Ever since its first formulation in Geeraerts/Grondelaers/Speelmann (1999), lectometry has been widely used to map distances between language varieties or “lects”. Often, these distances are given a geometrical representation in a low-dimensional space. Examples are the use of Multidimensional Scaling in Speelman/Grondelaers/Geeraerts (2003) and Ruette et al. (2014) and of Correspondence Analysis in Plevoets (2008), Delaere/De Sutter/Plevoets (2012), Prieels et al. (2015) and Ghyselen (2016). Usually, the number of dimensions of the geometrical space is chosen on the basis of representativeness, leading to an approximate picture of the linguistic variation. However, the spatial dimensions can also be interpreted as underlying factors governing the variability of the data. This methodological paper will explore this functional interpretation of the geometrical dimensions by establishing the link between lectometry and Latent Variable Models. It will be shown that the dimensions of the lectal space can be considered as hidden variables which lay bare specific causal mechanisms. In particular, analyses of translation and interpreting data will demonstrate that the lectometrical dimensions can be made to correspond to various socio-cultural determinants. That opens up the possibility for lectometrical studies of determining the “social meaning” of linguistic varieties and variants.