Why is it reasonable to believe that the Grimm's Law is true? How can those things be scientifically investigated?
An obvious answer is that we can take an dictionary of native Gothic words and the corresponding Latin words, and see whether Gothic 'f' indeed corresponds to Latin 'p' (and other correspondences that the Grimm's Law predicts) at a rate greater than chance and calculate the p-value based on that.

However, there are two problems with that.
First, that, as far as I know, hasn't been done (or, at the very least, that having been done is not the reason linguists actually believe the Grimm's Law is true).
But the more serious problem is that that would be calculating the p-values post-hoc, which goes against the scientific method. In the scientific method, you come up with a function to calculate the p-value before you do the experiment. In our case, the Gothic words and the corresponding Latin words are known data, as if the experiment has already been done. So, calculating the p-value based on that goes against the scientific method.

Yet, almost everybody agrees it is reasonable to believe the Grimm's Law is true, or at least an approximation (in the same sense that Newton's Laws of Motion are an approximation). Why is it so?

  • 1
    This question belongs on another site in the Stack Exchange network: linguistics.stackexchange.com
    – J D
    Jun 9 at 18:40
  • 1
    I think the question itself is sound philosophically. What you're asking for is about the empirical nature of investigations in linguistics. But you're doing it, apparently, according to speculation on method. "as far as I know, hasn't been done" Thus, what you really should be after is the method used to defend the notion of Grimm's Law as there is probably a literature around the sub-discipline of sound changes within the greater auspices of phonology...
    – J D
    Jun 9 at 18:45
  • IOW, when you ask "How can those things be scientifically investigated?" this is a question of determining the methodologies before it is a question of defending the empirical character of the methodologies. You might catch a break and find a phonologist who is lucid in philosophy of language and philosophy of science, in which case I'm happy to retract my vote for closure.
    – J D
    Jun 9 at 18:47
  • Where exactly does "in the scientific method, you come up with a function to calculate the p-value before you do the experiment" come from? The point is to test a hypothesis against the null hypothesis, whether it concerns an experiment yet to be performed, or observational data already available. What difference does it make whether the hypothesis is formulated before or after? In cases of astronomy, geology archeology or, indeed, philology, it is typically after.
    – Conifold
    Jun 9 at 23:21
  • 1
    A problem with this question is that individual sound laws cannot in practice be viewed in isolation from other sound laws, nor without consideration of other linguistic changes. Grimm’s law has known exceptions, but these are the result of interaction with other changes, also regular, which happened before and after it. You could statistically analyse a hypothesis about which order all the various laws (and other changes in e.g. morphology) happened in and how they affected each other, but you can’t analyse just one sound law this way between languages as far apart as Gothic and Latin.
    – dpk
    Jun 10 at 12:41

2 Answers 2


I'm not sure that Grimm's law should be viewed a natural law akin to the laws of thermodynamics, rather it should be viewed as a condensation of empirical observations into a structure that aids in the understanding, communication and interpretation of historical linguistic data. It is more akin to morphological cladistics in biology than many of the results in physical sciences.

That being said it is predictive. If we obtained evidence of a previously undiscovered Germanic language that did not incorporate the sound changes predicted by Grimm's law, it would provide evidence against it. Based on the abstract, this study looked for deviations from Grimm's law and obtained a negative result. So it can play the role of other theories/laws/models* in science.

Grimm's law is widely accepted because almost all of the available data are consistent with (or interpretable within) that framework of understanding things. Different fields of science have different methodologies for assessing the significance an validity of claims; the statistical requirements for reportability in particle physics are very different from those accepted in psychology. The kinds of statistical analysis feasible in some areas of science are not always applicable to the assessment of hypotheses about singular historical events, like the evolution of proto-Germanic.

* Law/theory/model My observation is that whether something is called "X's law" vs. "Theory of X" vs. the "X model" is mainly dictated by the time at which the idea is formulated. Though there is some correlation in these terms with the "scope" of the idea, some things that today would probably be called models would probably have been called laws if formulated in the 18th or 19th century.


"In the scientific method, you come up with a function to calculate the p-value before you do the experiment." Quite right. But in this case, it is not possible to do any experiments, so you need to do something else. This is not unique to linguistics. This particular law was developed by noting regularities in the way that sounds changed in certain languages and systematizing them. The criterion then was whether the system accurately described the observed sound changes. That sounds exactly like the classic scientific method to me.

The reason you can't apply probability in this case is that there is no way to identify all the possible outcomes at the point of any change.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .