Figure 1.
Normalized frequency distribution of language sizes and language areas (home ranges) over the world.
When a language is spoken in two or more disconnected domains, the total area (sum of all disconnected polygons) is represented. The data set includes all living languages plus about 700 recently extinct languages listed in the Ethnologue [22]. Curves stand for Gaussian fittings on the logarithmic variables. Both fittings are statistically highly significant, yielding correlation coefficients for the distribution of population sizes and
for the distribution of areas.
Figure 2.
Linguistic diversity in Papua New Guinea and surrounding islands.
Each polygon corresponds to a domain where a single language is spoken. The matrix where polygons are embedded (see Materials and Methods) has been mapped to an indexed image using transformation of the population sizes for each language, as catalogued by SIL in the Ethnologue DB ver.15. The index values are colored according to the scale in the right, which indicates the absolute size of the group speaking each language. Unhabited landmass of Papua New Guinea is colored grey. The inset contains the population size histogram, with six bins per decade from 1 to
.
Figure 3.
Scatter plot of the logarithm of the area () and the logarithm of the population size (
) for all languages.
Six different representative regions are plotted: (a) Africa (2314 languages), (b) Asia (2333 languages), (c) Europe (260 languages), (d) Papua New Guinea (813 languages), (e) North America (585 languages), and (f) South America (517 languages). North and South American languages have been separated according to whether their centroids laid to the north or to the south of the N parallel, which stands for the northernmost point of South America. The corresponding values of
and
are compiled in Table 1. The value of
for each case coincides with the slope of the line drawn in each panel. Axes labels and sizes are the same in all cases.
Table 1.
Parameters characterizing the distribution of population size and area covered by human languages in several representative world regions.
Figure 4.
(a) Exponent and (b) correlation
yielded by the model as a function of the parameters respectively characterizing spontaneous retreat and conflict outcome,
and
. The population-area relationship is recovered with different parameters for all world regions, except for North America (not shown in the plots). Data for the world considers all points represented in Fig. 3. Though
and
are, in principle independent variables (in terms of an arbitrary BND), the model establishes a quantitative relationship between them. The allowed values of
when
remains fixed can be read from this figure, and vice versa. As an example, note that, if
,
is to be found in the range
, approximately.