Download - Complexity in Language

Transcript
Page 1: Complexity in Language

Complexity in Language

Why don’t languages evolve toward efficiency?

Page 2: Complexity in Language

“As they evolve, things become more efficient.”

• Efficient operations, tools, methods, etc. should drive out those that are difficult and costly.

• Language has been around long enough that it should have shed arbitrary, encumbering, opaque, redundant, and just plain weird features and honed those that contribute to precision and clarity.

Page 3: Complexity in Language

What counts as `arbitrary’?

• The tale of the human children and the chimp children

Page 4: Complexity in Language

Most linguistic categories match semantic notions

• a ‘cat’ is a cat• tense is concept of time• pronouns map to persons• agents and objects are agents and objects• et cetera

Page 5: Complexity in Language

Languages choose and differ in which categories will be required

• Some languages mark tense (actual time: English) while others mark aspect (the way time unfolds: Yoruba).

• Some languages build into words the relations between agent and object (English) while others mark them overtly (Salish).

• et cetera

Page 6: Complexity in Language

Arbitrary categories are not grounded in semantics

• Gender• Verb classes

Page 7: Complexity in Language

• Gender is the compulsion to place nouns into classes. It is not necessary.

• Genders that are putatively based on some semantic notion (e.g. natural sex) collapse into arbitrary assignment fairly early.

• Gender nearly always involves recruiting other categories to display evidence of ‘agreement’.

Page 8: Complexity in Language

Spanish gender agreement

• mesa (fem)• la mesa amarilla • mano (fem)• la mano sucia• hombre (masc)• el hombre guapo• artista (masc)• el artista generoso

Page 9: Complexity in Language

Verb classes

• Verb classes are the arbitrary divisions of verbs into groups with different morphology, sometimes startlingly so, that marks the same linguistic category.

Page 10: Complexity in Language

Some English verb classes

• The default past tense affix –ed is used with the largest verb class

• Smaller classes often share phonological features.

• swim-swam-swum and its classmates (‘sing’)• bring-brought and its classmates (`think’)• Note how these classes are conflated in non-

standard dialects: brang; thunk

Page 11: Complexity in Language

• What is the past tense of ‘sneak’? (a fun group exercise)

Page 12: Complexity in Language

Cherokee is scary

• Cherokee has an impressive amount of arbitrary complexity.

• Phonologically: tone, nasalization, vowel length, stress, besides funky consonant clusters.

• Morphologically: 10 person/number pronoun distinctions with more than 30 outputs based on subject-object relations and conjugation class.

Page 13: Complexity in Language

• 5 verb stem classes yielding 28 alternations (depends on which linguist is counting)

• Position affixes reproduce 5 stem classes• Change of pronoun type depending on verb class,

conjugation class, and type of tense marker. (These are clearly independent semantically.)

• Operations are marked by using TOGETHER tone + stem class + conjugation class + type of affix + position of affix.

Page 14: Complexity in Language

• Position 1 = lexical root• + Position 2 +Position 3 (optional) +Position 4• Aspects: Aspects: Aspect:• imperfective andative habitual • perfective duplicative• punctual incipient Fused aspect: • iterative punctual past• Fused aspect: completive • imperfective/present venitive Other inflections:

• ambulative• experienced past (tense/evidential)

• Other inflections: reported past (tense/evid)• infinitive/hortative Valences: imperative (mood)• applicative future imperative (tns/mood) • causative future (tense)

Page 15: Complexity in Language

Examples of aspect manipulation

• (1) uu- áakhuyáthan-iílóòsk -vvʔi• 3SG-burp:PERF -ITER/IMPERF-

EXPER.PST• `He was burping repeatedly.’• (2) uùnii-wóonis -éesti• 3P -talk:PERF-FUT• ‘They will have talked.’ (Montgomery-

Anderson 2008)

Page 16: Complexity in Language

• (3) ini- -wóoniisk -óʔi• 1DU-talk:IMPERF-HAB• `You and I talk habitually.’ (Feeling &

Pulte 1975)

• But, Cherokee does not have gender!!

Page 17: Complexity in Language

Evidence from pidgins and creoles

• A pidgin is a language that is created quickly, by force, and by ADULTS.

• Speakers retain some grammatical and phonological features of their native (substrate) language, replacing vocabulary with the dominant (superstrate) language.

• In modern times, pidgins arose due to the slave trade, giving us a chance to examine what happens to human language in a peculiar laboratory.

Page 18: Complexity in Language

• Pidgins are characterized by radical loss of complexity: syntactic, phonological, and especially morphological.

Page 19: Complexity in Language

• A creole is a pidgin that has evolved, gaining child speakers.

• While very many African languages have tone, no creole does. Tone is an example of highly complex phonology.

• Creole languages have fewer vowels than either their substrate or superstrate languages.

Page 20: Complexity in Language

• Syntactically and morphologically, pronouns are often reduced to a single form (me, him) while tense and aspect are rendered with adverbs:

• ‘Him go yesterday.’

Page 21: Complexity in Language

Clawing back?

• Research question: are creoles gaining complexity?

• Answer so far: not much, not yet.• Children will perfectly acquire any language

that they are exposed to. Including pidgins.• Hypothesis: Creoles don’t gain arbitrary

categories because they have achieved efficiency.

Page 22: Complexity in Language

The social dimension

• Adults can’t acquire languages efficiently, but they are great inventors of minutiae.

• Adults, especially young ones, seek to imitate those they perceive to be powerful, attractive, and correct.

• Example: the pronunciation of ‘Iranian’ in the US.

• That god-awful creaky voice that young women use. (Alert – spreading to young men.)

Page 23: Complexity in Language

• While phonology and vocabulary change at a rapid rate, morphological and syntactic change is much slower.

• However, an innovative (read ‘incorrect’) form introduced by a person/group of status may gain currency and exist in tandem with an older form.

• Very often, an older form becomes less used and may eventually disappear, even becoming `wrong.’

Page 24: Complexity in Language

• English borrowed the Cornish (Celtic) use of an auxiliary ‘do’, eventually making it mandatory in both questions and negation, radically changing surface word order.

• Old form: Knowest thou John?• New form: Do you know John?• Old form: I know him not.• New form: I do not know him.• Adults did this. It probably took several centuries. The

actual syntactic difference is smaller than it looks.

Page 25: Complexity in Language

Some inferences

• We hypothesize that there is a limit to the complexity of natural language that children can acquire, but we haven’t found it yet.

• Deep arbitrary complexity is a sign of a very old language.

• There’s no reason to get rid of complexity if children are the ones who learn a language.

Page 26: Complexity in Language

• Speakers, especially adults, will incrementally change a language through serendipitous means, sustained by social pressure.

• When groups gain large numbers of adults who speak another language, there is pressure to lose arbitrary complexity.

Page 27: Complexity in Language

Predictions

• Creole languages will gain complexity, some of it arbitrary, slowly.

• Some very old language families (Algonquian, Iroquoian) will lose arbitrary complexity as child speakers become fewer and innovation disappears, while in-mixing of other groups becomes common.

• Very old but isolated languages (Georgian) will retain mind-boggling complexity.

Page 28: Complexity in Language

New research question

• Efficiency is not especially relevant to human language once it is acquired and is used by persons who speak the same one.