PACSCL Hidden Collections Processing Project

Courtney and I both felt strongly, from the very beginning of the project, that sacrificing description for speed was a risk in this project.  Although we know that every collection could still use additional work, we worked hard to make it so that the repository did not feel that additional work was necessary before they made the collection public.  Moreover, we knew from the start, that many of the collections would NEVER be worked on again.  Unfortunately, that is just how it is.

So what have we learned about description?  We learned that description takes a lot of time—in fact, that is probably the first thing we learned in this project when we tested the manual and discovered that even an experienced processor could not arrange and describe a fairly straightforward collection from start to finish in 2 hours per linear foot.  As a result, Courtney and I created processing plans that included a preliminary biographical/historical note before processing started.  In general, we have learned that it generally takes roughly the same amount of time to describe a collection as it does to arrange a collection.

I’m not going to lie … I am pro description … few things give me more professional pleasure that a beautifully crafted folder title or a paragraph in a scope and content note that I know will help a user determine if this collection is going to help them with their research.  That is the whole point—letting researchers know that we have the stuff that they need.  As a result, the PACSCL/CLIR team took it seriously.  Description is the one part of training that has probably evolved most over the course of the project.  We developed exercises to help our processors write better and more descriptive folder titles and structure notes so that they are both concise and informative.  The project didn’t have a lot of time, so we tried to make our processors think like a user and learn to quickly assess the contents of a folder.  For the most part, we are really pleased with our finding aids and I think, nine times out of ten, researchers will be able to determine by the finding aid if the collection is worth their time in looking at it.

One of the really interesting things we learned is, to me, still the most counter-intuitive.  A collection with extremely tidy existing arrangement usually results in a collection with less thorough description.  I am going to use two specific collections to illustrate this issue.

The first collection is the Dillwyn and Emlen family correspondence, 1770-1818, housed at the Library Company of Philadelphia (unquestionably one of my favorite collections in this project—as well as being one of my biggest disappointments, archivally speaking).  When I sat down to process this collection, I was really confident—the collection was 2 linear feet and was already arranged.  At one point in time, it had been bound in volumes and at another point in time, the letters were removed from the volumes and placed in very acidic folders.  Every letter had a catalog number written on the document.  While a few of the letters were out of chronological order, the vast majority of the collection was arranged very effectively; each folder containing letters from a span of dates.

This collection desperately needed to be re-foldered.  Not only were the folders highly acidic, but they were too small and some of the letters were showing a bit of damage.  I re-foldered the 130 folders in the collection which took about 2.5 hours.  Then I entered the folder list into the Archivists’ Toolkit which probably took only about 15 to 20 minutes.   So in roughly 3 hours (three quarters of my allotted time), I had the collection rehoused and the folder list in the Archivists’ Toolkit, which left me 1 hour to write a scope and content note.  Should have been easy, right? Well, no. Because this collection was perfectly arranged, I did not need to look at even one document in order to create the container list.  Moreover, the container list is not very helpful to a researcher.  All it contains is a list of dates which means that the scope and content note should be full of the subjects addressed in the correspondence.  Problem is, I did not know anything about the letters.  There was no way that I could read enough of the letters in an hour to discover all the topics addressed in the letters that will almost certainly be interesting to researchers.  I did my best—I valiantly scanned through as many letters as I could and wrote down key topics that popped up more than once or twice, and as each minute passed, my heart sank just a little more—I knew perfectly well that I could never do this extraordinary collection justice, even with twice the time.  Prior to beginning processing, I had performed my research for the biographical note and I had discovered that several authors had used portions of the collection in their published works … so I turned to them for expertise on this collection.  They wrote about only a tiny portion of the collection, Susanna Dillwyn Emlen’s bout with breast cancer.  I soaked up every bit of information in their books and included it in my scope note in order to give users the most information possible, but I feel like the project failed this collection.  Perhaps I feel this so strongly because I had been so confident in significantly improving access to it.

