Fig 1.
Map of New Guinea languages groups and principal component analysis.
(A) Linguistic areas (Polygons) of Trans New Guinea (TNG) and Austronesian language groups (identified by index numbers given in C; * includes groups 1–7, see Materials and Methods.) and location of villages (dots) used as occurrences for Eco-Linguistic Niche Modeling; map background generated by using R::Raster—CRAN Repository, public domain software. (B) Principal Component Analysis of Eco-Linguistic Niches of New Guinea language groups and Eigen values of the most explanatory axis. Red arrows show environmental variable contributions. Environmental variable codes are explained in Table 2. Ellipses represent the inertia distributions for groups belonging to TNG (green) and Austronesian (blue) language groups. Grey background represents the available environment in New Guinea.
Table 1.
Eco-Linguistic Patterns (ELPs) identified according to environmental space position and geographical distribution of the modeled Eco-Linguistic Niches (ELNs).
Index labels (id) for linguistic groups are given in Fig 1. Number (N) of languages counted from the Glottolog database [27]. Geographical regions are indicated on the map in Fig 6.
Fig 2.
Examples of four Eco-Linguistic Niches (ELNs), Growing Season (GS), and Eco-linguistic patterns (ELPs).
(A) Asmat-Kamoro, (B) Roro, (C) Mek, and (D) New Britain language groups ELNs. Colour shades reflect probability of niche presence. Green colour is used for Trans New Guinean ELNs and blue for Austronesian ELNs. The purple lines delimit linguistic areas. (E) Map of New-Guinea with GS values calculated according to the ecological risk formula (see Materials and Methods). Coulour scale indicates the length of the GS in months. (F-J): ELPs of New-Guinea. Each pattern (Highlands, Center, South, South-East and North) corresponds to the sum of similar Eco-Linguistic Niches according to their environmental space positions and their geographical distribution. Colour shades indicate prediction probabilities. Polygons indicate linguistic areas of modelled linguistic groups. The Highland pattern (F) includes from West to East the West Trans New Guinea, Uhunduni, Mek, Engan, Chimbu-Waghi, Kainantu-Goroka and Angan language groups. The Center pattern (G) includes the Ok and Bosavi, the South pattern (H) the Marind and Gogodola-Suki, the South-East pattern (I) the Mekeo, Roro, Motu and the South-East Papuan, and the North pattern (J) New Britain and New Ireland. Map background generated by using R::Raster—CRAN Repository–public domain software.
Fig 3.
Heatmap of Eco-Linguistic Niche environmental distances computed according to the Schoener’s D and Hellinger’s I overlap scores.
Distance corresponds to 1 –the overlap score (Table 2 in S9). Values vary between 0 and 1. Low values (purple) correspond to strong overlaps and environmental similarity, high values (dark red) to marked environmental differences. Numbers correspond to the index numbers of language groups given in Fig 1.
Fig 4.
Principal component analysis and map of New Guinea Eco-Linguistic Niches.
(A) Trans New Guinea language family, (B) Austronesian language family. Dots represent the distributions of ELN predictions, and the colour shades depict the number of languages. Grey background corresponds to the available environment in New Guinea. Map background generated by using R::Raster CRAN Repository–public domain software.
Fig 5.
Principal component analysis of New Guinea language groups’ Eco-Linguistic Niches.
Arrows show environmental variable contributions. Environmental variable codes are explained in Table 2. Ellipses represent the inertial distributions for groups belonging to Trans New Guinea (green) and Austronesian language groups (blue). Background colour scales correspond to the sum of annual precipitation (A), annual mean temperature (B), altitude (C) and Growing Season (D). Map background generated by using R::Raster CRAN Repository–public domain software.
Fig 6.
Environment suitability according to Eco-Linguistic Niche (ELN) modeling for Trans New Guinean (TNG) and Austronesian language families.
A: Geographic areas in mainland and island New Guinea with relevance to the text. B: Environmental suitability map corresponding to the difference (delta) between the mean predictions of TNG ELNs and Austronesian ELNs calculated with formula ΔELN (cf. Methodology section). Map background generated by using R::Raster CRAN Repository–river shapes obtained from Natural Earth–public domain.
Fig 7.
Geographical distributions of language families.
Geographical distribution of the Austronesian (blue) and Trans New Guinean language families according to Ross [19] (A), Glottolog versions 2.7 (B) and 3.0 (C) [27], the selected language groups for which Eco-Linguistic Niche modeling (ELNM) was performed (D) and the results of the ELNM (E). Dots in (B) and (C) correspond to the center of the geographical location of the constituent languages of each language family according to Glottolog versions 2.7 (B) and 3.0 (C) [27]. Map A after Ross 2005 [19]; Maps B-E generated by using R::Raster–CRAN Repository–public domain software.
Fig 8.
Ecosystems of New-Guinea according to the World Wide Fund For Nature (WWF) ecosystem database [58].
Map generated by using R::rgdal package—CRAN Repository–public domain software.
Table 2.
Environmental variables used for Eco-Linguistic Niche Modeling.
Climatic variables obtained from the WorldClim model [74]; topographic variables are derived from the altimetric model ETOPO1 [75].