Milk!

• Fortson makes a claim that it is not clear that PIE had a word for the noun milk, although it had a verb milk.

• Wiktionary 'milk' provides the following:
• NOUN's etymology:From Middle English milk, mylk, melk, mulc, from Old English meolc, meoluc (“milk”), from Proto-Germanic *meluks, from Proto-Indo-European *h₂melǵ-.
  
• VERB's etymology:Old English melcan, from Proto-Germanic *melkaną, ultimately from Proto-Indo-European *h₂melǵ-, the same root as the noun

• Let's list out all the cognates for the verb milk that we can find. OED gives us a lot of them, to which I've added what Wiktionary provides for the verb 'milk':
• First, note Wiktionary's suggestion: English 'milk' both verb and noun are from Proto-Germanic *meluks, from Proto-Indo-European *h₂melǵ-
• Old English melcan (the strong verb)
  
• ancient Greek ἀμέλγειν 'to milk'
  
• classical Latin mulgēre 'to milk'
  
• Early Irish bligid ( < mligid ) verb 'milk'; 
• compare also Early Irish melg  noun 'milk'
• Colton suggests: Irish bleacht < Old Irish mlicht < Proto-Celtic *mlixtus
• Tocharian A mālk- verb 'milk'
  
• Balto-Slavic: Russian Church Slavonic mlěsti and Old Russian mlěsti (stem mŭlz- ), Bulgarian mălzja , Lithuanian melžti , milžti , all verbs 'milk'
• compare Serbian mlaz , Croatian mlaz jet, spurt (originally of milk)
• compare Russian molozivo , Czech mlezivo , Slovene mlezivo , all in sense ‘colostrum’
• Note that Wiktionary provides Russian молоко́ (molokó): seems different from the above. I don't know Russian and cannot speak to this right now.
  
• Albanian mjel ,
  
• an original sense ‘to rub’ has been suggested for the Indo-European base on the evidence of Sanskrit mṛj- , Iranian marǰ-, both in sense ‘to rub off, clean’

• Now, turning to Fortson chapter 15, 'Germanic,' we try to find an account of each of the sounds in milk and its cognates in Germanic. We find:
• The switch from PIE to Germanic (the basic features of the branch and how it differs from PIE are always the first section in Fortson's treatment):
• First, phonology (always first in Fortson)
  
• At 15.6, Grimm's Law stage II (top of P. 341) suggests that the voiceless velar k in most of those Germanic languages was a voiced velar g in PIE.
• At 15.13, Fortson reports that resonants stayed intact in the switch from PIE to Germanic, so the m and the l are likely from PIE.
• At 15.15, Fortson reports that *e tended to change to *i, so perhaps the PIE root of milk had an e. 
• Other aspects of the vowels cannot be tracked with Fortson: that is because vowels are often much messier than consonants, and Fortson has to pick and choose what to report to us.
  
• The differentiation of the various daughter languages:
• Nothing seems super helpful here in Fortson, although the -ch endings in some cognate forms seems to be explained by a palatalization of the velars (15.56 reports that velar k changed to a /ch/ sound near i and e : it moved from the velum to the palate and also affricated (came to involve not a stop but a partially obstructed flow of air, some friction)). That might work for the change to Old High German miluh , milih  which then changed to Middle High German milich , milch , German Milch (the velar k changed first to a h and then to ch).
• On the various vowels, nothing super helpful in Fortson.
• What about the cognates? How do they measure up using Fortson's information?
• For ancient Greek ancient Greek ἀμέλγειν (amelgein in our alphabet) 'to milk': Fortson 12.13 reports that
  is a centum language, meaning the plain velar *g and the palatal ǵ merged into plain velar g from PIE to Greek, 12.20 says that the liquids and nasal resonants remained intact, 12.23 reports that h2 changed to a in Greek, and 12.29 reports that Greek preserved PIE vowels more faithfully than any other branch, which accounts for ἀμέλγειν (amelgein) nicely.
  
• For classical Latin mulgēre 'to milk', Fortson 13.6 reports that Latin is a centum language, meaning the plain velar *g and the palatal ǵ merged into plain velar g from PIE to Italic, 13.8 says the resonants also remained unchanged, and 13.10 says that the laryngeal consonants were lost. So *h₂melǵ- > mVlg- is the result, which is what we find in Latin mulgēre.
• For Early Irish bligid < mligid, Fortson reports at 14.4 that Irish is also a centum branch and so *ǵ becomes g, 14.6 says the resonants remained the same, but a syllabic l sometimes became li, which could be relevant here.
• In other words, the forms check out nicely when we look to Fortson.

So why does Fortson say there was no noun for 'milk' in PIE?

• Consider Old Irish mlicht > Proto-Celtic *mlixtus as well as the lack of an ending on the English (Germanic) side.
• So far, we have been ignoring the endings after the root mVlk, but that t and -tus are endings that indicate that this is a noun. Those are particular noun endings, not just any noun ending. And they differ from the endings that we find in the OED for the Old English and other old Germanic forms of 'milk.' The fact that they differ tells us that the endings were probably added after PIE, because if there had been a noun in PIE, it would have had characteristic endings (all nouns do) and they would usually have been handed on in the daugher languages.
• How do we know that the endings would have stayed the same from PIE to daughter languages for nouns? Because in the case of words like PIE *ph₂tēr 'father' and *h₃rḗǵs 'king,' all the daughter languages preserve the same endings as were already there in PIE (Latin pater, Greek patēr, Sanskrit pitṛ, Avestan pitar, Irish athair, German Vater). What is more, both of those words are just like 'milk': they came from verbs. So if the PIE verbal root *h₂melǵ- had a PIE noun associated with it, it would ordinarily have stayed the same sort of noun in the daughter languages. But the two differ in the daughter languages for which we have an attested noun derived from *h₂melǵ- like 'milk,' so the nouns seem likely to be later developments.
  

There's a bit more to the picture, however: there are nouns for 'milk' in Greek γάλα and Latin lact-. Since we are relatively certain that *h₂melǵ- became μέλγω and mulgeō in Greek and Latin, it just seems unlikely that somehow Greek γάλα and Latin lact- are from *h₂melǵ-. If they were, then we would need an explanation for why the verbs and the nouns went such radically different ways, and how they lost their initial m. Normal sound changes don't account for that at all.

• Just for fun: ignore this if you don't like confusing data dumps (I suppose what I did above might be considered an already confusing data dump, but it does get even more confusing, so be warned).
• A deeper dive: WHAT IS FOUND IN THE Oxford English Dictionary
  
• Starting from English milk, the noun, we find in the Oxford English Dictionary, a useful resource for you, available through the Howe Library website, that it had many forms in Old English and has cognates in many Germanic languages:
• Old English forms: OE mealc (rare), OE meluc (rare), OE meoloc, OE miolc (rare), OE mioloc (rare), OE–eME meolc, OE–eME meoluc
• Germanic Cognate forms: "Old Frisian melok (West Frisian molke ), Middle Dutch melc (Dutch melk ), Old Saxon miluk (Middle Low German melk , melik ), Old High German miluh , milih (Middle High German milich , milch , German Milch ), Old Icelandic mjólk , Old Swedish miolk , miölk (Swedish mjölk ), Old Danish mielch , mialk (Danish mælk ), Gothic miluks.
• That's a lot of forms, and it's hard to know what to make of all those vowel differences, but the consonants seem fairly stable.
  
• They also report:
  
• " The synonymous Slavonic forms (Old Church Slavonic mlěko , Russian moloko , Czech mléko , etc.) probably represent early borrowing < Germanic, as they have k instead of z (the expected Slavonic development of Indo-European palatal g ; compare the Slavonic forms with sibilants in the list of Indo-European cognates s.v. milk v.). Similarly, classical Latin melca dish made from sour milk ( > Hellenistic Greek μέλκα , in the same sense), probably also represents an early borrowing < Germanic. These borrowings, if not directly from the Germanic base of Old English melcan , perhaps reflect a form of the Germanic base of the noun before the development of an epenthetic vowel."
• Here's what I notice about that: at the end, they say that there was likely a Germanic base of the noun before a vowel was inserted (that's what epenthetic means), which tells me that the forms in Old English and cognates that have a vowel between l and k are likely to be developments in Germanic, not original to the PIE form.
  
