Menu Close

Modesty and caution

Theories and right answers

If you are worried about a child’s speech and / or language, and you probably are or you wouldn’t be here, you quickly encounter a bewildering range of ideas and theories. Obviously, they can’t all be correct. Are they all equally suspect? Or is there merit in a refusal or reluctance to take sides on controversies? The issues may seem arcane. Within the broad framework of what used to be known as ‘Transformational Generative Grammar’ and is now known by many proponents as ‘Biolinguistics’, I make the case here for being selective, but with caution.

What can and can’t be proved

A supposedly high standard of argumentation is set by the notion of proof. But whereas claims can be proved in mathematics, this is impossible where experimentation is involved, as in the study of medicine, health care, and education, or where there is just data to be analysed, as in cosmology and the study of human language. A theory can be disproved by showing that it has preconditions which are impossible or untrue. But proving that a theory is true is another matter. Experiments and analyses can never be definitive. Another experiment or analysis may point in the opposite direction, or even disprove it.

The search for evidence and approximations to proof is a difficult one. This does not mean that it should not be attempted.

I assume that with respect to any claim relating to health care, the most that can be said is that it is supported by evidence which is as complete as possible, internally consistent, parsimonious, and such that at least some predictions can be confirmed or disconfirmed. Over time, as evidence accumulates, this can approximate to a proof, but no more.

For my own part, in Nunes (2002) I set out to show that the distribution of children’s speech errors is highly asymmetric. There are many cases where one sound in a word is clearly influenced by another sound and there is no obvious reason for the influence to be in one direction or the other. For instance, many children say the word magnet as MAGNIK with the back of the tongue feature of the G clearly inflencing the tongue tip T. But I did not find one child doing the opposite and saying the word as MADNIT. If these asymmetries are widespread, this deserves explanation. And this explanation is likely to be relevant to children who can be hardly understood and may go through life being unable to speak understandably.

By the framework here, these asymmetries fall out from the organisation of a ‘Universal Grammar’ as the consequence of fact that speech and language have evolved in the species.

It is not the case that Biolinguistics is a monolith. It is developing.

Opposing perspectives

Biolinguistics has many critics.Some address what they see as gaps or explanatory failures. Others appeal to the principle of verification by experience, and reject the ‘logical empiricism’ which arose in response to the mathematical logic of Bertrand Russel and Gottlieb Frege (see Richard Creath, 2022). Especially in the English-speaking world, this has led to what seems to me like an obsession with methodology and a correspondingly cavalier approach to theory.

The proof of the pudding

Side-lining fastidiousness, it is sometimes claimed that a given intervention is justified by the fact that “It works” or that “The proof of the pudding is in the eating.” But there is a disingenuity in the word proof. Like folk medecines from before the time of science, such claims can’t be queried or compared with alternatives or modified to meet any more extreme condition. With a treatment for every condition, evidence is replaced by myth.


Burrhus Skinner (1957) proposed that language was ‘verbal behaviour’ and could be reduced to exchanges between individuals. He later told how he was motivated in the 20 year task of writing Verbal Behavior by a challenge from the English philosopher, Alfred Whitehead: Could behaviourism account for one of them saying “No black scorpion is falling on this table.” Whitehead’s point critically involved the quantifier no. How, he implicitly asked Skinner, could quantification, as by words like some, all and no, be reduced to the terms of a conversation? Skinner seems to have thought that he had responded to Whitehead’s challenge. In 1959 Chomsky reviewed Verbal Behavior for the New York Times, exposing the irremediable gap where language goes beyond conversation. At least for generatively-minded linguists, behaviorism has never recovered.

The problem of planning

Milder forms of empiricism live on in the investigation of conversation, as by a theory of speech disorder, known as ‘Childhood Apraxia of Speech’, CAS. Both the investigation of conversation and CAS involve the notion of planning, in the one case, planning the utterance, and in the other planning its pronunciation. But the notion of planning is not clearly defined by either of these approaches.

Oral Motor Therapy and discrimination training

From two perspectives, there are abilities which are obviously essential for speech, the coordination of the articulators and the auditory processing of what is heard, by memory and discrimination. The focus on articulators is particularly emphasised by ‘Oral Motor Therapy’, OMT. It appears to stem largely from the work of 19th century elocutionists. It has been largely discredited by Greg Lof (2006). Similarly there are programs which claim to sharpen the auditory discrimination of speech in particular, by contrasting speech with classical music. Both are of these approaches are peripheral in the sense that neither involves the central relation between sound and meaning. Both are defended by their proponents on the basis of what they claim to be their positive experiences or testimonials. But these are not backed up by evidence connecting the particular practices with any measurable results.

Radical sociology

An idea which I shall refer to here as ‘radical sociology’ views speech and language as inseparably embedded into human society, disregarding any structures defined only by grammar. Disallowing meaning in words, there are just reactions and interpretations between individuals. Radical sociology shifts the research focus to methodologies for establishing what these reactions and interpretations are. The search for uniformity is misguided. There is just infinite variation, defined either on individuals or on social groups defined by class, locality, gender, ethnicity, age. It is not obvious to me how such a research focus bears on children’s speech and language problems.

What the National Curriculum leaves out

From another perspective, there are the exigences of everyday teaching. The English National Curriculum strings together reading, writing, understanding  and speech as four irreducible components of what needs to be taught in school. But this leaves out dreaming, imagination, the actor’s mental rehearsal before going on stage, and more. And the place of grammar remains uncertain. It seems to me that dreaming, imagination, and rehearsing things in the mind are things which at the very least need to be explored in education.

Linguistic critiques

From a series of linguistic perspectives, such as those of Maurice Gross (1979), Robin Dixon (1997), William Croft (2002, 2004, 2010) Mike Tomasello (2010), Ben Ambridge (2004, 2010), Eva Dobroska (2015), the failures of Biolinguistics outweigh the merits. These perspectives are broadly what are known as ‘constructionist’. They prioritise the constructions by which words are commonly reordered in sentences over any more general considerations. There are four main thrusts to their criticisms, each vulnerable to counter-criticisms:

  • There are phenomena in language, such as the less common word orders or various degrees of freedom in word order, which are not adequately described or explained. The point here is the stock in trade of discussion between generative linguists of all persuasions. The point is turned against generative linguistics as  whole by Ambridge who writes: “The reality is that children do not appear to be operating with abstract categories at all” (Ambridge, 2004, p. 44).But Julie Anne Legate (2001) shows that the most extreme case of ‘free’ word-order has an explanation by the movement of all the seeming freely ordered constituents from strictly ordered positions;
  • The abstract categories are not adequately or reliably evidenced. By the latest formulation of the biolinguistic model in Chomsky et al (2023), there is only one operation in narrow syntax, namely the process of ‘Merging’ two elements, an operation which is surely irreducible;
  • There are predictions which are not born out, or not born out accurately. But this is so for any theory. And there are more concrete categories which are both better evidenced and which make more reliable predictions. The issue is not whether all possible predictions are confirmed, but whether some predictions are confirmed. It is a matter of judgement which predictions are the most revealing and significant;
  • The notion of ‘Universal Grammar‘ or UG is baseless. By the proposal here, the emergence of UG in acquisition can be convincingly demonstrated. Most linguists working in some version of biolinguistics believe in some version of UG. It seems to me that the notion of UG defines an irreducible skeleton for analysing defects in child speech and their asymmetric patterning.

Consider the case of passives. By most theories there is a relation between active forms like “Mummy is kissing her” and passive forms like “She is being kissed by Mummy”. Passives are commonly used in English. They involve changes in word order, in the case from her to she, in the form of the verb from kissing to being kissed, and in the subject being reworked as part of a by phrase. As shown by Legate (2014), languages vary in how much of this structure they use. Some languages don’t allow any of it, and have no equivalent expressions. English passives represent the most complex case. Children learning English are only just starting to learn the simplest aspects of the passive before they are three. If human language evolved construction by construction, it would be necessary to account for the evolution of each combination of forms separately. A spine-based, modular theory of the passive – by the Proposal here – offers a general account. On account of its generality it seems preferable to any constructionist account.

By the proposal here, contra Ambridge, there is evidence of the most abstract of biolingustic phenomena, namely Phase, which gradually becomes evident between the ages of two and three. If this point can be sustained – and the evidence is tricky – the constructionist critique collapses.

It seems to me that the greatest weakness of construction- based theories is that they cannot easily or naturally account for the evolution of human language.

Large language modeling

However, the most complete rejection of Biolinguistics is by Stephen Piantadosi’s (2023) ‘large language’ model. Piantadosi demonstrates convincingly the ability of current artificial systems, such as ChatGPT, to generate plausible and convincing replicas of natural language. These are, in my view, an extraordinary, engineering achievement. But the training is on data sets far larger than those available to the child. And these datasets are not compromised by incomplete or plainly ungrammatical sentences or inconsistencies such as those between parents speaking different varieties of a language, or even parents speaking different languages, communicating with one another in some sort of pigeon, and so on. Piantadosi’s model says nothing about how or why human language with all of its extraordinary complexity evolved to have the precise and particular forms which it demonstrably has. The Universal Grammar postulated here is motivated by the need to explain the evolvability and learnability of natural language in the imperfect and wildly variable circumstances into which children are born, all having randomly different linguistic experiences. Somehow they all converge on a single grammar. (See Roni Katzir (2023), and Noam Chomsky, Ian Roberts, and Jeffrey Watumull (2023) for more extended versions of this negative view of Piantadosi proposal).

From all of these perspectives, introspective data are just not data. Biolinguistics is dismissed out of hand – if it is considered at all.

A proposal

It might be asked: with opposition like this, isn’t Biolinguistics doomed? But like Chomsky and many other linguists, I assume here that Biolinguistics is very much alive and kicking. Here and elsewhere, I argue that it has more to say about the special problems of delays and disorders of language development than the sum of the criticisms.

A comparison with medicine

In medicine, there is a long tradition of querying any innovation with a view to destroying it. In 1847 the newly qualified Dr Ignaz Phillipp Semmelweis observed that maternal mortality was ten times greater where the delivery was by doctors than where it was by midwives. The discrepancy was detected by the population at large, and women started avoiding the doctors’ wards.  Semmelweis proposed that the critical variable was hand-washing, He introduced regular, compulsory hand-washing on his wards, And the discrepancy in the mortality disappeared, But Semmelweis’s idea of PR was to tell doctors who disagreed with him that they were murdering their patients. Unsurprisingly this was not a success, And he was sent to a lunatic asylum where he died in 1865 with his work unrecognised. Eminent voices were not convinced of the importance of topical hygiene for another 20 years.

200 years before Semmelweis, following a quite different career path, William Harvey qualified in medicine at the then world-famous University of Padua in 1602, where he was elected as the leader of the English students. Back in England, Harvey began a stellar career in medicine, in 1615 becoming the Lumleian Lecturer at the College of Physicians and in 1618 Physician Extraordinary to the king. But privately Harvey was following his own research agenda. One of his teachers at Padua had drawn attention to the fact that the circulatory system contains valves which allow the blood to flow through them in only one direction. It was believed at the time that the main function of the heart was to keep the body warm. But if so, why did the heart contain one-way valves?  Harvey undertook a long series of experiments with the hearts of animals and dead people, including some who had just been executed, experiments which would not get past an Ethiccs Committee today. In 1628 he published his Anatomical Exercise of the movement of the heart in Animals. He showed that more blood passed through the heart in half an hour than was contained in the body. The heart was a pump. As was standard practice at the time, this was in Latin. The reference to animals in the title may have been to discourage any theological reactions. To be on the safe side, Harvey published his book in Germany. He would have known of the fate of Michael Servetus who had mixed a novel theology and a theory of the circulation of the blood, and been burnt at the stake in 1553 in Geneva. Harvey’s new theory contradicted the theory of Galen from the second century that blood was continually produced by the liver. Galen had based his theory on his observations of the still-beating hearts of dying gladiators. Harvey’s theory meant that the widespread use of bleeding for various disorders was unjustifiable. Despite Harvey’s evidence from what was at the time modern experimentation, his theory met opposition from colleagues, patients, and others including René Descartes. Harvey’s book was not translated into English for another 20 years.

In brief

Theories of all sorts have to be proposed with a degree of modesty, humility, and caution. I try to do so here. But I see no merit in not taking sides where there is a clear conflict, as there is between those who believe that linguistic structures are GENERATED and those who prefer to see language as BEHAVIOUR, that is planned, but with no intrinsic meaning and no allowance of appeal to introspection. As far as I am concerned, theories which seek to reduce language to behaviour (with or without the implicit notion of planning) just miss whatever it is that makes language the stuff of love, hope, legend, invention, science, comedy, and law.

Theories about language and linguistics are impossible to prove and difficult to disprove. In my view, especially where children are involved, as they are here, it is appropriate to be cautious and to take stock of the weight of the evidence. I try to do this in my proposal here which I have been working on since Nunes (2002). Plainly and rather obviously, speech and language have evolved as one of the distinguishing characters of the human species. I propose that this evolution probably began some time before modern human ancestors diverged from Neanderthals about 650,000 years ago. I propose that this evolution involved seven steps, each taking tens or hundreds of thousands of years to disperse through the population, contributing one by one to the modern human genome. The last of these steps would seem to have been completed by around 150,000 years ago when humans almost became extinct, becoming reduced to a population which may have consisted of only 1,000 or so individuals.

Doubtless, the proposal here will develop, as research proposals are apt to do.