Module talk:ang-pron

Initial discussion

@Erutuon New module that is mostly complete. It still needs work but Module:ang-pronunciation can probably be deleted. This module has a list of prefixes and stress rules for the prefixes, and various other things. It has a pos= parameter, and one of the things I haven't fixed yet is what happens if pos= is omitted. Benwing2 (talk) 02:46, 20 November 2019 (UTC)

@Benwing2:Thanks for your work on this! I'm interested in the topic of Old English pronunciation and I wanted to add some words to the test cases for this new module. However, that page doesn't seem to have been set up yet, and I was unsuccessful in my attempts to adjust the Module:ang-pronunciation test cases code to work here. Could you let me know when that can be established?--Urszag (talk) 03:30, 20 November 2019 (UTC)

@Urszag I will add a testcases page. Your help will be greatly desired as I'm not perfectly familiar with Old English pronunciation, esp. as regards stress placement. Benwing2 (talk) 03:34, 20 November 2019 (UTC)

Well, the testcases are working for the phonemic transcription at least. Here is a list of all existing transcriptions in {{IPA}}, which might be useful. — Eru·tuon 04:34, 20 November 2019 (UTC)

@Urszag I added phonetic transcription tests as well. Feel free to edit Module:ang-pron/testcases and add new cases. Benwing2 (talk) 04:51, 20 November 2019 (UTC)

Thanks for getting that set up so quickly! I will start out by adding some more fricative test cases.--Urszag (talk) 04:52, 20 November 2019 (UTC)

Phonemic identity of <h> [h~x~ç]

A minor thing that might be changed about the phonemic transcription. Although the transcription /h/ is closest to the spelling, I've seen the phoneme transcribed as /x/ instead in some sources. I think the reason is to indicate that it is interpreted as being underlyingly a velar consonant, with allophonic debuccalization in syllable-initial position. (I've seen a similar analysis applied to the present-day Hungarian phoneme spelled as <h>, which shows a similar pattern of allophony to that hypothesized for the Old English phoneme.) For example, the chapter "Phonology and Morphology", by Richard M. Hogg, says "We can further distinguish a third allophone of /x/, for initially <h> represented the glottal fricative [h], as in hand 'hand', and this is best treated phonemically as /x/" (The Cambridge History of the English Language, Volume 1, p. 92).--Urszag (talk) 19:14, 5 December 2019 (UTC)

Pronunciation of -þu with an unconditionally voiceless fricative

(Notifying Benwing2): Some of the sources that I've read indicate that the fricative in the suffix -þu (also in ordinal -þa) is always voiceless, regardless of which segment precedes it. (I don't know how we know this; I think the the present-day pronunciation of words like "length" and ""strength" is considered relevant evidence.) Here is what is said about the topic in Donka Minkova 2010, "Phonemically Contrastive Fricatives in Old English?": "[...] The fricative in the suffix [-þu or -þa] is unvoiced at a morpheme boundary, for which two plausible explanations exist, cited in Laker (2009: 214): fricative voicing was blocked if the flanking vowels were unstressed, e.g. streng-þ(u) <∗strangi-þu, originating with Luick (1914–40) and supported by Fulk (2001), or -þu was exempt from voicing due to the open nature of the juncture between the base and the suffix – open junctures block the phonotactic processes (Dietz 1997)" (pages 33-34). Based on this, I am changing the phonetic transcription of wyrgþu to [ˈwyrˠɣ.θu].--Urszag (talk) 07:00, 10 December 2019 (UTC)

I found cēnþu listed with [ð], although some IP created the pronunciation. I'll implement your suggestion in the module. Benwing2 (talk) 07:04, 13 December 2019 (UTC)

Fricative voicess between unstressed vowels?

(Notifying Leasnam, Lambiam, Urszag, Hundwine): @Urszag, you mentioned somewhere that fricatives are voiceless between unstressed vowels, as in adesa, ċiefese. I found an apparent counterexample, earfeþe, whose pronunciation is explicitly given with [ð] (but earfoþe is explicitly given with [θ]). The pronunciation for earfeþe was inserted by an IP, maybe it's wrong? Benwing2 (talk) 07:42, 13 December 2019 (UTC)

I cannot find any source that discusses the pronunciation of the specific word earfeþe, and neither its spelling nor the spelling of forms descended from it in later varieties of English seems to indicate the voicing of the fricative one way or another. So I don't think it functions as a genuine counterexample: the pronunciation with [ð] that we currently give looks like someone's hypothetical reconstruction. (And on the other hand, I don't think the pronunciation that we currently give for "earfoþe" functions as genuine positive evidence for a voiceless fricative in this position.) I think earfoþe and earfeþe must descend from PG *arbaiþiz (the form *arbaidiz mentioned in the etymology section of earfoþe would have regularly resulted in an Old English form with /d/, as PG *d regularly corresponded in all positions to the Old English plosive /d/, not to a fricative). Based on the general rule given by Ringe and Taylor 2014, §6.7.2, pages 261-263 (which I mentioned/linked to in this earlier Tea room discussion), I think the þ in this word should have been voiceless θ in the standard reconstructed Old English form. I did not know about this special condition for the use of voiceless fricatives before I started looking for sources about the particular topic of fricative voicing in Old English; it wouldn't surprise me if the editor who gave the transcription with [ð] was aware of the general patterns for the use of voiced and voiceless fricatives, but not aware of this supposed exception. There might be scholars that are aware of the stress-distinguishing rule that Ringe and Taylor give but explicitly disagree with it, but I haven't read any sources yet that present that viewpoint.--Urszag (talk) 08:39, 13 December 2019 (UTC)

The IP made the analogous change at earfoþe, but Hundwine later changed the [ð] to [θ]. Both were created with broad pronunciation /ð/.  --Lambiam 08:45, 13 December 2019 (UTC)

@Urszag, Lambiam Another such case: eġesian given as [ˈejeziɑn]. If the above rule applies correctly, eġesian should be [ˈejesiɑn] but eġsian should be [ˈejziɑn]. Also, what about the suffix -else, currently given as [elze]? Here, the s is between two unstressed syllables but next to an l instead of directly next to two vowels. Benwing2 (talk) 17:30, 13 December 2019 (UTC)

My understanding is that -sian verbs are thought to have had voiceless [s] in general, including the contracted ones (so [s] in both eġesian and eġsian). Some sources explicitly say that fricative voicing occurred in the appropriate environments before contraction, but not in new environments of the same type that were created by contraction (although there is some argument about this). Thus, the t in forms like ‎blētsian is explained as resulting from devoicing of /d/ before the voiceless fricative [s] (Ringe and Taylor 2014 page 263). Ringe and Taylor also say that hālsian (where "ls" results from contraction) therefore had [ls], but healsas had [lz] (same page). But there is one very notorious exception to this account: ‎clǣnsian is thought to have ended in [ziɑn], based I think on the evidence of present-day English "cleanse". By the rule given in Ringe and Taylor, I believe -else should be [else]: a preceding resonant is not a stronger voicing environment than a preceding vowel with regard to Old English. (The s in -els(e) would have been voiced later in Middle English by the same process that voiced plural -s.) Laing and Lass 2019 argues that voicing occurred after contraction, using the single example [klæ:nziɑn] as evidence, but that article doesn't seem like as good a source to me. (One minor thing that bothered me while I was skimming it is that it doesn't seem to take into account that many cases of Old English non-initial /f/ come from spirantization of originally voiced *b, not from voicing of an originally voiceless fricative *f.) So I would support Ringe and Taylor's ordering of the rules. --Urszag (talk) 22:40, 13 December 2019 (UTC)

@Urszag OK. I'll have to implement a mechanism to override the automatic voicing/devoicing of fricatives. If clǣnsian is thought to have had [z] but other verbs in -nsian to have had [s], then pretty much you have to conclude that voiced fricatives were at least marginally phonemic. Likewise, hālsian [ls] vs. healsas [lz]. Given that many people don't seem willing to accept this, Ringe/Taylor's rule may be wrong. Benwing2 (talk) 05:23, 16 December 2019 (UTC)

I feel a little uncertain about the accuracy of Ringe/Taylor's rule, but I think that accepting it doesn't necessarily require treating the voiced/voicelessness distinction as phonemic for fricatives. For comparison, I think very few linguists analyze the use of [t] vs [ɾ] in present-day American English as a phonemic split, even though the conditioning factors in some cases are hard to explain and some speakers may have apparent near-minimal pairs like mili[t]aristic vs. capi[ɾ]alistic. Likewise, the /t/ in "mistake" is unaspirated while the /t/ in "misteach" is aspirated: this is typically analyzed as resulting from different syllable structure ("mi.stake" vs. "mis.teach") or the presence or absence of juncture rather than from the existence of different phonemes for aspirated and unaspirated plosives in English. A difference between [z] in clǣnsian, healsas vs. [s] in hālsian could potentially be attributed to a difference in prosody or phonological structure, like the first prosodic feet being [clǣns]ian, [heals]as vs. :::[hāl]sian.--Urszag (talk) 05:49, 16 December 2019 (UTC)

@Urszag Hmm, I never thought about mili[t]aristic vs. capi[ɾ]alistic but you are absolutely right about this one; it seems to be influenced by military with secondary stress after the t vs. capital with no secondary stress, although that falls outside of what's normally allowed in allophonic rules. BTW, I implemented a way to force a given allophone, using the respelling notation clǣn[z]ian, hāl[s]ian. I haven't yet added a rule to make -sian have [s] in the general case; I'll probably do that tomorrow. Benwing2 (talk) 06:40, 16 December 2019 (UTC)

Thanks for adding that system for specifying [s] or [z]. Looking through the verbs ending in "-sian", I think the rule could be improved by making it default to voiceless [s] only after consonants, not after stressed vowels. Since the voicelessness is supposed to be a result of contraction from suffixed words, it wouldn't have applied in many words where -sian is preceded by a vowel, such as neosian and drusian. I would also make -sian have [z] by default after a vowel with secondary stress, as in ġīemelēasian.--Urszag (talk) 03:32, 21 December 2019 (UTC)

@Benwing2 Oops, I forgot to include a ping in my previous edit--see above.--Urszag (talk) 03:35, 21 December 2019 (UTC)

Syllabic resonants and their distribution

@Benwing2 I'm still researching this topic, but it seems that not all word-final resonants after an obstruent counted as syllabic in the phonology of Old English. Rather, resonants in this position could be either syllabic or non-syllabic, with some variability depending on resonant in question, the structure of the preceding syllable, and based on the dialect/era or perhaps sometimes based on poetic considerations. "Post-consonantal resonants in word-final position", by B. Richard Page (Insights in Germanic Linguistics I: Methodology in Transition, 1995), gives the following summaries of when syllabic resonants can occur: "Fulk (1989), whose aim is to show that the occurrence of nonsyllabic resonants in Old English verse is a reliable dating criterion, examines the metrical evidence for nonsyllabic postconsonantal resonants following both heavy and light syllables. Fulk finds only four scansions in all of Old English verse that indicate a syllabic postconsonantal resonant after a light syllable. In contrast, there are numerous examples of unambiguously syllabic postconsonantal resonants after heavy syllables" (page 237). "After heavy syllables, postconsonantal r and n are sometimes scanned as syllabic and sometimes as nonsyllabic; l is never syllabic after d, t, f, or s but may be syllabic or nonsyllabic after other obstruents; and m is never syllabic regardless of the preceding consonant (although the only consonants which precede m in this environment are þ, t, and s)" (page 239).--Urszag (talk) 07:06, 14 December 2019 (UTC)

@Urszag OK. Can you boil this down to a set of rules? If so I can implement them. I'm surprised though that m in blōstm and fæþm isn't syllabic, esp. given later blossom and fathom. Note that the scheme I've adopted so far displays the syllabic marker e.g. m̩ under resonants that follow vowel + obstruent in a single syllable, but doesn't change the syllable division to include the syllabic resonant in a new syllable. Perhaps something like this happend in Old English? Even in Modern English, a word like rhythm is pronounced with a syllabic m but is often though of as one syllable. Benwing2 (talk) 07:13, 14 December 2019 (UTC)

The rule I would recommend is to not use the syllabic marker in any transcriptions. Page doesn't mention any contexts where a syllabic scansion is obligatory, and it seems that when syllabification did occur, it often led to the eventual insertion of an epenthetic vowel like <o> (yielding a different spelling). Present-day English forms like blossom and fathom attest to eventual syllabification of word-final resonants after obstruents, but they say nothing about when the syllabic pronunciations originated or became widespread. Words like bible and center/centre show that a process of syllabification or vowel insertion was active in post-OE periods; that could account for the current English forms of blossom, fathom, etc. even if Old English had non-syllabic resonants in this position. Our current transcription of blostm is [ˈbloːstm̥]; the voiceless [m̥] seems plausible as a possible phonetic realization of non-syllabic /m/ in this context, but is perhaps overly narrow given that I don't think we know about these kinds of details. I think [ˈbloːstm] is sufficient and satisfactory. Syllabification isn't a very phonetic concept anyway; the distinction between [m] and [m̩] can't be clearly or consistently made by phoneticians purely from listening to a recording or observing a spectrogram without knowledge of the language in question.--Urszag (talk) 09:32, 14 December 2019 (UTC)

