Summarine

Computer simulations of language change notes

This website collects my personal notes on Computer simulations of language change. These notes are provided to bring full transparency to my research process. Of course, since they are only notes, they do not reflect my final thoughts on a topic, and should not be interpreted as such. To read finished papers, please consult my website. Do not use these notes as a basis for your own scientific research. Start from high-quality, peer-reviewed scientific literature instead.

Does innovation need reanalysis?

Hendrik De Smet

Syntactic reanalysis has been claimed to be the only mechanism capable of explaining syntactic change. However, the concept of syntactic reanalysis is flawed. It insufficiently accommodates gradience in synchronic grammar and in language change, and depends too heavily on ambiguity as a cause of change. Alternative mechanisms exist to account for innovation that do not suffer from these problems. At the same time, the problem of explaining syntactic innovations is partly tied to models of language that overstate the role of syntax. Part of the problem therefore disappears under different theoretical starting assumptions.

Introduction

language change (p. 23)

  • happens as a side-effect of language users’ pursuit of other goals (Keller 1990; Joseph 1992; Berg 1998; Haspelmath 1999; Fischer 2000; Croft 2000)
  • a specific innovation is likely only to the extent that it can somehow arise from how language is routinely used
  • i.e. how language is acquired, how it is mentally represented, how utterances are processed, or how they are produced and interpreted …

not one explanation (p. 23)

  • different models of language disagree on [how language changes] and, as a result, predict different types of change
  • ⇒ different kinds of diachronic explanations for phenomena

↓ here

syntactic reanalysis (p. 24)

  • the privileged mechanism to account for syntactic change

The explanatory value [of syntactic reanalysis] has been called into question, however, both on empirical and on theoretical grounds (Haspelmath 1998; Fischer 2007; De Smet 2009; Bybee 2010).

Reanalysis

Definition

precondition (p. 24)

  • there exists such a thing as abstract syntactic representation

↓

syntactic reanalysis (p. 25)

  • a mechanism of change that assigns a new underlying structure to a surface sequence without overtly modifying that sequence

Langacker’s definition

I will define “reanalysis” as a change in the structure of an expression or class of expressions that does not involve any immediate or intrinsic modification of its surface manifestation. Reanalysis may lead to changes at the surface level […] but these surface changes can be viewed as the natural and expected result of functionally prior modifications in rules and underlying representations. (Langacker 1977, p. 58)

Effects

surface changes? (p. 25)

  • can result from reanalysis
  • not immediate, but can result from the deeper structural changes
  • ⇒ actualisation

The reanalysis itself is covert until some recognizable modification in the forms reveals it. (Hopper & Traugott 2003, p. 50)

Properties

1. abrupt (p. 26)

  • reanalysis is abrupt (in a qualitative sense)
  • there is a shift between two representations that are radically distinct, with no in-between alternatives
  • suddenly, the structure has changed

in contrast, actualization is commonly seen as gradual, in the sense that it potentially involves multiple small surface changes that do not take place simultaneously. (p. 26)

2. ambiguity (p. 26-27)

  • reanalysis takes place through ambiguity
  • some surface sequences allow two alternative analyses that the shift from one analysis to the other can take place

↳ why does it happen? (p. 27)

  • language acquisition
  • balance between transparency and code simplicity (Langacker 1977)
  • avoidance of opaque grammars (Lightfoot 1979)
  • the application of linguistic universals (Andersen 1973; Anttila 1989)

Problems

Abruptness

Grammaticalisation

Key idea: a surface form switches abruptly from one syntactic representation to another

grammaticalization (p. 28)

  • implements structural change gradually rather than abruptly
  • ⇒ this implies that many purported cases of reanalysis (i.e. those involving grammaticalization) may violate the abruptness associated with reanalysis

For word class changes, Haspelmath (1998) argues that transitions from one word class to another tend to be stepwise, with an item assuming the formal properties of its new word class one by one.3 His most elaborate example comes from Kortmann & König’s (1992) study of prepositions developing out of verbs in European languages. The study demonstrates that some of those prepositions (such as English during or pending) show many prepositional properties whereas others (such as English preceding or facing) show few.

