I'm having a philology test next week. One of the questions will be to "translate" an indoeuropean word into a germanic word, like: i.e. Agros -> germanic Akraz (i.e. "g" --> germ. "k" for Grimm's Law, excetera). I understood that there are some exceptions to this law, like the Verner's law. Are there other exceptions or can i say that every i.e. "p" turns ALWAYS into "f", or that every i.e. "b" turns ALWAYS into "p"?
Verner's Law is not an exception, since Germanic did not emerge in one sudden leap from PIE, it is a complication, i.e. there is another law that has to be factored in. Grimm's Law happened, and then Verner's Law also happened. You can't say "except for further developments under Verner's Law, p t k always become f θ x", because there are other complications. PIE *steigh "climb" shows up in OE forms such as gestīgan, Icelandic stíga and do on – there is another complication (GL does not apply after a fricative). Sanskrit téjate "to be sharp" and OE sticca are related, which illustrates another kind of complication (not an exception) – this IE root illustrates the phenomenon of "variable s", that s just sort of shows up sometimes. You could say that the Germanic (Greek etc) form reflects the root steig and the Indo-Iranian forms reflect the root teig, i.e. the reconstructed root itself is not totally fixed.
The word "path" seems to correspond to Sanskrit path-, constituting what looks like a serious exception to the claim that PIE p turns into Gmc. f. It is, however, thought that this word was borrowed from some Iranian language into Germanic, thus escaping the effects of GL. The word kitchen likewise corresponds to PIE pekʷ, exemplifying an extreme "irregularity", except the Germanic word is borrowed from Latin, cf. coquīna. That is, sometimes words are borrowed. There are also "exceptions" (though I cannot remember a specific one) where a word is iconic / onomatopoeic, and the correspondences seem to be off (the expected sound change does not happen).
It has been a virtual axiom in historical linguistics that sound change is regular, but recognizing that regularity requires knowing all of the complicating factors.
I would like to briefly add more info on how consonant clusters were affected by Grimm's law (based on a number of sources, e.g. Arsenieva et al. 2003; Ringe 2017, vol. 1 etc.). All the Proto-Germanic reconstructions are from Ringe. All of the exceptions below are well-known and usually studied in undergrad (at least, that was in my case).
- s followed by an obstruent (crucially, this is not only restricted to s-mobile):
Gothic stairno, OE steorra < PGmc *sternan- < PIE *h2stēr-s (cf. Latin stella)
Gothic fisks < PGmc *fiskaz < PIE *pisk- (cf. Latin piscis)
- two obstruents: only the first obstruent - if it's a stop - is affected:
Gothic ahtau, OE ahta, eahta < PGmc *ahtōu < PIE *h3(e)ḱtéh3 (cf. Latin octo)
OE nift, Old High German nift < PGmc *niftī < PIE *h2nep-t-iH (cf. Latin neptis)
- the coronal cluster *tst becomes ss:
OE sess < PGmc *sessaz < PIE *sedstós
geminates pp, tt, kk (often explained with Kluge's law, if you accept it)
Gothic fidwor, OE fēower (PGmc *fedwōr) seems to defy Grimm's law, cf. PIE *kʷétuōr
Several proposals have been put forward, analogy (Ringe) or /kʷ/>/p/>/f/ (Krahe).
word-initial PIE *gʷʰ > PGmc *b (Ringe 2017: 127-128), Gothic banja, OE benn, cf. PIE *gʷʰen-;
PIE *gʷʰ followed by a nasal was not affected: Gothic siggwan, ON syngva, cf. PIE *sengʷʰ - etc.