Hello, I'm Ryan Sandel, and welcome to our second session on the Gothic phonological system. In this video, we will move beyond segmental phonology and look into ways that Gothic organizes segments into larger prosodic entities. Among these entities, we count prosodic structures such as syllables and feet.

In carrying out this investigation, we will consider how the prosodic prominence, likely stress, was determined in Gothic, and introduce some phonological processes that interact with these prosodic entities. The first question with which we will concern ourselves in this session is how sequences

of segments are grouped into units that may include more than one segment, namely, syllables. The specific questions that should be answered are, what type of segments can serve as the head of a syllable? Namely, does Gothic allow syllables only headed by vowels, or can also certain classes of consonants occur as a syllable nucleus?

Then we can ask ourselves, which segments may or must occur in the onset and coda of syllables? That is, does Gothic require onsets for big codas or somehow restrict their content? Then comes the question, how many segments can occur in onsets and codas? Are complex codas and onsets permitted, or are they restricted in any way?

And must segments in syllable onsets and codas follow specific rules as to their order? Or, does Gothic allow syllable violations of the sonority sequencing principle, or strictly adhere to it? And where do syllable boundaries occur in the middle of the word? Then does Gothic like to prefer complex onsets over the presence of a coda?

And are there certain groups of consonants allowed only at the beginning of a word, for instance? Finally, should certain consonant clusters at word edges be potentially analyzed as containing extra syllabic segments? Before we can try to evaluate how groups of segments were structured in Gothic,

we should briefly consider the question of syllabic weight and length contrasts. Based on the comparative evidence from other older Indo-European languages such as Sanskrit and Greek, as well as other Germanic languages such as Old High German and Old English, the following prosodic structures would be entirely expected for Gothic.

First, a contrast between long and short vowels, and second, a contrast between heavy and light syllables. Evidence from the behavior of certain morphological categories appears to support just such distinctions in Gothic. In particular, the difference between the formation of the past tense that characterizes class 6 and class 7 strong verbs

is most easily understood as a difference between roots containing a short a vowel followed by just a single consonant in class 6, versus roots containing a long vowel or a short vowel followed by two consonants in class 7.

Thus, we can say that class 6 consists of prosotically light roots, while class 7 consists of prosotically heavy roots, as illustrated through the following examples. Similarly, Ya stem nouns exhibit different inflectional forms, again depending upon whether the root to which the stem is built is monomeraic or bimoraic.

In this case, we see that roots containing two short vowels behave like roots with one heavy root syllable. That's the case for the word for counselor in its genitive singular form. We will return to the phonological behavior of these nouns under the heading of Sievers' law later in this session.

These different morphological patterns thus lead us to adopt two generalizations about Gothic phonology, which were in part presented without justification in the last session. Gothic vowels may be long or short, and consonants following a vowel, presumably in the coda of a syllable,

may contribute to the weight of that syllable. We've just seen that an analysis of differences in the morphology of certain forms in Gothic depends upon the assumption of certain properties about prosodic features of segments in the language. Similar arguments can be adduced to help make determinations about the structure and composition of Gothic syllables.

But on this point, we are not wholly dependent upon linguistic analysis alone, but evidence from spelling practices in Gothic manuscripts may, in fact, give some indications of the intuitions of native speakers concerning syllabic structure.

Specifically, examining the preferred behaviors in the division of words at line ends may indicate where divisions between syllables typically fell. The following generalizations fall out from studies on the manuscripts. First, sequences of two consonants between vowels typically place a line break between the consonants.

More specifically, whenever such a sequence consists of a sibilant followed by a stop, the line break always occurs between the sibilant and the stop. In groups of three consonants word internally, the rightmost consonant is a liquid,

and the break typically falls after the leftmost consonant. But in other tri-consonantal clusters, especially if the leftmost consonant is a nasal and the rightmost is not a liquid, then a break before the rightmost consonant is typical. As regards groups of two consonants in which the second is a liquid, a generalization is difficult to establish.

However, just in case a long vowel precedes the consonant plus liquid sequence, a break typically occurs after the vowel. While these patterns of word division very well may indicate syllable divisions,

it should be kept in mind that morphological structure may have influenced scribes' choices regarding where to place a line break as well. When we turn to the evidence from phonological alternations internal to the language, we encounter evidence concerning both the preferred division of syllables and the content of syllable edges.

