Its natural ground

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Hilary Kornblith

Kornblith has two main questions that he seeks to tackle in his book Inductive Inference and Its Natural Ground. They are: What is the world that we may know it? And, what we that we may know the world?[1] The book is set up in two parts, the first pertaining to the first question and the second being tackled in the later half. The book explores the claim that inductive knowledge is possible and inductive inference is reliable, indeed Kornblith states this explicitly at the conclusion of the first chapter, claiming his book is "an attempt to explain in some detail how inductive knowledge is possible, and it is against that standard that its success or failure should be measured."[2] He primarily seeks to support this by the claim that there is a fit between the structure of our minds and the structure of the world. The book itself is formatted around these two argumentative points, the first half of the book being dedicated to the structure of the world and the second to the structure of our minds.

The underlying aim of Kornblith's Inductive Inference and Its Natural Ground is two fold. First and foremost is the intent that Kornblith explicitly states to be the purpose of his essay, "to explain in some detail how inductive knowledge is possible." However there is also the more underlying consideration of Kornblith using his book to defend the use of the Law of Small Numbers in inductive inferences. However it seems that the inferences that Kornblith attempts to justify are not inductive inferences at all, but are rather misconceptions.

As the second half of the book is heavily grounded and reliant upon the success of the first, any instability in the first half of the book will inevitably cause the seconded to loose its argumentative basis. Therefore it is the metaphysics of Kornblith's argument that should first be considered, for if there can be found any real inconsistencies, then there need not be any more work done. It is important to note that there is a significant portion of the later part of the text devoted to the defence of the use of the Law of Small Numbers in inductive inferences, and, although this defence is dependent upon the metaphysics of the first section of the book, it non the less warrants a short mention. The Law of Small Numbers refers to a judgmental bias which occurs when it is assumed that the properties of a sample population can be estimated from a small number of observations.

Kornblith attempts to justify inferences using this law that are not, in matter of fact, to be considered inductive inferences. An example Kornblith uses is that "Upon noting that one sample of copper conducts electricity, we infer that all samples of copper conduct electricity."[3] However, this inference is actually knowledge based deductive reasoning, and an inference of this kind is made with specific background knowledge. In this case the background knowledge would be that "All samples of the same metal share the same resistivity or conductivity value."[4] However, whether or not Kornblith's defence of the Law of Small Numbers succeeds, it is but a factor to the over all purpose of defining inductive knowledge and it is with this particular point that we must concern ourselves.

The first section of Kornblith's book focuses on the structure of the world. Kornblith approaches this concept by examining and furthering the earlier studies of Locke. He claims that Locke's stance on natural kinds was overly sceptical as a result of the progresses in the sciences at that period of time, he takes the metaphysics posited by Locke and advances them for an argument for a realist view of the world. Kornblith makes the claim that natural kinds are real, and that they are homeostatic property clusters responsible for the observable properties corresponding to the kind in the world, borrowing from the works of Richard Boyd. He further takes an anti-reductionalist account that higher lever natural kinds are real, which is shown by their successful employment in special sciences, as well as lower level natural kinds. The possibility of induction requires this realist metaphysics as a foundation to succeed.

A Homeostatic Property cluster, as put forward by Boyd, is a definition of a "category of objects such that there is no uniform set of necessary and sufficient conditions that all and only the member objects satisfy. Though it is a conceptual matter which properties are included, there is room for ambiguity or indeterminacy in the set of objects satisfying the definition."[5] A commonly used example of a homeostatic property cluster would be the concept of health. Health is determined by a person possessing a sufficient number of properties that reside in the property cluster, but the absence of some will not necessarily mean the person being unhealthy. Health is the higher order property of the homeostatic property cluster while properties like the state of someone's immune system, their cardiovascular system, their diet and general fitness all consist of lower level properties. A person can have any combination of the lower level properties and still be deemed healthy. Health is therefore the natural kind and has the predictive power of the generalisation that is to be made from the higher order term, unlike the lower level properties which make it up. Kornblith uses this definition as the basis for his account of real natural kinds.

