Draft:Substance-Free Logical Phonology: Difference between revisions

 

Line 7: Line 7:

<!– Important, do not remove anything above this line before article has been created. –>

<!– Important, do not remove anything above this line before article has been created. –>

”’Substance-Free Logical Phonology”’ (”’SFLP”’) is an approach to [[generative phonology]] that models phonological computation over [[Segment (linguistics)|segments]] represented as sets of [[feature (linguistics)|distinctive features]], with [[Phonological rule|rule]] effects stated using formalisms from [[formal logic]] and [[set theory]].<ref name=”HaleReiss2000″>{{cite journal |last1=Hale |first1=Mark |last2=Reiss |first2=Charles |title=”Substance Abuse” and “Dysfunctionalism”: Current Trends in Phonology |journal=Linguistic Inquiry |volume=31 |issue=1 |pages=157–169 |year=2000 |doi=10.1162/002438900554334}}</ref> SFLP is part of ”substance-free phonology”, which holds that phonological computation is insensitive to [[phonetic]] substance ([[articulatory phonetics|articulatory]] or [[acoustic phonetics|acoustic]] properties).<ref name=”Chabot2022″>{{cite journal |last=Chabot |first=Alex |title=On substance and Substance-Free Phonology: Where we are at and where we are going |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=429–443 |year=2022 |doi=10.1017/cnj.2022.37}}</ref> On this view, phonetic interpretation is handled at the phonology–phonetics interface, which maps the output of phonological computation to speech; explanations that appeal to phonetic substance are typically located outside the phonological grammar (e.g., [[historical linguistics|diachrony]], [[language acquisition|acquisition]], or [[speech production|speech performance]]).<ref name=”HaleReiss2000″ /><ref name=”Chabot2022″ />

”’Substance-Free Logical Phonology”’ (”’SFLP”’) is an approach to [[generative phonology]] that models phonological computation over [[Segment (linguistics)|segments]] represented as sets of [[feature (linguistics)|distinctive features]], with [[Phonological rule|rule]] effects stated using formalisms from [[formal logic]] and [[set theory]].<ref name=”HaleReiss2000″>{{cite journal |last1=Hale |first1=Mark |last2=Reiss |first2=Charles |title=”Substance Abuse” and “Dysfunctionalism”: Current Trends in Phonology |journal=Linguistic Inquiry |volume=31 |issue=1 |pages=157–169 |year=2000 |doi=10.1162/002438900554334}}</ref> SFLP is part of ”substance-free phonology”, which holds that phonological computation is insensitive to [[phonetic]] substance ([[articulatory phonetics|articulatory]] or [[acoustic phonetics|acoustic]] properties).<ref name=”Chabot2022″>{{cite journal |last=Chabot |first=Alex |title=On substance and Substance-Free Phonology: Where we are at and where we are going |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=429–443 |year=2022 |doi=10.1017/cnj.2022.37}}</ref> On this view, phonetic interpretation is handled at the phonology–phonetics interface, which maps the output of phonological computation to speech; explanations that appeal to phonetic substance are typically located outside the phonological grammar (e.g., [[historical linguistics|diachrony]], [[language acquisition|acquisition]], or [[speech production|speech performance]]).

== Terminology and scope ==

== Terminology and scope ==

Line 62: Line 62:

}}</ref>

}}</ref>

The label ”Logical Phonology” (LP) is used for to emphasize a formally explicit rule calculus and a restricted inventory of operations.<ref name=”Reiss2021″ /><ref name=”BaleReiss2018″ /> In this usage, ”Substance-Free Logical Phonology” is often treated as LP pursued within a substance-free research program, but LP may also be used as a shorter label for work that combines both commitments.<ref name=”LRV2024″ /><ref name=”Reiss2021″ />

The label ”Logical Phonology” (LP) is used for to emphasize a formally explicit rule calculus and a restricted inventory of operations.<ref name=”Reiss2021″ /><ref name=”BaleReiss2018″>

⚫ ⚫ ⚫ ⚫ ⚫ ⚫ ⚫ ⚫ ⚫ ⚫

In this usage, ”Substance-Free Logical Phonology” is often treated as LP pursued within a substance-free research program, but LP may also be used as a shorter label for work that combines both commitments.<ref name=”LRV2024″ />

== Theoretical foundations ==

== Theoretical foundations ==

SFLP combines two theoretical commitments: (i) a ”substance-free” view that phonological computation does not access phonetic information, and (ii) a ”logical” theory that formalizes phonological rules and representations explicitly.<ref name=”HaleReiss2000“>{{cite journal

SFLP combines two theoretical commitments: (i) a ”substance-free” view that phonological computation does not access phonetic information, and (ii) a ”logical” theory that formalizes phonological rules and representations explicitly.<ref name=””>{{cite

|last1=Hale

|first1=Mark

⚫ ⚫

|title=”Substance Abuse” and “Dysfunctionalism”: Current Trends in Phonology

|journal=Linguistic Inquiry

|volume=31

|issue=1

|year=2000

|pages=157–169

|doi=10.1162/002438900554334

}}</ref><ref name=”Chabot2022″>{{cite journal

|last=Chabot

|first=Alex

|title=On substance and Substance-Free Phonology: Where we are at and where we are going

|journal=Canadian Journal of Linguistics / Revue canadienne de linguistique

|volume=67

|issue=4

|year=2022

|pages=429–443

|doi=10.1017/cnj.2022.37

}}</ref><ref name=”Reiss2017″>{{cite book

|last=Reiss

|last=Reiss

|first=Charles

|first=Charles

Line 103: Line 92:

|doi=10.4324/9781315675428-15

|doi=10.4324/9781315675428-15

|isbn=978-1-315-67542-8

|isbn=978-1-315-67542-8

}}</ref><ref name=”BaleReiss2018″>{{cite book

}}</ref><ref name=”BaleReiss2018″>

⚫ ⚫

|last2=Reiss

|first2=Charles

⚫ ⚫ ⚫ ⚫ ⚫

}}</ref><ref name=”Reiss2021″>{{cite journal

|last=Reiss

|first=Charles

|title=Towards a complete Logical Phonology model of intrasegmental changes

|journal=Glossa: A Journal of General Linguistics

|volume=6

|issue=1

|year=2021

|pages=107

|doi=10.16995/glossa.5886

|doi-access=free

}}</ref><ref name=”VolenecReiss2017″>{{cite journal

|last1=Volenec

|first1=Veno

|last2=Reiss

|first2=Charles

|title=Cognitive Phonetics: The Transduction of Distinctive Features at the Phonology–Phonetics Interface

|journal=Biolinguistics

|volume=11

|year=2017

|pages=251–294

|doi=10.5964/bioling.9089

⚫

=== Substance-free phonology ===

=== Substance-free phonology ===

In substance-free phonology, distinctive feature specifications and their organization into segments, are treated as formal symbolic abstractions for the purposes of computation.<ref name=”Reiss2017″ /><ref name=”Chabot2022″ /> The central claim is that phonological computation is ”insensitive” to phonetic substance, that is, that phonological rules apply without reference to articulatory, acoustic, or perceptual properties.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> Phonemic segments are then simply bundles of distinctive feature specifications. For example, a palatalization rule can be stated as operating only over the {{math|[±F]}} feature specification in the abstract representation, without reference to ease of articulation.

In substance-free phonology, distinctive feature specifications and their organization into segments, are treated as formal symbolic abstractions for the purposes of computation.<ref name=”Reiss2017″ /><ref name=”Chabot2022″ /> The central claim is that phonological computation is ”insensitive” to phonetic substance, that is, that phonological rules apply without reference to articulatory, acoustic, or perceptual properties. Phonemic segments are then simply bundles of distinctive feature specifications. For example, a palatalization rule can be stated as operating only over the {{math|[±F]}} feature specification in the abstract representation, without reference to ease of articulation.

On this view, phonetic interpretation is supplied by a separate component at the phonology–phonetics interface, which maps the output of phonological computation to representations usable by production and perception systems.<ref name=”VolenecReiss2017” /><ref name=”Reiss2017” /> As a result, explanations that appeal to phonetic “naturalness”, for example why certain patterns recur cross-linguistically, are often attributed to factors outside the synchronic grammar, such as diachronic change, biases in acquisition, or properties of speech planning and execution.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> Common cross-linguistic patterns such as final obstruent devoicing or place assimilation are treated as computable rule types whose frequency may reflect extra-grammatical pressures rather than being enforced by grammar-internal phonetic constraints.<ref name=”HaleReiss2000” /><ref name=”Reiss2017” />

On this view, phonetic interpretation is supplied by a separate component at the phonology–phonetics interface, which maps the output of phonological computation to representations usable by production and perception systems.<ref name=”VolenecReiss2017″ /> As a result, explanations that appeal to phonetic “naturalness”, for example why certain patterns recur cross-linguistically, are often attributed to factors outside the synchronic grammar, such as diachronic change, biases in acquisition, or properties of speech planning and execution.<ref name=”Reiss2017″ /> Common cross-linguistic patterns such as final obstruent devoicing or place assimilation are treated as computable rule types whose frequency may reflect extra-grammatical pressures rather than being enforced by grammar-internal phonetic constraints.<ref name=”HaleReiss2000″ />

Substance-free work therefore typically argues against treating substance-based markedness or naturalness constraints as primitives of phonological computation.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /><ref name=”Reiss2021″ /> In this view, the grammar can compute both phonetically natural and phonetically arbitrary rule types, and typological frequency is explained by extra-grammatical factors rather than by phonetic naturalness encoded in Universal Grammar.<ref name=”HaleReiss2000” /><ref name=”Reiss2017” />

Substance-free work therefore typically argues against treating substance-based markedness or naturalness constraints as primitives of phonological computation.<ref name=”Reiss2021″ /> In this view, the grammar can compute both phonetically natural and phonetically arbitrary rule types, and typological frequency is explained by extra-grammatical factors rather than by phonetic naturalness encoded in Universal Grammar.<ref name=”HaleReiss2000″ />

=== Logical phonology ===

=== Logical phonology ===

Logical phonology (LP) develops a formally explicit rule theory intended to make the structure of phonological computations precise, often using set-theoretic and logic-inspired definitions of representations and operations.<ref name=”BaleReiss2018″ /><ref name=”Reiss2021″ />

Logical phonology (LP) develops a formally explicit rule theory intended to make the structure of phonological computations precise, often using set-theoretic and logic-inspired definitions of representations and operations.<ref name=”Reiss2021″ />

In this approach, segments can be modeled as sets of feature-value specifications, and [[natural class]]es are defined in terms of logically stated conditions on those sets, such as membership conditions based on containing particular specifications.<ref name=”BaleReiss2018” /><ref name=”Reiss2021” />

In this approach, segments can be modeled as sets of feature-value specifications, and [[natural class]]es are defined in terms of logically stated conditions on those sets, such as membership conditions based on containing particular specifications.<ref name=”BaleReiss2018″ />

A central goal is to treat familiar generative rule schemata (e.g. {{math|A → B / X __ Y}}) as ”cover descriptions” that can be decomposed into (i) an explicit structural description that identifies a target and context and (ii) an explicit operation that determines the output representation.<ref name=”BaleReiss2018” /><ref name=”Reiss2021” />

A central goal is to treat familiar generative rule schemata (e.g. {{math|A → B / X __ Y}}) as ”cover descriptions” that can be decomposed into (i) an explicit structural description that identifies a target and context and (ii) an explicit operation that determines the output representation.<ref name=”BaleReiss2018″ />

Within LP, many rule effects are analyzed as compositions of simpler operations rather than as primitive ”feature change” statements; for example, intrasegmental alternations can be derived from formally defined operations that remove and add feature specifications, which are often modeled as ”subtraction” and ”unification” over feature sets.<ref name=”Reiss2021″ />

Within LP, many rule effects are analyzed as compositions of simpler operations rather than as primitive ”feature change” statements; for example, intrasegmental alternations can be derived from formally defined operations that remove and add feature specifications, which are often modeled as ”subtraction” and ”unification” over feature sets.<ref name=”Reiss2021″ />

Line 162: Line 119:

=== Natural classes ===

=== Natural classes ===

In set-theoretic implementations of SFLP, a natural class is defined as a set of segments whose feature sets satisfy a stated membership condition (for example, containing a particular specification).<ref name=”BaleReissShen2019″>{{cite journal |last1=Bale |first1=Alan |last2=Reiss |first2=Charles |last3=Shen |first3=David Ta-Chun |title=Sets, rules and natural classes: [ ] vs. { } |journal=Loquens |volume=6 |issue=2 |year=2019 |pages=e065 |doi=10.3989/loquens.2019.065}}</ref> This makes class membership a formally defined property of representations and provides an explicit basis for stating rule targets and contexts.<ref name=“BaleReissShen2019” />

In set-theoretic implementations of SFLP, a natural class is defined as a set of segments whose feature sets satisfy a stated membership condition (for example, containing a particular specification).<ref name=”BaleReissShen2019″>{{cite journal |last1=Bale |first1=Alan |last2=Reiss |first2=Charles |last3=Shen |first3=David Ta-Chun |title=Sets, rules and natural classes: [ ] vs. { } |journal=Loquens |volume=6 |issue=2 |year=2019 |pages=e065 |doi=10.3989/loquens.2019.065}}</ref> This makes class membership a formally defined property of representations and provides an explicit basis for stating rule targets and contexts. “”

For example, a segment like /n/ might be represented as the set {{math|{[+nasal], [+coronal], [−continuant], …}}}. A natural class such as “nasal consonants” can then be defined as the set of all segments whose feature sets include {{math|[+nasal]}}, formalized as:

<math>\{x \in S \mid \mathrm{[+nasal]} \in x\}</math>

<math>\{x \in S \mid \mathrm{[+nasal]} \in x\}</math>

where <math>S</math> is the segment inventory and <math>x</math> ranges over segments.

where <math>S</math> is the segment inventory and <math>x</math> ranges over segments.

Line 169: Line 125:

=== Phonological operations ===

=== Phonological operations ===

A central strategy in SFLP analyses is to model complex rule effects as compositions of simpler operations over feature sets. Processes traditionally analyzed as feature changes are derived from primitives such as feature deletion (set subtraction) and feature insertion (unification).<ref name=”Reiss2021″ /> Insertion is typically defined only when it yields a consistent set (that is, when it does not introduce a contradiction such as both {{math|[+F]}} and {{math|[−F]}} for the same feature).<ref name=”Reiss2021″ /> Treating feature change as composite rather than primitive is intended to restrict the inventory of expressible rule types and thereby constrain the hypothesis space available to learners and analysts.<ref name=”Reiss2021″ />

A central strategy in SFLP analyses is to model complex rule effects as compositions of simpler operations over feature sets. Processes traditionally analyzed as feature changes are derived from primitives such as feature deletion (set subtraction) and feature insertion (unification).<ref name=”Reiss2021″ /> Insertion is typically defined only when it yields a consistent set (that is, when it does not introduce a contradiction such as both {{math|[+F]}} and {{math|[−F]}} for the same feature). Treating feature change as composite rather than primitive is intended to restrict the inventory of expressible rule types and thereby constrain the hypothesis space available to learners and analysts.

For example, a rule that voices a consonant before a vowel would traditionally be written as changing {{math|[−voice]}} to {{math|[+voice]}}. In SFLP, this can be decomposed into (i) deletion of {{math|[−voice]}} from the target segment’s feature set and (ii) insertion of {{math|[+voice]}} into that set.<ref name=”Reiss2021″ />

Reiss proposes ”priority union” as an additional set operation that formalizes the contrast between feature-filling and feature-changing effects.<ref name=”Reiss2022″>{{cite journal |last=Reiss |first=Charles |title=Priority Union and Feature Logic in Phonology |journal=Linguistic Inquiry |volume=53 |issue=1 |pages=199–209 |year=2022 |doi=10.1162/ling_a_00400}}</ref> Priority union combines two feature sets while assigning precedence to one set in case of conflict; depending on which input has priority, the result can model adding specifications to an underspecified segment (feature-filling) or overriding an existing specification (feature-changing).<ref name=”Reiss2022″ />

Reiss proposes ”priority union” as an additional set operation that formalizes the contrast between feature-filling and feature-changing effects.<ref name=”Reiss2022″>{{cite journal |last=Reiss |first=Charles |title=Priority Union and Feature Logic in Phonology |journal=Linguistic Inquiry |volume=53 |issue=1 |pages=199–209 |year=2022 |doi=10.1162/ling_a_00400}}</ref> Priority union combines two feature sets while assigning precedence to one set in case of conflict; depending on which input has priority, the result can model adding specifications to an underspecified segment (feature-filling) or overriding an existing specification (feature-changing).<ref name=”Reiss2022″ />

Line 209: Line 164:

=== Learnability and constraints ===

=== Learnability and constraints ===

A stated consequence of limiting rule operations is that certain rule types may be excluded from the theory. In Logical Phonology, exchange rules (processes that swap feature values between segments) are argued to be ungenerable from the available operations.<ref name=”GormanReiss2025″ /> This has been linked to acquisition in the claim that learners will not posit exchange rules, because the grammatical formalism does not provide the resources to represent them.<ref name=”GormanReiss2025″ />

A stated consequence of limiting rule operations is that certain rule types may be excluded from the theory. In Logical Phonology, exchange rules (processes that swap feature values between segments) are argued to be ungenerable from the available operations. This has been linked to acquisition in the claim that learners will not posit exchange rules, because the grammatical formalism does not provide the resources to represent them.<ref name=”GormanReiss2025″ />

This restrictiveness is presented as a desirable property of the approach: if exchange rules are absent from natural language phonology (or extremely rare and learnable only through special mechanisms), then their exclusion from the formal system is treated as an explanatory success.<ref name=”GormanReiss2025″ /> More generally, the approach aims to derive typological gaps from formal limitations rather than from substantive constraints or functional pressures.<ref name=”GormanReiss2025″ />

This restrictiveness is presented as a desirable property of the approach: if exchange rules are absent from natural language phonology (or extremely rare and learnable only through special mechanisms), then their exclusion from the formal system is treated as an explanatory success. More generally, the approach aims to derive typological gaps from formal limitations rather than from substantive constraints or functional pressures.<ref name=”GormanReiss2025″ />

== Criticism and debate ==

== Criticism and debate ==

Line 222: Line 177:

Reiss and Volenec defend the position that features are innate components of Universal Grammar with fixed phonetic correlates accessed during transduction at the phonology–phonetics interface.<ref name=”ReissVolenec2022″ /> On this view, learners need only determine which features are contrastive in their language rather than discovering what features exist.

Reiss and Volenec defend the position that features are innate components of Universal Grammar with fixed phonetic correlates accessed during transduction at the phonology–phonetics interface.<ref name=”ReissVolenec2022″ /> On this view, learners need only determine which features are contrastive in their language rather than discovering what features exist.

Odden proposes an alternative “radical” substance-free approach in which feature inventories are not innately specified by Universal Grammar.<ref name=”Odden2022″>{{cite journal |last=Odden |first=David |title=Radical substance-free phonology and feature learning |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=500–551 |year=2022 |doi=10.1017/cnj.2022.10}}</ref> He argues that substance-free proposals require explicit accounts of feature learning and the interface to phonetics, addressing challenges such as how learners can converge on a coherent feature system without phonetic guidance if features are arbitrary symbols.<ref name=”Odden2022″ />

Odden proposes an alternative “radical” substance-free approach in which feature inventories are not innately specified by Universal Grammar.<ref name=”Odden2022″>{{cite journal |last=Odden |first=David |title=Radical substance-free phonology and feature learning |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=500–551 |year=2022 |doi=10.1017/cnj.2022.10}}</ref> He argues that substance-free proposals require explicit accounts of feature learning and the interface to phonetics, addressing challenges such as how learners can converge on a coherent feature system without phonetic guidance if features are arbitrary symbols.

Mielke documents cross-linguistic evidence of phonetically heterogeneous segment classes that pattern together phonologically, arguing that features cannot have fixed universal phonetic correlates if segments with different articulatory and acoustic properties form natural classes in different languages.<ref name=”Mielke2008″>{{cite book |last=Mielke |first=Jeff |title=The Emergence of Distinctive Features |publisher=Oxford University Press |year=2008 |isbn=9780199207916}}</ref> His Emergent Feature Theory proposes that features arise from phonological learning rather than being prespecified by Universal Grammar.<ref name=”Mielke2008″ />

Mielke documents cross-linguistic evidence of phonetically heterogeneous segment classes that pattern together phonologically, arguing that features cannot have fixed universal phonetic correlates if segments with different articulatory and acoustic properties form natural classes in different languages.<ref name=”Mielke2008″>{{cite book |last=Mielke |first=Jeff |title=The Emergence of Distinctive Features |publisher=Oxford University Press |year=2008 |isbn=9780199207916}}</ref> His Emergent Feature Theory proposes that features arise from phonological learning rather than being prespecified by Universal Grammar.

These disagreements reflect broader divisions in linguistics between rationalist and empiricist approaches to language acquisition and fundamentally different views about what Universal Grammar must contain.<ref name=”Chabot2022″ /><ref name=”Odden2022″ />

These disagreements reflect broader divisions in linguistics between rationalist and empiricist approaches to language acquisition and fundamentally different views about what Universal Grammar must contain.<ref name=”Chabot2022″ /><ref name=”Odden2022″ />

Line 230: Line 185:

=== The phonology–phonetics interface ===

=== The phonology–phonetics interface ===

While substance-free approaches agree that phonetic interpretation occurs outside phonological computation, they disagree about the nature of the interface itself. Volenec and Reiss propose that the interface operates through universal, deterministic transduction with fixed mappings from features to phonetic targets.<ref name=”VolenecReiss2017″ /> On this view, there is no language-specific phonetics. All cross-linguistic variation in phonetic realization traces to differences in phonological representation.<ref name=”VolenecReiss2017″ />

While substance-free approaches agree that phonetic interpretation occurs outside phonological computation, they disagree about the nature of the interface itself. Volenec and Reiss propose that the interface operates through universal, deterministic transduction with fixed mappings from features to phonetic targets.<ref name=”VolenecReiss2017″ /> On this view, there is no language-specific phonetics. All cross-linguistic variation in phonetic realization traces to differences in phonological representation.

Scheer challenges this position, arguing that the interface is arbitrary and language-specific, analogous to morphological spell-out rules such as “past tense ↔ -ed”.<ref name=”Scheer2022″>{{cite journal |last=Scheer |first=Tobias |title=What substance-free phonology means for substance-free phonology |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=444–469 |year=2022 |doi=10.1017/cnj.2022.31}}</ref> If features are truly arbitrary symbols during computation, Scheer argues, there is no principled basis for claiming their phonetic interpretation must be universal.<ref name=”Scheer2022″ />

Scheer challenges this position, arguing that the interface is arbitrary and language-specific, analogous to morphological spell-out rules such as “past tense ↔ -ed”.<ref name=”Scheer2022″>{{cite journal |last=Scheer |first=Tobias |title=What substance-free phonology means for substance-free phonology |journal=Canadian Journal of Linguistics / Revue canadienne de linguistique |volume=67 |issue=4 |pages=444–469 |year=2022 |doi=10.1017/cnj.2022.31}}</ref> If features are truly arbitrary symbols during computation, Scheer argues, there is no principled basis for claiming their phonetic interpretation must be universal.

This debate has implications for understanding phonetic variation, coarticulation, and the division of labor between phonology and phonetics in explaining speech patterns.<ref name=”Scheer2022″ /><ref name=”VolenecReiss2017″ />

This debate has implications for understanding phonetic variation, coarticulation, and the division of labor between phonology and phonetics in explaining speech patterns.<ref name=”Scheer2022″ /><ref name=”VolenecReiss2017″ />

Line 242: Line 197:

From the SFLP perspective, Hale and Reiss attribute cross-linguistic tendencies toward phonetically natural patterns to factors external to the phonological grammar, such as articulatory and perceptual biases that influence sound change and language acquisition.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> They introduce the term “dysfunctionalism” for their position that functional phonetic explanations are systematically redundant with independently motivated phonetic and diachronic explanations.<ref name=”HaleReiss2000″ /> Since grammars must be capable of encoding phonetically arbitrary processes (such as English intrusive-r or rules documented in languages like Southern Pomo), and any mechanism capable of generating arbitrary processes can also generate natural ones, they argue that adding substantive components to account for natural processes violates Occam’s Razor.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ />

From the SFLP perspective, Hale and Reiss attribute cross-linguistic tendencies toward phonetically natural patterns to factors external to the phonological grammar, such as articulatory and perceptual biases that influence sound change and language acquisition.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> They introduce the term “dysfunctionalism” for their position that functional phonetic explanations are systematically redundant with independently motivated phonetic and diachronic explanations.<ref name=”HaleReiss2000″ /> Since grammars must be capable of encoding phonetically arbitrary processes (such as English intrusive-r or rules documented in languages like Southern Pomo), and any mechanism capable of generating arbitrary processes can also generate natural ones, they argue that adding substantive components to account for natural processes violates Occam’s Razor.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ />

The phonological grammar itself is argued to be neutral, capable of encoding any pattern that can be formalized within its operations, with naturalness emerging from the historical and acquisitional contexts in which grammars develop.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ />

The phonological grammar itself is argued to be neutral, capable of encoding any pattern that can be formalized within its operations, with naturalness emerging from the historical and acquisitional contexts in which grammars develop.

Critics respond that the existence of arbitrary patterns does not necessarily demonstrate that phonetic substance plays no role in constraining or biasing the space of possible patterns, and that dismissing cross-linguistic tendencies as epiphenomenal may eliminate useful generalizations about human language.<ref name=”Hall2014″ />

Critics respond that the existence of arbitrary patterns does not necessarily demonstrate that phonetic substance plays no role in constraining or biasing the space of possible patterns, and that dismissing cross-linguistic tendencies as epiphenomenal may eliminate useful generalizations about human language.<ref name=”Hall2014″ />

Line 249: Line 204:

SFLP stands in opposition to Optimality Theory’s incorporation of substantive markedness constraints.<ref name=”HaleReiss2000″ /> Hale and Reiss argue that constraints encoding phonetic naturalness (such as constraints against voiced obstruents in coda position or requiring voicing agreement) conflate synchronic computation with diachronic and acquisitional explanations for typological patterns.<ref name=”HaleReiss2000″ />

SFLP stands in opposition to Optimality Theory’s incorporation of substantive markedness constraints.<ref name=”HaleReiss2000″ /> Hale and Reiss argue that constraints encoding phonetic naturalness (such as constraints against voiced obstruents in coda position or requiring voicing agreement) conflate synchronic computation with diachronic and acquisitional explanations for typological patterns.<ref name=”HaleReiss2000″ />

They further argue that Optimality Theory’s reliance on typological justification for its constraint set is circular: the theory aims to explain typological patterns by appealing to constraints whose existence and formulation are justified by those same patterns.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> Additionally, they note that Optimality Theory must accommodate phonetically arbitrary processes through arbitrary constraint formulations, which they argue reveals that substantive constraints are not necessary components of the theory.<ref name=”HaleReiss2000″ />

They further argue that Optimality Theory’s reliance on typological justification for its constraint set is circular: the theory aims to explain typological patterns by appealing to constraints whose existence and formulation are justified by those same patterns.<ref name=”HaleReiss2000″ /><ref name=”Reiss2017″ /> Additionally, they note that Optimality Theory must accommodate phonetically arbitrary processes through arbitrary constraint formulations, which they argue reveals that substantive constraints are not necessary components of the theory.

Proponents of phonetically grounded Optimality Theory respond that incorporating phonetic motivations into constraints captures genuine cross-linguistic generalizations and provides a more explanatory account of recurrent patterns than attributing them entirely to external factors.<ref name=”Hayes1999″>{{cite book |last=Hayes |first=Bruce |chapter=Phonetically Driven Phonology: The role of Optimality Theory and inductive grounding |title=Functionalism and Formalism in Linguistics, Volume 1: General Papers |editor-last=Darnell |editor-first=Michael |publisher=John Benjamins |year=1999 |pages=243–285 |isbn=9789027273857}}</ref>

Proponents of phonetically grounded Optimality Theory respond that incorporating phonetic motivations into constraints captures genuine cross-linguistic generalizations and provides a more explanatory account of recurrent patterns than attributing them entirely to external factors.<ref name=”Hayes1999″>{{cite book |last=Hayes |first=Bruce |chapter=Phonetically Driven Phonology: The role of Optimality Theory and inductive grounding |title=Functionalism and Formalism in Linguistics, Volume 1: General Papers |editor-last=Darnell |editor-first=Michael |publisher=John Benjamins |year=1999 |pages=243–285 |isbn=9789027273857}}</ref>

This may take 5 weeks or more, since drafts are reviewed in no specific order. There are 2,048 pending submissions waiting for review. If the submission is accepted, then this page will be moved into the article space. If the submission is declined, then the reason will be posted here. In the meantime, you can continue to improve this submission by editing normally. Where to get help If you need help editing or submitting your draft, please ask us a question at the AfC Help Desk or get live help from experienced editors. These venues are only for help with editing and the submission process, not to get reviews. If you need feedback on your draft, or if the review is taking a lot of time, you can try asking for help on the talk page of a relevant WikiProject. Some WikiProjects are more active than others so a speedy reply is not guaranteed. Improving your odds of a speedy review To improve your odds of a faster review, tag your draft with relevant WikiProject tags using the button below. This will let reviewers know a new draft has been submitted in their area of interest. For instance, if you wrote about a female astronomer, you would want to add the Biography, Astronomy, and Women scientists tags.

Approach to generative phonology combining substance-free representations with logic-inspired rule formalism

Substance-Free Logical Phonology (SFLP) is an approach to generative phonology that models phonological computation over segments represented as sets of distinctive features, with rule effects stated using formalisms from formal logic and set theory.^[1] SFLP is part of substance-free phonology, which holds that phonological computation is insensitive to phonetic substance (articulatory or acoustic properties).^[2] On this view, phonetic interpretation is handled at the phonology–phonetics interface, which maps the output of phonological computation to speech; explanations that appeal to phonetic substance are typically located outside the phonological grammar (e.g., diachrony, acquisition, or speech performance).

Several partially overlapping labels are used for work that combines a substance-free stance about phonological computation with an explicitly formal approach to phonological rules. The label Substance-Free Logical Phonology is used for this combined approach.^[3][4][5]

The broader label substance-free phonology is an umbrella term for a family of approaches that share a commitment to substance-free computation but differ in other theoretical assumptions.^[2] Within this family, Substance-Free Logical Phonology is a prominent position or cluster of closely related proposals.^[2][1] This particular approach to substance-free phonology is also sometimes referred to as the Concordia School for its association with proponents at Concordia University.^[6]

The label Logical Phonology (LP) is used for to emphasize a formally explicit rule calculus and a restricted inventory of operations.^[7][8]
In this usage, Substance-Free Logical Phonology is often treated as LP pursued within a substance-free research program, but LP may also be used as a shorter label for work that combines both commitments.^[3]

SFLP combines two theoretical commitments: (i) a substance-free view that phonological computation does not access phonetic information, and (ii) a logical theory that formalizes phonological rules and representations explicitly.^{[2][9][8][7][10]}

In substance-free phonology, distinctive feature specifications and their organization into segments, are treated as formal symbolic abstractions for the purposes of computation.^[9][2] The central claim is that phonological computation is insensitive to phonetic substance, that is, that phonological rules apply without reference to articulatory, acoustic, or perceptual properties. Phonemic segments are then simply bundles of distinctive feature specifications. For example, a palatalization rule can be stated as operating only over the [±F] feature specification in the abstract representation, without reference to ease of articulation.

On this view, phonetic interpretation is supplied by a separate component at the phonology–phonetics interface, which maps the output of phonological computation to representations usable by production and perception systems.^[10] As a result, explanations that appeal to phonetic “naturalness”, for example why certain patterns recur cross-linguistically, are often attributed to factors outside the synchronic grammar, such as diachronic change, biases in acquisition, or properties of speech planning and execution.^[9] Common cross-linguistic patterns such as final obstruent devoicing or place assimilation are treated as computable rule types whose frequency may reflect extra-grammatical pressures rather than being enforced by grammar-internal phonetic constraints.^[1]

Substance-free work therefore typically argues against treating substance-based markedness or naturalness constraints as primitives of phonological computation.^[7] In this view, the grammar can compute both phonetically natural and phonetically arbitrary rule types, and typological frequency is explained by extra-grammatical factors rather than by phonetic naturalness encoded in Universal Grammar.^[1]

Logical phonology (LP) develops a formally explicit rule theory intended to make the structure of phonological computations precise, often using set-theoretic and logic-inspired definitions of representations and operations.^[7]
In this approach, segments can be modeled as sets of feature-value specifications, and natural classes are defined in terms of logically stated conditions on those sets, such as membership conditions based on containing particular specifications.^[8]

A central goal is to treat familiar generative rule schemata (e.g. A → B / X __ Y) as cover descriptions that can be decomposed into (i) an explicit structural description that identifies a target and context and (ii) an explicit operation that determines the output representation.^[8]
Within LP, many rule effects are analyzed as compositions of simpler operations rather than as primitive feature change statements; for example, intrasegmental alternations can be derived from formally defined operations that remove and add feature specifications, which are often modeled as subtraction and unification over feature sets.^[7]

Because the inventory of primitive operations is restricted, proponents argue that the formalism can limit the hypothesis space of possible phonological rules and thereby bear on learnability and typological gaps.^[7]
For example, some work argues that exchange rules (rules that swap feature values between segments) are not generable from the available operations, with proposed implications for acquisition and typology.^[11]

SFLP describes phonological representations and rule application in an explicitly stated formalism in which distinctive feature specifications (e.g., [+nasal]) are treated as elements of sets, segments are modeled as sets of specifications, and rules are analyzed as functions that map one set-based representation to another.^[12][7] In set-theoretic implementations, natural classes are defined by membership conditions on segments’ feature sets, and complex rule effects are analyzed as compositions drawn from a restricted inventory of set operations.^[12][7]

In set-theoretic implementations of SFLP, a natural class is defined as a set of segments whose feature sets satisfy a stated membership condition (for example, containing a particular specification).^[12] This makes class membership a formally defined property of representations and provides an explicit basis for stating rule targets and contexts. For example, a segment like /n/ might be represented as the set {[+nasal], [+coronal], [−continuant], …}. A natural class such as “nasal consonants” can then be defined as the set of all segments whose feature sets include [+nasal], formalized as:
{ x ∈ S ∣ [ + n a s a l ] ∈ x } {\displaystyle \{x\in S\mid \mathrm {[+nasal]} \in x\}}
where S {\displaystyle S} is the segment inventory and x {\displaystyle x} ranges over segments.

A central strategy in SFLP analyses is to model complex rule effects as compositions of simpler operations over feature sets. Processes traditionally analyzed as feature changes are derived from primitives such as feature deletion (set subtraction) and feature insertion (unification).^[7] Insertion is typically defined only when it yields a consistent set (that is, when it does not introduce a contradiction such as both [+F] and [−F] for the same feature). Treating feature change as composite rather than primitive is intended to restrict the inventory of expressible rule types and thereby constrain the hypothesis space available to learners and analysts. For example, a rule that voices a consonant before a vowel would traditionally be written as changing [−voice] to [+voice]. In SFLP, this can be decomposed into (i) deletion of [−voice] from the target segment’s feature set and (ii) insertion of [+voice] into that set.

Reiss proposes priority union as an additional set operation that formalizes the contrast between feature-filling and feature-changing effects.^[13] Priority union combines two feature sets while assigning precedence to one set in case of conflict; depending on which input has priority, the result can model adding specifications to an underspecified segment (feature-filling) or overriding an existing specification (feature-changing).^[13]

Work in SFLP provides formal accounts of how rules interact with underspecified segments. Bale, Papillon, and Reiss develop a set-theoretic analysis that distinguishes feature-filling from feature-changing effects via differences in lexical representations and in how rule operations apply to specified versus unspecified targets.^[14]
If a segment lacks a specification for a feature, a rule can add the relevant value without conflict (feature-filling). For instance, if a vowel is underspecified for nasality and a rule inserts [+nasal] in the context of a following nasal consonant, the result is a filled specification. If the target is already specified with the opposing value (e.g., [−nasal]), the rule must first remove that specification before inserting [+nasal], yielding a feature-changing effect.^[14]

Some analyses also distinguish between a segmental position associated with an empty set of specifications and the complete absence of a segment in the representation.^[12] Underspecification therefore affects both the set of potential targets for a rule and the sequence of operations required for the rule to apply.^[14]

In some SFLP models, rule application is formulated as a search procedure over representations. Dabbous, Leduc, Reiss, and Shen argue that both local and long-distance effects can be derived by varying parameters of the search, including the natural classes that define targets, potential blockers, and stopping conditions that determine where the search terminates.^[15]
In their account, apparent adjacency restrictions can arise when intervening segments block the search because they fail to meet a rule’s structural description, producing an opacity-like effect. In the absence of such blocking, the same mechanism can yield long-distance dependencies such as vowel harmony across a word. The proposal is intended to unify local and non-local processes under a single formally specified mechanism rather than positing distinct rule types for each.^[15]

SFLP is typically formulated within a modular architecture in which phonological computation and phonetic implementation are distinct components: phonology is a domain-specific symbolic system that manipulates feature-based representations, while the sensorimotor systems responsible for speech production and perception lie outside the phonological grammar.^[10]

Volenec and Reiss develop an interface model (“Cognitive Phonetics”) that uses the notion of transduction as a deterministic mapping between symbolic representations and perceptuomotor systems. In this account, phonological computation outputs sequences of feature bundles (sets of feature–value specifications), which are mapped by transduction to what are termed “True Phonetic Representations”—temporally organized neuromuscular patterns executable by the motor system and relatable to perceptual input.^[10]

A consequence of this division of labor is that phonetic detail, such as gradient coarticulation and other implementation effects, is not treated as part of phonological computation. Instead, such effects are attributed to the interface and to external factors such as speech planning and execution, diachronic change, or acquisition, rather than to constraints stated in the synchronic phonological grammar itself.^[10]

Reiss and Volenec argue that distinctive features are universal and innate components of Universal Grammar and that their phonetic correlates are fixed by Universal Grammar and accessed at the interface during transduction; on this view, these correlates do not guide phonological computation itself, where features function as arbitrary symbols manipulated by the grammar.^[16]

An SFLP analysis of Votic vowel harmony models harmony using explicit rule operations and set-based representations.^[3] The analysis avoids appeals to markedness-based primitives, deriving patterns of harmony, transparency, and opacity from the interaction of feature specifications and formally defined spreading rules.
In this account, vowels described as transparent are those that lack the relevant feature specification, allowing the harmonic feature to spread across them. Opaque vowels are specified with a conflicting value and block spreading. The analysis shows how these patterns can be modeled using basic operations such as feature deletion and insertion, together with underspecification in lexical representations.^[3]

A Logical Phonology model of intrasegmental alternations derives such alternations from combinations of subtraction (feature deletion) and unification (feature insertion), rather than treating feature change as a primitive operation.^[7] The stated goal is to constrain the space of possible rule types by limiting the inventory of primitive operations. In this model, processes affecting a single feature of a segment, such as vowel reduction or consonant lenition, are analyzed as sequences of feature deletion and insertion rather than as atomic operations.^[7]

A stated consequence of limiting rule operations is that certain rule types may be excluded from the theory. In Logical Phonology, exchange rules (processes that swap feature values between segments) are argued to be ungenerable from the available operations. This has been linked to acquisition in the claim that learners will not posit exchange rules, because the grammatical formalism does not provide the resources to represent them.^[11]
This restrictiveness is presented as a desirable property of the approach: if exchange rules are absent from natural language phonology (or extremely rare and learnable only through special mechanisms), then their exclusion from the formal system is treated as an explanatory success. More generally, the approach aims to derive typological gaps from formal limitations rather than from substantive constraints or functional pressures.^[11]

SFLP is discussed within broader debates about the role of phonetic substance and functional explanation in phonological theory. The approach has generated both support and criticism from phonologists with different theoretical commitments.

A major division within substance-free phonology concerns whether features are innate and universal or emergent and language-specific.^[2] Hall surveys positions on feature content and argues for a “moderate” role for substance in understanding feature content, contrasting fully substance-free and substance-informed approaches.^[17]

Reiss and Volenec defend the position that features are innate components of Universal Grammar with fixed phonetic correlates accessed during transduction at the phonology–phonetics interface.^[16] On this view, learners need only determine which features are contrastive in their language rather than discovering what features exist.

Odden proposes an alternative “radical” substance-free approach in which feature inventories are not innately specified by Universal Grammar.^[18] He argues that substance-free proposals require explicit accounts of feature learning and the interface to phonetics, addressing challenges such as how learners can converge on a coherent feature system without phonetic guidance if features are arbitrary symbols.

Mielke documents cross-linguistic evidence of phonetically heterogeneous segment classes that pattern together phonologically, arguing that features cannot have fixed universal phonetic correlates if segments with different articulatory and acoustic properties form natural classes in different languages.^[19] His Emergent Feature Theory proposes that features arise from phonological learning rather than being prespecified by Universal Grammar.

These disagreements reflect broader divisions in linguistics between rationalist and empiricist approaches to language acquisition and fundamentally different views about what Universal Grammar must contain.^[2][18]

While substance-free approaches agree that phonetic interpretation occurs outside phonological computation, they disagree about the nature of the interface itself. Volenec and Reiss propose that the interface operates through universal, deterministic transduction with fixed mappings from features to phonetic targets.^[10] On this view, there is no language-specific phonetics. All cross-linguistic variation in phonetic realization traces to differences in phonological representation.

Scheer challenges this position, arguing that the interface is arbitrary and language-specific, analogous to morphological spell-out rules such as “past tense ↔ -ed”.^[20] If features are truly arbitrary symbols during computation, Scheer argues, there is no principled basis for claiming their phonetic interpretation must be universal.

This debate has implications for understanding phonetic variation, coarticulation, and the division of labor between phonology and phonetics in explaining speech patterns.^[20][10]

Hall raises questions about what explanatory work remains for phonology if phonetic substance is excluded from phonological computation and how phonetic regularities should be accounted for.^[21] He suggests that while substance-free approaches correctly note that phonetic properties alone do not determine phonological patterning, some role for phonetic content in constraining or motivating patterns may be warranted. Hall questions whether a completely substance-free grammar can adequately explain why certain phonetically natural patterns recur across languages.^[21] In later work, Hall similarly argues that substance-free approaches can be right to reject phonetics as a direct determinant of phonological patterning while still allowing a limited role for substance in theorizing feature content.^[17]

From the SFLP perspective, Hale and Reiss attribute cross-linguistic tendencies toward phonetically natural patterns to factors external to the phonological grammar, such as articulatory and perceptual biases that influence sound change and language acquisition.^[1][9] They introduce the term “dysfunctionalism” for their position that functional phonetic explanations are systematically redundant with independently motivated phonetic and diachronic explanations.^[1] Since grammars must be capable of encoding phonetically arbitrary processes (such as English intrusive-r or rules documented in languages like Southern Pomo), and any mechanism capable of generating arbitrary processes can also generate natural ones, they argue that adding substantive components to account for natural processes violates Occam’s Razor.^[1][9]

The phonological grammar itself is argued to be neutral, capable of encoding any pattern that can be formalized within its operations, with naturalness emerging from the historical and acquisitional contexts in which grammars develop.

Critics respond that the existence of arbitrary patterns does not necessarily demonstrate that phonetic substance plays no role in constraining or biasing the space of possible patterns, and that dismissing cross-linguistic tendencies as epiphenomenal may eliminate useful generalizations about human language.^[21]

SFLP stands in opposition to Optimality Theory’s incorporation of substantive markedness constraints.^[1] Hale and Reiss argue that constraints encoding phonetic naturalness (such as constraints against voiced obstruents in coda position or requiring voicing agreement) conflate synchronic computation with diachronic and acquisitional explanations for typological patterns.^[1]
They further argue that Optimality Theory’s reliance on typological justification for its constraint set is circular: the theory aims to explain typological patterns by appealing to constraints whose existence and formulation are justified by those same patterns.^[1][9] Additionally, they note that Optimality Theory must accommodate phonetically arbitrary processes through arbitrary constraint formulations, which they argue reveals that substantive constraints are not necessary components of the theory.

Proponents of phonetically grounded Optimality Theory respond that incorporating phonetic motivations into constraints captures genuine cross-linguistic generalizations and provides a more explanatory account of recurrent patterns than attributing them entirely to external factors.^[22]

De Lacy characterizes the broader program of Hale and Reiss as a strongly modular, rule-based, and rationalist approach to phonological theory.^[23] He notes that it represents a return to certain aspects of classical generative phonology, updated with contemporary formal tools. The approach stands in contrast to frameworks such as Optimality Theory, which incorporate substantive constraints and parallel evaluation mechanisms.^[23] The broader program has also been discussed in independent reviews outside the journal literature, including a review of *The Phonological Enterprise* in the LINGUIST List.^[24]

SFLP shares with Blevins’s Evolutionary Phonology a rejection of substantive constraints in synchronic grammar and an emphasis on explaining typological patterns through phonetics and diachrony.^[25] However, the frameworks differ fundamentally in their commitments to modularity and innate linguistic structure: SFLP maintains a formal, modular computational system as part of Universal Grammar, while Evolutionary Phonology rejects domain-specific grammatical machinery in favor of usage-based emergence.^[25]

Proponents of SFLP argue that its rigor and restrictiveness offer a more scientifically tractable theory of phonology, one that can be integrated into broader cognitive architecture alongside syntax and other symbolic systems.^[8] Critics question whether the exclusion of phonetic substance eliminates useful explanatory resources and whether the formal system as currently developed can handle the full range of phonological phenomena, including prosodic structure, tone, and gradient effects.^[21][18]

The theory’s treatment of typological patterns remains contentious. SFLP attributes typological asymmetries (such as the rarity of word-final voicing processes compared to word-final devoicing) to factors outside the grammar, particularly perceptual biases and sound change.^[1][9] Critics argue this approach may underpredict the systematicity of cross-linguistic tendencies and make the theory less empirically constraining.^[21]

Category:Phonological theories
Category:Generative linguistics