Monday 31 March 2008

on the purpose of grammar

While reading some new Language Log posts on this guy named Kilpatrick, I came up with some lines of thought I'd like to run through.

1. Planned economics != socialism. In this case, it's planned linguistics. But of course I am picky on this because I call myself a libertarian socialist.

2. One ought to be careful about prematurely assigning "purposes" to grammar. It seems to me for example, many elements of universal grammar arise because of the common structure to the way we think (without a need for an explicit Chomskyian "module" and all its parameter-setting). The language universal for languages to use prepositions (as opposed to postpositions) for SVO word order may arise not because it's easier to understand, but just how we tend to be guided by existing mental structure when we create words.

3. We ought not to confuse "pain" with necessity. Speakers of pidgin English tend to be mostly comprehensible, if not totally so, save that alarm bells go off in our ears, declaring, "this isn't right!" Some language pedants are observant enough to notice the alarm bells do go off. Yet trying to rationalise the presence of the alarm bells as, "oh, grammar is important for clear communication," is a bit like saying, "oh, the inflammation and irritation associated with peanut consumption is biologically necessary to keep my body free of nutritional decadence." (Yes, I am allergic to peanuts.) The allergy may have developed for a reason, but note that G6PD deficiency (which yes, I also have) confers a protection against malaria. As an allergy may develop for purposes of immunity, not nutrition, aversion to non-standard grammar may have developed for purposes other than fostering clear communication. (A wild guess would be to foster an aversion to outsiders, which might have been useful for telling different groups apart, back in our hunter-gatherer days for example.)

4. Language contains many fascinating features and patterns, many of which might be unrelated to clear communication. These make it very interesting to study, but simply because these are very salient features of language doesn't mean these features are required for effective communication (as I have hinted above and will develop below).

5. Grammar may be a "crytographic" measure to check the integrity of the message. This can occur in two contexts.

6. In the first one, you might be communicating across an unreliable medium (such is the nature of speech, in which one phoneme might be confused for another), so features of grammar help rule out ambiguous messages. I noticed that until I got into Bizet's Carmen fairly recently, I had never really picked up on the French lyrics. Opera is an extreme example because the sound formants at high notes are especially distorted from the ones in speech, which results in a higher number of confused phonemes, but it will be good for illustrating my point. Until I actually realised that the townspeople people in Habanera were actually saying something ("prends garde à toi!"), I thought they were a background chorus, because all I heard previously (in various cultural contexts, like the part where the film plays Carmen when Hitler goes on his campaign trail in the film Hitler: the Rise of Evil") were the vowel sounds: honn har a hah!. Notably, when I didn't realise it was speech (as opposed to an "effects" chorus), many portions of the linguistic part of my brain remained unactivated, such that all the consonants (which also happened to be all plosives, which have the lowest audibility compared to the other consonants like sibilants, liquids, nasals, etc.) sounded like /h/. (This sort of makes interesting sense, because plosives have an "aspirated" [h] element to them, even somewhat for voiced consonants like /g/.)

Note that in spectrographs, it's often hard to tell the different kinds of plosives apart! Indeed, unless our linguistic portion of our brain is switched on, the normal auditory portion of our brain (and it is sensitive enough to pick up the nuances of music!) can't tell them apart. In this case, the patterns of language help with clear communication, but it wasn't so much about making semantics clear directly as ruling out message possibilities that didn't comply with the rules, as "improbable to have been sent". Note that it's been shown with the McGurk effect, where patterns from usual visual cues like the shape of the lips of the mouth have an influence upon perception. Why would it exist? Given that many consonants have quite fine distinctions (note that we are often unable to differentiate between many languages' exotic consonants if we haven't learnt them explicitly), it seems likely that it exists as part of an integrity check, confirming that a /g/ is in fact, /g/, and not a /b/, or a /d/, by ruling them out.

My example so far has been mainly phonological, not morphological -- but it can be extended there too. Consider a sentence like, "I'm right here at the bedside." If you heard the /m/ intact, then you know you can rule out the verb for writing because it would be an ungrammatical sentence. Many pedants are asinine when it comes to, "Oh! But you must recognise the fine distinctions between this verb and that verb!" But it doesn't have to be that way at all. When pedants are thinking about confusions in meaning, they might be on your backs on how "blame X on Y" is 'confused usage'. But they might never think that a sentence like, "I'm but a rock by the wayside" could pose a problem. How do you know for example, that you didn't really hear, "I put a rock by the wayside?" /p/ and /b/ are both bilabial plosives, save that one is unvoiced and semi-strongly aspirated while the other is simply voiced. They are phonetically similar.