Alternatively, if it seems better to make explicit note of the (seemingly somewhat rare) situations when a resonant can be syllabic, I would recommend doing that by showing multiple pronunciations, one with a syllable marker and one without one, in the following circumstances:

1. For word-final r and n when preceded by either any two-consonant sequence, or by any single consonant and a long vowel. Examples: bēacn, tācn, sōcn
2. " For word-final l when preceded by any single consonant other than "dtfs" (or allophones "vz") preceded by any other consonant or a long vowel. Examples: tungl (seems rare compared to tungol, with epenthetic <o>); templ

--Urszag (talk) 09:55, 14 December 2019 (UTC)

@Urszag Thanks for your response. I am fine removing the syllabic markers; it's just a few lines of code. Benwing2 (talk) 05:31, 16 December 2019 (UTC)

12 questions

(Notifying Leasnam, Lambiam, Urszag, Hundwine, Mnemosientje): I've been going through and preparing to replace all manual {{IPA|ang}} pronunciations with {{ang-IPA}}. This has led to a whole host of questions. To avoid overwhelming people, I'll give them in batches. Here's 12 questions:

1. Should /an/, /on/ be pronounced [ɒn]? Same for /am/, /om/.
2. Should final /ɣ/ be rendered as [x]? (It is currently.)
3. Should word-final double consonants be simplified in phonetic representation? Maybe also syllable-final except obstruents before [lr]?
4. Should we use /x/ instead of /h/?
5. Should fricatives be voiced before voiced sound across word boundary? (dæġes ēage [ˈdæːjez ˈæːɑ̯ɣe]?)
6. In bræġn, is the n syllabic? It's given as /ˈbræjn̩/. Similarly, seġl given as /ˈsejl̩/.
7. How is ġeond- pronounced? It's consistently given as [jeond], but English yond- suggests [jond]. Can it be either?
8. Is final -ol pronounced [ul] e.g regol [ˈreɣul]? Hundwine has created entries this way. What about final -oc etc.?
9. Is final -ian pronounced [jan] or [ian]? Cf. sċyldigian given as IPA^(key): /ˈʃyldiɣiɑn/, /ˈʃyldiɣjɑn/. What about spyrian given as /ˈspyr.jɑn/?
10. seht given as /seçt/ but sehtlian given as /ˈsextliɑn/. Which one is correct?
11. There are a ton of entries that list "early Old English" variants with initial [ɣ] and "late Old English" variants with initial [ɡ]. Examples: āgān, angenġa, gang, gāst, gold, glōwan, gnætt, unglēaw, etc. More entries list "early Old English" variants with final [ɣ] corresponding to written -g, and "late Old English" variants with final [x] corresponding to written -g. Examples: bearg, burg, dweorg. I'd like to have one standard. Should we adopt the "late Old English" standard and not list the early variants? That's effectively what I've done, with initial [ɡ] and final [x].
12. Where is the stress in the following words? Hierusalem, Iohannes, Israhel, Matheus, Nazareth, Sodoma/Sodome/sodomisċ/sodomitisċ?

Thanks for you responses! Benwing2 (talk) 07:07, 14 December 2019 (UTC)

Thanks for the great work you've done on this module and on Old English entries! I can't completely answer these questions right now, but here are some quick replies:

1. I would lean against transcribing [ɒn], but I don't know much about this issue yet. I am still in favor of transcribing /ɑn/ /ɑm/, /on/, /om/ as [ɑn], [ɑm], [on], [om]); an article that I just read seems to me to support that system. "Old English an ∼ on: a New Appraisal," by Betty S. Phillips (1980), says "Because the fluctuation between spelling in a and o before nasals does not appear to be random, but to be based on a combination of word frequency and word class, the spellings in the Hatton manuscript should be accepted as phonetic, indicating fluctuating pronunciations within the community" (p. 22)--in other words, the variation between spellings with "a" and "o" in this position was probably not just an orthographical convention, but was correlated with variation in pronunciation with /ɑ/ vs /o/. The fact that the variants spelled with "o" were pronounced with a rounded vowel therefore doesn't imply that the variants spelled with "a" were pronounced with a rounded vowel. So I would use [ɑ] in the transcriptions on pages for "a" spellings and [o] in the transcriptions on pages for "o" spellings.
2. Transcribing final /ɣ/ as [x] seems OK to me, as I think I have seen sources that give final <g> = [x] as a possible spelling-sound correspondence in Old English. Based on Hundwine's answer below, I now think [ɣ] would be better.
3. I have read that geminate consonants were phonetically simplified in word-final position, so I think that we ought to transcribe them that way (although it does make a somewhat odd situation with medial [dːʒ] vs. final [dʒ], which we currently treat as /jj/ (both the phonemic and phonetic identity of this sound are debated by sources on OE)). I'm not very sure.
4. Per my previous comment higher up on this page, I'm in favor of /x/ instead of /h/ since it can be argued that this fricative has an underlying velar place feature in Old English rather than a glottal place feature, but I'm not strongly opposed to /h/. It is the same phoneme however it is transcribed.
5. Word-final fricatives should always be transcribed as voiceless. I haven't seen any source that says that they were voiced before vowels or before voiced consonant sounds and I have seen at least one (I forget which, but I can look for it if it's helpful) that explicitly says that voicing did not occur across a word boundary in these contexts. There is a hypothesis that, like some forms of present-day German, Old English used an epenthetic glottal stop [ʔ] at the start of phonologically vowel-initial stressed syllables, which makes it seem particularly unlikely to me that voicing occurred before vowel-initial words. I can't think of a similarly convincing reason for voicing to be absent before voiced consonant sounds, but as I mentioned above, the consensus in the literature on Old English seems to be that word-final fricatives were always voiceless, so I'd go with that in our transcriptions.
6. I think both bræġn and seġl were pronounced as a single syllable, based on the source I cited in the section immediately above this one.
7. I don't know how ġeond- was pronounced. Considering the given etymology and the usual distribution of <eo>, I'm confused about how the vowel even developed to <eo>. That said, I think Old English [jeond] would not be entirely inconstent with the form of present-day yond; there are other cases where a diphthong "eo" appears to have an o-like reflex rather than the expected e-like reflex (e.g. shoot, choose). (This alternate development of eo in some words has traditionally been attributed to a "stress shift" in the diphthong from the first to the second element, but there are arguments about the validity of that explanation.)
8. Many reductions of unstressed vowels were already usual or in progress in Old English, so regol = [ˈreɣul] looks completely plausible to me. However, Hundwine says below that it was just a mistake. I don't think there is any need to treat ⟨ol⟩ as anything but /ol/ in this pronunciation module.
9. I think that some words ending in -ian have variant spellings with "ge" instead of "i", which would seem to imply a pronunciation with [j] existed for those words. However, I would guess that when a form is spelled with "-ian", syllabic [i] was at least an additional option in most, if not all cases, so I think the module should generate only pronunciations in /ian/ [ian] from words spelled with ⟨ian⟩. Etymologically, the i in spyrian comes from [j], but I don't think the etymology counts for much: it is very possible for etymological [j] to develop into syllabic [i] in a context like this. I'm not sure whether there is any way to distinguish /ˈspyr.jɑn/ and /ˈspy.ri.ɑn/ in terms of metrical behavior. I'm updating my recommendation for <ian> to agree with Hundwine below. Since my first post, I've seen some sources that describe ⟨ri⟩ in class I weak verbs and in the strong verb swerian as a spelling of /r.j/ (e.g. "On the Consonantal Phonemes of Old English", Kuhn 1970, p. 45). An alternative spelling ⟨swerigan⟩ seems to exist, which I think does indicate a pronunciation with /i/, but that seems parallel to variation between ⟨g⟩ /j/ and ⟨ig⟩ post-consonantally in other words.
10. As far as I know, there is little or no evidence about the distribution or even the existence of [ç] in Old English. Before a consonant, it would be a realization of the same phoneme as [x], so the slash transcriptions of seht and sehtlian should be given as /seht/, /ˈsehtlian/ or /sext/, /ˈsextlian/. The Wikipedia articles Phonological history of Old English and Old English phonology say that when followed by /h/, the front vowels *i, *e, *æ developed through breaking into io, eo, ea, which strongly implies that coda /h/ was velar rather than palatal after all vowels at that time and in that variety (and also makes the existence and distribution of "ih", "eh", "æh" in the various dialects a bit puzzling to me; for many words, I'm not sure which forms correspond to which dialects).
11. Word/foot-initial [g] vs. [ɣ] in Old English is debated; I don't know of any convincing evidence one way or another, or even evidence that would allow us to confidently date a shift from [ɣ] to [g]. So giving two transcriptions labelled "early OE" and "late OE" seems unwarrantedly precise to me. I agree with giving only a single transcription, with [g], for glōwan, āˈgān, etc.

