Thanks Christine, Robin, and John for such excellent papers! *Insert virtual round of applause here 👏*

I'm Melanie Walsh, and I'm pleased to be the discussant for this panel. I'll start the conversation off with a few questions for the group and for each panelist, but I hope that you, our workshop participants and panelists, will feel free to steer the discussion in whatever direction moves you.

What can library holding records — in addition to, or alongside, circulation records — tell us about a library's priorities, biases, limitations, etc.? 

Though these two papers explored libraries in different places and time periods, I was struck by their common interest in library holding records, in addition to or alongside library circulation records. Around 23:36, Christine makes the case that circulation records cannot necessarily be taken as evidence of reader preference and (implicitly) that they cannot be understood without considering library holding records: "Historians cannot necessarily interpret patrons choices of public library books as indicated in circulation records as evidence of their preferred tastes in reading. Circulation records can only indicate patrons choices from among what was on the shelves." She makes this point after discussing how many mid-century libraries failed readers who were seeking out books by Black authors, such as the Booklovers' Club.

Around 23:38, Robin and John similarly discuss how Chicago Public Library branch holdings of a particular book influence patron check-outs of that book, a phenomenon that they coin the "shelf effect." They show that, for many books, more holdings lead to more check-outs, but they also show that that isn't always the case. Apparently for certain books, such as The Book Thief, patrons requested the book even if it wasn't on their own branch's shelves.

So I'm wondering if all the panelists might expand on the significance of library holding records to their research and to the study of library circulation more broadly. I also have a few specific questions.

Robin and John: Based on your research, what do you think are some of the reasons that a patron might — or might not — be likely to request a One Book One Chicago book if it's not on the shelf? Are CPL's holding records useful for studying patterns or disparities within the library system itself? Also a small clarifying question. Do you have access to CPL circulation data for books beyond the One Book One Chicago Books?

Christine: How do holding records play into your emphasis on libraries as infrastructure? Do you know where the Booklovers' Club members were getting access to books by Black authors that weren't likely to be found at local public libraries? 

Hi Melanie, thanks for these great questions, and thanks to Robin and John for the introduction to their extremely interesting OBOC research. 

First, your “holdings” question. I see library acquisitions as the flip side of library circulation. The two go together, along with other organizational factors. Historians may latch with enthusiasm on to circulation data as a proxy for readers’ demands, but it’s important to point out that those demands must be framed within the limits of the library as a system that includes a number of different elements. The “shelf” factor might be one of those elements, especially for the later 19^th century onwards.  But we also need to avoid anachronism when envisaging libraries of the past. In the United States, direct user access to shelves (“open shelves”) cannot be assumed before the end of the 19^th century, and not necessarily even then, especially in academic libraries. Before open shelves became normal, readers would have to request books from library staff, perhaps by first consulting some kind of catalogue—creating layers of mediation between readers and texts. What constraints did these access practices place on readers, especially women, for example? Would they hesitate to ask in person from a male librarian for particular titles?  Even when readers did gain direct access to the shelves, and could therefore browse for their next title, women could be constrained by physical factors: did they need to climb a ladder, or walk around a gallery where those below might see up their skirts?  Were/are they at risk of being accosted in the unsupervised stacks?  (Not a joke, actually—readers at one American university still have to show ID to enter the stacks because of an attack by an axe man decades ago.)  In short, my plea is for historians to look at the whole picture of the library, before making inferences from one particular set of library records.

Your second question is about how the Book Lovers obtained their books. The short answer is, we don’t know. But we can speculate that they might have learned about suitable titles for review from a number of sources. In the 1930s Black interest magazines like The Crisis (NAACP) and Opportunity (Urban League) reviewed books, but sometimes after the Book Lovers, so this is not the whole story. The main Black newspaper in Iowa was the Bystander, but it didn’t do a lot of book reviewing. Other possibilities are through presentations at Black churches, or one of the many other African American organizations the Book Lovers belonged to. There is some small evidence that they were in touch with Black bookstores. They also worked with librarians at the Public Library of Des Moines, then one of the very few US public libraries concerned to foster reading books by and about African Americans, to produce specialized book lists, though it’s possible that the flow of information was from the Book Lovers to the librarians, rather than the other way around. The 135^th Street branch of the NYPL also distributed such book lists.  

