See the Zellig Harris bibliography for complete citation details.
[S]ince it is impossible to define the elementary entities and constraints of a language by recourse to its metalanguage (since the metalanguage is itself constructed from those entities by means of those constraints), it follows that the structure of language can be found only from the non-equiprobability of combinations of parts. This means that the description of a language is the description of contributory departures from equiprobability, and the least statement of such contributions (constraints) that is adequate to describe the sentences and discourses of the language is the most revealing.
"The background of transformational and metalanguage analysis", The Legacy of Zellig Harris: Language and information into the 21st Century Vol. 1, pp. 8-9.
With an apparatus of linguistic definitions, the work of linguistics is reducible, in the last analysis, to establishing correlations. […] And correlations between the occurrence of one form and that of other forms yield the whole of linguistic structure. The fact that these correlations may be grouped into certain patterned regularities is of great interest for psychology; but to the pattern itself need not be attributed a metaphysical reality in linguistics. Gray speaks of three aspects of language […], basing himself on the langue-parole dichotomy of de Saussure and many Continental linguists. This division, however, is misleading, in setting up two parallel levels of linguistics. 'Parole' is merely the physical events which we count as language, while 'langue' is the scientist's analysis and arrangements of them. The relation between the two is the same as between the world of physical events and the science of physics. The danger of using such undefined and intuitive criteria as pattern, symbol, and logical a prioris, is that linguistics is precisely the one empirical field which may enable us to derive definitions of these intuitive fundamental relationships out of correlations of observable phenomena. (Harris 1940:704)
Review of Louis H. Gray, Foundations of Language (New York: Macmillan, 1939). Language 16.3:216-231; p. 704 of repr. in Papers in Structural and Transformational Linguistics (Harris 1970:695-705).
The interest […] is not in investigating a mathematically definable system which has some relation to language, as being a generalization or a subset of it, but in formulating as a mathematical system all the properties and relations necessary and sufficient for the whole of natural language.
Mathematical Structures of Language (Harris 1968:1)
There is an advance in generality as one proceeds through the successive stages of analysis [from structural linguistics, to transformational analysis, to operator grammar]. This does not mean increasingly abstract constructs; generality is not the same thing as abstraction. Rather, it means that the relation of a sentence to its parts is stated, for all sentences, in terms of fewer classes of parts and necessarily at the same time fewer ways ('rules') of combining the parts, i.e. fewer constraints on free combinability (roughly, on randomness). But at all stages the analysis of a sentence is in terms of its relation to its parts - words and word sequences - without intervening constructs.
Papers on Syntax (Harris 1981:v)
[We must avoid] the undesirable effect of forcing all languages to fit a single Procrustean bed, and of hiding their differences by imposing on all of them alike a single set of logical categories. If such categories were applied […] it would be easy to extract parallel results from no matter how divergent forms of speech.
Methods in Structural Linguistics (1951[1946]:2)
We cannot describe a language without knowing how our description can in turn be described. And the only way to avoid an infinite regress is to know a self-organizing description, which necessarily holds also for the language we are describing even if we do not use this fact in our description.
"The background of transformational and metalanguage analysis", The Legacy of Zellig Harris: Language and information into the 21st Century Vol. 1, p. 10.
Operator grammar reveals a sharper relation between the structure of a sentence and its information by specifying and ordering each departure from equiprobability. Some of these departures (mostly universal ones) can be seen by their structure to be information bearing (in a sense related to that of mathematical information theory); others (mostly in particular languages or language families) create regularities and irregularities that are not substantively informational. By-products of the whole theory are the status of the operator-argument system as a mathematical object (a necessity for the stability of language structure), and the picture of language as a self-contained, self-organizing, and evolving system. All of these properties, including the lack of an external metalanguage, can be looked upon as being expectable, given the use (function) of language, and the fact that it was not created in any conscious plan.
"The background of transformational and metalanguage analysis", The Legacy of Zellig Harris: Language and information into the 21st Century Vol. 1, p. 13.
[G]iven what we know about the status of 'truth' in logic and about the alternative descriptions of structure, a theory should not be thought of as presenting the final truth, but only as organizing the results of certain methods of analysis, 'true' as far as it goes.
"The background of transformational and metalanguage analysis", The Legacy of Zellig Harris: Language and information into the 21st Century Vol. 1, p. 9.
Others have made hypotheses about language, more or less plausible; Harris made discoveries. This is derived loosely from "That the picture of language presented here is similar in some respects or others to various views of language […] does not obviate the need for reaching the present theory independently by detailed analyses of language. This is so partly because a more precise and detailed theory is then obtained, but chiefly because we then have a responsible theory in direct correspondence with the data of language rather than a model which is speculative no matter how plausible."
Language and Information(Harris 1991:6n.1)
Compare also the characterization by Sager and Nhàn (in The Legacy of Zellig Harris: Language and Information into the 21st Century Vol. 2, p. 79):
Zellig Harris ’s work in linguistics placed great emphasis on methods of analysis.His theoretical results were the product of prodigious amounts of work on the data of language,in which the economy of description was a major criterion.He kept the introduction of constructs to the minimum necessary to bring together the elements of description into a system.His own role,he said,was simply to be the agent in bringing data in relation to data.
Outsiders could see the genius and great insight into the workings of language that guided the application of rigorous methods of analysis,leading as they did to the formulation of grammatical systems,and ultimately to a penetrating theory of language and information (Harris 1982,Harris 1991). But it was not false modesty that made Harris downplay his particular role in bringing about results,so much as a fundamental belief in the objectivity of the methods employed.Language could only be described in terms of the placings of words next to words.There was nothing else,no external metalanguage.The question was how these placings worked themselves into a vehicle for carrying the ‘semantic burden ’of language.
Yet Harris ’s work did not start with a big question and search directly for the answer.His commitment to methods was such that it would be fair to say that the methods were the leader and he the follower.His genius was to see at various crucial points where the methods were leading and to do the analytic work that was necessary to bring them to a new result.