Skip to navigation – Site map

HomeIssues23PapersPhonological templates and the le...


Phonological templates and the lexicon

Semra Baturay-Meral


In past phonology literature, diacritics, brackets and other extra-phonological objects have been employed to identify morpheme boundaries and to differentiate words from affixes. In the present study, we argue that all of these extra-phonological items bring arbitrariness to phonological theory since phonology is only concerned with identifying phonological objects. In this respect, the present study proposes a new account in order to identify word boundaries in phonology and to explain phonological processes which show sensitivity to morphological boundaries without referring to any extra-phonological objects. Accordingly, we argue for a novel template model and propose that bases (e.g. words), productive suffixes and prefixes are listed in the lexicon with their own unique phonological templates, ONO, NO and ON, respectively. This means that their morphological categories are recognizable when they come to phonology. Also, the morphological boundness of prefixes and suffixes is visible in our template model: The absence of a final onset in prefixes (ON) and the absence of an initial onset in suffixes (NO) render them phonologically bound to a base because only bases exhibit an (O...O) structure. Accordingly, these morphemes come to phonology with their own templates, and phonological operations (government, licensing, etc.) apply to them when necessary. Also, we put forward two new parameters, the Final Onset Parameter and the Initial Onset Parameter, in order to explain the word-final/word-initial differences among languages. We argue that phonological processes, the phonology-morphology interface and their relation to the lexicon are non-arbitrarily explainable in our model thanks to the templates and novel parameters.

Top of page

Full text

I sincerely thank the editor and the anonymous reviewers for taking the time to review the manuscript and for providing constructive feedback to improve it.


  • 1 The present study has been developed from Baturay-Meral [2020].

1There are certain phonological processes observed in languages, such as devoicing and vowel harmony, which show sensitivity to morphological boundaries (e.g. word boundaries), i.e. devoicing tends to appear in word-final position, while vowel harmony respects word boundaries and cannot jump into the next word. Previous phonological studies have tried to explain those phonological processes and their relation to boundaries with the help of extra-phonological objects, rules and/or levels/stages. In this respect, various mechanisms have been developed to differentiate words and affixes from each other and to give an account of their derivations, i.e. diacritics (+, #) (SPE (The Sound Pattern of English): Chomsky & Halle [1968]); levels (Lexical Phonology: Booij & Rubach [1987], Kaisse & Shaw [1985], Kiparsky [1982], Mohanan [1986]); and brackets (Government Phonology: Kaye [1995], Kaye, Lowenstamm & Vergnaud [1990]). However, the presence of extra-phonological objects and mechanisms implies arbitrariness in phonological theory since phonology can interpret only phonological ones (Scheer [2012]). In contrast to previous approaches, the present study1 proposes an analysis by formulating a new model of constituent structure for phonological theory. This model aims to account for phonological processes without extra-phonological objects. The model is based on the templates, in which a base (a form (e.g. a word) to which affixes of any kind can be attached (Bauer [1983: 20])) is recognizable by its unique constituent structure (1b) and is distinguishable from a suffix (1c) and/or a prefix (1a), each of which also has its own specific constituent structure in phonology:

Image 10000201000001F00000009019EC6D2C0C990535.png

2As a novel attempt, we argue that the base template always begins with and ends in an O(nset), as given in (2a). The suffix templates, on the other hand, must begin with an N(ucleus) and end in an O(nset) (2b), while the prefix templates are to begin with an O(nset) and end in an N(ucleus) (2c):

Image 10000201000001920000006B6CDFF2FCC0F57468.png

3According to the model, the labels (base, prefix, suffix) are deemed redundant because the morphological information they convey is already inherent within the constituent structure. In the model, the proposal is that the bases and productive affixes are all stored in the lexicon with their own templates and come to phonology with these ones.

4The structure suggests that a prefix (ON) typically affixes to the left side of the base (ONO), while a suffix (NO) attaches to the right, maintaining the phonological sequence in the ON...O format, as seen in (3):

Image 10000201000000F3000000B97AE22FDBBB3F4829.png

5This is because a combination of prefix+base, base+suffix, or prefix+base+suffix ultimately forms a base in the ON...O pattern. Besides, phonology can identify whether an adjacent nucleus-onset belongs to the same base or is a part of an affix since this information is also encapsulated in the templates: each template specifies its own constituents and is stored in the lexicon with this information. Note that nothing more can be transferred to phonology other than this information in the framework of our New Template Model.

6Within the model, we adopt the assertions that (i) every nucleus licenses its onset (in ON-pair) as proposed by Harris [1994], and (ii) a licensed constituent (onset) is likely to support more melodic material (Scheer [2004]). Consider the base template in (4):

Image 100000000000012900000060C4E50D28508790EC.png

7As given in (4), the base templates do not end in a nucleus but in an O, which means that in our model, every onset position is linked with a nucleus that acts as a licensor, except for the final position (Ox). As a result, we deduce and propose that in certain languages (e.g., Italian, Vata, Zulu, etc.), consonants might be absent in the final onset position due to their weak nature, attributed to the lack of licensing. On the other hand, the final O may be melodically realized in other languages (English, French, Malayalam, Polish, Turkish, etc.) despite the lack of licensing by a nucleus. Compare (5a–b):

Image 10000000000001E700000061954E05D5BF94F0C3.png

  • 2 In the present study, the unlicensed onset means the onset which is not licensed by any nucleus (e. (...)

8In (5a), no melodic material can be present in O3 because unlicensed onsets cannot contain any melodic content in Italian. Conversely, in Turkish (5b), a consonant can appear in the base-final position (O3) without being licensed. Therefore, it is evident that certain languages (Italian, Vata, Zulu) exhibit intolerance towards the occurrence of a consonant in an unlicensed onset position, whereas others (Turkish, Polish, etc.) do not2. Within the model, we account for this variation among languages by introducing the Final Onset Parameter (6):

Image 10000201000001E600000097C38AAE3041AF41DA.png

9It is important to note that setting the Final Onset Parameter OFF does not necessarily imply the occurrence of any consonants in the base-final position of these languages. According to the Final Onset Parameter, the preferred scenario for the base-final onset is silence, as it is considered an unlicensed onset. Therefore, in Final Onset Parameter-OFF languages, lenition (weakening) may take place in the base-final onset, despite the opportunity for melodic filling in this position.

10Accordingly, there are different restrictive effects of unlicensing on the base-final onset in each language: i.e. for the base-final onset (i) only a placeless N is possible in Japanese; (ii) obstruents are not allowed in Malayalam; (iii) base-final branching onsets cannot occur in English; and (iv) base-final obstruent devoicing appears in Dutch, Turkish, Polish and so forth.

  • 3 In the present study, our discussion of GP is centered around KLV [1985, 1990] and Kaye [1995, 2014 (...)

11The details of the New Template Model and the unlicensed final O will be discussed in Section 1. Section 2 will compare and contrast the New Template Model with Standard Government Phonology (GP) (Kaye, Lowenstamm & Vergnaud [1985, 1990], Kaye [1995, 2014])3 in terms of domains, boundaries, domain-final p-licensed nucleus and lexicon modelling. The final section will suggest a revision of the Empty Category Principle (ECP) proposed by Kaye [1990, 1992] and give our concluding remarks with further study suggestions.

1. The new template model and the constituent structure

1.1. The basic premises of the new template model

12The present study aims to both propose a new template account and analyze phonological processes such as devoicing and vowel harmony, which exhibit sensitivity to word boundaries. A crucial aspect of this study is to develop an explanation without referring to diacritics, brackets or other extra-phonological objects, which have been commonly employed in previous phonology literature to identify morpheme boundaries and to distinguish words from affixes. With this aim, the study makes two major points: (i) it argues for a novel universal template model for bases, prefixes and suffixes; (ii) it puts forward two new parameters, the Initial Onset Parameter and the Final Onset Parameter, to explain the differences among languages regarding the base-initial and final positions ((non-)obligatory word-initial and/or final consonants).

13As the first aspect, we introduce a novel constituent structure model built upon templates. This model enables the identification of morphological categories like bases (e.g., words), prefixes, and suffixes within the phonological component. The base template (7b), which begins with and ends in an onset, accommodates prefixes (ON) (7a) to its left and suffixes (NO) (7c) to its right:

Image 10000201000000F60000008BBDCAC0895C4FFFAA.png

14In the present model, the phonological shape of prefixes and suffixes are in fact redundant. Given that they attach to a base and make it a larger one (O … O) (7a-c), a prefix template must begin with an onset (O) and end in a nucleus (N); suffixes must begin with a nucleus (N) and end in an onset (O) in their phonological template. Otherwise, a clash will appear (*OO/*NN) on the base, which is banned on the same phonological string (the same constituent structure), according to the Clash Principle (8) in our model:

Image 10000201000002D90000004E1BDDEF53123A300A.png

The clash avoidance idea has been previously proposed in phonology, particularly in metrical phonology concerning stress patterns (Gráf [2001], Nespor & Vogel [1979, 1986], Varga [1998, 2002]. According to this principle, it is prohibited for two or more consecutive stresses to occur (**). We adapt this clash avoidance idea into our model in order to give an account for the avoidance of identical phonological constituents in the same constituent structure (*OO / *NN).

15The phonological structure of the morphemes provided in (7a–c) corresponds to each morphological category in phonology, prefix, base and suffix, respectively. This implies that the templates inherently contain this morphological information (base, prefix, suffix) when there is productive morphology. Therefore, there is no need to introduce external morphological markings, diacritics, brackets, etc. In other words, we argue that the bases, productive suffixes, and prefixes are listed in the lexicon and computed by an affixation operation, consistent with the Dual-Route Lexicon Model proposed by Pinker & Prince (1991). In this model, productivity is defined as the capacity to generate and understand an unlimited number of new forms. Accordingly, our model suggests that these listed productive prefixes, suffixes, and bases are stored in the lexicon with their unique phonological templates: ON, NO, and ONO, respectively. Therefore, their phonological form inherently indicates their morphological labels. In essence, their morphological categories are identifiable when they are processed in phonology. Consider (9) to visualize how the system works in general within our model:

Image 100002010000026800000092ED20CB39538A4B66.png

  • 4 Note that the different subscripts belonging to bases (O1), suffixes (Na0) and prefixes (Nz1) given (...)