The second half of Inductive Inference and Its Natural Ground, however, focuses specifically on the structure of our minds, "what are we that we may know the world."[6] Using the metaphysics elaborated by his exploration in Part 1, Kornblith then goes on to justify the psychological tendency to understand the world in terms of essential kinds. He appeals to the results of certain psychological experiments made of children which he believes to provide evidence supporting psychological essentialism. He suggests that there is the natural tendency to make a firm distinction between the inside and outside of objects, and to classify things by their essential kind rather then their appearance, if there is a confliction, present in very young children. This tendency supports the concept of an innate tendency to think about the world in terms of essentialist kinds. This is interesting as Kornblith's essentialism deviates from Quine's philosophy, of which Kornblith's project is based, significantly in that Quine uses natural selection to justify our inductive practices, while Kornblith justifies induction with the success of science, but uses evolution as the explanation for induction.

As stated earlier, Kornblith uses the works of Locke as a basis for his projects investigation. According to Kornblith Locke has three main metaphysical pictures that he simultaneously presents, even though these pictures may not necessarily be consistent with one another. The first view is the conventionalist view that there are no real kinds in nature, the second is that there may well be real kinds in nature but what they are is entirely unknowable to us, and the final view is that there are real kinds in nature and that they are know to us by their boundaries. Kornblith provides criticism for each of the three metaphysical claims, which he then follows up with his own based on those criticisms. His claim is that natural kinds are real, completely knowable and that they provide the basis for the reliability of induction.[7]

Korblith makes an assumption that induction is reliable beyond doubt based on a comparison between the reliability of induction and the success of science.[8] He does not necessarily specify what he means by reliability, rather posits that the success of sciences on their own, which are needless to defend, warrant the reliability of induction. This line of arguing is not intended to defend against Humean scepticism but rather seeks to come up with the best possible explanation for why induction is reliable, given that it already is. In other words, Kornblith seeks to provide people who already believe induction with a verifiable explanation as to why they believe it.

If scepticism is immediately taken off the table as a realistic possibility, then it is a fairly logical step that there must be something accountable for the constant conjunction of certain observable properties in nature. From there it is reasonable to assume that the best possible explanation is that there are real natural kinds and unobservable properties that form homeostatic clusters. However it would appear that Kornblith is in fact making two simultaneous claims about the nature of real natural kinds. The difficulty lies in the shift of meaning that takes place when Kornblith refers to the word 'real', and the distinction between non-conventionalism and realism. This shift, while apparent through the breadth of his discussion on conventionalism, is embodied when Kornblith states that "Whatever our conceptual activity, and indeed, even if we were to have no conceptual activity, the relationship between hydrogen and oxygen in water would form a homeostatic unit."[9] There are in fact two claims being made independently of one another in this sentence. The first is that "whatever our conceptual activity, the relationship between hydrogen and oxygen in water would form a homeostatic unit", which is a non-conventionalist claim, where as the second position, that "even if we were to have no conceptual activity, the relationship between hydrogen and oxygen in water would form a homeostatic unit" is a realist one.

Initially, when Kornblith uses the word 'real', he refers to natural kinds existing, not being the result of human conventions and being independent of human conceptual activities. This is a non-conventionalist position, however, as the argument progresses, the meaning of the word shifts to that of realist one. The realist one is a metaphysical position of real that implies that natural kinds are objectively real in all possible worlds. This is a claim that is unsubstantiated. The distinction is that realism holts that natural kinds exist in the same way in all possible worlds, while the non-conventionalist that natural kinds do exist in this world, but that may not exist or exist in a different form in other possible worlds.

