english obstruent inventory

Upload: eunson

Post on 04-Apr-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 English Obstruent Inventory

    1/24

    - 1 -

    The Development of English Obstruent Inventory: An

    Optimality-Theoretic Account

    Chang Yong Sohn

    (Seoul National University)

    Chang Yong Sohn. 2010. The Development of English Obstruent Inventory: An

    Optimality-Theoretic Account. Studies in Modern Grammar XX, XXX-XXX.

    The historical origin of obstruent phonemes in English traces back to the

    Indo-European system. This paper attempts to account for two important

    developments (Grimm's Law and Middle English phonemicization) within the

    framework of Optimality Theory. The proposal is crucially based on the

    rejection of a serial interpretation of Grimm's Law and the substitution of the

    Maintain Contrast constraint with the interactions between symmetry in the

    system and the minimality of phonological change. In addition symmetry is

    claimed to function as a guiding principle whose manifestation can be extended

    to Middle English phonemicization. It is also shown that the application of

    nonsystemic faithfulness constraints to the systemic domain enables us to state

    a formal definition of the types of phonemic change.

    Key words: Grimm's Law, symmetry, phonemicization, chain shift, phoneme

    inventory

    1. Introduction

    The systemic change in the phoneme inventory tends to be marginalized in

    mainstream Optimality Theory (McCarthy 2002, 2008, Prince and

    Smolensky 2004). As pointed out in McCarthy (2002: 226), there existfundamental differences between evaluating systems or grammars and

    evaluating linguistic forms. System evaluation involves systemwide

    generalizations of a set of forms whereas form evaluation is restricted to

    specific forms. As a model for incorporating phonemic contrast at the

  • 7/30/2019 English Obstruent Inventory

    2/24

    - 2 -

    system level, Flemming (2002) proposes a systemic approach to synchronic

    inventory structure, building upon the dispersion theory. Recently, Grimm's

    Law, which denotes the change in the obstruent phoneme inventory from

    Indo-European (IE) to Germanic, has also been analyzed in terms of a

    constraint that controls the contrast between the input and the output

    (Petrova 2000, Ahn 2004, Lee 2007).

    Another systemic principle in the phoneme inventory is symmetry

    between phonemes. A more balanced system with a symmetrical distribution

    of phonemes within an inventory is preferred to an imbalanced or

    asymmetrical system and sound change is often driven by or tends toward a

    symmetrical system (Saussure 1959, Bynon 1977, Hock 1991, McMahon

    1994, Campbell 1999, Sohn 2009). In the present paper, we propose an

    account within the framework of Optimality Theory (OT) of the changes in

    the English obstruent inventory in which the notion of symmetry is directly

    encoded.

    2. Obstruent Inventories

    This section introduces the obstruent phoneme inventories from IE to

    Middle English (ME) and brings to the fore two important transitions

    between inventories which serve as the focus of our analysis.

    2.1 Phoneme Inventories from Indo-European to Middle English1)

    (1) IE Obstruent Inventory2)

    1) The major sources for the inventories are Scargill (1951), Voyles (1992) and

    Ringe (2006) for IE and Germanic and Campbell (1959), Brunner (1963), Pylesand Algeo (1993) and Millward (1996) for OE and ME.

    2) The glottalic theory proposed by a group of scholars (e.g., Emonds 1972,

    Hopper 1973, Kortlandt 1985) contrasts with the traditional view adopted here in

    that it denies the existence of voiced aspirates, with the IE obstruent series being

    reconstructed as voiced stops, voiceless stops and voiceless glottals (D, T, T').

    The controversy regarding the reconstruction of the IE consonant inventory has

  • 7/30/2019 English Obstruent Inventory

    3/24

    - 3 -

    Manner Place labial alveolar Palatal velar

    Stopsvoiceless p t k

    voiced b d g

    Fricativesvoiceless s

    voiced

    Aspiratesvoiceless

    voiced bh dh gh

    (2) Germanic

    Manner Place labialInterdental

    / alveolarPalatal velar

    Stopsvoiceless p t k

    voiced b d g

    Fricativesvoiceless f /s h

    voiced

    (3) Old English (OE)

    not been definitively resolved in favor of either view. However, Vennemann

    (2006) provides evidence supporting the traditional view based on the

    phonological adaptation of Semitic loanwords in Germanic. Calabrese and Halle

    (1998) also defend the traditional view on the grounds that the rarity of voiced

    aspirates without voiceless counterparts alone is not a sufficient reason for

    rejecting it. In fact, as Garret (1991) observes, the glottalic system entails highly

    unnatural developments in various daughter languages.

