@inproceedings{serrell_property_1956,
	address = {Washington, {D.C.}},
	title = {On a Property of Natural Language and Its Use for the Design of Improved Machine Languages (Associative Machine Languages)},
	language = {English},
	booktitle = {Symposium on Advanced Programming Methods for Digital Computers: Washington, {D.C.}, June 28, 29, 1956},
	publisher = {Office of Naval Research, Department of the Navy},
	author = {Serrell, Robert},
	year = {1956},
	pages = {77--83}
}

Serrell really does mean “machine languages”: the sample “Experimental Associative Machine Language” that he provides in the paper uses “words” of 2 to 3 digits each.

  • The need to improve machine languages: In two ways: first, implementing better usage of internal memory capacity; second, implementing conversion processes so that programs developed on one type of computer can be used on another (77).

  • Machine languages do not include associative meanings: Machine languages do not include “associative meanings,” as natural languages do; rather, each instruction word into machine language has a meaning that is “unique, fixed, and quite independent of any association with other words in the language” (78).

  • Meaning in natural language determined by association (context): the meaning of words in a natural language is largely determined by its association with other words — that is, by context (78).

  • A correlation between relative frequency of occurrence of words in English and range of different meanings for each word: the most frequently used words in the English language have a wider range of meanings per word then less frequently used words: that is, there is a statistical relationship between the relative frequencies of words in English and the number of meanings preferred (77-78). The meaning of words in a natural language is largely determined by its association with other words — that is, by context (78). The most frequently used words in the English language have a wider range of meanings per word then less frequently used words: that is, there is a statistical relationship between the relative frequencies of words in English and the number of meanings preferred (77-78).

  • A correlation between relative frequency of occurrence of English words used in applied mathematics and range of different mathematical meanings for each word: if we consider the English words used in applied mathematics (such as the word “add” or the word “multiply,” both of which can be used to describe multiple processes: add two numbers, or add many numbers…), it is evident that like other natural-language words, they too have multiple meanings, and that here, too, there is a relationship between frequency of occurrence and number of meanings, so that we can describe the language of applied mathematics as an associative language as well (78).

  • The need for “associative machine languages”: machine languages with natural-language-like associative properties “should appreciably shorten the programs prepared the language,” since meanings currently supplied by individual, fixed instruction words could be supplied merely by associations (79). To design such an “associative machine language,” one might define a set of individual words with individual meanings, along with definitions of associative meanings, as set of syntactic rules, and perhaps a punctuation patrol (79).