Bryn Mawr College

Center for Science In Society

To facilitate the broad conversations, involving both scientists and non-scientists, which are essential to continuing explorations of
  • the natural world and humanity's place in it,
  • the nature of education,
  • the generation, synthesis, and evaluation of information,
  • technology and its potentials,
  • the relationships between forms of understanding.

Language: A Conversation

Meeting Notes
18 March 2002

Paul Grobstein (Biology), Ruth Guyer (General Studies, journalist), Eric Raimy (Linguistics), George Weaver (Philosophy), Sharon Burgmayer (Chemistry) 

One Summary View (prepared by Eric Raimy; views by other participants encouraged and can be sent either by email or posted using our working group forum area):

The meeting started late (1:00pm) with a smaller than usual group attending most likely caused by spring break 'catch up'... [Note:  I think this summary reflects the topics that arose accurately but I'm sure that the temporal sequence of the topics is wrong.  I don't think this affects the content of the discussion too much though.  Feel free to post to the forum where this is inaccurate.- Eric]

The entire discussion revolved around the following question posed by Paul:

"Does a thermostat desire to keep the room a certain temperature?"

The gut reaction of the group was to answer 'No'.  Paul was expecting this and proposed that his differing judgement (Paul thinks thermostats do desire to control the temperature in a room) is based on how he views the world.  Paul explained since he is a biologist he views all behavior in a 'mechanistic' way and this is why he feels that thermostats can have 'desires'.  According to this view, 'desire' means something along the lines of 'able to cause things to happen' so since the thermostat can cause the heat to rise or fall, it is appropriate and correct to attribute 'desire' to thermostats according to Paul.

The point that Paul was trying to make was that the groups restriction of 'desire' to only 'animate' things was wrong.  Furthermore, the fact that Paul has a different definition of 'desire' than the rest of us was meant to show that the Pinker style analysis of language was fundamentally flawed and thus doomed to failure. Both George and Eric responded to this challenge by explaining how linguistics treats and analyzes language.  George explained that most of what we were arguing about with respect to the correct meaning of 'desire' is the nature of the selectional restriction that 'desire' has.  Most words have selectional restrictions that require other words to have certain characteristics.  Selectional restrictions also generally deal with semantic qualities in that sentences that violate selectional restrictions are grammatical but semantically anomolous.  Most cases of anthropomorphization result from a selectional restriction being violated.

Eric picked up on this point and explained that part of the problem we were having in discussing this question was the fact that we were using English to analyze English.  The idea that we need a meta-language in order to analyze English was put forth.  In using a meta-language, a precise and formal definition of the selectional restrictions on 'desire' could be formulated which would not be subject to the confusion that was occuring in the group.

Paul at this point reiterated his objection to this approach because he felt that the difference in judgement on 'what can desire' was still not met.  The notion of 'ideolect' was then brought forward.  Linguistics recognizes that in all likelihood the grammar in each person's head is most likely distinct at some level.  Consequently everyone speaks an ideolect (e.g. an individual dialect).  The relevance of this point is two fold.  First, since Paul has a different ideolect than every one else in the room (and vice versa) it should not be a surprise that we can disagree about what a particular word means.  From a Pinker perspective, this is not a problem as long as we can describe each of the individual ideolects.  Paul's ideolect allows machines to have 'desires' while Ruth's ideolect requires a cognizant being to hold a desire.

The fragmenting of the meaning of particular words to be dependent on specific ideolects did not convince Paul of much.  As Ruth said, "Words need us", in that words do not have a life of their own.  It is only since we have language and have learned a particular word that it has any meaning.

Paul was still not satisfied.  George then explained the Russian view of language as just a set of objects which can be described through a model.  The important aspect of this view of language is that language is not a communication device but just a set of objects to describe.  This approach to language basically develops a model that describes occurring statements or grammatical constructions and rules out non-occurring ones.  This is one general goal of linguistic theory, to provide a model that describes all known human languages, rules out non-human languages, identifies what is common to all human languages and identifies what is unique about each human language.

Paul was still not satisfied with this answer and further pointed out that he believed that by attempting to make a model of language some essential aspect of language would necessarily be lost in the endeavor.  Eric replied to this that what linguists were doing was just simple scientific method.  Linguists would observe language, build a model to explain the particular phenomena, observe language again to evaluate the model.  Inevitably, the model is wrong so it is modified to be made less wrong and this scientific method is repeated.

The fact that much of the scientific analysis of language was mirrored in immunology was raised by Ruth.  The immune system has similar characteristics as language in that it is generative.  The immune system is able to produce novel anti-bodies in response to any new pathogen (PG: actually the immune system is CONSTANTLY producing novel antibodies; it amplifies the production of particular, appropriate ones in response to a new pathogen).  It thus appears that the immune system has an 'innate universal grammar' of pathogens which allows it to respond to any novel situation.  Paul noticed that in this discussion Ruth made the comment that "The immune system can forget that it has seen a pathogen" which raised his original point once again.  Ruth then proceeded to explain that immunologists don't really say that the immune system 'forgets' a pathogen but instead the field has a technical term for this phenomena.  It was noted that this was basically the same response to Kathryn and Anne's question about 'words desiring things' that Eric and George provided two weeks ago.  The general question about whether the form of the message and the message itself could be separated  was noted to have come up again but left aside for the present time.

Paul reiterated his concerns about this approach to language.  Eric then asked if Paul's objection was about language specifically or about scientific method in general.  The objection could be specifically against language since the study of and observation of language is done using language and it is not clear that the development of a meta-language would remove this fact.  It may be impossible to study language because we have language as our only tool to use here.  It was pointed out though that Paul's objections could be modified and tailored to an argument against the scientific method being used in immunology though.  Paul was unsure of exactly where his objection lay.

Time had expired at this point.  The group agreed to pick up with Pinker again in two weeks and to also discuss the color and language.