Glottochronology

Several factors can skew results: borrowing inflates the cognate count (suggesting recent divergence when divergence is actually older), uneven rates of change, and loss of entire semantic domains due to cultural shift. Additionally, the assumed retention rate (0.81-0.86 per millennium) is an average; real languages vary. Always compare glottochronological estimates to archaeology, written records, and comparative linguistic evidence.

The Swadesh list is a standardized set of 100 or 200 basic vocabulary items (like 'I,' 'you,' 'water,' 'fire,' 'earth') chosen because they are thought to be relatively resistant to replacement and borrowing. The idea is that core vocabulary changes slower than specialized vocabulary, so using a standard list makes comparisons across language families more uniform. However, which items count as 'core' varies by language, and some items are more stable in some cultures than others.

Yes, but the calculations become more complex. You can compute pairwise cognate percentages for all language pairs, then use phylogenetic methods (like neighbor-joining or maximum parsimony) to construct a tree showing the probable order of divergences. Modern computational approaches treat this as a Bayesian phylogenetic problem and account for uncertainty in the data.

The method itself does not provide a formal confidence interval; you must estimate it from the quality of data and external checks. A rough rule of thumb is ±10-20% of the calculated date for recent divergences (< 2000 years) and wider margins (±30-50%) for deeper time. Always support glottochronological dates with other evidence before publishing them as established fact.