I had a new thought the other day regarding how language is conveyed and perceived. When a language is conveyed visually in English, it is conveyed in symbols (such as text or sign language). However when it is perceived aurally, it is perceived as a raw stream of acoustic frequencies.

Is there any linguistic distinction between conveying ideas/language via discrete chunks (such as symbols/text) and conveying via a stream of data that must be intellectually parsed (such as a stream of acoustic frequencies)?

I'm very new to the field of linguistics, so if anyone can think of any more appropriate tags, please feel free to edit them in.

In fact, I'm having trouble understanding exactly what I'm trying to distinguish in my mind. Is there an existing language or set of categories I could put this in? It seems like what I'm describing is a level of familiarity with neural input.

Text or symbols are always coming in visually as a stream of raw frequencies of data which must be parsed by the mind to be understood to be chunks of recognized data.

So, a visual language where conversation is visual only, but presented as animated waveform graphs of sound is not an already understood convention for my mind at least, so it is data that has no existing mental framework to fit in.

I don't know, I'm confused about it. It's like there are high level computer programming languages that use human words as code (like "print" or "if x= 10 then print") and then lower level programming languages, like machine code that use commands in hexadecimal (like "8B542408 83FA0077 06B80000 0000C383"). High level languages are convenient for human conscious awareness to be involved with (we are familiar with the word "print"), but lower level languages like binary or machine code, using commands like "8B542408", are something our waking state consciousness is not readily familiar with.

Similarly, letters of an alphabet are easy for our conscious minds to interact with, but the raw neural data stream of visual perception presented to the nervous system by the retina is something our conscious self is not easily involved with.

So, are there levels of language data like this for human languages? High level formats and lower level formats? If so, is that what I'm asking about when describing symbol-based communication and frequency-based communication?


Firstly, spoken language is the linguist's primary object of study, although there are (imperfect) analogues of written language in linguists' conception of language. I think these are similar to what you have in mind when you mentioned levels of language.

In linguistics, at the lowest level, we have phonetics, which studies the acoustic stream, how it's produced and how it's perceived. Abstracting away from phonetics, phonology looks as sounds as discrete chunks, called phonemes, so e.g. the word 'happy' is made of four chunks (roughly h, a, p, i). It's more abstract than phonetics because sounds in complementary distribution are regarded as different realisations of the same sound. The classic example is the 't' in 'star' and the 't' in 'tea', which are pronounced differently but regarded as different realisations of the same /t/ phoneme.

Above the phoneme layer, we have morphology, which looks at morphemes, the smallest meaningful unit. For example, there are two morphemes in the word 'lights': 'light' and 's'. Building meaningful units from meaningless sound segments is an important property of language known as duality of patterning, one of the defining features of language. Morphemes are concatenated to form words, which combine to form phrases and sentences, which are the object of study of syntax, one layer above morphology.

So, we have already looked at four different layers of language, and one could 'compile' a higher level of representation into a lower one. A sentence can be reduced to a string of morphemes, then to a string of phonemes, and finally an acoustic stream. That is (somewhat) analogous to, say, a database language compiling to C, then compiling to assembly, then finally turns into machine language (I'm no computer scientist; correct me if I'm wrong here).

Written language, which you've mentioned, conveys information roughly parallel to the phonology tier, though to a lesser extent in Chinese than in English and French, which in turn code phonology to a lesser extent than Korean and Spanish. But Korean and Spanish encode less morphology than English and French, which in turn encode less morphology than Chinese. So although you can easily tell between 'meet' and 'meat' in English, 'might' as a modal and 'might' as a noun are written in the same way, whereas in the Chinese writing system, which is nearly one symbol = one morpheme, homophones are usually written with characters that look different.

So your symbolic tier seems to map to two different 'levels' in linguists' conception of language: Morphology and syntax.

| improve this answer | |
  • Thank you. I'm going to ask another question now to see if I can follow up on this idea. – Thom Blair III Dec 30 '16 at 18:38
  • This is the underlying idea I was trying to think about. What would linguists call a language that cephalopods might develop with their skin color changing ability? – Thom Blair III Dec 30 '16 at 18:53

The premise of a significant difference between auditory vs. visual perception in terms of "raw streams" vs. discrete symbols is questionable. There is no difference, and all of language is, at the perceptual level, in terms of discrete symbols. The body operates on continuous visual or acoustic signals which are sensed and present the mind with discrete units for manipulation. The mind really can't handle continuous data streams, so to hold it in memory, it has to be somehow reduced to one thing (and then another thing... which will be organised hierarchically in some other way). I think you'd be better off not assuming that pre-perceptual processing involves a "language" of any kind.

What you describe in terms of computer operation is still entirely symbolic, whether it involves a statement like "print", or a sequence of assembler mnemonics, or even the hex data stream that corresponds to a "print" command. Even the stuff that seemed to be about voltages, which you might have read about in old Intel manuals, was really the symbolification of the continuous physical process behind a computation. Anything that involves "data" and "levels" is already in the domain of symbols, and not raw physical stimuli (e.g. continuous changes in voltage over time, as occur within a CPU).

There are many examples of how conceptual structure is imposed on language structure, in order to make it mentally manipulable. For example, you hear someone say something, where a waveform hits your ear, and what you get in your mind is a sequence of discrete phonemes [ˈsʌmθɪŋlɑikˈðɪs]. This long string of phoneme things can be organized into a shorter stream of syllable things, and those syllables can be arranged into even fewer feet. Ideas can be arranged into hierarchies that expand and contract, so you can either lump together "cat", "dog", "cow" etc into "mammal", or you can subdivide "dog" into "poodle", "malemute", "shepherd" and so on.

| improve this answer | |

I don't think there is much agreement about the fundamental nature of human language communication. IMO, it is a stream of articulations -- states of the human vocal apparatus. For background, look at the Wikipedia article on The motor theory of speech perception. The bridge to symbolic communication that uses an alphabetic system is through the description of phonemes as combinations of states of the various human articulatory organs, as given in generative phonology.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.