Hereafter, abbreviate ad hoc modifications to AHM and hypothesis to HT.

Source: p 636, A Concise Introduction to Logic (12 Ed, 2014) by Patrick Hurley.
To minimise this post's length, I do not quote the definition on 'hypothesis' on p 610, but please request it if needed.

The problem with ad hoc modifications is that their purpose is to shore up a failure of evidentiary support in the original hypothesis.

Abbreviate the Original HT to HT1. Then HT1 + first set of AHM = HT2,
HT2 + 2nd set of AHM = HT3,   etc...

Any modification to a HT1 implies rejection of HT1; otherwise you would not have modified HT1!
1. So AHM are identical to proposing a new HT? Why does shor[ing] up a failure of evidentiary support injure, rather than improve?

As more and more modifications are added, the hypothesis becomes self-supporting; it becomes a mere description of the phenomenon it is supposed to explain.

  1. HT should describe 'the phenomenon it is supposed to explain' : What is wrong about this?

  2. All HT are benefited by AHM that improve a hypothesis. So How can AHM make a HT self-supporting? The question of AHM differs from the question of testing, because independent of (the number of) AHMs, any modified hypothesis must still be tested.

2 Answers 2


Hurley's use of the word "evidentiary" suggests that he is talking about real world hypotheses in this section of the text, so in order to get a feel for what Hurley is talking about here let's look at a real world example.

In the 1990's observations were made that undermined the accuracy of our Big Bang model of the universe. Firstly, it was observed that the rate at which the universe is expanding is accelerating. Secondly, it was observed that the rotational characteristics of galaxies are uniform. This is not at all what the prevailing cosmological model predicted.

Cosmologists were faced with a choice. They could make ad hoc modifications to the existing cosmological model or they could extend the existing model by introducing new auxiliary hypotheses in order to explain these unexpected observations.

The introduction of ad hoc modifications is not desirable for a number of reasons. The existing model had been working in a satisfactory way up until then. Making ad hoc modifications may have unforeseen consequences leading to the need for further ad hoc modifications, and then more ad hoc modifications, etc.. This would undermine the desired simplicity of our scientific model and may also restrict its fruitfulness and accuracy. As the number of ad hoc modifications increases, you run the risk of ending up with a model that features an ad hoc modification to describe every observed feature rather than a model that predicts observed features.

Instead, cosmologists chose to add two new auxiliary hypotheses to the model, namely dark energy and cold dark matter. This results in simpler model than a model based on ad hoc modifications. It maintains its fruitfulness and accuracy. Also, importantly it is a more elegant solution to the problem.

(Also, tut-tut for not taking a break.)

  • +1. Thanks again. I will review this after a few days of break. I was too close to finishing the book to take a break today; I have finished the book now and so shall commit myself to a break.
    – user8572
    Jan 10, 2016 at 6:05
  • @LePressentiment I forgot to mention that there was a third option - namely to chuck out the model and start again. Obviously this is not a desirable option since the model had been working well up until the unexpected observations.
    – nwr
    Jan 10, 2016 at 6:14

An HT or a theory would feature an idea, a model or a narrative about what is happening, how it's happening and "why" it's happening (in a technical sense).

Now if one of these predictions ends up being false, this means that somewhere within this idea, there's a flaw. Now you could of course patch that with an AHM and declare "It's all of the before except for that strange anomaly". And that would work reasonably well in terms of describing what you already knew (as that is what HT already was supposed to do) as well as describing the anomaly (you added that part).

But it would leave you with a lot more doubt as to the reliability of your predictions. Because you know for a fact that there is a problem with the HT, yet it's still this part (HT) that does the heavy lifting of your HT1. So while there's always a level of doubt towards the predictions made by a theory if you already know that it's definitely, and not just possibly, wrong that makes it's predictions kinda dubious.

And apparently to Karl Popper those predictions and the acceptance of defeat are what distinguishes science from pseudoscience. Because the more AHMs you accumulate the more it will look like as if you're just interested in people doing what HT proposes (as the default first step) regardless of whether that is working or not. Like think of what it would mean if your HT is some religion. Then you would never actually question the premises of that religion but for every attempt to refute it you would just argue "yeah but that's an isolated incidence that doesn't mean anything".

And if you accumulate AHMs for long enough you just end up with a list of exceptions which not only loses all of it's predictive power but also makes it impractical for application.

So ideally you want to go back to a consistent theory that is able to make predictions and has no exceptions. So either you'd start anew if there are really fewer and fewer redeeming qualities of the original HT or you'd aim to find a general theory of which HT is just a special case or argue that it's a superposition of HT and an auxiliary HT.

Either way you'd like it to be a new HT and not just HT' with exceptions that you can't explain within that theory.

You must log in to answer this question.