lpetrich
Contributor
Vyacheslav Vsevolodovich Ivanov of the IE glottalic theory, which he worked on with Tamaz Valerianis dze Gamkrelidze, has an article on protolanguages as objects of scientific description, as opposed to systems of sound correspondence. Rather arcane, I must say. Much of it was about the possibility that Proto-Nostratic distinguished between volitional and non-volitional verbs, being "active-passive", and that this distinction had traces in some of its descendants.
Then Aharon Dolgopolsky's "A Probabilistic Hypothesis Concerning the Oldest Relationships Among the Language
Families In Northern Eurasia"
That's a GREAT article.
2. "Statistical selection of semantic values represented by morphemes (word parts) which are relatively impervious to change." -- that's what's behind the Swadesh list and similar lists. AD essentially rediscovers that kind of list.
The first step is to find those stable word forms. As an example of different stability, AD compared words for "star" and "lightning" in several language families. "Star" was much more stable than "lightning" in most of them.
AD then looked in several language families and for each meaning, counted replacements. Using language histories would have been ideal, but not many langs have long written histories.
He investigated 140 Eurasian langs, and he wanted to also use langs from elsewhere, but their data was sometimes limited, no words noted for "heart", "louse", "horn". Their stability trends were similar, however.
He then consulted Carl Darling Buck's Dictionary of Indo-European Synonyms, using only those with at most 10 replacements, and Robert B. Lees's list of relatively stable meanings for langs with long written histories. These, along with some others that it might seem good to check, were collected into a list of 250 meanings.
He found some 40 meanings with highly stable word forms, with at most 12 replacements for each one. He then trimmed his list down further, using various criteria, and he decided that "who" and "what" deserve only one entry, unlike in the Swadesh lists. That left him with 15 meanings.
Not much for good statistics, it seems to me. Martine Robbeets used 254 meanings for Transeurasian (Macro-Altaic, Broad Altaic), discovered in a similar manner.
Then Aharon Dolgopolsky's "A Probabilistic Hypothesis Concerning the Oldest Relationships Among the Language
Families In Northern Eurasia"
That's a GREAT article.
1. "Comparison of several languages" -- that gives better statistics.A primary difficulty in resolving this issue resides in the lack of a mathematically rigorous procedure to substantiate linguistic relationships. When dealing with closely related languages whose genetic affinity is rather obvious, as for example, in the case of Turkic vs. Indo-European, we may dispense with such rigorous proofs. It is quite another matter, however, with distance linguistic relationships. Here, we require mathematically rigorous methodologies which will permit us to distinguish between languages that are actually genetically related and those that are merely fortuitously similar.
2. "Statistical selection of semantic values represented by morphemes (word parts) which are relatively impervious to change." -- that's what's behind the Swadesh list and similar lists. AD essentially rediscovers that kind of list.
The first step is to find those stable word forms. As an example of different stability, AD compared words for "star" and "lightning" in several language families. "Star" was much more stable than "lightning" in most of them.
AD then looked in several language families and for each meaning, counted replacements. Using language histories would have been ideal, but not many langs have long written histories.
He investigated 140 Eurasian langs, and he wanted to also use langs from elsewhere, but their data was sometimes limited, no words noted for "heart", "louse", "horn". Their stability trends were similar, however.
He then consulted Carl Darling Buck's Dictionary of Indo-European Synonyms, using only those with at most 10 replacements, and Robert B. Lees's list of relatively stable meanings for langs with long written histories. These, along with some others that it might seem good to check, were collected into a list of 250 meanings.
He found some 40 meanings with highly stable word forms, with at most 12 replacements for each one. He then trimmed his list down further, using various criteria, and he decided that "who" and "what" deserve only one entry, unlike in the Swadesh lists. That left him with 15 meanings.
Not much for good statistics, it seems to me. Martine Robbeets used 254 meanings for Transeurasian (Macro-Altaic, Broad Altaic), discovered in a similar manner.