In a language such as Greek where the morphological and syntactic phenomenon of nominal composition is extensively developed, the interpretation of a compound needs to take into account as much as possible its sense and the phraseological associations that could lead to the creation of the word in question. To illustrate this fact, five examples are considered: [Greek] ablecheros/blechros, 'weak, feeble' <*'unable to hit' ([Greek] ablechros, cf. ballo),aganos, said in particular of heralds, not 'illustrious, noble' but perhaps 'he who cries loud', i.e.,[Greek] aga-, (*mgh(2)-) + -au-o-s, cf. [Greek] ano, akmenos 'fasting from food' <*'who has not been looked after, cared for', i.e., *n-kmh(2)-no-s, cf. *kem-h(2)- in [Greek] komizo; same for [Greek] thamnos...akmenos Od. 23, 190-191 = [Greek] aneu komides, telugetos perhaps 'born far away (from his parents)': [Greek] tel(e) + ugetos, cf. *h(2)eug- 'grow'. Finally, [Greek] chaliphron can be compared to [Greek} chairein phrena rather than [Greek] chalao, the original sense having been 'who rejoices in his [Greek} phrenes, hence 'carefree'. It is evident that if we mainly have archaic texts, certain analyses would be confirmed or weakened: for the ancient languages, the linguistic interpretation remains dependent on textual data.