SPANISH ‘ DEPALATALIZATION ’ : THE SYNCHRONIC , DIACHRONIC AND PERCEPTION PERSPECTIVES

Spanish has a restriction on palatal nasals and laterals in the coda so that they are normally realized as dental or alveolar coronals. In the onset position, the palatal point of articulation is retained, bello (be.[ʎ]o) ‘beautiful-masc.’, beldad (be[l].dad) ‘beauty’; doña (do.[ɲ]a) ‘Madam’, don (do[n]) ‘Mister’. Alternations such as these led phonologists to propose a rule of depalatalization that turns an underlying palatal nasal or lateral into a coronal (Contreras 1977; Harris 1983). More recently, within an optimality-theoretic framework and on the basis of loan word evidence, Lloret and Mascaró (2006) argue again in favor of an active process of depalatalization in the phonology of Modern Spanish. Taking the proposal of Lloret and Mascaró as its point of departure, this paper expands the discussion on depalatalization to consider diachronic data, the role of the underlying representation and the perception grammar, in particular regarding loanword phonology. Historical data support depalatalization as an active phenomenon in Old Spanish and Medieval Spanish; yet the morphophonological alternations cannot be considered active/productive synchronically. Unlike previous serial models of phonology, an OT framework allows for the incorporation of diachronic data into the analysis, through constraints, constraint re-ranking and Lexicon Optimization, thus explaining how the current situation came about and shedding light on apparent synchronic alternations. OT also provides a formalization of the role of the underlying representation in diachronic change and in synchronic loanword evidence, in connection with the perception grammar, thus providing support for depalatalization as an active phonotactic restriction in Spanish.


Introduction
The existence, or not, of a process of depalatalization of nasals and laterals in Spanish has been long debated in phonology (Contreras, 1977;Harris, 1983;Pensado, 1997;Harris, 1999;Bermúdez-Otero 2006;Lloret & Mascaró, 2006;Eddington 2012).Harris, who originally favored the idea of depalatalization in Spanish, admittedly argued against such a process in his latter work.Harris (1999) set to debunk the depalatalization process in Spanish, and even attempted to discourage future analyses on the topic, concluding with this foreboding statement, "May Harris (1983) and the specter of Nasal Depalatalization not come back to haunt us again" (Harris, 1999, p. 64).However, the discussion did not end with this remark.Lloret & Mascaró (2006) provide an analysis under Optimality Theory that argues in favor of depalatalization.Starting from Lloret & Mascaró's analysis (2006), this paper expands the analysis to consider diachronic data and the role of the underlying representation in language change.It also delves deeper into the loanword evidence presented by Lloret & Mascaró by considering the perception grammar.Historical data supports the view that depalatalization was active in Old Spanish and Medieval Spanish; yet, contra Lloret & Mascaró and in agreement with Bermúdez-Otero (2006) and Eddington (2012), it is argued that the morphophonological alternations cannot be considered active synchronically.Nonetheless, loanword evidence shows that a phonotactic restriction depalatalizes any palatal sonorant in the acoustic input.No morphological alternations exist because of the lack of patrimonial inputs to which depalatalization can apply.An optimality theoretic framework allows for the incorporation of diachronic and synchronic evidence, including loan words and perception data into the same analysis, thus offering a broader view of the problem.It is shown that restricting the analysis to synchronic data, as has been done by previous studies, limits our ability to fully understand the complex phenomenon known as depalatalization.

