ŋaren crîþa 9 vlefto: Ŋarâþ Crîþ v9

Project Elaine

Elaine is a name shared by several different female characters in Arthurian legend.

Elaine (legend) in Wikipedia

Reforming Ŋarâþ Crîþ morphophonology.

The current state of Ŋarâþ Crîþ morphophonology

The phonotactics of Ŋarâþ Crîþ distinguish between simple and complex codas, where complex codas may occur only at the end of a word. If morphology causes a complex coda to occur word-medially (such as by adding a suffix), and the bridge containing this coda cannot be reinterpreted as one without a complex coda, then a set of coda simplification rules takes place to remove the word-medial complex coda.

During Project Caladrius, it has been observed that some bridges, such as ⟦-cð-⟧ and ⟦-fp-⟧ rarely occur, to the degree that stem fusion with stems ending in such bridges are difficult to define. During Ŋarâþ Crîþ Corner #0x00, I have proposed the idea of restricting the set of valid bridges to a subset of all coda–onset combinations. To find what bridges are worth saving, I decided to analyze frequencies of bridges in a corpus of ŊCv9 text, but this has not yet provided much insight.

Originally, Project Caladrius was meant to revise the noun declension system, but the proposal of an operation that would make a stem have a valid coda such that a suffix beginning with a consonant could be attached to it, and the refinement of the idea into the current notion of stem fusion, has expanded the scope of the project and made me question if the current morphophonology of Ŋarâþ Crîþ was adequate.

Burning bridges

If a statistical analysis is not insightful, then reasoning from first principles to find which bridges should be valid might be. Here are the first three that come up in my mind:

  1. Lenition in the onset does not affect the validity of a bridge.
  2. The presence of ⟦j⟧ in the onset does not affect the validity of a bridge.
  3. All bridges that can be interpreted as having a coda that is null, ⟦-r⟧, or ⟦-l⟧ are valid.

That is, if ⟦-lt-⟧ is valid as an onset, then ⟦-lt·-⟧, ⟦-ltj-⟧, and ⟦-lt·j-⟧ are valid as well. Note that this does not say anything about the validity of the homophonous ⟦-lþ-⟧. Indeed, there might exist invalid bridges that are homophonous with valid bridges.

After that, there are two ways to approach the problem:

Odds and ends

The different types of boundaries.
Figure 1: The different types of boundaries in Ŋarâþ Crîþ: syllabic between syllables, glide between the initial and the medial, onset between the onset and the nucleus, and nuclear between the nucleus and the coda.

Project Caladrius has formalized the notion of a stem – namely, as a sequence of one or more syllables plus an onset. That is, the beginning of the stem marks a syllable boundary, while the end of a stem marks the boundary between an onset and the following nucleus.

Some operations are possible on some types of ends but not others. For instance, ending a word or adding a suffix such as ⟦-ten⟧ is not possible on an onset end, but adding a suffix such as ⟦-a⟧ is. (In this case, stem fusion is supposed to provide equivalents of the first two operations on an onset end.)

Affixes also have starts and ends. Prefixes always have syllabic starts, and suffixes always have syllabic ends. A sequence of letters making an affix might be compatible with multiple types of starts or ends; for instance, ⟦-a⟧ may have either a syllabic or onset start, depending on the context.

Additionally, there is another type of end called a terminal end, to which morphemes with syllabic starts cannot be attached. This end is produced by the presence of a complex coda at the end - that is, sword-medial complex codas are never produced in the first place.

Detailed rules for valid bridges

We reproduce the three initial rules proposed here, with some elaborations:

  1. Lenition in the onset does not affect the validity of a bridge. (This rule, however, does not establish a further relationship between the results of resolving two bridges that differ only by lenition in the onset. It also does not prevent lenition from affecting canonicality.)
  2. The presence of ⟦j⟧ in the onset does not affect the validity of a bridge.
  3. All bridges that can be interpreted as having a coda that is null, ⟦-r⟧, or ⟦-l⟧ are valid.
  4. New: If a bridge with a complex initial I is valid, then the bridge with an initial containing only the first consonant of I is also valid.

The following subsections describe situations in which bridges are invalid and must be repaired during concatenation. A bridge that is repaired might need further repair by later rules.

Importantly, bridge repair might cause earlier segments to change, and it is not idempotent: ⟦-sð-⟧ is repaired to ⟦-ss-⟧, but ⟦-ss-⟧ is repaired to ⟦-þ-⟧.

Also note that ⟦-cþ⟧ is now promoted to a simple coda.

Coalescence of ⟦-tš-⟧

The bridge ⟦-tš-⟧ is changed to ⟦-č-⟧.

Fortition of ⟦h-⟧ and ⟦ħ-⟧

The onset ⟦h-⟧ is fortited to ⟦c-⟧ after ⟦-s⟧, ⟦-þ⟧, ⟦-rþ⟧, ⟦-t⟧, ⟦-c⟧, ⟦-f⟧, or ⟦-cþ⟧. ⟦hr-⟧ and ⟦hl-⟧ are fortited analogously.

The onset ⟦ħ-⟧ is fortited to ⟦g-⟧ after ⟦-t⟧, ⟦-c⟧, and ⟦-f⟧. ⟦ħr-⟧ and ⟦ħl-⟧ are fortited analogously.

Metathesis of ⟦t⟧ before ⟦c⟧ or ⟦g⟧

⟦-tc-⟧ and ⟦-tg-⟧ are metathesized to ⟦-ct-⟧ and ⟦-cd-⟧, respectively. Likewise, ⟦-tcr-⟧, ⟦-tcl-⟧, ⟦-tgr-⟧, and ⟦-tgl-⟧ are metathesized to ⟦-ctr-⟧, ⟦-ctl-⟧, ⟦-cdr-⟧, and ⟦-cdl-⟧.

Similar bridges with lenited onsets, such as ⟦-tc·-⟧ and ⟦-tg·r-⟧ are treated analogously, with the resulting onset remaining lenited.

⟦t⟧ is deleted before ⟦cf⟧, ⟦cþ⟧, ⟦cš⟧, ⟦cš⟧, ⟦gv⟧, and ⟦gð⟧, devoicing the last two of these.

Nasal assimilation

For these rules, ⟦m·⟧ is counted as a nasal, even though it is pronounced as a fricative.

⟦-t⟧ before a nasal onset assimilates to ⟦-n⟧.

⟦-c⟧ before a nasal onset assimilates to ⟦-ŋ⟧. This is further corrected: ⟦-aŋ⟧, ⟦-oŋ⟧, and ⟦-uŋ⟧ to ⟦-or⟧, and ⟦-eŋ⟧ and ⟦-iŋ⟧ to ⟦-jor⟧ (coalescing the ⟦j⟧ if necessary), and analogously with the hatted vowels. As a special case, ⟦-cŋ-⟧ is repaired to ⟦-ŋ-⟧ instead.

Denasalization of ⟦ŋ-⟧

After ⟦-s⟧, ⟦-þ⟧, ⟦-rþ⟧, ⟦-f⟧, and ⟦cþ⟧, ⟦ŋ-⟧ is denasalized to ⟦g-⟧.

Devoicing of ⟦v-⟧ and ⟦ð-⟧

After ⟦-þ⟧, ⟦-rþ⟧, ⟦-t⟧, ⟦-c⟧, ⟦-f⟧, and ⟦-cþ⟧, ⟦v-⟧ devoices to ⟦f-⟧ and ⟦ð-⟧ devoices to ⟦þ-⟧. Additionally, ⟦ð-⟧ is devoiced after ⟦-s⟧.

This process occurs analogously for the onsets ⟦vr-⟧, ⟦vl-⟧, ⟦ðr-⟧, and ⟦ðl-⟧, except that ⟦-þ⟧ is deleted before ⟦vr-⟧ and ⟦vl-⟧ instead. Additionally, ⟦-rþCR-⟧ onsets (with R = ⟦r⟧ or ⟦l⟧ and C = ⟦v⟧ or ⟦ð⟧) are corrected to ⟦-RC-⟧.

As usual, similar rules apply to lenited onsets: ⟦v·⟧ devoices to ⟦f·⟧, and ⟦ð·⟧ is replaced with a copy of the preceding consonant.

Assimilation of ⟦s⟧ after ⟦þ⟧

After a ⟦þ⟧, ⟦s⟧ is replaced with ⟦þ⟧. Additionally, ⟦ss⟧ is coalesced into ⟦þ⟧, unless it is not followed by a consonant and the latter ⟦s⟧ arose from a ⟦ð⟧ in the previous step.

Degemination before another consonant

⟦þ⟧, ⟦t⟧, ⟦c⟧, and ⟦f⟧ are degeminated before another consonant in the onset; for instance, ⟦-ffr-⟧ is corrected to ⟦-fr-⟧, and ⟦-ccs-⟧ is corrected to ⟦-cs-⟧. ⟦td⟧ is degeminated to ⟦d⟧, and ⟦cg⟧ is degeminated to ⟦g⟧.

This rule also applies when the second instance of the degeminated consonant is lenited, in which case the first copy of the consonant is elided: ⟦-tt·l-⟧⟦-t·l-⟧.

Partial coda elision of bridges with ⟦-rþ⟧ and ⟦-cþ⟧ codas

If the coda is ⟦-rþ⟧, then it becomes ⟦-r⟧ before a fricative followed by ⟦r⟧ or ⟦l⟧, or before the onsets ⟦cf-⟧, ⟦cþ-⟧, ⟦cs-⟧, ⟦cš-⟧, ⟦tf-⟧, and ⟦dv-⟧. Before any other two-letter onset, it becomes ⟦-þ⟧.

If the coda is ⟦-cþ⟧, then it is maintained before the onsets ⟦þ-⟧, ⟦š-⟧, ⟦m·-⟧, ⟦t-⟧, ⟦ħ-⟧, ⟦m·-⟧, or ⟦t·-⟧. Before ⟦cf-⟧, ⟦cþ-⟧, ⟦cs-⟧, ⟦cš-⟧, or ⟦tf-⟧, or before any of the onsets consisting of ⟦þ⟧, ⟦š⟧, or ⟦ħ⟧ followed by ⟦r⟧ or ⟦l⟧, the onset loses its first consonant, and ⟦cs-⟧ additionally becomes ⟦þ-⟧. In all other cases, the coda becomes ⟦-þ⟧.

A revised view of the layer-0-to-1 pipeline

Morphemes are no longer viewed as raw MGP sequences; rather, they are viewed structurally. Some constraints apply to all morphemes:

When morphemes are joined together, repair processes must maintain these invariants at the join boundary. Morphemes that end in complex codas are simply forbidden from having anything appended to them. The second invariant is maintained by applying bridge correction when two morphemes are joined at a syllable boundary, and the third invariant is maintained by eliding the ⟦j⟧ if necessary when joining at an onset boundary.

In addition, there are environments that may naturally occur within a morpheme but are repaired away when created by appending morphemes. For now, these include deduplication targets such as ⟦-vav-⟧ and ⟦-ðoð-⟧, which arise when joining morphemes at a glide, onset, or nuclear boundary.

Complex coda simplification will no longer need to occur from layer 0 to 1, since complex codas cannot have anything appended to them in the first place.

More formally, we adopt the following notation:

Stem fusion

The difference from Project Caladrius is that the stem and the result are treated structurally instead of linearly. The notation is also different – given a stem τΤ:

The following variables are used: Σxy𝒮xy, γΓ, ιΙ, νΝ, κΚ, ωΩ, θ{𝗇,𝗍,þ}.

As usual, earlier rules take precedence over later ones. In addition, we use a shorthand for C-invariant rules: a rule such as τσ, where τΤ and σ𝒮ss is interpreted as the rules τε=σsω and τθ=σ:θ.

Another shorthand used in this document is ττ, which implies τε=(τ)ε and τθ=(τ)θ.

Stems ending in ⟦j⟧


From now on, any explicit instances of εΜ will be omitted.

Onset aliasing

(Σssι)ε=(Σss:𝗌)ε if ι{𝗍,𝖽}Alias

Valid codas

(Σsnγ)ε=Σsn:γΩ if γΩ(Σsnγ)θ=Σsn:γΚ:θ if γΚValidCoda


ΣsnκιΣsn:δ(κ) if κ=ι and |κ|=1Degeminate

where |κ| is the number of manifested grapheme phrases in κ and

δ(κ)={𝗅 if κ=𝗋þ if κ=𝗌κ otherwise

Vowel epenthesis

ΣsnκιΣsnεΚκι:𝖾Ν:ιΚ if ιΙΚ, κ{𝗋,𝗅}Epenthesis-LC

Nasal merging

Σsn𝗇Κ𝖽ΙΣsn𝗇ΙΣsn𝗇Κ(ι1ι2)ΙΣsn𝗇(ι2)Ι if ι1=𝖽Σsn𝗇Κ𝖽ΙΣsnŋΙΣsgμν𝗇Κ(ι1ι2)ΙΣsgξ(μν)𝗋(ι2)Ι if ι1{𝖼,𝗀}NasalMerge1(ΣsnεΚ𝗆)𝗇=Σsn𝗇𝗆(Σss𝗆)𝗇=Σss:𝗈ˆ𝗆(Σss𝗇)𝗇=Σss:𝖾𝗇𝗇(Σsoνκι)𝗇=Σsoχ(ν)κ:ι𝖽ð𝗏𝗇𝗇𝗆 if ι{𝖽,ð,𝗏}NasalMerge2

ξ:𝒮gn0𝒮gn0 is the ξ-transformation; i.e.