  • 7/30/2019 English Obstruent Inventory

    4/24

    - 4 -

    Manner Place labialInterdental

    / alveolarPalatal velar

    Stopsvoiceless p t k

    voiced b d g

    Fricativesvoiceless f /s h

    voiced

    Affricatesvoiceless

    voiced

    (4) Middle English (ME)

    Manner Place labialInterdental

    / alveolarPalatal velar

    Stopsvoiceless p t k

    voiced b d g

    Fricativesvoiceless f /s h

    voiced v /z

    Affricates voiceless

    voiced

    A cursory glance at the inventories reveals two general characteristics that

    hold throughout the developments. First, except for the change from IE to

    Germanic, the number of phonemes constantly increases with each stage. In

    addition, the phoneme inventory tends towards a more "balanced" system.

    Thus, as time passes, gaps in the inventory are filled and the distribution of

    the filled gaps shows a consistent pattern.

    Note that the phoneme inventories given in (1~4) presuppose the

    hypothetical and nonlinguistic demarcation of the periods. In fact, the actual

    phoneme inventories may vary depending on the specific period (e.g., early

    or late OE) under consideration. Moreover, the time depth from IE to ME

    exceeds 5,000 years. Thus, a substantial generalization is likely inevitable to

  • 7/30/2019 English Obstruent Inventory

    5/24

    - 5 -

    elicit an inventory for each period.

    Among the changes in the phoneme inventories of English obstruents, we

    will focus on two transitions: one from IE to Germanic and the other from

    OE to ME. We limit ourselves to these two transitions because they

    represent the typical changes attested in phoneme inventories, that is, a

    chain shift and a phoneme split. In addition, to explore how a chain shift

    and a phoneme split (or merger), which are not amenable to formalization

    in traditional theories, can be interpreted in Optimality Theory is by itself of

    theoretical significance.

    2.2 Grimm's Law and ME phonemicization

    The shift from IE in (1) to Germanic in (2) is better known as Grimm's

    Law (or The First Consonant Shift) and the shift from OE in (3) to ME (4)

    is known as the ME phonemicization of voiced fricatives. These two shifts

    differ in several respects even though both of them result in a change in the

    phoneme inventory.

    First, as shown in 2.1, Grimm's Law (GL) incurs no change in thenumber of total phonemes, whereas with the ME phonemicization of voiced

    fricatives (MEP) the number of phonemes increases. GL involves a chain

    shift in which a series of changes are interleaved with each other and MEP

    is a phoneme split in which conditioned allophones become contrastive. In

    other words, while GL is a sound change by itself, MEP is the result of a

    group of phonological processes.

    Second, the triggering elements of GL are quite different from those of

    MEP. Aside from language-internal causes (as discussed here), the

    language-external or "ultimate" causes of GL range from highly

    psychological to ethnological explanations, most of which are purely

    stipulative and, thus, easily reputable (Prokosch 1939, Hock and Joseph

    1996, Campbell 1999). On the other hand, MEP is often claimed to be

    triggered by a variety of language-internal sound changes that occurred

    during the time span from the late OE to the early ME (Millward 1996,

  • 7/30/2019 English Obstruent Inventory

    6/24

    - 6 -

    Laker 2009).3)

    3. Review of Previous Analyses

    3.1 Previous OT accounts of Grimm's Law

    Grimm's Law, as a whole, refers to the three changes depicted in (5) and

    (6).

    (5) IE Germanic

    a. b, d, g p, t, k

    b. p, t, k f, , h4)

    c. bh, dh, g b, d, g

    s s

    (6) a. Devoicing: b, d, g p, t, k

    b. Spirantization: p, t, k f, , h

    c. Deaspiration: bh, dh, gh b, d, g

    Several analyses have been conducted within the framework of OT to

    account for GL (Petrova 2000, Ahn 2003, and Lee 2007). Despite the

    different assumptions on the trigger and the intermediate stages of GL, they

    have much in common with respect to having a perception-based constraint

    at the core of their analysis. Building on the functional constraints proposed

    by Flemming (1996, 2002, 2005), which in turn draws on the adaptive

    dispersion theory, they employ a constraint that forces the contrast in the

    3)

    French loans, dialect mixtures with Kentish, final vowel deletion, weakening infunction words (Millward 1996) and Celtic influence (Laker 2009) all lead to the

    neutralization of conditioning environments and, as a result, voiced fricatives

    become contrastive. Teedle (2006) offers an OT account of MEP in terms of

    misperception in the transmission from speaker to listener.

