between congruent (i.e., PGS The contribution of SPLap in reaching-only and reach- and WGHL) and incongruent (PGL and WHGS) grasping to-grasp actions was not surprising. This result is consis- actions could be revealed only in terms of univariate anal- tent with those of Fabbri et al. (2014) and it can be ysis, whereas no activation pattern within wider brain explained by the fact that this area is sensitive to the areas encoded this difference. This might stem from a direction of visual, tactile and auditory stimuli (Bremmer limit of MVPA in distinguishing patterns of activity et al., 2001). Indeed, in our experiment, participants were across a large set of voxels (i.e., large-size ROIs) when the informed on the type of action to be performed (e.g., discriminating information is encoded in a small percent- reach-to-grasp vs. reaching-only) by auditory cues. age of the input voxels. MVPA is more sensitive to dis- The fact that BA1/2/3ab, bilaterally, was involved in the tributed coding of information whereas univariate discrimination between reaching-only and reach-to-grasp analysis is more sensitive to global engagement in ongo- actions could be explained by a sensitivity to different ing tasks (Jimura and Poldrack 2012). Another possible somatosensory feedback provided by the two actions explanation for the lack of the congruence effect, could toward the object. In addition, this was indexed by higher rely on the fact that we did not apply spatial smoothing accuracy in discriminating between reaching-only and to fMRI data before MVPA. As recently shown in a study reach-to-grasp using WHG rather than PG, probably mir- based on simulated data (Stelzer et al. 2014), the com- roring a greater difference in hand configuration, and bined use of spatial smoothing and cluster based correc- hence in somatosensory feedback. tion could increase the number of false positives and false The involvement of BA44/45 in discriminating between negatives, respectively. Thus, both univariate and multi- reaching-only and reach-to-grasp actions is consistent variate approaches could introduce possible limitations, with the most recent findings of Fabbri et al. (2014) and and their combination should be more informative than Gallivan et al. (2011), both using multivariate approaches the use of a single approach (see also Gallivan et al. 2011 for analyzing fMRI data. The first study highlighted that for a similar argument). reach direction and grip type are both represented in left PMv, whereas the second one showed that left PMv was involved in the discrimination between precision grasping Reach-to-grasp versus reaching-only actions and touching, in both the planning and the execution Here, we showed that it was possible to discriminate phase of the actions. between reach-to-grasp and reaching-only actions from The contribution of bilateral BA 6 in distinguishing the selected frontoparietal network. Interestingly, a between reach-to-grasp and reaching-only actions is not prominent role in characterizing the reaching–grasping surprising, since this area has been firstly suggested to network is played by bilateral SPOC and right hAIP. code only for the transport phase of the hand toward an These areas were not sensitive in decoding grasp type, but object (i.e., reaching) (Begliomini et al. 2014; Culham played a significant role in discriminating between reach- et al. 2006; Vesia and Crawford, 2012) and has been to-grasp and reaching-only actions. shown to be involved in the representation of both the Our results on SPOC suggest that the contribution of transport and the hand preshaping components of reach- these areas might be more crucial for reaching-only than ing-only and reach-to-grasp actions, respectively (e.g., shaping the fingers for different grip types, which is con- Fabbri et al. 2014). The bilateral involvement of PMd in sistent with the findings of Cavina-Pratesi et al. (2010). coding direction and amplitude of reaching-only has been These authors reported that the human SPOC showed shown by Fabbri et al. (2012), thus it was not surprising stronger activation during reach-to-grasp action toward that different activity patterns are present in these areas far rather than near locations, suggesting a preference for for reaching-only and reach-to-grasp actions. the transport rather than the grasp component. However, Finally, the involvement of primary motor area in our results are in contrast with those reported by Fabbri reach-to-grasp versus reaching-only discrimination was et al. (2014), where left SPOC did not show any effect in expected, as well as its involvement in distinguishing finer discriminating between reach-to-grasp and reaching-only aspects of the grasping action (i.e., grasp type classifica- actions. tion). These results are consistent with the most recent Our results on right hAIP suggest that this area con- neuroimaging studies in humans (Gallivan et al. 2011; tributes to the representation of both reaching-only and Fabbri et al. 2012, 2014). Interestingly, the novelty of these reach-to-grasp actions, but it does not appear to be criti- results relies on the right (ipsilateral) contribution of cally involved in the finer distinctions between grasp BA4p. As in the case of the grasp type classification, we types. This latter result might indicate a major role of the found a bilateral involvement of BA4p in discriminating ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (13 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. between reaching-only and reach-to-grasp actions, even if a clear-cut distinction between a dorsomedial (e.g., SPOC, the left (contralateral) hemisphere played a prominent medial intraparietal area MIP, and PMd) and dorsolateral role, in terms of classification accuracy. (e.g., hAIP and PMv) pathways, specialized for reaching- In conclusion, our results showed significant hemi- only and reach-to-grasp actions, respectively, as reported spheric asymmetries in discriminating reaching-only from in a series of recent studies on human and nonhuman pri- reach-to-grasp actions and PG from WHG, which con- mates (Fattori et al. 2009, 2010; Cavina-Pratesi et al. 2010; sisted of a left (i.e., contralateral) hemisphere dominance. Monaco et al. 2011). Our results are also consistent with This is consistent to our expectations, since participants the findings of Grol et al. (2007), which argue against the were using the right hand to perform the actions. Fur- presence of dedicated cerebral circuits for reaching-only thermore, we found that somatosensory and dorsal pre- and reach-to grasp actions, suggesting that the contribu- motor areas were more responsive in distinguishing tions of the dorsolateral and the dorsomedial circuits are a between reaching-only and reach-to-grasp actions, with function of the degree of online control required by the respect to all the other areas within the selected network. movement. Finally, our results are perfectly consistent with Finally, within the selected network, decoding accuracy the theory of a dorsomedial visual stream involved in was higher when discriminating reaching-only from reach-to-grasp actions, suggested by Galletti et al. (2003) reach-to-grasp action, when using WHG rather than PG. in nonhuman primates, and well documented by Fattori This result, together with the fact that no critical role was et al. (2009, 2010). Reaching-only and reach-to-grasp played by object size, could suggest that different activa- actions could be better characterized by temporal, rather tion patterns underlying reach-to-grasp and reaching-only than spatial criteria across planning and execution stages of actions could be mainly due to a physical difference in the action, as also suggested by a recent study of Beglio- hand configuration. The fact that this information was mini et al. (2014). Here we showed that several areas of the probably guiding the discrimination within all the selected human reaching–grasping network are involved in process- network (including parietal areas) indicates that hand ing aspects related to both reach-to-grasp and reaching- preshaping begins in early stages of action planning (i.e., only actions. Crucially, the precise nature, in terms of tim- action preparation), as also suggested by Gallivan et al. ing and direction (causality—Davare et al. 2010; Grol et al. (2011) and Begliomini et al. (2014). 2007) of the relations between the involved brain areas remains to be clarified by future studies. Altogether, the findings provided by the integrated Conclusion approach adopted in this work enrich the current knowl- To summarize, in our study no critical role of object size edge regarding the functional role of key brain areas emerged for both reaching-only and reach-to-grasp involved in the cortical control of reaching-only and actions. This result runs against the hypothesis that the reach-to-grasp actions in humans, by revealing novel fine- intrinsic object properties (e.g., object size) could play a grained distinctions among action types within a wide key role in both reach-to-grasp and reaching-only actions. frontoparietal network. Here we showed, for the first time that grasp type (i.e., PG vs. WHG), independently from object size, can be reliably Acknowledgments discriminated by a linear classifier within a wide fron- This work was supported by a grant from the European toparietal network distributed across both the hemispheres, Research Council (grant no. 210922) and the University with the exception of SPOC areas and right hAIP. The left of Padova (Strategic Grant NEURAT) to M. Zorzi. (i.e., controlateral) hemisphere, however, played a crucial role in terms of decoding accuracy. No significant interac- tion between the grasp type and the object size (i.e., our Conflict of Interest congruence classification) emerged within the considered None declared. network, despite the fact that univariate analysis of the same data set (Begliomini et al. 2007b) showed that activ- References ity of few voxels within PMd and M1 areas was modulated by congruence. This highlights the importance to perform Amunts, K., and K. Zilles. 2001. Advances in cytoarchitectonic data analysis from a more comprehensive perspective, mapping of the human cerebral cortex. Neuroimaging Clin. combining both univariate and multivariate analyses. This N. Am. 11:151–169. integrated approach could provide more informative Amunts, K., A. Schleicher, U. Burgel, € H. Mohlberg, H. B. results and a deeper understanding of the neural dynamics Uylings, and K. Zilles. 1999. Broca’s region revisited: underlying the cognitive processes of interest. Finally, our cytoarchitecture and intersubject variability. J. Comp. results provided further evidence against the hypothesis of Neurol. 412:319–341. Brain and Behavior, doi: 10.1002/brb3.412 (14 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Begliomini, C., M. B. Wall, A. T. Smith, and U. Castiello. produces fMRI activation in dorsal but not ventral stream 2007a. Differential cortical activity for precision and whole- brain areas. Exp. Brain Res. 153:180–189. hand visually guided grasping in humans. Eur. J. Neurosci. Culham, J. C., C. Cavina-Pratesi, and A. Singhal. 2006. The 25:1245–1252. role of parietal cortex in visuomotor control: what have we Begliomini, C., A. Caria, W. Grodd, and U. Castiello. 2007b. learned from neuroimaging? Neuropsychologia 44:2668– Comparing natural and constrained movements: new insight 2684. into the visuomotor control of grasping. PLoS One 2:e1108. Davare, M., M. Andrei, G. Cosnard, J. L. Thonnard, and E. Begliomini, C., C. Nelini, A. Caria, W. Grodd, and U. Olivier. 2006. Dissociating the role of ventral and dorsal Castiello. 2008. Cortical activations in humans grasp-related premotor cortex in precision grasping. J. Neurosci. 26:2260– areas depend on hand used and handedness. PLoS One 3: 2268. e3388. Davare, M., J. Duque, Y. Vandermeeren, J. L. Thonnard, and Begliomini, C., T. De Sanctis, M. Marangon, V. Tarantino, L. E. Olivier. 2007. Role of the ipsilateral primary motor cortex Sartori, D. Miotto, et al. 2014. An investigation of the neural in controlling the timing of hand muscle recruitment. circuits underlying reaching and reach-to-grasp movements: Cereb. Cortex 17:353–362. from planning to execution. Front. Hum. Neurosci. 8:676. Davare, M., R. Lemon, and E. Olivier. 2008. Selective Binkofski, F., G. Buccino, S. Posse, R. J. Seitz, G. Rizzolatti, modulation of interactions between ventral premotor cortex and A. Freund. 1999. Fronto-parietal circuit for object and primary motor cortex during precision grasping in manipulation in man: evidence from an fMRI-study. Eur. J. humans. J. Physiol. 586:2735–2742. Neurosci. 11:3276–3286. Davare, M., J. C. Rothwell, and R. N. Lemon. 2010. Causal Boroojerdi, B., K. Diefenbach, and A. Ferbert. 1996. connectivity between the human anterior intraparietal area Transcallosal inhibition in cortical and subcortical cerebral and premotor cortex during grasp. Curr. Biol. 20:176–181. vascular lesions. J. Neurol. Sci. 144:160–170. Di Bono, M. G., and M. Zorzi. 2008. Decoding cognitive states Bremmer, F., A., Schlack, N. J., Shah, O., Zafiris, M., from fMRI data using support vector regression. PsychNol. Kubischik, K. P., Hoffmann, K., Zilles, and G. R. Fink. 2001. J. 6:189201. Polymodal motion processing in posterior parietal and Di Lazzaro, V., A. Oliviero, P. Profice, A. Insola, P. Mazzone, premotor cortex: a human fMRI study strongly implies P. Tonali, et al. 1999. Direct demonstration of equivalencies between humans and monkeys. Neuron interhemispheric inhibition of the human motor cortex 29:287–296. produced by transcranial magnetic stimulation. Exp. Brain Carpaneto, J., M. A. Umilta, L. Fogassi, A. Murata, V. Gallese, Res. 124:520–524. S. Micera, et al. 2011. Decoding the activity of grasping Duhamel, J. R., C. L. Colby, and M. E. Goldberg. 1998. Ventral neurons recorded from the ventral premotor area F5 of the intraparietal area of the macaque: congruent visual and macaque monkey. Neuroscience 188:80–94. somatic response properties. J. Neurophysiol. 