I have beheld the second collection, the Belfield collection, 1697-1977, housed at the Historical Society of Pennsylvania, with equal amounts of awe, excitement and horror since I first laid eyes on it.  Never have I seen such a mess of a collection—please see just a few photographs as words cannot effectively describe the condition of this collection.  Courtney and I spoke with Matthew Lyons of HSP and he said that he was not expecting much more than good box level descriptions of the contents.  Even with these reduced expectations, we thought it wise to double our forces and therefore, Michael, Celia, Courtney and I all worked together on this collection.  I am happy to say that this collection will, for quite a few series, contain folder level description, but even more than that, the scope and content note for this collection is rich, deep and full of the flavor of the four generations of family who lived at Belfield.

So why does a collection that was the biggest (filthiest) mess of all time result in a better finding aid than a small and beautifully arranged collection?   I know it is because we were forced to sift through the messy collection in order to create any order, and it is amazing how much one absorbs simply by looking at the collection.  In the end, I feel that this is one of the biggest rapid maximal processing successes of the entire project.  We took the collection from utterly unusable chaos to an order that could certainly be refined, but is beyond serviceable.

When selecting collections for a minimal/rapid maximal processing project, consider your time frames and what result you want from the project.  If you want a container list in a hurry, select a well-organized collection.  If you want fuller description, a collection that needs some arrangement will probably be the best choice.  From a purely selfish perspective, I would pick a wreck of a collection over a tidy one every time—the sense of accomplishment and success is so much sweeter than that despair I still feel when I think of Dillwyn and Emlen letters.

I mentioned in an earlier blog post that there are about 3 collections that I don’t feel enormously benefited from this project.  In every case, the collections had existing arrangement that I felt either prevented me from starting from scratch or were in good enough order that I did not learn valuable content that I could then share with researchers.

Fairly early in this project, Courtney and I determined that “MPLP 2 Hours” was not going to be a wholesale success—most collections simply cannot be processed in that time frame, regardless of the shortcuts taken (our average across the board is 3.2 hours per linear foot).  And in some cases, those shortcuts resulted in a product that we did not feel was more useful to a researcher post-processing.  What we have determined is essentially this … it is difficult, if not impossible, to say that collections can be processed in a set or determined amount of time, but it is possible to make educated estimates allowing us to allocate human resources to process collections efficiently.

There are several factors that allow us to better determine a time frame for the processing of collections:  age, type of collection, and original arrangement of the collection are the three biggies. None of these factors work independently—they are all intertwined to help determine the time frame.  So, based upon the data collected for 125 collections, processors have physically processed collections with the oldest material dating from the:

17^th century at an average of 4.1 hours per linear foot;

18^th century at an average of 3.3 hours per linear foot;

19^th century at an average of 3.4 hours per linear foot;

20^th century at an average of 2.9 hours per linear foot.

Processors have processed:

artificial collections at an average of 3.6 hours per linear foot;

institutional/corporate records at an average of 2.5 hours per linear foot;

personal papers at an average of 3.7 hours per linear foot;

family papers at an average of 4.2 hours per linear foot.

Age seems like it should be the most logical factor, but in fact, it has proven to be the least certain factor in our ability to judge the time frame for processing.  We thought originally that old collections (pre 1850s for certain) would take us significantly longer to process, but this is not necessarily the case.  The age does not seem to deter us in being able to efficiently process an “old” collection.  Age does, however, quite frequently deter us from describing the collections well.  Quickly skimming for content in folders of 17^th, 18^th and 19^th century handwritten material is not easy—and it absolutely results in less thorough description.  However, if the collection is arranged and available for research use, perhaps this is where we ask for help … as researchers use the collections, we can ask them to provide more robust description of what the correspondence, journals, etc. contain.  Finding aids CAN be iterative … especially with technology such as the Archivists’ Toolkit.  “Newer” collections may or may not be easier to process … certainly there is more typewritten material that makes it immediately easier to categorize series/subseries/folders and describe the contents of the folders more thoroughly.  However, in the end, the ease of the processing relies more heavily on the type of collection more than the age.

