The University of Manchester

Ricardo Bermúdez-Otero

Linguistics and English Language, University of Manchester





Stratal Optimality Theory: an overview

The John Rylands University LibraryLinguistics and English language at ManchesterE-mail me


The chain of parallel and successive operations is what builds complexity.

Sean B. Carroll, Endless forms most beautiful, ch. 4




Pending the publication of [16], this page presents a selection of my writings, arranged so as to provide a general survey of Stratal OT. The page describes the main features of the theory as I currently understand it myself; other scholars pursuing similar lines of inquiry will no doubt disagree with me frequently on matters of detail, and occasionally on matters of principle. I focus on the leading ideas and major architectural assumptions of Stratal OT, and on the empirical results that follow directly from them; work on the detailed implementation of the theory remains ongoing.



Those looking for a brief schedule of readings covering the essentials of Stratal OT may wish to work through items [15], [8], [9], [13], and [2], in that order.



This page was created on 11 December 2010. It was initially based on the syllabus of a course on ‘The phonological cycle’ which I taught at the 2010 LOT Summer School in Nijmegen; I am grateful to the organizers of the LOT Summer School and to the students on the course for that invaluable opportunity to clarify and develop my thought on Stratal OT.


The page was last updated on 5 September 2016.



Leading ideas

Stratal OT arises from the combination of three powerful ideas, each with a long and complex intellectual history of its own:

  1. the phonological cycle, which dates back to the earliest days of generative phonology;

  2. phonological stratification, which has roots in the Prague School distinction between phonologie du mot and phonologie de la phrase, but is most closely associated with Lexical Phonology;

  3. parallel constraint-based computation in the manner of OT.

When combined, these elements react with one another in interesting ways: e.g.

  1. Replacing the ordered rules of Lexical Phonology with the ranked constraints of OT strengthens the concept of cyclic domain: see [11], [12], and below.

  2. Adopting a stratal-cyclic architecture for phonology enables one to dispense with many of the constraint types and correspondence relations posited in mainstream OT: see [8], [15], and below. A stratal-cyclic architecture also allows one to restore a modular and local conception of the morphosyntax-phonology interface within OT: see [9], [12], [14], and below.


The cycle


The hypothesis of the phonological cycle [15] asserts that certain constituents in the morphosyntactic structure of a linguistic expression define domains for phonological computation. Phonology applies iteratively over these domains, starting with the smallest, least inclusive cyclic domains, and moving progressively outwards to larger, more inclusive cyclic domains.


In [8] and [15] I provide arguments for a cyclic approach to morphosyntactic conditioning in phonology and against output-output correspondence. The cycle is shown to make several correct predictions, including the following, which is a corollary of the Russian Doll Theorem:


If a phonological process exhibits cyclic misapplication within a certain phonological configuration created by affixation, then it must also exhibit cyclic misapplication if the same configuration arises by word concatenation.


In the version of Stratal OT that I currently envisage, the phonological cycle is an emergent phenomenon, and cyclic effects within stem-level domains have a particular aetiology: see [9] and [11].





The hypothesis of phonological stratification [15] asserts that, for the purposes of phonological interpretation, morphosyntactic constituents are divided into three types: stem-level, word-level, and phrase-level. Each type is associated with its own ranking of phonological constraints.


Phonological stratification enables one to state systematic correspondences between morphosyntactic constituents and cyclic domains: see [4] and [15].

  1. The highest node in a linguistic expression triggers a cycle of the phrase-level phonology.

  2. Each inflectionally complete grammatical word triggers a cycle of the word-level phonology.

  3. Constituents created by root-to-stem derivation, and by an idiosyncratic set of derivational (stem-to-stem) and inflectional processes, trigger cycles of the stem-level phonology. However, internal cyclic effects within stem-level domains exhibit irregularities as part of the stem-level syndrome

  4. Roots do not trigger cycles. 

I regard it as a long-term goal of the theory to account for the emergence of these correspondences: see [9] and [11] for some steps in this direction.


Stratal OT does not impose formal limits on the extent to which constraint rankings may differ across levels within the grammar of a single language. Rather, the life cycle of phonological processes keeps level divergence at bay through diachronic change: see [1], [8], and [13] for discussion of this point.


Stratal OT as I envisage it acknowledges the fact that word-level affixes can and do occur inside stem-level affixes; but it nonetheless upholds the claim that word-level phonological cycles never precede stem-level cycles. Take, for example, English devélop ~ devélop-ment ~ devèlop-mént-al. The stem-level suffix -al attaches outside word-level -ment. Nonetheless, the base of -al is a stem rather than an inflectionally complete grammatical word (cf. *development-s-al), and it therefore does not trigger a word-level cycle: see [15].



Parallel constraint-based computation


Stratal OT assumes that, in each phonological cycle, the mapping from input to output is effected by means of parallel constraint-based computation in the manner of OT. 


However, Stratal OT disavows many features of mainstream OT:

  1. Stratal OT rejects many of the constraint types and correspondence relationships to which mainstream OT resorts when dealing with morphosyntactic conditioning effects. Notably, there are no output-output identity constraints: see [8] and [15].

  2. In line with its modular approach to the morphosyntax-phonology interface, Stratal OT imposes strict limits on the ability of phonological constraints to refer to extraphonological information: e.g. lexically indexed constraints are banned [9].

  3. Stratal OT's emphasis on cyclic locality reverses classical OT's trend towards excessively global analyses: see [9], [12], and [14].


Stratal OT vs Lexical Phonology

Although Stratal OT draws heavily on the tradition of Lexical Phonology, the two frameworks differ in important respects:

  1. It is not normally possible to turn a Lexical Phonology analysis into a Stratal OT one simply by replacing the ordered rewrite rules of the former with the optimality-theoretic constraints of the latter. Principles such as Strict Cyclicity and Structure Preservation play no role in Stratal OT; and, unlike Lexical Phonology, Stratal OT bans devices such as the extrinsic ordering of segmental transformations before prosodification within a cycle. In consequence, the concept of cyclic domain has far greater empirical content in Stratal OT than in Lexical Phonology: see [11] and [12].

  2. Stratal OT need not assert that the stratal-cyclic architecture of phonology is specified innately by Universal Grammar. On the contrary, in the version of the theory that I currently envisage, stratification and cyclicity emerge from fundamental storage and processing mechanisms, and from timing effects in the child’s linguistic development: see [9] and [11].


The architecture

Papers [9], [13], and [7] together provide an overall conception of the place of phonology in the architecture of grammar, and of the interactions of phonology with the lexicon, morphosyntax, and phonetics.



The classical modular feedforward architecture


Stratal OT assumes a classical modular feedforward architecture of grammar: see [9] and [13]. In a grammatical architecture of this sort, morphology, phonology, and phonetics constitute separate modules, each possessing its own proprietary set of representations:

  1.   morphology performs computations over morphs;

  2.   phonology performs computations over discrete phonological categories;

  3.   phonetics performs computations over continuous articulatory and auditory parameters.

These three modules interact serially: morphology precedes phonology within each cycle; all phonology precedes all phonetics; morphology and phonetics do not share an interface.



Phonology and the lexicon

Items [9] and [11] distinguish between two types of lexical listing: complex constructs may be

  •  listed nonanalytically (as whole output forms) or

  •  listed analytically (as concatenations of input pieces).

These papers argue that stem-level constructs are listed nonanalytically, whereas word-level constructs are either unlisted or listed analytically. This postulate explains the stem-level syndrome.


In turn, items [12] and [14] demonstrate that, in a stratal-cyclic architecture, the size of lexically listed exponents makes precise predictions about locality conditions on phonologically driven allomorph selection. The evidence of allomorphic locality supports an approach to morphology that is stem-driven, rather than root-driven. [12] and [14] describe the format of lexical entries in this framework.



Phonology and morphology

Item [9] lays out a system of four hypotheses regulating the modular interaction between morphology and phonology: the Four-Hypothesis Programme. This consists of the following four postulates:

  1. The Morph Integrity Hypothesis asserts that morphology selects and inserts morphs as integral units, and does not have the power to operate directly upon elements of phonological representation such as features, segments, nodes, or association lines.

  2. The Indirect Reference Hypothesis asserts that phonological constraints other than those on prosodic alignment cannot refer to morphosyntactic information.

  3. The Phonetic Interpretability Hypothesis asserts that output phonological representations do not contain diacritics of morphosyntactic affiliation.

  4. As noted above, the Cycle Hypothesis asserts that phonology applies cyclically over morphosyntactically defined domains.

The Four-Hypothesis Programme entails an approach to nonconcatenative morphology labelled Generalized Nonlinear Affixation. In this view, the role of morphology in apparently nonconcatenative exponence reduces to the insertion of morphs whose phonological content is nonsegmental: i.e. floating pieces of feature geometry or bare fragments of prosody. Item [9] works out the implications of Generalized Nonlinear Affixation for the analysis of reduplication. Item [15] provides a list of references to recent work in Generalized Nonlinear Affixation.


Moreover, the cycle gives rise to a range of locality effects. Items [12] and [14] explore cyclic locality constraints on phonologically driven allomorph selection. In addition, the depth of morphosyntactic information that phonology can access within a cyclic domain is restricted by Phonetic Interpretability: thus, a form of ‘Bracket Erasure’ emerges for free, as shown in [9].



