Welcome to the third session about phonology in our module about the Old Irish language. In this session, I will first speak briefly about aspects of stress in Old Irish, and I will then turn to a couple of morphophonemic rules of the language that are really crucial for understanding its synchronic operation. Accent in Old Irish is by stress. It is a lexical stress,

non-contrastive stress, and it cannot be used to award prominence to a phrase. It is fundamentally fixed on the first syllable of a word. That this initial stress was also the case in the prehistory of the Irish language follows from the massive reduction and deletion

processes that affected word forms in the period before the emergence of Old Irish as we know it. That is to say, when we compare how words are realized in Old Irish with what their protocaltic preforms looked like. For example, a word that would have been quenowindo, or quenowindos, white-headed in protocaltic,

appears as kengan in Old Irish, or the verbal form ambisoomos, which means something like we turn around, is found actually as impuff, so very strongly reduced. There are systematic exceptions to the rule of stress placement on the first syllable, but there are exceptions only on the surface.

Adverbs that originate from the merger of the article of prepositions and the nominal form bear the stress on the first syllable of the latter part. So, for instance, in due, today, which is put together from in the article plus the word for day, or

anous from above, which consists of an directional prefix plus a word for above. More importantly, verbal forms that have any element before the root syllable, and any element can be conjunctions, verbal particles, lexical preverbs, they bear the stress on the second element of the entire, well, what we call the verbal complex,

whether this be the root or a preverb. So, if you take the examples here, first we have the verb berith carries, but as soon as we put the negative ni in front of it, the verb is reduced to ni ber. If we put a lexical preverb do in front of it,

we get do ber, and if we put an, again, the negative before do ber, we get ni tavith. So the verb berith is ultimately reduced to plain ver. This will become even more important when we talk about the verbal complex and the verbal morphology in one of the later,

one of the later sessions. In modern, normalized orthography, the position after which the verbal stress falls is indicated by a raised dot or by similar means. Now, take this example from the Milan glasses. The scribe wrote it at the very margin of the manuscript.

He started writing the ass, but then it ran out of the space, because this is really the very end of the manuscript here, of the page. Then he continued with nakha above, but ran out of space again, and then he had to jump the two lines below to finish the verbal form, and we get asnakha tukketh.

Asnakha tukketh is one word, a verbal form that means out of which you would not have been able to bring them. In old Irish, this is expressed by a single word, whereas in English, we need 12 words to express the same thing.

The stress is on the two of the tukketh I hear. The element before this, which actually consists of several morphemes, has no accent, so you could say, oh, well, tukketh isn't actually the second element. Well, strictly speaking, you're correct, but we're doing old Irish

linguistics here, not mathematics. Function words, namely articles, prepositions, conjunctions, and various types of pronouns and pronominals, are for the greatest part unstressed. They can be regarded as proclitic and enclitic to stressed words. Reflecting this, early Irish scribes had the tendency to write unstressed elements without separation from

adjoining stressed words, as you can see in some of the examples cited on this slide. So, if we look at the first example for the article, in Dintruth, the article is actually the NT. Here, and this has been merged with the preposition G, and it's also written together with the noun Sruth.

Modern editors do not follow this practice. In parenthesis, therefore, I show you how modern editors would actually write those phrases. We can take the same example for the example for the prepositions. Here, it is the G which has been merged with the article and the word.

Or, if we look at conjunctions, the first example is kvælinche, so that I might die. The k, the thing with the underscore, is a conjunction meaning so that, but again, it is merged in writing with the

with the verb. So, this shows that the conjunction has been merged in writing with the verb. Here, we have an example for pronouns, and the first example actually contains a number of pronouns.

Fondum Tawartische, contains an infixed pronoun here, and a suffixed nota augens, as we call it at the end, but again, the whole thing is written as one. It is noteworthy that the verb to be here, in its use as a copula, is also always unstressed.

