In the case of operation with the stems, we may see the same methodological process as in the case of the phonetic variations, which is made with the intention of extending pseudo-scientific possibility of comparing a great number of words from different languages. With the same view the semantic parallels are extremely extended. In fact, if one confines the comparison of the words with the restricted meaning, e.g., the pine-tree, leg, cloud, etc., the number of possible words will not be very great. For this there is a semantic reason. The fact is well established that the words (sounding starters) change their meaning (complex of conditioned reflexes) and vice versa in the most capricious manner — from part to total and vice versa, from one to another complex, etc. If a certain semantic complex is well established for a certain group of dialects (language) in its historic aspect, and if it may hold for a series of languages, it may be and must be used in the analysis, but the same complex cannot be transferred into another language without being checked up. From this point of view, the analogy with the phonetic variations is very close. The semantic connexions may have no reverse force. It does not mean that one must not try, for instance, to find a common word in one of the semantic modifications, but it means that the finding not supported by other evidences, e.g., historic, is not convincing at all, and when largely used it may bring one to an absolutely erroneous inference. So we may say that the closer the meaning, the surer the result of comparison. It is true that such a limitation results in a great limitation of material which may be compared, but at the same time it makes the comparison reliable. These elementary rules are ignored by A. Sauvageot, who freely transplants the semantic complexes from one language to another in a great number of his cases [140]. In some cases, hundreds of notions, verbs, nouns, and adjectives are covered, and they are supposed to be alike in all languages. Of course, without such a methodological liberty, the number of cases would be greatly reduced.
140. At the basis of this methodology is the same theory of evolution which recognizes a similarity of semantic variations as one of the fundamental characters of the psycho-mcntal complex organically conceived.