79:126–136. Castiello, U. 2005. The neuroscience of grasping. Nat. Rev. Ehrsson, H. H., A. Fagergren, T. Jonsson, G. Westling, R. S. Neurosci. 6:726–736. Johansson, and H. Forssberg. 2000. Cortical activity in Castiello, U., and C. Begliomini. 2008. The cortical control of precision- versus power-grip tasks: an fMRI study. J. visually guided grasping. Neuroscientist 14:157–170. Neurophysiol. 83:528–536. Cavina-Pratesi, C., M. Goodale, and J. C. Culham. 2007. FMRI Ehrsson, H. H., E. Fagergren, and H. Forssberg. 2001. reveals a dissociation between grasping and perceiving the Differential fronto-parietal activation depending on force size of real 3D objects. PLoS One 5:1–14. used in a precision grip task: an fMRI Study. J. Cavina-Pratesi, C., S. Monaco, P. Fattori, C. Galletti, T. D. Neurophysiol. 85:2613–2623. McAdam, D. J. Quinlan, et al. 2010. Functional magnetic Fabbri, S., A. Caramazza, and A. Lingnau. 2012. Distributed resonance imaging reveals the neural substrates of arm sensitivity for movement amplitude in directionally tuned transport and grip formation in reach-to-grasp actions in neuronal populations. J. Neurophysiol. 107:1845–1856. humans. J. Neurosci. 30:10306–10323. Fabbri, S., L. Strnad, A. Caramazza, and A. Lingnau. 2014. Chen, Y., P. Namburi, L. T. Elliott, J. Heinzle, C. S. Soon, M. Overlapping representations for grip type and reach W. Chee, et al. 2011. Cortical surface-based searchlight direction. NeuroImage 94:138–146. decoding. NeuroImage 56:582–592. Fagg, A. H., and M. A. Arbib. 1998. Modeling parietal- Choi, H. J., K. Zilles, H. Mohlberg, A. Schleicher, G. R. Fink, premotor interactions in primate control of grasping. Neural E. Armstrong, et al. 2006. Cytoarchitectonic identification Netw. 11:1277–1303. and probabilistic mapping of two distinct areas within the Fattori, P., R. Breveglieri, N. Marzocchi, D. Filippini, A. anterior ventral bank of the human intraparietal sulcus. J. Bosco, and C. Galletti. 2009. Hand orientation during Comp. Neurol. 495:53–69. reach-to-grasp movements modulates neuronal activity in Culham, J. C., S. L. Danckert, J. F. DeSouza, J. S. Gati, R. S. the medial posterior parietal area V6A. J. Neurosci. Menon, and M. A. Goodale. 2003. Visually guided grasping 29:1928–1936. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (15 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Fattori, P., V. Raos, R. Breveglieri, A. Bosco, N. Marzocchi, Grafton, S. T., M. A. Arbib, L. Fadiga, and G. Rizzolatti. 1996. and C. Galletti. 2010. The dorsomedial pathway is not just Localization of grasp representations in humans by positron for reaching: grasping neurons in the medial parieto- emission tomography. Exp. Brain Res. 112:103–111. occipital cortex of the macaque monkey. J. Neurosci. Grefkes, C., S. Geyer, T. Schormann, P. Roland, and K. Zilles. 30:342–349. 2001. Human somatosensory area 2: observer-independent Fattori, P., R. Breveglieri, V. Raos, A. Bosco, and C. Galletti. cytoarchitectonic mapping, interindividual variability, and 2012. Vision for action in the macaque medial posterior population map. NeuroImage 14:617–631. parietal cortex. J. Neurosci. 32:3221–3234. Grol, M. J., J. Majdandzic, K. E. Stephan, L. Verhagen, H. C. Filimon, F. 2010. Human cortical control of hand movements: Dijkerman, H. Bekkering, et al. 2007. Parieto-frontal parietofrontal networks for reaching, grasping, and pointing. connectivity during visually guided grasping. J. Neurosci. Neuroscientist 16:388–407. 27:11877–11887. Filimon, F., J. D. Nelson, R. S. Huang, and M. I. Sereno. 2009. Hagberg, G. E., G. Zito, F. Patria, and J. N. Sanes. 2001. Multiple parietal reach regions in humans: cortical Improved detection of event-related functional MRI signals representations for visual and proprioceptive feedback using probability functions. NeuroImage 14:1193–1205. during on-line reaching. J. Neurosci. 29:2961–2971. Hinkley, L. B., L. A. Krubitzer, J. Padberg, and E. A. Disbrow. Frey, H. S., D. Vinton, R. Norlund, and S. T. Grafton. 2005. 2009. Visual-manual exploration and posterior parietal Cortical topography of human anterior intraparietal cortex cortex in humans. J. Neurophysiol. 102:3433–3446. active during visually guided grasping. Cogn. Brain Res. Jeannerod, M. 1981. Specialized channels for cognitive 23:397–405. responses. Cognition 10:135–137. Galletti, C., D. F. Kutz, M. Gamberini, R. Breveglieri, and P. Jeannerod, M., M. A., Arbib, G., Rizzolatti, and H., Sakata. Fattori. 2003. Role of the medial parieto-occipital cortex in 1995. Grasping objects: the cortical mechanisms of the control of reaching and grasping movements. Exp. Brain visuomotor transformation. Trends Neurosci 18:314–320. Res. 153:158–170. Jimura, K., and R. A. Poldrack. 2012. Analyses of regional- Gallivan, J. P., D. A. McLean, K. F. Valyear, C. E. Pettypiece, average activation and multivoxel pattern information tell and J. C. Culham. 2011. Decoding action intentions from complementary stories. Neuropsychologia 50:544–552. preparatory brain activity in human parieto-frontal Kriegeskorte, N., W. K. Simmons, P. S. F. Bellgowan, and C. I. networks. J. Neurosci. 31:9599–9610. Baker. 2009. Circular analysis in systems neuroscience: the Gardner, E. P., K. S. Babu, S. D. Reitzen, S. Ghosh, A. S. dangers of double dipping. Nat. Neurosci. 12:535–540. Brown, J. Chen, et al. 2007. Neurophysiology of prehension. Kroliczak, G., C. Cavina-Pratesi, D. A. Goodman, and J. C. I. Posterior parietal cortex and object-oriented hand Culham. 2007. What does the brain do when you fake it? behaviors. J. Neurophysiol. 97:387–406. An FMRI study of pantomimed and real grasping. J. Geyer, S. 2003. Brodmanns areas. Pp. 482–496 in M. J. Neurophysiol. 97:2410–2422. Aminoff and R. B. Daroff, eds. Encyclopedia of the Kuhtz-Buschbeck, J. P., R., Gilster, S., Wolff, S., Ulmer, H., neurological sciences. Academic Press, San Diego. Siebner, and O. Jansen. 2008. Brain activity is similar during Geyer, S., A. Ledberg, A. Schleicher, S. Kinomura, T. precision and power gripping with light force: an fMRI Schormann, U. Burgel, € et al. 1996. Two different areas study. Neuroimage 40:1469–1481. within the primary motor cortex of man. Nature 382:805– Lawrence, D. G., and D. A. Hopkins. 1976. The development 807. of motor control in the rhesus monkey: evidence concerning Geyer, S., A. Schleicher, and K. Zilles. 1999. Areas 3a, 3b, and the role of corticomotoneuronal connections. Brain 99:235– 1 of human primary somatosensory cortex: 1. 254. Microstructural organization and interindividual variability. Luppino, G., A. Murata, P. Govoni, and M. Matelli. 1999. NeuroImage 10:63–83. Largely segregated parietofrontal connections linking rostral Geyer, S., T. Schormann, H. Mohlberg, and K. Zilles. 2000. intraparietal cortex (areas AIP and VIP) and the ventral Areas 3a, 3b, and 1 of human primary somatosensory premotor cortex (areas F5 and F4). Exp. Brain Res. cortex: 2. Spatial normalization to standard anatomical 128:181–187. space. NeuroImage 11:684–696. Mars, R. B., S. Jbabdi, J. Sallet, J. X. O’ Reilly, P. L. Glover, S., R. C. Miall, and M. F. S. Rushworth. 2005. Parietal Croxson, E. Olivier, et al. 2011. Diffusion-weighted rTMS disrupts the initiation but not the execution of on- imaging tractography-based parcellation of the human line adjustments to a perturbation of object size. J. Cogn. parietal cortex and comparison with human and macaque Neurosci. 17:124–136. resting-state functional connectivity. J. Neurosci. Godschalk, M., R. N. Lemon, H. G. Nijs, and H. G. Kuypers. 31:4087–4100. 1981. Behaviour of neurons in monkey peri-arcuate and Matelli, M., and G. Luppino. 2001. Parietofrontal circuits for precentral cortex before and during visually guided arm and action and space perception in the macaque monkey. hand movements. Exp. Brain Res. 44:113–116. NeuroImage 14:S27–S32. Brain and Behavior, doi: 10.1002/brb3.412 (16 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Matelli, M., G. Luppino, and G. Rizzolatti. 1985. Patterns of Raos, V., M. A. Umilta, A. Murata, L. Fogassi, and V. Gallese. cytochrome oxidase activity in the frontal agranular cortex 2006. Functional properties of grasping-related neurons in of the macaque monkey. Behav. Brain Res. 18:125–136. the ventral premotor area F5 of the macaque monkey. J. Matelli, M., G. Luppino, and G. Rizzolatti. 1991. Architecture of Neurophysiol. 95:709–729. superior and mesial area 6 and the adjacent cingulate cortex Rice, N. J., E. Tunik, and S. T. Grafton. 2006. The anterior in the macaque monkey. J. Comp. Neurol. 311:445–462. intraparietal sulcus mediates grasp execution, independent Moll, L., and H. G. Kuypers. 1977. Premotor cortical ablations of requirement to update: new insights from transcranial in monkeys: contralateral changes in visually guided magnetic stimulation. J. Neurosci. 26:8176–8182. reaching behavior. Science 198:317–319. Rizzolatti, G., and M. A. Arbib. 1998. Language within our Monaco, S., C. Cavina-Pratesi, A. Sedda, P. Fattori, C. Galletti, grasp. Trends Neurosci. 21:188–194. and J. C. Culham. 2011. Functional magnetic resonance Rizzolatti, G., and G. Luppino. 2001. The cortical motor adaptation reveals the involvement of the dorsomedial system. Neuron 31:889–901. stream in hand orientation for grasping. J. Neurophysiol. Rizzolatti, G., L. Camarda, L. Fogassi, M. Gentilucci, G. 106:2248–2263. Luppino, and M. Matelli. 1988. Functional organization of Monaco, S., A. Sedda, C. Cavina-Pratesi, and J. C. Culham. inferior area 6 in the macaque monkey. Exp. Brain Res. 2015. Neural correlates of object size and object location 71:491–507. during grasping actions. Eur. J. Neurosci. 41:454–465. Rizzolatti, G., G. Luppino, and M. Matelli. 1998. The Muir, R. B., and R. N. Lemon. 1983. Corticospinal neurons organisation of the cortical motor system: new concepts. with a special role in precision grip. Brain Res. 261:312–316. Electroencephalogr. Clin. Neurophysiol. 106:283–296. Murata, A., L. Fadiga, L. Fogassi, V. Gallese, V. Raos, and G. Rizzolatti, G., L. Fogassi, and V. Gallese. 2002. Motor and Rizzolatti. 1997. Object representation in the ventral cognitive functions of the ventral premotor cortex. Curr. premotor cortex (area F5) of the monkey. J. Neurophysiol. Opin. Neurobiol. 12:149–154. 78:2226–2230. Sartori, L., E. Straulino, and U. Castiello. 2011. How objects Murata, A., V. Gallese, G. Luppino, M. Kaseda, and H. Sakata. are grasped: the interplay between affordances and end- 2000. Selectivity for the shape, size and orientation of goals. PLoS One 6:e25203. objects for grasping in neurons of monkey parietal area AIP. Scheperjans, F., K. Hermann, S. B. Eickhoff, K. Amunts, A. J. Neurophysiol. 83:2580–2601. Schleicher, and K. Zilles. 2008. Observer-independent Oldfield, R. C. 1971. The assessment and analysis of cytoarchitectonic mapping of the human superior parietal handedness: the Edinburgh Inventory. Neuropsychologia cortex. Cereb. Cortex 18:846–867. 9:97–113. Stelzer, J., G. Lohmann, K. Mueller, T. Buschmann, and R. O’Toole, A. J., F. Jiang, H. Abdi, N. Penard, J. P. Dunlop, and Turner. 2014. Deficient approaches to human neuroimaging. M. A. Parent. 2007. Theoretical, statistical, and practical Front. Hum. Neurosci. 8:462. perspectives on pattern-based classification approaches to Taira, M., S. Mine, A. P. Georgopoulos, A. Murata, and H. the analysis of functional neuroimaging data. J. Cogn. Sakata. 1990. Parietal cortex neurons of the monkey related Neurosci. 19:1735–1752. to the visual guidance of hand movement. Exp. Brain Res. Passingham, R. E. 1987. Two cortical systems for directing 83:29–36. movement. Ciba Found. Symp. 132:151–164. Tarantino, V., T. De Sanctis, E. Straulino, C. Begliomini, and Pereira, F., T. Mitchell, and M. Botvinick. 2009. Machine U. Castiello. 2014. Object size modulates fronto-parietal learning classifiers and fMRI: a tutorial overview. activity during reaching movements. Eur. J. Neurosci. NeuroImage 45:S199–S209. 39:1528–1537. Pitzalis, S., M. I. Sereno, G. Committeri, P. Fattori, G. Galati, Tosoni, A., S. Pitzalis, G. Committeri, P. Fattori, C. Galletti, A. Tosoni, et al. 2013. The human homologue of macaque and G. Galati. 2014. Resting-state connectivity and area V6A. NeuroImage 82:517–530. functional specialization in human medial parieto-occipital Pitzalis, S., P. Fattori, and C. Galletti. 2015. The human cortex. Brain Struct. Funct. 220:3307–3321. cortical areas V6 and V6A. Vis. Neurosci. 32:E007. Tunik, E., S. H. Frey, and S. T. Grafton. 2005. Virtual lesions Prado, J., S. Clavagnier, H. Otzenberger, C. Scheiber, H. of the anterior intraparietal area disrupt goal-dependent on- Kennedy, and M. T. Perenin. 2005. Two cortical systems for line adjustment of grasp. Nat. Neurosci. 8:505–511. reaching in central and peripheral vision. Neuron 48:849– Tunik, E., N. J. Rice, A. Hamilton, and S. T. Grafton. 2007. 858. Beyond grasping: representation of action in human anterior Raos, V., M. A. Umilta, V. Gallese, and L. Fogassi. 2004. intraparietal sulcus. NeuroImage 36:T77–T86. Functional properties of grasping-related neurons in the Vesia, M., and J. D. Crawford. 2012. Specialization of reach dorsal premotor area F2 of the macaque monkey. J. function in human posterior parietal cortex. Exp Brain Res. Neurophysiol. 92:1990–2002. 221:1–18. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (17 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Weinrich, M., and S. P. Wise. 1982. The premotor cortex of Zorzi, M., M. G. Di Bono, and W. Fias. 2011. Distinct the monkey. J. Neurosci. 2:1329–1345. representations of numerical and non-numerical order in Xia, M., J. Wang, and Y. He. 2013. BrainNet Viewer: a the human intraparietal sulcus revealed by multivariate network visualization tool for human brain connectomics. pattern recognition. NeuroImage 56:674–680. PLoS One 8:e68910. Brain and Behavior, doi: 10.1002/brb3.412 (18 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Brain and Behavior Wiley

Probing the reaching–grasping network in humans through multivoxel pattern decoding

Download PDF
18 pages

Loading next page...