    4) As is customary in most accounts, minor changes in the place of articulation

    (labials to bilabials and alveolars to interdentals) are disregarded.

  • 7/30/2019 English Obstruent Inventory

    7/24

    - 7 -

    input to be preserved in the output. The definition of the constraint is given

    in (7).5)

    (7) Maintain Contrast (Ahn 2003: 52)

    The phonemic contrast of the input should be maintained in the

    output.

    When applied to GL, the constraint in (7) prohibits neutralization among

    phonemes. At the same time, it prevents phoneme splits. Without a merger

    or a split in the phoneme inventory, the number of phonemes is maintained

    and the phonemic contrast of the input remains intact.

    Another property that all previous OT analyses have in common is that

    they view GL as a set of intermediate stages, each of which is activated

    stepwise by the constraint banning the loss of contrast. Although they all

    interpret GL as a series of steps, these analyses radically differ in terms of

    the order and content of each step. The stages assumed in each account are

    summarized in (8).6)

    (8) Intermediate Stages of Grimm's Law

    5) A more concrete constraint (3-way VOT CONTR) which prefers the 3-way

    contrast in stops in Petrova (2000) and PreserveContrast in Lee (2007) play a

    role similar to that in (7).

    6) There are other versions of intermediate stages (e.g., Prokosch 1939, Cser

    1994), which makes any attempts to derive a unified serial account more unlikely.

  • 7/30/2019 English Obstruent Inventory

    8/24

    - 8 -

    Stages Petrova (2000) Ahn (2004) Lee (2007)

    1b, d, g

    p, t, k

    bh, dh, gh

    b, d, g

    bh, dh, gh

    , ,

    2p, t, k

    ph, th, kh

    b, d, g

    p, t, k

    p, t, k

    f, , x

    3ph, th, kh

    f, , h

    p, t, k

    ph, th, kh

    b, d, g

    p, t, k

    4 b

    h

    , d

    h

    , g

    h

    b, d, gp

    h

    , t

    h

    , k

    h

    f, , h

    , ,

    b, d, g

    The differences in the posed stages in (8) appear to result from their

    assumptions pertaining to two issues with regard to GL. One issue centers

    on the trigger of GL, whereas the other assesses whether or not the notion

    of Germanic enhancement (aspiration) as proposed by Iverson and Salmons

    (2003) is relevant to GL. In Petrova's account, GL is initiated by marked

    (word-initial) voiced stops and in Ahn and Lee, it is triggered by marked

    voiced aspirates. As argued in Ahn, Petrova's treatment of word-initialvoiced stops as the initiator of GL faces a serious problem when we

    consider the fact that relatively more marked voiced aspirates (bh, dh, gh)

    are deaspirated at a later stage. In the other two accounts, the markedness

    of voiced aspirates functions as a trigger of GL.7)

    The inclusion of Germanic enhancement as a viable intermediate stage, as

    in Petrova and Ahn, appears to lack any substantive evidence. If we assume

    Germanic enhancement as one stage of GL, a conflict arises with the end

    result of GL in which aspirated obstruents, voiced or voiceless, disappear

    from the subsequent inventories.

    To summarize, the intermediate stages posited in previous accounts are

    too varied for any convergence to be possible. In addition, as regards a

    7) Kim (2004) also argues that Grimm's Law begins with the deaspiration of

    marked voiced aspirates and that successive changes should be regarded as a

    push-chain rather than a pull-chain.

  • 7/30/2019 English Obstruent Inventory

    9/24

    - 9 -

    change in two prehistoric languages (IE and Germanic), finding convincing

    evidence that supports the existence of intermediate stages independent of

    language-internal and ad-hoc arguments is challenging. The dubious status of

    detailed stages of GL can be highlighted by the following excerpt:

    (9) Hock and Joseph (1996: 135)

    For Grimm's Law, for instance, it is possible to cook up three

    or four different scenarios, all of them chain shifts. (Some of

    them may be more likely than others, but which of them

    actually took place remains anybody's guess.)

    Viewing GL as a series of stages also raises a theoretical problem as to

    how it can fit into the OT framework. Ahn (2004) assumes the same

    ranking between identical sets of constraints throughout all stages, with each

    stage having as its input the output of the previous stage. Lee's (2007)

    account, on the other hand, uses reranking as well as a different input for

    each stage.8) Under the basic principles and architecture of OT, reranking

    and lexicon optimization are legitimate formal mechanisms of sound changes

    (Holt 2003). Thus, in any version of a multiple-stage model of GL, the

    interactions between reranking and lexicon optimization should be spelled

    out. However, neither Ahn nor Lee provides testable and precise

    assumptions regarding this point.

    3.2 Maintain Contrast

    The theoretical importance of the constraint Maintain Contrast (MC) (or

    PreserveContrast in Lee (2007)) as defined and employed to account for GL

    is linked to its abstractness. First, it does not specify particular features of

    individual phonemes, for instance, [asp] or the noise frequency. As long as

    the number of the input contrasts is preserved, the output inventory satisfies

    8) Petrova (2000) does not offer constraint interactions for stages other than the

    first stage of Grimm's Law.

  • 7/30/2019 English Obstruent Inventory

    10/24

    - 10 -

    the constraint. As Flemming (2005: 71) observes, this very nature of

    Maintain Contrast complements the fundamental restriction on faithfulness

    constraints which only block but cannot motivate change. In the case of GL,

    the deaspiration of marked voiced aspirates (bh, dh, gh) yields voiced stops

    (b, d, g), which are already present in the phoneme inventory. Without the

    Maintain Contrast constraint which prohibits mergers between phonemes, no

    further changes are necessary. And the actual changes are governed by the

    interactions between the markedness and faithfulness constraints.

    On the other hand, this abstractness may allow a massive change from

    the set (b, d, g, p, t, k, bh, dh, gh) to the set (, , , f, , h, b, d, g) in

    the phoneme inventory. Moreover, if we interpret the Maintain Contrast

    constraint as sensitive to more specific levels such as the content of the

    contrasts, then GL cannot be regarded as a chain shift in which the

    phonemic contrast of the input is preserved in the output. In IE the

    phonemes contrast with each other in a four-way fashion: voiced aspirates

    (bh, dh, gh), voiced stops (b, d, g), voiceless stops (p, t, k) and a voiceless

    fricative (s). In Germanic, however, to which GL applies, there is actually a

    three-way distinction: voiced stops (b, d, g), voiceless stops (p, t, k), and

    voiceless fricatives (f, , s, h). Thus, in order for the Maintain Contrast

    constraint to work properly, we should assume that the maintenance of the

    contrasts is equivalent to the maintenance of the number of phonemes.

    The MC constraint as defined in (7) differs from that originally proposed

    by (Flemming 1996) in two important aspects. First, in most cases,

    Flemming's MC does not regulate the relationship between the input and the

    output. In particular, MC with respect to the phoneme inventory is restricted

    to the output evaluation. Second, according to Flemming's theory, the MC

    constraint is simply a formal implementation of the functional goal tomaximize the number of contrasts and it never functions as an active

    constraint that evaluates the faithfulness between the input and the output.

    (10) a. Functional goal (Flemming 2005: 4)

  • 7/30/2019 English Obstruent Inventory

    11/24

    - 11 -

    Maximize the number of contrasts

    b. Formal implementation (Flemming 1996: 89)

    Maintain 1 contrast >> Maintain 2 contrasts >> ...

    >> Maintain n contrasts

    Thus, given its rather restricted interpretation and somewhat deviant

    application, we find the MC constraint as in (7) to be marginally relevant at

    best to the account of GL.

    3.3 ME Phonemicization of Voiced Fricatives

    Sohn (2009) proposes the following set of constraints (11) and the ranking

    (12) for the ME phonemicization of voiced fricatives.

    (11) Constraints9)

    a. Symmetry (SM)

    Obstruents should be symmetric with regard to voice.

    b. DEPOutput phonemes must have input correspondents.

    c. MAX

    Input phonemes must have output correspondents.

    d. *Voiced/Velar-Fricative (*V/VF)

    Velar fricatives must not be voiced.

    (12) a. OE ranking

    DEP, MAX, *V/VF SM

    9) Apparently, applying the concept of correspondence reserved for segments in

    OT (Kager 1999, McCarthy 2002, 2008) to phonemes in an inventory is a

    novelty. In particular, the formal significance of the related constraints such as

    LINEARITY or CONTIGUITY is difficult to determine in this domain. However,

    it is obvious that the input phonemes are not wholly independent of the output

    phonemes and that some version of faithfulness constraints should be invoked to

    express the relationship between them.

  • 7/30/2019 English Obstruent Inventory

    12/24

    - 12 -

    b. ME ranking

    MAX, *V/VF SM DEP

    The change as a result of the ME phonemicization of voiced fricatives can

    be described in (13) and (14), in which the interactions between the

    markedness constraints (SM and *V/VF) and the faithfulness constraints

    (MAX and DEP) are shown.

    (13) OE phoneme inventory (fricatives)

    f, /s, , h DEP MAX *V/VF SM

    a. f, /s, , h *****

    b. f, /s, , h

    v, /z, , *!**** *

    c. f, /s, , h

    v, /z, *!*** *

    d. f *!*** *

    e. fv

    *! ****

    f. f, /s,

    v, /z, *!*** *

    (14) ME phoneme inventory (fricatives)

  • 7/30/2019 English Obstruent Inventory

    13/24

    - 13 -

    f, /s, , h MAX *V/VF SM DEP

    a. f, /s, , h *!****

    b. f, /s, , h

    v, /z, , *! *****

    c. f, /s, , h

    v, /z, * ****

    d. f *!*** *

    e. f

    v*!*** *

    f. f, /s,

    v, /z, *! ****

    The crucial constraint in this approach is SM, which evaluates an

    obstruent set with voiceless and voiced obstruents as more harmonic than

    those without either voiceless or voiced obstruents. In addition, DEP and

    MAX regulate the insertion of a phoneme (split) and the deletion of a

    phoneme (merger).

    4. Symmetry and Minimality

    4.1 Grimm's Law

    A new account of GL builds on two assumptions. First, to circumvent the

    problem of unsubstantiated intermediate stages, we interpret GL as a single

    change. This type of parallel interpretation has been well documented in

    numerous works (King 1969, Hock 1991, Hock and Joseph 1996, Calabrese

    and Halle 1998, Noske 2009).10) Second, the Maintain Contrast constraint,

    whose definition and implementation are not firmly grounded, must be

    substituted by more justifiable constraints.

    10) Unlike other accounts, Noske (2009) adopts the glottalic theory and proposes

    an analysis which combines Grimm's Law with Verner's Law.

  • 7/30/2019 English Obstruent Inventory

    14/24

    - 14 -

    To capture the nature of a chain shift, we assume the following

    constraints in addition to DEP and MAX and the ranking between them.

    (15) Constraints

    a. *ASP11)

    Aspiration must not be allowed

    b. IDENT(F)

    The feature (F) of the output segments must be identical

    with that of the input correspondents.

    (16) *ASP >> MAX, DEP >> IDENT(F)

    The ranking in (16) guarantees that the voiced aspirates of IE become

    something else while the number of phonemes remains the same. Thus, any

    candidate with a merger or with aspirated phonemes involves more

    violations than a candidate with no merger and no aspirated phonemes. This

    is shown in (17).

    (17)

    11) Since Lass and Anderson (1975) the idea that *ASP is the trigger of Grimm's

    Law has been prevalent in the literature (Calabrese and Halle 1998, Ahn 2004,

    Lee 2007, among others).

  • 7/30/2019 English Obstruent Inventory

    15/24

    - 15 -

    b, d, g

    p, t, k

    bh, dh, gh

    s

    *ASP MAX DEPIDENT

    (F)

    a.No

    Change*!**

    b. DA *!** ***

    c. DV *!** *** ***

    d. SP *!** ***

    e. GL******

    ***

    For the sake of exposition, the candidates in (17) are named after the

    process involved. The actual candidates are given below.

    (18) a. No Change

    b, d, gp, t, k

    bh, dh, gh

    s

    b. Deaspiration (DA) only

    change: bh, dh, gh b, d, g

    phoneme inventory: b, d, g

    p, t, k

    s

    c. Devoicing (DV) only

    change: b, d, g p, t, k

    phoneme inventory: p, t, k

    bh, dh, gh

    sd. Spirantization (SP) only

  • 7/30/2019 English Obstruent Inventory

    16/24

    - 16 -

    change: p, t, k f, , h

    phoneme inventory: b, d, g

    f, , h

    bh, dh, gh

    s

    e. Grimm's Law (DA, DV, SP all applied in one sweep)

    change: bh, dh, gh b, d, g / b, d, g p, t, k /

    p, t, k f, , h

    phoneme inventory: p, t, k

    f, , h

    b, d, g

    s

    The ranking in (16) triggers the elimination of voiced aspirates from the IE

    inventory but maintains the number of phonemes.

