Academia.eduAcademia.edu

Figure 18 – uploaded by Russell Gray

See full PDFdownloadDownload figure
Figure S12: Varying Priors for Penalizing Complexity in the INLA-analysis.

Figure 18 S12: Varying Priors for Penalizing Complexity in the INLA-analysis.

Related Figures (77)

phylogenetic effects, however the same is not true for those predicted to be areal (see Fig. S9) Figure 1: Variance explained by phylogeny and geography. Each point is a Grambank feature. The
phylogenetic effects, however the same is not true for those predicted to be areal (see Fig. S9) Figure 1: Variance explained by phylogeny and geography. Each point is a Grambank feature. The
Figure 2. Grammatical similarity in the Grambank sample of languages. The color coding represents
Figure 2. Grammatical similarity in the Grambank sample of languages. The color coding represents
Language families, then, can be both distinct and diverse samples from the design space.
Language families, then, can be both distinct and diverse samples from the design space.
Southern Sierra Miwok (ISO 639-3: skd, Glottolog: sout2985) Examples
Southern Sierra Miwok (ISO 639-3: skd, Glottolog: sout2985) Examples
Chalcatongo Mixtec would be coded as 1. A plural pronoun that is unmarked for clusivity can be derived from the polite or familiar first person pronouns with a prefix, but the inclusive first person is inherently plural. There is no first person plural pro-form that is marked for exclusivity. The existence of an inclusive form is sufficient to trigger a 1 and the lack of an exclusive form has no impact on this. Chalcatongo Mixtec (ISO 639-3: mig, Glottolog: sanm1295)
Chalcatongo Mixtec would be coded as 1. A plural pronoun that is unmarked for clusivity can be derived from the polite or familiar first person pronouns with a prefix, but the inclusive first person is inherently plural. There is no first person plural pro-form that is marked for exclusivity. The existence of an inclusive form is sufficient to trigger a 1 and the lack of an exclusive form has no impact on this. Chalcatongo Mixtec (ISO 639-3: mig, Glottolog: sanm1295)
Southern Sierra Miwok would be coded as 1. The lack of a first person dual exclusive form does not affect this designation, nor does the fact that the language has first person plural markers in Series 1 and 2 that do not mark clusivity.
Southern Sierra Miwok would be coded as 1. The lack of a first person dual exclusive form does not affect this designation, nor does the fact that the language has first person plural markers in Series 1 and 2 that do not mark clusivity.
Table S9:1. Example of theoretical metric calculation For our fusion metric we assigned a value of 0.5 to features that are consistent with the typological pattern of expressing information through phonologically bound morphs but which do not necessarily indicate that grammatical information is expressed by phonologically fused elements. For example, GBO75: “Are there postpositions?” encodes whether languages use an element that follows a noun to express adpositional meanings. Both postposition words (which are phonologically independent) and postpositional enclitics (which are phonologically fused)
Table S9:1. Example of theoretical metric calculation For our fusion metric we assigned a value of 0.5 to features that are consistent with the typological pattern of expressing information through phonologically bound morphs but which do not necessarily indicate that grammatical information is expressed by phonologically fused elements. For example, GBO75: “Are there postpositions?” encodes whether languages use an element that follows a noun to express adpositional meanings. Both postposition words (which are phonologically independent) and postpositional enclitics (which are phonologically fused)
Figure S1. Comparison of coverage per language and feature in WALS and Grambank.This plot shows that the amount of missing data per language is much lower in Grambank compared to WALS. The total number of languages is 2,430 for Grambank and 2,435 for WALS. The numbers are derived on the dialect-aggregated dataset, see Supplementary Material 1:7.
Figure S1. Comparison of coverage per language and feature in WALS and Grambank.This plot shows that the amount of missing data per language is much lower in Grambank compared to WALS. The total number of languages is 2,430 for Grambank and 2,435 for WALS. The numbers are derived on the dialect-aggregated dataset, see Supplementary Material 1:7.
Figure S2. Barplot showing the coverage of languages per Glottolog macroarea. Light blue represents languages which do not yet have a grammar as indexed by Glottolog, dark blue indicates languages that are already in the Grambank database and orange denotes languages which have a grammar indexed in Glottolog but which are not (yet) in the Grambank dataset. Languages in the light blue category are most likely not possible to include in Grambank, whereas the orange category could be included in future. The numbers are derived from the dialect-aggregated dataset, see Supplementary Material 1:7.
Figure S2. Barplot showing the coverage of languages per Glottolog macroarea. Light blue represents languages which do not yet have a grammar as indexed by Glottolog, dark blue indicates languages that are already in the Grambank database and orange denotes languages which have a grammar indexed in Glottolog but which are not (yet) in the Grambank dataset. Languages in the light blue category are most likely not possible to include in Grambank, whereas the orange category could be included in future. The numbers are derived from the dialect-aggregated dataset, see Supplementary Material 1:7.
