For our testing purposes , we have been using the numbered <c>'s. In doing so, we have made some assumptions about how we use them that address a number of issues recently raised on the list. (While all three Yale repositories are using numbered levels, the following assumptions apply primarily to the way in which Manuscripts and Archives is approaching the issue.) - We are only using the level attribute available in the <c>'s for "series" and "subseries." Below that, we rarely have or identify formal sub-subseries, treating entries in the finding aid simply as a <unittitle>. These <unittitle>s continue to form a hierarchy embedded within enumerated <c>'s, but without a level attribute. This reflects an earlier discussion on the list with Alvin Pollack and Janice Ruth. Each enumerated <c> level represents an ordered set of <unittitle>'s arranged according to a common principle. Thus all <c03>'s might be listed alphabetically (or chronologically) by <unittitle>. - Within each numbered <c> level that does not have a level attribute, we make no assumptions about what level the entry represents, and I am not sure that it matters. They are simply entries within an hierarchical list. Logically, I am not sure that one could identify and name all of the various levels consistently down to the item level and if one could, I am not sure what one gains by it. I can see a potential need to extract/index/display or otherwise manipulate series, subseries in a particular way to the exclusion of other levels (not sure about files or items), but the levels in between probably represent conveniences for grouping materials, rather than true levels. To go back to Leslie Morris's earlier (3/26) metaphor, if the species of animal is important, define it in an attribute. I should note that there is not complete consensus on this point here and others from Yale may post an alternative view. - Steven Mandeville-Gamble's suggestion, seconded by Kim Brookes, that a "container" attribute be added to the <c>'s mixes logical and physical groupings. "Box" and "folder" are not hierarchically subordinate to sub-series or to any other of the values for the (logical) level attribute in <c>. The design of many of our finding aids, with the container numbers on the left as though they were the organizing principle for the collection or record unit, reinforces this confusion. Once you get beyond series and subseries, the organizing principle (whether alpha, chron, or some other) is really - except for unprocessed materials without any order other than how they are boxed - the <unittitle>. The design choice was to treat the intellectual/logical arrangement as primary rather than the physical/container setup and I would encourage us to continue in this direction. Rich Szary Manuscripts and Archives Yale University Library