 
/lp/wiley/probing-the-reaching-grasping-network-in-humans-through-multivoxel-p3HUFQj0qT

References (166)

Publisher
Wiley
Copyright
© 2015 Published by Wiley Periodicals, Inc.
ISSN
2162-3279
eISSN
21623279
DOI
10.1002/brb3.412
pmid
26664793
Publisher site
See Article on Publisher Site

Abstract

Functional magnetic resonance imaging, Introduction: The quest for a putative human homolog of the reaching–grasp- multivoxel pattern decoding, reaching-only ing network identified in monkeys has been the focus of many neuropsycholog- action, visuomotor reach-to-grasp action ical and neuroimaging studies in recent years. These studies have shown that Correspondence the network underlying reaching-only and reach-to-grasp movements includes Maria Grazia Di Bono, Dipartimento di the superior parieto-occipital cortex (SPOC), the anterior part of the human Psicologia Generale, University of Padova, via intraparietal sulcus (hAIP), the ventral and the dorsal portion of the premotor Venezia 8, 35131 Padova, Italy. Tel: +39 049 cortex, and the primary motor cortex (M1). Recent evidence for a wider fron- 8276642; Fax: +39 049 8276600; toparietal network coding for different aspects of reaching-only and reach-to- E-mail: [email protected] grasp actions calls for a more fine-grained assessment of the reaching–grasping and Marco Zorzi, Dipartimento di Psicologia network in humans by exploiting pattern decoding methods (multivoxel pattern Generale, University of Padova, via Venezia analysis—MVPA). Methods: Here, we used MPVA on functional magnetic res- 8, 35131 Padova, Italy. Tel: +39 049 onance imaging (fMRI) data to assess whether regions of the frontoparietal net- 8276618; Fax: +39 049 8276600; work discriminate between reaching-only and reach-to-grasp actions, natural E-mail: [email protected] and constrained grasping, different grasp types, and object sizes. Participants were required to perform either reaching-only movements or two reach-to- Funding Information grasp types (precision or whole hand grasp) upon spherical objects of different This work was supported by a grant from the European Research Council (grant no. sizes. Results: Multivoxel pattern analysis highlighted that, independently from 210922) and the University of Padova the object size, all the selected regions of both hemispheres contribute in coding (Strategic Grant NEURAT) to M. Zorzi. for grasp type, with the exception of SPOC and the right hAIP. Consistent with recent neurophysiological findings on monkeys, there was no evidence for a Received: 4 May 2015; Revised: 27 July clear-cut distinction between a dorsomedial and a dorsolateral pathway that 2015; Accepted: 13 September 2015 would be specialized for reaching-only and reach-to-grasp actions, respectively. Nevertheless, the comparison of decoding accuracy across brain areas Brain and Behavior, 2015; 5(11), e00412, highlighted their different contributions to reaching-only and grasping actions. doi: 10.1002/brb3.412 Conclusions: Altogether, our findings enrich the current knowledge regarding the functional role of key brain areas involved in the cortical control of reach- ing-only and reach-to-grasp actions in humans, by revealing novel fine-grained distinctions among action types within a wide frontoparietal network. activity of single neurons is recorded with techniques Introduction allowing a high level of spatial and temporal resolution. In the domain of motor control great attention has been These studies have identified the main cortical structures given to reaching-only and reach-to-grasp actions, appar- involved in the control of visually guided reach-to-grasp ently simple and straightforward behaviors which are part movements. They are the primary motor cortex (F1), the of our everyday life motor repertoire, and fundamental premotor cortex (area F5), and the anterior part of the for our interaction with the environment. intraparietal sulcus (AIP; Murata et al. 1997, 2000). The A great extent of our knowledge regarding the cortical ability to perform a successful reach-to-grasp action control of reach-to-grasp movements is rooted in neuro- depends primarily on the integrity of F1; indeed, lesions physiological studies on behaving monkeys, in which the of this area in macaques produce a remarkable deficit in ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (1 of 18) This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Multivoxel Pattern Decoding M. G. Di Bono et al. the control of individual fingers, bringing to a loss of monkeys (Cavina-Pratesi et al. 2007; Culham et al. 2006; coordination abilities (Lawrence and Hopkins 1976). Area Kroliczak et al. 2007; Tunik et al. 2007; for reviews see F5, which forms the rostral part of the macaque ventral Castiello 2005; Castiello and Begliomini 2008; Filimon premotor cortex (PMv) and AIP, a small zone lying 2010). Overall, reach-to-grasp fMRI studies converge in within the rostral part of the posterior bank of the intra- considering the anterior part of the human intraparietal parietal sulcus (Matelli et al. 1985; Luppino et al. 1999; sulcus (hAIP), a likely homolog of monkey AIP (Grafton Matelli and Luppino 2001) are directly connected and are et al. 1996; Culham et al. 2003; Frey et al. 2005; Beglio- involved in converting intrinsic object properties (e.g., mini et al. 2007a; Hinkley et al. 2009). The key role of shape, size) into a proper hand conformation for grasping hAIP in the dynamic control of reach-to-grasp move- the object (Jeannerod et al., 1995). ments has also been confirmed in a series of TMS studies In macaques trained to grasp various objects, activity (Glover et al. 2005; Tunik et al. 2005; Rice et al. 2006). of F5 and AIP neurons show not only strong similarities, Tunik et al. (2005) have shown that applying TMS to the but also important differences (Rizzolatti et al. 1988, hAIP induces a delay in grasp adaptation, suggesting that 2002; Taira et al. 1990; Rizzolatti and Arbib 1998). On this area performs a sort of iterative comparison between one side, both F5 and AIP neurons code for reach-to- the incoming sensory information and the motor grasp actions (Murata et al. 1997, 2000). However, AIP command during the ongoing movement. neurons seem to represent the entire action, whereas F5 The quest for the human homolog of macaque F5 has neurons seem to be concerned with a particular segment identified the ventral part of the premotor cortex (PMv) of it (Rizzolatti et al. 1998; Murata et al. 2000). Another as a plausible candidate. However, neuroimaging studies important difference is that visual responses to three-di- investigating brain activity during a reach-to-grasp move- mensional objects are found more frequently in AIP than ment do not provide a coherent picture regarding the in F5 (Murata et al. 2000). This suggests that AIP, involvement of the PMv. Some fMRI studies have although part of a parieto-frontal network dedicated to reported PMv activation during multidigit visually guided hand movements, also contains a population of neurons reach-to-grasp actions (Grol et al. 2007; Cavina-Pratesi that code three-dimensional objects in visual terms. et al. 2010), object manipulation (Binkofski et al. 1999), Building upon this knowledge, Fagg and Arbib (1998) and isometric grasping (Ehrsson et al. 2001), whereas suggest that AIP could store the objects’ sensory proper- other studies found no evidence of PMv involvement dur- ties (Taira et al. 1990; Murata et al. 1997, 2000). These ing visually guided reach-to-grasp action (Culham et al. representations influence the ventral premotor area F5 2006; Begliomini et al. 2007a,b). A possible explanation and also the dorsal premotor area F2, which is involved for this controversial finding, which contrasts with the in visual guidance of the hand (Moll and Kuypers 1977; clear involvement of PMv for reach-to-grasp movements Godschalk et al. 1981; Weinrich and Wise 1982; Passing- in macaques (e.g., Rizzolatti et al. 1988), could be due to ham 1987; Rizzolatti et al. 1988; Raos et al. 2004, 2006). the fact that interspecies differences in the organization of Area F5 plays a primary role in selecting the most appro- the PMv, as well as the development of a motor speech priate type of grip on the basis of the object affordances area in humans, may have changed the location of the provided by AIP, thereby activating a motor representa- human functional homolog of monkey area F5 (Amunts tion of that object. This motor representation is then sup- and Zilles 2001). Moreover, it is worth noting that in the plied to F2, which keeps memory of it and combines it majority of studies, grasping-related activity has been iso- with visual information provided by cortical areas of the lated by subtracting activations obtained during the superior parietal lobe to continuously update the configu- reaching-only from the reach-to-grasp task (Grafton et al. ration and orientation of the hand as it approaches the 1996; Culham et al. 2003; Frey et al. 2005; Begliomini object. The final output is then sent to the F1 for motor et al. 2007a,b). Because in these studies both the reach- execution (for review see Castiello and Begliomini 2008). ing-only and the reach-to-grasp tasks required specific Moreover, the same role of F2 is played by area V6A, motor goals—triggering premotor activity—it might well which is strongly and reciprocally connected with the be that activations within premotor areas could have can- dorsal premotor cortex controlling arm movements, and celed one another when compared (Grafton et al. 1996; elaborates visual information, motion and space, for con- Culham et al. 2003; Frey et al. 2005; Begliomini et al. trolling both reaching-only and reach-to-grasp move- 2007a,b). ments (Galletti et al. 2003; Fattori et al. 2009, 2010). The dorsal part of the premotor cortex (PMd) has been In humans, both functional magnetic resonance imag- suggested as the human correspondent of macaque area ing (fMRI) and transcranial magnetic stimulation (TMS) F2 (Matelli et al. 1991). As demonstrated in macaques studies have demonstrated the existence of localized corti- (Raos et al. 2004), in humans the contribution of PMd to cal reach-to-grasp areas similar to those described in reach-to-grasp action is that of an online monitoring dur- Brain and Behavior, doi: 10.1002/brb3.412 (2 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding ing the execution phase of the action. A study comparing wide frontoparietal network adapted to both object size reach-to-grasp movements with different levels of com- and location. Furthermore, in an electroencephalogram plexity, underlined bilateral PMd involvement in associa- (EEG)/event-related potentials (ERP) study, Tarantino tion with conditions that required higher levels of et al. (2014) showed that the kinematics of reaching-only, accuracy in implementing the action (Begliomini et al. as well as the amplitude and the latency of P300 and 2007b). N400 ERP components in parietal and prefrontal sites, Although the studies reviewed above significantly con- respectively, were modulated by object size, consistent tributed to sketch an overall picture of the neural sub- with physiological findings on nonhuman primates (Fat- strates of reaching-only and reach-to-grasp in humans, a tori et al. 2012). The possibility to shed further light on crucial issue that requires further investigation is how the these issues is offered by a multivariate approach that different areas specifically contribute to the coding of exploits multivoxel pattern analysis (MVPA; e.g., Di Bono grasp type (e.g., precision grasping [PG], whole hand and Zorzi 2008; O’Toole et al. 2007; Pereira et al. 2009; grasping [WHG]) with respect to object size. This knowl- Zorzi et al. 2011). A study by Gallivan et al. (2011) edge is fundamental in order to fully define the paral- showed distinct activity patterns coding different preci- lelism between the monkeys and the human grasping sion grasping actions toward two differently sized objects network. Indeed, Rizzolatti et al. (1988; see also Rizzolatti positioned at two different spatial locations (i.e., the and Luppino 2001) showed that in monkeys, neurons smaller cube on the top of the larger one). The authors within AIP and F5 areas code for grasping actions in rela- claimed that it was possible to decode two different types tion to the type of object to be grasped. More in detail, of grasping, but it was unclear whether this result could F5 neurons seem to be mainly involved in selecting the be related to the object size or to a different direction in most appropriate motor act from a “motor vocabulary.” reaching-only toward the bottom or top object. Gallivan For instance, the act of grasping a raisin (which requires et al. (2011), also showed that voxel pattern activity the opposition of the index finger with the thumb) is within multiple frontoparietal areas during movement encoded by neurons different from those that encode the planning allowed discrimination between reach-to-grasp grasping of an apple (which requires the opposition of and reaching-only actions. More evidence against a clear the thumb with all fingers). distinction between a dorsomedial (e.g., superior parieto- In humans, fMRI studies that directly contrasted PG occipital cortex [SPOC], medial intraparietal area MIP, versus WHG using conventional analysis, revealed activa- and PMd) and a dorsolateral (e.g., hAIP and PMv) path- tion differences between the two grasping actions in con- way, specialized for reaching-only and grasping, respec- tralateral M1 (WGH > PG), bilateral PMv and hAIP tively, was provided by Fabbri et al. (2014). These recent (PG > WHG) (Ehrsson et al. 2000, 2001; Begliomini findings in humans are consistent with the theory of a et al. 2007a). More recent studies have confirmed these dorsomedial visual stream (e.g., V6A) involved in reach- findings, suggesting that grasp types (PG vs. WHG) have to-grasp actions, suggested by Galletti et al. (2003). distinct representations within a wide frontal–parietal net- Indeed, this has been documented by Fattori et al. (2009) work subserving reach-to-grasp movements (Begliomini and more directly by Fattori et al. (2010), who showed et al. 2014). This issue, however, remains controversial evidence of grasping neurons in the medial parieto-occip- given that other studies failed to detect such differences ital cortex of the macaque monkeys. The abovementioned (e.g., Kuhtz-Buschbeck et al., 2008). results about macaque area V6A suggested SPOC area as Another interesting question that requires further its putative homolog in humans (Pitzalis et al. 2013, investigation is the role of object size in both reaching- 2015; Tosoni et al. 2014). The human homolog of V6A only and reach-to-grasp actions. The visuomotor channel has been also identified as the parieto-occipital junction hypothesis of Jeannerod (1981) states that the grasping by Prado et al. (2005) and as the superior end of the action is composed of grip and transport components, parieto-occipital sulcus (sPOS) by Filimon et al. (2009). which rely on intrinsic (e.g., object size) or extrinsic (e.g., The recent findings on different aspects of reaching- location) object properties. According to this view, object only and reach-to-grasp actions call for a thorough and size and location have to be processed independently in fine-grained assessment of the reaching–grasping network separate visual channels. However, the recent neuroimag- in humans. We exploited pattern decoding methods for ing findings of Monaco et al. (2015) have suggested that, investigating the following key questions: (1) whether in humans, the cortical processing of object size and loca- there are distinct representations for different grasp types tion does not conform to a strict segregation between grip (i.e., PG vs. WHG); (2) whether there are distinct repre- and transport components of the reach-to-grasp action. sentations of object size during reaching-only action; (3) In an fMRI adaptation paradigm, the authors found that whether object size could modulate each grasp type action left aIPS showed adaptation only to object size, whereas a in a congruent/incongruent action setting (e.g., PG ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (3 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. toward a small object and WHG toward a large object an angle of ~30° and they were supported by a foam [congruent] vs. PG toward a large object and WHG wedge permitting direct viewing of the stimulus without toward a small object [incongruent]). Moreover, we mirrors. The apparatus was placed at a natural reaching aimed at: (4) replicating the findings of Gallivan et al. distance (~15 cm) above the participant’s pelvis for (2011) and Fabbri et al. (2014), which provided evidence avoiding further movements of the upper part of the of distinct representations for reaching-only and reach-to- trunk. grasp actions, distributed across a wide frontoparietal net- work; (5) replicating the findings of Monaco et al. (2015) Stimuli and task procedures on the representation of the object size during reach-to- grasp actions. The stimuli consisted of two spherical plastic objects of To address these issues, we reanalyzed the fMRI data of different dimensions (small stimulus: 3 cm diameter; large Begliomini et al. (2007b) using MVPA for investigating stimulus: 6 cm diameter). Participants were requested to the specific contribution of each brain area belonging to perform three different actions toward either the small or the reaching–grasping network in humans. To this end, the large stimulus: (1) grasping the stimulus with a PG; we selected anatomically defined regions of interest (2) grasping the stimulus with a WHG; (3) only reach the (ROIs) within a wide frontoparietal network involved in stimulus (R), by touching it with the hand knuckles, reaching-only and reach-to-grasp action representation maintaining the hand closed like in a fist. Participants (e.g., Gallivan et al. 2011; Fabbri et al. 2014). We then were informed about the type of movement to perform trained a support vector machine (SVM) classifier (see through a sound delivered by pneumatic MR-compatible Pereira et al. 2009, for a tutorial overview) with linear headphones: (1) PG—low tone (duration: 200 msec; fre- kernel on the voxel pattern activity of those ROIs for quency: 1.7 kHz); (2) WHG—high tone (duration: decoding (1) object size in both reach-to-grasp and reach- 200 msec; frequency: 210 Hz); R-double tone (duration: ing-only actions, (2) grasp type, (3) the congruence 70 msec each, staggered by a 60 msec silence period; fre- between grasp type and object size, and (4) the action quency: 445 Hz) and they were instructed to start their type (i.e., reach-to-grasp vs. reaching-only actions). action toward the stimulus only when the sound was delivered. Materials and Methods Experimental design Participants The experiment was conducted by using an event-related Nineteen right-handed participants (12 female; 19– design. inter stimulus interval (ISI) varied from 3 to 8 sec 30 years old) participated in the experiment. All gave writ- with a “long exponential” probability distribution (Hag- ten informed consent before entering in the scanner room. berg et al. 2001). ISIs distribution was fully randomized According to Begliomini et al. (2007b), three participants across trials in each run for each subject. Action toward were not included in the analysis due to the presence of the stimulus (PG, WHG, R) and stimulus dimension head motion. The cut-off used for motion correction tol- (small or large) were manipulated as to create six differ- erance was the size of the voxel (3.3 9 3.3 9 3 mm). In ent conditions (see Fig. 1): (1) “PG toward the small other words, if motion exceeded these measures in transla- object” (PGS); (2) “PG toward a large object” (PGL); (3) tion and/or rotation, the participant was not included in “WHG toward a large object” (WHGL); (4) “WHG the analysis. All participants were right-handed as mea- toward a small object” (WHGS); (5) “reaching-only sured by the Edinburgh Handedness Inventory (Oldfield toward a small object” (RS); (6) “reaching-only toward a 1971). The experimental procedures were approved by the large object” (RL). There were 45 trials for each experi- ethics committee of the University of Padua (see Beglio- mental condition, grouped into mini-blocks of five trials mini et al. 2007b, for all details). belonging to the same condition. Trials were divided in four runs, with a short rest between each run. In the odd runs the object was small, whereas in the even runs the Apparatus object was large. Participants were requested to perform either reaching- only or reach-to-grasp actions toward stimuli presented Imaging parameters by using a metal-free apparatus, which was composed of a table mounted on a plexiglass structure that allowed the Images were acquired with a whole-body 3T scanner (Sie- presentation of real 3D stimuli to participants lying mens Magnetom Trio, TIM system, Siemens, Erlangen, supine in the scanner. Participants had their head tilted at Germany) equipped with a standard Siemens 12 channels Brain and Behavior, doi: 10.1002/brb3.412 (4 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Figure 1. Experimental conditions (adapted from Begliomini et al. 2007b). Participants viewed one of the two stimuli (i.e., a spherical object of two different sizes) and performed three different tasks (i.e., reaching-only and two types of reach-to-grasp actions). The experimental conditions involved either precision grasp (PG), whole hand grasp (WHG), or Reaching-only (R) actions. Participants were instructed about the movement to perform (PG, WHG, and R) with a sound delivered through headphones. According to the size of the object to be grasped, the reach-to-grasp action was defined as congruent (PG toward a small object—PGS; WHG toward a large object—WHGL) or incongruent (PG toward a large object —PGL; WHG toward a small object—WHGS). All actions had to be performed with the right hand. coil. Functional images were acquired with a gradient- rological Institute (http://www.mni.mcgill.ca/) and echo, echo-planar (EPI) T2*-weighted sequence in order distributed with the software SPM. To avoid any circular- to measure blood oxygenation level-dependent (BOLD) ity issue in ROI selection (Kriegeskorte et al. 2009), we contrast throughout the whole brain (47 contiguous axial did not rely on the functional data but selected six ROIs slices acquired with descending interleaved sequence, that were defined on purely anatomical grounds (using 64 9 64 voxels, 3.3 9 3.3 9 3 mm resolution, the SPM Anatomy toolbox; http://www.fil.ion.ucl.ac.uk/ FOV = 210 9 210 mm, flip angle = 90°,TE = 30 msec). spm/ext/#Anatomy). One additional ROI, selected on the Volumes were acquired continuously with a repetition basis of the results of Fabbri et al. (2012), was obtained time (TR) of 3 sec; 117 volumes were collected in each through a spherical image mask using the SPM Sim- single scanning run (5:51 min; four scanning runs in pleROIBuilder toolbox (http://www.fil.ion.ucl.ac.uk/spm/ total). High-resolution T1-weighted images were acquired ext/#SimpleROIBuilder). for each subject (3D MP-RAGE, 176 axial slices, data The seven ROIs were defined as follows: matrix 256 9 256, 1 mm isotropic voxels,  ROI-1: bilateral superior parieto-occipital cortex TR = 1859 msec, TE = 3.14 msec, flip angle = 22°). (SPOC) defined according to the functional study by Fabbri et al. (2012). We extracted a sphere of 8-mm radius, centered on the Talairach coordinates (SPOC Regions of interest LH: 17, 72, 37; SPOC RH: 21, 73, 31). The functional images were preprocessed using the soft-  ROI-2: bilateral superior parietal lobe (SPLap), defined ware package SPM (Wellcome Department of Imaging according to the anatomical study by Scheperjans et al. Neuroscience, University College of London, http:// (2008). We used two different subregions of SPL (la- www.fil.ion.ucl.ac.uk/spm/). For each participant, images beled as SPL 7A and SPL 7P in the Anatomy toolbox) underwent motion correction and unwarping, and each to create this anatomical mask. volume was realigned to the first volume in the series.  ROI-3: bilateral hAIP, defined according to the The mean of all functional images was then co-registered anatomical study by Choi et al. (2006) on the human IPS. We used three different subregions of the anterior to the anatomical scan, previously corrected for intensity inhomogeneity. EPI images were then normalized adopt- IPS (labeled as hIP1, hIP2, and hIP3 in the Anatomy ing the MNI152 template, supplied by the Montreal Neu- toolbox) to create this anatomical mask. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (5 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. ROI-4: bilateral Brodmann area (BA) 1/2/3ab, accord- PG vs. WHG) as input to the classifier. In order to main- ing to the anatomical studies of Geyer et al. (1999, tain sample independence for SVM training and testing, 2000) and Grefkes et al. (2001). for each mini-block (i.e., five trials from the same condi- ROI-5: bilateral primary motor cortex, defined accord- tion), we discarded the first four volumes to capture a ing to Geyer et al. (1996), but selecting only the poste- stable fMRI signal without incorporating any noise from rior part of the primary motor cortex (bilateral BA 4p) trials within the previous mini-block and then created one sample averaging the remaining volume images (e.g., to focus on the hand representation. Pereira et al. 2009). Consequently, the target condition, ROI-6: bilateral premotor area BA 6, defined according relative to each contrast, was coded in a way to have a to the anatomical study of Geyer (2003), roughly vector T {+1, 1} , where i refers to the sample corresponding to the PMd. Specifically, BA 6 is a i i =1,...,N and N is the number of samples relative to both condi- rather large area that includes not only PMd laterally, tions in the classification (e.g., N = 36 in PG vs. WHG but also supplementary motor area (SMA) and classification), in which all the samples corresponding to pre-SMA medially. ROI-7: bilateral BA 44/45, according to Amunts et al. one target condition (e.g., PG) were labeled with +1, (1999), roughly corresponding to the PMv. whereas all the other samples (e.g., WHG) with 1. Cross-validation was used to estimate the test generaliza- To test classifier performance outside our selected net- tion performance. The SVM classifier was trained on the work, we defined one additional control ROI in which no data set using a modified version of leave-one-out cross- BOLD signal was expected and then no consistent classifi- validation. At each step of the cross-validation loop, two cation performance should be possible (see Gallivan et al. samples (one for each condition) were excluded from the 2011, for a similar methodological procedure). Therefore, training set and used to test generalization performance we selected a (8 mm) cubic region outside the skull of (see Zorzi et al. 2011). Classifier accuracy, computed the brain (centroid MNI coordinates: [63, 63, 75]). across the entire cross-validation loop on the test set, was used as statistical measures of binary classification. Preprocessing Statistical analysis on the classifier After ROI extraction, the voxel time series were prepro- performance cessed through a series of commonly used steps: stan- dardization, detrending, and temporal filtering. For each Previous studies (e.g., Chen et al. 2011; Gallivan et al. participant, each of the four runs was processed sepa- 2011) showed that t-test group analysis, with respect to rately. The time series were first standardized in order to nonparametric randomization tests, is a rather conserva- have zero mean and standard deviation 1. Then, linear tive estimate of significant decoding accuracy. Therefore, trends in each time series were removed, and a high-pass we conducted a set of one-tailed t-tests, one for each filter (0.01 Hz) was applied in order to remove low ROI, on the classifier accuracy (against the chance level of frequency drift in the signal. 