Figure S3. Tree plot of GB133, the Grambank feature with the highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (verb-final is a pragmatically unmarked constituent order for transitive clauses), uncolored = no (verb-final is not a pragmatically unmarked constituent order for transitive clauses), gray = missing data. Branch colors represent probability estimates: yellow = higher probability that verb-final is a pragmatically unmarked constituent order for transitive clauses, purple = lower probability that verb-final is a pragmatically unmarked constituent order for transitive clauses.
Figure S3. Tree plot of GB133, the Grambank feature with the highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (verb-final is a pragmatically unmarked constituent order for transitive clauses), uncolored = no (verb-final is not a pragmatically unmarked constituent order for transitive clauses), gray = missing data. Branch colors represent probability estimates: yellow = higher probability that verb-final is a pragmatically unmarked constituent order for transitive clauses, purple = lower probability that verb-final is a pragmatically unmarked constituent order for transitive clauses.
Figure S4. Tree plot of GB074, the Grambank feature with the second highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (there are prepositions), uncolored = no (there are not prepositions), gray = missing data. Branch colors represent probability estimates: yellow = higher probability that there are prepositions, purple = lower probability that there are prepositions.
Figure S4. Tree plot of GB074, the Grambank feature with the second highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (there are prepositions), uncolored = no (there are not prepositions), gray = missing data. Branch colors represent probability estimates: yellow = higher probability that there are prepositions, purple = lower probability that there are prepositions.
Figure S5. Tree plot of GB090, the Grambank feature with the third highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (the S argument can be indexed by a prefix or proclitic on the verb in simple main clauses), uncolored = no (the S argument can not be indexed by a prefix or proclitic on the verb in simple main clauses), gray = missing data. Branch colors represent probability estimates: yellow = greater probability that the S argument can be indexed by a prefix or proclitic on the verb in simple main clauses, purple = lower probability that the S argument can be indexed in this way.
Figure S5. Tree plot of GB090, the Grambank feature with the third highest phylogenetic effect in the INLA (dual) model. Tip point colors represent observed values: black = yes (the S argument can be indexed by a prefix or proclitic on the verb in simple main clauses), uncolored = no (the S argument can not be indexed by a prefix or proclitic on the verb in simple main clauses), gray = missing data. Branch colors represent probability estimates: yellow = greater probability that the S argument can be indexed by a prefix or proclitic on the verb in simple main clauses, purple = lower probability that the S argument can be indexed in this way.
Figure S6. Map of GB038, the Grambank feature with the highest spatial effect in the INLA (dual) model Purple indicates languages that have demonstrative classifiers; Orange indicates languages that do not have demonstrative classifiers.
Figure S6. Map of GB038, the Grambank feature with the highest spatial effect in the INLA (dual) model Purple indicates languages that have demonstrative classifiers; Orange indicates languages that do not have demonstrative classifiers.
Figure S8. Map of GB136, the Grambank feature with the third highest spatial effect in the INLA (dual) model. Purple indicates that fixed word order occurs in the language; Orange indicates that fixed word order does not occur in the language.
Figure S8. Map of GB136, the Grambank feature with the third highest spatial effect in the INLA (dual) model. Purple indicates that fixed word order occurs in the language; Orange indicates that fixed word order does not occur in the language.
Figure S10 Spatial decay in precision matrices for spatiophylogenetic analysis. This figure shows the relative decay in covariance based on the various parameterisations of the Matérn function. The x-axis shows Haversine distance ("as-the-crow-flies" distances, taking into account the curvature of the earth), and is shown on the y-axis with the black line for reference. The red line indicates the parameterization of spatial covariance used in the main text. Blue, and green lines show parameterizations that iteratively increase the relationship of geography between languages in the model. Vertical dotted lines ground the covariance functions in real-world distances to give a sense of at what point geographic relationships are no longer statistically relevant in this model.
Figure S10 Spatial decay in precision matrices for spatiophylogenetic analysis. This figure shows the relative decay in covariance based on the various parameterisations of the Matérn function. The x-axis shows Haversine distance ("as-the-crow-flies" distances, taking into account the curvature of the earth), and is shown on the y-axis with the black line for reference. The red line indicates the parameterization of spatial covariance used in the main text. Blue, and green lines show parameterizations that iteratively increase the relationship of geography between languages in the model. Vertical dotted lines ground the covariance functions in real-world distances to give a sense of at what point geographic relationships are no longer statistically relevant in this model.
Figure $13: Simulation results for the 12 conditions (four levels of phylogenetic signal, for three different proportions of traits). Each column of graphs contains the results for a particular model structure, each row of graphs contains the results for a particular proportion of traits, and within each graph shows the results across the four levels of phylogenetic signal. The dual process model contains two boxplots per level of phylogenetic signal, one representing the posterior mean for the phylogenetic effect, and one for the posterior mean of the spatial effect.
