Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Linguistics. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 20 January 2021

Morpheme Orderingfree

  • Patrik ByePatrik ByeDepartment of English Language and Literature, Nord University


Morpheme ordering is largely explainable in terms of syntactic/semantic scope, or the Mirror Principle, although there is a significant residue of cases that resist an explanation in these terms. The article, we look at some key examples of (apparent) deviant ordering and review the main ways that linguists have attempted to account for them. Approaches to the phenomenon fall into two broad types. The first relies on mechanisms we can term “morphological,” while the second looks instead to the resources of the ‘narrow’ syntax or phonology. One morphological approach involves a template that associates each class of morphemes in the word with a particular position. A well-known example is the Bantu CARP (Causative-Applicative-Reciprocal-Passive) template, which requires particular orders between morphemes to obtain irrespective of scope. A second approach builds on the intuition that the boundary or join between a morpheme and the base to which it attaches can vary in closeness or strength, where ‘strength’ can be interpreted in gradient or discrete terms. Under the gradient interpretation, affixes differ in parsability, or separability from the base; understood discretely, as in Lexical Morphology and Phonology, morphemes (or classes of morphemes) may attach at a deeper morphological layer to stems (the stronger join), or to words (weaker join), which are closer to the surface. Deviant orderings may then arise where an affix attaches at a morphological layer deeper than its scope would lead us to expect. An example is the marking of case and possession in Finnish nouns: case takes scope over possession, but the case suffix precedes the possessive suffix. Another morphological approach is represented by Distributed Morphology, which permits certain local reorderings once all syntactic operations have taken place. Such operations may target specific morphemes, or morphosyntactic features characterizing a class of morphemes. Agreement marking is an interesting case, since agreement features are bundled as syntactically unitary heads but may in certain languages be split morphologically into separate affixes. This means that in the case of split agreement marking, the relative order must be attributed to post-syntactic principles. Besides these morphological approaches, other researchers have emphasized the resources of the narrow syntax, in particular phrasal movement, as a means for dealing with many challenging cases of morpheme ordering. Still other cases of apparently deviant ordering may be analyzed as epiphenomena of phonological processes and constraint interaction as they apply to prespecified and/or underspecified lexical representations.

It is widely assumed that the order in which the morphemes of a word appear is syntactically and semantically natural, with the possible exception of a small residue of ‘deviant’ cases that resist analysis in these terms. In this article, I will discuss what this syntactically and semantically natural order consists in, and illustrate some of the mechanisms linguists have posited in order to account for the residue.

The remainder of this article is structured as follows. Section 1 anchors the concept of natural morpheme ordering in semantic scope and the hierarchy of functional categories. Section 2 examines claims that there is a distinct morphological contribution to morpheme ordering, encoded in morphological templates, morphological boundaries and domains, or a post-syntactic morphological component. Section 3 looks at phrasal movement as a way to approach morpheme ordering, including apparently ‘deviant’ cases. Section 4 examines the role of the ‘narrow’ phonology in creating deviant order effects. Section 5 concludes.

1. Natural Morpheme Ordering

One clear demonstration of the way semantic scope may determine affix order comes from cases where the scope relation between two heads X and Y may be reversed, and this is mirrored in the order of their exponents. Baker (1985, p. 375) formulates this relationship, conceptualized as a synchronic constraint on grammars, as the Mirror Principle, which states that “morphological derivations must directly reflect syntactic derivations (and vice versa).” For example, in Yup’ik (esu; Eskimo-Aleut; Alaska; Mithun, 2000), the diminutive {-cuar} and augmentative {-paɡ} may alternately scope over the other, giving rise to compositionally distinct pairs, as shown in (1).


Another demonstration comes from recursive structures where a derivational affix may take scope over another instance of the same affix. Korotkova and Lander (2010) furnish an example from Adyghe (ady; North Caucasian; Adygea/Russia), which has a simulative (sml) suffix {-ŝw(a)} meaning ‘be like X’, or ‘be X-ish’. Recursion of the simulative suffix results in an attenuated comparison meaning roughly ‘similar to being similar to X’. This is shown in (2).


So far our examples have been drawn from derivational morphemes. When we turn to functional categories, there is a growing consensus that they occur in a strict sequence. Bybee (1985), for example, argues that the order of functional morphemes reflects their ‘relevance’ to the verb, and proposes the hierarchy in (3).


In generative grammar, the order in which functional heads are merged in the syntactic representation continues to be the focus of considerable research. After early work by Larson (1988) demonstrating the need for more articulated functional structure within the VP, and Pollock (1989) showing that inflection (INFL, or I) may be split into separate heads for Tense and Agreement, the Minimalist Program (Chomsky 1993, 1995) has helped foster further research on the detailed inventory of functional heads and their relations, a program generally known as cartography. Cinque (1999) has mapped out the functional space of the IP, and similar work has been undertaken for the DP (Cinque, 2005; Svenonius, 2008), and the left periphery of the CP (Rizzi, 1997).