    At this stage, we have to deal with how voiced aspirates turn into voiced

    stops and nothing else and how other phonemes are changed into the

    desired outputs. In order to account for the actual changes, we propose the

    two additional constraints given in (19).

    (19) a. MINIMALITY (MN)12)

    More than one feature cannot be affected.

    b. SYMMETRY (SM)

    SM(stops) >> SM(fricatives)

    (19a) assigns a violation mark if the change involves more than one feature

    of the input. As pointed out by Picard (1999), the insight behind the MNconstraint is that any sound change should be gradual, affecting one feature

    12) Admittedly, the constraint as it stands in (18) does not fit well into the usual

    formation of OT constraints despite the fact that its motivation is well justified.

    In the current OT, there is no formal mechanism established to measure the

    degree of deviation from the identity between the input and the output.

  • 7/30/2019 English Obstruent Inventory

    17/24

    - 17 -

    at a time. Bynon (1999) offers a remark on the general principle motivating

    the MN constraint.

    (20) Bynon (1977: 86, recited from Picard 1999)

    The first of these requirements, namely that changes in the

    realization of particular sounds must be small enough for

    speakers using both the old and new realizations still to be

    able to recognize lexical items, is self-evident. This is why it is

    usual for change to proceed in small steps which involve the

    alteration of only one feature at a time, ...

    (19b) is a subset of the SM constraint of (11a). It requires that the

    symmetry of voice be enforced differently depending on the categories of

    the obstruents. The symmetry of stops are preferable to the symmetry of

    fricatives.13)

    (21) MN, SM(stops) >> SM(fricatives)14)

    13) For expository reasons, candidates violating *ASP >> MAX, DEP >>

    IDENT(F) and with [s] are not included in the tableau.

    14) The phonemes that undergo a change are marked with shading in (21).

  • 7/30/2019 English Obstruent Inventory

    18/24

    - 18 -

    b, d, g

    p, t, k

    bh, dh, gh

    MNSM

    (stops)

    SM

    (fricatives)

    a.

    p, t, k

    f, , h

    b, d, g

    ***

    b.

    b, d, g

    p, t, k

    , ,

    *!** ***

    c.

    b, d, g

    p, t, k

    f, , h

    *!** ***

    d.

    f, , h

    b, d, g

    p, t, k

    *!***** ***

    e.

    p, t, k

    , ,

    b, d, g

    *!** ***

    f.

    , ,

    f, , h

    b, d, g

    *!**

    All of the candidates in (21) satisfy *ASP, MAX and DEP, which indicates

    that there are no aspirated phonemes, with the number of phonemes

    remaining intact. Candidate (a) is selected as the optimal candidate because

    it incurs no violations of the superordinate constraints MN and SM(stops).

    Candidates (b) and (c) differ from (a) as regards the manner in which

    voiced aspirated banned by *ASP are eliminated. In (b), the voiced aspirates

    become voiced fricatives and two features [asp] and [cont] are affected. In

    (c) the voiced aspirates turn into voiceless fricatives affecting three features

    [asp], [voice] and [cont]. (d) and (e) are candidates in which the MN

    constraint is violated either in IE voiced stops or in IE voiceless stops. In

  • 7/30/2019 English Obstruent Inventory

    19/24

    - 19 -

    both cases, more than one feature are affected, resulting in violations of the

    MN constraint. Candidate (f) is crucial for motivating the reduction of the

    SM constraints and the ranking between the SM(stops) and the

    SM(fricatives) constraints. As a series of changes (spirantization of voiced

    and voiceless stops and deaspiration of voiced aspirates) involve only one

    feature ([cont] for spirantization and [asp] for deaspiration), it incurs the

    same number of violations with regard to the MN constraint. However,

    satisfaction of the MN constraint is achieved by creating a symmetric

    fricative set instead of a symmetric stop set. In contrast, candidate (a), the

    output of GL, satisfies the superordinate SM(stops) constraint with all stops

    having voiced and voiceless pairs.

    4.2 Grimm's Law and ME Phonemicization

    Depending on their nature, the constraints proposed thus far can be

    classified into two groups.

    (22) a. Faithfulness constraintsDEP, MAX, IDENT(F)

    b. Markedness constraints

    SYMMETRY, *ASP, MINIMALITY

    With the constraints in (22) GL can be accounted for by the reranking (or

    promotion) of the markedness constraints *ASP, SM(stops) and MN over

    the faithfulness constraints DEP and MAX.

