Glottochronology - how far do we get in eLinguistics.net?

The data in this study represents attractive candidates for trying to date the common ancestors of languages (Glottochronology) and thus to trace when language families and sub-families separated from each other. This sounds exciting, but we have to be extremely careful about the use of the data.
 
As an automated system, eLinguistics.net assumes that the rate of language erosion is constant in both the genetic proximity calculation and the trees. From recent language history, we know that the erosion is definitly not constant.
 
Let's consider following results to sharpen both the potential of glottochronology and its limits:
 

Language relationship Genetic proximity Date common ancestor Comments
Latin to Italian (Detail score | Lexical details) 19,7 +/- 1500 years Italian is one of the most direct descendents of Latin.
Old English to English (Detail score | Lexical details) 20,4 +/- 1000 to 1200 years Old English is approximately equidistant to English and other Germanic languages like Dutch, German, Frisian and Old Norse.
Old Norse to Swedish (Detail score | Lexical details) 20,1 +/- 1000 years Same degree of erosion as Old English to English - similar age.
Old Norse to Icelandic (Detail score | Lexical details) 3 +/- 800 years Not much younger separation than from Swedish but very close genetic proximity. The slow pace of evolution made by Icelandic is well documented and reflected here.

 
As we can see, some languages can evolve very slowly, as did Icelandic which is still the closest Northern Germanic language to Old Norse, the language of the Vikings. This is a well known issue - there are many reasons why languages evolve at different paces - wether the speakers where isolated (island, mountain,...) or in frequent contact with other languages, etc. An other, underestimated issue, is that the very concept of language as we know them today is more tricky than considering language as a clear entity. Today's languages are either an earlier dialect which has established as lingua franca or some kind of compromise between dialects which arose to make use of a common language in communities where dialectal differences made Intercomprehension almost impossible between speakers of the same language sub-family, living a few villages away. The history of each language is specific and this specificity makes it difficult to let the computer date the age of language separation.
 
Of course, this doesn't mean we can simply ignore the quantified relationships between languages. They do give an indication - but we have to understand that what we get is an approximate direction, not an exact date.
 
An other interesting point - from the eLinguistics.net's data point of view - is wether the correlation between the genetic proximity value as calculated from the lexical comparison and the age of the common ancestor is linear or not. It seems that languages lose from 5 to 20 points in the first 1000 years, 20 to 35 within 2000 years and then less and less per 1000 years - as the "root" of Indo-European is located at appr. 65 to 70, which should reflect about 7000 years BP.

An experimental timeline of the Indo-European languages on the basis of the data from this study can be downloaded as Pdf. It is a first step and should be regarded as a 'beta version'!