Hiatus Resolution Mechanisms in Old English

Old English has several strategies of hiatus resolution, which have received a lot of attention in diff erent theoretical approaches. This article discusses these strategies from a constraint-based approach within Optimality Theory. The analysis relies mainly on the solutions proposed by Opalińska (2002; 2004; 2006), and reveals the hierarchy of preference between 4 strategies of hiatus resolution: contraction, diphthongisation, gliding, and glide insertion. It is shown that all mechanisms are a result of the interaction of diff erent constraints. The article reveals that diphthongisation is the most optimal mechanism, whereas contraction occupies the last position due to the violation of weight preservation principle. The goal of this article is to advance Opalińska’s solution by demonstrating that the preference for a given hiatus resolution strategy results from the fact which particular subset of constraints is needed to activate this strategy.


Introduction
Old English belongs to the western branch of Germanic languages.Like other languages in this group, Old English is weight-sensitive with a strong preference for fi lled onsets, which in turn results in diff erent strategies of hiatus resolution.This pattern has been long known in linguistics.For example, Kiparsky (1998, 1) claims that in Germanic languages Syllable structure is also constrained more directly by a preference for simple onsets, which entails avoidance both of hiatus and syllable-initial consonant clusters.Processes of syllabifi cation, deletion, shortening and lengthening in the Germanic languages favour those quantitative and syllabic patterns that fi t these prosodic conditions… As a member of the Germanic language family, Old English shares the same basic phonological conditioning.Similarly to Modern English, its ancestor maximises onsets, as shown in the words: ['kniçt] 'knight', ['provjan] 'to prove', ['druɣoθ] 'drought'.However, intervocalic clusters are heterosyllabic, for example in the [vj] cluster in [provjan] the consonant falls into the coda of the preceding syllable whereas [j] occupies the onset of the following syllable.Such syllabifi cation results from a diff erent organisation of rules within the Syllable Structure Algorithm (SSA).Specifi cally, Old English right after the erection of the syllable proceeds to the erection of the rhyme, instead of maximising the onset.Technically, these facts are translated into the statement that in Old English Coda Rule precedes Complex Onset Rule, unlike in Modern English, where the ordering is reverse (Opalińska 2002, 56).
As shown above, the syllable structure in Old English is predictable and subject to SSA.Consequently, it is assumed that it cannot be underlying.However, as for the moraic structure, it needs to be encoded in the UR.To clarify, Old English has a contrast between long and short vowels, which in turn has implications for the stress placement.Additionally, a syllable is counted as long when its vowel is followed by more than one consonant (Campbell 1959;Hogg 1992).This fact suggests that consonants can contribute to syllable weight.However, onsets do not bear any moras in Old English (Opalińska 2002, 59).Another important feature is that Old English permits maximally bimoraic syllables (Dresher and Lahiri 1999) -a fact that has implications for hiatus resolution strategies, which have received much attention in the literature.SPE phonology (Chomsky and Halle 1968) has analysed them linearly in terms of language-specifi c rules (see Keyser 1975;Lass and Anderson 1975;Hogg 1992), whereas Optimality Theory approach (OT; Prince and Smolensky 1993;McCarthy and Prince 1995) has discussed contraction in terms of universal constraints (Opalińska 2002;2004;2006).The following sections present the details of the latter analysis, specifi cally that of Opalińska (2002).
(1) Long and short diphthongs in Old English (a) eolh Long diphthongs may be formed due to a merger of two vowels.This is illustrated on the development of the Old English verb bēo ('I am').Historically, the underlying representation is //be//+//o// 1 .In order to arrive at the form bēo, the two forms needed to coalesce.This did not, however, lead to the deletion of a mora, which is a manifestation of the principle of weight preservation (Opalińska 2002, after Hayes 1989), according to which the moraic structure in Old English has a strong tendency to preserve the same number of moras between the input and the output.Consequently, the coalesced bēo still has two moras, thus achieving the optimal bimoraic structure.The motivation behind coalescence lies in the syllable structure.Without diphthongisation, there would emerge an onsetless syllable in /be.o/: a prohibited structure in Old English.