50%) to obtain group statistics regarding the discrimina- tion between the two conditions included in each classifi- cation. We used false discovery rate (FDR) for correcting Classifier analysis for multiple comparisons. Furthermore, for each classifi- We used SVM with linear kernel (the C parameter was cation we assessed the possible differences between ROIs fixed to 1, which is the default value) as multivoxel pat- and hemispheric asymmetries by performing an ANOVA tern classifier. We performed six classifications: (1) Object on the classifier accuracy using ROI (SPOC, SPLap, hAIP, size in Reach-to-grasp (i.e., PGS + WHGS vs. BA 1/2/3ab, BA 44/45, BA 6, BA 4p) and hemisphere (left PGL + WHGL); (2) Object size in Reaching-only (i.e., RS vs. right) as factors. Finally, to assess the sensitivity of vs. RL); (3) Grasp type (i.e., PGS + PGL vs. each ROI for each classification, we performed a repeated WHGS + WHGL); (4) Congruence between grasp type measure (RM) ANOVA on the classifier accuracy, using and object size (i.e., PGS + WHGL [Congruent] vs. classification as a within-subject factor. PGL + WHGS [Incongruent]); (5) PG vs. Reaching-only (i.e., PGS + PGL vs. RS + RL); (6) WHG vs. Reaching- Results only (i.e., WHGS + WHGL vs. RS + RL). For each partic- ipant, we trained a linear classifier on the voxels within In this section we report, for each classification (i.e., Object each selected ROI, separately for each hemisphere. We size in reach-to-grasp action, Object size in reaching-only used only the fMRI volumes corresponding to the experi- action, Grasp Type, Congruence, PG vs. Reaching-only, mental conditions for each classification (e.g., grasp type: WHG vs. Reaching-only) the results obtained by training Brain and Behavior, doi: 10.1002/brb3.412 (6 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding linear SVM classifiers on each selected ROI, separately for bilateral SPOC, right hAIP, and the control ROI (see the left and the right hemisphere. For each ROI, the results Table 1; Fig. 2, panel B), revealing an hemispheric asym- are expressed in terms of classification performance on the metry for the hAIP. test set. To investigate possible interhemispheric asymmetries for the ROIs, we performed RM-ANOVA on the classifier accuracy using ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA Object size in reach-to-grasp action 44/45, BA 4p, and BA 6) and hemisphere (left vs. right) Independently from the grasp type, it was not possible to as within-subject factors. The analysis revealed a main discriminate between grasping a small and large object effect of ROI (F(6, 90) = 4.03, P = 0.001, g = 0.21) and from all the selected ROIs in both hemispheres, Control hemisphere (F(1, 15) = 8.12, P = 0.012, g = 0.35). The ROI included (mean accuracy = 0.47  0.02 SEM, all two-way interaction was not significant (F = 1.49). ts < 0.59). Decoding accuracy was higher when decoding from the left (M = 0.59  0.02 SEM) than from the right (M = 0.57  0.02 SEM) hemisphere. Paired t-tests (FDR Object size in reaching-only action corrected, corrected a = 0.007) showed higher decoding It was not possible to discriminate between reaching-only accuracy in the somatosensory cortex (BA 1/2/3 ab) a small and large object from all the left and right selected (M = 0.63  0.02 SEM) with respect to SPOC (M = ROIs, Control ROI included (mean accu- 0.54  0.02 SEM, t(15) = 3.73, P = 0.002), hAIP (M = racy = 0.53  0.03 SEM, all ts < 2.3). 0.54  0.02 SEM, t(15) = 4.78, P < 0.001), and the selected motor areas (BA 4p) (M = 0.56  0.02 SEM, t (15) = 4.2, P = 0.001) (see Fig. 3, panel A). No further Grasp type significant results were observed. Results for grasp type classification are summarized in Table 1. Congruence The classifier analyses showed that it was possible to linearly decode the type of grasp from the voxel pattern From none of the left and right selected ROIs (Control activity of all the selected ROIs with the exception of ROI included), it was possible to discriminate between congruent and incongruent conditions (mean accu- racy = 0.48  0.02 SEM, all ts < 0.78). Table 1. Grasp type classification. Results obtained by training linear SVM classifiers on each selected ROI, separately for the left and the Precision grasping versus reaching right hemisphere. For each ROI, the results are expressed in terms of classification performance on the test set (M  1 SEM) and the t Results for reach-to-grasp using PG versus reaching-only statistics for assessing classification significance. classification are summarized in Table 2. The classifier analyses showed that, independently from ROI Left hemisphere Right hemisphere the object size, it was possible to linearly discriminate SPOC .52  .03 .55  .03 between PG and Reaching-only from the voxel pattern t(15) = 0.75, ns t(15) = 1.49, ns activity of all the selected ROIs with the exception of the SPLap .61  .02 .54  .03 control ROI (see Table 2; Fig. 2, panel C). t(15) = 4.55, P < .001 t(15) = 3.55, P < .01 hAIP .57  .03 .51  .02 To investigate possible interhemispheric asymmetries, t(15) = 2.32, P = .017 t(15) = .48, ns we performed an RM-ANOVA on the classifier accuracy BA 1/2/3ab .67  .02 .59  .02 using ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA 44/45, t(15) = 7.31, P < .0001 t(15) = 3.92, P < .0001 BA 4p, and BA 6) and hemisphere (left vs. right) as BA 4p .56  .02 .56  .02 within-subject factors. The analysis revealed a main effect t = 2.4, P = .015 t(15) = 2.54, P = .015 of ROI (F(6, 90) = 9.38, P = 0.001, g = 0.39) and hemi- BA 6 .6  .2 .56  .03 sphere (F(1, 15) = 7.71, P = 0.014, g = 0.34). The two- t(15) = 2.95, P < .001 t(15) = 2.14, P = .025 p BA 44/45 .58  .03 .57  .03 way interaction was not significant (F = 1.31). Indepen- t(15) = 2.99, P < .005 t(15) = 2.42, P= .015 dently from the selected ROI, classifier accuracy was higher Control ROI .5  .02, when decoding from the left (M = 0.67  0.01) than from t = .17, ns the right (M = 0.63  0.01) hemisphere. Paired t-tests (FDR corrected, corrected a = 0.031) showed that decod- SVM, support vector machine; ROI, regions of interest; SPOC, superior ing accuracy from BA 1/2/3ap (M = 0.73  0.02 SEM) parieto-occipital cortex; SPLap, superior parietal lobe; BA, Brodmann area; hAIP, anterior part of the human intraparietal sulcus. and BA 6 (M = 0.72  0.02 SEM) was higher with ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (7 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. (A) (B) (C) (D) Figure 2. (A) Regions of interest (ROIs) used in the multivariate classifier analyses, transparently superimposed on top, lateral and mesial view of a standard template using BrainNet Viewer (http://www.nitrc.org/projects/bnv/) (Xia et al. 2013). ROI-1 (yellow) includes SPOC areas (Fabbri et al. 2012). ROI-2 (violet) includes SPLap areas (Scheperjans et al. 2008). ROI-3 (red) includes three subregions in the hAIP (Choi et al. 2006). ROI-4 (pink) includes BA 1/2/3ab (Geyer et al. 1999, 2000; Grefkes et al. 2001). ROI-5 (blue) includes the posterior part of the BA 4 (Geyer et al. 1996). ROI-6 (green) includes BA 6 (Geyer 2003). ROI-7 (orange) includes BA 44/45 (Amunts et al. 1999). (B) Mean linear SVM classification accuracy for grasp type decoding as a function of the involved ROIs in the left (L) and right (R) hemisphere. (C) Mean linear SVM classification performance for discriminating (independently from the object size) between PG and Reaching-only conditions as a function of the involved ROIs in each hemisphere. (D) Mean linear SVM classification performance for discriminating (independently from the object size) between WHG and Reaching-only conditions as a function of the involved ROIs in each hemisphere. Error bars indicate one standard error of the mean. Asterisks assess statistical significance with one-tailed t tests across subjects with respect to 50% (significance levels: *P < .05; **P < .01; ***P < .001 ). respect to all the other ROIs (all ts ≥ 4.5, all Ps < 0.015). 45 (M = 0.63  0.03 SEM, t = 1.6) (see Fig. 3, panel B). Furthermore, higher accuracy was observed when decod- No further significant results were observed. ing from SPLap (M = 0.65  0.02 SEM) with respect to hAIP (M = 0.6  0.02 SEM, t(15) = 5.63, P < 0.0001). Whole hand grasping versus reaching Finally, decoding accuracy from SPOC areas was lower than those obtained from all of the other ROIs (all Results for reach-to-grasp using WHG versus Reaching- ts ≤ 2.8) with the exception of hAIP (t = 0.84) and BA 44/ only classification are summarized in Table 3. Brain and Behavior, doi: 10.1002/brb3.412 (8 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding (A) (B) (C) Figure 3. Results of the RM-ANOVA on the decoding accuracy. (A) Grasp type: independently from the hemisphere, decoding from somatosensory areas was significantly more accurate than from SPOC, hAIP, and BA 4p. (B) PG versus Reaching-only: independently from the hemisphere, decoding from somatosensory areas and BA 6 was significantly more accurate than from SPOC and hAIP. Moreover, decoding accuracy from voxel pattern activity of BA 6 was significantly higher than from SPLap. (C) WHG versus Reaching-only: independently from the hemisphere, decoding from somatosensory areas was significantly more accurate than from all the other ROIs. In contrast, decoding accuracy from SPOC areas was significantly lower than that from all the other ROIs. Moreover, decoding from BA 6 was significantly more accurate than from hAIP and BA 44/45. For all the three classifications, independently from the selected ROI, the decoding accuracy was significantly higher in the left (contralateral) hemisphere than in the right (ipsilateral) hemisphere (see the bottom part of each panel). Error bars indicate one standard error of the mean across subjects. Asterisks assess statistical significance levels, as reported in the Result section. The classifier analyses showed that, independently form accuracy from SPOC (M = 0.6  0.02 SEM) areas was the object size, it was possible to linearly discriminate lower than that from all of the other ROIs (all Ps < 0.002) between the WHG and the Reaching-only conditions except for BA 44/45 (M = 0.67  0.03 SEM, t = 2.27). from the voxel pattern activity of all the selected ROIs Moreover, decoding from BA 4p (M = 0.7  0.02 SEM) with the exception of the control ROI (see Table 3; Fig. 2, was more accurate that from hAIP (M = 0.66  0.02 SEM, panel D). t(15) = 2.52, P = 0.023) (see Fig. 3, panel C). No further For investigating possible interhemispheric asymmetries significant results were observed. for the selected ROIs, we performed an RM-ANOVA on the classifier accuracy using ROI (SPOC, SPLap, hAIP, BA Classification comparison 1/2/3ab, BA 44/45, BA 4p, and BA 6) and hemisphere (left vs. right) as within-subject factors. The analysis revealed a As a final step we compared the decoding accuracies main effect of ROI (F(6, 90) = 14.66, P < 0.001, among the three possible classifications (i.e., Grasp type, g = 0.49) and hemisphere (F(1, 15) = 10.96, P = 0.005, PG vs. Reaching-only, and WHG vs. Reaching-only). We g = 0.42). The two-way interaction was not significant computed a RM-ANOVA on the classifier accuracy using (F = 0.82). Independently from the selected ROI, classifier Classification (three levels) as within-subject factor, sepa- accuracy was higher when decoding from the left rately for each ROI (SPOC, SPLap, hAIP, BA 1/2/3ab, BA (M = 0.72  0.01) than from the right (M = 0.68  0.01) 44/45, BA 4p, and BA 6). The analysis revealed for all the hemisphere. Paired t-tests (FDR corrected, corrected ROIs, except for SPOC areas (F = 2.44, P = 0.1), a main a = 0.033) showed higher decoding accuracy from BA1/2/ effect of Classification (all Fs ≥ 6.58, all Ps < 0.004, all 3ap (M = 0.79  0.02 SEM) with respect to all of the other g ≥ 0.31). A significant linear contrast (all Fs ≥ 13.14, all ROIs (all ts > 3.81, all Ps < 0.002) except for the BA 6 Ps < 0.002, all g ≥ 0.47) for all the ROIs, suggests that (M = 0.77  0.02 SEM, t = 1.004). Moreover, also decod- the decoding accuracies linearly increased from the Grasp ing from BA 6 was more accurate than from all of the other type toward PG versus Reaching-only and WGH versus ROIs (all ts ≥ 2.4, all Ps ≤ 0.029). In contrast, decoding Reaching-only classifications. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (9 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Table 2. PG versus reaching classification. Results obtained by train- Discussion ing linear SVM classifiers on each selected ROI, separately for the left and the right hemisphere. For each ROI, the results are expressed in Here, we exploited the potential of MVPA for better char- terms of classification performance on the test set (M  1 SEM) and acterizing the specific contribution of brain areas belong- the t statistics for assessing classification significance. ing to the reaching–grasping network in humans. We ROI Left hemisphere Right hemisphere performed MVPA on the activation patterns detected within ROIs of a wide frontoparietal network, for investi- SPOC .58  .03 .57  .03 gating three main aspects characterizing reaching-only and t(15) = 2.72, P = .016 t(15) = 1.96, P = .034 reach-to-grasp actions: the role of object size, the grasp SPLap .67  .03 .64  .02 t(15) = 6.65, P < .001 t(15) = 9.34, P < .001 type, and the congruence between the grasp type and the hAIP .63  .03 .56  .02 object size. In addition, to better define a possible differen- t(15) = 4.41, P = .001 t(15) = 2.26, P = .039 tial contribution of grasping-related areas, we also directly BA 1/2/3ab .75  .02 .71  .02 compared reach-to-grasp and reaching-only actions. t(15) = 12.9, P < .0001 t(15) = 1.27, P < .0001 Results showed no critical role of object size in per- BA 4p .69  .03 .62  .03 forming both reach-to-grasp and reaching-only actions. It t(15) = 7.39, P < .0001 t(15) = 4.68, P < .0001 was possible, however, to discriminate between grasp BA 6 .69  .02 .62  .03 t(15) = 8.22, P < .0001 t(15) = 4.68, P = .019 types (PG and WHG) regardless of the object size from BA 44/45 .62  .03 .63  .03 activation patterns within all the selected ROIs, with the t(15) = 4.68, P = .019 t(15) = 4.02, P < .0001 exception of bilateral SPOC and right hAIP. No effects Control ROI .52  .03, were found concerning the congruence between grasp t = .85, ns type and object size. Distinctions between reach-to-grasp SVM, support vector machine; PG, precision grasping; ROI, regions of (PG and WHG separately) and reaching-only actions interest; SPOC, superior parieto-occipital cortex; SPLap, superior pari- emerged from all the selected