The data
Before discussing the previous studies, it is important to consider the distribution of the data.Example (1) shows that palatal nasals and laterals in the onset (a) surface as coronals in word-final position in patrimonial Spanish words.
(1) Depalatalization of [ʎ] and [ɲ] in word-final position in patrimonial words a. Onset [ʎ] Coda [l] ello 'that' el 'he' aquello 'that-pro' aquel 'that-adj' doncella 'fem. virgin' doncel 'male virgin' clavellina 'carnation-like plant' clavel 'carnation' b. Onset [ɲ] Coda [n] doña ' Madam' don 'Mister' desdeñar 'disdain (v)' desdén 'disdain (n)' (Pensado, 1997, p. 595-596;Lloret & Mascaró 2006, p. 77-78) The examples in (2) show that the ban on palatals is not limited to word-final position, but takes place also in the coda position in which it adopts the point of articulation of the following consonant, rather than a default one (i.e.alveolar).1Each term listed in (2a) shows a palatal consonant in onset position, while (2b) contains a morphologically related form with the lateral in the coda position.In preconsonantal coda position (vs.absolute word-final position) palatals share the point of articulation of the following consonant as a way to license point of articulation, rather than adopting a default coronal point of articulation.This allows for the complete articulation of the lateral, while also avoiding a palatal one in the coda.
(2) Depalatalization of [ʎ] and [ɲ]  tinte 'tint (n)' (Pensado, 1997, p. 595-596;Lloret & Mascaró 2006, p. 77-78) Additionally, the process is not limited to patrimonial words, but also affects loans, as can be seen in the examples in (3) which keep an orthographic spelling similar to the language from which there are adapted, but are produced with an alveolar consonant.
(3)  (Lloret & Mascaró 2006, p. 79) The examples in (3) demonstrate that palatal nasals and laterals are not permitted in the coda in Spanish.The palatal/alveolar alternation seen in ( 1) and (2) seems to indicate that at one point the words went through a process of depalatalization.The contrast between bello 'beautiful' and beldad 'beauty', for instance, shows that when in the coda, the palatal [ʎ] is realized as alveolar [l].Whether or not the connection can be clearly made between morphologically related words is also a matter of debate (see Harris, 1999 andLloret &Mascaró, 2006 for differing opinions).

