Language, Communication, and Animals – dantsdritic branching – Jan 12

Language, Communication and Animals

Okay, this is going to probably a series of posts because there’s a lot of background that needs to be done before I can adequately cover the area. I might do a post on animal cognition and various aspects of that later on if there’s interest (this’d be all about spatio-temporal encoding in bees, theory of mind in birds, and related stuff like that). To start off, I want to talk about what language is, its relation to communication, and the common question of whether we think in language.

So what is language?

A language can be defined as a set of sentences each finite in length and constructed out of a finite set of elements. To put this more simply, a language is the collection of all the sentences that can be made out of the individual components that go together. What allows us to combine and ‘generate’ the sentences is what we call the grammar of a language. So we can see language as a computational system that has a grammar that allows us to generate sentences out of finite components like lexical items (words). There are different types of grammars and their ability to generate different types of sentences depends on how complex they are. This is an important point, and will feature later.

Now, I’m going to quickly digress and address another common question – do we think in language? This is a topic of hot debate, but I believe they we don’t, and furthermore I think there’s excellent evidence and arguments to show this. Something you may have noticed in your day to day life is that sometimes you know what you want to say, but you can’t find the word for it. This is some basic evidence to suggest that maybe we don’t think in language, because if we did, then we should never be loss for words like that.

If we investigate language, we actually find that there’s certain manipulation of thoughts that we can’t quite express in words. I’m just going to block quote Chomsky here:

Well, again, to try to make the matter concrete, let’s take an actual example, it’s simple enough so we don’t need a blackboard. So, just take the sentence:

‘He wondered whether the mechanics fixed the cars.’

And just consider two questions that you could ask about that.

The two questions are:

‘How many cars did he wonder whether the mechanics fixed?’

Answer: ‘Three cars’

‘How many mechanics did he wonder whether fixed the cars?’

Answer: ‘Three mechanics’

But the trouble is, you can’t say that for some reason. You can’t ask: ‘How many mechanics did he wonder whether fixed the cars?’

In technical terms it’s called an ECP violation, but it doesn’t matter what that is.

The thought is fine — fine thought — but you have to express it in some kind of paraphrase. There’s something about the language design which poses a barrier to communication. You just can’t express a simple thought like that, you need a circumlocution. Many such examples.

Language and the Cognitive Science Revolution(s) – Noam Chomsky

So here we have an example of thoughts that can’t be quite expressed in language, and furthermore that there’s something peculiar about the way language is set up. This is where structure dependence comes in.

Structure Dependence, Linerisation and Optimisation of Language

Structure dependence is all about the ‘hidden’ internal order in language.

The really interesting thing is that children don’t make mistakes about structure dependence when learning language – for example, take the sentence “The dog in the corner is hungry.”. If you were to form the question of this, you’d make “Is the dog in the corner hungry?” which looks simply like moving the first ‘is’ to the front of the sentence. But say if we have the sentence “The dog that is in the corner is hungry.”, then if we apply the rule we get: “Is the dog that in the corner is hungry?” which is obviously incorrect. The correct version being: “Is the dog that is in the corner hungry?”.

Another example is the sentence “Do eagles that fly swim?”. Now, it’s obvious what the question is asking, but it’s not at all obvious about how we understand that, or why the question is phrased the way it is. The ‘do’ is associated with ‘swim’not ‘fly’, which is interesting. If we have to express language in a linearly processed form, how is it that we know the question is about the second verb and not the first? Why wouldn’t, if language was optimised for communication, it be that the first verb is associated with the question?

Now, kids don’t make the first error, and people know that the second question is about whether eagles can swim, not whether they can fly. This suggests that they learn the structure dependent version from the very beginning, which is interesting because there’s practically no evidence for them just to learn this by the input alone. This is highly contentious in the field of linguistics, and I’ll probably end up addressing it in (yet another) post, but the point is that there’s some sort of structure to language that sits beneath the surface, yet we are immediately able to process it. Furthermore, it doesn’t seem to be optimised for communication, which gives further credence to the idea that language is not for communication. This is a very hard and fast version of the argument, and I’ll return to it later if people are interested.

Animals and Language

Okay, so hopefully for whatever reason you’re willing to take it on faith that there’s a difference between a system of communication and a language – animals can communicate, but this does not necessarily mean they have language. Now I’m going to talk a bit about different types of grammars. Different types of grammars are required to produce different types of languages. This is a whole field unto itself, but today I’m just going to talk about Finite State Grammars (FSGs) and Phrase Structure Grammars (PSGs). Now, to get the sort of structure dependence we need in human languages, we need a grammar that is at least a PSG – FSGs just can’t generate the types of sentences that PSGs can, and so if you don’t have or have the ability to learn a PSG, then more or less you don’t have language.

There’s research be done on tamarins (a type of monkey) that show that they are quickly able to master FSGs, but are unable to do so for closely related PSGs, and that this inability seems reflected in the majority of animal research so far conducted:


“[…] a finite state grammar […] is inadequate to capture the structure of any human language. More powerful grammars, called phrase structure grammars, are required to capture the long-distance dependencies and phrasal nesting found in natural languages […]. The necessity for phrase structure in language is one of the few areas of apparently complete agreement among syntacticians of various theoretical backgrounds […].” (Fitch 2005a, p.g. 206)

“Comparative data on nonhuman primate cognition and communication, as well as communication in other groups such as birds, have allowed us to focus on mechanisms that play a key role in language but appear to be lacking in our nearest living relatives […] [such as] phrase structure parsing.” (Fitch 2005a, p.g. 209)

“Phrase structure seems to be critical not just for human language but many other aspects of our cognition […]. Monkeys exposed to auditory output from a finite state grammar, with only flat structure, easily learn it, spontaneously generalizing to novel grammatical stimuli, but fail to do so when exposed to a carefully-matched phrase-structure grammar (Fitch & Hauser, 2004). Although too few species have been examined to reach any broad conclusions […] hierarchical processing does not seem as widespread in animals as in humans.” (Fitch 2005b, p.g 392)

“Interestingly, despite the well-documented serial processing abilities of some nonhuman primates, recent data suggests that their ability to process phrase structure are limited […]. [P]erceiving phrase structure and relationships between distant items – may be difficult or impossible for other primates.” (Fitch 2005a, p.g. 206-7)

Fitch, W. (2005a). ‘The evolution of language: A comparative review’. Biology and Philosophy, 20, 193-230

Fitch, W. (2005b). Computation and Cognition: Four distinctions and their implications. In A. Cutler (Ed.), Twenty-First Century Psycholinguistics: Four Cornerstones (pp. 381-400). Mahwah, NJ: Lawrence Erlbaum.

You may have heard of projects like Nim Chimpsky, Washoe, Koko and the like, but these have been heavily studied and reviewed and they’ve been found that they don’t really display evidence that apes have a grasp of language. In most cases, the responses were prompted, not formed correctly and weren’t novel nor showed any real similarity to what a child of 3 or so can accomplish.