ROIs. Overall, decoding etal lobe; BA, Brodmann area; hAIP, anterior part of the human intra- accuracy was higher in distinguishing reach-to-grasp from parietal sulcus. reaching-only than in distinguishing PG from WHG actions. In both cases the left (controlateral) hemisphere played a prominent role in terms of decoding accuracy. Table 3. WHG versus reaching classification. Results obtained by Object size in reach-to-grasp action training linear SVM classifiers on each selected ROI, separately for the left and the right hemisphere. For each ROI, the results are expressed The evidence that object size did not play a relevant role in terms of classification performance on the test set (M  1 SEM) in reach-to-grasp action is consistent with the findings of and the t statistics for assessing classification significance. the reference study by Begliomini et al. (2007b), where the GLM did not reveal a modulation of the BOLD activ- ROI Left hemisphere Right hemisphere ity induced by object size. This allows us to discard the SPOC .63  .02 .57  .03 hypothesis that object size may account for the differen- t(15) = 5.99, P < .001 t(15) = 2.64, P < .005 tial activations within key areas concerned with visuomo- SPLap .7  .02 .7  .02 tor reach-to-grasp actions. This is, however, in contrast t(15) = 1.32, P < .001 t(15) = 13.47, P < .001 with a very recent finding of Monaco et al. (2015), where hAIP .69  .03 .63  .02 t(15) = 6.42, P < .001 t(15) = 5.45, P < .001 the authors used fMRI adaptation for investigating BA 1/2/3ab .76  .02 .73  .03 whether object size and location play a significant role in t(15) = 1.27, P < .0001 t = 7.76, P < .001 reach-to-grasp actions. Specifically, left hAIP showed BA 4p .73  .03 .67  .03 adaptation effect only to object size, whereas left SPOC, t = 7.76, P < .001 t(15) = 6.004, P < .001 primary somatosensory and motor areas (S1/M1), PMd BA 6 .79  .03 .75  .03 and SMA were sensitive to both object size and location. t(15) = 11.55, P < .001 t(15) = 2.07, P < .001 This discrepancy could be ascribed to several factors. BA 44/45 .68  .03 .66  .03 t(15) = 7.09, P < .001 t(15) = 5.86, P< .001 First, the paradigm of Monaco and colleagues was specifi- Control ROI .52  .03, cally conceived to highlight adaptation phenomena. t = .77 Indeed, the systematic variation intrinsic properties (e.g., object size) of the stimulus is crucial for adaptation SVM, support vector machine; ROI, regions of interest; SPOC, superior mechanisms. In contrast, in the study of Begliomini et al. parieto-occipital cortex; SPLap, superior parietal lobe; BA, Brodmann (2007b) this aspect was manipulated in a different way area; WHG, whole hand grasping; hAIP, anterior part of the human intraparietal sulcus. (i.e., object size was kept constant within each run). Cru- Brain and Behavior, doi: 10.1002/brb3.412 (10 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding cially, participants were informed about the size of the within the human hAIP (Begliomini et al. 2007a,b), object to be grasped at the beginning of each run (i.e., whether the human hAIP contains neural populations small object in the odd runs, and large object in the even selectively involved in the coding of different grasping ones). schemata remained to be clarified. Here we demonstrate that only left hAIP can discriminate between grasp types. This result is in agreement with the study by Gallivan Object size in reaching-only action et al. (2011), the first using a decoding method for dis- The fact that no critical role of object size emerged for criminating between different types of precision grasping the reaching-only action is in contrast with recent find- (toward a small vs. a large object stacked in a top and ings by Tarantino et al. (2014). The authors registered bottom location, respectively). However, Gallivan et al. kinematic and evoked related potentials while participants (2011) did not include the right hAIP in their decoding were asked to reach-only for differently sized objects. analysis, neglecting a possible role of the ipsilateral hemi- Results showed that the kinematics of reaching-only sphere in coding for different grasp types. Their ROI action, as well as the amplitude and the latency of P300 selection procedure was relying on the results of the GLM and N400 ERP components in parietal and prefrontal group random effects voxelwise analysis. Despite the fact sites, respectively, were modulated by object size, consis- that they avoided the “double dipping” problem tent with physiological findings on nonhuman primates (Kriegeskorte et al. 2009) by performing this analysis on a (Fattori et al. 2012). The discrepancy between these and different data set that was not used for decoding analysis, our findings could rely on the better temporal resolution this ROI selection procedure suffers from the limitations provided by ERPs with respect to fMRI, and might sug- of the GLM and it implies discarding all regions that do gest that object size, or more precisely the level of accu- not show significant effects at the level of single voxel racy of the movement determined by it, could modulate analysis. However, in our study, MVPA showed that no reaching-only actions on a temporal, rather than a spatial grasp type discrimination is possible from right hAIP. basis. Critically, we did not find any involvement of SPOC areas in distinguishing PG from WHG. Evidence of the involvement of left SPOC in discriminating between two Grasp type different precision grasping comes from the study of Gal- Concerning the grasp type, here we showed that discrimi- livan et al. (2011). Because in that study also object posi- nation between PG and WHG is possible from several tion was manipulated, it is unclear whether this result is areas of our selected network. In particular, decoding due either to the object size or to a different direction in accuracy was higher within the left (contralateral) rather reaching toward the bottom or top cube. The spatial than the right (ipsilateral) hemisphere. Conventional uni- aspect is crucial since it has been demonstrated that variate analyses performed by Begliomini et al. (2007b) SPOC activity is strictly related to the transport compo- revealed only an effect of grasp type (i.e., nent of the reach-to-grasp action (Cavina-Pratesi et al. [PGS + PGL] > [WHGS + WHGL]) in the left hAIP. 2010). In contrast, our results are consistent with the This discrepancy could be ascribed to the differences study of Fabbri et al. (2014), in which the comparison between univariate and multivariate analysis and to the between PG and WHG actions toward a spherical object fact that the findings by Begliomini et al. (2007b) were of constant size did not reveal any grasp type selectivity obtained by means of a subtraction procedure (reach-to- for the left SPOC. However, Fabbri et al. (2014) focused grasp—reaching-only) which is conventionally adopted by their attention only on the left hemisphere, discarding studies focusing on visuomotor transformation compo- possible results within the right hemisphere, whereas here nents underlying grasping (Culham et al. 2003, 2006). we show that the lack of grasp type selectivity character- Here we confirmed the involvement of left AIP in coding izes both contralateral and ipsilateral SPOC. differences between the two types of grasp, also at the The contribution of the SPLap in discriminating preci- level of voxel patterns. Moreover, MVPA revealed that sion versus whole hand grasp actions is consistent with other ROIs were involved in grasp type coding (all but the findings of Fabbri et al. (2014). Specifically, SPLa bilateral SPOC and right hAIP), because activity modula- broadly corresponds to monkey ventral intraparietal area tion within the voxel patterns related to the two condi- (VIP; Mars et al. 2011) and has been reported to be sen- tions were linearly separable within each ROI. sitive to the spatial congruency between visual and tactile Evidence that neurons within hAIP can selectively code information (Duhamel et al. 1998). for different grasp types comes from neurophysiological The involvement of bilateral BA 1/2/3ab in coding the studies (Murata et al. 2000). Although there is evidence grasp type could be explained by the sensitivity to differ- for different levels of activity depending on type of grasp ent somatosensory feedback provided by the two grasping ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (11 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. actions (i.e., PG and WHG). This peculiarity is confirmed or grasped (Ehrsson et al. 2000, 2001) objects. Bilateral by the results obtained for the bilateral dorsal premotor involvement of PMv during grasping movements has been (BA 6) and motor (BA 4p) cortices: somatosensory infor- also observed in TMS studies (Davare et al. 2006, 2008) mation from the hand should be integrated with motor revealing that lesioning either the left or the right PMv commands from frontal motor areas specifying the type modifies fingertip positioning, which is a prerequisite to of movement necessary to achieve the goal of grasping grasp an object properly (Sartori et al. 2011). (Gardner et al. 2007). Previous neurophysiological data report grasp type On the basis of neurophysiological and neuroimaging specificity within M1. M1 neurons active during WHG are studies, the role of the PMd for distal forelimb move- silent during PG (e.g., Muir and Lemon 1983). Although ments is becoming increasingly established (Raos et al. in humans different levels of activity in M1 for PG and 2004; Begliomini et al. 2007b). Here we extend this litera- WHG (Ehrsson et al. 2001; Begliomini et al. 2007a) have ture by demonstrating that within the left BA 6 different been reported, different spatial distributions of activity patterns of activity associated to different grasp types are associated with different grasping schemata had yet to be evident. This is in agreement with neurophysiological demonstrated. Here we showed that the bilateral BA 4p findings showing that F2 and F5 share similar functional significantly discriminated among grasping schemata. properties and act in concert for the control of grasping Since the movement is performed with the right hand, (Raos et al. 2004, 2006). In particular, F5 would be one might have expected this functional property to be mainly devoted to grasp selection, while F2 would moni- evident solely in left BA 4p. In general, however, the con- tor hand shaping during the ongoing movement, assuring tribution of the ipsilateral hemisphere could be hidden movement accuracy. Therefore, it might well be that the when using traditional GLM analysis, since in this case the discrimination ability shown here by left BA 6 indicates a research question is based on searching where in the brain differential hand shape monitoring depending on grasp there is a significant greater BOLD activity for an experi- types. Grasp type classification was also possible within mental condition with respect to a second one. This the right BA 6: as demonstrated by previous findings, this assumption could discard the involvement of brain areas result could be explained in terms of learning new motor where the experimental manipulations produce an effect sequences or by high requirements in terms of precision at the level of activation patterns rather than at the level and coordination, independently from the hand used of single voxel activity. In contrast, MVPA is intended to (Davare et al. 2006; Begliomini et al. 2008). In this uncover whether and to what extent a brain area is coding regard, PG requires high precision in positioning the two differential voxel pattern representations for two experi- fingers on the opposite sides of the object, whereas WHG mental conditions. We found that also the ipsilateral requires coordination among phalanxes of all fingers. hemisphere has a role in representing different grasp types, Therefore, it is conceivable that the right BA 6 acts in but the decoding accuracy was significantly higher in left concert with the left BA 6, in order to fulfill the accuracy than in the right BA4p. Recent findings show that admin- and coordination requirements intrinsic to the considered istering rTMS (repetitive TMS) on ipsilateral M1 affects types of grasp (Begliomini et al. 2007b). the timing of muscle recruitment, resulting in a loss of Neurophysiological data suggest a key role for PMv in coordination during hand movement (Davare et al. 2007). selecting the most appropriate motor configuration on This phenomenon potentially occurs on the basis of recip- the basis of 3D analysis provided by AIP (Fagg and Arbib rocal connections between cortices via the corpus callosum 1998). In this respect, human neuroimaging findings have (Boroojerdi et al. 1996; Di Lazzaro et al. 1999). provided mixed results. Whereas isometric grasping tasks Overall, we showed, for the first time, that grasp type detected PMv activity (Ehrsson et al. 2001), visually could be decoded from a wide frontoparietal network in guided tasks did not (Culham et al. 2006; Begliomini both hemispheres, with the left (controlateral) hemisphere et al. 2007a,b). Therefore, it was unclear whether the playing a more informative role with respect to the right human PMv really holds a function of “motor vocabu- (ipsilateral) one. However, since participants were able to lary” similarly to macaque F5. Our results extend this lit- see their own movements, results about grasp type could erature by showing that in humans bilateral BA44/45 be also interpreted as different representations mediated exhibits a differential activation pattern in association by the vision of a different movement. with different grasp types and supports the parallelism between macaque and humans in grasp type selectivity at Congruence the level of premotor cortices (Murata et al. 1997; Carpa- neto et al. 2011). Furthermore, several functional imaging Despite the fact that in the study by Begliomini et al. studies have shown activation in both the left and right (2007b) the contrast between natural and constrained PMv when subjects manipulated (Binkofski et al. 1999) reach-to-grasp actions (i.e., our Congruence classification) Brain and Behavior, doi: 10.1002/brb3.412 (12 