This week we start looking at how language knowledge is acquired. Readers should revisit LS7 to LS10 to ensure they understand what the blog means when it talks about concepts, propositions, the mental network and progression through it.
Relating phonology to meaning
In all the posts up to now it’s been assumed that the hearer decodes language into conceptual propositions by identifying valid junctions between phonological words.
This is the simplest case – for example, P2 is beautiful, P3 is Poppaea and the M3 / R / M2 proposition delivered to cognition is POPPAEA / HASP / BEAUTIFUL. But more complication is common – for example the three propositions (for AGENT, THEME and GOAL) needed when a give form is encountered. And some syntactic phenomena can’t easily be explained unless it is assumed that the C for a particular word is changed to a different C if the word forms particular junctions (see LS26).
The simplest case is good enough for the purposes of this post.
If the hearer doesn’t already know that Poppaea is beautiful, hearing the phrase adds to his knowledge – perhaps conditional upon how trustworthy he believes the speaker to be.
The hearer does need to know the propositions PHONOLOGICAL beautiful / MEANS-X / BEAUTIFUL, PHONOLOGICAL Poppaea / MEANS-Y / POPPAEA and MEANS-Y / HASP / MEANS-X (where X signifies adjective and Y proper-noun).
If one or other of these three propositions is missing from the hearer’s knowledge, then the fourth – POPPAEA / HASP / BEAUTIFUL doesn’t register.
But suppose the hearer already has that proposition – let’s say in the form EMPEROR’S WIFE / HASP / BEAUTIFUL. One of the other three propositions could be missing and still the communication could be effective. For example, the emperor’s wife is salient to the interlocutors and her good looks are manifest to both but the hearer doesn’t know her name. Knowledge is still communicated to the hearer – not POPPAEA / HASP / BEAUTIFUL but PHONOLOGICAL Poppaea / MEANS-Y / EMPEROR’S WIFE.
P2 is Poppaea and M2 is EMPEROR’S WIFE. Maintaining the convention that shaded triangles are existing knowledge and dotted/unshaded triangles are what is delivered to cognition, the hearer has learned the lady’s name.
More rarely a rule of combination may be learned. For example, suppose it’s claimed the imperial couple are both gorgeous, someone may react incredulously saying Nero BEAUTIFUL? A hearer still learning about prosodic subtlety in the language may infer NERO / CONSIDERED NOT / BEAUTIFUL and thus picks up a C / R / C rule in which a stressed adjective dissociates its concept from the preceding noun, expressing conflict with the speaker’s preconceptions.
So, if the hearer has any three of the propositions in the P / M / C / R / C / M / P arrangement, then the fourth can be understood.
The most usual case is where the two P / M / C words and the C / R / C rule are known and what is communicated is a piece of real-world information. The information may not be profoundly true but it at least represents something about the speaker’s mental state.
Less usual is where it’s one of the other propositions that is missing from the hearer’s knowledge. It’s almost certain that the new knowledge imparted by the speaker is ‘true’ (at least for the language community that speaker and hearer belong to): using false lexical or syntactic information would be a very difficult, very unreliable way for a liar to operate.
Of course, language knowledge is rather abstract because the C and R concepts are difficult for a linguist to express and impossible for anyone else. A real-world M / R / M proposition might be learned in one encounter. But a P / M / C word or a C / R / C rule may take repetition before it is firmly established in the language user’s mind.
It’s not difficult to see how an infant acquires a vocabulary – PHONOLOGICAL beautiful / MEANS / BEAUTIFUL etc. An arbitrary phonological sign is understood to relate to a concept that has been experienced. Higher animals may also have the ability to acquire pairings from their experiences; for example, a dog may be trained to respond to its keeper’s commands.
What seems to distinguish homo sapiens is the ability to relate two different instances of MEANS (the C concept) in order to compose or to comprehend an ad hoc proposition. This poses a couple of questions.
The first is about the ability to relate: how are the R concepts acquired? The ones we’ve already seen in LanguidSlog are concepts like HAS PROPERTY and INSTANTIATED by; and AGENT, THEME and GOAL. Arguably these are innate. They certainly must have been in the minds of our pre-linguistic ancestors for language to have evolved to express them. But that doesn’t mean all R concepts must be innate. The example CONSIDERED NOT that was used above might not be universal and could be constructed just as well as LABRADOODLE can.
The second question is harder. The compositionality of language depends upon the generalisation of C concepts – not all into one but into manageable groups. For example, the infant learns POPPAEA / HASP / BEAUTIFUL and NERO / HASP / MAD, and knows that NERO / HASP / BEAUTIFUL may not be true but can be represented by a grammatical sentence. The infant also knows that MAD / HASP / NERO is nonsensical and would be ungrammatical in language. How is that done?
We might jump to the conclusion that Cs are acquired individually but are soon grouped. It’s conceivable that some groups are innate – old-fashioned ‘parts of speech’ perhaps. But there is a lot of evidence – synonymy, homonymy, idioms etc – that grouping is much more subtle than that. Surely it is learned rather than innate.
At this point we must be careful not to imply that, going from the particular to the more general, there’s a ghost in the machine making the necessary judgments. That would be very un-NG. So please tune in again next week.