In order to arrive at the form with a diphthong, in OT terms there has to be a constraint promoting diphthongisation.This function is fulfi lled by O (Syllables must have onsets), the position of which must be high in the ranking, crucially higher than N D (N ), a constraint prohibiting diphthongisation 2 (Opalińska 2002, 170).As argued above, the number of moras in the output was preserved, which could be ensured by a high-ranked M μ (Input moras must have output correspondents).Consequently, the ranking for bēo is as follows.
(2) O , M μ >> N D It is not only important to ensure that no moras were deleted but also that no segmental content was erased.Consequently, the ranking should include M -V (Input vowels must have output correspondents).This is shown in the tableau below.
(3) Tableau for bēo The faithful candidate (3a) is eliminated by undominated O due to having an onsetless syllable.Candidate (3c) commits a fatal violation due to the deletions of a vowel and a mora.The deletion of a mora in candidate (3d) also proves fatal, and thus leads to its elimination.The optimal candidate (3b) incurs the least costly violation of N D and thus wins in the evaluation.It is worth noting that thanks to diphthongisation the optimal candidate actually avoids the hiatus of two vowels.It is done at the lowest possible cost to faithfulness because no material is deleted in the output.Both moras and vowels are preserved.Consequently, diphthongisation was suggested to be the most optimal method of avoiding onsetless syllables (Opalińska 2002, 173).

Gliding
Despite its advantages, diphthongisation turns out not to be the most frequent strategy of hiatus resolution in Old English, especially in verbal declension.Historically, these verbs are built of the following morphemes: a root, a stem extension suffi x -i-and an infl ectional ending -an.Given this structure, the surface /j/ derives from the underlying //i//.Consequently, a hiatus arises at the stem-suffi x boundary.For example, in the word [nerjan] 'save', the morphological structure is as follows /ner+i+an/ and yields a vocalic sequence /ia/.Since there is no */ia/ diphthong in OE, diphthongization is precluded.As a matter of fact, the absence of /ia/ in the inventory is not accidental because Old English disprefers diphthongs whose constituents are of diff erent feature specifi cation for height (Opalińska 2002, 171).In OT terms, this implies an undominated constraint D H (DH: Constituents of a diphthong must agree in height).The change from /i/ to /j/ suggests that class I weak verbs resolve hiatus by gliding.
To further complicate the image, a change from a vowel to a glide implies the loss of a mora.Perhaps, the motivation for gliding and at the same time for preserving the vowel [a] lies then in the fact that it is more costly to delete a vowel than to delete a mora.Technically, such a statement would imply a dominance of M -V over M μ, and would thus promote gliding.In [nerjan] no vowel is actually deleted because //i// corresponds to /j/.Consequently, neither /nerian/ nor /nerjan/ would violate M -V but it seems that the latter would incur a violation of M μ.However, in actuality there would be no violation of M μ due to some additional facts of moraic conditioning.Namely, Old English is a weight-sensitive language which has Weight-by-Position (see Opalińska 2006 and references therein); in other words, the coda in Old English receives a mora (5).
What is more, Old English disprefers clusters of a consonant and [j]: *Cj (Barber 2013, and references therein), which results from OE syllabifi cation.Consequently, the syllable division in ner.jan compels the /r/ into the coda, which in turn triggers mora addition by Weight-by-Position.Importantly, the word-fi nal consonants cannot receive moras due to an undominated restriction in Old English not to attract stress to fi nal syllables.This is translated into OT as N (N F ) (No moras on the word-fi nal consonant) (Prince and Smolensky 1993).As a result, both */nerian/ and /nerjan/ have 3 moras each, as shown in (6) below.The graphs in (6) show that all input moras are retained in both candidates, so M μ cannot decide which is the optimal candidate.The diff erence lies in the association of moras to segments.Namely, in (6a) the high vowel /i/ loses its mora, which turns out to be a more optimal solution than preserving it, as in *[nerian].This would technically imply the dominance of M μ over I μ (The association between the underlying mora and the segment it dominates must be retained in the output form.). To sum up our discussion so far, the ranking of constraints is given in (7).For expository purposes, the undominated constraint N is excluded from the rankings in this article.