But ah! Because of the presence of the copula (as the phoneme /m/) you know it cannot have involved the action of putting. What if you didn't pick up on the /m/? Ah, but you might pick up on the numerous other linguistic patterns too, from other sentences, or suprasegmental features like stress. In this case, the presence of rules is sort of like a linguistic version of an md5 hash or a cyclic redundancy check: for a message to have integrity (i.e. you are sure you're hearing the right message), rules might develop to develop information redundancy in case one portion of a message gets obscured. In language, this redundancy is stored in grammar, and can then be part of grammar's purpose. In fact, we rely on these rules so automatically that we don't realise how frequently we rely on a variety of cues to comprehend language.

In the case where we hear ungrammatical pidgin sentences, the "jarring" effect might come from the fact that our minds have automatically ruled out every possibility, and yet, our conscious mind knowing there is a message there, must "override" the automatic processes, declaring that, "yes, there is actually a valid message here." (This is where it would be pertinent to discuss Freudian theories about repression and the subconscious, but I will refrain from it.) Note for example cases where non-native English speakers might use a syllable-timed rhythm and put the primary stress on the wrong syllable -- or perhaps a more pertinent case for Singaporeans, where a Western-born teacher tries to use Singlish and fails miserably because he didn't realise how important tone was. If you inflect "lah" wrongly (excluding question inflections, which would be even more awkward), it's not like you've affected the meaning or you've made yourself unclear -- it just sounds weird. What does it tell us? The speaker is an outsider. This leads me to my seventh point.

7. In the second type of integrity check, grammar may exist for security purposes. Suppose you are competing with a nearby tribe. Maybe it's even Neanderthal! Before they invented military codes, it might have been advantageous to come up with grammar in language. It would be a sort of shibboleth: firstly, it obscures information from the enemy. (It might be hard to comprehend a simple enemy sentence like, "we will attack them at dawn," if you have no idea what their locative or accusative cases are.) Secondly, the enemy can't "pose" among you very easily if he uses sentences that don't comply with your rules. Perhaps this would have been sufficient cause for our genome to code for "alarm bells" whenever one heard an ungrammatical sentence. Now in the modern day globalised-world, it's just a pain in the you-know-what, but it could have been critical in the evolution of our species.

8. Drawing from such a need, a reason for constant language change might be a need to change the code constantly. Note that in real life, implementing a code change is a real pain. In language, you rely on spontaneous order, and how magical it is! I am also reminded of the Tower of Babel story -- perhaps grammar was invented to divide up the human population and set them into disassociating nations.


9. Pragmatics and ease of organisation. Stuff influencing the development of grammar might not only be comprehension, but the efficiency in storing information. Creating a system with elaborate patterns increases processing time, but decreases the space and memorisation required to use the signs required for language. If messages must be of finite length, and yet the signs in them must be capable of representing an infinite number of thoughts, the tradeoff for more processing time might be attractive. It is recognised through many recorded incidents that strict Skinner behaviouralism fails to fully explain language acquisition, and hence the attractiveness of Chomskyian nativism. I like Chomsky's UG theory somewhat (though I think that the idea of explicit modules and parameters is an overly elaborate explanation for a much more subtler reason for UG), but this can be noted in the way children can recognise productive inflections and create them for words they might have never heard of before. In more powerful examples, they can synthesise their own grammars and creoles and create regular systems where pidgin-using migrant parents might not. My point here is to point out that the development of grammar might also have arisen out of a convenient way to acquire language, as opposed to comprehension.



10. We ought not to be so quick to accept arbitrariness of sign as an axiom. My tenth and last point -- phonetic signs in language are often referred to as arbitrary -- e.g. outside of their association they have no real correlation with their meaning. But I think it would be rather hasty to call this an axiom. There are many words which I suspect have a degree of correlation with their sound. It just so happens for example, that "hostile," "harsh" and "hiss" are well --- harsh sounding. Of course, one must issue the caveat that it may sound harsh to us because the semantic association with the sounds of the words developed *after* we learnt the sign, but at the same time, I must also issue a counter-caveat that just simply because the patterns following the correlation between sound and sign are too complex to be analysed so far doesn't mean they don't exist -- there might be just so many rules and interferences, and so forth. The debate over the origin of language, which as of yet we have no convincing mainstream explanative theory for, and how the first words originated, perhaps tells us we should not be quick to dismiss current signs as arbitrary, for they might just have a very complex development.