--Urszag (talk) 08:58, 14 December 2019 (UTC). Updated : Urszag (talk) 05:09, 15 December 2019 (UTC)

Here are my thoughts:

1. [ɒ] was an allophone of /ɑ/, not /o/. The distribution of [ɒ] is somewhat unclear to me, since the sources I've read say somewhat different things. Some commenters simply say [ɒ] occurred "before a nasal." Don Ringe says it was the outcome of earlier nasalized /ɑ/, and that /ɑ/ had been nasalized before a nasal except when it was unstressed and the nasal was in a different syllable. Neither of those explanations seems convincing to me. The whole evidence for [ɒ] existing is that ⟨o⟩ was often substituted for ⟨a⟩ before a nasal — but this only regularly occurred in stressed syllables! That's why you don't usually see ⟨on⟩ for /ɑn/ in the infinitive or in the inflectional endings of weak nouns.
2. imo final /ɣ/ should be rendered [ɣ]. The change of final [ɣ] to [x] took place in Late West Saxon, and the standard dialect on this website is Early West Saxon. Even the verb conjugations in the module you wrote are like 90% Early West Saxon.
3. Word-final geminate consonants probably were simplified by the written period, since they were often written with a single letter, so I'm fine with transcribing them as such in IPA.
4. Most academics seem to use /x/ instead of /h/, also /x/ looks cooler imo
5. I don't recall reading anything saying /s f θ/ were voiced between word boundaries.
6. Richard Hogg seems to think the final consonants in words like brægn and setl were syllabic. Don Ringe says they were often scanned as nonsyllabic in "Beowulf and other early verse," which tells me that they were probably syllabic by the time of written prose, but I'd have to do more research to be sure.
7. Not sure how geond was pronounced.
8. Final ⟨ol⟩ was pronounced /ol/, not /ul/. I've been trying to correct the entries I made that say otherwise, but apparently haven't gotten them all.
9. ⟨ian⟩ is pronounced differently depending on whether the verb is weak class II or weak class I. In weak class II verbs like scyldigian and a million others, it's pronounced /i.ɑn/, with syllabic /i/. In the few weak class I verbs that end in -ian like nerian and spyrian (plus the strong verb swerian), it's pronounced /jɑn/, with non-syllabic /j/.
10. /x/ was [ç] after a stressed front vowel, so sehtlian was /ˈsextliɑn/ [ˈseçtliɑn].
11. No one really knows when initial ⟨g⟩ changed from [ɣ] to [ɡ], except that at the beginning of the Old English period it was [ɣ] and by the end it was [ɡ].
12. I always assume Latin loans were accented like they were in Latin, though there may be no way to confirm how often this was actually true. Hundwine (talk) 03:21, 15 December 2019 (UTC)

Regarding question 12, I would default to transcribing such terms with initial stress unless it is specifically known that they had stress on another syllable. Kuhn 1970, p. 45 says Hierusalem/Gerusalem and Iuliana could alliterate on g/ġ; since alliteration is based on the onsets of stressed syllables, this implies that the first syllables of those names were stressed.--Urszag (talk) 05:09, 15 December 2019 (UTC)

Thanks both of you for your detailed responses. Here's what I'm thinking of doing:

1. Render /ɑn/ as [ɑn], /on/ as [on].
2. I will render final /ɣ/ as [ɣ]. I agree we should be using Early West Saxon as our base, which is also why I've been moving terms with y to the ie spelling whenever it's etymologically correct and the spelling with ie is attested (is attestation of spelling variants like this important, or can we just normalize the spelling in all cases?). User:Hundwine, are you sure that the change of final /ɣ/ to [x] occurred only in Late West Saxon? This is based on spelling alternations between final g and h; do they not occur in Early West Saxon?
3. I will implement simplification of final geminates in the phonetic version.
4. I will use /x/.
5. Fricatives will not be voiced across a word boundary; I assume dæġes ēage -> daisy with [z] was due to early univerbation, presumably occurring in speech in Late West Saxon before [z] became a phoneme.
6. I'll leave bræġn and seġl with non-syllabic n and l for now. I haven't yet figured out what to do about syllabic resonants in general.
7. I'll leave ġeond- as /jeond/. As for "shoot" and "choose", I always assumed that "shoot" had its vowel from the noun sċot (similar to how the verb "work" was probably influenced by the noun). Same thing seems to have happened in "lose", which even in Old English had a variant losian.
8. I'll leave written -ol as /ol/.
9. I'll make verbal -rian be /rjɑn/ after a short vowel. What about other weak I verbs in -ian, e.g. dynian, hrisian, temian, þenian? I'm guessing these are late forms that came about at a time when weak I and weak II started to be confused, but the module needs to do something with them. My instinct is to leave them as /ian/, and require that the user respell the pronunciation as -ġan to get /jan/.
10. I'll leave in place the current code that renders /x/ after a front vowel as [ç].
11. I'll go with initial /g/ as [g]. I've already removed the "early Old English" pronunciations with initial /g/ as /ɣ/.
12. There are definitely some entries for Latin borrowings whose existing pronunciation indicates non-initial stress, e.g. declī́nian, discípul, Ebrḗisċ, Iūdḗisċ, Rōmā́ne. I have kept the stress here. If these words exist in poetry, it should be possible to figure out the stress position based on alliteration. For the remainder, it is easy if the Latin stress is on the first syllable as in Nazareth, Sodoma, but unclear when the stress is elsewhere, e.g. Iohannes, Matheus, Israhel. The entry for modern Israel says it derives from Old English, and it has initial stress, but that may not mean much, and modern John and Matthew appear to be derived from French rather than Old English. For now I'll adopt Urszag's suggestion of putting initial stress, but we can revisit it later if need be.

Benwing2 (talk) 21:18, 15 December 2019 (UTC)

Answering your questions and adding a few comments:

1. imo we should normalize spelling in all cases. Slavishly adhering only to attested spellings is wrong in principle unless we all agree to stop using ⟨w⟩ instead of wynn along with other monstrosities, and the only reason to do it is pedantry at the expense of simplicity and consistency.
2. [ɣ] to [x] is attested sporadically in Early West Saxon, where spellings with ⟨g⟩ were still predominant. Spellings with ⟨h⟩ only become the norm in Late West Saxon.
3. ⟨rian⟩ wasn't always /rjɑn/, even after a short vowel. It was /rjɑn/ in words like werian and herian which result from Gmc *-rjan (almost all weak class I), and /riɑn/ in words like carian and sparian which result from earlier -ōjan (i.e. weak class II verbs). The difference is from etymology, not vowel length.
4. You're right about forms like trymian, þenian, etc., except they actually begin appearing in early Old English, where they still have their inherited weak class I endings in the past tense and the pres. ind. 3sg. In late Old English they become much more common and also start to completely merge with weak class II. There's a brief explanation in this PDF, pages 358 to 359.
5. Since it's purely a judgment call whether to have initial [ɣ] or [ɡ], my own view is that we should use [ɣ] because it's more archaic and also cooler. Nylt þū lā þæt Ealdenglisċ sīe cōl? Hundwine (talk) 00:13, 16 December 2019 (UTC)

My current thoughts:

I feel like there is a difference between normalization in the sense of replacing a littera with a standardized symbol that represents the same sound (e.g. wynn to w, k [k] to c, ð to þ, ae to æ), which I think is always unproblematic when selecting a primary form for a headword, and a process of replacing a spelling that represents a sound associated with a particular dialect or stage of the language with one that represents the expected correlate of that sound in another dialect or stage (which is what seems to be the situation with stuff like "ie" vs. "y" vs. "i"). It is conceivable that dialect mixing, incomplete sound changes, or irregular or analogical developments could prevent a hypothetical regular form from ever being widely used. So for the particular case that you're talking about, I think it is good to have attestation.

I think there are a few ways "daisy" could have gotten /z/; aside from univerbation in Old English, as you mention, if the first part continued to be recognized as the genitive singular of "day" up until the sound change that voiced the sibilant genitive suffix, it would have regularly developed /z/ the same as the independent word "day's". The OED has a quote indicating that the formation of the word was noticeable to at least some Middle English speakers: "c1385 G. Chaucer Legend Good Women 184 Wele by reson men it calle may The dayeseye, or ellis the eye of day."

By "existing pronunciation" do you mean the present-day English pronunciations of words like "disciple"? I think that in many cases, these pronunciations aren't directly derived from the Old English pronunciations, so the present-day English stress is not clear evidence of the Old English stress. For example, the /s/ in present-day English "disciple" is not consistent with a derivation from Old English "sc" (either palatal or velar); nor is the /aɪ/ consistent with a derivation from Old English short /i/.--Urszag (talk) 04:42, 16 December 2019 (UTC)

@Urszag I am inclined to agree that we should have attestation of cases of ie spellings. However, I'm willing to accept weak evidence for this, e.g. if Köbler has an entry under ie, I am currently accepting that as evidence of an attested ie spelling. Perhaps I should not accept this; Köbler does have a notation to indicate "this particular normalized spelling isn't attested" (a * following the headword), but may not be using it consistently.

As for "existing pronunciation", I am referring to the manually-entered pronunciations in the Wiktionary entries (using {{IPA|ang}}), before I replaced them with {{ang-IPA}}). Benwing2 (talk) 05:26, 16 December 2019 (UTC)

@Hundwine I'm inclined to use initial [ɡ] because it's not clear when the transition from initial [ɣ] to [g] happened, and it may well have happened before Early West Saxon. Also, the use of [g], when occurring inside of a word and not after /n/, makes it clearer that there's a hard morpheme break before it. Whether one spelling is "cooler" or not is a subjective criterion (even if enticing). Benwing2 (talk) 05:30, 16 December 2019 (UTC)

A question about un-

(Notifying Leasnam, Lambiam, Urszag, Hundwine, Mnemosientje): This is especially directed towards User:Hundwine. Your phonetic transcriptions list un- as unstressed in infinitives but stressed in past participles like unsawen and derivatives of present and past participles like unasundrodlic and ungenemnendlic. This seems unlikely to me; all other prefixes behave the same for infinitives and other verbal forms, including participles and nouns/adjectives derived from them. (But for some reason not stressed in ungewemmed and ungewemmednes, maybe a mistake.) What is your evidence for this? Benwing2 (talk) 08:47, 14 December 2019 (UTC)

The solution to this paradox is that un- was actually two different prefixes. There was the stressed prefix un- < PGmc un-, used with adjectives including those identical to participles, and the unstressed prefix un- < earlier on-, used with verbs. As for sources, you can look at page five of this study or page 51 of this book. Hundwine (talk) 07:28, 15 December 2019 (UTC)

There are two etymologically unrelated un- prefixes in modern English (and apparently also in Old English--I wasn't familiar with the situation there until checking just now). The negative prefix un-, which is supposed to take the primary stress in Old English (I think evidence for this is from alliteration), is typically not attached to verbs. The unrelated "reversative" prefix un- (apparently in variation with on-, and thought to be derived from and-) is typically attached to verbs, and I expect derivatives of those "reversative" verbs would follow the same stress patterns as derivatives of verbs with other prefixes. "Category:Old English words prefixed with un" seems to currently conflate both prefixes.---Urszag (talk) 09:03, 14 December 2019 (UTC)

Dutch and Middle Dutch ont- likewise have two etymologies – one of which corresponds to reversive un- – but only one prefixed-with cat each. Category:Ancient Greek words prefixed with ἀ- also makes no distinction between alpha privans and other alpha prefixes.  --Lambiam 10:24, 14 December 2019 (UTC)

OK, I will fix the module to distinguish between verbs and verbal derivatives, with un- having primary stress in the latter. Benwing2 (talk) 20:41, 15 December 2019 (UTC)

This is done, along with most of the other changes I've been intending to make. I still need to implement a mechanism to specify particular allophones (esp. of fricatives), and remove syllabic resonants (and maybe other things). Benwing2 (talk) 05:33, 16 December 2019 (UTC)

I added the "explicit allophone" mechanism, made -rian be /rjɑn/ after short front vowels in verbs, made -sian be [siɑn] in verbs, and removed syllabic resonants. Benwing2 (talk) 08:10, 16 December 2019 (UTC)

The prefix ed- and stress

(Notifying Benwing2, Leasnam, Lambiam, Hundwine, Mnemosientje): In "Prefixation and stress in Old English," by Donka Minkova (2008), I came across the surprising statement that the prefix ed- takes primary stress even when attached to verbs: "Two prefixes, and- ‘against’ and ed- ‘again’, attract stress uniformly across the entire derivational paradigm.

"(2) Uniformly stressed: and-, ed-: ándswàrian ‘to answer’, ándswàru ‘answer’, édnìwe ‘renewed’, édnìwian ‘to renew’

" yrre and anræd / ageaf him ándswàre Maldon 44

"Him þa ellenrof / ándswàrode Bwf 340

"and swa édnìwe / eft gewiorðan MB 11 39

"ealde æfþoncan / édnìwedan Juliana 485

"In this group primary stress on the prefixes reduces the stress on the etymological root to a secondary level of prominence. In this group too, the prefixes are inseparable from the etymological root" (p. 22). Is this correct? I don't know enough about Old English poetry to confidently evaluate the quotes from primary sources. What Minkova says here seems to contradict the form "ed-nī́wan" given in "Prosodically conditioned morphological change:preservation vs. loss in Early English prefixes", by Benjamin J. Molineaux (p. 31). Ed- also is included in a list of unstressed verb prefixes in "Phrasal Verbs: The English Verb-Particle Construction and its History," by Stefan Thim (2012), p. 148, and in "A Grammar of Old English, Volume 1: Phonology", by Richard M. Hogg (2011), §2.88. It does seem that the unstressed prefix ge- can be added before ed- in many verbs; e.g. forms like "geedníwodest" and "geedníwode", "geedstaðelode", "geedwyrped" and more are attested.--Urszag (talk) 08:15, 15 December 2019 (UTC)

@Urszag Interesting. The existing pronunciation given in andswarian does indeed have primary stress, but the pronunciations of other entries in and- suggests that this was an exception, and other verbs in and- do not have stress on and-. I did not know about ed-, but the two quoted examples definitely alliterate on ed- not on -niwe-. I wonder if that word is also an exception, are there similar examples for other verbs in ed-? Benwing2 (talk) 20:40, 15 December 2019 (UTC)

@Benwing2 Minkova doesn't give other specific examples for ed-. I will look and see if I can find some. By the way, I think the sources are in consensus that ed- was stressed when prefixed to a noun or adjective, so I am planning to edit the pronunciations for entries like edcierr and edstaþelig. I assume you left it unstressed to preserve what the previous editor had entered, but I can't find a clear basis for that editor's pronunciations.--Urszag (talk) 20:00, 20 December 2019 (UTC)

@Urszag This is fine with me. Yes, I left it unstressed because of what was there before. Benwing2 (talk) 20:29, 21 December 2019 (UTC)

Questions about individual words

(Notifying Leasnam, Lambiam, Urszag, Hundwine, Mnemosientje): I have converted all (or almost all) cases of {{IPA|ang}} to {{ang-IPA}}. But in the process of doing this, I found several words for which I have questions:

• beġeondan, ġeonre: Pronunciation given as beġondan, ġonre. I preserved them but it relates to the above issue of ġeond.
• cusċeote, ġenerstede: Two pronunciations, with and without secondary stress. I gave them secondary stress. Correct?
• fēasċeaft: Pronunciation given as /-ʃæft/, while sċeaft was given as /ʃæɑ̯ft/. I converted fēasċeaft to have /ʃæɑ̯ft/, is that correct?
• firgen: Should this be firġen? fyrġengāt has ġ.
• forewarnian, geondferan, onwrēon, twidǣlan: Given with primary stress on the prefix. I normalized them; correct?
• forewarnian, infēran, mānswerian, ymbhabban: Each one given with two pronunciations, one with primary stress on the prefix. I normalized this; correct?
• -ġē: Is the length correct?
• ġee: Pronounced ġē: Is this a real word?
• ġī: Is this a real word?
• gæġn: Given with initial [g-]: I preserved this; correct?
• Israhel: Given with [ç]. I normalized this; correct?
• offēstre: Given as offḗstre, with the stress on ē. I preserved this; correct?
• onġemang: Given as onġemáng. I normalized this; correct?
• orsāwle: Given as orsā́wle. I normalized this; correct?
• -raþ: Secondary stress or no? I assumed no. Correct?
• sċeadwian, sċeafa, sċeaga, sċeaþa: Can/should these be pronounced with [ɑ]? I assumed no in the pronunciations I generated.
• -sian: How pronounced? I think this is discussed above.
• sperehealf: Given as sperehéalf. I normalized this; correct?
• spinelhealf: Given as spinélhealf. I normalized this; correct?
• sumerhūs, winterhūs: Each given with two pronunciations, one with initial stress, one with stress on -hūs. I normalized them; correct?
• tohyht: Given as tohýht. I normalized this; correct?
• twirǣdness: Given as twìrǣ́dness. I preserved this; correct?
• unforht, untīme: Given as unfórht, untī́me. I normalized them; correct?
• unġewemmed, unwemme, unwemming: Given as unġewémmed, unwémme, unwémming. I normalized them to ún-; correct?
• ūþwitegung, ūþwitian: Given as ū́þwitegung, ū́þwitian. I normalized them; correct?
• ǣwignes: Should this be ǣwiġnes?
• þrīe: Given with two pronunciations, þrīe and þrī. I normalized to just þrīe. Correct?

Benwing2 (talk) 01:33, 16 December 2019 (UTC)

The "ea" cases all seem to be related to the topic of palatal diphthongization and a-restoration. In sċeafa, sċeaga, sċeaþa, I think a-restoration would have definitely applied because of the a in the final syllable, so [ʃɑ] looks correct. Compare with [ɑ] in cnafa, draca, fana. I don't fully understand the details on sċeadwian, but sċeadu seems to have back a. I can't figure out what's correct for sċeaft/-sċeaft.

I don't know of any reason to think [ç] was used in Israhel.

onġemang: the pronunciation of PDE among seems to support final stress. The first-syllable stress rule applied to nouns and adjectives; since onġemang is a preposition, I'm not sure whether first-syllable stress would be expected.

-ġē: On the one hand, it is supposed that Old English only had long vowels and diphthongs in stressed syllables. On the other hand, the situation with elements that are sort of like suffixes but also look sort of like the second elements of compounds seems to be difficult to determine.

I don't know of any reason why non-initial stress would be present in sperehealf, spinelhealf, sumerhūs, winterhūs. They look like compounds, so I agree with your changes.

ūþwitian: The etymology section for that entry says that the verb is derived from the noun ūþwita. If that is true, it would seem to explain initial stress, so I'm not certain about the pronunciation.--Urszag (talk) 06:30, 16 December 2019 (UTC)

@Urszag Thanks, I updated ūþwitian + derives, onġemang, sċeadwian etc., sċeafa, sċeaga, sċeaþa. Benwing2 (talk) 06:49, 16 December 2019 (UTC)

Regarding the "un-" prefixed words: I haven't seen any mention of a verb "unwemman" so far, and the prefix seems to have a negative rather than a reversative meaning. So I'd mark the stress on the prefix. Likewise for unforht and untīme. There was a tea room discussion about "ǣwignes" in May. --Urszag (talk) 16:51, 16 December 2019 (UTC)

Answering what I can answer that hasn't been covered by others:

• Hogg seems to think geond, geonre, and begeondan were pronounced /jɑnd/ [jɒnd], /ˈjɑnre/ [ˈjɒnre], /beˈjɑndɑn/ [beˈjɒndɑn], with the ⟨e⟩ marking the palatal pronunciation of the preceding ⟨g⟩ and the ⟨o⟩ marking rounded /ɑ/ [ɒ].
• You got cusċeote, ġenerstede, winterhūs, sperehealf etc. all right. The second elements of compounds were normally secondarily stressed.
• fēasċeaft should have /ʃæɑ̯ft/ from palatal diphthongization.
• firgen should be [ˈfirˠ.ɣen] with a hard ⟨g⟩, since it was *feorguni when palatalization happened. Obviously this means it also had a hard ⟨g⟩ in fyrgengāt.
• I'm pretty sure your renderings of forewearnian, inferan, etc. are correct, except that twidælan should be [ˈtwiˌdæːlɑn]. The prefix twi- dates all the way back to Proto-Indo-European, so when pre-Germanic shifted to word-initial primary stress, words beginning with twi- were included. The reason most other OE prefixes could be unstressed is that, unlike twi-, they were still separate words at the time this shift happened. (This also applies with twirǣdness, obviously.)
• The vowel length for -ġē should be right.
• ġī is a genuine attested spelling of ġēa, according to the Bosworth-Toller.
• Both G's in ġæġn should be palatal, Anglo-Frisian brightening occurred before palatalization.
• There's no way Israhel had [ç], which occurred only after stressed front vowels. Probably it was really pronounced [iz.rɑˈhe(ː)l], with the stress on the last syllable like in the original Latin. Intervocalic /x/ was lost except when it carried primary stress, and the Anglo-Saxons never restored it even through leveling (instead /ɡ/ was leveled into such words, e.g. hēage beside hēa 'high' masc.nom.pl.), so it's quite possible that [ˈiz.rɑ.hel] would just sound wrong to them, or even that they couldn't pronounce it.
• sċeafa, sċeaga, and sċeaþa all had /ɑ/ from a-restoration, which was not affected by palatal diphthongization. sċeadwian < *skadwōjan should have /æɑ̯/ because a-restoration didn't occur before two non-geminate consonants, but I guess it could have developed /ɑ/ by analogy to sċeadu /ˈʃɑdu/.
• Though ūþ- normally had primary stress (oþ- being its unstressed equivalent), this study gives ūþwítian. So I guess that word could be an exception.
• ǣwignes should indeed be [ˈæː.wij.nes]. [j] was never de-palatalized by contact with a following consonant: only palatal stops were. Thus þencst, þencþ, mengst, mengþ, but wȳsċst, wȳsċþ and sæġst, sæġþ. Hundwine (talk) 06:30, 17 December 2019 (UTC)

@Hundwine Thanks! Note that Campbell says firġin- "mountain-" [p. 59, footnote 3]. Benwing2 (talk) 07:15, 17 December 2019 (UTC)

@Hundwine Did Latin "Israhel" have final stress? Usually the last syllable of Latin polysyllables is unstressed (although there are supposed to be some exceptions). Perhaps the written "h" might correspond to hiatus rather than a consonant sound in the pronunciation of this name. Why is [ç] restricted to stressed syllables--wouldn't [ç] also occur in the coda of an unstressed syllable in a word like sceadiht or þorniht?--Urszag (talk) 10:42, 17 December 2019 (UTC)

Campbell "Old English Grammar" on stress placement

(Notifying Leasnam, Lambiam, Urszag, Hundwine, Mnemosientje): Campbell's Old English grammar has a bunch of interesting things to say about stress placement. It has a long section on Latin loans, which says among other things:

• "Late loan-words, like early ones, transferred the stress to the first syllable, but a strong half-stress remained on the syllable which had borne the main stress in Lat., so that the accentuation mágìster arose. Now in the native system, a half-stress on a medial formative syllable only existed after long syllables, and hence, if the half-stress was maintained, the accented vowel had to be lengthened to give the normal native pattern mā́gìster like ṓþèrne. Accordingly we find verse evidence for long vowels as māgister "master', and in cālend "calend", gīgant "giant", sācerd "priest", following the inflected forms cā́lèndas, &c., and also in Rōmane Romans, Lat. Rōmānī, where the first syllable would be shortened pretonically and then re-lengthened in OE. [in an earlier discussion, Campbell says that Vulgar Latin shortened pretonic syllables.] ... Conversely, a syllable short in Lat. receives a half-stress in OE because a long syllable preceds in áspìde "asp", Lat. aspidem."
• Three syllables: Ā́gùstus, Ṓmèrus, Nṓuèmbris, Dḗcèmbris, vs. Babilon, Elene, Nineue, Lucifer, Salomon with short first syllable and no secondary stress.
• Four syllables: Ā́gamèmnon, Ā́gustī̀nus, Hṓlofèrnus, Abimēlech, Benedictus, basilica [no stress marks in these three], Ápollī̀nus, Fílistī̀na, Hiérusā̀lem.
• Three syllables with hiatus: Mā́thèas, Mā́rìa; contrast Bắbilon, Iulius, Iunius, Libia, Siria.
• Four syllables with hiatus: Árriā̀nus, Cȳ́riā̀cus, Iū́liā̀na; Grḗgṓrìus, Ī́tā́lìa.
• [etc]

