Simulation Suggests Steps to Speaking of the Unseen

Language-users have a special problem in linguistic unreliability, i.e., the meaning the speaker intended may not match the meaning the listener takes. This problem was discussed at the conference on protolanguage (Torun, Poland) in a presentation led by Tao Gong, James Minett, and William S.-Y. Wang. In particular, the problem is most acute when language breaks from the here-and-now to address absent things (a phenomenon known technically as displacement). The speakers wondered if displacement existed at the protolanguage stage of evolutionary history and, if so, how was the reliability of linguistic information established? As there is no direct empirical evidence bearing on the question, the speakers presented simulated results and a theoretical framework for studying the question.

Language requires that both speaker and listener share a “common ground.” When speaking about the here-and-now they can both look about themselves, but in a displaced circumstance the common ground must be more explicit. If I say, “Barack Obama was on David Letterman this week,” the listener must know who Obama and Letterman are. The meaning of that simple on will be lost to anyone who does not know that Letterman has a television show. Was Obama riding some guy named Letterman? (Literally, on Letterman.) Was Obama berating Letterman? (Figuratively, on Letterman’s case.)

Non-linguistic (sensory) information is also unreliable. A snake might be so well camouflaged that I don’t see it.

The speakers proposed that displacement only becomes possible when linguistic information becomes more reliable than non-linguistic information. If I say to you, “Look, there’s a snake,” and you then see the snake, language has done more than your senses did by themselves.

Tao Gong’s team ran computer simulations determining when languages become reliable enough to refer to things outside the here-and-now. The critical variable in these simulations was the reliability of a cue (RC). For speakers, the RC is the probability with which they can express what is going on in the common ground. For listeners, the RC is the probability with which they can use events in the common ground to interpret the message. The simulations found that once an RC reaches a probability of 40%, a highly-understandable displaced language can emerge.

Languages do not have to originally have this RC. If generations of language-use leads to an increased reliability, then a language could acquire the ability to speak of things beyond the here-and-now. This simulation thus suggests a scenario:

1. Speech heavily dependent on here-and-now.
2. Speech becomes more intrinsically understandable.
3. Speech becomes so understandable it can displace the here-and-now.

The presentation was interesting because it offers a way to be precise. Now it is time to see if this math really describes what is happening when children acquire language or when people learn a second language. If observation and simulation can come together, we will be on to something.