16Both the prefix {un-} and the suffix {-ful} given in (9) are productive affixes since they can attach to a large number of bases (Bauer [2001], Horn [2005], Plag [1999]). It means that the three morphemes {un-}, {faith}, {-ful} given in (9) are listed in the lexicon with their unique constituent structures according to our model. These morphemes come to phonology with their own templates, and phonological operations (government, licensing, etc.) apply to them when necessary. The suffixal Na0 or prefixal Nz1, as given in (9), is recognized by phonology as a part of a suffix or a prefix, respectively, but not as the integral part of the base thanks to that encapsulated lexical information available to phonology4. Also, note that no other morphological information is transferred into phonology via these templates. This means that phonology cannot see which affix is derivational, inflectional, a case marker, etc.: there is only limited access to morphological information via the templates.

17Another important point regarding affixes is that the ON structure for prefixes and the NO structure for suffixes allow for the consecutive attachment of further prefixes and suffixes, respectively, as demonstrated in (10):

Image 1000020100000218000000440DFFCF9422FE1771.png

As indicated in (10), when a prefix and/or a suffix is attached to a base, it forms part of a larger O…O structure, enabling further prefixation and/or suffixation in languages characterized by extensive affixation.

18To sum up, the preceding discussion suggests that the only viable templates for prefixes and suffixes are ON and NO, respectively. In morphology, prefixes and suffixes are not independent entities in languages but rather attached to a base. This morphological attachment is evident in our template model: the absence of a final onset in prefixes (ON) and the absence of an initial onset in suffixes (NO) render them phonologically bound to a base, as only bases exhibit an O…O structure.

  • 5 We do not present further discussion on Denwood [1998] and Charette [2004], which follow the domain (...)

19Furthermore, in our model, there is no upper limit on the number of template-internal ON pairs, unlike the proposals put forth by Denwood [1998] and Charette [2004]. They advocate for a four-position (ONON) template for Turkish words, introducing artificial domains to preserve domain-final p-licensing in their analysis5. Conversely, our model establishes a natural lower limit (ONO) since the base template must begin and end with an O. This also satisfies the minimum word requirement, which is either (C)V or (C)VC. (Dobrovolsky [1987], Inkelas & Orgun [1995], Itô & Hankamer [1989], Kabak [2014], Kabak & Vogel [2001], Kenstowicz [1994], McCarthy & Prince [1986] among others). Now let us go into the details of the base-final onset in the New Template Model.

1.1.1. The final onset parameter against the domain-final p-licensing parameter

20One of the points which differentiates our model from Standard GP approaches (Kaye, Lowenstamm & Vergnaud [1985, 1990], Kaye [1995, 2014] and the Strict CV version of GP (Lowenstamm [1996]) is related to the domain-final p-licensed nucleus and the Domain-Final P-licensing Parameter. In addressing the variations across languages regarding the presence or absence of word-final consonants (evident in Turkish, English, Polish, etc., versus languages like Vata, Italian, etc.), the present model eliminates the domain-final p-licensed nucleus and the Domain-Final P-licensing Parameter from phonological theory. Instead, it puts forward a new parameter, the Final Onset Parameter, which groups languages depending on their behavior with respect to final position. Consequently, our argument posits that each onset position is accompanied by a licenser nucleus, except for the final one (Ox). According to our model, in certain languages like Italian, Vata, and Zulu, the absence of a licensing nucleus means that no consonant can occur in the word-final position. In contrast, languages such as English, French, Malayalam, Polish, and Turkish may realize the final O melodically even if there is no licensing. Our model accounts for this difference among languages via the Final Onset Parameter, given in (6) and repeated in (11):

Image 1000020100000236000000ACB1C58A4A8830082F.png

  • 6 See Baturay-Meral [2020] for the theoretical account of the exceptions to the base-final restrictio (...)

21According to the Final Onset Parameter, in certain languages such as Italian, Vata, and Zulu, the base-final onsets must remain silent as specified by the parameter (Final Onset Parameter-ON). This implies that these languages do not permit any melodic filling in the base-final onset position. However, in languages like Japanese, Malayalam, English, French, Turkish, and Polish, the Final Onset Parameter is deactivated (OFF), allowing for the presence of consonants in the base-final position6.

22To gain a deeper understanding of our theory, let us contrast the Final Onset Parameter in our template model with the Domain-Final P-licensing Parameter of GP while analysing a hypothetical Italian word. We will show how a Standard GP-based analysis (12) and our model (13) explain the ungrammaticality of a consonant-final word in Italian:

Image 100002010000029F000001224B25FAD4908DCACF.png

23In Italian, all words must end in a vowel. Standard GP explains the absence of the final consonant with the domain-final p-licensing parameter, given in (14):

Image 10000201000002B10000008E8259FA78F6CE1787.png

The final empty nuclei are not domain-finally p-licensed in Italian since it sets the parameter in (14) OFF. Consequently, in (12), N2 never remains silent in that language; it must be pronounced. This implies that the word ending in a consonant, as seen in (12), is unacceptable in Italian.

24In contrast, the New Template Model elucidates why the word ending in a consonant, as represented in (13), is unacceptable in Italian without necessitating the mechanism of domain-final p-licensing. We elucidate why *cas is unacceptable in Italian through our Final Onset Parameter: Italian activates this parameter (ON), meaning that unlicensed onsets cannot be phonetically realized in this language. Thus, the final onsets must be phonetically silent in Italian.

25In our current study, we challenge the concept of domain-final p-licensing since the origin of licensing remains ambiguous in the theory. There exists a causality dilemma regarding the relationship between the domain boundary and the domain-final p-licensed nucleus. Is the final empty nucleus p-licensed simply because it is domain-final, or is the domain-final boundary determined by the presence of a p-licensed nucleus?

  • 7 See Baturay-Meral (2020) for a detailed critical review of Dienes & Szigetvári [1999] and Szigetvár (...)
  • 8 See Ziková & Scheer [2010] for the alternative view and discussion.

26Another reason for us to argue against domain-final p-licensing is related to markedness. Regarding the definition of markedness given in Ulfsbjorninn [2017], the empty categories, which need extra licensing such as the domain-final p-licensed nucleus, are more marked than their non-empty counterparts. Ulfsbjorninn [2017: 164] argues that this extra licensing increases structural complexity, the length of description and thereby markedness. Accordingly, compared to the ONON of Standard GP or to strict CVCV accounts, which may end in a domain-finally p-licensed nucleus, the ONO template for bases given in our model is less marked since the final O needs no licensing to occur in the constituent structure: the inherent property of consonants is muteness, which means that an onset does not need to be licensed to be silent (Dienes & Szigetvári [1999], Szigetvári [1999]).7 The domain-finally p-licensed empty nucleus in ONON or CVCV templates, on the other hand, is highly marked since it is licensed by an unidentified licensor.8 Thus, we argue that if there is a (base) template proposed for all languages, then it must be the less marked and less complex one, namely the ONO of the New Template Model.

27Regarding our final onset analysis, the relevant question might be whether the final unlicensed onset displays any deficiency or weakness compared to the licensed onset positions even if that unlicensed onset is allowed to be melodically filled out in certain languages according to the Final Onset Parameter. Let us discuss the unlicensed final onset position in more detail in Section 1.1.2, which will refer to final onset restrictions and lenition processes observed among languages in light of the Final Onset Parameter as the side effect of a lack of licensing.

1.1.2. The unlicensed base-final O among languages

28In Section 1.1.1, we argued that the base-final onset position may be melodically filled out in such languages as Japanese, Malayalam, English, French, Turkish and Polish since these languages fix the Final Onset Parameter in (11) OFF. In our model, following Szigetvári [1999], the onset does not have to be licensed to survive in the constituent structure. A nucleus only licenses its onset to support the maintenance of its melodic material. The licensed positions are always stronger than the unlicensed ones. Consider (15a–b) for the definitions of government and licensing, respectively, adapted from Szigetvári [1999].

Image 10000201000002E2000000D4197D601B51F3BDD3.png

29In accordance with Dienes & Szigetvári [1999] and Szigetvári [1999], we posit that there is no hindrance to the realization of an unlicensed C position (onset), even though the preferred base-final onset is phonetically silent, as observed in Italian (16), in accordance with the Final Onset Parameter. This preference arises due to its status as an unlicensed position:

Image 100002010000016F000000C90896F62759EB679E.png

30On the other hand, the unlicensed C position in the Final Onset Parameter-OFF languages is inclined to undergo lenition as there is no licensor nucleus which strengthens its melodic content. It means that there may be certain restrictions on the base-final onset of the Final Onset Parameter-OFF languages in that every consonant occurring in the consonant inventory of a language may not occur in the base-final onset position due to lack of licensing. For instance, the bases may end only in a placeless N in Japanese (Nasukawa [2005]), as given in (17):

Image 1000020100000119000000CB0011E7DC063BB732.png

  • 9 We consider final devoicing as an instance of lenition in light of Brockhaus [1995], Jessen & Ringe (...)

31Similar to the Japanese case above, Dutch, German, Polish and Turkish fix the Final Onset Parameter OFF. Hence, the final onset position may be filled out in those languages, but this comes with a cost: the final voiced (non)-continuant obstruents must be devoiced.9 Consider the word-final devoicing instances from Dutch given in (18a) (van Oostendorp [2007, 2008: 127]), from German given in (18b) (Grijzenhout [2001: 202]), Polish in (18c) (Cyran [2008: 450]) and Turkish in (18d):

Image 1000020100000265000000EB87A4CB82FA0B8A26.png

  • 10 The present study adapts the Binarity Theorem and the branching constituents proposal of Kaye [1990 (...)