The long section on word accent says, among among other things:

• and- is always stressed, including in verbs like ándwyrdan, ándswarian that are derived from nouns. on- is the unstressed equivalent.
• for- can be stressed in fórwyrd "ruin" (also forwýrd), note also béhat/bēot "vow".
• æfter-, fore-, from-, mid-, (on)ġēan- are always stressed, even in verbs, because they are "prepositional adverbs" that form "pseudo-compounds". Examples of this type: æfter-spyrian "inquire", fore-ġesċrīfan "pre-ordain", from-hweorfan "turn away", mid-wesan "be with", onġēan-þingian "speak against", æt-wunian "dwell with", bī-standan "stand by", inn-gangan "enter", to-liċġan [sic, no length on to-] "lie near", of-adrīfan [sic, no length on a-] "drive away", on-lōcian "look at", wiþ-sprecan "speak against". [I don't know if I believe this]
• "Compound adverbs of which the first element is a preposition are stressed on the second, whether it be a noun or an adverb: todǽġ, onwéġ, onbǽċ [sic], beǽftan, befóran, beġéondan, behíndan, beínnan, benéoþan, onúfan, onúppan, wiþǽftan, wiþfóran, wiþínnan, wiþnéoþan, wiþū́tan, undernéoþan, tofóran [sic], togǽdere [sic], ætfóran, ætgǽdere."

Benwing2 (talk) 07:10, 17 December 2019 (UTC)

With regard to "half-stress" on medial syllables in Latin loans, I wonder whether this is to be regarded as equivalent to the secondary stress found in English compounds, or whether it's a less prominent kind of stress that we can ignore for the purposes of our transcriptions. Apparently the same kind of "half-stress" is found according to Campbell on the second syllables of certain inflected forms like singende ("Stress Assignment Rules for Old English Poetry," by Suksan Kim, p. 733).

As I mentioned somewhere above, there seems to be a fairly common idea that long vowels did not exist in unstressed syllables in Old English; I think that might be the reason why Campbell doesn't mark length on the first syllable of words like adrīfan, to-liċġan, todǽġ. Fulk 2002 has a footnote saying "Campbell (1959: §90 n. 4) prescribes that vowel length occurs only in stressed syllables in Old English" (p. 82), although Fulk says the situation with suffixes or suffixoids is difficult. It seems plausible to me that some words might have had variable reduction of vowels, as there is in the present-day pronunciation of "today".

The distinction between prefixes and prepositional adverbs (or "particles") is also brought up in Minkova 2008 (p. 24), as I mentioned in the tea room discussion on Old English pseudo-prefixes, but I haven't tried to understand this topic yet.--Urszag (talk) 09:34, 17 December 2019 (UTC)

@Urszag Campbell talks about a "strong half-stress" on p. 216 ("a strong half-stress remained on the syllable which had borne the main stress in Lat., so that the accentuation mágìster arose"). Campbell thinks that half-stresses require a preceding long foot, hence the first syllable of magister must be lengthened, producing mā́gìster. Conversely on p. 217 he talks about a "light half-stress" in the middle syllable of inflected forms like Ā́dàmes, Iā́còbes, "like native Bḗowùlfes, which is similarly neglectable in verse". Page 34 says "A half-stress always fell on teh second element of a compound when both the elements retained full semantic force: géoldwlànc proud with gold, wǽldrḕor blood of slaughter, gámolfèax grey-haired, fȳ́rhèard hardened by fire, fýrġenstrḕam mountain-stream." A nearby footnote says "The types of word which can have a half-stress are classified by J. Huguenin, Secondary stress in Anglo-Saxon (Baltimore, 1901)." Finally, p. 35 says "The half-stresses described in §89 were clearly very light when they fell on a syllable which was itself short (i.e. did not end in two consonants). Such half-stresses are often neglected in verse, and in late Old English syllables bearing them are frequently subject to change and loss, like fully unaccented syllables."

From all this I conclude that (1) "half-stresses" are inferred based on Old English metrical patterns in poetry (as first laid out by Sievers), and (2) they may or may not correspond to secondary stress. Presumably, if there's no half-stress, there's no secondary stress, but the opposite inference cannot be made. For now I'm sticking with the principles I've established in this module, which are (by default, but of course can be overridden):

1. Each element of a compound gets a stress, generally on the first syllable; the first element gets primary stress, the others secondary stress.
2. No secondary stress in obscured compounds like hlā́ford (but yes in the variant hlā́fwèard, which is still segmentable into its parts).
3. Nominal prefixes with primary stress automatically trigger a secondary stress on the following element (únclǣ̀ne, únġewèmmed, tṓcỳme, etc.).
4. "Verbals" (participles in -en/-ed/-od/-end/-ende and verbal nouns in -ing/-ung, possibly followed by -līċ(e), -nes + variants, or līcnes + variants, e.g. ācennedlīċ, tōmearcodnes and unġedafenlīcnes) follow verbs in their stress pattern, except for un-, which takes primary stress with such "verbals" but is unstressed in actual verbs.
5. Most verbal prefixes are unstressed, except for multisyllabic prefixes, which generally have secondary stress on the first syllable.
6. Most nominal prefixes have primary stress, except for be-, for- and ġe-.

Benwing2 (talk) 03:13, 18 December 2019 (UTC)

partially palatalized geminate [ç.x] should probably be [ç.ç]

On the pages hliehhan, hlehhan we find the /x.x/ transcribed phonetically as [ç.x]. This seems unlikely as -hh- is usually analyzed as a geminate consonant, and geminates are generally expected to have the same value throughout their double duration. Urszag (talk) 08:45, 18 January 2023 (UTC)

Voicing of intervocalic fricative

@Benwing2: At hūsum, the template generates the surface pronunciation [ˈhuː.sum]. Shouldn't it be [ˈhuː.zum]? I know I can explicitly specify [z], but I feel like it shouldn't be necessary in this case. —Mahāgaja · talk 11:28, 23 March 2024 (UTC)

Based on some experimentation, it produces [z] as expected for hūsa or hūsam. I think what's going on is that the module assumes that words ending in /sum/ (or some subset of these? I haven't looked at the module to see the exact criteria) will be compounds/suffix forms ending in -sum. I'm not necessarily a fan of hard-coding lexical rules like this in pronunciation modules but I think either path is a reasonable choice in cases like this.--Urszag (talk) 22:17, 23 March 2024 (UTC)

IMO the default assumption should be that words ending in -sum do not end in that suffix, and it's the ones that do that should be marked explicitly in the template parameter. —Mahāgaja · talk 08:16, 24 March 2024 (UTC)