WORKING BACKWARDS FROM MODERN LANGUAGE TO PROTO-GRAMMAR

1
WORKING BACKWARDS FROM MODERN LANGUAGE
TOPROTO-GRAMMAR

• Sverker Johansson
• School of Education Communication
• University of Jönköping, Sweden

2
Our early ancestors lacked syntax today we
have it.Somehow it emerged during our evolution.
How?
Plenty of proposed precursors, from social
scripts to navigation. May provide cognitive
tools for syntax but what came next?
3
Start at the other end of the problem!

• Possible proto-languages can be identified by
  disentangling modern language, removing one
  component at a time.
• Which aspects of modern language can be removed,
  without the whole system becoming non-functional?
• Removability from modern language entails
  addability to a proto-language.

4
What are the components of modern language?

• Imprudent to assume the truth of any particular
  grammar theory too many of them out there.
• Better to start with a few fundamental properties
  of language, on which there is consensus across
  different theories of grammar.

5
Universal properties of modern language

• 1. Structured. Not just a random bunch of words
  relations between words somehow indicated.
• 2. Hierarchical. Levels of structures within
  structures.
• 3a. Flexible, in the transformational sense that
  structures can be moved around.
• 3b. Recursive. The same structure may recur
  inside itself at different hierarchical levels.

Properties ordered with least removable first.
6
Structured

• Without Structured, there is no grammar.
• All the other properties require Structured in
  order to be meaningful.
• Language with Structured but none of the other
  properties is conceivable e.g. child language
  at the two-word stage.
• Necessary early addition to proto-language.

7
Hierarchical

• Hierarchical requires Structured, obviously, as
  Hierarchical operates on structures.
• Grouping words into headed units, and having
  structural rules operating on units as a whole,
  may be the main point with Hierarchical.
• Having headed units works fine without units
  containing the same kind of units as substructure
  Recursive not needed for Hierarchical.
• Having headed units works fine without having the
  option to move around units Flexible not needed
  for Hierarchical.
• Intermediate-level addition to proto-language.

8
Flexible

• Flexible may add spice and expressivity to
  language, but nothing fatal happens if its
  removed.
• Flexible requires Hierarchical, as entire headed
  units are typically moved around.
• Flexible is pointless without Structured.
• Candidate late addition to proto-language

9
Recursive

• Recursive likewise can be removed, without the
  rest of the grammar collapsing. A recursion-less
  language is simpler, without subordinate clauses
  and stuff, but still functional.
• Recursive requires Hierarchical, with multiple
  levels of structure within structure.
• Candidate late addition to proto-language

10
Recursive II

• Recursive key feature in the model of Hauser
  Chomsky Fitch (2002) but see also Pinker
  Jackendoff (2005), as well as Parkers talk here.
• A modern language apparently without recursion
  actually exists Pirahã (see Parkers talk here)
  proof positive that Recursive is a removable
  feature.

11
Recursion in theory and in practice

• In linguistic theory
• Structures may be recursively nested an arbitrary
  number of levels.
• Analysed top-down.
• Theoretically unlearnable.
• Either you have recursion or you dont.
• In practice
• Rarely more than 2-3 nested levels in natural
  speech.
• More levels difficult to process.
• Commonly built bottom-up.
• Different types of recursion.

12
Types of recursion

• Central-embedded true recursion
• The rat the cat the dog the man owned chased bit
  squeaked.
• Grammatical in theory but who would ever say
  anything like that? And whod understand it?
• We do use true recursion but sparingly.
• Edge-attached tail recursion
• The man owned the dog that chased the cat that
  bit the rat that squeaked.
• Easier to process, and more common.
• Iteration repetition without embedding.
• Round and round and round she goes.
• Not really recursion, even though it may be
  described by recursive phrase structure rules.
  (cf. Parkers talk)
• Any reason to believe that these all emerged
  together?

13
Possible sequence of proto-languages

• Non-structured. Isolated words. Relations
  between words from semantics only.
• Structured. Sentences with a few words, but
  without substructure. Relations between words
  indicated by grammatical markers (most likely
  word order).
• Hierarchical structure. Words grouped in headed
  units, but no recursive embedding.
• Recursive syntax (possibly in several substeps).
• Flexible may be added either before or after 4.
• Grammaticalization processes will become
  operative at some stage, adding morphology and
  function words. (cf. Hurford (2003) Deutscher
  (2005)
• 7. Modern human language.