• I also notice that the Balto-Slavic branch and the Italic branch and the Greek branch are held to have borrowed from the Germanic branch! That's interesting, because our evidence for the Italic and Greek branch is usually a lot earlier than the Germanic branch. So folks think there was borrowing between those branches way back then already! That makes thing potentially really messy! But it offers a way for linguists to basically dismiss some things that would be counter-evidence to their claims: here they justify rejecting the Balto-Slavic evidence on the grounds that a sound change law in Balto-Slavic would have had a different result and so claim that the word was borrowed into Balto-Slavic from German after that sound change took effect.
• The OED also reports:
  
• " The word is originally a Germanic feminine athematic consonant stem (compare book n., borough n., goose n., louse n., etc.), which in Old English would be expected to show variation between on the one hand the nominative and accusative singular form meoluc , meoloc (or, with syncopation, meolc ), and on the other the genitive and dative singular form milc (with Germanic mutation of e to i before an i -suffix). However, the word was widely assimilated to other declensions with consequent levelling of forms: in West Saxon the stem meol(u)c , meol(o)c was levelled to all cases, and in Anglian the stem milc ."
• So that explains the vowels a bit more. "Levelling of forms" means that speakers took one pattern as a model and changed a bunch of other forms to match it: in this case, OED claims, the genitive and dative had an i and the other forms were changed to match that. It also means that OED thinks the earlier forms had an eo.
• It also might help you understand what I was talking about above when I said that the endings of the Irish and German words for milk show that they are different kinds of nouns and so don't go back to one PIE noun. A lot more on this to come in the chapter on nouns.
  

• So what about the English verb 'milk'?
• Here's what OED says:
• Forms: 

  α. OE melcan, OE melkan (rare).

  β. OE melcian, OE meolcian, OE mylcian, lOE milcian.

• A merging of two distinct words:
  
• (i) (represented by the α. forms) an Old English strong verb of Class III (melcan ), cognate with Old Frisian melka (West Frisian melke ), Middle Dutch melken (Dutch melken ), Middle Low German melken , Old High German melcan , melchan (Middle High German melken , melchen , German melken ) (in all of these languages the verb is now in the process of becoming weak) < an Indo-European base meaning ‘to milk’,
  
• compare (with various ablaut grades) ancient Greek ἀμέλγειν , classical Latin mulgēre , Early Irish bligid ( < mligid ), Tocharian A mālk- , Russian Church Slavonic mlěsti and Old Russian mlěsti (stem mŭlz- ), Bulgarian mălzja , Lithuanian melžti , milžti , Albanian mjel , all in sense ‘to milk’ (compare also Early Irish melg milk, Serbian mlaz , Croatian mlaz jet, spurt (originally of milk), Russian molozivo , Czech mlezivo , Slovene mlezivo , all in sense ‘colostrum’);
  
• an original sense ‘to rub’ has been suggested for the Indo-European base on the evidence of Sanskrit mṛj- , Iranian marǰ- , both in sense ‘to rub off, clean’;
  
• and
  
• (ii) (represented by the β. forms) an Old English weak verb (meolcian) < milk n.; compare Old Saxon milcon, Old Icelandic mjólka, Swedish mjölka, Danish †melke; compare also ( < the zero-grade of the Germanic base of the strong verb) Old Icelandic molka (also mylkja to suckle, give suck), Old Swedish molka (Swedish regional molka), Old Danish molchæ (Danish malke).
• In Old English the prefixed form of the weak verb, (Northumbrian) gemilciga, is also attested; compare also amelcan, to milk, prefixed form of the strong verb.
• So OED finds clear evidence that there was a strong verb that goes back to PIE and a weak verb that came from the noun we examined above (Old English forms: OE mealc (rare), OE meluc (rare), OE meoloc, OE miolc (rare), OE mioloc (rare), OE–eME meolc, OE–eME meoluc), and that noun came from the strong verb.