If reanalysis is abrupt, then why is actualisation gradual?

grammaticalisation is gradual (p. 30)

  • e.g. ENG semi-modals: gonna, gotta, hafta, wanna
  • assumed to be gradual

↓

syntactic reanalysis (p. 30)

  • gradual as well?

Frequency effects

frequency-driven decrease in compositionality (p. 30)

  • give a likely explanation for certain structural changes (p. 33)
  • ⇒ could apply to reanalysis as well!

Repeated exposure to a particular phonological pattern (be it one we classically call a morpheme, a word, or even a sequence of words) increases speed and fluency of processing of the pattern. […] As this process is repeated, any tendency toward compositionality within the pattern is gradually reduced, leading to words and word sequences losing their compositionality if they are of high absolute or relative frequency. (Bybee & McClelland 2005, p. 396)

  • there is no cut-off point at which want to ceases in the eyes of the language user to consist of two separate elements and is fused into one (see also Bybee 2004, 2006, 2010)
  • evidence from grammaticalization primarily demonstrates the relation between frequency and gradual formal fusion (Krug 2000)
  • That the loss of underlying compositionality is also a gradient phenomenon is mainly supported by research in other domains, notably morphology (Hay & Baayen 2005)

(p. 30)

Hybrids

hybrids (p. 31)

  • manifest conflicting behavioural features
  • evidence that some surface forms instantiate different underlying structures at the same time
  • commonly found in actualization processes, indicating that instances of a reanalyzed item need not resort either under its old or its new analysis
  • old and new analyses can simultaneously influence the item’s behaviour (De Smet 2012a)

The idea that a form instantiates either one structure or another is thereby contradicted. (p. 31)

A lot of

  1. original sense: head noun → a set of articles to be sold in batch
  2. new sense: a quantifier → indicating a large amount of something

↓

  1. a lot of goods is to be fitted out for auction (1853, COHA)
    • agreement with head noun → singular
  2. a lot of parchments about the chair / Are handed to witnesses then and there (1859, COHA)
    • agreement with head noun → plural

(p. 31)

↳ hybrid form (p. 31)

  • if lot is part of a quantifier, it should not combine with adjectives
  • if lot is a noun, it should not relinquish its head status yet
  • but: look in 👁 ↓
  1. an awful lot of people are on medication who don’t need it. (1993, COHA)
    • behaves as a noun in selecting a modifying adjective (awful) but as a quantifier in being transparent to agreement

(p. 31)

Ambiguity

Ambiguity cannot exist by itself

problem with ambiguity (p. 33)

  • a surface element can only be ambiguous between two interpretations if those interpretations are independently supported (Fischer 1988, 2007; De Smet 2009)

The form [soʊl] in (6) is ambiguous between the meanings ‘bottom of a shoe’ and ‘spiritual part of man’ because both those meanings are separately attested for [soʊl] outside the context in (6).

  1. Truly, sir, in respect of a fine workman, I am but, as you would say, a cobbler. […] a mender of bad soles. (Shakespeare, Julius Caesar 1.1.10–14)
  • ↳ Supposing [soʊl] in English meant only ‘bottom of a shoe’, the pun in (6) would not work, since the ambiguity with ‘spiritual part of man’ would simply never arise.

↓ syntactic domain

  • predicative adjective, for-NP and to-infinitive in (7a) is structurally ambiguous because both of its readings are independently supported
  1. a. But in general it was good for mankind to look outward to the stars. (BNC)
    • … iti was [good for mankind] [to look outward to the stars]i.
    • … iti was [good] [for mankind to look outward to the stars]i.

b. Which we believe is good for our customers. And good for our business.

c. [a horse’s] sense of smell must function for it to assess horses it has not
met before (BNC)

  • ↳ from contexts like (7b–c) (and various others) it follows that the for-NP in (7a) could be the benefactive of good or the subject of the following to-infinitive. But without the contradictory evidence from (7b–c), (7a) would not be ambiguous at all

↓

chicken and egg solved (p. 34)

  • a form can only be ambiguous if its conflicting interpretations already exist
  • ⇒ ambiguity cannot be a source of innovation

More likely, it is innovation that is a source of ambiguity. (p. 34)

Syntactic theoretic explanations

1. language acquisition reanalysis (p. 35)