(7) DH, M -V, M μ >> I μ It is worth noting that like in the case of diphthongisation, gliding is also triggered by the need to fi ll the onset position.*[nerian] has an onsetless syllable [an], and that is the reason why it is dispreferred.In OT terms this implies the dominance of O .The glided candidate /nerjan/ still incurs a violation of faithfulness because /i/ loses its mora, but it is still a more optimal solution than leaving an onsetless syllable.
With reference to the constraints on weight, it must be emphasised that the ranking should include the constraint W --P (W P), which must be ranked higher than faithfulness D μ, a constraint militating against any addition of a mora.W P in turn should be ranked lower than undominated N because it would be a more costly operation to insert moras on all codas than to leave the word-fi nal coda moraless.To sum up, the partial ranking should take the shape of N >> W P >> D μ.For expository purposes, the highest and the lowest ranked constraints will not be included in the tableau as they would be never violated in candidates which do not add moras.No mora-insertion is in line with the principle of weight preservation (see section 2).
As a fi nal point, it must be also emphasised that the discussed ranking of constraints permits the possibility of both gliding and glide insertion.In other words, both candidates [nerjan] and *[nerijan] would be selected as optimal.Neither of them violate O or weight preservation principle.Worse still, the candidate *[nerijan] turns out not to violate I μ, as opposed to the candidate with gliding.Consequently, the candidate with glide insertion seems to be more harmonic than the desired output.In order to ensure the selection of [nerjan], the ranking should include a constraint which bans insertion of segments.This function could be achieved by a faithfulness constraint D ( ) (No insertion of segments).Importantly, such a constraint should be ranked higher than I μ.In other words, it should be a more costly operation to insert a segment than to change the association of a mora.To conclude, the ranking of constraints presents itself as shown in (8), and the evaluation follows in (9).The tableau in ( 9) illustrates an OT evaluation of nerian.As shown above, the candidate which imposes gliding (9b) to satisfy O is optimal.It violates I μ, but such violation of faithfulness is the least costly option in a given ranking.The faithful candidate (9a) is eliminated as a result of committing a fatal violation of O .Candidate (9c) uses gliding but builds an illicit onset.Consequently, it is eliminated by *Cj.Candidate (9d) uses a diphthong /ia/, which is forbidden in the system.Other candidates, namely (9e) and (9f), fatally violate M -V M μ, whereas candidate (9g) violates W --P (W P) due to the lack of a mora in the coda of the fi rst syllable.Finally, candidate (9h) with glide insertion is eliminated by a higher ranked D ( ).
To conclude, the discussion so far has revealed several interesting facts about hiatus resolution in Old English.Firstly, Old English has a dispreference against onsetless syllables, manifested as the constraint O . This constraint has been shown to be the driver for both strategies discussed above: diphthongisation and gliding.Secondly, both strategies show that Old English has a preference for mora preservation, which is technically expressed as the ranking of M μ >> I μ.Thirdly, in the verbal declension pattern of class I weak verbs a strategy of gliding is used due to high-ranked constraints DH, *Cj, and D ( ), which ensure that neither diphthongisation nor glide insertion are applicable to nerian.

Glide insertion
Heavy stems of class II weak verbs instantiate yet another scenario of hiatus resolution in Old English.The relevance of this verbal category to the problem of hiatus in Old English fi rst came to light in Kiparsky and O'Neil (1976), whilst the fi rst OT analysis was presented in Opalińska (2002), upon which is the following discussion based.Let us consider the data in (10).It is interesting to observe that similarly to class I weak verbs, the verbs have trimorphemic structure.They consist of a root, a stem extension suffi x -i-and an infl ectional ending -an.Therefore, structurally they are parallel to the class II light stems.The diff erence lies in the position of /j/.Specifi cally, in [nerjan] /j/ comes from the suffi x /i/.In [loːkijan] however /i/ is still on the surface together with /j/ following it.A question arises whether this /j/ could belong to the underlying form of any of the suffi xes.Such a hypothesis can be rejected with the form [nerjan], where there would be no independent rule in Old English to delete the /i/ suffi x.There would also be no reason for gliding, which would in turn cause complications to the moraic structure.Consequently, it must be assumed that /j/ is inserted in [loːkijan], and that the UR is resultingly //loːk+i+an//.Interestingly, the insertion of /j/ resolves the ia hiatus.