Figure $13: Simulation results for the 12 conditions (four levels of phylogenetic signal, for three different proportions of traits). Each column of graphs contains the results for a particular model structure, each row of graphs contains the results for a particular proportion of traits, and within each graph shows the results across the four levels of phylogenetic signal. The dual process model contains two boxplots per level of phylogenetic signal, one representing the posterior mean for the phylogenetic effect, and one for the posterior mean of the spatial effect.
Figure S14. Feature loadings onto PC1 and PC2, including only the top 40 most contributing features.
Figure S14. Feature loadings onto PC1 and PC2, including only the top 40 most contributing features.
Figure S16. Scatterplot of Indo-European languages (green) among all other languages (grey) and their position given PC1 and PC2 with specific languages highlighted with names. Figure S17. Scatterplot of Austroasiatic languages (blue) among all other languages (gray) and their position given PC1 and PC2 with specific languages highlighted with names. The two major clusters in the Austroasiatic family correspond to languages inside and outside of the Indian subcontinent.
Figure S16. Scatterplot of Indo-European languages (green) among all other languages (grey) and their position given PC1 and PC2 with specific languages highlighted with names. Figure S17. Scatterplot of Austroasiatic languages (blue) among all other languages (gray) and their position given PC1 and PC2 with specific languages highlighted with names. The two major clusters in the Austroasiatic family correspond to languages inside and outside of the Indian subcontinent.
Figure S18. Pairwise Cultural Fixation scores over macroareas in the Grambank dataset. The pair with the lowest score, and therefore most likely to be similar, is North and South America.
Figure S18. Pairwise Cultural Fixation scores over macroareas in the Grambank dataset. The pair with the lowest score, and therefore most likely to be similar, is North and South America.
Figure S19. Pairwise Cultural Fixation scores over AUTOTYP-areas in the Grambank dataset.
Figure S19. Pairwise Cultural Fixation scores over AUTOTYP-areas in the Grambank dataset.
Figure S21. Comparison between different unusualness probability density estimation approaches. Each column/row corresponds to individual estimators. Lower triangle panels show smooth loess curves. Panels on the diagonal show probability densities. Upper triangle panels show Spearman correlation values. Figure S21. Comparison between different unusualness probability density estimation approaches. Each
Figure S21. Comparison between different unusualness probability density estimation approaches. Each column/row corresponds to individual estimators. Lower triangle panels show smooth loess curves. Panels on the diagonal show probability densities. Upper triangle panels show Spearman correlation values. Figure S21. Comparison between different unusualness probability density estimation approaches. Each
Figure 825. Histogram of Manhattan distances between languages in Grambank. Pairwise Manhattan distances show for each pair of languages in the dataset how many times they had different values, in absolute terms. The height of the bars show how many pairs of languages had that particular Manhattan distance. A Manhattan distance of 0 means that there were no features for which that language pair had different values. The mean Manhattan distance in the entire dataset is 39.
Figure 825. Histogram of Manhattan distances between languages in Grambank. Pairwise Manhattan distances show for each pair of languages in the dataset how many times they had different values, in absolute terms. The height of the bars show how many pairs of languages had that particular Manhattan distance. A Manhattan distance of 0 means that there were no features for which that language pair had different values. The mean Manhattan distance in the entire dataset is 39.
Table S1 Model fit scores (WAIC) of BRMS models with a beta-distribution prediction the mean spatial and phylogenetic effects of Grambank features.
Table S1 Model fit scores (WAIC) of BRMS models with a beta-distribution prediction the mean spatial and phylogenetic effects of Grambank features.
Table S2. Phylogenetic and spatial effect in INLA model per feature.
Table S2. Phylogenetic and spatial effect in INLA model per feature.
Table S3: Correlation coefficients of association between Principal Components and Theoretical scores, as calculated by PGLS.
Table S3: Correlation coefficients of association between Principal Components and Theoretical scores, as calculated by PGLS.
Table S5. Table of binarised Grambank features
Table S5. Table of binarised Grambank features
Table S6. Grambank features with information on theoretical scores and predictions from Nichols (1995).
Table S6. Grambank features with information on theoretical scores and predictions from Nichols (1995).
Table S7: Cultural Fixation Scores between AUTOTYP-areas
Table S7: Cultural Fixation Scores between AUTOTYP-areas
Table S8. Coefficients and associated error estimates for the spatiophylogenetic Bayesian regression mode predicting Unsualness scores.
Table S8. Coefficients and associated error estimates for the spatiophylogenetic Bayesian regression mode predicting Unsualness scores.
Table S9: Language pairs with a Manhattan distance of 0
Table S9: Language pairs with a Manhattan distance of 0
Related topics:
Languages and LinguisticsEndangered LanguagesLanguage DiversityLanguage Loss
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved