Or: When is an identifier not an identifier
Or: Oh yeah, Google is fallible
We have gotten in the habit of using ISBN’s as identifiers for titles (or more accurately, based on how ISBN’s are used, for manifestations as determined by the ISBN authority, which are not quite the same thing as manifestations as determined by us. Anyway).
For instance, the Google Books API allows you to query by ISBN and determine if a book exists in GBS, and at what service level.
Using this service, I (or one of my librarians doing testing, actually) recently ran into an ISBN that seemed to be giving me the wrong book from GBS. At first I thought that GBS had bad metadata, certainly not out of the question. But then I remembered hearing something somewhere about how publishers can re-use ISBNs for different titles. Both of these books were from the same publisher, hmm…
ISBN: 0804710198
Is in my catalog, on a record for a book titled “Women and marriage in Kpelle society / Caroline H. Bledsoe”, Stanford University Press, LCCN 78066170. (My catalog record has no oclc num in this case). I look up that record in LC cat though, and sure enough it’s got this ISBN in LC too.
According to Google, that ISBN belongs to “Daily Life in Spain in the Golden Age By Marcelin Defourneaux”, Stanford University Press, 1979. Now, oddly that ISBN does NOT show up in the Google displayed metadata for that record. Yet Google returns that hit from that ISBN to the GBS javascript API. Hmm.
In fact, I can’t find any source that lists 0804710198 as a valid ISBN for “Daily Life in Spain…”. Worldcat doesn’t list 0804710198 as an ISBN for the 1979 edition of the Defourneaux book. Neither does LC cat.
So maybe this isn’t a case of a re-used ISBN at all, maybe this is just bad data or a bug in GBS? What do you guys think? Yes, error in GBS?
You think there’s any way for me to report this to GBS, or anyone at Google who would care? Anyone?
Bibliographic Wilderness 
haven’t looked closely, but when searching manually in Google Book Search, I get both of those books in the results. However, the top match and the one you’re talking about shows the number as a snippet …ie
Daily Life in Spain in the Golden Age – Page 19
by Marcelin Defourneaux – History – 1979 – 256 pages
… 0804710198 …
Limited preview – About this book – Add to my library – More editions
so the number seems to be matching something in the text, not in the metadata…
do you have the link that caused the problem?
It’s entirely possible it’s based off a bad record that’s since been fixed. One nice thing about GBS is that we’re certainly discovering a lot of records where the isbns were wrong…seems to outweigh the few times GBS might screw something on their end so far.
Hi,
Interesting post and quandry you have. According to the ISBN standard, ISBNs cannot be “re-used” for a manifestation of a work or a different title. There have been cases where publishers have attempted to do so, but major bibliographic databases such as Books in Print will not allow for such assignments to become public.
When searching this ISBN in Books in Print, the following results came up:
Title: Women and Marriage in Kpelle Society
Author: Caroline H. Bledsoe
Publication Date: July 1980
Publisher: Stanford University Press
Country of Publication: United States
Market: United States
ISBN: 0-8047-1019-8
ISBN13: 978-0-8047-1019-0
Jonathan, not sure he can do anything, but I do know that he’d certainly care: Kurt Groetsch. Recent UIUC grad and cataloger and metadata junkie works at GBS.
Several ways to contact him at one of his blogs: http://kurt.vox.com/profile/
Interesting comment from Andy about BIP not allowing a doubly used ISBN to become public. Uh, they may not allow it in their database but having seen re-used ISBNs physically printed in/on books, including in CIP data, and also in OCLC records, it seems a bit silly to say they “will not allow for such assignments to become public.” They are already public.
And no, publishers are not supposed to re-use ISBNs but I’m not supposed to jaywalk either.
Andy, thanks, good to know that re-use of ISBNs violates the standard, that sets my mind at ease.
Jon, here’s the GBS API call I was making. I note that it has both ISBN and LCCN in it.
http://books.google.com/books?jscmd=viewapi&callback=gbscallback&bibkeys=ISBN%3A0804710198%2CLCCN%3A78066170
But okay, I’ll try one with only the ISBN too. Yes, same behavior:
http://books.google.com/books?jscmd=viewapi&callback=gbscallback&bibkeys=ISBN%3A0804710198
So I guess I should let Google know about a bug of some kind?
It probably is a problem with Google metadata. We have found a handful of identifiers that return GBS links to the wrong book. Most of those have been LCCN numbers.
I believe there is a feedback form buried somewhere on the GBS site, but you might have better luck pursuing Kurt as a contact to let them know.