Previous studies
As previously mentioned, the existence of a depalatalization process as the explanation for the data in section 1.1 has been a matter of debate in Spanish phonology for more than three decades with prominent studies in favor (Contreras 1977, Harris 1983, Lloret & Mascaró 2006) and against (Pensado 1977, Harris 1999).Contreras (1977) argues that the data in ( 1)-( 3) can be explained through a series of depalatalization rules whereby a palatal nasal or lateral is converted into an alveolar at the end of a word or when followed by a consonant (4a) and (4b).Following the analysis of Contreras, Harris (1983) also argues in favor of a depalatalization rule.He incorporates syllabic conditions (i.e., the syllable rhyme) in the context of the rule in (5).
(5) Nasal and lateral depalatalization (Harris 1983 The first to argue against a depalatalization rule was Pensado (1997).She conducted a psycholinguistic study on the realization of palatal nasals and laterals in Spanish in which subjects were given a series of nonce words and asked to derive morphologically related forms.Subjects were presented with a verb or a noun (e.g., enapillar, enapil) and were asked to form the plural by adding -es, an adjective by adding the suffix -oso, and a past participle by adding -ado.Pensado concluded that the results of the study demonstrate that there is no process of depalatalization in modern day Spanish because there are enough cases where the palatal in the base is not depalatalized in the derived nonce form.She explains that while depalatalization may be "surface true", appearing in isolated nonce words and being influenced by previous word forms, it is not morpho-phonologically true (p.605).However, when considered carefully, the results do not necessarily go against the idea of depalatalization.When given the verb form of the nonce word enapillar followed by the noun form enapil, subjects were divided on whether the adjective formed by -oso should be enapilloso (which received 14 responses) or enapiloso (which received 17) (601).Furthermore, it appears that the responses were influenced by the last form of the nonce word participants were given.For instance, if the last form presented was enapilar, the adjective tended to be enapiloso, but if it was enapillar the most common answer was enapilloso.Another interesting finding was that when participants were asked to create a word for tratar de don 'address as sir', doñear was chosen by twelve participants, a possible indication that for those participants there is still a clear association between don and doña.
In the wake of the evidence provided by Pensado, Harris (1999) sets out to debunk his original depalatalization rule.His evidence comes from the fact that, according to him, desdén 'disdain (n.sgl.)', desdenes 'disdains (n.pl.)', and desdeñes 'disdains (v.2nd sing.)' is the only good example of depalatalization in Spanish.The other clear examples are also what he denotes as "xenonyms" and claims they cannot be part of the analysis.He defines "xenonyms" as loan words (from Latin or other languages) and argues that they must not be included in an analysis because they vary from speaker to speaker and "come and go at the whim of popular culture" (p.57).Once the vast majority of modern day Spanish examples (the entire set found in example (3) above) have been removed for being xenonyms, Harris contends that depalatalization is not an active process.He explains that desdén/desdenes both have an underlying /n/.Examination of the word classes of Spanish reveals that words ending in ill-formed coda consonant or consonant cluster are followed by epenthetic -e (e.g., nub-e 'cloud', part-e 'part').Therefore, Harris concludes that if the stem-final nasal were palatal, the stem would have an epenthetic -e, in which case, the singular should be desdeñe and plural desdeñes.Since this is not the case, Harris argues that there is no depalatalization rule at work.In order to account for the verb form desdeñes, Harris invokes readjustment rules, which he explains "are called upon when a few vocabulary items have similar but phonologically distinct underlying forms the differences among which cannot be attributed to independently motivated phonological rules" (p.63).His rule for desdeñes can be seen in ( 6) below.
Within a generative framework, Bermúdez-Otero (2006) also argues that the alternations above are "lexically listed rather than synchronically derived" (288) and sees depalatalization as a principle of allomorph selection in which there is no common underlier for the allomorphs.This is also the position taken by Eddington (2012) who contends that palatalization is not productive word-internally.
The latest in the series of fully-fledged analyses of depalatalization in Spanish comes from Lloret & Mascaró (2006) who provide a counter argument to Haris (1999) and argue for depalatalization.They base their argument on the fact that /ʎ/ and /ɲ/ are absolutely not allowed in coda or word final position in Spanish, and that the treatment of loan words (to follow the pattern of depalatalization) is enough evidence to suggest that depalatalization is an active process in Spanish.
Lloret & Mascaró proceed to provide an Optimality Theory account that is based on the constraints presented in Baković ( 2001) on neutralization of nasals in Spanish.The key constraints presented are explained in ( 7).On the basis of Kager (1999), Lloret & Mascaró define the base as "a freestanding output form --a word", that "...contains a subset of the grammatical features of the derived form" (p.87).We further define the base as a fully-inflected, independent, morphological word to which no inflectional or derivational morphemes can be added other than those shown to attach at the word level (i.e.plural and adverbial -mente).In other words, the base is the output of the lexical phonology (in serial terms).
Through the constraints listed in (7) Lloret & Mascaró account for the data presented in (1)-(3).Their final constraint hierarchy is: ID-BA (PL) >> ID-ONS(PL) >> *-COR >> ID-SON(PL).An example of how this constraint hierarchy functions is provided in (8).In (8a) an input with a palatal surfaces as coronal (i.e., depalatalized) because the candidate with a coronal nasal is better than the one with the palatal, since the latter contains a violation of the higher-ranked constraint against palatals (*-COR).ID-BA (PL) is vacuously satisfied because there is no base.In (8b), the top ranked constraint ID-BA (PL), which seeks identity in place with the singular base, prevents the palatal nasal from surfacing in the plural form of the noun, even though the palatal would otherwise be allowed in the onset of a syllable.In (8c), since desdeñes (V) does not have desdén as its base, ID-BA (PL) is vacuously satisfied and the form with the palatal is selected as the winner.It must be noted that an input with a palatal nasal needs to be considered, since in OT no restrictions can be imposed on the form of the input, according to the postulate known as Richness of the Base (McCarthy 2002).Positing an input with a coronal would of course still result in the selection of the same output (cf.8a-b and 14 below).
Additional clarification regarding the nature of the base and derivational morphology may be useful here.A denominal verb form such as desdeñes does not have a base with a palatal (i.e. it is not derived from a base with a palatal).Although it seems possible that desdén could be posited as the base for a denominal verb such as desdeñ-ar/*desden-ar, denominal verbs are not formed on independent words, but on a derivational base without any class markers or suffixes (i.e., forr-o 'lining' forr-ar 'to cover with lining'; nata 'cream', des-nat-ar 'to remove the cream'3 ).Desdén is homophonous with a complete word because it is a class three, consonant-final noun (Harris 1999), thus it needs no wordmarker/derivational suffix added to the base to form a complete word; yet, the morphological constituent/level to which -ar is attached is not a word, but a derivational base.The same can be said of morphophonological alternations, such as mil, millar, piel, pelar, etc. and therefore ID-BA is satisfied vacuously.
In section 3 we will take the optimality-theoretical proposal and the constraints in Lloret & Mascaró (2006), as defined here, as a point of departure for the diachronic and synchronic analysis proposed.

