I began this project back in 2005 as a side interest to my main research into the cultural history of the Irish language. Since that time, I've made halting progress on a few central problems that need to be solved in order to get to my goal: a clearer understanding of the timing and demographic features of Ireland’s nineteenth-century language shift. Last summer, after a few months of soaring success as I applied what appeared to be clear solutions to these problems, work came to a screeching halt when the numbers I was producing stopped making sense (more on this in coming posts).
Since that time, other teaching and researching responsibilities this fall have prevented me from regrouping and moving forward. However, this gives me the opportunity to provide some backstory on the knotty issues at the heart of this census investigation. Doing this will also enable me to continue to fulfill the overall purpose of these online posts as I see it, namely, to draw back the curtain a little on the research process prior to publication. Historians tend to present final conclusions as if conjured abruptly in their entirety, with perhaps only extended footnotes to hint that years of thought and wrong turns went into such work. This insistence on presenting research in its finished and definitive form, rooted no doubt in the lingering concept of the historian as a type of chronicler, has its strengths in that it (ideally) prioritizes final conclusions over premature conjectures and underdeveloped speculation. At the same time, it seems that much of what makes the historical process interesting to historians are the twists and turns of investigation. A more transparent account of that process, as presented here in something approximating real time, therefore may be of interest, too. As a bonus, should anyone want to helpfully weigh in on the investigative process, there is an opportunity here to do so long before I—with any luck—send this off to readers and publishers.
With that, there is no better place to begin than with the primary difficulty inherent in Ireland's nineteenth-century censuses. This is the fact that the 1881 censuses recorded a higher percentage of Irish speakers in the country than had been recorded in the 1851, 1861, and 1871 counts. This naturally contradicted what was obvious to all--that the proportion of the population speaking Irish was declining, not increasing--and it was immediately recognized by census officials even at the time that earlier counts were not capturing the full number of Irish speakers.
The possibility that Irish speakers were under-reporting their own abilities in the language arises here, of course, as does the question of undercounting due to the characteristics of many nineteenth-century Irish-speaking communities (generally in rural, more isolated upland areas, or in the poorer and densely settled districts of cities). This may have made them less prone to accurate counting. But, as was also recognized at the time of the counts, a more basic hindrance to accuracy had likely been created by the way the schedule for the 1851, 1861, and 1871 household forms presented its request for information on languages. Enumerators were instructed in a footnote at the bottom of the form to add the word "Irish" in the education column to all persons who were monoglot Irish speakers, and "Irish and English" to bilinguals. Here is what the 1851 form looked like:
Click here for a detail of the footnote in question.
The good chance that this footnote was often overlooked by enumerators has been identified by a number of historians, including G.Brendan Adams and Garret FitzGerald, as a contributor to the undercounts in 1851-1871. To this we might therefore emphasize that as far as these forms were concerned, an incorrectly recorded Irish speakers (i.e. a speaker of Irish for which the words "Irish" or "Irish and English" were not marked on the forms) and a correctly recorded English monoglot "look" the same in terms of the final census count. Some of the "English speakers" in the 1851-1871 counts, in other words, are actually misrecorded Irish speakers. Accounting for this fact and, if possible, trying to estimate the size of this error, is a central concern for these censuses.