The Invisible Constitution in Comparative Perspective

The Centrality and Diversity of the Invisible Constitution 

155

The lesson that is relevant for present purposes is this: the rules of inference 

that underpin the validity of arguments are not themselves to be considered 

among the constituent elements of the argument.

When we turn from formal, logical systems to informal and natural language 

systems such as legal systems, the notion of a rule of inference becomes less 

clear. Nevertheless, a version of Carroll’s point continues to apply. Consider 

some putative law L which is (at least hitherto) unexpressed, and which has 

been derived from some relevant legal source by application of rules of legal 

inference; and let R be the rule of recognition, S the source from which L has 

been derived, and I the legal rules of inference. On the ‘thick’ Hartian picture 

of validation, for L to actually be a law it must be the case that R validates S 

(as a source of law) and L (as the putative law derived from S). Given that 

(ex hypothesi) L is not expressed by S, validating S will not straightforwardly 

validate L. Rather, it will have to be the case that L follows via I from the law 

that is expressed in S.

Can  I be written down, as principles of the legal system in question? 

Presumably yes: just as we can write down the rule of inference that underpins 

the argument from (1) and (2) to (3) – it is written down, above, as (4) – so there 

is no reason to suppose that we cannot, at least in principle, write down I.  

However, it seems that our inference to L from what is expressed in S could 

not depend upon our application of a written I, for the reason that Carroll’s 

essay points us to: this would then require application of some further princi-

ple of legal inference (call it X) via which L follows from both I and the law 

that is expressed in S. And ex hypothesi X would be unexpressed.

It might be objected that, in the informal reasoning that is characteristic 

of a legal system, that the same I should be able to govern both: (a) the infer-

ence to L from what is expressed in S; and (b) the inference to L from I and 

what is expressed in S; in other words, it might be objected that X need not 

be any different from I. However, this possibility suggests a new consideration 

983, 984. J. F. Thomson characterises the infinite regress as ‘just an infinitely long red herring’:  

‘What Achilles Should Have Said to the Tortoise’ (1960) 3 Ratio 95. His point is this: suppose an  

argument to (3) taking (1) and (2) as premises were not valid because, in general, no argument 

from a finite set of premises to a conclusion could be valid if not supplemented by an addition-

al premise stating the rule of inference for such an argument (call this general principle the 

supplementation requirement). Then the premises as supplemented would equally fail to yield 

a valid argument, because we would still have nothing more than an argument from a finite 

set of premises to a conclusion – which fails to satisfy the constraint stated in the supplemen-

tation requirement! Hence the supplementation requirement states a constraint on validity 

that necessarily cannot be satisfied, hence is absurd, and hence is false; and we do not need to 

appeal to the infinite regress to see that this is so. Nevertheless, the idea of the infinite regress 

has heuristic utility, and is commonly used to explain the point of Carroll’s paper (e.g., both 

Coffa and Musgrave use the device), and hence I have used it in the text.

---

156