First, class VII strong verbs once again present useful evidence. For roots beginning with S plus stop clusters, such as the root stald, both the S and the stop appear in the reduplicating syllable of the past tense stem.

But when a root begins with an S or stop plus a liquid, then only the first consonant appears in the reduplicant, as with the root sleip. The copy of both S and the stop into the reduplicant in forms like gastei stald, seen here,

may plausibly be attributed to a dis-preference for syllables containing two identical consonants and no other consonants. But if building a complex syllable onset of S plus stop is worse than copying more consonants, then a syllable built with the root initial S in the syllable coda can be built.

With roots having a liquid after an S or stop, a complex onset with an obstruent plus a liquid may be assumed then to be acceptable. Certain consonants arranged in certain orders likewise appear to be unacceptable at the end of a syllable.

I analyze ja and jostem nouns as containing a stem formant having an underlying glide, which surfaces in the nominative singular of these classes as a vowel e, or rather the accusative singular, I should have said. Rather than build a syllable with a coda consisting of a stop followed by a glide,

the glide becomes syllabic. The jostem noun mawi, girl, also shows a helpful example of the syllabic interaction between two glides. In the nominative, the word final glide becomes syllabic,

but before a vocalic inflectional suffix, as in the genitive singular form, the sequence of a vowel a plus a glide w becomes a monophonic o, as we see in the first syllable there. This would speak against a complex onset with two glides.

Conversely, the class one weak verbs, which can be analyzed as having an underlying vowel e as their derivational formant, their vowel hiatus is systematically avoided by glide formation or vowel contraction. Before vowels other than e, the derivational suffix e in this class becomes a glide on the surface.

On vowel final roots, the underlying e becomes a glide before any vowel, but otherwise before another e it contracts to produce a long vowel, as we see in the second singular present form.

Apparently onsetless syllables are not entirely excluded in Gothic, however, as the following examples show us. Orthographically, we find sequences of multiple vowels. Moreover, the process of hiatus lowering, which was discussed in the previous session,

precisely depends upon the assumption of directly adjacent vowels. Let's recall the example of the infinitive of the verb meaning to sew. Whether unwritten non-phenemic segments such as a glottal stop might have possibly been inserted to fill hiatus is simply unknowable.

The external evidence and phonological evidence mustered on the preceding slides leads to the following generalizations about syllable structure in Gothic. First, any syllable will be built around any underlying mid or low vowel. Underlying high vowels can, however, be potentially subject to glide formation.

As in virtually all other older Indo-European languages, and as would be typologically expected, a single consonant between two syllable heads will be treated as an onset. More interesting is the fact that sequences of two intervocalic consonants would appear to be parsed into two separate syllables.

This applies not only to sequences of falling sonority, but also sequences of rising sonority. Occasionally, complex onsets of an obstruent plus a non-nasal sonorant may have been built if the preceding vowel was long.

Consonant sonority also plays a role when sequences of more than two consonants arise word internally. In such cases, the manuscript evidence suggests that either a complex onset or a complex coda was allowed so as to avoid either falling sonority onsets. Thus, a word like bänsta, barn,

probably built a complex coda in the first syllable to avoid a rising sonority onset. To avoid a falling sonority onset, I meant. Otherwise, most clear violations of sonority sequencing are limited to sequences involving an S at the beginning or end of a word.

One might potentially posit the extrametricality of an S in such cases. To summarize, the maximal gothic syllable could have had up to two onset segments and two coda segments. If the syllable contains a long vowel, coda consonants might have been avoided, however, as we've seen earlier.

One substantive issue with respect to syllabification that remains, however, is the question of whether gothic may have permitted sonorants, that is, nasals and glides, to function as syllable heads. As a consequence of vowel syncope that occurred between proto-Germanic and gothic,

gothic possesses a number of lexemes that, in certain inflectional forms, show a sequence of a stop or a fricative followed by a sonorant at the right edge of the word. Given that gothic syllable structure seems to have been sensitive to considerations of consonant sonority, it is entirely conceivable that in such cases,

the sonorant served as a syllable nucleus, effectively replacing the vowel which had earlier undergone syncope. Whether the sonorants in word-final position in such cases were truly syllabic cannot be definitively proven or disproven.

Arguments one way or another essentially depend upon the development of similar cases in West and North Germanic, which exhibit different behaviors. In West Germanic, we find epenthetic vowels, but in North Germanic, the sonorants are not syllabic.

And so in those cases in North Germanic, like the word for bird in Old Norse, there are sonority sequencing violations that are permitted. In gothic, the only readily available evidence that potentially bears on this question is the vocalization of the glide j in word-final position. Such behavior might speak in favor of syllabic sonorants.

On the other hand, we have words like warst, work, seen here, where we precisely do not see vocalization of an underlying glide w to u. With a general sense of how segments typically combine themselves into syllables in gothic,

we can now turn our attention to higher-level prosodic structures, in particular, the foot and the prosodic word. As a way of approaching the discussion on the phonological construction of feet and prosodic words, I would like first to have a look at the position of prosodic prominence on syllables, or stress.

Unlike for Vedic Sanskrit, for instance, gothic offers no direct indication in its writing system of stress of any kind, nor of any other suprasegmental phonological information. The evidence concerning stress delivered from the later Germanic languages,

which the gothic material does not directly contradict, is that a primary stress in proto-Germanic typically fell on the leftmost, or stem initial, syllable of a word, as we see in the derivation of a proto-Indo-European form, the word for father, down into gothic.

Synchronically in gothic, the processes of vowel deletion might be taken to give evidence of differences in the position of stress. As an example, consider nominal and verbal forms that contain the prefix anda. In the verbal forms on the right of each pair, the prefix takes the monosyllabic form

with an apparent syncope of the second a, that is, we have simply and in the verbal forms andnemon and anhathjan. In the nominal forms, the prefix instead has the form anda. As with prefixed non-separable verbs in modern German,

primary stress fell on the verbal root, then we can posit a process of vowel deletion in an open syllable immediately preceding a primary stress. One way of accounting for the stress differences between nominal and verbal forms is to presume that, in the case of prefixed verbal forms,

a prosodic boundary is maintained between the prefix and the verbal root when primary stress gets assigned. As a result, the prefix and verbal root are initially treated as independent prosodic words and are both assigned a primary stress, which then leads to syncope of the unstressed a vowel

of the prefix anda. The first general rule concerning gothic stress is then that the first vowel following a word boundary will receive the primary stress. Some further possible evidence concerning the position of stress in gothic may be derived from some apparent exceptions

to the process of lowering that applied to short high vowels E and U, which was discussed in the previous section. In the first three exceptions listed, where we find a short E or U in a context where lowering to E and O would be expected,

the exceptional high vowel comes immediately after the first syllable, which plausibly bears primary stress. That's the case, for instance, of the U in Fidur or the I in Thargis. Given the philological evidence for possible vowel centralization in gothic

adduced by Rauch in 1981, it may be the case that stressless short vowels were centralized and therefore not subject to lowering. On the other hand, the two prefixes fair and far exhibit no sort of alternation between forms with and without lowering, depending upon whether they occur

in verbal or nominal forms. This means that in verbal forms where the primary stress probably fell on the verbal root, the prefix probably bore a secondary stress. If this interpretation is correct, then a stress clash between a primary and a secondary stress would not be illicit in gothic.

To account for the occurrence of secondary stress on pre-verbs, I will posit that a primary stress to the left of a word boundary, regardless of whether any other syllables intervene between this primary stress and the word boundary to its right, becomes a secondary stress. This is expressed in the rule of pre-verbal stress reduction.

In this fashion, pre-verbs will be assigned a primary stress, which will subsequently be transformed into secondary stress because the boundary between the prefix plus clitic sequence persists. Such persistence is shown by the assimilation of the consonant in the clitic ugh and

to the following consonants, as we see in following examples, like bithangitanda, and we don't see assimilation of this clitic ugh to the first consonant of the verbal root. That is, we find forms like

gahmelida, and I wrote not gammelida with assimilation. The evidence from vowel deletion and vowel lowering leads us to conclude that Gothic distinguished three phonologically relevant degrees of stress, primary and secondary stress, alongside unstressed syllables.

The principle stress rule in Gothic, in accord with other Germanic languages, then seems to have been that primary stress was assigned to the leftmost syllable of a morphological word. When a semantic word was constructed with only one free morpheme as a base, the pattern of a leftmost stress and no other primary stress results naturally. When multiple free morphemes occur,

as in compounds, the situation is more complicated. Let's look here at the compound feeder dochs. Under the presumption that stress was cumulative, namely that only one primary stress could occur per word, prosodic words with multiple primary stresses are subject to processes of stress reduction,

whereby primary stresses become secondary stresses. We've seen this applied to morphological sequences of a preverb and verb on the preceding slide. In cases of nominal compounds or nouns with a prefix, any primary stress to the right of an initial syllable will get reduced to a secondary stress.

From the case of forms like feeder, for, we can conclude that the second in a sequence of two light syllables was stressless. But how did stress apply in sequences of multiple heavy syllables? Was every heavy syllable assigned stress? Were adjacent secondary stresses deleted?

Or could a sequence of secondary stresses potentially accumulate? To take the class three weak verb salvo vedina as an example, we find a sequence of four heavy syllables. Did stress here alternate after the first syllable with a secondary stress only on the third syllable? Or did a secondary stress fall on

every heavy syllable after the first? These are the two possibilities with which we are presented. In order to hypothesize about this remaining issue, we must consider how syllables were grouped into smaller units, into units smaller than the word. That is, into feet.

Cross-linguistically, stress patterns, when not determined by purely lexical information or the morphological structure of a word, tend to depend upon groupings of syllables into larger units and the distance of those syllables from word edges. Stress also tends to organize itself rhythmically into either trochaic or iambic patterns and may be either dependent upon

or independent of syllable weight. That gothic tendentially places stress at the very beginning of a sequence of syllables and probably leaves immediately following light syllables at least as unstressed strongly indicates that gothic

feet and in turn stress followed a trochaic pattern. That is, there was a strong member on the left part of the foot. Less immediately clear, however, is whether feet simply followed binary groupings of syllables or whether they cared about syllable weight.

Already the evidence from syllabification that a consonant was perhaps avoided after a long vowel points to the relevance of syllable weight in phonological computation in gothic. For the word andaafs answer this would point towards a foot being constructed around the heavy syllable at the left edge of the word and around the third

syllable which is also heavy. Thus we see the following two feet being built. To say with greater certainty how syllables organized themselves into feet, a closer look at further phonological phenomena is necessary. In particular we will look at the alternations between a high glide ya and a vowel i

preceding inflectional endings beginning with the vowel i. These alternations are traditionally labeled Seaver's law. The classic examples of the generative singular form of ya stem nouns we have met on an earlier slide. Here we can see that a root containing a short

vowel followed by a single consonant shows a glide i as in hargis while a root with a short vowel followed by two consonants becomes vocalic as in hardis, as in herdis excuse me. The same applies

to roots with two short vowels as in raginis with a long i. In the following the analysis of these alternations which has been put forward by Kiparski will see that the pattern is best understood when we assume that gothic preferred to build feet consisting of precisely two moras, that

is moraic trochis. The further elements of Kiparski's analysis consist of the following constraints. First that onsets consisting of a consonant plus a glide ya are avoided and that syllables without an onset consonant are avoided.

Meanwhile these preferences cannot be satisfied by inserting or deleting any segments not present in the underlying form. For the genitive singular of army the form hargis is the best option because it builds a bimoraic foot at the left edge of the word.

Three other plausible alternatives result in either a consonant plus yod onset a syllable without an onset consonant as in ha ri is or a monomoraic foot at the left edge of the word, that is

the syllable ha parsed into a single foot on its own. For the genitive singular of shepherd the preferred form hargis is ideal hargis, excuse me. It builds two bimoraic feet. The alternatives again

contain either a consonant plus yod onset in onsetless syllable or if the coronal stop the is syllabified into the following syllable a super heavy foot containing three morai since each coda consonant contributes to syllabic weight. Precisely this avoidance of a super heavy syllable

in the form hargis can be used to clarify the difference between the genitive hargis versus herdis. Attention to foot structure and not merely the number of segments at the right

edge of a root explains why disyllabic roots obtain a form with a long vowel e rather than a glide in the last syllable. Specifically if the n in the root ragin is parsed into the coda of a syllable built around a short e the result will be either a trimariac foot resulting

from the light first and heavy second syllable or an initial syllable that fails to be parsed into a foot. The best option to build two perfectly bimariac feet is to create a long vowel in the final syllable preceded by two light syllables. Similarly

disyllabic stems with a long vowel in the second syllable like the genitive singular of siponi, disciple will have a long vowel e in the final syllable as we see here and that likewise occurs in order to avoid a super heavy trimariac second syllable where an n

would be parsed into a syllable already having a long vowel. In a subsequent session on verbal morphology we will meet similar alternations of vowels and glides conditioned by the same factors in the inflection of class one weak verbs. The evidence from Sievers law alternations thus demonstrates that

gothic strove to organize syllables into bimariac feet, more specifically into moraiic trochees. This conclusion then leads to the prediction that in the form salbodeidina mentioned above, every heavy syllable constituted a foot as well and thus in principle could have borne a stress. Given that preverb

plus verbal root sequences suggests that a secondary stress did not produce a stress clash with primary stress I tend then to think that the strong part of every trochaic foot following the second foot bore a secondary stress but this is not entirely certain. Thus the output would be so

where we have feet built around every heavy syllable and after the first syllable a secondary stress assigned to each foot. To wrap up our look at foot structure and stress we should ask one more question. If stress depended upon foot structure, how can a primary stress be guaranteed at the left

edge of a morphological word? In particular, stems with sequences of a necessarily light followed by a necessarily heavy syllable would have a left would have a left syllable a light syllable, excuse me left at the beginning of a word on footed and thus we would have a syllable that in principle

would be unstressable which cannot have been the case. We will therefore adopt the position found in Goering 2016 and in older literature as well that the initial syllable in a morphological word constituted the head of a foot regardless of its weight. In other words the preference to align the left

most foot of a word with the left edge of the word outranked the preference for all feet to be bimeraic and thus an underlying sequence light heavy at the left edge of a word would build two feet and a primary stress would occur on that left most light syllable.

Our last topic for this session will be the treatment of clitic lexemes from the phonological point of view clitic lexemes stand out in that they cannot build independent prosodic words and hence are incorporated into the prosodic domain of linearly adjacent words that can independently project a

prosodic word. There are two immediate problems with identifying any such prosotically deficient lexemes in Gothic. First is that since stress is not directly indicated one cannot easily say whether a given lexeme is stressed on its own or builds a stress domain with another lexeme. Second since word breaks are not given

in Gothic manuscripts to say whether two adjacent lexemes belong to the same phonological word or phonological phrase is difficult. However in the case of at least two common clitics, the question particle u and the conjunction u and the application

of certain phonological processes that are otherwise unknown across word boundaries gives evidence that these two lexemes are endclitic. First when either u or u stand after a lexeme that ends in a vowel either the last vowel of the host lexeme may be deleted

as in fatu or at least with the conjunction u the vowel u may be deleted as we see in the form nich. Even clearer evidence for incorporation into the same word is the behavior of fricatives at the right edge which typically

become voiced between u and u. This shows that these fricatives cannot be at the right edge of a word where the process of fricative devoicing would necessarily apply. Having then established that u and u are prosodically deficient

clitics we find further evidence for their clitic behavior in that they may intervene between a preverb and a verb where sometimes some of the same segmental effects already mentioned apply. Thus we find forms such as gahmelida and he wrote with the conjunctive endclitic after the prefix gah showing vowel deletion as well

as uzuhida and I came forth. This feature allows yet further prosodically deficient clitics to be identified. Other lexemes that intervene between a verbal prefix and the verbal root must also be prosodically deficient. Thus in the sentence

fra ina go hwe sehua the indefinite pronoun hwa can be identified as a clitic as well following both the prefix ga and the question particle u. As the examples here show

Gothic endclitics thus either attach to the end of a word or after a verbal prefix. In this session we have explored the construction of prosodic structure in Gothic examining syllabification, stress, foot structure and clitics.

We thank you for your attention. In the next session we will look at some morphophonological processes including reduplication.