The historical data
In this section, historical data of the distribution of nasal and lateral palatal consonants is revisited.The data is divided up into two main periods of time, Old Spanish and Medieval Spanish.

Old Spanish
In Old Spanish there was an alternation between palatals in onset position, and coronals in the coda, as seen in ( 9). ( 9 The palatal would also surface across words: when a word-final palatal was followed by a vowel-initial world, it was resyllabified in the onset, giving such examples as mill omnes (mi[ʎom]nes) (Penny, 2002, p. 82).However, when the following word began with a consonant, the palatal was depalatalized, e.g., mil cavallos (mi[l-k]avallos) (Penny, 2002, p. 82).
The alternations in (9), and most certainly, the patterns of resyllabification explained above, suggest that in Old Spanish depalatalization was an active phonological process.The alternations in the surface form of singular and plural forms of the noun will be revisited in section 3.

Medieval Spanish
The pattern in Old Spanish gives way in Medieval Spanish to a more generalized depalatalized system: instead of an alternation between palatal and alveolar forms, the singular and plural forms of a noun always agreed in point of articulation, as does a word-final palatal when resyllabified into the onset of a following word.Example (10) (Penny, 2002, p.82)It is important to note here that, while the singular and plural form of the noun always agree in point of articulation (both contain a palatal or alveolar), there is disagreement in which allophone was selected.In cases like piel --pieles the alveolar always surfaced; however, in cases like valle --valles the palatal always surfaced.This is an important piece of evidence that further suggests a depalatalization process.If the singular form was taken to include the -e like in valle, then the palatal consonant was allowed to remain, and did so in both singular and plural forms.However, if the singular was formed without the -e, then the resulting form became alveolar and this was then generalized to the plural as well.
In sum, the evidence provided in this section supports the view that depalatalization was an active phonological process in Old Spanish exhibiting a surface alternation between a palatal and an alveolar allophone within and across the word (i.e., singular/plural and across-the-word resyllabification).The data from Medieval Spanish mimics modern day Spanish, in that the allomorphy between the singular and plural form and in across-the-word resyllabification was levelled in favor of the non-palatal.In section 3 we show that, despite the lack of alternation in patrimonial forms, depalatalization continues to be an active process as revealed by the data and the optimality-theoretic analysis.

OT analysis 3.1 Diachronic analysis in OT
For Optimality Theory, the history of a language consists of a diachronic series of synchronic grammars and language change involves a series of re-rankings of the constraint hierarchy (Holt 2003(Holt , 2006)).Any given sound change can be described as the promotion or demotion of one or more constraints.Once these constraints are re-ranked, the surface form (the output in OT) of a word can also change.Given an absence of surface alternations, speakers may then posit a different underlying form (the input in OT) through a process of Lexicon Optimization which selects as the input the form that most closely resembles the output.With change in the underlying form, additional re-ordering of constraints may take place.Newer listeners are not aware of the changes that were made in the hierarchy; they only learn the new one.
An optimality-theoretic view of language change offers new insights and justification for considering historical data.If there is evidence that depalatalization was an active process at one point in the history of the Spanish language, an account of how the constraints and constraint ranking evolved into the current hierarchy of constraints in present day Spanish may shed light on the current process.