But why glide insertion is preferred in this case over gliding?The same sequence in section 3 was resolved precisely by the latter strategy.The answer needs to be sought in the moraic structure of the stems.For example, in [ner.jan] the stem //ner// is monomoraic in the UR, whereas in [lo:.ki.jan] the stem //loːk// contains two underlying moras due to a long vowel.After syllabifi cation and the application of Weight-by-Position, both syllables [ner] and [loː] have a bimoraic structure.Were lōcian to be realised as *[loːk.jan], the syllable [loːk] would contain three moras, thanks to the additional mora attached to the coda.Such a scenario would be impossible in Old English, which is a language allowing at most a bimoraic structure of syllables (Opalińska 2006, 120).This requirement translates in OT into the constraint *μμμ, as shown in ( 11): (11) *μμμ: No trimoraic syllables.
Consequently, it is a less costly solution to resyllabify /k/ into the following syllable than to allow a trimoraic syllable.For this reason, gliding is forbidden because /k/ needs to be an onset of the following syllable.In this scenario, glide insertion violates a faithfulness constraint against any segment insertion, D ( ), but as argued above it is the least costly violation.If gliding had actually taken place, there would have emerged an onset /kj/, which would in turn violate a highranked ban on Cj clusters, *Cj.
Interestingly, with glide insertion it becomes apparent that there must be 3 levels of constraints.By this we mean that so far in the discussion of gliding and diphthongisation it has been necessary to introduce only 2 levels; hence there has been only 1 dominance relationship.To clarify, diphthongisation is essentially based on the relationship O >> N D , whereas gliding relies on the ranking O , M μ >> I μ.In glide insertion there is a need to introduce yet another level.From the discussion of nerian, it is necessary to have a dominance relationship D ( ) >> I μ in order to stop glide insertion.Conversely, to promote this strategy, there is a need to outrank these constraints with *μμμ.Consequently, the essential ranking for glide insertion is *μμμ >> D ( ) >> I μ.The three levels of constraints have already been shown to be relevant for the constraints on weight: N >> W P >> D μ.This argument will be developed with the help of the new data in the following section.However, it is relevant for the current discussion on glide insertion to revise the position of O in the ranking.Specifi cally, the faithful and the winning candidates incur just a single violation.The former violates O , whereas the latter does D ( ), as shown in the tableau below.If these two constraints were at the same level in the ranking, there would be no way to select the desired output with glide insertion.Consequently, O must be ranked higher than D ( ), thus joining the level of other nonviolable constraints: *μμμ and N .For convenience, a revised ranking of constraints is presented in ( 12) and the evaluation follows in (13) below.The evaluation of lōcian shows that only glide insertion is the optimal strategy of hiatus resolution in class II weak verbs.Gliding is forbidden by a ban on trimoraic syllables, as in candidate (13c), and by the constraint *Cj, as shown in candidate (13d).At the same time, gliding is not suffi cient if the coda lacks a mora, as in candidate (13f), due to the violation of W P. Vowel deletion is also penalised (13e) due to the fatal violation of M -V and mora preservation principle.The faithful candidate (13a) lacks gliding but has an onsetless syllable, for which it is eliminated.Candidate (13b) turns out to fare best in the evaluation.It satisfi es all high ranked constraints and incurs a less costly violation of faithfulness D (S ).Consequently, it is selected as the optimal candidate.
To conclude, the discussion of glide insertion has confi rmed the leading role of O in hiatus resolution.Moreover, it has been shown that, unlike in diphthongisation or gliding, the dominance relationships need to be revised.As a result, there emerged a class of constraints which turn out to be inviolable in outputs.

Contraction
The fourth strategy of hiatus resolution in Old English is contraction.The phenomenon is based on the principle that "a word… is reduced by removal of some internal material, possibly with application of some other phonological processes" (Trask 1996, 92).This is exactly the case of Old English, where descriptive sources show that the deletion of a fricative resulted either in compensatory lengthening (Hogg 2011, 170) or additionally in contraction (Hogg 2011, 179).In OT the analysis of contraction (and of compensatory lengthening) has been presented in the works by Opalińska (2002;2004;2006), upon which the following discussion is based.