The first position, that "whatever our conceptual activity, the relationship between hydrogen and oxygen in water would form a homeostatic unit" is true by virtue of holding that a natural kind is a non-conventional natural kinds, such as water being composed of H2O. They are independent of our particular conceptual activities, but doesn't prevent them existing in different ways in other possible worlds. The second position, however, posits that a natural kind, like water, is a real natural kind, and exists the same way in all possible worlds. In a possible world where there is no human conception at all, water would still be H2O. The natural kinds would be objectively real, which is the position Kornblith takes, that the world is independent of any observing perspective and purely objective.

This is unverified as, even though Kornblith takes this realist position, he does not explicitly defend it, rather all his supporting evidence instead provides a defence for a non-conventionalist stance. When Kornblith does confront the notion of possible worlds, or as Kornblith puts it "the world without human conceptual activities at all"[10], he does so with the same human temporal and spatial conceptions.

One of the key features to Kornblith's definition of natural kinds is the notion of stability. What he considers to distinguish natural kinds form non-natural kinds is whether the properties that make up the kind are considered to be stable. Stability, or homeostasis, refers to whether or not the properties of a kind can be combined in a stable way and maintain themselves in a stable state. Kornblith supports this definition with his argument about gaps and chasms, that in nature we find certain combinations of properties to be stable and the co-variance of certain properties form these gaps and chasms. When confronting the prospect of the recognisability of stability Kornblith makes the assumption that stability can be defined in an objective sense in all possible worlds, however what he doesn't explain is how we can tell whether something is stable or not. In recognising the existence of unstable property clusters Kornblith creates the need to justify the reasoning behind them being stable or unstable and our capacity to recognise them as such. An unstable property cluster could reasonably be a stable property cluster that is objective stable in all possible worlds that we are just unable to recognise. Or, a better argument for the same point would be how do we know that an unstable property cluster is objectively unstable in all possible worlds? There are many examples prevalent in chemistry of chemicals that are unstable and instantly break down into two stable chemicals, which it is entirely possible to posit a world in which such a chemical does not break down and remains stable. Even if this is not the case, it seems that the evidence of an unstable property cluster on its own is not enough to justify the belief that it is objectively unstable across all possible worlds.

Water, H2O, is a stable chemical bond. So too is Hydrogen Peroxide, H2O2 and Hydroperoxyl, HO2. However not all combinations of hydrogen and oxygen atoms will be stable. Suppose we have HmOn, where 'm' and 'n' correspond to the amount of atoms respectively. Some examples of HmOn will be unstable, but that is not to say that they are objectively stable. For one thing, they may be stable for such an insignificantly small time for us to deem it meaningless, but this does not necessarily posit that the property cluster is objectively unstable, rather it reflects our own temporal perspectives and perceptions. We can only comprehend and define stability in a non-conventionalist way in this world, and are unable to generalise it to all possible worlds.

Therefore the problem faced by Kornblith in his metaphysics is that there is, inescapably, still the subjective temporal human perspective contaminating his objective realism. This impacts the rest of his work by suggesting a closer relationship between the mind and the world, although non the less Kornblith has admirably explored the question of induction and shed a great deal more light onto Quine's original project, it does imply that he has not quite yet reached a fully defensible answer to his project.


  • Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts
  • Sankey, Howard. Melbourne University Lectures
  • Pust, Joel (1996), 'Induction, Focused Sampling and the Law of Small Numbers', Synthese, 108:1 pp. 89 - 104
  • Hetherington, Stephen Cade (1997), 'Book reviews', Australasian Journal of Philosophy, 75:1, pp. 122 - 124
  1. Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts. pg 2
  2. Ibid. pg 10
  3. Ibid. pg 95
  4. Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts. Pg 96
  6. Ibid. pg 2
  7. Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts
  8. Sankey, Howard. Melbourne University Lecture.
  9. Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts
  10. Kornblith, Hilary (1993), Inductive Inference and Its Natural Ground, MIT Press, Cambridge, Massachusetts