A relation between order and scope is also seen with functional categories, such as Tense (T) and Aspect (Asp). For example, Julien (2002) shows that, where both precede the verb, the order is Tense > Aspect, but when they both follow, the order is the opposite: Aspect > Tense. Haitian French Creole (Damoiseau & Gesner, 1975) exemplifies the first order, shown in (4). Here, the T and Asp heads are spelled out by function words.


Turkish (Lewis, 2000) illustrates the mirror order in which the T and Asp heads are realized as suffixes, and the Aspect marker precedes that for Tense. This is shown in (5). (Note: epv = epenthetic vowel)


Observations like these paved the way for the analysis of suffix order in syntactic terms, as head movement. Movement of a head X to a target head Y results in a complex head [Y X Y ] (cf. the Right-hand Head Rule; Williams, 1981). Head movement is restricted by the Head Movement Constraint (Travis, 1984, p. 131), which prevents skipping of heads. Thus, a head can only adjoin to the nearest head, and takes any head to which it is already adjoined with it on successive moves (roll-up). Given the base order Z > Y > X, shown diagramatically in (6a), we can thus derive the orders Z-Y-X (the base order) (6a) Z-X-Y (6b), and X-Y-Z (6c), but not *X-Z-Y (6d), since this would involve raising X past the intervening head Y.


The structure in (6a) parallels the situation in Haitian French Creole, where Tense and Aspect heads are spelled out in their base positions, as shown in (7) (cf. (4c)).


The head movement analysis of Turkish begins with the same base order, as shown in (8a). The mirror order is then derived in two successive steps. In the first step, V adjoins to Asp to derive a complex head V-Asp as in (8b). In the second and last step, V-Asp adjoins to T, giving (8c).


As Head Movement would predict, the order Tense > V > Aspect is also attested, corresponding to the structure shown in (6b). An example is Turkana (tuv; Nilotic; Kenya; Dimmendaal, 1983), e.g., ɛ̀-à-ɪ̀mʊ̀j-ɪ ̀ (3-pst-eat-prf), ‘he ate’ (p. 130). On a head movement account of Turkana, V would raise as far as Asp, but not to T.

Facts on the ground are of course not always quite so simple as described in this section, and mirror violations are, at least apparently, common enough that morphological factors should be seriously considered.

2. Morpheme Ordering and Morphology

In this section we review some of the forms a distinct morphological contribution to morpheme ordering might take. Section 2.1 begins with the idea that the morphological component may simply stipulate the linear ordering of morphemes by means of a template. Section 2.2 looks at how the boundary or domain involved in a morphological process may affect morpheme order. Finally, section 2.3 considers how post-syntactic interpretative processes may affect morpheme ordering, as in Distributed Morphology.

2.1 Linear Morphology

One radical possibility is to reject the Mirror Principle as a synchronic constraint on morphology. This is the position of Givón (1971), who argues that the order of morphemes in a word reflects not synchronic principles but the diachronic order in which independent words were grammaticalized (cf. also Comrie, 1980; Jacques, 2013). The rules that govern morpheme order are according to this view language-specific, and may not reflect more general principles such as scope. Hyman (2003, p. 245) similarly proposes that “languages can impose specific morphotactic constraints for which there is no synchronic extra-morphological explanation.”

Since Simpson and Withgott (1986), it has been common to draw a distinction between layered morphology, which has hierarchical structure, and position class morphology (a.k.a. slot-filler morphology), which is linear, or flat, and defined in terms of positions, or slots (Anderson, 1992; McDonough, 2000; Spencer, 1991; Stump, 1993, 1997).

The sequence of slots is generally referred to as a template. Inkelas (1993, p. 560) characterizes position class morphology as a system where “morphemes or morpheme classes are organized into a total linear ordering that has no apparent connection to syntactic, semantic, or even phonological representation.” The position class template is widely used as a descriptive device, but some researchers argue on the basis of the more challenging cases that word formation may be inherently templatic. Since position classes must be stipulated, however, much current research seeks to reanalyze these cases in layered morphological terms (Hale, 2001; Rice, 2000). For two recent defenses of the template, see Nordlinger (2010) and Crysmann and Bonami (2016).1

One much discussed example of a template is the Bantu CARP template, which refers to the ordering of Causative, Applicative, Reciprocal and Passive suffixes in the Bantu verb. The template enforces certain affix orders that go against what would be predicted by scope. In Chichewa (nya; Niger-Congo; Bantu; Zambia, Malawi, Mozambique, Zimbabwe; Hyman 2003), for example, the causative may take scope over the applicative and vice versa, but the causative suffix {-its} must precede the applicative {-il} regardless of semantic scope, as shown in (9). In (9a) the applicative scopes over the causative, and the order of the suffixes conforms to the expectations of the Mirror Principle. In (9b), however, the order of the suffixes is the same, despite the fact that the causative is interpreted lower than the applicative.


