because the claim itself is incoherent and second because what is meant by a generalisation is over-specified. Barrow appears to treat a generalisation based on educational research as a universally quantified proposition of the form:
For all x, if Fx then Gx
which is refutable by a case of an F which is not G. But although this is how generality is treated in textbooks on deductive logic, it is not necessarily how it is thought of in EER or indeed in many other contexts. We can tolerate some exceptions to a generalisation often because generalisations are implicitly understood to mean ‘Nearly all cases of F are G’. Second, they are very often hedged with an implicit contextual limitation so that they only have a limited range of application. Barrow, however, seems to go further than this and to wish to deny that a teacher (and by implication a researcher) could develop an explanation that applied to more than one pupil at a time. This move, however, raises an issue which we will consider in more detail in the next chapter. A successful explanation of a phenomenon is usually one that allows some purpose to be achieved. This does not entail that it has to describe, in minute detail, all the particulars of a chain of events which purport to be covered by the explanation. It suffices that the phenomenon is explained in such a way as to usually lead to greater understanding of or to the improvement of practice. There are ways of assessing the quality of explanations in this respect which any explanation will need to take account of. To take an example, if one is trying to explain why a method of teaching reading is relatively successful compared to alternative methods, one is required to show that it is a better explanation of why than any other that is practically feasible. It does not require that the explanation is an exhaustive account of every detail of each individual child’s learning to read.
The key point is that explanations have to address particular problems or concerns raised by researchers, not that they either confer certainty or exhaustiveness in relation to the phenomena with which they are concerned.5 If we were to take such objections as the above seriously then educational explanation would be impossible in most cases and thus our knowledge of educational practices could not be attained through systematic investigation. A good case could therefore be made out for saying that it is unknowable. These considerations should lead us to provisionally dismiss objections 2 and 3.
Difficulties Related to Context and Identity
Related objections that occur repeatedly in the work of Andrew Davis (e.g. 1995, 2015) concern the complexity and stability of educational phenomena. The context-dependent nature of educational phenomena has already been remarked on in relation to Barrow’s work. However, Davis also claims that transferability is also a major problem. Transferability applies to knowledge and know-how. Thus something learned in one context might not be applied in another (e.g. formal mathematics in the classroom applied to practical situations in household budgeting). However, the evidence for non-transferability is weak and if it were largely true, would jeopardise the rationale for much formal education. The second part of the objection, that concept-dependent educational entities like schools have very weak criteria of identity and cannot thus be meaningfully spoken about outside very specific contexts is, perhaps, a more serious objection.
Certainly, if the criteria for identity of such things as schools or teaching methods had to be as robust as those for spatio-temporal objects then the objection would have some force. However, there is no reason to suppose that this is so. We do not find difficulty in talking about the same school or the same teaching method over a period of time, either in ordinary lay discourse or in a research context. Nevertheless, the objection does point to something important, namely that we need to take care in elucidating criteria of identity when talking about educational practices and institutions. This takes care of objection 4.
Most Educational Findings are False
There are two ways in which we can approach this claim. The first is through a generalised falsificationist approach (e.g. Popper 1959). In this sense, no finding of educational research, however robustly conducted, could be true. But this point, dependent on Popperian views of the provisional nature of scientific assertions, is just as true of the results of any scientific enquiry as it is of educational research. Underlying it are two questionable assumptions. The first concerns the unreliability of induction, a Humean concern which Popper appears to endorse (Popper 1959; Hume 1978). Induction, by its nature, cannot deliver deductive certainty, which is not to say that we cannot be certain about many matters established by induction, just that certainty in the cases of robust inductive inferences does not exclude the appearance of counter-examples.
The second objection is closely related to a conception of truth that we had cause to question in the previous chapter, namely that truth attributions are only really justified if they are true for all time. We had cause to reject this view because it appeared to rely on an untenable substantive account of truth, the correspondence theory, whereby correspondence rests on an alignment with a proposition and a (timeless) fact. It also rests on the idea that ‘true’ means ‘true for all time’. The two views are connected. If a proposition is true it is so because it corresponds with reality, and reality in the form of facts is fixed and cannot be changed. However, we are always making new discoveries about reality, and sometimes we are required to revise our views about what constitutes reality. Thus our attributions of correspondence with reality are nearly often fallible and hence invalid. To understand why this objection is not convincing we need to remind ourselves of the criterial account of truth outlined in the previous chapter. Truth attributions are indefeasible but criteria for attributions of truth are revisable (Ellenbogen 2003). This applies to any sort of enquiry, including any sort of scientific enquiry and thus EER is no special case.
However, it may legitimately be objected that EER is especially unreliable, not just for the reasons considered and dismissed earlier on, but also because as a matter of fact it is of poor quality, poorly designed and executed and often of a lofty ambition which is by no means matched by evidence and analysis. This kind of objection must be taken more seriously by anyone concerned with the quality of EER, particularly as there are many instances of poor-quality work which suffers from conceptual confusion, methodological inappropriateness, inadequacy of empirical base and poor analysis. In fact, EER is littered with large-scale theories which suffer from one or more such problems – the theory of IQ, various forms of developmentalism, verbal deficit theory, theories of language acquisition and concept formation and so on. Does the fact that so many educational research programmes have failed, either comprehensively or substantially, tell us anything?
The argument of this chapter is that it does, but not in a way as to justify such practices which, hopefully, belong to the infancy and youth of EER. The very fact that there is good reason to believe that they are false tells us something valuable about the phenomenon that they purport to describe and explain. False educational theories have a great deal of value, even if the intention of those who generated them was not to produce either complete or partial falsehood. Consider a false theory6 P. If P is false then its contradictory, not-P is true. This is an important result, which, however, is only the beginning of the need for further investigation.
A theory like P will not only contain many propositions, but some of them may well be true, even though the conjunction of the whole is false.7 However, the position is more complex than this since it is highly unlikely that there are no inferential relationships between the propositional components of P. These may be deductive relationships, as when theorems are derived from axioms as in geometry or axiomatic logic, they may be material inferences, as when conceptual relationships between elements of a theory are expanded (a star which is part of a solar system has planets within its gravitational sphere of influence) or inductive, when well-established empirical connections are embedded within the theory (societal economic breakdown is followed by famine). It is thus often the case that the falsity of one element of a theory has implicational ramifications for the truth of other parts.8 In order to understand the real significance of a falsified component of a theory, the theory needs to be well understood