A couple of quick notes: 

For Robin and John:

You might take a look at Alex Leslie's attempts (Panel 5) to construct a predictive model for borrowing using WMR data.  I'd be curious to hear how it compares to yours.  I'm also generally hearing more from both you and Alex about what you see as the value of  "predictive insight" derived from circulation efforts. In your DHQ article you discuss the utility it offers to CPL staff as they consider choices for OBOC.  What are the benefits of such an approach for humanities scholars interested in, for instance, other kinds cultural interventions or trends?

On a more specific note: are you able to make any observations yet about the demographics associated with the neighborhood/branch-specific borrowing patterns you detect?  The DHQ article explains how you determined the social character of branch neighborhoods, so I am in whether you see any patterns worth discussing.

For Christine:

It's interesting to me that it's precisely because you are examining a group outside the circulation records (the Book Lovers Club) that key elements of the library infrastructure became visible. I'm reminded of the workingmen's library effort we came across while researching WMR, which threw into sharp relief what the public library was not doing. Is that something you think is necessary in general in the kinds of circulation-based studies we've both done?  A subscription or social library has a more obviously circumscribed set of borrowers but for public libraries this seems essential, at least as a complement to analyzing circulation data.  Attention to the infrastructure you describe is not just a means of understanding the "what" in "who read what," but also the who, at least obliquely.

Christine, thank you for a wonderfully lucid presentation.

One question, however (I do apologize if you covered this and I missed it): to what extent is a public library an industrial system and institution? As opposed to being agrarian, or post-industrial, or embodying some other set of economic organizing principle? By "industrial" I mean something that has calculable capital requirements, predictable and standardized inputs and outputs, and a standardized staffing model. This sounds hyperbolic, but: an assembly line for reading.

I can't quite get my head around this, but it seems like we as a group might be thinking about three different kinds of libraries: the sort of early or pre-industrial library described in "Books and Borrowing Across Scotland", a post-industrial library of "Modeling Contemporary Reading Behavior at City-Scale", and the Muncie Public Library, which is what I think of when I think of a "library" with the features outlined in your talk.

Am I overthinking this, or is this really a thing?

@jim-connolly 

Jim, yes, that’s a good point about the failure of libraries and the Book Lovers.  One of the key insights into the concept of infrastructure is that we tend to only notice it when it breaks down.  But I do think we need to be generally conscious of the operation of infrastructure, because it can bake in effects and qualities that otherwise don’t come to our attention, or that we gloss over (think of the airy “cloud” rather than the reality of power guzzling data farms occupying vast areas of farmland). And yes, I do think that circulation studies need to consider the entire institutional framework in which the records are produced. There is a risk in using any organizational records for purposes other than those for which they were originally created, and circulation records are no different. This is particularly true, I feel, when the researcher is inferring from organizational records to individual behavior, intentions, preferences, and so on.  I agree that understanding more about the “who” is of course a benefit of looking at the various interlocking parts of the system.

I am perhaps sticking my neck out here when I suggest that historians are generally very comfortable with the idea of libraries—libraries are familiar, and historians can be lulled into thinking they know a lot about them. But this confidence can be misplaced, and they can make assumptions that don’t hold in different periods and places. As an example, one of the things that I find students can’t get their heads around is the notion of textual scarcity. Living at a time of textual abundance themselves, they can’t conceive of a time when reading materials were rare and precious. Yet even in the mid-twentieth century we find accounts of people—Americans!—willing to read anything they could get their hands on. At the very least this complicates the notion of demand and choice in selecting books.

@steve_pentecost 

Steve, first, I think that the concept of infrastructure is useful, no matter whether one is talking about pre-industrial, industrial or post-industrial systems. But yes, the modern, publicly-funded public library, which began in the mid-nineteenth century and certainly lasted through the C20 (and maybe still now, but that’s a different question) was thought of in industrial terms even in the C19. Melvil Dewey, who had an enormous influence on the development of western librarianship in many different ways, was very smitten with Taylor’s ideas of scientific management, and for most of C20 library service was conceived very much in terms of inputs and outputs. Library improvement was thought to be achievable through the application and achievement of standards. There was also an impulse to grow. Library units became bigger, and amalgamated. And, of course, libraries were early adopters of automation and digital technology. With the development of computerized bibliographic utilities beginning in the 1960s, their bibliographic control more and more followed well-defined paths. The notion of the assembly-line delivering books to readers is not at all hyperbolic (I have used it myself!). Your over-arching tri-partite scheme sounds plausible; I’d like to hear more about “post-industrial,” in particular. 

