How new words become part of a language 


Repeated over and over again, this process reflects how people invent and share new words for objects: they constantly invent new words, yet can only use ones that others understand, so it keeps a lid on the number of words in use. 

How new words become part of a language 
• 15 October 2005 
• NewScientist.com news service 
• Mark Buchanan
 
WHEN unwanted email first came along, people invented different words for it, such as unsolicited email and junk email. But eventually "spam" became the word of choice to describe the phenomenon. 
It's a process that happens each time a new thing needs a name, but language researchers have struggled to model how it happens without a central decision maker. Now a computer model shows the process at work - and may give insights into how the first human languages emerged. 
Luc Steels of the Sony Computer Science Laboratory Paris in France and his colleagues studied the "naming game", a simple computer model that reflects how people invent words and use them. In the game, a group of "agents" live in a virtual environment with a number of "objects". Each agent makes up random names for the objects, and the agents then interact in pairs, trying to "talk" about those objects. 
In each interaction, one agent (the speaker) says its word for an object, while the second agent (the hearer) listens. If the hearer fails to recognise the word, it memorises it as a possible name for the object. But if the hearer understands the word, both agents retain this word in memory and ditch any others they have made up or heard. 
Repeated over and over again, this process reflects how people invent and share new words for objects: they constantly invent new words, yet can only use ones that others understand, so it keeps a lid on the number of words in use. 
The simulations showed that this is enough for the emergence of a unique shared vocabulary. In the model, each object always ends up being described by just one word (www.arxiv.org/physics/0509075). "The model is as simple as possible," says Steels. "But it captures the main ingredients of how a population develops an efficient communication system." 
So could a similar process have helped the historical emergence of human languages? 
"Absolutely," says linguist James Hurford of the University of Edinburgh, UK. But he emphasises that in addition to common words, human language also requires richer structures such as grammar, the emergence of which the model cannot yet explain.
“Repeated over and over again, the process reflects how people invent and share new words for objects” 
While Steels and colleagues hope to develop more complex models capable of evolving grammar, they already see potential applications in computing. For instance, programmers currently have to establish standards to get commercial or scientific databases to communicate effectively. It may soon be possible to get computers to talk to one another by letting them evolve a common language on their own.
From issue 2521 of New Scientist magazine, 15 October 2005, page 16 

Posted: Sat - November 19, 2005 at 09:58 PM          

©