ξ(μν)={μ𝗈 if ν{𝖺,𝗈,𝗎}μ𝗈ˆ if ν{𝖺ˆ,𝗈ˆ}𝗃𝗈 if ν{𝖾,𝗂}𝗃𝗈ˆ if ν{𝖾ˆ,ıˆ}

χ:ΝΝ inverts the tone of a vowel; i.e.


Obstruent merging

Σsn𝗋þιΣsn𝗋:ι if φ(þ,ι)Σsn𝖼þιΣsn𝖼:ι if φ(þ,ι)ΣsnκιΣsnε:ι if φ(κ,ι)FricMerge1(Σss(ι1ι2)Ι)θ=(Σss:ι1)θ if φ(ι2,θ)(Σssι)θ=Σss:θ if φ(ι,θ)FricMerge2



Final devoicing

(Σssι)ε=[Σssh(ι)]ε if ι{𝗏,𝗆,𝖽,ð}(Σssι)θ=[Σssh(ι)]θ if ι{𝗏,𝗆,𝖽,ð} if θ{𝗍,þ}FinalDevoice



Stems ending in consonant–liquid onsets

Let ρ{𝗋,𝗅} and πΠ. Then πρΙ.

ΣsnεΚινεΚ(πρ)ΣsnιΚ:πΙ:ν:ρΚ if ι{ε,𝗌,𝗇,𝗅}Cl-Meta(ΣsnεΚ(𝗍Πρ))𝗍=Σsn:𝗋Κ:𝗍(ΣsnεΚ(πρ))𝗍=(ΣsnεΚπΙ)ε𝖽Cl-NoCoda-T(ΣsnεΚ(πρ))𝗇=Σsn(ι𝖽𝖼𝗋𝗌)Κ:ŋ if π{𝖽,𝖼}(ΣsnεΚ(πρ))𝗇=(ΣsnεΚπΙ)𝗇Cl-NoCoda-N(ΣsnεΚ(πρ))þ=Σsn:𝗋Κ:π if π{þ,ð}(ΣsnεΚ(πρ))þ=(ΣsnεΚπΙ)εðCl-NoCoda-Þ(Σss(π𝗋))ε=(ΣssπΙ):𝗈ˆ𝗋(Σss(π𝗅))ε=(ΣssπΙ):𝖾ˆ𝗅Cl-Nil(Σss(πρ))θ=(ΣssπΙ):𝖾ˆgs:θ if θ{𝗍,þ}Cl-TÞ(Σss(πρ))𝗇=(ΣssπΙ):𝗈ˆgs:𝗇Cl-N

where for any consonant c and coda ωΩ,


and :ΩΚ denotes the operation of taking the maximal prefix of a coda that is a simple coda.

Stems ending in ⟦r⟧ or ⟦l⟧

(Σss𝗋)ε=(ΣssεΙ):𝗈ˆ𝗋(Σss𝗅)ε=(ΣssεΙ):𝖾ˆ𝗅L-Nil(ΣssρΙ)θ=(ΣssρΙ):𝖾ˆgs:θ if θ{𝗍,þ}L-TÞ(Σsn𝗇ΚρΙ)𝗇=(Σsn𝗇Κ):𝗇(ΣssρΙ)𝗇=(ΣssρΙ):𝗈ˆgs:𝗇L-N