    (23) a. IE Obstruent Inventory

    DEP, MAX >> *ASP, SM, MN

    b. Germanic Obstruent Inventory

    *ASP, SM(stops), MN >> DEP, MAX, SM(fricatives)

    >> IDENT(F)

  • 7/30/2019 English Obstruent Inventory

    20/24

    - 20 -

    MEP is a typical case of a phoneme split in which the phonemicization

    of voiced fricatives (except for the velar //) leads to a symmetric fricative

    set. The change in the obstruent phoneme inventory from OE to ME can be

    viewed as the result of reranking the SM(fricatives) over the faithfulness

    constraint DEP as in (24).15)

    (24) a. OE Obstruent Inventory

    DEP, MAX, *V/VF SM(fricatives)

    b. ME Obstruent Inventory

    MAX, *V/VF SM(fricatives) DEP

    The faithfulness constraints MAX, DEP and the other markedness

    constraints on symmetry along with the MN constraint play a crucial role in

    our account of GL and MEP.

    An additional benefit of introducing MAX and DEP to the domain of the

    phoneme inventory is that we can now derive a formal definition of the

    types of change in the phoneme inventory.

    (25) Types of Phoneme Change

    a. Chain Shift (Grimm's Law)

    markedness constraints >> MAX, DEP >> ...

    b. Phoneme Split (ME phonemicization)

    MAX, markedness constraints >> DEP ...

    c. Phoneme Merger

    DEP, markedness constraints >> MAX ...

    5. Conclusion

    The present paper argues that Grimm's Law, the systemic change in the

    15) For other minor changes from OE to ME (symmetry in palatal fricatives and

    affricates), refer to Sohn (2009).

  • 7/30/2019 English Obstruent Inventory

    21/24

    - 21 -

    phoneme inventory from IE to Germanic need not resort to the Maintain

    Contrast constraint whose definition and implementation in previous

    accounts are not well motivated. Instead the notion of symmetry in

    inventory in conjunction with minimality of change are employed. It is also

    demonstrated that a serial approach to Grimm's Law should be replaced by

    a parallel analysis which obviates the postulation of stipulative intermediate

    stages. By adopting symmetry as a viable constraint, we can integrate

    Grimm's Law with the later ME phonemicization of voiced fricatives. In

    addition, a formal description of the types of phonemic changes at the

    system level is made possible. The replacement of the Maintain Contrast

    constraint by symmetry presupposes the extension of the faithfulness

    constraints reserved for segments in the current version of Optimality

    Theory to the systemic domains of phoneme inventory, though this requires

    further research.

    References

    Ahn, Sang-Cheol. 2004. Towards an Optimal Account of Diachronic Chain

    Shifts. Studies in Phonetics, Phonology and Morphology 10-1, 43-67.

    Brunner, Karl. 1963. An Outline of Middle English Grammar. Oxford: Basil

    Blackwell.

    Bynon, Theodora. 1977. Historical Linguistics. Cambridge: Cambridge

    University Press.

    Calabrese, Andrea and Morris Halle. 1998. Grimm's and Verner's Laws: A

    New Perspective. Jasnoff J., C. Melchert and L. Olivier (eds.), Mir

    Curad: Studies in Honor of Calvert Watkins, 47-62. Innsbruck:

    University of Innsbruck.

    Campbell, Alistair. 1959. Old English Grammar. Oxford: The Clarendon

    Press.

    Campbell, Lyle. 1999. Historical Linguistics: An Introduction. Cambridge,

    Massachusetts: The MIT Press.

  • 7/30/2019 English Obstruent Inventory

    22/24

    - 22 -

    Cser, Andrs. 1994. Constraints Operating on Germanic Geminates. Vienna

    English Working Papers 3, 62-74.

    Emonds, Joseph. 1972. A Reformulation of Grimm's Law. Brame, M. (ed.)

    Contributions to Generative Phonology, 108-122. Austin: University of

    Texas Press.

    Flemming, Edward. 1996. Evidence for Constraints on Contrast: The

    Dispersion Theory of Contrast. ULCA Working Papers in Phonology 1,

    86-106.

    Flemming, Edward. 2002. Auditory Representations in Phonology. New Yorkand London: Routledge.

    Flemming, Edawrd. 2005. Speech Perception and Phonological Contrast.

    Pisoni, D. and R. Remez (eds.) The Handbook of Speech Perception,

    156-181. Oxford: Blackwell Publishing.

    Garrett, Andrew. 1991. Ind-European Reconstruction and Historical

    Methodologies. Language 67.4, 790-804.

