Tanja Säily: Sociolinguistic Variation in English Derivational Productivity. Studies and Methods in Diachronic Corpus Linguistics

Helsinki: Société Néophilologique 2014 (Mémoires de la Société
Néophilologique de Helsinki Nr. XCIV). VII + 284 Seiten
Lea Kawaletz1
Published online: 29 April 2015
© Springer Science+Business Media Dordrecht 2015
In recent decades, it has become almost impossible to investigate the productivity
of word formation processes through corpora without taking into account statistical
measures of some kind. In the field of historical linguistics, however, this has been
problematic due to small corpus sizes and the lack of suitable methods. This book
proposes such methods, making a strong point for data-driven approaches while exploring
the diachronic and sociolinguistic variation of productivity in word formation.
Specifically, the study focuses on the native suffix -ness and its Romance-based
counterpart -ity, analysing data from 15th to 18th letters, 18th century letters, and the
BNC. In order to obtain valid results, the author combines statistical and visualization
methods which are already established in other research traditions, and complements
these with original software. Apart from presenting interesting findings in the field
of historical sociolinguistics, she provides methodological insights which are relevant
for research based on historical (or otherwise problematic) corpora, especially for but
not restricted to those interested in productivity and word formation.
The book consists of three parts. Part I summarizes the theoretical background and
previous research as well as the statistical methods employed, and does so in detail
but not excessively. The frame is set by five well-motivated research questions, four
of which are methodological:
1. Is there sociolinguistic variation and change in the productivity of -ness and -ity
in the history of English?
2. How can we study productivity in small corpora which contain a great deal of
spelling variation?
B L. Kawaletz
lea.kawaletz@uni-duesseldorf.de
1 Department of English Language and Linguistics, Institute of English and American Studies,
Heinrich-Heine-University, Universitaetsstr. 1, 40225 Duesseldorf, Germany
346 L. Kawaletz
3. How can we study variation and change in corpora which may not be completely
comparable over time and across genres?
4. Are the productivity measures proposed in previous research valid in and applicable
to sociolinguistic data of this kind?
5. What are the requirements for a usable tool for studying variation in productivity
in data of this kind?
Given that the book is a stapled dissertation, Part II comprises six distinct papers,
each dealing with one or more of these research questions. For instance, Chapters 7
and 8 concentrate largely on research question (3) and thus on methodology, while
Chapter 11 has a linguistic focus and tackles research question (1) by applying the
methods described in the preceding chapters to 18th century letters. It is obvious that
this dissertation is quite complex, basically dealing with any problem you might encounter
when trying to answer research questions such as (1) by consulting historical
corpora. However, with Part I setting the scene and the otherwise independent chapters
arranged in a way that they build logically upon each other, the reader can keep
track of the big picture. Part III then ties up all loose ends, evaluating both the linguistic
results and the methods used. With this frame provided by Parts I and III
the reader can also get an overview of the whole subject matter while being able to
quickly identify which chapters are relevant to read up on specific details. The book
is completed by two appendices: a very helpful glossary of statistical terms and a list
of sociolinguistic parameters applied.
The author makes a number of valuable methodological observations. For instance,
she shows that hapax-based measures of productivity are not reliable when
applied to small corpora, and proposes a combination of statistical methods and visualization
to overcome this problem. Thus, the rather strict method of resampling
by permutation testing is combined with type and hapax accumulation curves, which
allow the researcher to discover near-significant tendencies visually. In collaboration
with computer scientists, the author has developed an open-source program (types1)
which not only computes the upper and lower bounds of these accumulation curves
by means of Monte Carlo sampling, but also provides a built-in measure of statistical
significance. The author weighs the pros and cons of every method applied and
conclusion reached, and argues convincingly for the usefulness of introducing unconventional
methods into a traditional field. Apart from the open-source software,
complete analysis files are also available online, aiding in the transparency of her conclusions
and offering others the possibility to build on her research. Her findings are
relevant primarily for diachronic studies, but they are also applicable to synchronic
research faced with similar problems, such as large but messy corpora compiled from
the Internet.
Given the solid methodology it can be safely said that the author makes the best
out of the available historical data and presents a number of interesting linguistic
findings: Overall, she finds an influence of social factors (gender, social rank and participant
relations) on the use of both -ness and -ity. Furthermore, diachronic change in
productivity can mostly be asserted for -ity, while the use of -ness has remained very
stable. To pick out one especially interesting finding with regard to social influences,
the author shows that gender plays a central role: Throughout history and up until
Tanja Säily: Sociolinguistic Variation in English Derivational Productivity 347
today, women appear to use -ity less productively than men, while there is no gender
difference with -ness. Why this is the case, however, remains to be seen.
Overall, both the results and the methodology are convincing, and my points of
criticism are few: First, although the author takes great care to introduce all statistical
methodology and terminology to the uninitiated reader, some technical terms remain
undefined (e.g. ‘NP-complete problem’). Second, the social variables investigated
are not always completely transparent. Thus, ‘domicile’, a variable investigated in
Chapter 6 which turns out as not significant, is not defined there, nor introduced
in the background section of the book. It could be interpreted as referring to the
parameter ‘Place of birth: main domicile’ which is listed in Appendix II. However,
the subcategory ‘other’, which figures in Chapter 6, is not mentioned there, so that it
remains unclear what exactly is meant by this term.
These marginal issues hardly diminish the overall good quality of this book. I believe
that anyone basing his/her research on historical (or comparably problematic)
corpora can greatly benefit from this work, especially when dealing with word formation
processes and/or the influence of sociolinguistic factors. Although this book
is heavy on statistics, every method is so well explained that, in combination with
the provided glossary of statistical terms, anybody who has a basic knowledge of
statistics can follow with ease.
References
Suomela, J. (2007). types1: type and hapax accumulation curves. Computer program. ZENODO. doi:
10.528/zenodo.9860.
The British National Corpus (BNC) (2007). Version 3 (BNC XML Edition). Distributed by Oxford University
Computing Services on behalf of the BNC Consortium. http://www.natcorp.ox.ac.uk/ (accessed
01 April 2015).