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding revealed a greater activation within few voxels belonging right hAIP in visuomotor reaching-only rather than to bilateral PMd and left M1, this was not the case with grasping action representation. MVPA. Thus, a difference between congruent (i.e., PGS The contribution of SPLap in reaching-only and reach- and WGHL) and incongruent (PGL and WHGS) grasping to-grasp actions was not surprising. This result is consis- actions could be revealed only in terms of univariate anal- tent with those of Fabbri et al. (2014) and it can be ysis, whereas no activation pattern within wider brain explained by the fact that this area is sensitive to the areas encoded this difference. This might stem from a direction of visual, tactile and auditory stimuli (Bremmer limit of MVPA in distinguishing patterns of activity et al., 2001). Indeed, in our experiment, participants were across a large set of voxels (i.e., large-size ROIs) when the informed on the type of action to be performed (e.g., discriminating information is encoded in a small percent- reach-to-grasp vs. reaching-only) by auditory cues. age of the input voxels. MVPA is more sensitive to dis- The fact that BA1/2/3ab, bilaterally, was involved in the tributed coding of information whereas univariate discrimination between reaching-only and reach-to-grasp analysis is more sensitive to global engagement in ongo- actions could be explained by a sensitivity to different ing tasks (Jimura and Poldrack 2012). Another possible somatosensory feedback provided by the two actions explanation for the lack of the congruence effect, could toward the object. In addition, this was indexed by higher rely on the fact that we did not apply spatial smoothing accuracy in discriminating between reaching-only and to fMRI data before MVPA. As recently shown in a study reach-to-grasp using WHG rather than PG, probably mir- based on simulated data (Stelzer et al. 2014), the com- roring a greater difference in hand configuration, and bined use of spatial smoothing and cluster based correc- hence in somatosensory feedback. tion could increase the number of false positives and false The involvement of BA44/45 in discriminating between negatives, respectively. Thus, both univariate and multi- reaching-only and reach-to-grasp actions is consistent variate approaches could introduce possible limitations, with the most recent findings of Fabbri et al. (2014) and and their combination should be more informative than Gallivan et al. (2011), both using multivariate approaches the use of a single approach (see also Gallivan et al. 2011 for analyzing fMRI data. The first study highlighted that for a similar argument). reach direction and grip type are both represented in left PMv, whereas the second one showed that left PMv was involved in the discrimination between precision grasping Reach-to-grasp versus reaching-only actions and touching, in both the planning and the execution Here, we showed that it was possible to discriminate phase of the actions. between reach-to-grasp and reaching-only actions from The contribution of bilateral BA 6 in distinguishing the selected frontoparietal network. Interestingly, a between reach-to-grasp and reaching-only actions is not prominent role in characterizing the reaching–grasping surprising, since this area has been firstly suggested to network is played by bilateral SPOC and right hAIP. code only for the transport phase of the hand toward an These areas were not sensitive in decoding grasp type, but object (i.e., reaching) (Begliomini et al. 2014; Culham played a significant role in discriminating between reach- et al. 2006; Vesia and Crawford, 2012) and has been to-grasp and reaching-only actions. shown to be involved in the representation of both the Our results on SPOC suggest that the contribution of transport and the hand preshaping components of reach- these areas might be more crucial for reaching-only than ing-only and reach-to-grasp actions, respectively (e.g., shaping the fingers for different grip types, which is con- Fabbri et al. 2014). The bilateral involvement of PMd in sistent with the findings of Cavina-Pratesi et al. (2010). coding direction and amplitude of reaching-only has been These authors reported that the human SPOC showed shown by Fabbri et al. (2012), thus it was not surprising stronger activation during reach-to-grasp action toward that different activity patterns are present in these areas far rather than near locations, suggesting a preference for for reaching-only and reach-to-grasp actions. the transport rather than the grasp component. However, Finally, the involvement of primary motor area in our results are in contrast with those reported by Fabbri reach-to-grasp versus reaching-only discrimination was et al. (2014), where left SPOC did not show any effect in expected, as well as its involvement in distinguishing finer discriminating between reach-to-grasp and reaching-only aspects of the grasping action (i.e., grasp type classifica- actions. tion). These results are consistent with the most recent Our results on right hAIP suggest that this area con- neuroimaging studies in humans (Gallivan et al. 2011; tributes to the representation of both reaching-only and Fabbri et al. 2012, 2014). Interestingly, the novelty of these reach-to-grasp actions, but it does not appear to be criti- results relies on the right (ipsilateral) contribution of cally involved in the finer distinctions between grasp BA4p. As in the case of the grasp type classification, we types. This latter result might indicate a major role of the found a bilateral involvement of BA4p in discriminating ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (13 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. between reaching-only and reach-to-grasp actions, even if a clear-cut distinction between a dorsomedial (e.g., SPOC, the left (contralateral) hemisphere played a prominent medial intraparietal area MIP, and PMd) and dorsolateral role, in terms of classification accuracy. (e.g., hAIP and PMv) pathways, specialized for reaching- In conclusion, our results showed significant hemi- only and reach-to-grasp actions, respectively, as reported spheric asymmetries in discriminating reaching-only from in a series of recent studies on human and nonhuman pri- reach-to-grasp actions and PG from WHG, which con- mates (Fattori et al. 2009, 2010; Cavina-Pratesi et al. 2010; sisted of a left (i.e., contralateral) hemisphere dominance. Monaco et al. 2011). Our results are also consistent with This is consistent to our expectations, since participants the findings of Grol et al. (2007), which argue against the were using the right hand to perform the actions. Fur- presence of dedicated cerebral circuits for reaching-only thermore, we found that somatosensory and dorsal pre- and reach-to grasp actions, suggesting that the contribu- motor areas were more responsive in distinguishing tions of the dorsolateral and the dorsomedial circuits are a between reaching-only and reach-to-grasp actions, with function of the degree of online control required by the respect to all the other areas within the selected network. movement. Finally, our results are perfectly consistent with Finally, within the selected network, decoding accuracy the theory of a dorsomedial visual stream involved in was higher when discriminating reaching-only from reach-to-grasp actions, suggested by Galletti et al. (2003) reach-to-grasp action, when using WHG rather than PG. in nonhuman primates, and well documented by Fattori This result, together with the fact that no critical role was et al. (2009, 2010). Reaching-only and reach-to-grasp played by object size, could suggest that different activa- actions could be better characterized by temporal, rather tion patterns underlying reach-to-grasp and reaching-only than spatial criteria across planning and execution stages of actions could be mainly due to a physical difference in the action, as also suggested by a recent study of Beglio- hand configuration. The fact that this information was mini et al. (2014). Here we showed that several areas of the probably guiding the discrimination within all the selected human reaching–grasping network are involved in process- network (including parietal areas) indicates that hand ing aspects related to both reach-to-grasp and reaching- preshaping begins in early stages of action planning (i.e., only actions. Crucially, the precise nature, in terms of tim- action preparation), as also suggested by Gallivan et al. ing and direction (causality—Davare et al. 2010; Grol et al. (2011) and Begliomini et al. (2014). 2007) of the relations between the involved brain areas remains to be clarified by future studies. Altogether, the findings provided by the integrated Conclusion approach adopted in this work enrich the current knowl- To summarize, in our study no critical role of object size edge regarding the functional role of key brain areas emerged for both reaching-only and reach-to-grasp involved in the cortical control of reaching-only and actions. This result runs against the hypothesis that the reach-to-grasp actions in humans, by revealing novel fine- intrinsic object properties (e.g., object size) could play a grained distinctions among action types within a wide key role in both reach-to-grasp and reaching-only actions. frontoparietal network. Here we showed, for the first time that grasp type (i.e., PG vs. WHG), independently from object size, can be reliably Acknowledgments discriminated by a linear classifier within a wide fron- This work was supported by a grant from the European toparietal network distributed across both the hemispheres, Research Council (grant no. 210922) and the University with the exception of SPOC areas and right hAIP. The left of Padova (Strategic Grant NEURAT) to M. Zorzi. (i.e., controlateral) hemisphere, however, played a crucial role in terms of decoding accuracy. No significant interac- tion between the grasp type and the object size (i.e., our Conflict of Interest congruence classification) emerged within the considered None declared. network, despite the fact that univariate analysis of the same data set (Begliomini et al. 2007b) showed that activ- References ity of few voxels within PMd and M1 areas was modulated by congruence. This highlights the importance to perform Amunts, K., and K. Zilles. 2001. Advances in cytoarchitectonic data analysis from a more comprehensive perspective, mapping of the human cerebral cortex. Neuroimaging Clin. combining both univariate and multivariate analyses. This N. Am. 11:151–169. integrated approach could provide more informative Amunts, K., A. Schleicher, U. Burgel, € H. Mohlberg, H. B. results and a deeper understanding of the neural dynamics Uylings, and K. Zilles. 1999. Broca’s region revisited: underlying the cognitive processes of interest. Finally, our cytoarchitecture and intersubject variability. J. Comp. results provided further evidence against the hypothesis of Neurol. 412:319–341. Brain and Behavior, doi: 10.1002/brb3.412 (14 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Begliomini, C., M. B. Wall, A. T. Smith, and U. Castiello. produces fMRI activation in dorsal but not ventral stream 2007a. Differential cortical activity for precision and whole- brain areas. Exp. Brain Res. 153:180–189. hand visually guided grasping in humans. Eur. J. Neurosci. Culham, J. C., C. Cavina-Pratesi, and A. Singhal. 2006. The 25:1245–1252. role of parietal cortex in visuomotor control: what have we Begliomini, C., A. Caria, W. Grodd, and U. Castiello. 2007b. learned from neuroimaging? Neuropsychologia 44:2668– Comparing natural and constrained movements: new insight 2684. into the visuomotor control of grasping. PLoS One 2:e1108. Davare, M., M. Andrei, G. Cosnard, J. L. Thonnard, and E. Begliomini, C., C. Nelini, A. Caria, W. Grodd, and U. Olivier. 2006. Dissociating the role of ventral and dorsal Castiello. 2008. Cortical activations in humans grasp-related premotor cortex in precision grasping. J. Neurosci. 26:2260– areas depend on hand used and handedness. PLoS One 3: 2268. e3388. Davare, M., J. Duque, Y. Vandermeeren, J. L. Thonnard, and Begliomini, C., T. De Sanctis, M. Marangon, V. Tarantino, L. E. Olivier. 2007. Role of the ipsilateral primary motor cortex Sartori, D. Miotto, et al. 2014. An investigation of the neural in controlling the timing of hand muscle recruitment. circuits underlying reaching and reach-to-grasp movements: Cereb. Cortex 17:353–362. from planning to execution. Front. Hum. Neurosci. 8:676. Davare, M., R. Lemon, and E. Olivier. 2008. Selective Binkofski, F., G. Buccino, S. Posse, R. J. Seitz, G. Rizzolatti, modulation of interactions between ventral premotor cortex and A. Freund. 1999. Fronto-parietal circuit for object and primary motor cortex during precision grasping in manipulation in man: evidence from an fMRI-study. Eur. J. humans. J. Physiol. 586:2735–2742. Neurosci. 11:3276–3286. Davare, M., J. C. Rothwell, and R. N. Lemon. 2010. Causal Boroojerdi, B., K. Diefenbach, and A. Ferbert. 1996. connectivity between the human anterior intraparietal area Transcallosal inhibition in cortical and subcortical cerebral and premotor cortex during grasp. Curr. Biol. 20:176–181. vascular lesions. J. Neurol. Sci. 144:160–170. Di Bono, M. G., and M. Zorzi. 2008. Decoding cognitive states Bremmer, F., A., Schlack, N. J., Shah, O., Zafiris, M., from fMRI data using support vector regression. PsychNol. Kubischik, K. P., Hoffmann, K., Zilles, and G. R. Fink. 2001. J. 6:189201. Polymodal motion processing in posterior parietal and Di Lazzaro, V., A. Oliviero, P. Profice, A. Insola, P. Mazzone, premotor cortex: a human fMRI study strongly implies P. Tonali, et al. 1999. Direct demonstration of equivalencies between humans and monkeys. Neuron interhemispheric inhibition of the human motor cortex 29:287–296. produced by transcranial magnetic stimulation. Exp. Brain Carpaneto, J., M. A. Umilta, L. Fogassi, A. Murata, V. Gallese, Res. 124:520–524. S. Micera, et al. 2011. Decoding the activity of grasping Duhamel, J. R., C. L. Colby, and M. E. Goldberg. 1998. Ventral neurons recorded from the ventral premotor area F5 of the intraparietal area of the macaque: congruent visual and macaque monkey. Neuroscience 188:80–94. somatic response properties. J. Neurophysiol. 