    Hock, Hans H. 1991. Principles of Historical Linguistics. Berlin: Mouton de

    Gruyter.

    Hock, Hans H. and Brian D. Joseph. 1996. Language History, Language

    Change, and Language Relationship. Berlin, New York: Mouton deGruyter.

    Holt, Eric. 2003. Remarks on Optimality Theory and Language Change. Holt,

    E. (ed.) Optimality Theory and Language Change, 1-30, Dordrecht:

    Kluwer Academic Publishers.

    Hopper, Paul. 1973. Glottalized and Murmured Occlusives in Indo-European.

    Glossa 7, 141-166.

    Iverson, Gregory and Joseph Salmons. 2003. Laryngeal Enhancement in

    Early Germanic. Phonology 20, 43-74.

    Kager, Ren. 1999. Optimality Theory. Cambridge: Cambridge University

    Press.

    Kim, Kee-Ho. 2004. Grimm's Law: Push or Drag Chain Revisited. English

    Language and Linguistics 18, 1-24.

    King, Robert. 1969. Historical Linguistics and Generative Grammar.

    Englewood Cliffs: Prentice-Hall.

  • 7/30/2019 English Obstruent Inventory

    23/24

    - 23 -

    Kortlandt, Frederik. 1985. Proto-Indo-European Glottalic Stops: The

    Comparative Evidence. Folia Linguistica Historica 6, 183-201.

    Laker, Stephen. 2009. An Explanation for the Early Phonemicisation of a

    Voice Contrast in English Fricatives. English Language and Linguistics

    13.2, 213-226.

    Lass, Roger and John. M. Anderson. 1975. Old English Phonology.

    Cambridge: Cambridge University Press.

    Lee, Sechang. 2007. First Consonant Shift in the Germanic Obstruent

    System. Korean Journal of Linguistics 32.3, 539-555.McCarthy, John. 2002. A Thematic Guide to Optimality Theory. Cambridge:

    Cambridge University Press.

    McCarthy, John. 2008. Doing Optimality Theory: Applying Theory to Data.

    Oxford: Blackwell Publishing.

    McMahon, April. 1994. Understanding Language Change. Cambridge:

    Cambridge University Press.

    Millward, Celia. 1996. A Biography of the English Language. Fort Worth:

    Harcourt Brace College Publishers.

    Noske, Roland. 2009. Verner's Law, Phonetic Substance and Form of

    Historical Phonological Description. Proceedings of JEL'2009(dis)continu, 6th Nantes Linguistic Meeting, 33-42.

    Petrova, Olga. 2000. Grimm's Law in Optimality Theory. Journal of

    Indoeuropean Studies Monograph Series 35, 45-68.

    Picard, Marc. 1999. On Spirantization and the Minimality of Phonological

    Change. Folia Linguistica Historica 20.1-2, 63-78.

    Prince, Alan and Paul Smolensky. 2004. Optimality Theory: Constraint

    Interaction in Generative Grammar. Oxford: Blackwell Publishing.

    Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia:

    Linguistic Society of America.

    Pyles, Thomas and John Algeo. 1993. The Origins and Development of the

    English Language. Fort Worth: Harcourt Brace Jovanovich.

    Ringe, Donald. 2006. From Proto-Ino-European to Proto-Germanic. Oxford:

    Oxford University Press.

    Saussure, Ferdinand de. 1959. Course in General Linguistics. New York:

  • 7/30/2019 English Obstruent Inventory

    24/24

    - 24 -

    McGraw-Hill.

    Scargill, Matthew. 1951. Notes on the Development of the Principal Sounds

    of Indo-European through Proto-Germanic and West Germanic into

    Old English. Toronto: Toronto University Press.

    Sohn, Chang Yong. 2009. Reranking and Sound Change. Studies in Modern

    Grammar 57, 1-14.

    Teedle, David. 2006. Phoneme Split in OT. ms. University of California,

    Santa Cruz.

    Vennemann, Theo. 2006. Grimm's Law and Loanwords. Transactions of thePhilological Society 104.2, 129-166.

    Voyles, Joseph. 1992. Early Germanic Grammar: Pre-, Proto-, and

    Post-Germanic Languages. San Diego: Academic Press.

    Department of English Language and Literature

    Seoul National University

    599 Gwanak-ro, Gwanak-gu

    Seoul 151-742, Korea

    (02) 880-6093, [email protected]

    Received:

    Revised version:

    Accepted: