__ _ ________ __ _______ ON A LANGUAGE OF THOUGHT by George Xixis May 3, 1987 George Xixis 1 In the continuing attempt to comprehend the nature of the mind and intelligence scholars from the fields of experimental psychology, linguistics and artificial intelligence have banded together in the studie of "cognitive science." Perhaps as a result of the type of investigators involved, the notion of thinking as symbol manipulation has permeated cognitive science. Intelligence, postulates Newell has to satisfy aset of obvious constraints. The most obvious and expedient way for this to occur is to argue for a language of thought. The arrguments for this language of thought come in two main flavors. Firstly, there is the systematicity argument. Given that one understands the sentence "Mary hit John" then one can understand the sentence "John hit Mary". One therefore postulates a combinatorial processing system that can explain this. The second, mildly compelling argument for a language of thought is the "Number argument". Take for example a twelve digit number and multiply that by another twelve digit number. The result is a twenty-four digit number. There are 10 ^48 thinkable sentences of this form. How can we represent 10 ^48 distinct thoughts in a limited capacity device such as the human brain without postulating some representational system of a language of thought. These arguments are far from perfect though. For the systematicity argument one might ask why is it that we understand the sentence "John hit Mary" iff we understand "Mary hit John". George Xixis 2 Fodor's answer is that this is true because we can think "Mary hit John" iff we can think "John hit Mary". This substantiation tends to reverse matters. It end up explaining language of thought in terms of public language mastery . Likewise, the number argument has similar problems. We know the arabic numeral system and that is why that particular argument holds. This development of "a language of thought" runs into more problems when one tries to apply the concepts to unconcious thought. When one has to deal with the reality of total internal repressentations that have no obvious external reference this interpretation breaks down. The argument is keyed to concious use of public representations, consequently nothing can be said about the internal milieu. Another problem with the system is the error question. Given that we do think in terms of a language of thought would we not expect errors in the system. Given that one has the sentence ______ ___ "where is my dog in their wonder box wouldn't be expected that at ___ least on occasion something of the form "I will look for my log appear in the relevant intention box. The counter to this argument is that we have no proof that internal language represenatation has to be exactly like the external/public domain represenattions. Errors do occur but they are more often of they type "where is my dog"--> "I will look for my cat" It is quite conceivable that the internal system is not phonetically arranged, but somehow semantically arranged.