A casual examination of the skaldic corpus suggests that a very large proportion of kennings are unique, that is, the nouns that form a kenning are very unlikely to appear elsewhere as a kenning, even when grammatical variation is taken into account. This enormous productivity is due to the principle of variability and substitution in kennings. The phenomenon is discussed in Fidjestøl’s 1974 structural analysis of the kenning system and presented there as a reversal of the principle of linguistic economy. This paper investigates this phenomenon by quantitatively analysing the kenning corpus as recorded in the skaldic database, in order to identify matching kennings within the corpus. The results are broken down according to the complexity of the kennings in question and are then examined in relation to the structural framework of the kenning system. It is shown, firstly, that a relatively small number of kennings are repeated (10–12 % of kennings are repeated), but that there is a fairly high amount of repetition among simple kennings (at least 25 % of simple kennings are found elsewhere in the corpus), suggesting that the kenning system has quantitative similarities with other linguistic systems, albeit with a much higher level of variability.