Old Spanish
The analyses contained in this section builds off the constraints provided by Lloret & Mascaró (2006) (cf.(7) for definitions) for Modern Spanish.As was seen in section 2, the data from Old Spanish suggests that depalatalization happened regularly (i.e. the vacillation between singular pie[l] and plural pie[ʎ]es forms), and that the constraint ID-BA(PL) was violated, allowing for palatals in the onset of plural forms and in prevocalic word-final position, and alveolars in the coda and preconsonantal word-final position.This suggests that, as it will be shown in (12), in Old Spanish ID-BA(PL)-the antiallomorphy constraint-was lower in the constraint hierarchy than ID-ONS(PL), the constraint responsible for retaining the palatal in the onset.We propose that the constraint ranking responsible for depalatalization in Old Spanish was ID-ONS(PL) >> *-COR >> ID-SON(PL).The singular form of piel is shown in (11).ID-BA(PL) is irrelevant at this point.
(11) Word-final depalatalization in Old Spanish (singular) The underlying representation of [piel] would have been /pieʎ/, as input /piel/ would make the wrong prediction for the plural (*[pieles]).The resulting output in (11) [piel] exhibits depalatalization due to the need to satisfy the high ranking of the *-COR constraint.Due to the fact that ID-BA(PL) is ranked lower in the hierarchy than ID-ONS(PL) (i.e, dominated by), the plural form seen in ( 12) does not depalatalize.Note that although ID-BA(PL) is shown at the bottom of the hierarchy in ( 12), it could also be ranked above *-COR or in between *-COR and ID-SON(PL), as its ranking is lower than ID-ONS(PL), but undetermined with respect to the other three constraints.
(12) No depalatalization in the plural in Old Spanish: antiallomorphy constraint low ranked

Medieval Spanish
In section 2 it was shown that the allomorphy in singular and plural forms was lost in Medieval Spanish, where the singular and the plural form both began to match in terms of place of the nasal or lateral.This leads to the conclusion that the constraint ID-BA(PL) moved up in the constraint hierarchy to dominate ID-ONS(PL), forming the new hierarchy of ID-BA(PL) >> ID-ONS(PL) >> *-COR >> ID(Son-PL).Example (13) shows that the singular form of /piel/ was unaffected at this stage.The high ranking of the ID-BA(PL) constraint caused the depalatalized form [pie-les] to surface in order to match the point of articulation of the singular form, which provided the base for the plural form.This new output of [pieles] entailed a change in the input in the new generations who would not have heard the palatal lateral from other speakers, through a process of Lexicon Optimization.Lexicon Optimization dictates that the learner will eventually opt for the underlying representation that most closely resembles the output (Prince & Smolensky 1993Inkelas 1995, Jarosz 2006, Krämer 2012, Tesar 2013).In this case, due to the reranking of ID-BA(PL), learners would have no evidence of a palatal lateral, because it did not surface in the output, and would therefore posit an alveolar in the underlying representation, thus restructuring it from a palatal to an alveolar.This change in the input can be seen ( 15).As explained in section 2.2, the -e in some plural forms like calles was considered part of the base of the singular noun form by speakers of Medieval Spanish when creating the new correspondence relationship between singular and plural forms of nouns.Candidate evaluation for the singular and plural of calle can be seen in ( 16a) and (16b) respectively.Deletion of input /e/ is ruled out by the high-ranking of the anti-deletion constraint MAX.The current account explains the alternations that gave rise to the controversy previously outlined in (1) as morphologically related, lexically listed forms that do not share a common input (e.g.él 'he ', ello 'that'; desdén 'disdain (n) ', desdenes 'disdains (n) ', desdeñes '(you) disdain, pres. subj.'), in agreement with previous proposals such as Bermúdez-Otero (2006).The base identity constraint is vacuosly satisfied in these forms, since there is no base-output correspondence, i.e., the verb desdeñes is not the base for the noun desdén (8c), contrary to what happens with piel and pieles.Desdén and desdeñes have two separate, independent inputs, namely /desDen/ /desDeɲes/; in other words, /desDen/ is not the base for desdeñes.

Synchronic depalatalization
After discussing Spanish depalatalization in Old Spanish and Medieval Spanish, this section revisits synchronic depalatalization.As seen in sections 3.2 and 3.3, depalatalization results in a change in the underlying representation in the subsequent generation.This means that once a form is depalatalized, the subsequent generations have no phonological intuition that the word used to consist of a palatal, and conceptually only interpret the alveolar form.As mentioned above, this also explains the discrepancy seen between the desdén -desdenes noun forms and the desdeñes verb form.The noun forms underwent the change from Old Spanish where desdén -desdeñes (n.sg.-n.pl.) would have existed as noun forms, to the new desdén -desdenes (n.sg.-n.pl.) in Medieval Spanish.Since the verb does not relate to the singular noun form in the same way that the plural noun form does, the ID-BA(PL) constraint did not apply, allowing for the alternation between desdenes (n.pl.) and desdeñes (v.).
Along the same lines, newly borrowed and foreign words should follow a similar pattern.This is borne out by the examples in (3) (i.e.Sabade[ʎ] > Sabade [l]) where loan words that end in a palatal nasal or lateral that are adapted to Spanish are depalatalized.Using Harris's argument in reverse, we can postulate that if depalatalization did not exist, loan words would be adapted using an epenthetic -e; however, this is not the case.In fact, there is solid evidence to support the view that depalatalization is one of several repair mechanisms employed by Modern Spanish to avoid illegal consonants and consonant clusters and that -e epenthesis is no longer productive in word-final position in Spanish (Colina 2003, Bonet 2006).Colina (2003Colina ( , 2009) ) and Bonet (2006), among others, explain that in earlier stages of the history of Spanish, an ill-formed consonant or consonant cluster was repaired through epenthesis (e.g., bote 'boat', nube 'cloud', parte 'part'), indicating that final epenthesis was active.In Modern Spanish, however, epenthesis is no longer an active process and final -e has been restructured as underlying.The current repair mechanism for parsing unsyllabifiable word-final consonants or consonant clusters is deletion or neutralization of various coda features (place of articulation for sonorants and continuancy and voice for obstruents, cf.Colina 2009), as seen in ( 17) (see Bonet (2006)  In order to better understand how loan words ending in a palatal nasal or lateral are incorporated into Spanish today, the hierarchy discussed in 3.3 can be applied.To illustrate the process, consider the word Sabadell [sə.bə.déʎ] as it is pronounced by a Catalan speaker.The way in which the Spanish speaker goes from the auditory cues to an input that is then used for production is not a trivial matter (related to the formalization of the processing grammar).For now, let us posit that there exist in principle two possible inputs for the loan word Sabadell, one with a palatal lateral and one with an alveolar.If the input consists of the palatal lateral in word-final position /sa-ba-deʎ/, the Spanish speaker would produce the depalatalized alveolar [l], showing the effects of the *-COR constraint.This stage is represented in (18).The singular depalatalized form becomes the base for the plural.ID-BA(PL), being high in the hierarchy, forces the plural form to agree with singular form in terms of palatalization, (19).Lexicon Optimization would then change the underlying form for speakers that hear the loan word through other speakers of Spanish, and for subsequent generations.This distinction is an important part of the argument here.The newly optimized form for the singular can be seen in ( 20) and the plural in ( 21).
(20) Depalatalization in loan words with an alveolar input: Singular forms (21) Depalatalization in loan words with an alveolar input: Plural forms As mentioned above, an alternative input to consider is one with an alveolar (rather than a palatal), i.e., /saBaDel/.Positing /saBaDel/ as the input has no bearing on the selection of the output, which would still contain an alveolar lateral in the singular and in the plural.Evaluation would proceed as in (20-21) without Lexicon Optimization.
While the analysis proposed works regardless of the form of the input, input selection is not a trivial matter and it has not been addressed in existing analyses of depalatalization.Positing a palatal input is problematic from the point of view of the learner because of the futility of selecting an input (18-19) that will be immediately changed (i.e, 'optimized') (20-21) on the grounds of lack of surface evidence.On the other hand, an alveolar input requires an explanation as to how the Spanish speaker sets up an alveolar lateral as the input for a palatal form, which requires a more developed analysis of the perception component of depalatalization.
Following Boersma and Hamann (2009), we argue that the underlying representation used for production is related to perception, which is in turn formalized as a bidirectional optimality-theoretic grammar (Figure 1).Boersma and Hamann see loanword adaptation as a mapping of sensory data to an abstract mental representation.In a parallel fashion to production, in which the grammar is the result of the interaction between faithfulness and structural (markedness) constraints, perception is described as the interaction between structural and cue constraints.Cue constraints evaluate the relation between the input of perception (the auditory/phonetic form) and the output of the perception process (the phonological form that represents what is heard), while structural constraints evaluate only the output of the phonological process.Cue constraints play a similar role to that of faithfulness constraints in production in that they evaluate the relation between the input and output of perception.Applying the model above to the current analysis, the relevant cue constraint is one that states that "the auditory cues for a palatal should not be heard as alveolar" (*[palatal]/+COR/).Spanish has a coronal/palatal contrast in the onset, but not in the coda, thus the structural constraint against palatal sonorants in the coda (*-COR, cf.7d) would dominate the cue constraint (*[palatal]/+COR/).In the case of an onset palatal, Lloret, the constraint *-COR, which only affects coda sonorants in the coda, is vacuously satisfied and thus a palatal lateral can surface in the onset, [ʎ]oret.5 Selection of the optimal phonological output of perception from the percept [səBəDeʎ] proceeds as shown in ( 22).To summarize, depalatalization is shown to be active in Spanish as a phonological restriction in the adaptation of loan words.The analysis of the historical developments presented reveals that due to input restructuring and constraint reranking, there exist no native inputs that depalatalization can apply to and therefore there are no active morphophonological alternations in patrimonial words.The high ranking of ID-BA(PL) has the same effect for singular/plural alternations in borrowed words (23)7 .New words ending in a palatal nasal or lateral consonant will follow the pattern outlined in (22)(23).