32Another example of base-final restrictions could be given from English, where the branching onsets, i.e. consonant clusters such as tr, br, etc., are banned from the unlicensed base-final onset position (19g–i), although they can occur word-initially (19a–c) and -medially (19d–f)10:

Image 10000201000002330000007539463E8D624ACA0A.png

33A more severe example of word-final restrictions can be observed in Spanish. Note that Spanish has /p, t, k, b, d, g, θ, s, tʃ, x, m, n, ɲ, l, ʎ, ɾ/ in its consonantal inventory (Martínez-Celdrán, Fernández-Planas & Carrera-Sabaté [2003]). In this language, while CC clusters are possible word-initially (20a–f) and word-medially (20g–l), they are totally out word-finally. What is more, only five consonants /l/, /ɾ/, /d/, /n/, /s/ can appear in the unlicensed base-final onset position, and the rest of the consonantal inventory are excluded from that position (20m–s) (Bedore [1999]):

Image 1000020100000304000000E648013627FCEBE194.png

34Another final restriction on the onset realization is observed in Malayalam, which has 38 consonants in its phonological system (Punnoose [2010]), but no consonants other than [m] and [n] can appear word-finally in informal patterns of the language (Mohanan [1989: 591]). Mohanan [1989] states that a morpheme must end either in a vowel or in an [m]/[n] in the language. If not followed by a vowel-initial word (21a), obligatory vowel epenthesis [ə] (21b) takes place in word-final position:

Image 100002010000030400000056CBA69F69126326E0.png

  • 11 See Baturay-Meral [2020] for a detailed discussion on what determines the type(s) of base-final res (...)

35Instances of final onset restrictions among the Final Onset Parameter-OFF languages may be increased when we bring more languages into consideration11. The crucial point in the discussion is that we do not need any stipulative domain-final p-licensed nucleus to explain the final differences among languages. In Section 2, we will try to be more explicit about why the domain-final p-licensed nucleus falls into disfavor in our model.

36A valid question regarding our template model and the discussion presented up to this point might be why the base template needs to begin with and end in an O. Section 1.2 will show that the initial and final Os are necessary in the structure since the cross-linguistic data analyzed indicate that a consonant is an obligatory component in word-initial and/or word-final positions in certain languages. Additionally, the data analyses show that initial/final consonant epenthesis is a must in the adaptation of vowel-initial/final loan words into languages such as the Mayan Languages, which prohibit words beginning with or ending in a vowel. Also, the need for a final consonant in some instances of English and Portuguese, though not obligatory, supports our analysis. This means that the base-initial and base-final O positions are necessary in phonology. Let us examine the details in Sections 1.2.1 and 1.2.2.

1.2. Empirical and theoretical support for the O…O template

1.2.1. The initial O on the base template: obligatory word-initial consonants

37Empirically speaking, there are a number of languages where word-initial consonant(s) are a must and which exist under the initial onset position in the constituent structure. As for the theoretical point of view, the obligatory initial onset position is connected to the issue of the default syllable type (CV) (Clements & Keyser [1983]), the Onset Principle of Itô [1989] and Prince & Smolensky [2004] and the onset satisfaction need of a nucleus. Let us begin with presenting empirical data for the initial O.

38There are a number of languages where the base-initial onset position cannot be empty, such as Modern Arabic (Abdoh [2010], Watson [2002]), Ute (Givón [2011]), Ingush (Москва [1965] cited in Nichols [2011]) and most of the Australian languages (Dixon [1970: 273]), such as Djapu (Morphy [1983]), Nyangumarta (Sharp [2004]) and Warlpiri (Turpin, Demut & Ngampart Campbell [2014]). This implies that words have to begin with a consonant in those languages.

39In the Mayan Languages also, word-initial consonant(s) are obligatory (Kaufman [2015], Bennet [2016]). Bennet [2016: 476–477] notes that the Mayan languages do not allow words to begin with a vowel: words which begin with an underlying vowel obligatorily get [ʔ] as an epenthetic consonant in Tektikek (22a–b), which is a Mayan language spoken by the Tektitan people of Huehuetenango, Guatemala (Simons [2017]):

Image 1000020100000212000000716458AFFBBF88F3B2.png

  • 12 Consonant epenthesis in the initial onset position is not limited to the cases above. There are oth (...)

40Similarly, underlyingly vowel-initial words undergo glottal stop epenthesis in Washo (23a–c) (Staroverov [2014], Yu [2008]), which is a Native American language spoken by the Washo people on the California-Nevada border (Golla [2011])12.

41The examples given in (22–23) show that the base-initial O position must exist in the constituent structure since certain languages, such as Arabic, Ute, Ingush, Djapu, Nyangumarta, Warlpiri, Tektikek, Washo and Xincan, need this position to be filled out with an obligatory (epenthetic) consonant:

Image 10000201000003020000006C55C98E13A46D9F6F.png

42As for the theoretical point of view, the requirement for the initial onset position in the base template is linked to the concept of the “default syllable type” CV, which is present in all languages (Clements & Keyser [1983]). Therefore, it is reasonable to keep the base-initial O as a part of the constituent structure. Also, the principles proposed in theoretical phonology literature support the existence of the obligatory base-initial onset position, i.e. according to the Onset Principle proposed by Itô [1989: 223], every syllable has an onset and every nucleus needs an onset in the structure. Similarly, Prince & Smolensky [2004: 106] argue for the existence of an onset to the left of the nucleus and note that the typical unmarked syllable has an onset consonant.

43It is important to note that according to both Itô [1989] and Prince & Smolensky [2004], an onset is a required component preceding the nucleus. This finding further supports our claim that bases should initiate with an O rather than an N (24):

Image 10000201000000C90000001EC8B7236D5EE74FEA.png

44In our current research, we posit that every nucleus necessitates an onset for onset satisfaction. In example (24), N1 lacks such a pairing, resulting in its onset requirement remaining unsatisfied and, consequently, being excluded. To gain deeper insights into onset satisfaction, we draw an analogy between a nucleus and a transitive verb in syntax. Similar to how the object (whether overt or covert) is a must for both the Theta Criterion and argument structure satisfaction in the presence of a transitive verb (as proposed by Chomsky [1981, 1986]), in our phonological framework, a nucleus resembles a transitive verbal head, necessitating an onset (whether empty or not) to its left as its complement. We term this necessity onset satisfaction. The theoretical rationale behind the presence of the base-initial O in our model can be attributed to the concept of onset satisfaction. Furthermore, another theoretical justification for the initial O may stem from the concept of licensing. Nuclei, according to Harris [1994], hold the potential to license onsets. This is illustrated in (25):

Image 100002010000020A0000004FE128536BFEDB3954.png

  • 13 For the details of VC constituent structure proposal, see Szigetvári [1999] and Dienes & Szigetvári (...)

45Accordingly, a nucleus needs an onset to do licensing, as in (25). Hence, the structure outlined in (24) is considered unattainable as N1 requires an onset for licensing. This reinforces our opposition to the concept of a base beginning with an N without an onset, which contrasts with the viewpoint advocated by Szigetvári [1999]13, who argues that the constituent structure must begin with a V (N) position, as in (24).

46Recall that cross-linguistic observations demonstrate that the initial onset position must be phonetically filled in some languages (such as Modern Arabic, Ute, Ingush, Djapu, Nyangumarta, Warlpiri, etc.), while it is not required to be filled with a consonant in others (such as Turkish, English, German, etc.). The question arises: what determines whether the initial O must be phonetically realized or not? We argue that these variations among languages arise from a parametric choice, namely the Initial Onset Parameter, given in (26), which is a novel attempt in phonological theory.

Image 100002010000027C000000C93F698C96FAED9983.png

47According to the Initial Onset Parameter, the base-initial O must be phonetically realized in languages such as Modern Arabic, Ute, Ingush, Djapu, Nyangumarta, Tektikek, Washo, Xincan, and Warlpiri, as these languages set the parameter (26) to ON. Conversely, languages like English, French, and Turkish set the Initial Onset Parameter to OFF, indicating that the base-initial O does not require a consonant to be filled out. This parameter allows us to differentiate languages based on their realization of the base-initial onset.

48Let us now examine the importance of final onsets within our base template (O…O) and seek empirical evidence to substantiate their presence in the structure.

1.2.2. The final O on the base template: obligatory word-final consonants

49In the present study, we also argue that the base-final onset is necessary on the template, based on both empirical and theoretical support. Empirically speaking, it has been observed in the literature that a number of languages have an obligatory final consonant. As for the theoretical point of view, the final O gets some support from the Final-C Constraint of McCarthy [1993]. Let us begin in discussing certain empirical data showing the necessity of the final O in the constituent structure.

50Certain languages have been observed to exhibit the obligatory word-final consonant, such as Cairene Arabic and Arabic (nominal stems) (McCarthy & Prince [1990]), Yapese (Piggott [1999], Goad & Brannen [2003]) and Choctaw (Broadwell [2006]). Likewise, in Tilburg Dutch (Swets [2004]), vowels do not occur in the final position of lexical items.

51For the Mayan languages, Bennet [2016: 489] argues that free morphemes usually follow the /CVC/ template and that there is an active preference for consonant-final roots in loanword adaptation. In Yukatek Maya, for instance, vowel-final Spanish words are adapted into the language with a final epenthetic /h/, as in (27a–b):

Image 100002010000025200000095C8DED60800D11A26.png

52Anderbois [2011: 43] and Lombardi [2002: 229] note that [h] does not occur in underlying representations of words like (27a–b) but that /h/-epenthesis appears in Yucatec loans to meet the requirement that words must end in a C.

53Further support for the base-final O comes with Tunica, which is a language spoken in the Central and Lower Mississippi Valley in the United States. Similar to Yukatek Maya, (phrasal-final) words must end in a consonant in Tunica and the epenthesis of [n] applies base-finally to satisfy this requirement (28a–b).

Image 10000201000002410000007708AE48A0C1A63E0B.png

54The instances of word-final consonant epenthesis given in (28a–b) are argued to be a prosodic boundary marker in the literature (Nespor & Vogel [1988], Selkirk [1980], Zygis [2010]).

55In addition to final consonant epenthesis, an obligatory consonant may appear in vowel-ending words of certain languages. In the Bristol dialect of English, [l] is observed after all final schwas. Consequently, words like area and aerial are homophonous [eəriəl] (Wells [1982]). This subsequent [l] appears to occupy the final onset position in our base template. Another example of an obligatory final consonant appearing in some vowel-ending words is found in certain European dialects of Portuguese where nasalized vowels may transform into sequences of vowel plus nasal consonant (denasalized), as noted by Trigo [1993]. Consider (29a–d):

Image 100002010000022B00000075910692AFF6BA2BAF.png

(29a–d) exemplify the cases where nasalized vowels tend to occur as vowels followed by a nasal segment. Accordingly, words ending in a nasalized vowel may be pronounced as ending in a consonant. Consider bõ vs. bɔŋ (29b), represented in (30a–b), respectively. The [ŋ] in b[ɔŋ] given in (29b) seems to occur under the final O (O2) in (30b):

Image 1000020100000151000000977061F9F3EA5B9B22.png

56The inclusion of the nasal consonant within the structure is enabled by the final O position within the template. Essentially, the final onset position is a necessary component of the template, as it hosts the required consonant(s) in specific languages.

57From a theoretical perspective, it is noteworthy to mention that the concept of obligatory final onset is not novel in earlier phonology literature. For instance, McCarthy [1993: 176] advocates for the requirement of a final consonant in phonological (prosodic) words through a constraint known as the Final-C, which demonstrates that a prosodic word cannot end in a vowel but rather it ends in a consonant or a glide. Swets [2004: 48] also supports the Final-C constraint and argues that it is not a violable constraint in Tilburg Dutch, except for certain functional words.

58In Section 2, we will discuss the New Template Model and the Final Onset Parameter in a more detailed fashion by comparing our model with Standard Government Phonology (GP) (Kaye [1995, 2014]) in terms of domains, domain boundary identification and lexicon.

2. The new template model vs. the standard government phonology

59In the present study, we propose a new account in order to identify word boundaries in phonology and to explain phonological processes in light of the New Template Model, which is based on the templates as the constituent structure, government and licensing mechanisms, elements and parametric variations. In this respect, we adapt certain basic principles and claims of Standard GP, such as elements and binarity, into our analysis, and follow the Principle of Non-Arbitrariness, which is based on a direct relation between a phonological process and the context in which it occurs (Kaye, Lowenstamm & Vergnaud [1990: 194]). Following Standard GP (in the sense of Kaye, Lowenstamm & Vergnaud [1990]), the model has no place for rules applying to different parts of derivation, no rule or level ordering, no diacritics or any other external rules and no bivalent features in the model, as stated in the Introduction and Section 1 above. However, Standard GP is not without problems when we consider domain-final p-licensing. In Sections 2.1, 2.2 and 2.3, we compare and contrast our New Template Model with Standard GP and argue against the domain-final p-licensing nucleus and bracketing accounts.

2.1. Standard GP and domain-final p-licensed nucleus

60In Standard GP, all roots, prefixes and suffixes begin with an O(nset) and end in an N(ucleus) position (31). As illustrated in (31), each onset comes with its nucleus in a domain, and the nucleus licenses its onset pair. Harris [1994: 160] labels this process Onset Licensing, which is that an onset head position must be licensed by a nuclear position:

Image 10000201000000DB000000537A462D7CE4848915.png

61In Kaye’s [1995] framework, it is argued that ON pairs occur within domains where phonological processes operate. Consequently, Kaye’s system identifies three distinct types of domains: (i) independent analytic, exemplified in (32a); (ii) dependent analytic, as in (32b); and (iii) non-analytic, as in (32c):

Image 10000201000001D7000000B45BABA61C8A718D2F.png

62Regarding the interface between phonology and morphology, Kaye [1995: 302] highlights that morphological structures can either have no effect (non-analytic) or minimal effect (analytic) on phonology. In analytic morphology, morphological complexity is phonologically evident, whereas non-analytic forms show indifference from morphologically simple forms since phonology treats them as monomorphemic forms. The bracket pairs in phonological representations signify each domain where phonology applies, as demonstrated in (32a–c) above. According to Kaye [1995], phonology initially applies to each inner domain and subsequently to the entire string after concatenation.

63Kaye [1995: 308] argues that internal phonological domains enable morphological information to be visible in analytic morphology. This implies that phonology perceives the internal domain(s) but does not discern the morphological category of the suffix. However, in non-analytic morphology, phonology lacks access to morphological information due to the absence of separate domains.

64Although phonological forms are assumed to have ONON structure in Standard GP, there are many consonant-final forms in languages. According to Kaye [1995], GP permits the final nucleus to remain empty based on the Domain-Final P-licensing Parameter outlined in (14). Consequently, in languages where the Domain-Final P-licensing Parameter is set to ON, the final nucleus position may be p-licensed and muted, as observed in German, Polish, and Arabic. Conversely, if the parameter is set to OFF in a given language, words are not permitted to end with a consonant, as seen in Italian, Japanese, and Vata.

65However, certain languages, such as Japanese, may allow base-final consonant appearance, as we have stated in Section 1.1.2, although Kaye puts it into the OFF-list.

66Regarding the Domain-Final P-licensing Parameter, Kaye [1995] also argues that a domain-final p-licensed nucleus is licensed since it is domain-final, while the properly governed one is licensed by a full nucleus. Accordingly, the domain boundary is marked with a p-licensed nucleus (33):

Image 10000201000000D9000000962C2D2A6983EAF186.png

67We argue that identifying domainhood via the presence of a p-licensed empty nucleus poses several problems. First, it remains unclear whether the domain boundary causes the emergence of the final p-licensed empty nucleus, or if the presence of the final p-licensed empty nucleus defines the domain boundary. Second, without the assistance of extra-phonological objects (such as brackets), the definition of the domain boundary remains ambiguous. Moreover, it is hard to identify the domain boundary in languages which do not allow domain-final p-licensing. In the following section, we compare our model with Standard GP and explore the limitations of the analytic versus non-analytic categorization proposed by Kaye [1995] in explaining the intricate dynamics of suffixation and its impact on the interface between phonology and morphology in Turkish and other languages.

2.2. The new template model vs. standard GP: boundaries

68In our template model, as opposed to Standard GP, all bases begin with an onset (empty [34b] or full [34a]) and end in an onset (empty [34b] or full [34a]), as given in the Turkish examples (34a–b):

Image 1000020100000331000000B1DF161CCB9467F54B.png

69The model demonstrates that every onset is paired with a nucleus, except for the final one in (34a–b). The initial and final onsets serve as indicators of the beginning and end of a new base, respectively. Therefore, these initial and final Os serve as natural, phonological, and non-diacritic “boundary markers”, indicating the beginning and end of a new base.

70The immediate theoretical consequence of this claim is that there are neither brackets nor domain-final p-licensed empty nuclei in our model, contrary to Standard GP.

71In our model, two identical constituents (*NN or *OO) cannot occur in the same template (in the same row), as stated in Section 1.1, since *NN or *OO occurrence violates the Clash Principle. The Clash Principle is significant in our model as it elucidates how morphological boundaries are visible within phonological structures. Now let us examine the example of Turkish compounding provided in (35) below, which will help us illustrate the boundaries of bases:

Image 10000201000001C3000000DAD87301AAE19875AF.png

  • 14 Vowel harmony (element spreading) is a phonological process which cannot exceed the word boundary t (...)

72According to the Clash Principle, O3 and O1 in (35) cannot be adjacent within the same constituent structure (phonological string). Therefore, the adjacency of O3 and O1 indicates that O3 is the final position of a word, while O1 signifies the beginning of another. In that way, word boundaries are naturally visible in phonology, thanks to our template model. As a result, the element spreading (I spreading) from N2 of the first base to N1 of the second base is blocked, i.e. *demir kepi is out.14 Thanks to the templates, our model offers an explanation for why the spreading of elements is restricted from crossing certain boundaries: two adjacent Os (O3 and O1) block element spreading from one nucleus to another by creating opacity between two nuclei (N2 and N1). In Kaye’s [1995] system, conversely, every phonological domain ends in a nucleus, whether it is empty or not. Thus, in the Standard GP framework, demir kapı would have a structure similar to the one illustrated in (36):

Image 10000201000001E2000000E80F80E0480D66F354.png

73In our model, the presence of two adjacent onsets in (35) signifies a boundary, preventing spreading. This mechanism allows for the automatic detection of word boundaries by the phonological constituent structure. However, in Kaye’s system, as illustrated in (36) above, the element (I) cannot identify boundaries and stop spreading without the aid of brackets, such as [[demir][kapı]]. It could be argued that N3 is licensed at the domain boundary in the representation (36). Nevertheless, domain-final p-licensing simply explains why the final nucleus is silent and does not serve as an indicator of a domain boundary. In other words, the preceding nucleus (N2 in (36)) lacks the capacity to know how N3 is muted; it is unable to recognize whether N3 is domain-finally p-licensed or properly governed. The evidence for this uncertainty is demonstrated by the suffixation illustrated in (37):

Image 10000201000001AC000000D8F73924F2A7769820.png

74The suffixation represented in (37) and the compounding in (36) share identical constituent structures within Standard GP. This indicates that N3 is domain-finally p-licensed in both (36) and (37). However, in (37), (I) spreads to the suffix N1, whereas it does not in (36). Consequently, Standard GP fails to address the boundary issue without resorting to brackets. This is evident in (38a–b), where compounding and suffixation are distinguished from each other through brackets in Standard GP:

Image 10000201000000AE0000004EB7AF91F42DCDAB82.png

75According to Kaye [1995], the brackets employed in the representations shown in (38a–b) are considered mere symbols and not integral parts of the constituent structure. However, the comparison between (37-38) and (38a–b) demonstrates that brackets function as genuine boundary markers. Consequently, without the use of brackets, suffixation and compounding share identical constituent structures within Kaye’s [1995] system. In contrast, our model distinguishes between the two, as in (39a–b) and shows why element spreading is impossible in compounding cases, as exemplified in (39a), but is possible in suffixation cases, as represented in (39b):

Image 10000201000003CD000000F7A21387664716E9B9.png

76The representations in (39a–b) demonstrate that bases begin and end in an onset, and that the boundaries on both sides of bases are inherent to the structure itself, as illustrated in (39a). On the contrary, suffixes initiate with a nucleus, indicating that they are not bases (39b), but rather dependencies of a base since the initial nucleus of the suffix attaches to the base’s final onset. In that way, phonology lets the element go from N2 to Na0 in (39b) but not from N2 to N1 in (39a) due to the blocking effect of O3O1.

77In Sections 2.1 and 2.2, we compared our model with Standard GP (Kaye [1995]) and outlined several fundamental distinctions between the two models concerning constituent structure, licensing relationships, and the phonology-morphology interface. Now, we will explore further into the discussion of Standard GP and conduct a comparative analysis with our model in Section 2.3., focusing on the constituent structure, domainhood and the lexicon view.

2.3. The new template model vs. standard GP: domainhood and lexicon modelling

78According to Standard GP (Kaye [1995, 2014]), phonology can see morphology [[A]B] only if there is a morphological footprint on phonology such as devoicing or weird sound sequences (e.g. a long vowel preceding consonant clusters, as in the case of the English seeped [[si:p]t]). If there is no morphological footprint on phonology, the entire structure is presumed to be stored as a single unit in the lexicon, i.e., [AB]: kept [kept]. This suggests that the output is interpreted as synthetic (non-analytic). Our inquiry into Standard GP (Kaye [1995, 2014]) revolves around whether all forms labeled as synthetic are indeed stored as a single unit in the lexicon. For instance, what if we attach the regular past tense marker to a base and get an output resembling a simplex word like massed? Note that massed features a VCC sequence [æst], which is typical in English simplex words like fast and cast. Consequently, the root+suffix combination in mass-ed does not provide any indication of the morphological boundary according to Kaye’s analysis. Consider (40a–b).

Image 100002010000008800000037C33DD7DF4423F28D.png

79If we get a morphological footprint on phonology as the only evidence for a morphological boundary, would we say that the whole structure massed was kept as a single unit (non-analytic), as in (40a), or would we assume that the output mass-ed was a result of analytic morphology, as in (40b), since its attachment may bear unusual sequences, as in other examples, such as seeped (V:CC)? Or might we argue that the regular past tense marker is analytic in certain cases, like in seeped [[si:p]t], but non-analytic in others, such as massed [mæst]?

80Regarding Kaye’s (1995) system, Scheer [2011: 297] also raises concerns about parsing in cases of hidden morphology (without phonological parsing cues), such as in words like massed [mæst] and stepped [stept]. He states that [stept] does not provide any phonological parsing cue, so “look-up of the entire item” is expected to be performed. By arguing against such an idea, Scheer [2011] rightly claims that there is no lexical item such as [stept] in the lexicon since it is derived via productive morphology.

81Also, in a language like Turkish, there are more than 104,000 words and 200 suffixes (Banguoğlu [1976], Gencan [1979], Göksel & Kerslake [2011], Kornfilt [2013] and Korkmaz [2014] among others). There are numerous productive suffixes, like the accusative {-(y)I} and the dative {-(y)A} markers, which do not leave any trace on phonology. Additionally, some suffixes, such as the plural marker {-lAr}, may exhibit morphological cues in certain phonological contexts but not in others. Consequently, according to the GP perspective (Kaye [2014]), we are led to the conclusion that kitap-lar ‘book-PL’ is morphologically complex due to devoicing, while araba-lar ‘car-PL’ is stored as a single unit in the lexicon owing to the absence of any phonological footprint. Considering this, we might question whether morphology plays an active enough role and whether our lexicon is burdened with an excessive number of forms, both mono-morphemic and synthetic. Do we truly maintain such an extensive list in our lexicon?

82Hankamer [1989: 401] posits that human word recognition in agglutinative languages involves parsing. Maintaining a complete listing of derivations in memory for languages like Turkish, with its agglutinative suffixation, becomes impractical due to the size, complexity, and sheer number of words. He illustrates this with the example of a nineteen-morpheme multi-morphemic word that can be formed through affixation in Turkish: muvaffak-iyet-siz-leş-tir-ici-leş-tir-iver-e-me-yebil-ecek-ler-imiz-den-miş-siniz-cesine ‘as if you were one of those we cannot easily make a maker of unsuccessful ones’. Hankamer [1989] argues that an educated native speaker of Turkish would need to store over 200 billion word forms, a capacity far beyond the capabilities of the human brain.

83In the present study, we are also against such huge storage in the lexicon and argue that productive suffixes combine with a base via morphology, following the Dual-Route Lexicon Model (Pinker [1991, 1994). In the Dual-Route Lexicon Model, as espoused in this investigation, a “mental lexicon” is posited, comprising memorized lexical items. Within this framework, irregular forms (e.g., ran) are retained within this mental lexicon and retrieved as needed, while regular forms (e.g., walk-ed) are generated via rules encoded within the mental grammar. These rules systematically combine base forms to construct larger words and are applicable in a productive manner to novel and non-sense forms (Berko [1958], Chomsky [1965, 1995], Pinker [1991, 1994], Pinker & Prince [1988]). Consequently, the inflection of regular forms is predictable, whereas irregular verb forms lack the same degree of flexibility within the Dual-Route Lexicon Model. Thus, irregular forms are specifically memorized and subject to associative memory influences, such as frequency and similarity (Pinker [1991, 2000], Pinker & Prince [1988], Ullman [1999]).

84In Pinker’s [2000] work, he highlights the challenge posed by language productivity. Productivity refers to the ability of a language to create new words or forms. Pinker [2000: 122] points out that while individual words like walk or duck can be easily stored in the lexicon, the sheer number of new words formed by combining prefixes, stems, and suffixes makes it impossible to memorize them all. He notes that languages such as Kivunjo or Turkish may have hundreds of thousands to millions of word forms, a feat that speakers could not realistically achieve during childhood.

  • 15 See Rumelhart & McClelland [1986], Smolensky [1995], Seidenberg & Gonnerman [2000], Daugherty & Sei (...)

85Accordingly, we adopt the view that all regular productive (inflectional and derivational) suffixes and prefixes in languages are listed in the lexicon with their phonological template and that they combine with bases, which also have their own templates, via morphological (or morpho-syntactic) operations rather than being listed in the lexicon as whole chunks. Unproductive and irregular forms are lexically stored in our model, too, as the Dual-Route Lexicon Model supports.15 This is also important for the application of phonological processes, as we noted before, since phonology knows the morphological categories (prefix, suffix) thanks to their templates kept in the lexicon, if there is productivity. Phonological processes apply according to this information.

86As the synthesis of our views of the lexicon and the phonology-morphology interface, our model suggests that constructions such as araba-lar ‘car-PL’, araba-yı ‘car-ACC’, araba-ya ‘car-DAT’, and others alike are formed through the rules of morphology rather than being stored as separate entries in the lexicon. This perspective is consistent with the idea of semantic compositionality, implying that synthetic morphology tends to lack transparent semantic meanings. In other words, the meanings of the individual components do not independently contribute to the overall meaning of the word. In our case, on the other hand, the productive affixes are highly analyzable units.

87Furthermore, Şafak [2015: 83], in her examination of Turkish suffixes using the Dual-Route Lexicon Model, supports the idea proposed in our model that productive word forms, both inflected and derived, are analyzed into their basic building blocks – stems and suffixes – during the process of recognizing words visually in Turkish.

88In light of our model’s perspective on the relationship between lexicon and phonology, we suggest that certain morphological combinations may affect phonological structures, while others may not leave a noticeable footprint. See (41a–c):

Image 1000020100000112000000546DF2352CC6051946.png

  • 16 Note that a phonological process may be a morphological boundary marker in one language but not in (...)

89In (41a–b), there is word-final devoicing, which is a footprint-marking word boundary in Turkish, while there is no devoicing in (41c).16 However, there is active suffixation both in (41b) and (41c), i.e. the plural and accusative markers are both productive in Turkish. According to Kaye’s [1995, 2014] model, on the other hand, the plural marker must be analytic in (41b) since word-final devoicing is a morphological footprint on phonology, appearing due to the domain-final p-licensed empty nucleus (N3) (lack of onset licensing power of N3) (42):

Image 1000020100000203000000DE09CAFAE2E74FF29E.png

90However, the accusative marker does not cause any devoicing in (41c). Accordingly, kitab-ı must be non-analytic and stored as a whole chunk in the lexicon in Kaye’s [2014] system. This means that it is indifferent from a simplex form.

91In our template model, we challenge the dichotomy between analytic and non-analytic approaches in Standard GP, as it leads to problematic outcomes. For example, consider the accusative marker, a highly productive suffix. It’s evident that kitab-ı is formed through a morphological process of base+acc, resulting in kitab-ı. The key distinction between kitap-lar and kitab-ı lies in how morphology affects phonology: while kitap-lar exhibits devoicing due to morphology, kitab-ı does not. However, the absence of a phonological cue does not negate the presence of morphology. If a phonological cue exists, it indicates morphology, whereas its absence does not necessarily imply the lack of morphology; it depends on the productivity of the affix. In our template model, we assume that all productive suffixes listed in the lexicon follow the same phonological template N…O:

Image 10000201000002FC000000FF11C6C2510348ED03.png

92In (43a), the suffix-initial Na0 constitutes an empty nucleus that is appropriately governed by Na1. However, it does not serve as the genuine counterpart to O3, as O3 and Na0 are not situated on the same template within the lexicon. Consequently, Na0 lacks the capacity to license O3, given its status as a properly governed empty nucleus distinct from the base. Thus, devoicing emerges as a result. It is essential to emphasize that Na0 lacks the capability to license O3 while it requires this onset since every nucleus needs an onset in the constituent structure (Onset Satisfaction). However, O3 does not need Na0 to exist in the structure.

93In (43b), on the other hand, Na0 is not properly governed, so it can license O3 though it belongs to the suffixal template. In short, there is a difference between the suffix-initial properly governed empty nucleus (43a) and the suffix-initial interpreted nucleus (43b) regarding the onset licensing in our model, as a result of which the unlicensed base-final onset (O3 in [43a]) is exposed to a weakening/lenition process (Brockhaus [1995], Jessen & Ringen [2002], Harris [1997, 2009], Lombardi [1995], Mascaró [1987]).

94Going back to final devoicing as an instance of lenition, it is a mark of morphology on phonology, which signals the word boundary (right). This means that morphological information is visible in the constituent structure via devoicing. Consider the Turkish examples in (44a–d), which involve an unproductive suffix {–lAk}, to get a better insight into productive vs. unproductive morphology:

Image 10000201000001060000006B3FB52E491EFF3F29.png

95{-lAk} in (44a–d) represents an unproductive suffix in Turkish, observed on a limited number of bases. According to Korkmaz [2014: 131], {–lAk} originated from the fusion of {–lA+k} and possessed a distinct property as a unit for deriving nouns and adjectives. Of significance for our discussion is the voicing status of the final consonant of the base to which the suffix historically attached. Upon examination of the data, it is evident that {–lAk} follows [b] and [d] in (44c–d) respectively. If {–lAk} were a productive suffix attached to the base via morphology, it would possess its own template, with the initial nucleus of this template being a properly governed empty nucleus. Consequently, it would not act as a licensing nucleus for the base-final O. In other words, ödlek would be expected to manifest as *öt-lek due to lenition, specifically devoicing. However, this is not observed. When ödlek (44d) is compared with simplex (45b–c) and complex (45a) words, it presents itself as a simplex word phonologically, rather than a complex one:

Image 10000201000001DF0000006BECA0C170C1245A9F.png

96(45a–d) above show that öd-lek is similar to the simplex forms abla ‘sister’ and medlul ‘sense’ (45b–c), respectively, since the voiced non-continuant obstruent [d] does not devoice in (45d) similar to the base internal ones, [b] and [d] respectively, in (45b–c). Note that ödlek was (historically) morphologically complex although its historical complexity is phonologically undetectable now. Accordingly, it is stored in the lexicon as a single unit (46a), similar to abla (46b), but unlike kitap-lar (45a), which is represented in (43a):

Image 10000201000002DF000000D684A5B05BDCB2C899.png

  • 17 Word-internal empty nuclei have licensing power in Turkish if one takes examples like abla “sister” (...)

97The representations given in (46a–b) demonstrate that both öd-lek and abla are preserved in the lexicon as individual entities. It is clear that the base-internal properly governed empty nucleus (N2 in (46a–b)) can license the voiced non-continuant obstruents. However, in cases like kitap-lar ‘book-pl’ (43a), where a suffix-initial properly governed empty nucleus (Na0) is unable to license a base-final voiced non-continuant obstruent, it becomes evident that the licensing dynamics between an onset and an empty nucleus vary based on their origin (base or suffix). In the current context, it can be argued that while the suffix-initial properly governed empty nucleus fails to license the voicing property of the base-final onset, the base-internal counterpart effectively manages to do so for the preceding onset. This distinction is evident in the behavior of properly governed empty nuclei (N2) as illustrated in (46a) and (46b), both of which exhibit similar patterns and successfully license O2. Consequently, in both ödlek (46a) and abla (46b), there is no occurrence of devoicing17. However, the unlicensed onset (O3) remains subject to specific constraints, such as devoicing in Turkish since phonology recognizes that the nucleus (Na0) does not belong to the base template but is suffix-initial, as in (43a).

98To sum up, the assertion that stem combined with unproductive suffix instances are stored as singular units is corroborated by phonological evidence, particularly the lack of devoicing observed in voiced non-continuant obstruents in Turkish.


99In the present study, we discussed our New Template Model (47a–c) in detail and presented empirical and theoretical support for the initial and final onset positions in the base template:

Image 10000201000000ED0000008C177CC70227015B80.png

100We argued that the base-initial and base-final Os, serving as natural boundary markers in phonology, are indispensable on the template due to the requirement of these positions to be invariably occupied by a (potentially epenthetic) consonant in various languages. As a new parameter, we put forward the Initial Onset Parameter to differentiate languages where the word-initial consonant is required from those where it does not have to be pronounced. The Final Onset Parameter is another parameter we put forward in the present study. This parameter distinguishes languages from one another according to (non-)occurrence of the consonant(s) in their base-final onset position.

101Also, we argued that the final O is theoretically essential to explain base-final phonological processes. We compared and contrasted our model with the Standard GP account (Kaye [1995, 2014]) in terms of the final position, boundary identification and lexicon view. As a novel attempt, we distinguished “bases” from “suffixes” and “prefixes” via phonological templates (ONO, NO and ON, respectively), which are empirically and theoretically supported.

102Since we argue that the phonological system has no domains or domain-final p-licensing in the sense of Standard GP, we now propose to eliminate (48a) “Domain-final (empty) categories are p-licensed (parametrized)” from the definition of the Phonological Empty Category Principle (ECP), given in (48a–c):

Image 10000201000002DD000000A00FBDCC75067E4EC8.png

  • 18 As for sCC clusters, we follow Baturay-Meral [2020], who leaves the Magic Licensing of Kaye [1992] (...)

Contrary to the definition of the ECP given in (48), the present study argued that only properly governed nuclei receive no phonetic interpretation18.

103In sum, the present study tries to provide a theoretical model for phonology (the phonology-morphology interface) and the lexicon, which welcomes further studies testing it. An instance of the further research points may be about the empirical coverage of the model.

  • 19 See the template proposal of Baturay-Meral [2023a–b] for infixes.

104We have focused solely on the primary morphological categories within our model: base, prefix, and suffix. It is worth noting that there exist other forms of affixation, including infixation, as observed in Tagalog (McCarthy & Prince [1993]), and circumfixation, as seen in Chickasaw (Fitzgerald [2016]), among other languages. Moreover, there are isolating languages devoid of affixation, such as Chinese (Shibatani & Bynon [1995]; Tai [1973]). These languages and types of affixation may be subject to investigation within the framework proposed here in future research endeavors19.

Top of page


Abdoh, Eman M. A. 2010. A Study of the Phonological Structure and Representation of First Words in Arabic. Ph.D. Dissertation, University of Leicester.

Anderbois, Scott. 2011. Strong positions and laryngeal features in Yukatek Maya. In Suzi Lima, Kevin Mullin & Brian Smith (eds.). North East Linguistic Society (NELS) 39.1. 41–54. Amherst, MA: GLSA.

Banguoğlu, Tahsin. 1976. Türkçe’nin Grameri [Turkish Grammar]. Ankara: TDK Publications.

Bauer, Laurie. 1983. English Word-formation. Cambridge & New York: Cambridge University Press.

Bauer, Laurie. 2001. Morphological Productivity. Cambridge: Cambridge University Press.

Baturay-Meral, Semra. 2020. The New Template Model and the Phonology Morphology Interface in Turkish: The Parametric Hierarchical System and Universal Implications. Ph.D. Dissertation, Radboud University Nijmegen.

Baturay-Meral, Semra. 2023a. Are infixes phonologically prefixes or suffixes? A constituent structure analysis. International Journal of Language Academy 11(4). 225-247.

Baturay-Meral, Semra. 2023b. Does Turkish Have Infixes? Why not? Korkut Ata Türkiyat Araştırmaları Dergisi Special Issue 1 (for the 100th Anniversary of the Republic of Turkey). 1339-1359.

Bedore, Lisa M. 1999. The acquisition of Spanish. In Taylor Orlando & Leonard Laurence (eds.), Language Acquisition across North America: Cross-cultural and Cross-linguistic Perspectives, 157–207. San Diego, CA: Singular.

Bennett, Ryan. 2016. Mayan phonology. Language and Linguistics Compass 10(10). 469–514.

Berko, Jean. 1958. The child’s learning of English morphology. Word 14. 150-177.

Booij, Geert & Jerzy Rubach. 1987. Postcyclic versus postlexical rules in lexical phonology. Linguistic Inquiry 18. 1–44.

Broadwell, George Aaron. 2006. A Choctaw Reference Grammar. Lincoln: University of Nebraska Press.

Brockhaus, Wiebke. 1995. Final devoicing in the phonology of German. Linguistische Arbeiten 336. Tübingen: Niemeyer.

Campbell, Lyle. 1997. American Indian Languages: The Historical Linguistics of Native America. New York: Oxford University Press.

Charette, Monik. 2004. Defining the structure of Turkish words. SOAS Working Papers in Linguistics 13. 49-79.

Charette, Monik & Aslı Göksel. 1996. Licensing constraints and vowel harmony in Turkic languages. SOAS Working Papers in Linguistics and Phonetics 6. 1-25. Also in Cyran Eugeniusz (ed.), Structure and Interpretation: Studies in Phonology, 65-88. Lublin: Folium, 1998.

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.

Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.

Chomsky, Noam. 1986. Knowledge of Language. London: Praeger.

Chomsky, Noam & Morris Halle. 1968. The Sound Pattern of English. Cambridge, Mass.: MIT Press.

Clements, George N. & Samuel Jay Keyser. 1983. CV Phonology. Cambridge, Massachusetts: MIT Press.

Cyran, Eugeniusz. 2008. Consonant clusters in strong and weak position. In de Carvalho Joaquim Brandão, Tobias Scheer & Philip Ségéral (eds.), Lenition and Fortition, 447-481. Berlin: Mouton De Gruyter.

Daugherty, Kim G. & Mark S. Seidenberg. 2001. Beyond rules and exceptions: A connectionist approach to inflectional morphology. In Susan D. Lima, Roberta Corrigan & Gregory K. Iverson (eds.), The Reality of Linguistic Rules, 353-388. Amsterdam: John Benjamins.

Denwood, Margaret Ann. 1998. A template for Turkish. SOAS Working Papers in Linguistics and Phonetics 8. 177–190.

Dienes, Péter & Péter Szigetvári. 1999. Repartitioning the Skeleton: VC Phonology. Ms. Budapest: ELTE.

Dixon, Robert M. W. 1970. Olgolo syllable structure and what they are doing about it. Linguistic Inquiry 1(2). 273-276.

Dobrovolsky, Michael. 1987. Why CVC in Turkish reduplication? In Pirkko Lilius & Mirja Saari (eds.), The Nordic Languages and Modern Linguistics 6, 131-146. Helsinki.

Fitzgerald, Colleen M. 2016. Morphology in Muskogean Languages. Language and Linguistics Compass 10(12). 681-700.

Gencan, Tahir Nejat. 1979. Dilbilgisi [Grammar]. Istanbul: TDK Publications.

Gibson, Mark. 2010. Fonología del Español: Enfoque desde la Teoría de Optimidad [Issues in Spanish Phonology from an Optimality-Theoretic Approach]. Ph.D. dissertation, Universidad Complutense de Madrid.

Givón, Talmy. 2011. Ute Reference Grammar. Amsterdam: John Benjamins.

Goad, Heather & Kathleen Brannen. 2003. Phonetic evidence for phonological structure in syllabification. In Jeroen van de Weijer, Vincent van Heuven & Harry van der Hulst (eds.), The Phonological Spectrum vol. II: Suprasegmental Structure. Amsterdam: John Benjamins.

Golla, Victor. 2011. California Indian Languages. Berkeley: University of California Press.

Göksel, Aslı & Celia Kerslake. 2011. Turkish: A Comprehensive Grammar. London: Routledge.

Gráf, Zoltán Benedek. 2001. Length in Metrical Phonology. Ph.D. Dissertation, Eötvös Loránd University.

Grijzenhout, Janet. 2001. Devoicing and voicing assimilation in German, English, and Dutch: A case of constraint interaction. In Katarzyna Dziubalska-Kolaczyk (ed.), Constraints and Preferences, 201-228. Berlin: Mouton de Gruyter.

Hankamer, Jorge. 1989. Morphological parsing and the lexicon. In William Marslen-Wilson (ed.), Lexical Representation and Process, 392-408. Cambridge, Mass.: MIT Press.

Harris, John. 1994. English Sound Structure. Oxford: Blackwell.

Harris, John. 1997. Licensing inheritance: An integrated theory of neutralization. Phonology 17. 315-370.

Harris, John. 2009. Why final obstruent devoicing is weakening. In Kuniya Nasukawa & Phillip Backley (eds.), Strength Relations in Phonology, 9-46. Berlin: de Gruyter.

Horn, Laurence. 2005. An un-paper for the unsyntactician. In Salikoko Mufwene, Elaine Francis & Rachel Wheeler (eds.), Polymorphous Linguistics: Jim McCawley’s Legacy, 329-365. Cambridge, Mass.: MIT Press.

Inkelas, Sharon & Cemil Orhan Orgun. 1995. Level ordering and economy in the lexical phonology of Turkish. Language 71. 763–793.

Itô, Junko. 1989. A prosodic theory of epenthesis. NLLT 7. 217-260.

Itô, Junko & Jorge Hankamer. 1989. Notes on monosyllabism in Turkish. In Junko Itô & Jeffrey Runner (eds.), Phonology at Santa Cruz 1, Syntax Research Center, 61-69. UC Santa Cruz.

Iverson, Gregory K. & Joseph C. Salmons. 2007. Domains and Directionality in the Evolution of German Final Fortition. Phonology 24. 121-145.

Jessen, Michael & Catherine Ringen. 2002. Laryngeal features in German. Phonology 19. 1-30.

Kabak, Barış. 2014. Pervasive syllables and phonological unity in words. In J. C. Reina & R. Szczepaniak (eds.), Syllable and Word Languages, 112-139. Berlin: De Gruyter.

Kabak, Barış & Irene Vogel. 2001. The phonological word and stress assignment in Turkish. Phonology 18. 315-360.

Kaisse, Ellen M. & Patricia A. Shaw. 1985. On the theory of Lexical Phonology. Phonology Yearbook 2. 1–30.

Kaye, Jonathan. 1990. Coda licensing. Phonology Yearbook 7. 301-330.

Kaye, Jonathan. 1992. Do you believe in magic? The story of s+C sequences. SOAS Working Papers in Linguistics 2. 293–313.

Kaye, Jonathan. 1995. Derivation and Interfaces. In Jacques Durand & Francis Katamba (eds.), Frontiers of Phonology, 289-332. London and New York: Longman.

Kaye, Jonathan. 2014. The Ins and Outs of phonology. In Sabrina Bendjaballah, Noam Faust, Mohamed Lahrouchi & Nicola Lampitelli (eds.), The form of structure, the structure of form: Essays in honor of Jean Lowenstamm, 255–270. Amsterdam & Philadelphia: John Benjamins.

Kaye, Jonathan, Jean Lowenstamm & Jean-Roger Vergnaud. 1985. The internal structure of phonological representations: A theory of charm and government. Phonology Yearbook 2. 305-328.

Kaye, Jonathan, Jean Lowenstamm & Jean-Roger Vergnaud. 1990. Constituent structure and government in phonology. Phonology 7. 193-231.

Kaufman, Terrence. 2015. Initial Glottal Stop in Mayan Languages. Ms., University of Pittsburgh.

Kenstowicz, Michael. 1994. Phonology in Generative Grammar. London: Blackwell.

Kim, Seon Jung. 1996. The Representations of Korean Phonological Expressions and their Consequences. Ph.D. Dissertation, University of London.

Kiparsky, Paul. 1982. From cyclic to Lexical Phonology. In Harry van der Hulst & Neil Smith (eds.), The Structure of Phonological Representations 1, 131–75. Dordrecht: Foris.

Korkmaz, Zeynep. 2014. Türkiye Türkçesi Grameri (Şekil Bilgisi) [Turkish Grammar: Morphology]. Ankara: TDK Publications.

Kornfilt, Jaklin. 2013. Turkish. London and New York: Routledge.

Lombardi, Linda. 1995. Laryngeal neutralization and syllable wellformedness. Natural Language and Linguistic Theory 13. 39-74.

Lombardi, Linda. 2002. Coronal epenthesis and unmarkedness. Phonology 19(2). 219-251.

Lowenstamm, Jean. 1996. CV as the only syllable type. In J. Durand and B. Laks (eds.), Current Trends in Phonology: Models and Methods, 419-441. CNRS, ESRI, Paris X.

Martínez-Celdrán, Eugenio, Ana Fernández-Planas & Josefina Carrera-Sabaté. 2003. Castilian Spanish. Journal of the International Phonetic Association 33(2). 255-259.

Mascaró, Joan. 1987. A Reduction and Spreading Theory of Voicing and Other Sound Effects. Ms. Universitat Autònoma de Barcelona.

McCarthy, John. 1993. A case of surface constraint violation. Canadian Journal of Linguistics 38. 169-195.

McCarthy, John & Alan Prince. 1986. Prosodic Morphology. Ms., University of Massachusetts & Brandeis University.

McCarthy, John & Alan Prince. 1990. Prosodic morphology and templatic morphology. In Mushira Eid & John McCarthy (eds.), Perspectives in Arabic Linguistics 2, 1-52. Amsterdam: John Benjamins.

McCarthy, John & Alan Prince. 1993. Prosodic Morphology I: Constraint Interaction and Satisfaction. Ms., University of Massachusetts & Rutgers University.

McClelland, James L. & Karalyn Patterson. 2002a. Rules or Connections in Past Tense Inflections: What does the evidence rule out? Trends in Cognitive Neurosciences 6(11). 465-472.

McClelland, James L. & Karalyn Patterson. 2002b. ‘Words or Rules’ cannot Exploit the Regularity in Exceptions. Trends in Cognitive Neurosciences 6(11). 464-465.

Mohanan, Karuvannur Puthanveettil. 1986. The Theory of Lexical Phonology. Dordrecht: Reidel.

Mohanan, Karuvannur Puthanveettil. 1989. Syllable structure in Malayalam. Linguistic Inquiry 20(4). 589- 625.

Morphy, Frances. 1983. Djapu, a Yolngu dialect. In Robert M. W. Dixon & Barry Blake (eds.), Handbook of Australian Languages 3, 1-188. Amsterdam: John Benjamins.

Nasukawa, Kuniya. 2005. A Unified Approach to Nasality and Voicing. Berlin: Mouton de Gruyter.

Nespor, Marina & Irene Vogel. 1979. Clash avoidance in Italian. Linguistic Inquiry 10(3). 467-482.

Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris.

Nespor, Marina & Irene Vogel. 1988. Prosodic Phonology. Berlin: Mouton de Gruyter.

Nichols, Johanna. 2011. Ingush Grammar. Berkeley: University of California Press.

Van Oostendorp, Marc. 2007. Derived environment effects and consistency of exponence. In Sylvia Blaho, Patrik Bye & Martin Kramer (eds.), Freedom of Analysis, 123–148. Berlin: Mouton de Gruyter.

Van Oostendorp, Marc. 2008. Incomplete devoicing in formal phonology. Lingua 118(9). 1362-1374.

Piggott, Glyne L. 1999. At the right edge of words. The Linguistic Review 16. 143-185.

Plag, Ingo. 1999. Morphological productivity: Structural constraints in English derivation. Berlin: Mouton de Gruyter.

Pinker, Steven. 1991. Rules of language. Science 253. 530-535.

Pinker, Steven. 1994. The Language Instinct. New York: Harper Collins.

Pinker, Steven. 2000. Words and Rules: The Ingredients of Language. New York: Harper Collins.

Pinker, Steven & Alan Prince. 1988. On language and connectionism: An analysis of a parallel distributed processing model of language acquisition. Cognition 28. 73-193.

Pinker, Steven & Alan Prince. 1991. Regular and irregular morphology and the psychological status of rules of grammar. In Susan D. Lima, Roberta L. Corrigan & Gregory K. Iverson (eds.), The Reality of Linguistic Rules, 353-388. Amsterdam: John Benjamins.

Pöchtrager, Markus A. 2009. Does Turkish diss harmony? Paper presented at the 6th Old World Conference in Phonology, January 21-24, 2009, Edinburgh.

Prince, Alan & Paul Smolensky. 2004. Optimality Theory: Constraint Interaction in Generative Grammar. Oxford: Blackwell. [Ms., Rutgers University and University of Colorado, Boulder, 1993.]

Punnoose, Reenu. 2010. An Auditory and Acoustic Study of Liquids in Malayalam. Ph.D. Dissertation, Newcastle University.

Rogers, Chris. 2016. The Use and Development of the Xinkan Languages. University of Texas Press.

Rumelhart, David E. & James L. McClelland. 1986. On Learning Past Tenses of English Verbs. In David E. Rumelhart & James L. McClelland (eds.), Parallel Distributed Processing 2: Psychological and Biological Models. MIT Press.

Scheer, Tobias. 2004. A Lateral Theory of Phonology: What is CVCV, and Why should it be? Berlin: Mouton de Gruyter.

Scheer, Tobias. 2011. A Guide to Morphosyntax-Phonology Interface Theories. Berlin: Mouton de Gruyter.

Scheer, Tobias. 2012. Direct Interface and One-Channel Translation: A Non-Diacritic Theory of the Morphosyntax-Phonology Interface. Berlin: Mouton de Gruyter.

Seidenberg, Mark S. & Laura M. Gonnerman. 2000. Explaining Derivational Morphology as the Convergence of Codes. Trends in Cognitive Sciences 4(9). 353-361.

Selkirk, Elisabeth O. 1980. Prosodic domains in phonology: Sanskrit revised. In Mark Aranoff & Mary-Louise Kean (eds.), Juncture. A Collection of Original Papers, 107-129. Saratoga: Anma Libri.

Sharp, Janet Catherine. 2004. Nyangumarta, A Language of the Pilbara Region of Western Australia. Canberra: Pacific Linguistics.

Shibatani, Masayoshi & Theodora Bynon. 1995. Approaches to language typology: A conspectus. In Masayoshi Shibatani & Theodora Bynon (eds.), Approaches to Language Typology, 1-25. Oxford: Clarendon Press.

Simons, Gary F. 2017. Ethnologue: Languages of the World, Twentieth edition. Dallas, Texas: SIL International.

Smolensky, Paul. 1995. Reply: Constituent Structure and Explanation in an Integrated Connectionist/Symbolic Cognitive Architecture. In Cynthia MacDonald & Graham MacDonald (eds.), Connectionism: Debates on Psychological Explanation. Basil Blackwell, Oxford.

Staroverov, Petr. 2014. Splitting Theory and Consonant Epenthesis. Ph.D. Dissertation. Rutgers University.

Swets, Francine. 2004. The Phonological Word in Tilburg Dutch. Government Phonology and a City Dialect. Ph.D. Dissertation, University of Amsterdam.

Szigetvári, Péter. 1999. VC Phonology: A Theory of Consonant Lenition and Phonotactics. Ph.D. Dissertation, Eötvös Lorand University.

Şafak, Duygu F. 2015. Processing of Inflected Words in Second Language. MSc Thesis, Middle East Technical University.

Tai, James H.-Y. 1973. Chinese as a SOV language. Chicago Linguistic Society 9. 659-671. Chicago: Chicago Linguistic Society.

Trigo, R. Lorenza. 1993. The inherent structure of nasal segments. In Marie K. Huffman & Rena A. Krakow (eds.), Phonetics and Phonology 5: Nasals, Nasalization and the Velum, 369-400. New York: Academic Press.

Turpin, Myfany, Katherine Demuth & April Ngampart Campbell. 2014. Phonological aspects of Arandic baby talk. In Rob Pensalfini, Myfany Turpin & Diana Guillemin (eds.), Language Description Informed by Theory, 49-80. Amsterdam: John Benjamins.

Ullman, Michael T. 1999. Acceptability ratings of regular and irregular past-tense forms: Evidence for a dual-system model of language from word frequency and phonological neighborhood effects. Language and Cognitive Processes 14. 47–67.

Watson, Janet C. E. 2002. The Phonology and Morphology of Arabic. Oxford: Oxford University Press.

Wells, John C. 1982. Accents of English. Cambridge: Cambridge University Press.

Varga, Laszlo. 1998. Rhythmical variation in Hungarian. Phonology 15. 227-266.

Varga, Laszlo. 2002. Intonation and Stress: Evidence from Hungarian. Palgrave Macmillan, Basingstoke.

Yu, Alan C. L. 2008. The phonetics of quantity alternation in Washo. Journal of Phonetics 36(3). 508–520.

Ziková, Markéta & Tobias Scheer. 2010. The Coda Mirror V2. Acta Linguistica Hungarica 57(4). 411-431.

Zygis, Marzena. 2010. Typology of consonantal insertions. In Melanie Weirich & Stefanie Jannedy (eds.), Papers from the Linguistics Laboratory, 111-140. Berlin: ZAS.

Top of page


1 The present study has been developed from Baturay-Meral [2020].

2 In the present study, the unlicensed onset means the onset which is not licensed by any nucleus (e.g. the base final O3 in (5a–b)). Note that the properly governed empty nuclei may be onset licensors. See Baturay-Meral (2020) for details.

3 In the present study, our discussion of GP is centered around KLV [1985, 1990] and Kaye [1995, 2014] since they introduced the basic premises of GP, as well as domain final p-licensing discussions, to the literature. However, the issue can also be extended to other GP work in the future. See Baturay-Meral (2020) for further discussion.

4 Note that the different subscripts belonging to bases (O1), suffixes (Na0) and prefixes (Nz1) given in (9) and (10) may be imagined as the formulization of different morphemes. These subscripts do not occur in phonology or the lexicon. They are just for the reader to follow the machinery, representing their differences in the lexicon and phonology.

5 We do not present further discussion on Denwood [1998] and Charette [2004], which follow the domain final p-licensing account of Kaye [1995] in their templatic system, since most of the domainhood and domain-final p-licensing criticism in the present study is also valid for them. See Baturay-Meral [2020] for further criticism on their domainhood and pseudo-morphology accounts regarding the lexicon, phonology and morphology.

6 See Baturay-Meral [2020] for the theoretical account of the exceptions to the base-final restrictions in the Final Onset Parameter-ON languages such as Italian.

7 See Baturay-Meral (2020) for a detailed critical review of Dienes & Szigetvári [1999] and Szigetvári [1999].

8 See Ziková & Scheer [2010] for the alternative view and discussion.

9 We consider final devoicing as an instance of lenition in light of Brockhaus [1995], Jessen & Ringen [2002], Harris [1997, 2009], Lombardi [1995], and Mascaró [1987]. See Kim [1996] and Iverson & Salmons [2007] for the alternative view that final devoicing is fortition.

10 The present study adapts the Binarity Theorem and the branching constituents proposal of Kaye [1990] and considers TR clusters as branching onsets, whereas RT clusters are in a coda-onset licensing relation, as proposed by Kaye [1990].

11 See Baturay-Meral [2020] for a detailed discussion on what determines the type(s) of base-final restriction(s) the Final Onset Parameter-OFF languages are exposed to.

12 Consonant epenthesis in the initial onset position is not limited to the cases above. There are other languages which have this phenomenon. See Rogers [2016] for more data.

13 For the details of VC constituent structure proposal, see Szigetvári [1999] and Dienes & Szigetvári [1999].

14 Vowel harmony (element spreading) is a phonological process which cannot exceed the word boundary though highly active in complex (suffixed) forms of Turkish (Charrette & Göksel [1996], Pöchtrager [2009]). Note that licensing conditions (Charrette & Göksel [1996], Pöchtrager [2009]) are as important as base boundaries in explaining element spreading in Turkish and in other harmonic languages.

15 See Rumelhart & McClelland [1986], Smolensky [1995], Seidenberg & Gonnerman [2000], Daugherty & Seidenberg [2001], McClelland & Patterson [2002a, 2002b] for alternative lexicon model(s) (e.g. the Single System Lexicon Model).

16 Note that a phonological process may be a morphological boundary marker in one language but not in others. Therefore, each language has its own morphological footprint on phonology, i.e. devoicing marks the word boundary in Turkish but not in English.

17 Word-internal empty nuclei have licensing power in Turkish if one takes examples like abla “sister” into consideration. In Turkish, there are no branching onsets, but /b/ and /l/ occur under separate onsets. [b] occurs under O2, which is followed and licensed by a properly governed nucleus (N2), as in (46b). However, note that the licensing power of the properly governed nucleus needs further investigation for languages which are assumed to have syllable-final devoicing (or just consonantal assimilation), such as German (Brockhous [1995]). Also, the potential reasons behind the different behaviors of the word-internal governed empty nuclei and the suffix-initial ones need further investigation. See Baturay-Meral (2020) for a detailed discussion of the issue.

18 As for sCC clusters, we follow Baturay-Meral [2020], who leaves the Magic Licensing of Kaye [1992] out of phonological theory and proposes a new type of empty nucleus to explain licensing of sCC and heavy clusters among languages.

19 See the template proposal of Baturay-Meral [2023a–b] for infixes.

Top of page


Electronic reference

Semra Baturay-Meral, “Phonological templates and the lexicon”Lexis [Online], 23 | 2024, Online since 25 April 2024, connection on 14 June 2024. URL:; DOI:

Top of page

About the author

Semra Baturay-Meral

Yıldız Technical University



Top of page



The text only may be used under licence CC BY-SA 4.0. All other elements (illustrations, imported files) are “All rights reserved”, unless otherwise stated.

Top of page
Search OpenEdition Search

You will be redirected to OpenEdition Search