Hyman (2003, p. 263) explains such templatic effects as emerging from “the pressure for affix ordering to be invariant.” We shall return to one possible explanation of this pressure in section 2.2. In section 3 we look at a possible account of the Chichewa facts that exploits phrasal movement.

2.2 Boundaries and Domains

Another potential morphological factor in morpheme ordering is the nature of the boundary or domain. In the linear conception of The Sound Pattern of English (SPE) (Chomsky & Halle, 1968), the boundary is conceptualized as an element in the string. SPE phonology operated with two types of boundary, a strong, or ‘word’, boundary (#), and a weak, or morpheme, boundary (+). This basic idea subsequently developed in two main directions.

The first major development of the SPE theory of boundaries is stratal approaches (Kiparsky, 1982; Siegel, 1974), which reconceptualize the word and morpheme boundaries of earlier linear approaches in terms of hierarchically organized morphological domains, such as the stem and word. In Lexical Morphology and Phonology (LMP; Kiparsky 1982), morphological and phonological processes are interleaved, and morphological processes may form stems or words. Since they belong to separate strata, all morphological processes that output stems precede all morphological processes that output words. Given LMP posits an autonomous morphological component, we might expect situations to arise where scope and morphological layering do not align. Such a situation would obtain if, given a scopal relation X > Y, the lower element Y triggers a word-forming process, but the higher element X triggers a stem-forming process. Kiparsky argues that such a case exists in Finnish, which marks case and possession in nouns. In terms of scope, case is outside possession, but the possessive suffix follows case, e.g. [ [ [ talo ] mme ] ssa ] → talo-ssa-mme (house-iness-1pl.poss) ‘in our house’. In Kiparsky’s stratal terms, this anti-scopal ordering follows from the morphological organization of noun inflection in Finnish: a possessive suffix takes a stem as input and forms a word, while a case suffix takes a stem and forms another stem. Since all stem-formation takes place before word-forming suffixes are added, however, the case suffix appears before possession, a situation Kiparsky dubs introfixation.

A second approach to boundary strength with implications for morpheme ordering derives from research on speech processing. Hay (2002, 2003) and Hay and Plag (2004) argue that boundary strength is not a binary, as in SPE and its descendants, but a gradient property grounded in factors in speech perception. Affixes differ depending on how easily they can be parsed out from the stem in processing, allowing them to be ranked in a complexity-based ordering. Hay and Plag identify a preference for more separable affixes to attach outside less separable affixes, which has potential implications for understanding cases of deviant morpheme ordering. For example, English has two nominal suffixes that differ in separability, the Latinate {-ity} and the Germanic {-ness}, the latter being easier to parse out than the former.2 The adjective-forming suffix -less (also Germanic), as in atomless, is similarly highly separable. Complexity-based ordering predicts that atomlessness is a possible word, but that *atomlessity is not, since the low-separability suffix in this case attaches outside the highly separable one. This is despite *atomlessity being semantically indistinguishable from atomlessness.

It is possible that differences in separability may result in disruptions of the compositionally determined order. In Athabaskan languages, for example, what Rice (2011, p. 182) terms “prosodically ‘small’ or ‘incomplete’ affixes (V, C, VC) [appear] closer to the stem than ‘full’ or ‘complete’ ones (CV, CVC, CVCV).” This can be seen in languages where the same meaning may be expressed using either a ‘small’ or a ‘full’ affix, but the linear position of the ‘small’ and ‘full’ affixes differ. In Slavey (a.k.a. Slave; den; Na-Dené, Athabaskan; Northwest Territories, Canada; Rice, 1989), inceptive meaning can be marked by the ‘full’ affix {tį-}, ‘start off’, or the ‘small’ affix {d-}. Scope would predict that these affixes should occur in the same position, but there are differences. The ‘full’ inceptive affix precedes the aspect marker {na-}, but the ‘small’ inceptive affix attaches closer to the stem, and so follows it, e.g., tį-na-neh-tɬa, ‘I started off again’ versus na-d[e]-h-tɬaah, ‘I started out on a trip again’ (Rice, 2011, p. 182).3