She illustrates the mechanism of contraction on the example feohes [fe:os] 'money (gen.sg.)ʼ.In the genitive the word takes the shape feoh [feox].Since the OE genitive morpheme is [es], the output lacks the suffi xal vowel and the velar fricative.Additionally, the diphthong gets lengthened.Consequently, she runs an analysis of both contraction and compensatory lengthening.For the purposes of this article, however, it is suffi cient to focus only on the former strategy.This can be achieved in an analysis of the example lēon [leːon] 'to lend' given by Lass and Anderson (1975).This move is also motivated by technical reasons; namely, only verbal declension can provide arguments to support (or remodel) the established ranking of constraints.
The verb lēon belongs to class I strong declension.Consequently, it needs to have an infl ectional ending -an at some stage.Moreover, the past form in the 1 st person sg. is lāh [laːx].This implies that there might be an underlying fricative in the form [leːon].This claim is further corroborated by Old Saxon and Old High German, where [leːon] corresponds to [li:han].It is also interesting that the verb 'to lend' in the ancestral language, Proto-Germanic, has the form ['liː.xʷɑ.nɑ ].To sum up, the form [leːon] in Old English needs to have an underlying fricative in the stem; specifi cally, this must be a velar fricative, as shown in the past form of [leːon].
At this stage a question arises with regards to the vowel in the stem.In related languages, this vowel is realised as [i:], whereas in Old English it is a diphthong [eːo].Furthermore, the declension pattern in present tense in Old English contains [i:], as shown in [li:ɣe] '(I) lend', [li:xst] '(you) lend'.Lass and Anderson (1975) claim that this is a result of Ablaut, regular vowel variations, as a result of which the form [eːo] in the infi nitive is related to the sound [i:] in the fi rst person singular.They conclude that a common denominator for these sounds is a diphthong [ei].Consequently, the underlying representation of [leːon] takes the shape of //le:ix+an//.At this stage it is possible to descriptively present the diff erences between the underlying and the surface form.Firstly, [leːon] lacks the velar fricative and the suffi xal vowel.Secondly, the quality (but not the quantity) of the vowel is diff erent.These observations can be expressed by means of a set of OT constraints.Let us fi rst consider [x]-deletion.In Old English, voiceless velar fricatives are not found in the onset position.This can be expressed by the constraint *O -, as shown in ( 14). ( 14) *O -: Voiceless velar fricative is disallowed in the onsets.
The constraint in ( 14) follows from a syllable-based generalisation which warrants complementary distribution between /h/ and /x/: the former appears in onsets, the latter in codas 3 (cf.Opalińska 2006, 117).
The deletion of the suffi xal vowel [a] is motivated by the main driver of hiatus resolution: the constraint O .Importantly, both O and *O -promote deletions; hence they need to be counterbalanced by faithfulness constraints, which preserve input segments in the output.O needs to dominate M -V (Input vowels must have output correspondents), whereas *O -must be ranked higher than M -C (Input consonants must have output correspondents).
The last diff erence between the input and the output is the qualitative change of the diphthong.It can be addressed with the constraint D H (DH) from the section on gliding.Specifi cally, the second part of the diphthong will have to adjust in height to [e], thus yielding a diphthong [eo].
It is also interesting to investigate an alternative scenario, where only the fricative is deleted, and the suffi xal vowel joins with a diphthong [e:i] to form a triphthong [e:ia].Such a candidate would not violate O , but it would incur two other serious violations.First, it would violate *μμμ due to the fact that the diphthong [e:i] is already bimoraic, so with an extra vowel the triphthong would contain 3 moras.Second, [e:ia] would incur a double violation of DH, since each vowel is of diff erent height.
To sum up our discussion so far, let us present the cumulative ranking of constraints relevant for contraction and the tableau for lēon.Certain constraints are not included in the tableau because they do not contribute to the evaluation of lēon.For example, since there are no consonantal clusters, *Cj cannot be violated.The same scenario holds for I μ, which cannot be violated in the absence of gliding.Additionally, W P is excluded because it assigns a single violation to all candidates; hence it has no discriminatory force in this evaluation.Before we proceed to the tableau there are still two issues that require discussion.The fi rst one concerns the dominance relationships in the given ranking.The topic has been already started in the discussion of glide insertion.It turns out that all undominated constraints are never violated in any output forms.Interestingly, this feature of non-violability is also shared by two other constraints in this ranking: DH and *Cj.Consequently, they should be by analogy counted as undominated constraints, all the more there is no counterevidence against it.As for the middle level, all remaining constraints are at some point violated in order to help promote a given hiatus resolution strategy.Finally, the lowest level contains constraints which counterbalance the constraints from the highest level.For example, O promotes gliding in nerian, whereas I μ militates against it.The second issue has been raised by a reviewer with respect to a candidate in the analysis of contraction.Specifi cally, the reviewer asked how the ranking of constraints can eliminate a candidate with a glide instead of a velar fricative.Importantly, a potential glide would not be a result of gliding; however, it could be potentially a glide from insertion, as in a candidate [le:ijan].Such a form would be eliminated by DH.A complication emerges in a potential candidate [le:ojan], where none of the discussed constraints are violated.In order to address such a problem, the ranking needs to be supplemented with an extra constraint I -C, which ensures that the velar fricative is preserved in the output 4 .Since [j] is a corresponding sound, it violates I -C.Importantly, the constraint needs to be ranked higher than all M constraints in order to allow the selection of the candidate [le:on].
In the light of the above discussion the ranking of constraints needs to be revised.It is later presented in the evaluation of lēon.The most faithful candidate (17a) incurs a fatal violation *O -, and is thus eliminated.Similarly, candidate (17c) fatally violates a high ranked O .Because no trimoraic syllables are allowed, candidate (17d) loses in the evaluation.Both candidates (17e) and (17f) contain a glide, and thus are eliminated by I -C.The optimal candidate (17b) incurs a double violation of faithfulness but due to the satisfaction of high ranked markedness, it wins in the given ranking of constraints, and is thus selected as the optimal output.
To conclude, the analysis of contraction has revealed additional facts of hiatus resolution strategies.It has turned out that mora preservation principle can be violated in order to satisfy a ban on trimoraic syllables.Interestingly, none of the other strategies can suspend this principle, which shows that contraction is the most demanding strategy of hiatus resolution (Opalińska 2002, 187).

Discussion
Old English is a language that applies multiple hiatus resolution strategies.This article has attempted to off er a unifi ed approach to all 4 strategies: diphthongisation, contraction, gliding, and glide insertion.To this end, the OT model has been employed, predominantly due to its potential to reveal the motivation behind various strategies.Similarly to other OT analyses, we have argued that the main driver for hiatus resolution is O (Opalińska 2002;2004;2006).This fact comes as no surprise, because Old English is a member of the family of Germanic languages, which are known for the dispreference against empty onset position.
This article has confi rmed the scale of preference, proposed by Opalińska (2002), for all 4 strategies of hiatus resolution in Old English: diphthongisation, contraction, gliding, and glide insertion.It turns out that diphthongisation functions as the most optimal strategy to resolve hiatus.The reason lies in the fact that the change takes place at the least cost to faithfulness.No melodic elements are changed but their association to higher levels of representation.Other strategies of hiatus resolution are activated when diphthongisation is blocked by other conditioning.Contraction is chosen in cases where otherwise diphthongisation would lead to the formation of a triphthong (/leːi.xan/→ *[leoan]), a highly marked structure in Old English.Gliding is a more optimal solution when diphthongisation is blocked by phonotactic restrictions.Namely, Old English lacks diphthongs where both constituents vary in value specifi cations for backness (Opalińska 2002, 171).Consequently, the diphthong /ia/ is prohibited (/ne.ri.an/ → *[ne.rian]).Due to the fact that the sequence contains a high vowel, gliding turns out to be a more preferred strategy.As far as glide insertion is concerned, it is also motivated by the restriction not to have /ia/ diphthongs (/loː.ki.an/ → *[loː.kian]).Similarly to gliding, it benefi ts from the presence of a high vowel.However, gliding cannot be used there due to two additional restrictions.First, Old English does not tolerate trimoraic syllables, and such a syllable would be formed if [k] became the coda of the syllable [loː].Second, even if a restriction on 3 moras were obeyed, gliding would yield a sequence [kj], which is an illicit cluster in Old English.Consequently, glide insertion turns out to be the only viable option.