79:126–136. Castiello, U. 2005. The neuroscience of grasping. Nat. Rev. Ehrsson, H. H., A. Fagergren, T. Jonsson, G. Westling, R. S. Neurosci. 6:726–736. Johansson, and H. Forssberg. 2000. Cortical activity in Castiello, U., and C. Begliomini. 2008. The cortical control of precision- versus power-grip tasks: an fMRI study. J. visually guided grasping. Neuroscientist 14:157–170. Neurophysiol. 83:528–536. Cavina-Pratesi, C., M. Goodale, and J. C. Culham. 2007. FMRI Ehrsson, H. H., E. Fagergren, and H. Forssberg. 2001. reveals a dissociation between grasping and perceiving the Differential fronto-parietal activation depending on force size of real 3D objects. PLoS One 5:1–14. used in a precision grip task: an fMRI Study. J. Cavina-Pratesi, C., S. Monaco, P. Fattori, C. Galletti, T. D. Neurophysiol. 85:2613–2623. McAdam, D. J. Quinlan, et al. 2010. Functional magnetic Fabbri, S., A. Caramazza, and A. Lingnau. 2012. Distributed resonance imaging reveals the neural substrates of arm sensitivity for movement amplitude in directionally tuned transport and grip formation in reach-to-grasp actions in neuronal populations. J. Neurophysiol. 107:1845–1856. humans. J. Neurosci. 30:10306–10323. Fabbri, S., L. Strnad, A. Caramazza, and A. Lingnau. 2014. Chen, Y., P. Namburi, L. T. Elliott, J. Heinzle, C. S. Soon, M. Overlapping representations for grip type and reach W. Chee, et al. 2011. Cortical surface-based searchlight direction. NeuroImage 94:138–146. decoding. NeuroImage 56:582–592. Fagg, A. H., and M. A. Arbib. 1998. Modeling parietal- Choi, H. J., K. Zilles, H. Mohlberg, A. Schleicher, G. R. Fink, premotor interactions in primate control of grasping. Neural E. Armstrong, et al. 2006. Cytoarchitectonic identification Netw. 11:1277–1303. and probabilistic mapping of two distinct areas within the Fattori, P., R. Breveglieri, N. Marzocchi, D. Filippini, A. anterior ventral bank of the human intraparietal sulcus. J. Bosco, and C. Galletti. 2009. Hand orientation during Comp. Neurol. 495:53–69. reach-to-grasp movements modulates neuronal activity in Culham, J. C., S. L. Danckert, J. F. DeSouza, J. S. Gati, R. S. the medial posterior parietal area V6A. J. Neurosci. Menon, and M. A. Goodale. 2003. Visually guided grasping 29:1928–1936. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (15 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Fattori, P., V. Raos, R. Breveglieri, A. Bosco, N. Marzocchi, Grafton, S. T., M. A. Arbib, L. Fadiga, and G. Rizzolatti. 1996. and C. Galletti. 2010. The dorsomedial pathway is not just Localization of grasp representations in humans by positron for reaching: grasping neurons in the medial parieto- emission tomography. Exp. Brain Res. 112:103–111. occipital cortex of the macaque monkey. J. Neurosci. Grefkes, C., S. Geyer, T. Schormann, P. Roland, and K. Zilles. 30:342–349. 2001. Human somatosensory area 2: observer-independent Fattori, P., R. Breveglieri, V. Raos, A. Bosco, and C. Galletti. cytoarchitectonic mapping, interindividual variability, and 2012. Vision for action in the macaque medial posterior population map. NeuroImage 14:617–631. parietal cortex. J. Neurosci. 32:3221–3234. Grol, M. J., J. Majdandzic, K. E. Stephan, L. Verhagen, H. C. Filimon, F. 2010. Human cortical control of hand movements: Dijkerman, H. Bekkering, et al. 2007. Parieto-frontal parietofrontal networks for reaching, grasping, and pointing. connectivity during visually guided grasping. J. Neurosci. Neuroscientist 16:388–407. 27:11877–11887. Filimon, F., J. D. Nelson, R. S. Huang, and M. I. Sereno. 2009. Hagberg, G. E., G. Zito, F. Patria, and J. N. Sanes. 2001. Multiple parietal reach regions in humans: cortical Improved detection of event-related functional MRI signals representations for visual and proprioceptive feedback using probability functions. NeuroImage 14:1193–1205. during on-line reaching. J. Neurosci. 29:2961–2971. Hinkley, L. B., L. A. Krubitzer, J. Padberg, and E. A. Disbrow. Frey, H. S., D. Vinton, R. Norlund, and S. T. Grafton. 2005. 2009. Visual-manual exploration and posterior parietal Cortical topography of human anterior intraparietal cortex cortex in humans. J. Neurophysiol. 102:3433–3446. active during visually guided grasping. Cogn. Brain Res. Jeannerod, M. 1981. Specialized channels for cognitive 23:397–405. responses. Cognition 10:135–137. Galletti, C., D. F. Kutz, M. Gamberini, R. Breveglieri, and P. Jeannerod, M., M. A., Arbib, G., Rizzolatti, and H., Sakata. Fattori. 2003. Role of the medial parieto-occipital cortex in 1995. Grasping objects: the cortical mechanisms of the control of reaching and grasping movements. Exp. Brain visuomotor transformation. Trends Neurosci 18:314–320. Res. 153:158–170. Jimura, K., and R. A. Poldrack. 2012. Analyses of regional- Gallivan, J. P., D. A. McLean, K. F. Valyear, C. E. Pettypiece, average activation and multivoxel pattern information tell and J. C. Culham. 2011. Decoding action intentions from complementary stories. Neuropsychologia 50:544–552. preparatory brain activity in human parieto-frontal Kriegeskorte, N., W. K. Simmons, P. S. F. Bellgowan, and C. I. networks. J. Neurosci. 31:9599–9610. Baker. 2009. Circular analysis in systems neuroscience: the Gardner, E. P., K. S. Babu, S. D. Reitzen, S. Ghosh, A. S. dangers of double dipping. Nat. Neurosci. 12:535–540. Brown, J. Chen, et al. 2007. Neurophysiology of prehension. Kroliczak, G., C. Cavina-Pratesi, D. A. Goodman, and J. C. I. Posterior parietal cortex and object-oriented hand Culham. 2007. What does the brain do when you fake it? behaviors. J. Neurophysiol. 97:387–406. An FMRI study of pantomimed and real grasping. J. Geyer, S. 2003. Brodmanns areas. Pp. 482–496 in M. J. Neurophysiol. 97:2410–2422. Aminoff and R. B. Daroff, eds. Encyclopedia of the Kuhtz-Buschbeck, J. P., R., Gilster, S., Wolff, S., Ulmer, H., neurological sciences. Academic Press, San Diego. Siebner, and O. Jansen. 2008. Brain activity is similar during Geyer, S., A. Ledberg, A. Schleicher, S. Kinomura, T. precision and power gripping with light force: an fMRI Schormann, U. Burgel, € et al. 1996. Two different areas study. Neuroimage 40:1469–1481. within the primary motor cortex of man. Nature 382:805– Lawrence, D. G., and D. A. Hopkins. 1976. The development 807. of motor control in the rhesus monkey: evidence concerning Geyer, S., A. Schleicher, and K. Zilles. 1999. Areas 3a, 3b, and the role of corticomotoneuronal connections. Brain 99:235– 1 of human primary somatosensory cortex: 1. 254. Microstructural organization and interindividual variability. Luppino, G., A. Murata, P. Govoni, and M. Matelli. 1999. NeuroImage 10:63–83. Largely segregated parietofrontal connections linking rostral Geyer, S., T. Schormann, H. Mohlberg, and K. Zilles. 2000. intraparietal cortex (areas AIP and VIP) and the ventral Areas 3a, 3b, and 1 of human primary somatosensory premotor cortex (areas F5 and F4). Exp. Brain Res. cortex: 2. Spatial normalization to standard anatomical 128:181–187. space. NeuroImage 11:684–696. Mars, R. B., S. Jbabdi, J. Sallet, J. X. O’ Reilly, P. L. Glover, S., R. C. Miall, and M. F. S. Rushworth. 2005. Parietal Croxson, E. Olivier, et al. 2011. Diffusion-weighted rTMS disrupts the initiation but not the execution of on- imaging tractography-based parcellation of the human line adjustments to a perturbation of object size. J. Cogn. parietal cortex and comparison with human and macaque Neurosci. 17:124–136. resting-state functional connectivity. J. Neurosci. Godschalk, M., R. N. Lemon, H. G. Nijs, and H. G. Kuypers. 31:4087–4100. 1981. Behaviour of neurons in monkey peri-arcuate and Matelli, M., and G. Luppino. 2001. Parietofrontal circuits for precentral cortex before and during visually guided arm and action and space perception in the macaque monkey. hand movements. Exp. Brain Res. 44:113–116. NeuroImage 14:S27–S32. Brain and Behavior, doi: 10.1002/brb3.412 (16 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. M. G. Di Bono et al. Multivoxel Pattern Decoding Matelli, M., G. Luppino, and G. Rizzolatti. 1985. Patterns of Raos, V., M. A. Umilta, A. Murata, L. Fogassi, and V. Gallese. cytochrome oxidase activity in the frontal agranular cortex 2006. Functional properties of grasping-related neurons in of the macaque monkey. Behav. Brain Res. 18:125–136. the ventral premotor area F5 of the macaque monkey. J. Matelli, M., G. Luppino, and G. Rizzolatti. 1991. Architecture of Neurophysiol. 95:709–729. superior and mesial area 6 and the adjacent cingulate cortex Rice, N. J., E. Tunik, and S. T. Grafton. 2006. The anterior in the macaque monkey. J. Comp. Neurol. 311:445–462. intraparietal sulcus mediates grasp execution, independent Moll, L., and H. G. Kuypers. 1977. Premotor cortical ablations of requirement to update: new insights from transcranial in monkeys: contralateral changes in visually guided magnetic stimulation. J. Neurosci. 26:8176–8182. reaching behavior. Science 198:317–319. Rizzolatti, G., and M. A. Arbib. 1998. Language within our Monaco, S., C. Cavina-Pratesi, A. Sedda, P. Fattori, C. Galletti, grasp. Trends Neurosci. 21:188–194. and J. C. Culham. 2011. Functional magnetic resonance Rizzolatti, G., and G. Luppino. 2001. The cortical motor adaptation reveals the involvement of the dorsomedial system. Neuron 31:889–901. stream in hand orientation for grasping. J. Neurophysiol. Rizzolatti, G., L. Camarda, L. Fogassi, M. Gentilucci, G. 106:2248–2263. Luppino, and M. Matelli. 1988. Functional organization of Monaco, S., A. Sedda, C. Cavina-Pratesi, and J. C. Culham. inferior area 6 in the macaque monkey. Exp. Brain Res. 2015. Neural correlates of object size and object location 71:491–507. during grasping actions. Eur. J. Neurosci. 41:454–465. Rizzolatti, G., G. Luppino, and M. Matelli. 1998. The Muir, R. B., and R. N. Lemon. 1983. Corticospinal neurons organisation of the cortical motor system: new concepts. with a special role in precision grip. Brain Res. 261:312–316. Electroencephalogr. Clin. Neurophysiol. 106:283–296. Murata, A., L. Fadiga, L. Fogassi, V. Gallese, V. Raos, and G. Rizzolatti, G., L. Fogassi, and V. Gallese. 2002. Motor and Rizzolatti. 1997. Object representation in the ventral cognitive functions of the ventral premotor cortex. Curr. premotor cortex (area F5) of the monkey. J. Neurophysiol. Opin. Neurobiol. 12:149–154. 78:2226–2230. Sartori, L., E. Straulino, and U. Castiello. 2011. How objects Murata, A., V. Gallese, G. Luppino, M. Kaseda, and H. Sakata. are grasped: the interplay between affordances and end- 2000. Selectivity for the shape, size and orientation of goals. PLoS One 6:e25203. objects for grasping in neurons of monkey parietal area AIP. Scheperjans, F., K. Hermann, S. B. Eickhoff, K. Amunts, A. J. Neurophysiol. 83:2580–2601. Schleicher, and K. Zilles. 2008. Observer-independent Oldfield, R. C. 1971. The assessment and analysis of cytoarchitectonic mapping of the human superior parietal handedness: the Edinburgh Inventory. Neuropsychologia cortex. Cereb. Cortex 18:846–867. 9:97–113. Stelzer, J., G. Lohmann, K. Mueller, T. Buschmann, and R. O’Toole, A. J., F. Jiang, H. Abdi, N. Penard, J. P. Dunlop, and Turner. 2014. Deficient approaches to human neuroimaging. M. A. Parent. 2007. Theoretical, statistical, and practical Front. Hum. Neurosci. 8:462. perspectives on pattern-based classification approaches to Taira, M., S. Mine, A. P. Georgopoulos, A. Murata, and H. the analysis of functional neuroimaging data. J. Cogn. Sakata. 1990. Parietal cortex neurons of the monkey related Neurosci. 19:1735–1752. to the visual guidance of hand movement. Exp. Brain Res. Passingham, R. E. 1987. Two cortical systems for directing 83:29–36. movement. Ciba Found. Symp. 132:151–164. Tarantino, V., T. De Sanctis, E. Straulino, C. Begliomini, and Pereira, F., T. Mitchell, and M. Botvinick. 2009. Machine U. Castiello. 2014. Object size modulates fronto-parietal learning classifiers and fMRI: a tutorial overview. activity during reaching movements. Eur. J. Neurosci. NeuroImage 45:S199–S209. 39:1528–1537. Pitzalis, S., M. I. Sereno, G. Committeri, P. Fattori, G. Galati, Tosoni, A., S. Pitzalis, G. Committeri, P. Fattori, C. Galletti, A. Tosoni, et al. 2013. The human homologue of macaque and G. Galati. 2014. Resting-state connectivity and area V6A. NeuroImage 82:517–530. functional specialization in human medial parieto-occipital Pitzalis, S., P. Fattori, and C. Galletti. 2015. The human cortex. Brain Struct. Funct. 220:3307–3321. cortical areas V6 and V6A. Vis. Neurosci. 32:E007. Tunik, E., S. H. Frey, and S. T. Grafton. 2005. Virtual lesions Prado, J., S. Clavagnier, H. Otzenberger, C. Scheiber, H. of the anterior intraparietal area disrupt goal-dependent on- Kennedy, and M. T. Perenin. 2005. Two cortical systems for line adjustment of grasp. Nat. Neurosci. 8:505–511. reaching in central and peripheral vision. Neuron 48:849– Tunik, E., N. J. Rice, A. Hamilton, and S. T. Grafton. 2007. 858. Beyond grasping: representation of action in human anterior Raos, V., M. A. Umilta, V. Gallese, and L. Fogassi. 2004. intraparietal sulcus. NeuroImage 36:T77–T86. Functional properties of grasping-related neurons in the Vesia, M., and J. D. Crawford. 2012. Specialization of reach dorsal premotor area F2 of the macaque monkey. J. function in human posterior parietal cortex. Exp Brain Res. Neurophysiol. 92:1990–2002. 221:1–18. ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. Brain and Behavior, doi: 10.1002/brb3.412 (17 of 18) Multivoxel Pattern Decoding M. G. Di Bono et al. Weinrich, M., and S. P. Wise. 1982. The premotor cortex of Zorzi, M., M. G. Di Bono, and W. Fias. 2011. Distinct the monkey. J. Neurosci. 2:1329–1345. representations of numerical and non-numerical order in Xia, M., J. Wang, and Y. He. 2013. BrainNet Viewer: a the human intraparietal sulcus revealed by multivariate network visualization tool for human brain connectomics. pattern recognition. NeuroImage 56:674–680. PLoS One 8:e68910. Brain and Behavior, doi: 10.1002/brb3.412 (18 of 18) ª 2015 The Authors. Brain and Behavior published by Wiley Periodicals, Inc.

Journal

Brain and Behavior – Wiley

Published: Nov 1, 2015

Keywords: ; ; ;

There are no references for this article.