This means that it also behaves like a function word, even though at the same time, it is also a strongly inflected word. There's a sharp phonological contrast between stressed and unstressed syllables. Whereas all phonemic oppositions can be realized in stressed syllables, a lot of neutralization goes on the further away we move from the stress.

The vocalism of unstressed syllables suffers many restrictions. We saw that already in the section on phonology. Consonants in pretonic words like the copula, prepositions, particles, etc., were de-palatalized in early Orage, and the opposition between voiced and voiceless fricatives is largely neutralized in unstressed syllables, and voiceless stops become voiced.

The different treatment of syllables in stressed and unstressed position is also one of the reasons for the complexity of the verbal system that we'll see in a later session. Stress on the first element of compounds normally entails phonetic reduction in the second element.

So this is shown, for example, nicely by the word gashketh, weapons, which is a compound of, on the one hand, gai, which means spare, and schieth, shield. Synchronically, this is no longer transparent as a dvandva compound, or you can also see similar things happening in anthroponyms and

personal names that have lost their connection with a common lexicon. Old Irish is characterized by a series of synchronic phonemic variations that are either concomitant with other phonomorphological processes, or that function as grammatical processes in their own right.

I lumped them here together under the heading morphophonemic variation, and I will briefly look at the most important ones. Several of the major sound changes that affected primitive Irish in prehistory acquired synchronic grammatical functions and put a stamp on the Irish which it retains to the present day.

Well, to all extents and purposes, must have been a language of fairly average Indo-European typology at the beginning of the first millennium AD was converted into a very different system, which is non-Indo-European, if I may say so, by outward appearance, in which modifications of initial and final consonants, as well as internal syllables, play a key morphological role.

Some of these sound rules have a wider structural implications. So, for instance, lenition and nasalization, which together form the two main pillars of an initial mutations, or palatalization. These processes have repercussions beyond phonology to the extent that what started out as allophonic variations in consonantial quality

acquired morphophonemic status when the conditioning factors disappeared with the loss of final syllables. Other changes like syncope or metophony, i.e., umlaut, have a rather restricted capacity of making morphological distinctions.

But they are still, nevertheless, all-pervading in the grammatical system of all Irish. The numerous diachronic and synchronic sound rules of Irish and concomitant phonetic processes cannot really be commented upon in detail in this survey, so I'll be rather brief about them. They cause a considerable amount of allomorphy,

which in turn is a source for morphemic splits and reanalysis. One of the most notorious features of all insular Celtic languages is the pervasive role of phonemic consonant mutations, i.e., of

variations in word initial position, which then carry morphological and syntactic information. All modern Celtic languages developed them, and their roots go back to the first half of the first millennium AD. They originate from prehistoric external sandy phenomena, which only had allophonic status until the massive loss of final syllables in all insular Celtic languages.

But nowhere in none of the insular Celtic languages are the mutations so fundamentally interwoven with all aspects of grammar as it is in Old Irish. Three types of mutations can be distinguished in Old Irish. We call them linition, nasalization, and aspiration.

When we talk about them in theoretical parlance, we indicate them by superscript L, N, and H, respectively. Only linition and nasalization find graphic expression in Old Irish at all, and they only partially.

While aspiration remains entirely unexpressed in writing. So how do we know in the first place that this mutation was there? Well, it can be inferred from Middle Irish and modern Irish orthographic practices. Aspiration is also much more limited in effect than the other two mutations, in that it prefixes

H or H to word initial vowels after a very small number of functional elements. Linition affects only consonants, and it turns a consonant into its linited counterpart. I have already talked about this in the session about phonology, so I won't repeat it here. Nasalization has effects on a smaller number of consonants, but it also affects vowels.

Nasalization of vowels is realized by simply prefixing N. Nasalization of consonants is a kind of synchronic misnomer, since only in the case of voice stops, a homo-organic nasal is prefixed.

So P becomes M-B, D becomes N-D, and G becomes N-G. In the case of voiceless stops and F, what actually happens is voicing. So P becomes B, T becomes D, K becomes G, F becomes V. Unfortunately, you can't actually see it in writing. You just have to know it.

Liquids and nasals are not affected by nasalization, but sometimes they are doubled in spelling. But this doubling must not be misunderstood as phonemic gemination. Mutations operate across word boundaries. They are triggered by the preceding words in lexical concatenations,

but they are usually restricted to within the phrase. So, for instance, in NPs and PPs, an overt element mutates the following element. Mutations inside the verbal complex are more complicated because the trigger may not always be overt. So if we take a look at the examples here, the first example is domarche,

overlooks me. The elation on the F is affected by the M, which indicates the me in this case, okay? Or the next example, da mithg pelted him or threw stones at him. The nasalization on the B is caused by the A, which goes before.

However, the other two examples show more complexity here. So we have Dober brings, but we have Dommer, what it brings. The nasalization on the B of Dommer is actually not caused by the D.

It's caused by an element which you can't see. It's hidden behind the D. The same is true for the next example here, voklad means digs, but voklad, who digs, again, the lenition is caused by an element which is virtually there, but which one cannot see. There are no simple rules for which elements trigger which mutations.

Instead, the mutations are governed by a large and diverse number of lexical, morphological and syntactical rules. The role of palatalization and the places where it applied were constantly expanding in the prehistory of Old Irish.

Particularly important are the traces that the apocobi or the loss of inherited inflectional endings left on the rest of the word, depending on whether the lost syllable had contained a back or a front vowel. While in the former case, the remaining consonant retained its neutral quality, in the latter case, the consonant acquired a distinct palatal, palatalized colour.

So let's have a look at the example here. In Primitive Irish, the difference between carach, cart and genitive cadi of a cart consisted simply in the ending. But when the endings were lost, this was transformed into a functionally loaded difference of quality,

namely Old Irish nominative singular car and genitive cas. In this manner, palatalization was established as a more phonemically relevant process. In some cases, difference in quality is also concomitant with overworked morphemes. So take the example knav, nominative singular, but genitive knava,

the loss of palatalization goes hand in hand with the adding of an ending. Or beredith, he or she carries with a palatalized R, versus beredith, they carry with a non-palatalized R. So again, going hand in hand with the ending.

Palatalization has thus acquired high phonological prominence as a morphological marker in Old Irish, and it has been spreading ever since to positions where it has no etymological or morphological justification. Another morphophonemic alternation that often, and in rather complicated ways, accompanies variation in palatalization is umlaut, or the English term would be metophony.

Metophony refers to changes of vowels, predominantly of stressed, short vowels. One of the fundamental aspects of Old Irish metophony is the alternation of short e and o with i and u. We call this alternation raising, and we have the exact opposite,

where i and u turn into e and o, and this is of course then called lowering. Such alternations are frequently concomitant to alternations in consonant quality. So the example, the first example here, fyr, a man, singular, but fys, man.

Not only has the R become palatalized, but also the vowel has been raised. Or the second example, fy, blood, so normally singular, but if we form the genitive, we get follow. So the u has been lowered to o.

Another frequent morphophonemic process is the insertion of u or w, which we call u-infection. So don't be afraid, it doesn't make you sick when you get it. This happens after another short vowel, for instance after an R. The example that I have here is canu, I sing, but if I negate it, I get nikaun, I don't sing. So an u has been inserted after the R.

Other methphonic alternations are much more restricted to the fact that sometimes they look like lexical properties, and I won't discuss them here. The triggers for these alternations are diverse morphological categories, which unfortunately elude a simple systematic description.

Another diachronic change that gained new life as a central synchronic rule is that of syncope. As a historical process, syncope meant that after the loss of inherited final syllables, the vowel of every second non-final syllable was deleted.

This rule operated really mechanically, and it failed to apply only rarely, for instance when the resulting cluster would have been too awkward to pronounce. In synchronic terms, this means that when an extra syllable is suffixed to a form, or when in verbal morphology a grammaticalist element is added before or even inside a form,

a new syllable count has to be made for the entire new form, and if it is found to have too many syllables, the vowels of all eligible syllables have to be elided. So let's take a look at some examples. We take the word cladev, sword.

When we form the accusative plural, which fundamentally consists in adding and ending u, we would expect cladev u, but what we get is clavvio, a shortened form. Or carim, I love. The first plural, synchronically, is carmi, which actually goes back to an earlier carami,

where again the second syllable has been lost. Or when we want to form new adjectives from a noun, so we take that noun diral, which means revenge. Adjectives are very often formed by adding ach, so we would expect diralach,

but no, what we get is dirach. The matter is rendered more complicated by the fact that the rule applies to the diachronically underlying forms, not to the synchronic surface representations. For example, the plural of the word ingen, nail or claw, is inga, as we would expect it with syncope.

But why is the plural of the homograph inren, here at the bottom, which looks exactly the same as the word for nail or claw, why is the plural of that ingena? You can ask the question, where is the syncope?

Well, first of all, the two words are only written with the same letters, but the sounds are quite different. Secondly, ingen, girl, goes back to an earlier enigena. In fact, syncope has already taken place regularly on the underlying form of ingen. So the old Irish plural ingena regularly continues the equally syncopated form enigenas.

Remember, you only syncopate once. Syncope is often dovetailed by several other rules and changes, the most important of which are palatalization and its counterpart depalatalization,

and diverse assimilation processes. One of them is homorganic delineation. This means that lenition, that is there in the word form, is automatically given up on the involved segments when two consonants that share the same place of articulation in the mouth, when they come into contact with each other.

So, let's have a look. For instance, the verbal stem ben means to strike, to hit, and the ending of the second plural is the. But because n and th are pronounced at the same place in your mouth, at the teeth, the outcome is not, as you would expect, bente, but bente.

Both the n and the t are delineated. Or, another nice example, we have the adjective trogh, which means miserable, ending in a r, and we have a compound element khar, which means loving, starting with a h. Now, of course, r and h are obviously homo-organic sounds.

If the two come together, we don't get trogh-har, or whatever we might expect, but what actually comes out is trogh-kar, charitable, with a k. While syncope overwhelmingly leads to a reduction of the number of syllables, in some instances, it actually creates clusters that were too difficult to pronounce,

even for speakers of old Irish. And believe me, speakers of old Irish could take up with pretty complex clusters. In such cases, what happens is appendices, i.e. the insertion of vowels, or insertion of support vowels. Take this example here, athglaðr, means converses,

he or she converses, talks to somebody. Again, when we put ní in front of it, syncope should actually produce something like, what you can see here at the very right, agldr. Now, that was too much even for speakers of old Irish, so a support vowel was inserted here before the l,

and what we actually get is ní agaldrðr. Syncope is an all-pervading phenomenon in the grammatical system of old Irish, operating in inflection and derivation alike. Throughout the history of the medieval Irish language, its rules were surprisingly faithfully adhered to,

despite the extremely opaque olamorphy that it produced. All of the mentioned processes interact with each other in complex ways, and their interaction can create quite drastic olamorphy in inflection, especially in verbal inflection.

For example, imsoat and impat, in ní impat, both reflect the same underlying form, ambisoant, they turn. But in the second form, the negative particle ní has been added in front, as you can see. But this simple addition of a particle in front

causes the accent to shift one position to the left, and this in consequence entails concomitant lenition, syncope, devoicing, and vowel reduction. Or compare böslundi, denies, the next example, with its negated counterpart, nijelti, does not deny.

Both reflect the same diachronically underlying structure, d'slomdith. The relationship between d'slundi and nijelti is absolutely irregular, even though it may not look like it. The variation is again triggered by the different stress pattern, and all associated changes.

With this overview of the most important morphophonemic phenomena, we have finished a section about Old Irish phonology, and we will turn to morphology in the next series. At lochor indit if dóilf. I thank you for your attention.