The above conditioning reveals that in terms of optimality diphthongisation is most preferable among all other strategies.Additionally, gliding turns out to be less restricted than glide formation.The place of contraction in this ranking of preference is at the last position, as has been shown by the interaction of constraints.Contraction is the only strategy that can violate mora preservation principle.In the case of other strategies, such a violation would lead to the elimination of candidates.In other words, contraction is used when other strategies are for some reasons inapplicable.Opalińska (20 02, 207) makes a similar observation in that she says that contraction is used when "the remaining [hiatus resolution strategies] are blocked by higher-ranked constraints".Consequently, she implies that contraction is the least preferable strategy.
As shown in section 5, contraction is the only strategy where the violation of M μ does not lead to the elimination of the candidate.This argument is mute in the works by Opalińska.Most possibly the reason is her choice of analysed examples.She chose only words where contraction is accompanied by compensatory lengthening.In other words, in such examples mora preservation principle is always obeyed: a point already raised by Hayes (1989).
Despite convergent conclusions, the analysis presented in this article diverges in certain aspects from that of Opalińska.There are diff erences with respect to the ranking of constraints.For example, we argue that the position of W P must be at the level of M constraints, partly due to the fact that it must be ranked lower than N F , an inviolable constraint.However, Opalińska (2002, 177) places W P unranked with respect to O , another inviolable constraint, thus elevating the position of W P. Another example of a diff erence in the ranking concerns the constraint D ( ).We argue that this constraint should be parallel to the M family.However, Opalińska (2002, 185) claims that its position in the ranking should be lower than M .Interestingly, in both cases of W P and D ( ), she manages to select the same outputs as in our analysis.This is because the same constraints but reranked would yield identical results.Such a scenario seems to be coincidental, all the more that ranking arguments come to light in an analysis of other examples that apply diff erent strategies, as we have shown in our article.
Probably, one of the reasons for these ranking diff erences results from the order in which diff erent hiatus resolution strategies were discussed.In this article, we have developed the discussion from the most preferable strategy: diphthongisation to the least preferable one: contraction.Opalińska (2002), however, started the discussion with contraction, later moved to diphthongisation, gliding and glide insertion.Consequently, the development of ranking arguments could diff er between our analyses.Another reason for the diff erences may be also sought in the style of the analysis.Specifi cally, with each strategy in this article there was building on the previously established ranking of constraints.A disadvantage of this method is that the complexity of the analysis is increased; however, it is possible to notice more generalisations.Opalińska (2002) does not apply many constraints in a single evaluation, thus achieving clarity of exposition.However, certain ranking arguments may remain mute.
One of the main goals of this article was to establish the cumulative ranking of constraints, which would technically illustrate the scale of preference for hiatus resolution strategies.For ease of reference, let us present the ranking below in (18).The ranking above contains constraints which function as conditions for the activation of a given hiatus resolution strategies.The core constraint which drives any hiatus resolution is O .Interestingly, the lower a given strategy is on the scale of preference, the more conditions it needs to fulfi l.This is illustrated in ( 19).The list above shows that the ranking of strategies is established on the basis of the subset of constraints activating a given strategy.Diphthongisation requires only one out of high-ranked constraints.When it is inapplicable due to some phonotactic restrictions (DH) or syllabifi cation parameters (*Cj), gliding is chosen as the next strategy.However, when gliding would lead to a trimoraic syllable, it is preferable to opt for glide insertion.The last strategy on the scale is contraction.
It additionally needs a ban on velar fricatives in the onset.
To conclude the discussion, the ranking of preference for hiatus resolution strategies takes the following shape: diphthongisation >> gliding >> glide insertion >> contraction.The ranking in this analysis is identical to the one presented in Opalińska (2002).However, the diff erence lies in the argumentation for the ranking.Specifi cally, Opalińska (2002) does not provide the cumulative ranking of constraints, and resultantly does not discuss the subsets of constraints.Consequently, the articulation of these arguments may cast new light on resolving hiatus in Old English data.