  • assumption: language users have access to syntactic structures beyond what is attested in their specific language
  • child assigns structure to the utterances it is exposed to, homing in on a possible grammar to match its parents’ linguistic output
    • but: may arrive at a grammar that differs from its parents’ albeit with approximately the same output
  • ↳ only really applicable if there are only a few structures, which is not the case

2. ambiguity between attested structural possibilities (p. 35)

  • only ambiguity between constructions that have been attested by the language user

↓

categorical incursion (p. 35)

  • one construction extends its range of application into the domain of (what used to be) another
  • uses analogical extension
 beep  beeped = weep x the book is brilliant  a truly brilliant book = the book is rubbish x \begin{aligned} \frac{\text { beep }}{\text { beeped }}&=\frac{\text { weep }}{\mathrm{x}} \\ \frac{\text { the book is brilliant }}{\text { a truly brilliant book }}&=\frac{\text { the book is rubbish }}{\mathrm{x}} \end{aligned}
  • ↳ does not seem to allow for genuine innovation (only mere “extension” of existing constructions)

↓

no ambiguity can arise unless both relevant interpretations are somehow already licensed by the grammar (p. 37)

Innovation

Innovation through analogy

analogy bad (p. 37)

  • poor reputation as a source of structural innovation
  • can at most account for items switching categorial allegiance, but not for the introduction of fundamentally new categories

↳ unjustified (p. 34)

  • any analogically-licensed addition to an existing category changes that category, to the point that similarity-based generalizations can give rise to wholly new categories
  • ⇒ the piecemeal addition of new members to a category can gradually work substantial changes (De Smet 2012b)

Innovation through other mechanisms

Other mechanisms of change can bring genuine innovation yet need not involve reanalysis!

1. frequency (p. 40)

  • can result in formal and structural coalescence

2. blending (p. 41)

  • the intertwining of parts of competing formulations (Coppock 2010, p. 38)
  1. that’s torrible! (terrible/horrible) (quoted from garrett 1980, p. 179).

Structural indeterminacy

Some apparent innovations are syntactically less innovative than they may seem. (p. 42)

↳ syntactic innovation as a theoretical construct (p. 42)

  • (cf. Haspelmath 1998)
  • need for reanalysis to a degree stems from the assumptions made by specific models of synchronic syntactic theory
    • assumptions are changed → the apparent logical necessity of reanalysis disappears
  • ⇒ changes may seem more innovative than they actually are simply because of the syntactic model used to describe them

Conclusions

Syntactic reanalysis is problematic (p. 43)

  1. it downplays the gradualness of change and overestimates the role of ambiguity
  2. it conflates different underlying mechanisms of change as well as different types of outcomes
  3. syntactic reanalysis is not the only way to explain grammatical innovation in language change
  4. syntactic reanalysis is at least in part tied to models of grammar that are intolerant of syntactic indeterminacy

↓

syntactic indeterminacy (p. 43)

  • allows a new interpretation of various grammatical changes
  • functionally innovative, but not necessarily syntactically innovative.

questions (p. 44)

  • 👁 ↓
  1. Should reanalysis be abandoned altogether?
    • It’s useful: it describes a recurrent constellation of change in which functional specialization can take place without immediate surface modification
    • Even so, the above also suggests that explanations for changes describable as reanalyses may have to take recourse to very different underlying mechanisms, which may ultimately challenge the usefulness of even a purely descriptive application of the term.
  2. The alternatives to reanalysis suggested here are only operational to the extent that (among other things) the relative role of similarity-based generalization and structural indeterminacy in synchronic grammar is clarified
    • The question is only beginning to be explored of how much similarity is needed for language users to spot a potential new generalization (Petré 2012)
  3. The type of alternative explanations proposed here have hardly been put to the test systematically
    • Their application to a much wider variety of documented changes is ultimately the only way to evaluate their validity and find any shortcomings

Computer simulations of language change notes

This website collects my personal notes on Computer simulations of language change. These notes are provided to bring full transparency to my research process. Of course, since they are only notes, they do not reflect my final thoughts on a topic, and should not be interpreted as such. To read finished papers, please consult my website. Do not use these notes as a basis for your own scientific research. Start from high-quality, peer-reviewed scientific literature instead.