philosophy is for everyone
and not just philosophers

philosophers should know lots
of things besides philosophy

PhiloSophos knowledge base

Philosophical Connections

Pathways to Philosophy programs

University of London BA

Pathways web sites

Philosophy lovers gallery

GVKlempner: complete videos

PhiloSophos home

Pathways to Philosophy

Philosophical Connections

Compiled by Anthony Harrison-Barbet


(1905 — 1997)



Born in Oranienburg, Germany Carl Hempel studied mathematics and physics at the Universities of Göttingen, Heidelberg, and Berlin, where he gained his doctorate. He became a member of the logical positivist group there in 1934. After a short time researching in Brussels he emigrated to the U.S.A. in 1937. He taught at Yale from 1948-55 when he was appointed to the professorship at Princeton. he was also Fulbright Senior Research fellow at Oxford (1959/60) and a Fellow of the Center for Advanced Study in the Behavioural Sciences at Stanford (1963/4). After retirement he continued to teach at the Universities of California (Berkeley and Irvine) and Jerusalem.



[1] While Hempel accepted the logical positivists' approach to the concept of meaning, he came to modify it in a number of respects. Initially he suggested that statements are meaningful only if they can be translated into an empiricist language. But later he allowed for degrees of meaningfulness between the extremes of significance and nonsense [a]. He also argued that meaning should properly be located in language systems as a whole rather than in separate statements. He was thus committed to a coherence theory of truth [b]. He recognised that there is a problem with theoretical terms utilized in the sciences in that they do not refer to entities which can be observed, and he suggested that an 'interpretive system' consisting of both theoretical and observational terms might allow for a partial interpretation which would produce consequences that could be tested empirically [c].



[2] Hempel developed the 'covering-law' model of scientific explanation (that is, as applied to the natural sciences) and sought to extend it to explanations in the social or human sciences [a]. According to the model an event (for example, the apparent bending of a stick when placed in water) is explained when statements describing it are deduced from statements of antecedent conditions in conjunction with general laws (which may in some cases be statistical and only probable), and which in turn are deducible from more widely ranging laws. Thus we have the following pattern: (a) a set of statements, C1, C2,... Ck, which describe particular facts; (b) L1, L2,... Lr, which are general laws; (c) the conclusion E, which describes the explanandum-event — (a) and (b) jointly forming the explanans. This deductive subsumption of the explanandum under principles which have the character of general laws Hempel calls a 'deductive-nomological' explanation. He also identifies a second type, which is probabilistic and therefore inductive (for example, an explanation of the occurrence and subsidence of a person's allergic attack) but he tries to show such explanations conform to the primary model and are still nomological (that is, appertaining to laws), although not deductive

How is evidence of conditions obtained? What are 'general' laws grounded in? Why do we rely on them? This is in effect the problem of induction: general laws are confirmed by their instances. Some philosophers of science have accepted a number of confirmation criteria: (1) 'All As are B' is confirmed by any (A and B); (2) 'All As are B' is disconfirmed by an (A and not-B); (3) 'All As are B' is neither confirmed nor disconfirmed by any non-A; (4) Whatever confirms one group in a set of logically equivalent hypotheses confirms the others. However, it was claimed that holding these criteria in conjunction leads to paradox. To deal with this Hempel suggested that condition (3) be rejected; and accordingly he allowed that anything other than an (A and non-B) would confirm 'All As are B'. Thus, for example, since 'All ravens are black' is logically equivalent to 'All non-black things are non-ravens', it would seem to follow that anything at all (a white swan, a black mouse) would confirm 'All ravens are black'. This is Hempel's paradox. What is at issue is how instances would be regarded as relevant; and any formulation of the inductive principle should take account of this requirement [b].

As for the social and human sciences, Hempel discusses ["Explanation in Science and History"] two kinds of explanation supposedly used by historians — 'genetic' explanations and appeals to 'motivating reasons'. If we seek to explain why a particular historical phenomenon occurred, or why someone performs a particular action, we seek to show that the phenomenon or action can be derived as a conclusion from sets of relevant facts and well-established generalizations [c]. In cases of specific actions of an individual we have some such pattern as this: (a) A was in a situation of type C; (b) B was disposed to act rationally; (c) any person who is disposed to act rationally will, when in a situation of type C, invariably (with high probability) do X — which, Hempel claims, conforms to the probabilistic version of the covering-law model. Again he thinks of these patterns of explanation as nomological — though they may be combined with some straight description.



While Hempel's approach to the concept of meaning arguably marks an advance on standard verificationist theories in that he would appear to mediate between physicalist and operationalist, or instrumentalist theories, his appeal to observational data as the criterion of significance (even in the context of 'partially interpreted' deductive systems) leaves him open to criticism. One reason in particular for this is that many terms used in our discourse cannot easily be linked to such data. Indeed they would often seem to be used meaningfully when it is not known whether properties designated by 'empirical' terms are, or even could be observed.

Hempel's main contribution to philosophy, however, probably lies in his work on methodology. But here too his attempt to apply covering-law models to historical and genetic explanations and his paradox of confirmation have engendered a great deal of debate. In relation to the former it can be argued, for example, that motivating reasons cannot be assimilated to causes in the way he requires; and that the model cannot cope with exceptions without either emptying of all meaning the concept of general explanation, or treating human behaviour as inevitable and completely predictable. As for the paradox of confirmation, some philosophers have responded by arguing that statistically observations of black ravens, for example, (given that all ravens are black) have greater confirmation value than those of non-ravens, say, white swans; and that background knowledge can increase probability. Another approach is to reject all attempts at confirmation. 'All ravens are black' is to be accepted only to the extent that there has so far been no observation instance of a non-black raven; and appeals to classes of non-ravens are irrelevant. Hempel's paradox, however, continues to cause puzzlement.



Hempel: Aspects of Scientific Explanation and Other Essays in the Philosophy of Science (1965) (contains many important essays); Philosophy of Natural Science (1966); see also "Explanation in Science and History" (1962), in The Philosophy of Science, ed. P. H. Nidditch.


W. Dray, Laws and Explanation in History.

J. H. Fetzer, The Philosophy of Carl G. Hempel: Studies in Science, Explanation, and Rationality.

I. Knopf, The Anatomy of Enquiry.






Note: the general influence on Rorty of Hempel's approach to philosophical problems.


[1a] Statements meaningful if translatable into empiricist language; [later] degrees of meaningfulness    Carnap [1a]


[1b] Meaning located in whole language system; coherence theory of truth







       [Neurath — see note in Carnap's Profile]


[1c] Theoretical and observational terms: partial interpretation — empirically testable consequences



[3c 4b]



[2a c] Covering-law model of scientific explanation (causal) — extended to social/ human sciences; phenomena, actions derived from facts and generalizations
















[4a b]



[2a b]


       [*Influence via neo-Wittgensteinians
   such as Peter Winch and Charles Taylor]


[2b] Induction — confirmation of instances; Hempel's paradoxes (concerning relevance of instances)




[3a 5a b]


[4c d]