Since idiolects and dialects may differ somewhat independently in any of their constituent systems or subsystems and also in the ways these systems articulate with one another, it behooves us to ask which kind of difference is more likely to be productive of mutual misunderstanding or mutual unintelligibility: difference within any one system or difference in the articulation of two systems with each other. Offhand, we expect that difference in the articulation of two systems -- say the phonological and the morphological, or the morphological and the semantic -- will be more quickly productive of misunderstanding than will difference within any one system itself. This judgment seems reasonable, also, in light of what we know about systems in general. Variation within any subsystem has less effect on the larger system of which it is a part than does variation in the way the several subsystems articulate with one another; for the structure of the larger system is most immediately characterized by the pattern of subsystem articulation.
Given little difference in the other systems, two speakers can differ considerably in their phonological systems without seriously impairing their ability to understand one another. We may have to take a little time to get used to one another, but most of us have little difficulty. understanding people who speak our language with quite thick foreign accents. People learning a second language tend to use the distinctive features of their first language as a basis for distinguishing and pronouncing the phonemes of the second. Consequently they miss some phoneme distinctions entirely, just as a native German speaker tends, when speaking English, to fuse the phonemes // (voiced th) with /d/ and /þ/ (voiceless th) with /t/.
Some differences in morphological systems can also have little effect on mutual intelligibility. The rules governing the height of final vowels in compound words illustrated earlier for the Romónum dialect of Truk vary considerably among Truk's several dialects. The same nine vowel phonemes (see footnote 6) appear in all of these dialects, but the rules of vowel harmonics differ from one to the next, so that we find sópwótiw ("lower district") as well as sópwu-tiw, and sópwo-wu ("outer district") and sapwo-wu as well as sópwu-wu or sópwu-u.
Misunderstanding is bound to develop rapidly, however, with differences in the semantic system, that is, with the way concepts are mapped into morphs, words, and other expressions. By assigning to ordinary words in the Trukese language a set of different denotations, members of a traditional group of political specialists in Truk are able to speak in public and convey messages to one another that are not understood by the uninitiated. Thus the Trukese word aaw ordinarily denotes the large tree Ficus carolinensis, but in this special argot it denotes the son of a chief, ordinarily referred to by another expression. Speakers of this argot use Trukese phonology, morphology, and syntax with only minor alterations, but by assigning special meanings to the words they use, they make themselves unintelligible to other speakers of Trukese. By the criterion of mutual intelligibility, they speak a different language.
When we think of learning a new language, although we recognize that it may involve learning some new rules of grammar, most of us think of the task as primarily one of learning a new vocabulary to represent the same old things. What we call a house in English is called maison in French and iimw in Trukese. We may later discover that the class of phenomena designated by house is not identical with the classes of phenomena designated by maison or iimw and that thinking in French or Trukese involves in each case somewhat different percepts and concepts than does thinking in English. But even if this were not the case, if French and English had the same phonology, the same patterns of morphological construction, and the same principles of syntax, and if the words in one language denoted the same things that the words in the other did, if at the same time the shapes of the words in the two were always different, the same shapes never designating the same things, we would regard them as different languages.
The matter of variance boils down to this, then: so long as we can recognize in the speech of another the code functions of our own idiolect, his speech is intelligible to us. If the denotations of his words are altered so as to have little correspondence with the denotations of the phonologically like words in our own idiolect, or if the phonological shapes of his words with the same denotations are altered beyond our ability to recognize them, in either case mutual intelligibility is lost. Considerable variance is possible within these limits without such loss. Two people speak the "same" language, then, if the variance between their idiolects does not exceed these limits.
But this is not the end of the matter. The problem of definition is actually more complicated. When space technicians start talking about space technology or linguists start talking about technical matters pertaining to language, a layman finds himself unable to understand what is being said. Does this mean that the layman and space technician speak different languages? In one sense it does. The layman recognizes that he has to learn the "language" of space technology, the special vocabulary and the concepts that comprise its denotations. But in another sense, the layman and the specialist both speak English (or whatever), for they communicate readily about nontechnical matters; and even when he talks about his specialty, the space technician uses ordinary English words according to English grammar intermixed with a specialized vocabulary. There is obviously a difference between the situation where two people both have knowledge of similar subjects but cannot communicate with each other about them and the situation where they can communicate about subjects of which they both have knowledge but not about other subjects. In the former case, the two speak different languages; in the latter case they speak the "same" language, but with varying degrees of competence in the several subject matters for which it serves as a code.