Parsability may also explain certain cases of multiple exponence, which raise some of the same issues as deviant ordering. For a comprehensive recent overview of the phenomenon of multiple exponence, I refer the reader to Harris (2017). As an example, we can take Breton (Stump, 1989), where certain nouns have both a simple and a double plural. In one pattern, the simple plural is formed by ablaut, e.g., louarn ~ lern, ‘fox ~ foxes’, gavr ~ gevr, ‘goat ~ goats’, troad ~ treid, ‘foot ~ feet’. In the double plural pattern, plural morphology reapplies to an ablauted stem to give forms like lerned, gevred, and treidoù. In an analysis of a similar phenomenon in varieties of German, Plank (1985) proposes that double exponence serves as a listener-oriented enhancement strategy, making it easier to perceive the presence of the category in the signal. In this case, the simple ablauted plural, whose marker has a very low degree of separability from the stem, is augmented with a suffix with high separability.4

We might extend this listener-oriented approach to understand cases of deviant affix ordering, such as the CARP template effects in Bantu discussed in section 2.1. Hyman (2003, p. 263) suggests that templatic effects arise because learners may resist decomposing affix sequences when “much of the scope relations is either trivially predictable from the lexical semantics or the discourse context or non-consequential, even indeterminate.” In the pair of examples from (9) discussed in section 2.1, -lil-its-il-, ‘cause to cry with (e.g., by throwing sticks at someone)’ versus -takas-its-il-, ‘cause to stir with (e.g., a spoon)’, ambiguity in scope is unlikely to arise in context given the semantics of the verbs involved, although we can imagine less plausible interpretations of each with the scope reversed, for example, ‘cause to cry with (e.g., a handkerchief, which the causee uses to dry their eyes)’, or ‘cause to stir with (e.g., a stick, which the causer uses to urge the causee to stir)’. Hyman thus imputes a certain cost of analysis to the learner, but this position may be taken to imply that the learner simply posits a portmanteau -itsil-, which spells out a combination of causative and applicative heads regardless of their scopal relation. An alternative interpretation would take the listener’s point of view while preserving Hyman’s essential insights. Under the circumstances Hyman is discussing, the cost of processing deviant ordering for the listener is low due to rich contextual cues that offset the risk of any ambiguity of interpretation. What this purchases, for speaker and listener alike, is greater predictability in the linear ordering of the two suffixes.

An affiliated proposal by Ryan (2010) is that learners extract information about the local sequences in which affixes occur and encode them as bigrams. Such bigrams encode preferences for pairs of affixes to surface adjacently. One striking prediction of the bigram approach is that, since affixes are ordered pairwise, we should find cases of non-transitivity, which obtains when, given affixes X, Y and Z, X ≺ Y, and Y ≺ Z, but Z ≺ X. Such a situation is argued to obtain in Chumbivilcas Quechua (Muysken, 1988, p. 263). In this language, inchoative {-ri} must precede the assistive {-schi}, and {-schi} must precede the reciprocal {-na}. Transitivity would then predict that {-ri} precede {-na} but, in fact, the only grammatical order is {-na-ri}.

Facts like these have also been taken to point to the existence of interpretative processes that can reorder syntactic terminals or lexical items once inserted. It is to these theories we now turn.

2.3 Distributed Morphology

In sections 3 and 4 we consider what strategies there may be in the ‘narrow’ syntax and the ‘narrow’ phonology for dealing with deviant orderings. The term ‘narrow phonology’ is not, to my knowledge, encountered in the generative literature. I use it here as a way to delimit the module involved in interpreting syntactic representations whose terminal nodes have been spelled out with the phonological shapes of particular vocabulary items. For practitioners of Optimality Theory (OT; McCarthy & Prince, [1993] 2001, Prince & Smolensky, [1993] 2004), the ‘narrow’ phonology would consist of the universal set of violable constraints Con, and the functions Gen, which generates candidate parses of a particular input, and Eval, the language-specific ranking of Con against which candidate analyses of the input are evaluated to produce a unique output form.5 This ‘narrow’ phonology is not to be equated with PF (‘Phonological Form’), which, at least in some approaches, is assumed to contain other post-syntactic processes with implications for the order in which morphemes appear. In Distributed Morphology (DM; Halle & Marantz, 1993, 1994, Siddiqi, 2010), most of the work of word formation is carried out by the syntax, through mechanisms such as head movement. In the PF branch, however, there is also a set of morphosyntactic operations that apply after linearization of the terminal nodes of the syntactic representation, but before Vocabulary Insertion (VI), or the ‘narrow’ phonology. These processes do not constitute a generative component; they are restricted to local editing of syntactic features, and include fusion, fission, and movement of heads, preparatory to Vocabulary Insertion. Distributed Morphology owes its name to the fact that morphology is split between a syntactic (generative) and a post-syntactic (interpretative, non-generative) component.

In addition to the morphosyntactic operations just mentioned, there is also a set of operations called readjustment rules that alter the phonological shape of particular morphemes. The overwriting of the substring -oy in destroy with -uct to give the stem destruct- would be an example. In practice, however, cases of readjustment are difficult to distinguish from suppletion, and some researchers argue that readjustment rules should be excluded from the theory (see, e.g., Haugen & Siddiqi, 2013). Nonetheless, both morphosyntactic and morpheme-specific readjustment processes have been invoked in the literature to explain deviant orderings.

