<h3> SUPPLEMENTAL MATERIALS </h3> <br> *All supplemental materials are available to be downloaded individually on the Files page* ---------- <h4> MODEL IMPLEMENTATION: </h4> - Behavioral data (in .mat format) used for the main analysis - Behavioral data (in .mat format) used for the independent analysis - MATLAB code (in .zip format) with the scripts used for model fitting - Output file with results (in .mat format) for the three critical analyses fit on the full dataset - Output file with results (in .mat format) for the three critical analyses fit on the restricted dataset - Output file with results (in .mat format) for the three critical analyses fit on the independent dataset <br> - GloVe vectors are available for download here: https://nlp.stanford.edu/projects/glove/ - word2vec vectors are available for download here: https://code.google.com/archive/p/word2vec/ <br> ***NOTE:** Use the unpack_results.R function to convert the .mat data structure to a list element in R* ---------- <h4> TABLES: </h4> - **Table SI-1a**: Pearson r coefficient values for the linear correlation analysis between the predicted and observed responses calculated on the full dataset. Values in bold indicate a significant positive correlation (p < 0.05). - **Table SI-1b**: Pearson r coefficient values for the linear correlation analysis between the predicted and observed responses calculated on the restricted dataset that excluded the words used in the construction of the composite vector. Values in bold indicate significant results (p < 0.05). - **Table SI-1c**: Pearson r coefficient values for the linear correlation analysis between the predicted and observed responses calculated on an individual dataset. Values in bold indicate significant results (p < 0.05). - **Table SI-2a**: Percentage of word pairs, for which the difference in ratings had the same sign in both human judgments and the semantic projection calculated on the full dataset. - **Table SI-2b**: Percentage of word pairs, for which the difference in ratings had the same sign in both human judgments and the semantic projection calculated on the restricted dataset that excluded the words used in the construction of the composite vector. - **Table SI-2c**: Percentage of word pairs, for which the difference in ratings had the same sign in both human judgments and the semantic projection calculated on an individual dataset. - **Table SI-3**: Likelihood values for the logistic decision model for the three semantic evaluation models fit on the size and animacy trials together fit on the full dataset and the restricted dataset excluding the trials that were used to construct the composite vector. Bolded values indicate the best performing models as determined by the wAIC. - **Table SI-4**: Likelihood values for the linear ballistic accumulator model for the three semantic evaluation models fit on the size and animacy trials together fit on the full dataset and the restricted dataset excluding the trials that were used to construct the composite vector. Bolded value indicates the best performing model as determined by the wAIC. ---------- <h4> FIGURES: </h4> - **Figure SI-1: Results of the logistic model.** Log likelihood values for the logistic decision model combined across the size and animacy tasks. Different colors represent two distributional semantic models. Different shades represent the two datasets on which the models were evaluated. Values closer to zero correspond to a closer fit. - **Figure SI-2: Results of the LBA model.** Log likelihood values for the LBA decision model combined across the size and animacy tasks. Different colors represent two distributional semantic models. Different shades represent the two datasets on which the models were evaluated. Values closer to zero correspond to a closer fit. - **Figure SI-3: Results of the correlation analysis.** Pearson r coefficients between the predicted responses and the mean human judgments for the size task (upper panel) and animacy task (lower panel). Different colors represent two distributional semantic models. Different line types represent the two datasets on which the values were calculated. - **Figure SI-4: Results of the pairwise order consistency analysis.** Pairwise order consistency values for the size task (upper panel) and animacy task (lower panel). Different colors represent two distributional semantic models. Different line types represent the two datasets on which the values were calculated.