Blair Bolles blog Babel's Dawn has just posted "More Evidence against Grammatical Universals" in which he continues the arguments about whether the human brain evolved with in-built rules for language grammar. In particular he quotes the letter "Evolved structure of language shows lineage-specific trends in word-order universal" recently published in Nature by Michael Dunn, Simon J. Greenhill, Stephen C. Levinson and Russell D. Gray, which carries out a statistical examination of word order in a large number of languages.
As the CODIL language is based on semantic associations rather than syntax the research is encouraging me to think about the possible features of CODIL which could be considered to model human thinking. As a result I posted the following comment to Blair's blog:
Thank you for bringing the letter in Nature Evolved structure of language shows lineage-specific trends in word-order universal to my attention as the findings are very relevant to my own re-examination of my earlier work on CODIL in terms of its relationship with evolution of human intelligence and language.
CODIL was designed as a language for communicating between humans and a proposed electronic information processing system with (in stored program computer terms) a very unconventional architecture. The key point was that the language was symmetrical - the language used by the human was to be directly interpreted by the hardware, which could tell the human what it was doing, and why, in the same language. A computer simulation test program was tested on a variety of applications and the approach was shown to be feasible for handling open ended information processing tasks where human action is essential and it was difficult or impossible to pre-define the problem in a conventional algorithmic manner. The approach was unsuitable for problems involving comparatively simple algorithms involving large arrays of numbers and no dynamic human interaction - i.e. for applications far removed from the information skills needed by pre-civilisation human beings.
At the time the work was perhaps too concerned with the computer aspects of the task and aspects relating to human linguistics were never explored, and it is only now, in retirement, that I have started to look as this aspect of the research.
The CODIL language was little more than a list of set names (which would be the human user's words) and partitions of sets (equivalent to saying a "House" is a "Building" or that the "Date" was after "1900"). These "items" were held as lists of lists within an associatively addressed memory. There was a very simple but highly recursive processing routine, called the "Decision Making Unit" which compared the "current item" with an active list of items called "The Facts" which was meant to model human short term memory.
The relevance is that the system was based of a very strongly semantic model of the task, with an virtually complete absence of syntax. The only syntax rule used by the Decision Making Unit was that the last item of a list was true if all the preceding items in the list were true in the context defined by the Facts. Meaning was conveyed by the choice of the set names. For instance a marriage might be described as "EVENT = MARRIAGE; BRIDE = JANE; GROOM = JOHN" which could be interpreted by a human as "John married Jane" while the information processor could recognise Jane as a "PERSON" because the recusrive structure allowed synonyms to be recognised.
In fact this comment highlights the point that one area of research which was not examined prior to the unfortunate abandonment of the CODIL research was to see how the CODIL language could be used to generate natural language sentences. Obviously different natural languages would need different "interpreters" to convert the underlying semantic model into statements which conformed to different word order rules, etc. However if it can be done there is no need to invoke any significant evolutionary advances in the brain to progress from a semantic to a syntax based language - as the advance could be explained by learning.