The representational space of observed actions

  1. Raffaele Tucciarelli
  2. Moritz Wurm
  3. Elisa Baccolo
  4. Angelika Lingnau  Is a corresponding author
  1. Royal Holloway University of London, United Kingdom
  2. University of Trento, Italy
  3. University of Regensburg, Germany
9 figures and 4 additional files

Figures

Figure 1 with 1 supplement
Stimuli and behavioral task.

(A) Stimuli used in the behavioral and the fMRI experiment, depicting static images of 28 everyday actions. To increase perceptual variability, actions were shown from different viewpoints, in …

Figure 1—figure supplement 1
All images used in the fMRI experiment.

Different actions are depicted in rows (for corresponding labels, see words printed in bold font in Supplementary file 3). Different exemplars for each action are shown in the columns. Exemplars …

Figure 2 with 1 supplement
Behavioral Representational Dissimilarity Matrix for the semantic model.

Bluish colors indicate high similarity between pairwise combinations of actions, whereas reddish colors indicate high dissimilarity.

Figure 2—figure supplement 1
Pairwise cross-correlation matrix across models.

For each participant, we computed the pairwise cross-correlation matrix across models. We then computed (A) the averaged correlation values across participants; and reported (B) the maximum and (C) …

Figure 3 with 4 supplements
Cluster analysis.

Clusters resulting from the K-means clustering analysis for the semantic task. The 2D-plot shows component 1 and 2, the corresponding labels of individual actions and the suggested labels for the …

Figure 3—figure supplement 1
Results silhouette analysis.

Silhouette analysis revealed that the optimal number of clusters for the Semantic model was 6. The analysis was performed using the fviz_nbclust function of the R package ‘factoextra’. The …

Figure 3—figure supplement 2
Eigenvalues obtained from the PCA of the semantic model.

The PCA shows that the first three components accounted for the largest amount of variance.

Figure 3—figure supplement 3
Results cluster analysis.

Clusters resulting from the K-means clustering analysis for the semantic model. (A) 2D-plot showing component 1 and 2, corresponding labels of individual actions and suggested labels for the …

Figure 3—figure supplement 4
First two principal components of the control models.

Clusters were distinguished using K-means clustering as implemented in R. Number of clusters used as input were 6 (based on the Silhouette analysis of the semantic model; see Figure 3—figure …

Standard RSA, Semantic model.

Group results of the searchlight-based RSA using the semantic model (standard RSA, that is second order correlation between neural data and behavioral model). Statistical maps only show the positive …

Model Representational Dissimilarity Matrices.

Group representational dissimilarity matrices for body, scene, movement, and object model, derived from inverse MDS carried out in the same group of participants after the fMRI experiment. The …

Figure 6 with 2 supplements
Multiple regression RSA (semantic model).

Group results of the searchlight-based multiple regression RSA, in which the ten different models (see Materials and methods, section Construction of Representational Dissimilarity Matrices, for …

Figure 6—figure supplement 1
Multiple regression based RSA results, together with Glasser parcellation.

Significant clusters resulting from the searchlight multiple-regression RSA (see Figures 6 and 7), with Glasser brain parcellation (Glasser et al., 2016) superimposed on the statistical maps. The …

Figure 6—figure supplement 2
Results for the body model, together with EBA coordinates.

Cluster identified by the multiple regression-based RSA using the body model, with coordinates of the extrastriate body area (EBA) reported by a number of studies that used a functional localizer to …

Searchlight-based multiple regression RSA.

Searchlight-based multiple regression RSA results for the body (A), movement (B), object (C), transitivity (D), distance (E), 1 vs 2 people (F) and the HMAX-C1 (G) model. The resulting beta …

Neural RDM from LOTC.

(A) Kendall rank correlation coefficient between neural and model RDMs in the cluster in the LOTC revealed by the multiple regression RSA using the semantic model (see Figure 6). Error bars depict …

Author response image 1
Comparison of the results of the standard RSA using Spearman correlation (top panel) and squared Euclidian distance (right panel) as distance metric.

Additional files

Supplementary file 1

Cluster table standard RSA (semantic model).

Cluster table standard RSA (semantic model). List of clusters resulting from the standard RSA for the semantic model which survived correction for multiple comparisons (cluster p-value<0.05; see Materials and methods and Figure 4). Coordinates are in MNI space. Labels are based on MRI scans that originated from the OASIS project (http://www.oasis-brains.org/) and were provided by Neuromorphometrics, Inc (http://www.neuromorphometrics.com/) under academic subscription provided in SPM12 and Glasser’s surface-based atlas (Glasser et al., 2016).

https://cdn.elifesciences.org/articles/47686/elife-47686-supp1-v1.docx
Supplementary file 2

Cluster table multiple regression RSA.

Cluster table multiple regression RSA. List of clusters resulting from the multiple regression RSA for the eight different models (semantic, body, movement, object, transitivity, distance, 1 vs 2 people, HMAX-C1) which survived correction for multiple comparisons (cluster p-value<0.05; see Materials and methods and Figure 6 and Figure 6—figure supplement 1). Coordinates are in MNI space. Labels are based on MRI scans that originated from the OASIS project (http://www.oasis-brains.org/) and were provided by Neuromorphometrics, Inc (http://www.neuromorphometrics.com/) under academic subscription provided in SPM12 and Glasser’s surface-based atlas (Glasser et al., 2016).

https://cdn.elifesciences.org/articles/47686/elife-47686-supp2-v1.docx
Supplementary file 3

List of actions identified using the online survey.

We initially kept actions that were mentioned by at least 20% of the participants. The final selected twenty-eight actions used for the study are the ones highlighted. See Materials and methods section for the procedure used to select these actions.

https://cdn.elifesciences.org/articles/47686/elife-47686-supp3-v1.docx
Transparent reporting form
https://cdn.elifesciences.org/articles/47686/elife-47686-transrepform-v1.pdf

Download links