For this project, we have divided collections into four basic types:  institutional/corporate records, personal papers, family papers and artificial collections.  Again, there is no one size fits all … each collection is unique (is that not why archival collections are so awesome?).  Generally speaking though, an institution or company’s records can be processed most quickly, followed by personal papers and then family papers.  Artificial collections are usually the fastest or the slowest depending entirely upon the collector.  Usually, they are speedy—the collector is in love with the topic they are collecting and as a result, they arrange the collection for their own personal satisfaction and use—all the letters of a children’s book author are arranged chronologically by date sent or alphabetically by the recipients’ names.  If this is the case, the artificial collection is a dream to process and it usually requires only description.  In a few instances, however, we have found collections where the collector simply collects … they probably know that the stuff is important, but they are not organizers.  At that point, trying to create a system out of a group of randomly acquired material can be quite difficult.

Institutional and business records are usually quick and easy and this is because the functions of a business or an institution generally follow the same basic structures and are fairly predictable.  Usually, you will find financial records, minutes, committee records, administrative records, subject files, correspondence, etc.  Because the function generates the records, it is logical and easy to determine a good organizational scheme for the papers.  But as always, the collections are unique and we have found that different creators generate different levels of tidiness, logical order, and structure.

Personal papers are the next quickest to process (generally speaking), especially if the creator was involved in several major movements, careers, and/or activities.  However, the ability to efficiently process a person’s personal collection often depends upon how intermingled those pursuits are with family, friends, and work.

Family papers have been, fairly consistently, the most time-consuming collections to process.  The problems that arise with family papers that generally do not exist with personal papers are the intertwining relationships that make determining to whom a certain group of materials belong challenging, and sometimes, impossible.  When every generation in a family has a woman named Sarah, determining generations becomes a trial.   Many a day passed at the Historical Society of Pennsylvania with the following conversation: “So wait, this is Sarah Logan Wister Starr?”  “No, this is Sarah Logan Starr Blaine!”  Or:  “Here is a letter to Grandma Sarah from Sarah …does that mean it is Sarah Logan Starr Blain?”  “No!  It could be Sarah Logan Starr Blain OR Sarah Logan Wister Starr OR Sarah Tyler Boas Wister!”  Egads … I wanted to buy a baby name book for this family!  Not surprisingly, this kind of questioning takes time … lots of time.

The third main factor in determining time for processing a collection is existing arrangement.  A collection of 20^th century business records thrown into boxes will take longer than a collection of 18^th century business records that are housed in volumes.  A collection of family papers organized by the donor into distinct family member’s papers can probably be processed more quickly than a collection of personal papers that are completely unsorted.  I have intentionally not used the term original order which implies that the order was generated the creator.  Existing arrangement may have been generated by the creator, but in many cases, it is generated by an archivist who starts processing the collection but does not complete the project.  Unfortunately, the hardest collections to process efficiently are often collections that someone else has started to process.  Trying to understand an undocumented order that has been imposed or continue with an arrangement scheme that does not seem logical is much more difficult than imposing order from absolute chaos.  And without a questions, the collections that take the absolute longest are ones in which parts of the collection have received item level treatment.  Addressed in the next blog post will be how this type of existing arrangement affects description of collections.

So, basically what we have said here is that every collection is different and unique and there is absolutely no way to say that one time will work even within a date frame or a type of record. Our observations are backed by Greene and Meissner who say that “MPLP … advises vigorously against adopting cookie-cutter approaches … and [recommends] flexible approaches,” (page 176).  In order to make educated estimates for allocating resources, we believe that a base-line starting time frame is needed:  institutional/corporate collections should be given 3 hours per linear foot.  Based upon the existing arrangement, tack on another hour per linear foot if it is in a shambles.  If the bulk of the material is from the 18^th century, tack on yet another hour per linear foot for increased perusal time which will result in more effective description.  So, in this case, your estimated processing time is 5 hours per linear foot.  Could you do it in three?  Yes, probably.  However, with allowances for age and existing arrangement, you will almost unquestionably have a better product, still at just over ½ the rate of traditional processing.

Based upon our experience, the PACSCL/CLIR project believes that the following base-line processing time estimates would work well:

Artificial collections:  3 hours per linear foot

Institutional/corporate collections:  3 hours per linear foot

Personal papers:  4 hours per linear foot

Family papers:  6 hours per linear foot

Our averages clearly show how quickly collections can be processed … but the base-line estimate with upgrades allows us to provide the best possible product while being mindful of available resources.