The fact that people are able to speak and write, and to comprehend texts (if often imperfectly), assures us that linguistic systems are able to operate for producing and comprehending texts. Therefore, a model of "linguistic structure" cannot be considered realistic if it cannot be put into operation in a realistic way. This principle, the requirement of operational plausibility, has also been mentioned by Ray Jackendoff (2002).
It is important to appreciate that the relational network hypothesis, though it is supported also by neurological evidence, can be arrived at and justified purely on the basis of linguistic evidence, including the fact that people are able to talk and to understand one another. Though obvious and abundant, this evidence tends to be neglected by many linguists, who work with theories of language that have no way of being put into operation for speaking and understanding.
Lamb (2001) now enumerates some pieces of this linguistic evidence, a baker’s dozen of items:
- Coexistent alternative analyses; e.g. "hamburger" (Lamb 1999: 233-236). The network allows ham - burger, and hamburg - er to both be present and to operate in parallel.
- Multiple parallel interpretation of (many) complex lexemes (cf. Müller 2000, Lamb 1999: 184-197). For example, the Chinese compound zhong 'central, middle' plus guo 'kingdom' is the name for China; but in its interpretation it also, simultaneously and in parallel, means 'middle kingdom'.
- Disambiguation of ambiguous words using linguistic and extralinguistic context. How connotations operate (Lamb 1999: 187-188).
- Context-driven lexeme selection (Lamb 1999: 190-194). For example, the selection of zoom (as opposed to the expected go) in the spontaneously produced Are you ready to zoom to the camera store? (Reich 1985).
- The interpretation of puns and other cases requiring simultaneous activation of double pathways; e.g. a talking duck goes into a bar, orders a drink, and says "Put it on my bill".
- Complex associations in slang lexeme formation. Eble (2000) gives the following example:
Sometimes sound provides the link in a set. With the popularity of African-American comedians came the form ho, a dialect pronunciation of whore, for 'a promiscuous woman'. The same sequence of sounds, spelled hoe, refers to 'an implement for tilling the earth', i.e. a garden tool. Thus ho and garden tool are current slang synonyms for 'a promiscuous woman'. (Eble 2000: 509)
- Phenomena involving association, such as literary allusions (e.g. to Hamlet by quoting) and Freudian slips. For example, the statement Something is rotten in the state of Florida conjures up Hamlet to people acquainted with this play.
- Degrees of entrenchment of idioms and other complex lexemes — accounted for by variability in the strengths of connections.
- Gradualness of learning — related to degrees of entrenchment. In the learning process, connections get strengthened.
- Slips of the tongue (cf. Dell and Reich 1980).
- Prototypicality phenomena. The conceptual category BIRD , for example, includes some members, like ROBIN, SPARROW that are more prototypical than others, like EMU, PENGUIN. The effects have shown up in numerous psychological experiments using such evidence as reaction time for deciding whether an item is or is not a member of the category. The relational network model provides a simple and direct means of accounting for the phenomena, by means of two devices that are needed anyway to account for other phenomena: variation in the strength of connections (thus the property of FLYING is strongly connected to the category BIRD), and variation in degrees of threshold satisfaction. Strength of activation, strength of connections, and number of activated connections all contribute to the speed and degree to which the threshold of a node is satisfied. It is important to notice that although these phenomena have been discussed in the literature for years, no means of accounting for them other than by means of a network model has ever been proposed.
- Realistic means of accounting for speaking and understanding. This one, of basic importance, covers a wide range of phenomena. The fact that people are able to speak and to comprehend one another cries out for explanation. The relational network model, whose origin in the early 1960s was motivated partly by this evidence, provides a simple and direct means of such accounting: by the 'travelling' of activation through the pathways provided by the network (Lamb 1999).
- On-line cognitive processing in conversation. This rich but neglected opportunity for study, again blessed by abundant but neglected evidence, has been explored by Cynthia Ford Meyer in three papers (1991, 1992, 2000), and in her dissertation. Strangely and sadly, her work has not yet encouraged others to undertake similar explorations. Here I will give one example, not from her work but from my own analysis (Lamb 1999: 202) of an actual courtroom exchange reported by Lederer (1987).
Attorney: Mrs. Jones, is your appearance here this morning pursuant to a deposition notice which I sent to your attorney?
Witness: No. This is how I dress when I go to work.
We can observe a number of phenomena that are readily accounted for by the relational network approach. The witness is evidently concerned about her appearance and believes that a woman's clothing contributes to her appearance. Beliefs are registered as conceptual subnetworks, and matters of present or ongoing concern register as weak activation in these networks. Such activation is increased by emotional stimulation. To this factor we add another: Unfamiliar lexemes or locutions are likely not to provide much conceptual activation if any, because the connections that would provide activation are weak or lacking. So the lexeme pursuant and the possibly unfamiliar expression pursuant to a deposition notice, although they were surely received by her phonological recognition system, probably didn't generate much activity in her lexico-grammatical system, therefore little or none in her conceptual system. In addition, any emotional affect aroused by someone's seeming to draw attention to her appearance would deflect attention that might otherwise be directed toward attempting to understand the passage beginning with pursuant. The factor of attention has a global effect on degrees of threshold satisfaction. As a result, that latter part of the sentence, which in an attorney's cognitive system provides strong contextual activation to one interpretation of the lexeme appearance (the intended one), fails to have such an effect in the woman's system, and the other interpretation has in any case already been activated by the time the phrase beginning with pursuant was received.
There is an opportunity for many more fruitful studies along these lines and those opened up by Meyer (1991, 1992, 2000). In any case, this brief survey suggests that considerable linguistic evidence exists for the hypothesis that the neurocognitive basis of a person's linguistic system is a relational network. These phenomena all support the network model, and no one has ever proposed an alternative means of accounting for them.
Lamb, Sydney M. "Linguistic structure: A plausible theory" in Language Under Discussion, published online June 2, 2016.