Thank you for this illuminating look at context to re-frame the question of reader agency. On the subject of scarcity, literacy, and (lack of) access, it's always instructive to reread Rølvaag's story of Midwester immigrants, _Giants in the Earth_.

@cpawley Thanks for your elaboration about acquisitions and how the "shelf effect" might have influenced patrons in earlier historical periods. I'm fascinated by your points about how women may have hesitated to ask for certain books from a male librarian or been reluctant to climb a ladder — those potential constraints and "layers of mediation" do seem so important to consider.

@jim-connolly I'm also curious about what Robin and John think of Alex Leslie's predictive modeling compared to theirs — as well as the humanistic insights that can be gleaned from predictive modeling in particular.

To zoom out from predictive modeling a bit, I've also been thinking about the affordances/constraints of computational methods more broadly. As Jim points out, Christine is able to make library infrastructure visible by studying a group (the Book Lovers Club) outside circulation records, but it also seems important that she has deeply researched, historicized, and contextualized this group.

For Robin, John, or anyone else, how do you think about the balance or tension between computational methods and closer analysis (close reading, historicization, interviews, ethnography, etc.) for the study of library circulation?

@melaniewalsh

Related to your last question, in Mark Towsey's Keynote, he describes circulation records as a tool for filling in the wider landscape and provide comparative context for analysis of individual reader choices and experiences.  Elsewhere in this workshop (Tatlock et al, for instance), individual examples provide illustration of patterns extracted from circulation data.  This difference, between data as the starting point and as background pattern, is interesting to consider.  What are pros/cons of each?

@melaniewalsh Thank you for the suggestions, and I too was struck by the connection in our presentations over what we could term "shelf effect." (Prof. Pawley's points about the consequences of lack of access to books by/for/about African Americans is important, and I also recommend Laura Helton's recent PMLA article about Dorothy Porter and cataloguing.)

In our case, each One Book One Chicago "season" had a city-wide launch via social media accompanied by physical advertising (posters, flyers, etc.) in branches. Speaking with the director of the OBOC program over the years, she indicated that some seasons were able to have a minimum standard physical infrastructure in place in all 80 branches but others seasons did not have 'full coverage' in all branches. In addition, some branches had multiple print copies of the season's book 'at hand' with advertising (at least at the start of the season) and other small branches did not. The difference in our early 21st c. environment, of course, is that patrons can place another branch's copy on hold through the CPL website or check out an e-copy and avoid waiting altogether. (We have these statistics too. This makes for an interesting new dimension to thinking about participation and patron 'fulfillment' (to highjack a term made a bit sinister by Amazon): in our case, patrons with an interest in a particular book/season topic have multiple ways of getting it, and on different timescales (travel to other local branch, request from other branch to one's home branch, check out e-copy, purchase one's own copy, etc.). Given the full library database records we have for about 400 books (i.e. the seasonal choices and 'if you liked ... ' recommended titles) over the past decade it is possible to tell stories of access and fulfillment on these scales. At what branches, and when (e.g. after book groups or film tie-in events at branches), are patrons more likely to place holds? and the like.

Our models of "shelf effect" in seven recent seasons is one way of trying to quantify, to commensurate, reader behavior given some things we know (since we do not know the individual patrons like Middletown/Muncie and Osage). We do know e.g. how many physical copies of a book was held at any particular time, we know in many cases, what branches did for seasons and on particular days (interviews and records from branches), and we have checkout and hold numbers. In our DHQ paper, at figures 8-10, you will see some ways to join these data to models of branch demographics.

Prof. Pawley's "missing middle" of organizations and infrastructure is a productive architecture for our analysis, and I am grateful for the selection of our papers together for this panel. In the case of Reading Chicago Reading, we have the Chicago Public Library system (org) and social media (infrastructure of a peculiar kind). The library as "platform" for which (as CPL advertises it) "the book is just the beginning."

A reflection on this social media "missing middle" in our case of library circulation records 2011-2021: social media is both data riches (#OBOC tweets, Goodreads reviews, all time-stamped and linkable outwards, etc.) but also proprietary. The Twitter feed can be scraped via API. But CPL's "One Book" program has years of Facebook data that we have not been able yet to scrape. For scholars of early 21st century reading publics, now and in the distant future, there are many complications to building up patron information when dimensions are paywalled and proprietary.

@cpawley When I mentioned the "post-industrial" library, I was gesturing toward the library plus social media.  We've heard a bit about that around the One Book, One Chicago program, and The Nerdfighter's Census. 

I'm not entirely convinced that the social media add-on, if you will, is always different in kind from older ways of shaping reader choices (book reviews, back of book publisher advertisements, etc), although perhaps it's different in that it can replace an older top-down, few-authorities-speaking-to-many-readers model with a networked, many-readers-talking-to-many-readers model, at least on some cases (heavens, what a messed up sentence!  did I leave out any punctuation marks?).

I don't think that One Book, One Chicago and The Nerdfighter's Census are both entirely peer-to-peer networked structures for taste formation, but the affordances of social media suggest that readers might be able to get outside the industrial model's ability to shape reader taste by shaping their sense of what's available.

Thanks, all, for the comments.

Re: Holdings. We were originally interested in holdings as a way to normalize branch circulation. It made sense that branches with more copies would generate more checkouts and we wanted to look at something like checkouts per copy as a more comparable measure of branch-level intensity of interest. But then it turned out that many branches had zero copies so we could do that. (We ended up normalization using checkout / visitors, instead.) But since we had the holdings data, we were interested whether it had an impact on circulation and it did, for some books as we note.

An interesting thing to look at would be to see if the shelf effect varies in some systematic way based on the demographics or branch size or something else. That is, holding constant the effect related to a particular book, can the branch-level variation in this quantity be predicted? It's an interesting question, but I'm a bit hesitant to do that just because those multi-level model effects are themselves uncertain. Building models around their output parameters seems  speculative.

@jim-connelly Lessons about demographics from the modeling (based on my a bit hazy recollections).

As a baseline characteristic, higher property values and higher percentages of white folks are correlated with higher involvement in One Book. Both of these effects varied considerably by book, but not enough to overcome the baseline trend. Another trend we found (maybe not surprising) was that these whiter, more affluent parts of the city were also the ones more interested in books geographically distant from the city, which were a bit less popular in branches with more diverse demographics.

I think one thing we learned was that the library has been generally doing a pretty good job of trying to find books appealing (over time) to a variety of audiences in the city. It was one of their express aims and they seem to be able to do it without any fancy data science. 🙂

Re: circulation data for other books

We were originally very limited in our ability to collect circulation data because of the library's LMS infrastructure (there's that word again). They went through a system migration and suddenly the queries became much faster to run. At that point, we were able to request circulation data on around 300 volumes that were promoted as recommendations associated with particular OBOC selections. We wanted to see if there were measurable ripple effects to these peripheral texts. That data is gathering dust right now, but there may be some interesting insights there. Sadly, the circulation data that covers the first 10 years of the One Book program was lost in a prior system migration in 2011 and is completely unrecoverable.

Re: comparison with Alex Leslie's modeling work

(Which is very interesting, and I made some comments over there, too.) I think the key difference here is that his work makes use of patron identity (or something like it) and so he can approximate the choices of individual readers and their taste for books. We were not able to collect patron identity information from the circulation database -- in fact the library doesn't even store it. So there is no way to know about the actions of any particular patron even in an anonymized form. It would be useful, for example, to know how many "repeat customers" there are, checking out the library's OBOC selection year after year, but there's no way to get to that. We can only model at the aggregate branch level and therefore we don't have an individual notion of "taste" but instead more of a collective one. Another key difference is that we are not looking at the normal reading diet of patrons longitudinally as Alex is able to. We don't even have profiles of all the books checked out at a particular branch over time. Rather we are sampling from patrons activities as they intersect CPL's civic initiatives. So, I would say our models are more about the intersection of the "neighborhood tastes" that emerge in library branches and the "civic tastes" of the library staff as demonstrated in their One Book selections.