Table 1.

Spatial autoregressive (SAR) error models for (a) log(language richness) and (b) log(threatened language richness). Results are shown for two univariate models (mammal richness and mean elevation), and one multivariate model (mammal richness + mean elevation + % land cover per grid cell). For each model, the Akaike information criterion (AIC) for the SAR model and the corresponding non-spatial ordinary least squares (OLS) regression model are given. Asterisk (*) indicates the models that provide a statistically significant fit to the data.

model	predictors	slope	p-value	AIC (SAR)	AIC (OLS)
(a) log(language richness)
1	log(mammal richness)	0.10	0.11	326.8	468.5
2	log(mean elevation)	0.0001	0.99	329.1	505.9
3	log(mammal richness)	0.09	0.26	329.0	471.5
	log(mean elevation)	−0.04	0.41
	% land cover	0.002	0.32
(b) log(threatened language richness)
1	log(threatened mammal richness)	−0.14	0.08	398.0	527.3
2	log(mean elevation)	−0.12	0.01*	394.2	519.2
3	log(threatened mammal richness)	−0.06	0.46	397.7	519.7
	log(mean elevation)	−0.11	0.04*
	% land cover	0.002	0.95