Dear Pragya and Hans,
Thank you for your support.
Across all the census language token data are very sparse mentions of people self-identifying as mother-tongue speakers of Vedic Sanskrit, Pali and Prakrit. One of the issues is that these instances appear in much earlier enumerations, only featuring a mere handful of tokens at best. Today, if someone were to self-identify as an L1–L3 speaker of Prakrit, etc., its token would no doubt end up in the amorphous Other category, rendering further analyses impossible.
The whole issue of whether or not there are 'Sanskrit-speaking' villages filled with people only speaking a pure, hybrid or imperfect learning variety of the language is largely a mute point. While there are fragmented pockets among the 'rural masses,' most of the villages I visited during my fieldwork – mentioned in multiple internet blogs, etc. – barely have anyone speaking Sanskrit.
What my book pivots toward is using the census data to reverse-engineer the analysis. The data demonstrate where the aspirations to identify as speakers of Sanskrit are located. Therefore, it doesn't matter whether someone can speak Sanskrit, or if a village in MP is filled with people who have become 'cultured and civilised' through Sanskritization. To a certain degree, empirical linguistic verification becomes obsolete. After all, what the census data confirm are indications of aspiration to become a Sanskrit speaker and a commitment to imagine a Sanskritland (Hindu nation) and world.
Importantly, the census data show that this aspiration is mostly found in urban areas across the Hindi Belt. Linked to this result is the affinity between Hindi, Sanskrit and English – regardless of their respective positions within the 3-language formulaic options (SHE, HSE, ESH, etc.) allowed by the census rules – sits at 95% + in almost every instance.
The following image shows the districts that returned more than 100 L1-Sanskrit tokens in 2011.
The next image shows the top 20 rural sub-districts that returned the highest L1-Sanskrit tokens (i.e, aspiration) in 2011. It includes the locations of the more infamous saṃskṛta-grāms. What is interesting is that none of these internet-famous villages are located within any rural sub-district that returned L1-Sanskrit tokens.
Diachronic analysis of state, district and sub-district data demonstrates token shifts beyond any natural propensity. One example is Sitapur District, UP.
Notice the shift between 2001 and 2011.
Thank you for your interest.