Notes on Paul Meehl’s “Philosophical Psychology Session” #04

These are the notes I made whilst watching the video recording of Paul Meehl’s philosophy of science lectures. This is the fourth episode (a list of all the videos can he found here). Please note that these posts are not designed to replace or be used instead of the actual videos (I highly recommend you watch them). They are to be read alongside to help you understand what was said. I also do not include everything that he said (just the main/most complex points).

  • Saying “it’s highly probable that Caesar crossed the Rubicon” is the same as “it’s true that Caesar crossed the Rubicon” (1st is object language, 2nd is meta).
  • Probability (about evidence): talking about the relation between the evidence and the theory that the evidence is for.
  • Verisimilitude (ontological concept, refers to whether the state of affairs obtains or not in eyes of omniscient Jones) is NOT a statement about the relationship to the evidence (can’t be equated with probability); it’s a statement with relation to the world, whatever is the case (on the correspondence view).
  • Caesar crossed the Rubicon is true if and only if he crossed the Rubicon.
  • Caesar crossed the Rubicon is probable if there is sufficient evidence in the Vatican Library for you to believe that he did. Caesar crossed the Rubicon is probable if there is sufficient evidence for the Centurion sat next to him as he crosses the Rubicon.
  • Difference between the content and the evidence in support of it
  • This distinction completely torpedoed the “meaning is the method of it’s verification”.
  • Verisimilitude is a matter of degrees (just as confirmation).
  • Science theories can differ in how true they are (non-binary).
  • Logically, can argue that a science theory is false if it contains ANY false statements.
  • Falsifying any conjunct in the argument immediately falsifies the conjunction.
  • T(false)=S1(true),S2(true),S3(false). Falsifying any conjunct in a conjunction, falsifies the conjunction.
  • But we have to talk about degrees of truth to get anything done. Unsatisfactory but there have been (unsuccessful) attempts to quantify degrees of truth (true statements-false statements/total statements). Probability is also unsatisfactorily defined (logicians can’t agree if there’s one or two types of probability).
  • In psychology (when using statistics), use the frequency concept/theory:
  • When we are evaluating theories, we may use the other kind.
  • Kinetic theory of gases & gases: explain equations about gases (their volume, temp etc.), you could derive it by principles of mechanics. Theory of heat reduced to non-thermal concepts (concepts of mass, velocity, collision). They did this by imaging a cylinder of gas which contains molecules. These molecules act like billiard balls (with mass, velocity etc.). Degree of heat (temperature) and amount of heat are different things.
  • Scientists like going down the hierarchy of explanations.
  • Kinetic theory doesn’t work under extreme conditions. According to strict Popperian falsification, an example modus tollens so kinetic theory must be rejected. But we don’t do that. To abandon the theory is not the same as falsification.
  • Instrumentalists don’t care about truth (only utility), realists would have to reject it (but could recognise that part of it is false and so won’t totally abandon it).
  • Thinking about how the kinetic theory was false (in idealised Popperian form) allowed researchers to further explore it and it becomes a corroborator as thinking about how it was false tells us how to rewrite the equation and fit this model to the facts much better. You can be far along enough with your theorising to know that the theory is idealised, you use that knowledge to change equation (don’t need theory powerful enough to generate parameters, can be done in psychology).
  • One kind of adjustment is to change the theory to fit the facts (as with above example). Other is to change belief about the particulars (e.g. Planets weren’t behaving as they should, hypothesised that there might be another planet in such and such a place. Point telescope there and voila, Neptune).
  • Primitive statements are more important in some sense than others.
  • We need an idea of the centrality of a postulate (ideas that are crucial to the theory and can’t be dispensed with) and the peripherality of a postulate (those that can be amended and you still agree with that theory). Any way of getting at the theories’ verisimilitude that doesn’t take this into account is unsatisfactory. Why you can’t measure verisimilitude by a nose count/sentence count (gives same weight to central and peripheral postulate).
  • Core and periphery isn’t particularly explicated (and you can’t do it). Attempt: Can I derive the theory language statement (which come pairing this theory language statement with the replicated experiments) from the postulates of the theory. If not, theory is incomplete. You look at the derivation chain from each of the postulates to the facts and see how many derivation chains contain common postulates. If there are postulates that are common to all derivation chains, can be said to be a central postulate.
  • In any theory, we have a set of theoretical postulates. We have a set of mixed postulates (postulates that contain a mixture of observational and theoretical words).
  • Derive from that statements that are all observational. Makes a theory empirical. What makes it empirical is there is at least one sentence that can be ground out by applying laws of logic and mathematics to the theory that does not contain theoretical words but only logic, maths, and observational words. Some words are obviously observational (black) others are not (libido) but some are fuzzy e.g. 237 amps. But we have an ampmeter and a theory about ampmeters and trust the measurements, so it’s observational (but can be disputed). You can link observations together (and therefore observational statements) via theoretical statements.


  • Kinds of theoretical entities: Russell said there was only 1 (events). Meehl’s ontology: substances (chemistry sense e.g. elements), structures (including simples e.g. quarks), events (e.g. the neuron spikes), states (e.g. Jones is depressed, I am thirsty. Difficult to distinguish between events and states, could say events are just states strung out over long time intervals), dispositions (if x then y, -ble e.g. soluble, flammable; they are dispositional predicates), fields (magnetic fields). Can be used to analyse any concept in social sciences. Important kinds of events are when structures or substance undergoes a change in state, then changes dispositions. Power of the magnet to attract is a first order disposition. Iron being able to become magnetic is a second order disposition. Supreme Dispositions are dispositions that an object must have in order for it to be that object. The list helps think about the laws and theories are present in science. Most laws turn out to be compositional, functional dynamic, developmental. Compositional theories state what something is made out of and how it’s arranged. Functional dynamic involve Aristotelian efficient causes; if you do this then this will happen. Changes in state will result in changes in disposition over time. When comparing theories in similarity, list the kinds of entities and compare. How do they connect (compositionally and functionally). If you’ve drawn functional connections or time changes in developmental statements, can ask what’s the sign of the first derivative? You don’t claim to know what the function is; does x go up or down with y. You’ve got a strand in the net connecting entities. What about the sign of the second derivative?
  • Continuous case: [partial]F(x1x2) [over] [partial] x1 is greater than [partial]F(x1x2) [over] [partial] x2 EVERYWHERE. Means x1 is a more potent influence on y than x2 but still didn’t tell you what function is.
    X1 and x2 = two inputs
  • Discontinuous case: when the influence of x2 is greater depending on x1 being small.
  • Allows you to order partial derivatives.
  • Interaction effect: y is the output. (y with a present, y with a absent)when b is present-(y with a present, y with a absent)when b is absent. The difference between them is not zero.
  • The effect of y when b is there is greater than the effect of y when b is not there.
  • Theories can look the same/have the same connections/have same entities, but this theory makes this influence much more powerful but this only appears when you look into the derivatives (you see that there is an interaction).
  • Fisher effects=partial derivatives for continuous case. Interaction=mixed partial derivatives for continuous case.


Yonce, J. L., 2016. Philosophical Psychology Seminar (1989) Videos & Audio, [online] (Last updated 05/25/2016) Available at: [Accessed on: 06/06/2016]