Given the number of languages being compared and the number of forms in many of the sources, forms that resemble one another in shape and meaning by chance will not be uncommon, and the decision as to whether a comparison that appears good is a product of chance must be based on criteria such as - how general the semantic category of the form is (e.g. phonologically corresponding forms meaning ‘cut’ are less diagnostic of relationship than phonologically corresponding forms for particular types of cutting), - how richly attested the form is (if it is found in just two witnesses the likelihood that it is a product of chance is greatly increased), - there is already a well-established reconstruction for the same meaning.