Stems ending in ⟦š⟧, ⟦ł⟧, or ⟦č⟧

ΣsnκιΣsnI(κι)𝗂gs if ι{𝗌ˇ,ł,𝖼ˇ,𝖼𝗌ˇ}Sˇłcˇ

where I:ΓΓ is defined as

I(γ)={𝗋þ if γ=𝗋þ𝗌ˇ𝗋𝗍 if γ=𝗋þ𝖼ˇþ𝖼ˇ if γ=𝖼þ𝗌ˇł if γ=𝗅ł𝖼ˇ if γ=𝗍𝖼ˇγ otherwise

Stems ending in ⟦c⟧ or ⟦g⟧

(Σsnκ𝖼)ε=Σsn:κ𝗌þ[𝖼þ][𝖼þ](Σss𝖼)θ=Σss:θ𝗍𝗇þ𝗍ŋ[𝖼þ]Cþ(ΣsoνεΚ𝗀)ε=ΣsoνεΓ𝗂εΩ if ν{𝖺,𝖾,𝖺ˆ,𝖾ˆ}(ΣsoνεΚ𝗀)ε=Σsoν𝗌G-Nil(Σss𝗀)ε=Σss:𝗂sωG(Σsnκ𝗀)θ=Σsnκ:𝗂ss:θ if κW(Σss𝖼)θ=Σss:θ𝗍𝗇þ𝖽ŋ[𝗀ð]Gð

where W is the set of codas that end in a voiceless consonant.

Stems ending in ⟦p⟧

ΣsoνεΚ𝗉Σsoν𝖿 if ν{𝖾,𝖾ˆ}ΣsoνεΚ𝗉ΣsoνεΓ𝖾εΚP-NilΣsn𝗇Κ𝗉ΣsnεΚ:𝗆𝖾ssΣss𝗉Σss:𝖾ssP

Stems ending in ⟦h⟧


Stems ending in ⟦ħ⟧


Stems ending in ⟦ŋ⟧

ΣsgμνκŋΣsgξ(μν)𝗋 if κ{ε,𝗇}Ŋ-N(Σsnκŋ)ε=Σsn𝗅𝗈ˆ𝗋nω if κ{𝗋,𝗅}(Σsnκŋ)θ=Σsn𝗅Κ:θ if κ{𝗋,𝗅}Ŋ-Rl

Stems ending in any other onset with two consonants


Coda-based rules

NB: by this point, the only possible onsets at the end of the stem are ⟦n s þ f⟧.

  • All two-consonant onsets have already been handled.
  • ⟦c ŋ š r l ł g p č h ħ⟧ handled by their respective rules.
  • ⟦v m d ð⟧ handled by final devoicing.
  • ⟦t⟧ handled by onset aliasing for fusion with ε and by obstruent merging for fusion with θ.

By observation, the only possible codas in the final bridge at this point are ⟦s n þ rþ t f⟧.

  • The empty coda is obviously eliminated, as all of ⟦n s þ f⟧ are valid simple codas.
  • ⟦r⟧: ⟦rþ⟧ is a valid coda; other cases handled by (Epenthesis-LC)
  • ⟦l⟧ handled by (Epenthesis-LC); in case of fusion with ε, ⟦lþ⟧ and ⟦lt⟧ are valid complex codas
  • ⟦c⟧: ⟦-cn-⟧ is an invalid bridge in the first place; ⟦-cs-⟧, ⟦-cþ-⟧, and ⟦-cf-⟧ are interpreted as having an empty coda and a complex onset.
  • ⟦cþ⟧: ⟦-cþn-⟧, ⟦-cþs-⟧, and ⟦-cþf-⟧ not valid. ⟦-cþþ-⟧ handled by obstruent merging.

Some codas are limited to certain onsets at this point:

  • ⟦t⟧ can be followed only by ⟦s⟧: ⟦-tn-⟧ not valid, ⟦-tf-⟧ canonicalizes to a null coda, and ⟦-tþ-⟧ handled by obstruent merging.
  • ⟦þ rþ⟧ followed only by ⟦n⟧: neither can precede ⟦s⟧. ⟦þþ rþþ⟧ handled by degemination or obstruent merging. ⟦þf rþf⟧ handled by obstruent merging.
  • If fusing with a null consonant, ⟦n⟧ is followed only by ⟦f⟧: ⟦-ns⟧ and ⟦-nþ⟧ are already valid complex codas, ⟦-nn⟧ handled by degemination.

The bridge ⟦-ts-⟧


The codas ⟦-s⟧, ⟦-þ⟧, ⟦-rþ⟧, and ⟦-f⟧

Σsoν𝗋þιΣsoη(ν)𝗋:ιCoda-RþΣsoνκιΣsoη(ν)ε:ι if κ{𝗌,þ,𝖿}Coda-Sþf

The coda ⟦-n⟧


Properties of stem fusion

Fusion with ⟦t⟧ is invariant only when the final onset of the stem is ⟦t-⟧.

