By the generative framework here, language is organised in three dimensions. One dimension is obvious in the fact that sentences have length and and left to right ordering. In the spoken word know, the N sound precedes the OH sound, and in the expression, “You know”, you comes before know. The sentence “You seem to know her” has five words. The linearity is a necessary consequence of the fact that  no matter whether the structure is heard in speech or seen in sign language, human articulation, perception, and memory work in a linear way. Sign is more complex in as much as it allows the two hands and head and eyes to do more things at once, but still by an inescapable sequence. But largely by the work of Noam Chomsky from 1955 to 2023, two hidden dimensions have been discovered. There is DOMINANCE in the vertical dimension and DERIVATION in time. The length of structures is an artifact of the derivational depth, rather than the other way round.

The hidden dimensions are captured by the branchedness in the diagram above, now commonly known as a ‘spine’, built from the bottom up. The branching can continue, as by the diagram above, indefinitely. One hidden dimension is from bottom to top. A second hidden dimension relates the elements by each branching according to their roles. This simple device makes language infinitely creative, using only one simple procedure or operation. Consider these:

• I know.
• Know her?
• You know her.
• Who do you know?
• You seem to know her.
• Who asked who you seem to know?
• We know who asked who you seem to know.
• You think we know who asked who you seem to know.
• I suspect you think we know who asked who you seem to know.
• I’m sorry to say I suspect you think we know who asked who you seem to know.

We can go on multiplying the variations any number of times. The structures of the sentences get harder to understand, but there is no point at which they cease to be English. And not just English, but any one of the other six or seven thousand or so languages spoken around the world. All human languages are generally said to have this infinite creative potential.

By the proposal here, human language started to evolve when the first gestures with meaning and some phsical expression were mentally defined on the basis of this relation of complete unlikeness between the members of this set, the set of the members of an expression. This was the beginning of the modern lexicon. This to have been the very first step in the evolution of speech and language. (These gestures may have been vocal or with the hands and body. My proposal is neutral about this.)

It might seem logically posssible for speech and language to have evolved from single words which were then combined by a simple function putting one word after another. The total of logical possibilities is finite, according to the number of words or signs. But as Marinus Huybrechts (2019) shows, a function with an infinite output can’t grow out of a system with a finite one. If this line of reasoning is correct, as I believe it is, what the small child is doing by his or her first word is rather more sophisticated than it might appear to be. Human speech and language development has to START with a potentially infinite output. This leaves a tree structure as the only possible starting point. Taking the Huybrechts conclusion to the limit, even the first word opens to door to an infinite output system. There is no way of characterising the pathway from single words to the five words in “You seem to know her” other than by postulating a potential branchedness in the simplest structures.

Immediately below I show only the first and second branches.

By the tree diagram above, a structure can grow from one branching to another. Consider “Did you do the washing up?” and “You did not do the washing up.” Did and do are formed from the same do root used in different ways in English. But in both cases do is pronounced in between did and washing up. And you in the first case and not in the other are prononnced in between did and do in what is clearly a verbal element. In a similar way, the equivalence between “Can’t you just do it?” and “Can you not just do it” can be characterised in terms of the position in the chain of branchedness of the negative form according to whether it is fully pronounced as not, or pronounced only by the consonants and without the vowel, as ‘NT. one branching further up the chain. Now these complex orderings could be described as different positions on different sorts of string. But it would be vastly more complex than a tree structure with the positions of the elements varying one at a time, as by the disagram above of “You seem to know her”.

Since 1995 Chomsky has charactersed this process as the ‘merging’ of one element with another. It applies ‘recursively’ – over and over again. These two hidden dimensions are not a figure of speech, but a representation of the underlying algebra of the cognitive process here. These dimensions are hidden, We are as unaware of them as we are of gravity. Like gravity, they can be detected only by induction. What we are aware of, words and signs, are the necesssary externalisation.

The hidden dimensions

In “You seem to know her”, the deepest relation is between know and her. The two words differ on the point that her picks out some individual female known to the speaker and listener, and the verb, know, plainly does not refer. The ‘know her’ clause, traditionally described as ‘subordinate’, is generated as the first derivational step.

Know is said to ‘project’ as the defining element of the branch, and shown by the heavier line in the diagram of a ‘Verb Phrase’ here. Know is thus the ‘head’ of the resulting expression. The history of how the expression was formed is preserved in the verb phrase label.

The notion of you as the knower is reflected in the next derivational step, by which the bare bones of a sentence and a meaningful proposition, either true or untrue, is created. But you is not pronounced at this position, as shown in the diagram below by greying.

The non-finiteness of the clause, the fact that the verb is not marked for tense is shown by the word to.

The main link between the clauses is by the special property residing in the verb, seem, calling into question the certainty of the knowing.

By the final step in the derivation, a slot is created for a nominal subject of seem, filled by you copied from its original position, and ‘raised’ to a higher position by a further grammatical event. For the purposes of pronunciation, you is ignored in the position where it originated, although it remains there for the sake of understanding.

There are two hidden dimensions here, one in the fact that the derivation is by steps, and the other in the fact that resulting structure has depth.

What does this quite intricate structure achieve? Would it not be simpler to say just that the ordering in “you seem to know her” is associated with a particular meaning? The problem with a strictly linear characterisation is in the definition of the ordering. There is no plausible way in which the necessarily very complex grammar could plausibly have evolved.

This situation has led to the postulation of what is commonly referred to as ‘Universal Grammar‘ or UG. The notion of UG does not propose (absurdly) that there is just one language, but just that there are deep and very significant commonalities across all languages, and that the commonalities are at least as important as the obvious differences. As Roger Bacon put it in the 13th century, “In its substance, grammar is one and the same in all languages, even if it accidentally varies.”

A counter claim

The frame work here is commonly criticised in relation to a much cited article by Hauser, Chomsky, and Fitch (2002) where it is proposed that the distinctively human property in language is the ‘recursion’ by which one structure, such as ‘know her’ is embedded inside another structure, in “You seem to know her”, with the raising verb, seem.

Daniel Everett (2009, 2013, 2018) claims to have found a language which lacks this capacity. He claims that he was not able to hear any instances of recursion in the language of the 500 members of the Amazonian tribe which he studied, recursion can’t be universal. The language known as, Pirahã, is spoken by one isolated tribe of hunter gatherers in the Amazon. Everett appears to have been the first outsider to make contact with them and learn their language. He claims that it is simply impossible in Pirahã to say anything like “You seem to know her.” But, as David Pesetsky, Andrew Nevins, and Cilene Rodrigues (2009) show, there are good reasons for doubting Everett’s claim. They show that Everett misrepresents his own data. The argument rumbles on. But Everett is in a very small minority in arguing that the framework here can be discredited in its own terms.

The direction of thought

It is sometimes thought that the main focus in the analysis of child speech should be on what children most often get wrong. But that does not answer the questions: Why do children get wrong what they do, not just individually, but generally? And how do they end up talking ‘correctly’, understanding one another, and agreeing about meanings? To find an answer, I propose to consider what learners HAVE to attend to. and how could human speech and languaage have evolved the structures they plainly have. The two considerations of what the language learner must be attending to and what must have happened in the evolution of speech and language are what justify and motivate the research program and the proposal here.