Looking at examples in the testdata, you find (spaces added for clarity)
336 $a text $2 rdacontent 337 $a unmediated $2 rdamedia 338 $a volume $2 rdacarrier which reveals that the item is a plain book. The same might, using the codes instead, also be recorded like this: 336 $b txt $2 rdacontent 337 $b n $2 rdamedia 338 $b nc $2 rdacarrier The codes suggest two things: 1. 337 is redundant (the letter is always the first letter of the 338 code) 2. The verbal terms are less useful for indexing. For if you use the codes, you can truncate nc to n, for example, to get all unmediated stuff, cx to c to get all computer usable stuff, and so on. This is not possible with the words. One step further: if you string it all together into txtnc, you get the idea what can be done with it. In addition, verbal terms are not international and, in the course of history, subject to change. We all know what that means for large databases. The use of verbal terms is, in other words, subadequate both in terms of sustainability and machine actionability. This means the method used in the test is a waste of time and space and produces inferior data. B.Eversberg