Viewed from a modern perspective, one obvious necessity is the experience of speech and language as the child hears people talking, the experience of the primary linguistic input as this is sometimes called. But experience is not enough on its own to resolve what might seem to be a number of anomalies, many not identified until the 1960s. From the perspective here, these apparent anomalies can be resolved in one of two sorts of ways, either by appealing either to what is known as Universal Grammar or UG, i.e. factors obviously specific to speech and language, or by appealing to general cognitions. Either or both of these factors necessarily involve the human genome at some level. But there are problems with both sorts of appeal in ways that I shall come to shortly. The anomalies occur in just one of what might seem to be parallel cases. Most of these involve what are known as ‘variables’, like the pronouns, you, he, she, we, they, nouns like house, bed, bus, toy, dinner, question words like what and which, words like that in “I know that it’s naughty” can all be conceived as ‘variables’. They can often be freely substituted for one another, “He had dinner”, “She had dinner”, Which did you break?” “What did you break?” obviously with changes in the meaning, but without compromising the essential meaningfulness of the structure. There are more anomalies with respect to what have come to be known as ‘islands’. In “There is some doubt about what he is supposed to be getting to the bottom of” it is clear what is meant, and there is a clear assumption that he is supposed to be getting to the bottom of something, but the Englishness or grammaticality is questionable. So “What are you supposed to be getting to the bottom of?” is okay. By the framework here, what is copied or extracted from the bottom of. But when this is part of a complex phrase headed by the word doubt, an island is created from which extraction is questionable or problematic.

I call these ‘apparent anomalies’ because the essence of the framework here is to appeal to general principles, rather than just listing rules, as by traditional frameworks. Detecting the very large issue here, in 1964 Noam Chomsky proposed a general principle which he called the ‘A over A principle’ which required that a variable is enclosed in another of the same sort, as in doubt about what he is supposed to be getting to the bottom of, where doubt and what he is supposed to be getting to the bottom of are both what are known as noun phrases, one a simple noun, doubt, the other a more complex phrase, only the topmost element, doubt, can be manipulated by the syntax. In proposing the principle, Chomsky noted points where it did not work correctly. But it was a first stab at the problem. Then in 1967 in a justly famous PhD dissertation, John Ross proposed that there was not one single problem here, but a whole number of separate problems, to which there had to be separate solutions.

The apparent anomalies take five main forms:

• Where the meaning of a variable is more or less restricted;
• Where a variable is blocked or never pronounced;
• Where, conversely, a variable has to be pronounced;
• Where a variable is freely pronounced or not pronounced;
• Islands, at least by some judgements, sometimes with opposite grammaticality judgements by speakers of the same dialect.

The restricted meanings occur in sentences like these:

• Before he made the bread, Daddy bought some rye flour.
• Before Daddy made the bread, he bought some rye flour.
• Daddy bought some rye flour, before he made the bread.
• He bought some rye flour, before Daddy made the bread.

In the first three of these sentences, he may be Daddy, but not necessarily. In the last, he is any male who is not Daddy. The seemingly anomalous behaviour of he, and other pronouns was brought within the scope of a powerful unifying analysis by Tanya Reinhart in 1981. But since 1981, more data has been found, considerably complicating the analysis.

The case of pronunciation being blocked occurs in structures like these where a question word like who or what ‘extracts’ the subject of an embedded ‘clause’, in the first example below, not the boy with the dirty face, but some unspecified individual in the second example:

• I know what you think the boy with the dirty face is going to say OR I know what you think that the boy with the dirty face is going to say
• I know who you think is going to talk first BUT NOT I know who you think that is going to talk first

The case of pronunciation being forced occurs in structures where a sentence with an embedded ‘clause’ occurs before the main clause. In the first examples below, that is freely pronounced or not pronounced. In the second example with the embedded clause before the main clause, that has to be pronounced.

• It is rather obvious that Daddy does not look good swimming with that hat on OR It is rather obvious Daddy does not look good swimming with that hat on;
• That Daddy does not look good swimming with that hat on is rather obvious BUT NOT Daddy does not look good swimming with that hat on is rather obvious.

The case where pronunciation is (mostly) free also occurs in what is known as ‘ellipsis’. Here the part of the structure which is not pronounced is shown in brackets. In the first two of these examples, what is pronounced is changed from what is not pronounced by one or two separate processes, one involving what is known as ‘case’.

• Who wants more? Me (I want more);
• Teddy was naughty. Not me (Iwas not naughty);
• I like everything and Daddy (likes) only fish and chips;
• Daddy likes rock music, and Mummy (likes rock music) too;
• I am working tomorrow, (I am) not (working) the day after tomorrow;
• You have had three helpings and your sister (hs had) two (helpings);
• You’ve done more than I thought you would (do).

In the case of ellipsis, it might be thought that the combination of experience and convenience and simplicity was sufficient for the totality of these structures to be quite easily learnable. But it is not obvious how ellipsis should be analysed. Are all of these cases formally the same? Or are there different factors at play in each one? And if there is just one factor, are the cases listed here by ‘non-pronunciation’ more simply explained on the basis what is not pronounced is simply never there to be pronounced? The same point arises with respect to the case of that, freely pronounced or not pronounced in many cases.

And finally there is the case of islands, first identified by Chomsky, and then named and studied in previously unparalleled detail by Ross.

There are two obvious questions about these apparent anomalies: How do children learn about  without being taught about them or having it explained? And how is it that there is broad agreement on these points between native speakers of English despite wide variations of culture and levels of education? Some children have many hours of careful, diligent instruction about their grammar. Other children have none whatsoever. There are other similar, seemingly anomalous phenomena in numerous other languages, both closely related to English like German, Dutch and Danish, and others clearly unrelated to English like Hebrew and Arabic. The simplest and most obvious explanation of phenomena common to across a wide sample of related and unrelated languages is by UG. children are somehow guided towards an appropriate analyis of what they hear said by their biological inheritance, in the same way that they inherit a body with two legs and two arms, with one bone in each of the parts closest to the body, two bones in the further part, and five fingers and toes. Children were not born knowing how to talk, but knowing how to interpret the evidence of what they heard, their linguistic experience. But an inheritance with respect to the anomalous behaviour of pronouns would be quite useless without a rich experience of grammatical structure and what are known as their ‘antecedents’, in the case of the examples above, Daddy, or what might be Mummy, or anyone else capable of buying flour, bread, or anything else.

The big problem with any sort of UG account of the anomalies is the complexity of the data. While there are similar anomalies in most, if not all, languages, there are large or small differences in how they apply from one language to the next, and sometimes, as noted above, from one speaker of what is seemingly the same dialect to the next. A UG account of such variability is quite implausibly complex. The explanation is as complex as what is being described. This does not explain anything.

By an alternative account, anomalies in phenomena which are seemingly specific to language like ellipsis, islands and the understanding of pronouns, should be explained as far as possible in terms of the categories of general cognition. Chomsky often refers to these as ‘third factor’ cognitions, where the first and second factors are experience and UG. Among these third factor cognitions, two of the most obvious are the notions of an entity and an event. The idea here is not new. It corresponds very loosely to the notion in many traditional grammar books for children of nouns as names of people, places and things, and of verbs as doing words. Without the notion of an entity, reference would seem to be impossible. And without the notion of an event it would seem hard to conceive of a verb or any sort of time scale as expressed by the -ED in talked or the difference between grow and grew. Similarly, without the notion of what is known as the ‘Theory of Mind ‘or TOM, or understanding that others can think about thoughts just as we can, it impossible to make sense of thinking, knowing, or saying anything in particular. The problem with third factor cognitions is that they are hard to define because they are even more difficult to test experimentally. Mike Tomasello has shown experimentally that at least some chimpanzees have at least some version of a TOM. But very few of those most familiar with non-human primates have anything resembling a modern human cognition. As far as I know, only one academic has ever cited a non-human as a co-author.

While it seems obvious that third factor cognitions necessarily underly much of grammar, the relative weightings of them and UG are highly uncertain. Both and the relation between them are the topic of intense ongoing debate.

But whatever the relative weightings of UG and general coginition, at least some aspects of general cognition and all aspects of UG seem to be necessarily part of the human genome, and clearly not part of the genome of any other known species.

From an evolutionary perspective, the weighting looks quite different. It is hard to see how any of the seven evolutionary steps by the proposal here could have got off the ground genetically without a degree of more general cognitive infrastructure. For instance, it is hard to see how the first combination of a primordial prototype noun and some contrasting element could have occurred without some foundational cognition.