Fusion with ⟦n⟧ is invariant only when the final bridge of the stem is ⟦-nn-⟧.

Fusion with ⟦þ⟧ is invariant only when the final onset of the stem is ⟦þ-⟧ or ⟦cþ-⟧.

Future work

The first complete draft of stem fusion is specified and implemented in f9i, but it needs testing with real-world input to assess what needs to be fixed.

Obsolete rules

Rules too complicated for their own good.

Stems ending in ⟦f⟧, ⟦v⟧, or ⟦m⟧ (first pass)

This pass applies only to onsets ending in these consonants that are not preceded by a nonempty coda.

(ΣsnεΚι)ε=Σsn𝖿 if ι{𝖿,𝗏,𝗆,𝗍𝖿,𝖽𝗏,𝖼𝖿,𝗀𝗏}(ΣsnεΚ𝖼𝖿)þ=Σsn𝖼þ(ΣsnεΚ𝗀𝗏)þ=Σsn𝗀ð(ΣsnεΚι)θ=Σsn𝖿:θ if ι{𝖿,𝗏,𝗆,𝗍𝖿,𝖽𝗏,𝖼𝖿,𝗀𝗏}Fvm1

Stems ending in ⟦p⟧, ⟦t⟧, or ⟦c⟧

For ι*{𝗉,𝗍,𝖼}:

Σsgμν𝗇Κ𝖼ΙΣsoξ(μν)𝗋Σsn𝖼Κ𝗍ΙΣsn𝖼þ(Σsn𝗇Κ𝗍Ι)ε=Σsn𝗇𝗌Ptc-SpecialΣsnεΚι*Σsn:H(ι*)Ptc-NoCodaΣsoνκι*Σsoχ(ν):H(ι*) if κ{𝗌,𝖿,þ}Ptc-SfthΣsoν𝗋þΚι*Σsoχ(ν):H(ι*)Ptc-Rth(Σssι*)þ=Σss:𝖼þPtc-Þ(Σssι*)𝗇=Σss:(ι*)𝗉𝗍𝖼𝗆𝖽ŋPtc-N(Σssι*)𝗍=Σss:𝗍Ptc-T(Σsnκι*)ε=ΣsnκΩ if κ{𝖼,𝗍,𝖼þ}Ptc-Nil1(Σsnκ𝗉)ε=Σsn𝗅𝗍 if κ{𝗋,𝗇,𝗅}Ptc-Nil2

where H,H:{𝗉,𝗍,𝖼}Κ are defined as


Stems ending in ⟦d⟧ or ⟦g⟧

For ι*{𝖽,𝗀}:

(ΣsnεΚι*)þ=ΣsnεΚ(ι*)𝖽𝗀ðħDg-Nil-ÞΣsnεΚι*Σsn(ι*)Κ𝖽𝗀𝗍𝖼Dg-Nil(Σsn𝗋Κ𝖽Ι)ε=Σsn𝗅(Σsn𝗋Κ𝖽Ι)θ=Σsn𝗅:B(𝖽,θ)D-r(Σsnκ𝖽Ι)ε=Σsn(κ𝗍)Ω if κ{𝗌,𝗅}D-sl(Σsnκ𝖽Ι)ε=ΣsnκΩD-WithNil(Σsnκ𝗀Ι)ε=ΣsnκΩ if κ{𝗋,𝗅}G-Rl(Σsnκ𝗀Ι)ε=(Σsnκ:𝖼)εG-WithNil(Σssι*)θ=Σss:B(ι*,θ)Gd


B(ι*,θ)={θ𝗍𝗇þ𝖽𝗇ð if ι*=𝖽θ𝗍𝗇þ𝗀ŋħ if ι*=𝗀

Stems ending in ⟦f⟧, ⟦þ⟧, or ⟦s⟧

(Σssι)ε=(Σss:ι[𝖼𝖿][𝖼𝗌][𝗍𝖿]𝖼𝗌𝗍)ε if ι{𝖼𝖿,𝖼𝗌,𝗍𝖿}(Σssι)θ=(Σss:ι[𝖼𝖿][𝖼𝗌][𝗍𝖿]𝖼𝗌𝗍)θ if ι{𝖼𝖿,𝖼𝗌,𝗍𝖿}SfcΣsnκιΣsnιΚ if κ{𝗌,𝖿,þ}Sfth1