Conclusion
This article argues in favor of depalatalization in Spanish by incorporating into the analysis data from the history of Spanish and loan word adaptation, including a formalization of perception data.Depalatalization is analyzed in terms of a phonotactic restriction on coda palatals and of Base Identity in the production and perception grammar (pro Lloret & Mascaró 2006;contra Pensado 1997, Harris 1999, Bermúdez-Otero 2006and Eddington 2012).In the perception grammar, a cue constraint is also relevant.Depalatalization has been active since Old Spanish when palatals were only depalatalized in the coda or word-final position, except before a word beginning in a vowel, in which case the palatal would be resyllabified to the onset of the following syllable and maintained.A change occurred in the ranking of the constraints affecting depalatalization in Medieval Spanish, where ID-BA(PL), rose in the hierarchy.Consequently, all allomorphs agree in place of articulation.In the case of words like piel, the alveolar form became the base and singular and plural forms both were realized as alveolar.The resyllabified consonant was also realized as an alveolar.Another outcome was the maintenance of the palatal, which can be found in words like valle where the base form included the -e and therefore allowed the palatal to be maintained.The underlying representations of these words were restructured by subsequent generations, given that these generations no longer had the phonological knowledge that the consonant was at one point a palatal.As a result, there are no morphophonological alternations involving palatal vs. alveolar sonorants (yet there is an active process of depalatalization in the coda).Harris (1999) was correct: there are no input palatals in the native lexicon to which depalatalization can apply because if there were, they would have been followed by -e, like calle.In other words, forms like desdén (n) and desdeñes (v.) are morphophonogically related, but listed separately in the lexicon (i.e. are not derived from a common input/base) and thus Base Identity has no relevance for them (contra Pensado 1997, Harris 1999, Bermúdez-Otero 2006, Lloret and Mascaró 2006and Eddington 2012).

( 7 )
Main constraints in Lloret & Mascaró (2006) a. Identity-Base(Place) (ID-BA (PL)): The place of articulation of the base form should correspond with the place of articulation of the affixed form.b.Identity-Onset(Place) (ID-ONS(PL)): The place of articulation of an onset in the input should match the place of articulation of this onset in the output.c.Identity-Sonorant(Place) (ID-SON(PL)): The place of articulation of a sonorant in the input should match the place of articulation of this sonorant in the output.d. *-COR: based on hierarchy of place markedness, *Labial, *Dorsal >> *Coronal; avoid non-coronal consonants in the coda.

( 13 )
Depalatalization in Medieval Spanish: Highly-ranked antiallomorphy constraint On the other hand, the plural form of piel was affected by the change in hierarchy found in Medieval Spanish.Example (14) shows this change.

( 22 )
Spanish perception of Catalan palatal sonorants (nasal or lateral) 22), the final consonant in Sabadell is heard as alveolar and the output of the phonology selected by the perception grammar is (b) /saBaDel/./saBaDel/ then becomes the input to the production module (23).(23)Loan adaptation: Output of perception is input for production a shows this new distribution.
for additional examples).