Phonology and phonetics

Papers [7], [10], and [13] address the phonology-phonetics interface: they provide arguments in favour of the classical hypothesis that gradient processes of phonetic implementation may not directly refer to lexical or morphological information. [7] and [13] list a number of phenomena that may create the appearance of direct morphological conditioning in phonetics. [10] and [13] discuss the diachronic processes of stabilization whereby gradient phonetic rules become categorical.



The life cycle of phonological processes


Research in the tradition of Lexical Phonology and Stratal OT pays close attention to the intimate connection between the architecture of phonology and phonological change. In particular, Stratal OT provides an insightful account of the life cycle of phonological processes, surveyed in [3], [10], and [13].

  1. Mechanical phonetic effects become phonologized as language-specific gradient processes of phonetic implementation: see [10], [13].

  2. Gradient processes of phonetic implementation become stabilized as categorical phonological processes applying across the board in phrase-level domains: see [10], [13].

  3. Analogy, operating by input restructuring, causes the cyclic domains of phonological processes to become progressively narrower, and so phonological generalizations climb up from the phrase level to the word level and from the word level to the stem level: see [10] and [13], with further examples in [8] and [12].

  4. Phonological rules eventually become morphologized or lexicalized: see [9].

  5. Each of the steps in this pathway may produce rule scattering, whereby an innovative avatar of an existing process enters a higher component of the grammar whilst the old process remains in situ: see [7], [13].


Some results


Stratal OT has implications for virtually every area of phonological research. The basic principles of the theory produce desirable outcomes when applied to a wide range of problems:  

  1. opacity,

  2. cyclic effects vs prosodic effects,

  3. contrast and Richness-of-the-Base paradoxes,

  4. the stem-level syndrome,

  5. phonological variation,

  6. phonological change.

This list is far from exhaustive.





Although there is a broad consensus on the intensional definition of phonological opacity, the extension of the concept is currently unknown because a phonological generalization can be classified as opaque or transparent only against a complex and usually disputed background of assumptions concerning the theory of representations, the constraint set, the contents of the lexicon, the role of morphosyntax, the role of phonetics, and the relative scope of synchronic and diachronic explanation: see [1] and [8]. Accordingly, Stratal OT does not set out to provide a comprehensive theory of opacity; no such theory exists at present, nor is one likely to come within our grasp in the near future.


Stratal OT does, however, make an important contribution in this area: a stratal-cyclic analysis of a misapplication effect arising from affixation or word concatenation automatically generalizes to instances of the same misapplication effect in morph-internal environments. [2] highlights the learnability advantages of this result.



Cyclic effects vs prosodic effects


One of the toughest problems in the analysis of morphosyntactically conditioned phonological phenomena is how to distinguish between the effects of morphosyntax on phonological representations (prosody) and the effects of morphosyntax on phonological derivations (what we may informally and pretheoretically call ‘synchronic analogy’). Stratal OT enjoys significant strengths in this area because if offers demarcation criteria that are not available to other theories: e.g.

  1. Prosody is visible to phonetics; cyclic domains are not.

  2. Cyclic domains are exactly coextensive with morphosyntactic constituents; prosodic units need not be.

  3. Cyclic effects obey cyclic locality; prosodic effects do not.

[6] illustrates the advantages of these criteria with an extended case-study; see also [8].



Contrast and Richness-of-the-Base paradoxes


In classical OT under Richness of the Base, all levels of phonological representation mix contrastive and predictable information. This can cause ranking paradoxes when phonological processes apply in a non-structure-preserving way: i.e. when tokens of a segment x arise predictably in some environment through allophony or neutralization, but x never occurs contrastively elsewhere. In such cases, classical OT may encounter a ranking paradox: the constraint *[x] must rank high enough to ban contrastive tokens of x, but it must simultaneously rank low enough to enable x to emerge as the output of allophony or neutralization.


Stratal OT does not face this problem because segments excluded from the phonemic inventory of a language are filtered out by the stem-level hierarchy, but may be permitted to arise again through allophony or neutralization at the word or phrase levels: see [5].



The stem-level syndrome


Stem-level phonological processes typically exhibit a cluster of characteristic properties, discussed in [11]:

  1. stem-level processes may or may not display cyclic reapplication within complex stem-level forms;

  2. stem-level processes may be neutralizing or purely allophonic within minimal domains;

  3. if a stem-level process does display cyclic reapplication within complex stem-level forms, then it cannot be purely allophonic in minimal domains (Chung's Generalization);

  4. in turn, cyclic reapplication within stem-level forms shows irregularities and exceptions of its own.

[9] and [11] show that this syndrome can be explained on the assumption that stem-level constructs are listed nonanalytically.



Phonological variation


Stratal OT with stochastic ranking retains Gregory Guy's insights into the effect of cyclicity on phonological variation.


For example, English t-deletion applies with greater frequency in mist (where the [t] belongs to the stem) than in miss-ed (where the [t] belongs to a word-level suffix) because mist meets the conditions for t-deletion both at the stem and at the word levels, whereas miss-ed only meets them at the word level: see [7], [13].



