Darwin used the artificial selection of domestic breeding to introduce the concept of natural selection. Dawkins follows this, in particular using the breeding of the various types of dog from the wolf to illustrate the range of characteristics that can be achieved. Indeed, he points out that with almost any plant or animal we can breed for almost any trait we wish, such as maize for high or low oil content (p67) or rats for low or high susceptibility to tooth decay (p68). He also observes that in all such cases, though rapid change can be achieved in the first few generations, before long it tails off - not just because e.g. low oil content trends to zero, and you can't get any lower than that, but also the cultivation for high oil content tends to plateau.
Dawkins recognises that domestic breeding is through losing or subtracting genes, comparing it with removing pieces of stone in making a sculpture (p37). And this, of course, is why there is a limit to the degree of change achievable - because once all of the genes available in the original species that favour the desired trait are retained, and all those that detract from the trait are bred out, then no further change is possible through conventional breeding (ignoring genetic engineering which is now available).
So he is completely unjustified to extrapolate from the changes possible through domestic breeding (which involves the loss of genetic information) to macroevolution (which would require the emergence of new genes), yet this is how he tries to mislead his readers:
If so much evolutionary change [referring to dog breeding] can be achieved in just a few centuries or even decades, just think what might be achieved in ten or a hundred million years. (p37)
And this isn’t just a momentary oversight or over-enthusiasm on his part, for he repeats this false extrapolation at the end of chapter 3:
if so much difference can be achieved in breeding such different breeds of dogs in just a few centuries, think how much can be achieved over geological time.
From what he has written in these first chapters he must surely be aware that this is a misleading comparison. Of course, he says that the gene pool is added to by mutation (p37); but this is merely the evolutionary dogma. It is clear that he does not see this as taking place in the course of domestic breeding (certainly not to the extent of producing new characteristics), because he recognises (p56):
Domestically bred songs are longer, louder and more frequent than the wild ancestral type. But all these highly prized songs are made up of elements that occur in wild canaries, just as the habits and tricks of various dogs come from elements found in the behavioural repertoire of wolves.
He claims to be a science educator and is not slow to berate creationists for any misrepresentation of scientific facts - yet this is exactly what he's doing here!
The title of his chapter 3 is 'The primrose path to macroevolution'; but it's nothing of the sort - all but one of his examples of natural selection are comparable with the changes achievable through domestic breeding - i.e. merely through selection from an existing gene pool - whereas macroevolution would require new genes to arise. (The exception relates to bacteria, which I will discuss subsequently.)
Unfortunately this is typical not only of Dawkins, but of many evolutionary writers: the unsupported presumptions required to support the overall (macro)evolutionary theory (notably that new useful genes arise through mutation) are tucked away among the facts of microevolution (that substantial morphological change can be achieved merely using existing genes) - no doubt to give the impression that the unsupported assumptions are valid too.
It's interesting that in chapter 1 he discusses the dictionary definition of a fact:
... a particular truth known by actual observation or authentic testimony, as opposed to what is merely inferred ... (p14)
and he takes issue with 'inferred', arguing that although the overall theory of evolution is inferred from limited observations, the inference is as sound as any observed fact (p16):
... I shall show the irrefragable power of the inference that evolution is a fact. (p16)
But I think the definition is right, and particularly apposite here: it highlights the fact that the whole theory of (macro)evolution (which would require new genetic information) is an inference from the observed facts of microevolution (which merely involves the use of existing genetic information).
Evolutionists are keen to promote the notion that macroevolution is nothing more than accumulated microevolution. But this is wrong. There is a fundamental distinction that microevolution is based merely on existing genes (even though sometimes this may result in large morphological changes, such as the different dog breeds) and macroevolution which would require new genes.