A Deepness in the Sky

A Deepness in the Sky (Zones of Thought)
By Vernor Vinge 2000
704 pages

Vinge tells the the story of the Qeng Ho, a spacefaring civilization of traders spanning thousands of lightyears.  This story centers on the adventures of one trading mission to a unusual star system ,following the exploits of the legendary founder of the Qeng Ho.  The traders discover intelligent alien life, treachery, revolutionary physics and focus, a dangerously powerful technology.  Vinge creates an awesome tale of plotting, survival, technological development and cultural contact on a galactic scale. Its Foundations good.

Language Structure Is Partly Determined by Social Structure

Lupyan & Dale (2010)

Language Structure Is Partly Determined by Social Structure

The authors study demographic factors including population size, area spoken, and linguistic contact for over 2000 languages included in WALS.  They conclude that all factors are negatively correlated with morphological complexity measured across a variety of features (type, case, verb morphology, agreement, evidentials/possibility, negation/plurality, tense, aspect, mood, possession, articles, demonstratives pronouns).  They argue that exoteric languages, those with larger populations, which are spoken over a large area, or have heavy language contact tend to shift complexity from morphology to lexical items.
The authors propose the Linguistic Niche Hypothesis which states that language structures are subjected to different evolutionary pressures in different social environments.  Specifically they claim that esoteric languages have mostly L1 learners which can benefit from the excess redundancy provided by complex morphology while exoteric languages have a large degree of L2 learners due to their roll as interface languages.  L2 learners have difficultly learning complex morphology and so do not benefit from the added redundancy.  Highly exoteric languages, therefore, tend to have reduced morphological complexity creating higher overall fitness for the language by affording L2 learners an easier learning experience.
Lupyan & Dale present compelling models showing the general trend they describe.  Their hypothesis relies on the strong link between language structure and learning which needs further exploration.  Their notion of language fitness and redundancy (adopted from Nowak et  al. 2002) is intriguing but relies on an incredibly simplified model of the relationship between form and meaning.  More sophisticated models based on Information Theory should supplant these simplistic models of language redundancy,complexity, and learnability.  This would allow an explicit connection to formal learning theory to be developed to further test the Linguistic Niche Hypothesis.

Basic Objects in Natural Categories

Basic Objects in Natural Categories – Rosch et al 1976

Experiment 9:
Children sorted pictures that were taxonomically related or grouped by superordinate sets.
Basic: 4 shoes, 4 chairs, 4 men’s faces, 4 cars
Superordinate: sock, shirt, pants, table, chair car etc.
Subject more likely to sort by taxonomy in basic condition especially at ages K and 1st grade.
Subjects given multiple tries to sort taxonomically .
Explanation for sorting lagged correct sorting by a few grades.
Young children struggle with superordinate sorting but accel at basic level sorting.

Experiment 10:
Subjects were given packets requiring labels at the Superoridnate, basic, or subordinate level to individually name each object
Overwhelmingly subjects responded with basic level names and not for the frequency of the name or lack of superordinate knowledge.

Experiment 11:
Corpus of language used by stage I language learner included almost exclusively basic level nouns for concrete terms.
Children age 3 identify object with basic level names.
Experiment 12:
Signers of ASL asked if a sign existed for superordinate, basic and subordinate level nouns.
Significantly greater number of signs for basic level categories than superordinate or subordinate categories.
More signs for superordinate categories than subordinate categories.
Something weird about biological categories where “superordinate” category behaves like the basic level category across most experiments.

How do we delineate the set of attributes used for categorization and cue validity?
Does taxonomic depth vary with information structure?