Phonological change


Stratal OT enjoys privileged insight into those phenomena reflecting the diachronic life cycle of phonological processes.


The allophony of liquids in present-day English provides a good example. Synchronic phonological processes reflecting early changes in a diachronic trajectory of liquid lenition tend to apply in smaller cyclic domains (at higher levels) than processes reflecting later changes: [10], [13]. For example, there are many English dialects in which

  •  r-lenition applies at the word level, and r-deletion at the phrase level;

  •  l-darkening applies at the word level, and l-vocalization at the phrase level.

This enables us to provide cogent answers to questions such as the rise of r-intrusion [8] and the history of English phrasal syllabification [10].


In addition, Stratal OT illuminates the problems of lexical diffusion, e.g. [3] and [9], and of secondary split [3].




  1. Bermúdez-Otero, Ricardo. 1999. Constraint interaction in language change: quantity in English and Germanic [Opacity and globality in phonological change]. PhD dissertation: University of Manchester & Universidad de Santiago de Compostela.

  2. Bermúdez-Otero, Ricardo. 2003. The acquisition of phonological opacity. In Jennifer Spenader, Anders Eriksson & Östen Dahl (eds), Variation within Optimality Theory: Proceedings of the Stockholm Workshop on `Variation within Optimality Theory´, 25-36. Stockholm: Department of Linguistics, Stockholm University. [For an update of my analysis of Canadian Raising, see here]

  3. Bermúdez-Otero, Ricardo. 2007. Diachronic phonology. In Paul de Lacy (ed.), The Cambridge handbook of phonology, 497-517. Cambridge: Cambridge University Press.

  4. Bermúdez-Otero, Ricardo. 2007. Morphological structure and phonological domains in Spanish denominal derivation. In Fernando Martínez-Gil & Sonia Colina (eds), Optimality-theoretic studies in Spanish phonology, 278-311. Amsterdam: John Benjamins.

  5. Bermúdez-Otero, Ricardo. 2007. Marked phonemes vs marked allophones: segment evaluation in Stratal OT. Paper presented at the Workshop on Segment Inventories, GLOW XXX, Tromsø, 11 April 2007.

  6. Bermúdez-Otero, Ricardo & Ana R. Luís. 2009. Cyclic domains and prosodic spans in the phonology of European Portuguese functional morphs. Paper presented at the Old World Conference in Phonology 6, Edinburgh, 24 January 2009.

  7. Bermúdez-Otero, Ricardo. 2010. Morphologically conditioned phonetics? Not proven. Paper presented at On Linguistic Interfaces II, Belfast, 2 December 2010.

  8. Bermúdez-Otero, Ricardo. 2011. Cyclicity. In Marc van Oostendorp, Colin Ewen, Elizabeth Hume & Keren Rice (eds), The Blackwell companion to phonology, vol. 4, 2019-2048. Malden, MA: Wiley-Blackwell.

  9. Bermúdez-Otero, Ricardo. 2012. The architecture of grammar and the division of labour in exponence. In Jochen Trommer (ed.), The morphology and phonology of exponence (Oxford Studies in Theoretical Linguistics 41), 8-83. Oxford: Oxford University Press.

  10. Bermúdez-Otero, Ricardo & Graeme Trousdale. 2012. Cycles and continua: on unidirectionality and gradualness in language change. In Terttu Nevalainen and Elizabeth Closs Traugott (eds), The Oxford handbook of the history of English, 691-720. Oxford: Oxford University Press.

  11. Bermúdez-Otero, Ricardo. 2013. The stem-level syndrome. Paper presented to the UPenn Linguistics Department, Speaker Series, Philadelphia, 11 April 2013.

  12. Bermúdez-Otero, Ricardo. 2013. The Spanish lexicon stores stems with theme vowels, not roots with inflectional class features. Probus 25(1): 3-103.

  13. Bermúdez-Otero, Ricardo. 2015. Amphichronic explanation and the life cycle of phonological processes. In Patrick Honeybone and Joseph C. Salmons (eds), The Oxford handbook of historical phonology, 374-399. Oxford: Oxford University Press.

  14. Bermúdez-Otero, Ricardo. 2016. We do not need structuralist morphemes, but we do need constituent structure. In Heidi Harley & Daniel Siddiqi (eds), Morphological metatheory, 387-429. Amsterdam: John Benjamins.

  15. Bermúdez-Otero, Ricardo. Forthcoming. Stratal Phonology. In S.J. Hannahs & Anna R. K. Bosch (eds), The Routledge handbook of phonological theory. Abingdon: Routledge.

  16. Bermúdez-Otero, Ricardo. In preparation. Stratal Optimality Theory (Oxford Studies in Theoretical Linguistics). Oxford: Oxford University Press.