Day 1: Using Parsed Corpora for Diachronic Research

Schedule:

• 9:00 - 10:30
  

  Tony Kroch and Beatrice Santorini (Penn)

  What a parsed corpus is and how to use it

  In this presentation, we will show what information syntactic annotation adds to electronic text corpora and how such annotation can be used in the analysis of historical material. Using the phrase structure annotation scheme developed for the Penn Parsed Corpora of Historical English (Kroch and Taylor 2000, Kroch et al. 2004, 2010) and the York Corpora (Pintzuk and Plug 2002, Taylor et al. 2003, 2006), also the basis of the annotation in sizable corpora of French (Martineau et al. 2008), Icelandic (Wallenberg et al. 2010), and Portuguese (Galves and Britto 2002) as well as smaller corpora of German, Ancient Greek, and Spanish, we will show what information the annotation provides and what it leaves out. We will discuss how the problem of structural ambiguity can be handled and why the trees assigned to the sentences of a corpus cannot provide a complete structural description. We will also give a brief overview the methods used in the creation of the corpora and how they can be effectively searched for grammatical options and for quantitative analysis.

  more details...

  Beatrice Santorini's slides

  Tony Kroch's slides

• 10:30 - 12:00
  

  Aaron Ecay (Penn)

  Do-support in the parsed EME corpora: beyond Ellegård (1953)

  Do-support is the phenomenon whereby, in modern English, the auxiliary verb do is inserted in negative sentences, emphatic affirmatives and questions when the sentence otherwise lacks an auxiliary verb. The emergence of this construction in the Early Modern English (EME) period (1300-1700) ranks among the most extensively studied syntactic changes in linguistics. Ellegård(1953), the first rigorous quantitative study of the change, constructed a large diachronic corpus of potential do-support sentences, which enabled patterns to be noticed in the evolution of the change. Some of these patterns were noticed by Ellegård himself - such as the fact that a specific lexical class of verbs (with know as the most prominent member) resists do-support. With modern parsed corpora, it is possible to revisit Ellegård's work, replicate it and extend it, as we will see in this presentation.

  more details...

  Aaron Ecay's slides

• 12:00 - 1:30
  

  Lunch

• 1:30 - 3:00
  

  Caitlin Light (York)

  The use of parsed corpora in information structural research

  This presentation will explore the application of diachronic parsed corpora to questions of information structure and the syntax-pragmatics interface. We present a comparative and diachronic study of the information structure of passive constructions, drawing data from parsed corpora of Icelandic (Wallenberg et al. 2011), English (Kroch et al. 2004; Kroch and Taylor 2000; Taylor et al. 2003) and German (Light 2011). The corpora in question are not annotated for information structural categories, and indeed, recent work suggests that information structural annotation is inconsistent and will require a deeper theoretical understanding of the categories in question (Bech 2013; Cook 2013). Instead, we will highlight several methods for using corpus data for research on information structure without relying on information structural annotation.

  More details...

  Caitlin Light's slides

• 3:00 - 4:30
  

  Joel Wallenberg (Newcastle)

  Cross-linguistic corpora and the theory of language change

  We aim to show how cross-linguistic comparison with multiple diachronic, parsed corpora can test and develop a detailed theory of linguistic variation and change in ways that data from a single corpus cannot. Using a diachronic parsed corpus of Icelandic (IcePaHC; Wallenberg et al. 2011) and four parsed diachronic corpora of English (Kroch et al. 2005, 2010; Kroch and Taylor 2000; Taylor et al. 2003), we present a comparison of long-term syntactic change in Icelandic and English of a type that was impossible to carry out before this kind of diachronic comparative data became available. In our case study, the same initial state of variation in embedded polar questions in Icelandic and English leads to two different outcomes in the two languages: replacement of one form by the other in Icelandic, and specialization of the two forms for different syntactic contexts in English. Furthermore, the change in English is slow enough that it is not observable at all without the remarkable time-depth and sample size afforded by the combined four corpora of historical English. The Icelandic and English data bear out Kroch (1994)’s prediction about the two possible outcomes of morphosyntactic variation, specialization and replacement, suggesting a particular theoretical direction in how variation must be represented in a Minimalist syntax.

  More details...

  Joel Wallenberg's slides

  Tutorial to reproduce the analysis

Day 2: Perspectives on Historical Syntax

Schedule:

• 9:00 - 9:40
  

  David Lightfoot (Georgetown)

  Nothing in linguistics makes sense except in the light of change

  Just as in biology nothing makes sense unless one can provide a plausible account of how it might have evolved (Theodosius Dobzhansky 1973), so in syntax no feature of I-languages makes sense unless one can show how it might have been first acquired by children exposed to new primary linguistic data. Diachronic syntacticians have been able to show how particular elements of new I-languages might have been triggered by new PLD and therefore have been able enrich models of acquisition beyond the ambitions of synchronic syntacticians. So English has elements not occurring in other European languages and we can show how those elements were acquired in response to new E-language PLD and how there were domino effects.

  David Lightfoot's handout

• 9:40 - 10:20
  

  William Snyder (Connecticut)

  Grammatical conservatism and the child's role in language change

  A growing body of evidence supports the claim (Snyder 2007 etc) that children are "Grammatically Conservative" during normal first-language acquisition: Children do not make productive, spontaneous use of a new syntactic structure until they have both determined that the structure is permitted in the adult language and identified the adults' grammatical basis for it. If this is correct, it has interesting implications for the likely roles of children in language change. Grammatical reanalysis of a surface form in child-directed speech should be quite rare, at least in normal, generation-to-generation transmission of an existing natural language, although the picture could be quite different in situations of language contact. On the other hand, simple loss of a structural option over time becomes more likely, if the conservative learner insists on robust evidence in the input. The hope in this talk will be to generate discussion of specific predictions and relevant historical data.

  William Snyder's handout

• 10:20 - 11:00
  

  Claire Bowern (Yale)

  Syntactic change within theories of change

  In this talk I explore some of the consequences of the different paths that research in phonological and syntactic change has taken. Although syntactic and phonological change is often conceptualized differently, I argue that many of the problems that apply to syntactic change are also applicable in the domain of phonology. I suggest that some of the differences between syntactic and phonological change are instances of what biologists call "individualist thinking" versus "population thinking."

  Claire Bowern's slides

• 11:00 - 11:40
  

  Marlyse Baptista (Michigan)

  E pluribus unum: Out of many voices, one language

  This presentation focuses on diachronic development of the Tense Mood and Aspect (henceforth, TMA) systems of two varieties of Cape Verdean Creole. The first variety spoken on Santiago emerged in the latter part of the 15th century and the second variety, spoken on São Vicente, developed in the latter part of the 19th century. In the first part of the paper, the synchronic examination of the TMA forms in the two varieties highlights their similarities and differences. In the second part, the diachronic analysis explores the etymological origins of these forms, the grammaticalization and reanalysis processes they have undergone and reveals that the traces of the founding languages may still be readily detectable in the Santiago variety today.

  Marlyse Baptista's handout

• 11:40 - 12:20
  

  Carmel O'Shannessy (Michigan)

  Age and contact-induced language change

  The sociolinguistic context of the genesis of Light Warlpiri (O'Shannessy 2012) suggests that young children provided radical innovations in the emerging mixed language. This may have been at least partly due to children spending a lot of time interacting with each other from an early age. These observations correlate with contexts of genesis of koines (e.g. Kerswill & Williams 2000), suggesting that children between the ages of 3 and 12 years may be those most influencing contact-induced language change.

  • Kerswill & Williams 2000. Creating a new town koine: Children and language change in Milton Keynes. Language in Society 29.1: 65-115.
  • O'Shannessy 2012. The role of code-switched input to children in the origin of a new mixed language. Linguistics 50.2: 305-340.

  Carmel O'Shannessy's handout

• 12:20 - 1:30
  

  Lunch

• 1:30 - 2:10
  

  Sally Thomason (Michigan)

  Deliberate (morpho)syntactic change

  For some time now, evidence has been accumulating that speakers can and sometimes do make deliberate changes in their language, including syntactic changes. It is not always possible to prove that completed changes were made deliberately and partly or entirely consciously; but one can at least show that they happened too fast to be the result of `normal' gradual language change. With innovations that have not (at least not yet) become changes in a language, it is often easy to prove that an element of deliberation was involved in producing the innovation. These phenomena raise a number of interesting questions. For instance: what, if anything, constrains the possibility of deliberate syntactic change? How common are such changes, as opposed to unconscious, gradual syntactic changes? And what challenges, if any, do deliberate syntactic changes pose for theories of syntactic structure?

  Sally Thomason's slides

• 2:10 - 2:50
  

  Emily Coppess (Chicago) & Acrisio Pires (Michigan)

  The residue of syntactic change: Partial pro-drop in Old English?

  Old English had null subjects, but the frequency of referential null subjects was very low, and they mainly occurred in restricted syntactic domains. We consider this apparent residual nature of null subjects in view of formal approaches to partial pro-drop.

  Emily Coppess and Acrisio Pires's slides

• 2:50 - 3:30
  

  Mark Hale (Concordia)

  What changes in 'syntactic change'? Syntactic reconstruction reconsidered

  This paper examines the nature and properties of the entities whose change is generally labeled 'syntactic change' and explores the degree to which such entities might be successfully reconstructed. The data will be drawn from archaic Indo-European languages.

  Mark Hale's handout

• 3:30 - 4:10
  

  Andrew Garrett (Berkeley)

  Syntax and reconstruction: Assessing the Parametric Comparison Method

  A novel connection between work in historical linguistics and syntactic theory has been developed by Longobardi & Guardiano (2009), Longobardi (2012), and related work. They propose a Parametric Comparison Method that analyzes traits derived from formal analysis of syntax using the computational tools of biological systematics, with the goal of reconstructing phylogenies. In this presentation I assess that work from the point of view of a historical linguist interested in diversification and phylogeny.

• 4:10 - 4:50
  

  Mark Liberman (Penn)

  From 500 passages to 50,000 books: Creating and using a large-enough historical corpus

  For well over a century, scholars have been using historical corpora in systematic studies of morphological and syntactic change. Recently, the development of large parsed corpora has made an enormous quantitative improvement in the efficiency of such research, thereby changing the field qualitatively as well. However, an important problem remains. Thus the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) has 1,737,853 words in 448 texts (between 75 and 16,593 words per text, or 3,879 on average), covering the 210 years from 1500 to 1710. This is 8,275 words per year, or 82,750 words per decade on average. Despite the heroic labor involved in creating this resource, there remains a problem with sample size for those phenomena that don't occur at rates high enough to give a clear signal with reasonable temporal resolution at this scale. Tackling the problem of sample size will require corpora of a different sort.

  more details...

  Mark Liberman's slides