Two processes that produce local reordering are local movement of heads (lowering) and local dislocation. While lowering manipulates morphosyntactic features, local dislocation applies after Vocabulary Insertion, to specific morphemes.

Lowering has been invoked in the analysis of English tense inflection to explain well-known facts about English word order. In finite clauses lacking an auxiliary, the verb is inflected for tense and agreement. On the distributional evidence, tense forms a category with modal auxiliaries like may. However, despite this syntactic equivalence, tense and agreement markers do not appear in the same linear position as modal auxiliaries. While modal auxiliaries precede VP-adjoined adverbs like usually, as in (10a), a verb follows, even when inflected for tense and agreement, as in (10b). The order in (10c) is ungrammatical in English, but would be grammatical in French, which, unlike English, raises V to T.


In the negated versions of (10), the auxiliary precedes negation, as in (11a). Where negation intervenes between tense and the verb, tense is spelled out in the same linear position as a modal auxiliary. This activates the insertion of the ‘dummy’ auxiliary do, as shown in (11b).


Developing the affix-hopping analysis of Chomsky ([1957] 2002), Halle and Marantz (1993) and Bobaljik (1994) argue that the displacement of the tense and agreement markers in English should be understood as a post-syntactic fusion, where T lowers and merges with the adjacent V head (the head of T’s complement). The adjacency restriction explains why T is unable to bypass a negative head to merge with V, as shown by the ungrammaticality of (11c). Insertion of the dummy auxiliary do is thus used as a repair when the locality requirements rule out lowering of T to V.

In contrast to lowering, local dislocation (Embick, 2007; Embick & Noyer, 2001) occurs after Vocabulary Insertion and targets specific items. For example, the Latin coordinate -que attaches to the right of the first word in the second conjunct, e.g., ob eās-que rēs (lit. because.of these- and things), ‘and on account of these achievements’, rather than *-que ob eās rēs (Hale & Buck, 1966, p. 165).

Another DM process with implications for ordering is fission, since it results in affixes whose ordering cannot be determined by appeal to syntactic or semantic relations. Phi- (φ‎-) features (person, gender, and number) are intrinsically present in an NP but, as agreement features, they are not interpretable at LF (‘Logical Form’). This raises interesting questions for how the morphemes expressing such features are linearized. There are two ways to conceptualize the place of agreement features in the overall model. The traditional view is that uninterpretable features are available to computation in the Syntax, and then deleted at LF (Chomsky, 1993, 1995). An alternative view is that they are introduced in the PF branch and hence do not have to be deleted for semantic interpretation (Bobaljik, 2008). According to this view, agreement highlights structural relations in the linearized expression but does not figure in the computation of syntactic structure. The mechanism underlying agreement is the copying of φ‎-features, which are later targeted by lexical insertion. This is suggested by some cases where agreement may take the form of inserting of an affix that is phonologically (near) identical to the nominal class marker, as in Swahili ki-sima ki-refu ( ac-deep), ‘(a) deep well’ (Perrott, 1957, p. 9), where the ki-class marker on the noun also marks adjectival concord (AC).

The morphology of agreement across languages indicates the existence of ordering principles other than those that operate in syntax. In DM, agreement is theorized as a unitary Agreement (Agr) head but with several φ‎-features. In some languages, however, the same Agr head may be spelled out with two or more Vocabulary items, indicating that the Agr head must split before VI. DM posits an interpretative process of fission, which takes unordered bundles such as [ 1 pl m ] and parses out, for example, person, giving [ 1 [ pl m ] ] (Noyer, 1997). The question is then how the affixes that spell out the fissioned components are ordered. According to Noyer, ordering is resolved by appeal to a universal feature hierarchy: 1 > 2 > dual > plural > . . . In a continuation of this line of inquiry, Trommer (2003) examines the placement of subject agreement affixes in a sample of languages. There are six possible orderings of Person, Verb, and Number, shown in (12).


Trommer observes that in systems where person and number marking are split, 80 languages in his sample, person tends to precede the verb, while number tends to follow. The preferred order is Pers > V > Num, which is found in Muna (mnb; Austronesian, Malayo-Polynesian; Sulawesi, Indonesia; Van den Berg, 1989), e.g., o-kala (2-go), ‘you (sg.) go’ versus o-kala-amu (2-go-pl), ‘you (pl.) go’. The Pers > V > Num order is found in 39 languages, with V > Pers > Num and Pers > Num > V accounting, respectively, for 22 and 9. There is thus a marked preference for Pers to precede Num, regardless of where V appears. The three remaining orderings, in which Num instead precedes Pers (V > Num > Pers, Num > Pers > V, and Num > V > Pers) account for the remaining 10 languages in the sample. Working within an OT framework, Trommer proposes constraints that drive the exponents of Pers to align maximally to the left, and Num maximally to the right. To account for those languages where Pers and Num appear on the same side of the verb, giving the next most preferred orders Pers > Num > V and V > Pers > Num, Trommer proposes there is a constraint (Coherence) requiring that exponents of the unitary Agr head appear adjacently.

In sections 3 and 4, we shall look at attempts to analyze deviant ordering in syntax and phonology.

3. Reordering in Syntax

A growing skepticism to head movement in recent years (Chomsky, 2001; Harley, 2013; Roberts 2011) has encouraged the development of alternative approaches to ordering phenomena, including those that seem to diverge from the predictions of the Mirror Principle. In this section we consider the approach that originates in work by Cinque (2005, 2009). Inspired by minimalist principles (Chomsky 1993, 1995), phrasal movement is an instantiation of Internal Merge.

Cinque’s point of departure is Greenberg’s Universal 20 (Greenberg, 1966, pp. 68f.), which concerns cross-linguistic orderings of nouns and their modifiers. It states: “When any or all of the items (demonstrative, numeral, and descriptive adjective) precede the noun, they are always found in that order. If they follow, the order is either the same or its exact opposite.” Schematically: Dem > Num > A > N, N > Dem > Num > A, or N > A > Num > Dem. With the benefit of a richer set of data, however, Cinque (2005) is able to show that there is in fact greater variation in the attested orders of {Dem, Num, A, N} than Greenberg described, although it is still circumscribed. There are 24 (4!) logically possible permutations of {Dem, Num, A, N}, but Cinque shows that only 14 are actually attested in natural human languages. The situation with nominal modifiers is paralleled by Cinque’s findings regarding the orderings of Mood, Tense, Aspect, and V (Cinque, 2009), which we look at in more detail here. Let us compare the predictions of phrasal movement with those of head movement, which yields the four orderings in (13).


While all of the orders in (13) are attested, there are ten other attested orderings that head movement fails to predict. Tested against the range of attested cross-linguistic variation, head movement turns out to be too restrictive.

Assuming a universal base-generated order Mood > Tns > Asp > V, phrasal movement generates the orders in (14). Unpronounced copies are shown in gray.6


The base order is shown in (14a). In (14b), the VP moves to a position between Tns and AspP. Orders (14b) to (14d) involve movement of the VP in successive cycles. Notice that orders (14c) and (14d) appear to show mirror violations, and could not be produced by head movement. On the other hand, phrasal movement is capable of mimicking the effects of head movement, as is the case for (14b), (14g), and (14m), which respectively correspond to (13b), (13c), and (13d).

The technology of phrasal movement has been used to explain deviant orderings analyzed in the literature using other mechanisms. Koopman (2017), for example, brings phrasal movement to bear on the analysis of mobile affixes in Huave (hue; isolate; Stairs & Hollenbach, 1981), which Embick and Noyer (2001) analyze using local dislocation. (See also Kim (2010, 2015) for a contrasting perspective to both.)

An application of phrasal movement to understand aspects of the CARP template discussed in section 2.1 is found in Muriungi (2008). Muriungi’s account, which focuses on Kîîtharaka (thk; Bantu; Kenya), may be generalized to account for the Chichewa facts discussed in section 2.1. According to Muriungi, the Applicative head is quasi-prepositional, and introduces an applied DP-argument, as shown in (15). This assumption is crucial for explaining why the applicative and causative have a fixed ordering, regardless of differences in scope. (Muriungi uses coerce to refer to the Causative head.)


The DP-argument itself is ultimately extracted, and realized outside the verbal domain, but we abstract away from this step here. The innermost (phrasal) constituent containing the verbal root is first copied to a position between coerce and the complex head containing the Applicative and its DP-argument, and then to the position above coerce, as shown in (16). This derives the observed ‘anti-mirror’ ordering of the causative and applicative morphemes (root-coerce-appl).


The derivation parallels the attested order of V > Tns > Asp in (14c), derived by successive cyclic movement of VP. When the scope is reversed (Appl > Caus/coerce), the root must ‘drag’ the causative suffix along with it, mimicking the effect of head movement (cf. the attested order in (14g)). This again produces the surface order root-coerce-appl, but in this case it mirrors scope. The question then becomes why, when Causative scopes over Applicative, as in (15), the root does not take the applicative suffix with it. The crucial difference according to Muriungi is the DP-argument in the representation of the applicative. It is the complexity of the applicative head that prevents it from being dragged along with the root constituent when it moves above the causative head.

4. Deviant Ordering as Phonological Epiphenomenon

In this section, we raise possibility that certain apparent cases of deviant ordering are the result of constraint interaction in the narrow phonology, which applies after lexical insertion, perhaps on a phase-by-phase basis. It is generally assumed that segmental phonology does not ‘see’ syntax at all, and that the syntactic information to which prosody has access is limited to the distinction between functional and lexical categories, and the distinction between word- (X), phrase-level categories (XP) (see Selkirk, 2011). Bye and Svenonius (2012) take a step in this direction, arguing that nonconcatenative effects result in general from some combination of featural underspecification and prosodic prespecification in the underlying shape of the affix, together with interaction of ranked, entirely general phonological constraints. There is no nonconcatenative morphology according to this view—there are only epiphenomenal nonconcatenative effects. This same general idea may be extended to deal with some cases of putative deviant ordering. In this section, we will review some of the ways in which purely phonological operations can give rise to effects that mimic deviant ordering.

Deviant ordering effects might result from three types of phonological process, the first two of which involve some kind of literal metathesis. First, metathesis may change the underlying order of elements in response to structural markedness (section 4.1). Second, metathesis may be a strategy for preserving underlying prosodic structure (section 4.2). The third type of case starts with the idea that even when scope is defined, the ordering of two exponents may be phonologically undefined because they occupy different tiers of the phonological representation. Deviant ordering effects may be the result (section 4.3).

4.1 Markedness-Driven Metathesis

Metathesis is a change in the underlying order of segments /xy/ to surface [yx]. On occasion, metathesis may result in the local reordering of segments that happen to be exponents, giving rise to deviant ordering effects. Paster (2009, p. 30) provides a case from Doyayo (a.k.a. Dowayo; Niger-Congo, Adamawa-Ubangi; Cameroon; Wiering & Wiering, 1994). In combination with another suffix, the augmentative {-m} always comes first. This is shown in (17).


As Paster points out, though, the pattern is motivated by straightforwardly phonological considerations, and occurs with a non-derived base as well. With consonant-final roots like /tus/, ‘spit’, and /kab/, ‘catch’, the augmentative is infixed, as in (18).


Aronoff and Xu (2010) analyze a case from Lezgian (Haspelmath, 1993) that works in a similar way.

4.2 Prespecification and Faithfulness-Driven Metathesis

Inkelas (1989) and Inkelas and Cho (1993) propose that some prosodic structure could be present underlyingly. Translating this idea into Optimality Theory requires positing the existence of Faithfulness constraints that demand preservation of such prespecified prosodic structure in the surface form. Metathesis is one possible strategy to achieve this. Assuming that edge orientation is given by the syntax, there are nonetheless cases where the edge to which an affix aligns is the opposite of what one would expect.

Jenks and Rose (2015) deal with a case of edge mobility in Moro (mor; Niger-Congo; Nuba Hills, Sudan) that may yield to this kind of analysis. In this language, verbs take object markers which by default follow the verb (as enclitics), but which may also appear preverbally under certain conditions involving tone. Jenks and Rose identify a domain they term the ‘macro-stem’ which requires a high (H) tone at its left edge, which we can interpret as a case of prespecified prosodic structure. Although this requirement is not always satisfied, the crucial point is that it can be met by attracting an H-tone-bearing clitic to the left edge of the macro-stem. Consider the examples in (19) and (20).7 The examples in (19) show the perfective and proximal imperfective forms without object clitics. The perfective form in (19a) is marked by an H-tone-bearing suffix {-ó}. The proximal imperfective (‘about to’), shown in (19b), has H tone on the root. This difference turns out to be crucial for understanding the difference in how the object markers attach.


Now consider the corresponding forms in (20) with an object marker attached, in this case, the second person singular object marker -ŋá-. In the perfective, the objective marker follows the root, as in (20a). In the proximal imperfective, however, the same objective marker with an H tone precedes the root. This is shown in (20b).


The objective marker precedes the root precisely in the environment where the verb stem surfaces with an H tone in the absence of an object marking prefix. Jenks and Rose propose that the ‘macrostem’, which includes the exponent of the verb and any iterative prefix, has an H tone aligned to its left edge. In the absence of an object marker, this H-tone is realized by associating to the first tone-bearing unit (vowel) of the verb root and spreading rightwards, as in (19b). This, however, entails a violation of a (faithfulness) requirement that tone-bearing units surface with their underlying tonal specifications (e.g. Ident[tone]). Where there is an object marker available that already bears an H-tone, as in (20b), the marker pivots about the macrostem to become a proclitic, thereby allowing the marker’s H-tone to merge with the left-aligned H-tone of the macrostem. This metathesis circumvents the need to introduce non-underlying associations between tones and vowels.

Continuing in this vein, Bye and Svenonius (2012) propose that certain affixes may be prespecified as repelling the edge they are introduced by the syntax, a property they dub ‘antitropism’. In Tamazight Berber (tzm; Afro-Asiatic; Morocco; Abdel-Massih 1968), for example, subject agreement on the verb is generally marked by a suffix, e.g., ʕum-n (, ‘they swam’, but certain other agreement markers are exceptionally prefixal, e.g., n-ʕum (1pl-swim), ‘we swam’.

Bye and Svenonius (2012) suggest further that antitropism may also account for infixation and that whether we find infixation or affixation to the opposite edge is decided by the phonology. Thus, the facts of Tamazight Berber may be analyzed as purchasing satisfaction of Contiguity, which punishes the intrusion of material between underlyingly contiguous segments, at the cost of Linearity, which punishes changes to the underlying order in pairs of segments. Compare Berber with the situation in Atayal (Egerød, 1965: tay; Austronesian; Northern Formosan; Taiwan), where animate actor focus is marked by infixing {-m-} immediately after the first consonant of the root, e.g., /m+qul/→qmul, ‘snatch’, /m+sbil/→smbil, ‘leave behind’. The Atayal pattern can be understood as an example of sacrificing Contiguity to Linearity, so that only minimal violation of Linearity occurs, consistent with the faithfulness requirement that the ‘antitropal’ actor focus marker never surface at the left edge.

4.3 Underspecification and Apparent Reordering

It is a well-established finding of phonological theory that some affixes may lack segmental content and consist solely of prosodic material, e.g., mora, syllable, or higher (Saba Kirchner, 2010; Trommer & Zimmermann, 2014; Zimmermann, 2017). Since most prosodic structure is not underlyingly specified, prosodic affixes may be expected to float, with implications for the way in which their segmental exponents are linearized. For example, Saba Kirchner (2013) discusses the excessive reduplicant (“too much”) in Kwak’wala (kwk; Wakashan; Queen Charlotte Strait, Canada; Boas, 1947), e.g., mə-miːχ‎kən, ‘sleep too much’. Saba Kirchner analyzes the reduplicant, here realized segmentally as /mə-/, as an underlyingly empty syllable, which acquires segmental content by epenthesis of schwa and copy of the nearest root consonant, in this case /m/.

One possible case of apparent reordering comes from Tagalog (Schachter & Otanes, 1972), where the positioning of the reduplicative affix (RA) may vary in ways that appear to deviate from what scope would lead us to expect. RA reduplication is used in a variety of processes, including the formation of the so-called ‘contemplated aspect’, or future. RA can occur before or after the exponent of a morphosyntactic head over which the aspectual head has scope (Carrier, 1979; French, 1988; Lieber, 1988; Marantz, 1982). Tagalog verbs are inflected with a topic marker which indicate which NP (subject, direct object, or indirect object) in the sentence is topic. An example is shown in (21), where all of the forms derive from [ ma [ RA [ ʔi [ pag [ línis ] ] ] ] ], ‘will manage to clean for’. The segmental exponent of RA is shown underlined.


The compositionally transparent order is (21a), where RA copies material from the object topic prefix ʔi- to its right. In (21b), the segmental exponent of RA is displaced to the right of the object topic prefix, and instead retrieves its segmental content from the transitive prefix pàg-. Finally, in (21c), the RA is displaced to the right of the transitive prefix, ending up adjacent to the verb root.

The facts of Tagalog stress suggest how a phonological account of the variability shown in (21) might work. According to French (1988), the root forms a phonological domain with following suffixes, while the prefix string forms a separate domain. French distinguishes foot- and word-level (main) stress. Main stress falls on either the final or penultimate syllable of the domain that includes the root and any suffixes that follow. With the exception of the RA morpheme, stress in the prefix domain is assigned according to two principles. First, a disyllabic prefix will generally have a foot-level stress on its first syllable, e.g., pà -ʔusáp-an (loc-request-loc) ‘makes a request of X’. Second, closed syllables also receive a foot-level stress. The RA morpheme behaves differently to the other prefixes with respect to stress. For one thing, RA always receives at least a foot-level stress, despite meeting neither of the usual phonological conditions (disyllabicity, closed syllable) for doing so. French argues that RA receives a foot-level stress when it immediately precedes and copies root material, as in (21c). When it immediately precedes and copies prefixal material, RA receives a word-level stress, as in (21a) and (21b). This suggests that the exponent of the RA morpheme is a prosodic unit higher than a syllable. If we posit that this unit is the prosodic word (ω‎), this would explain why RA receives word-level stress, at least when it surfaces non-adjacently to the root. A phonological implication of French’s analysis of RA is that it splits the morphological word into two prosodic words—essentially making it a compound.

Since RA does not contain any segments of its own, its prosodic exponent (ω‎) is not properly ordered with respect to other segmental material in the input. This may explain how RA is permitted to float, without the assumption that there is a literal mirror violation. The deviant ordering may instead be analyzed as an effect of phonological constraint interaction.

One way to do with this is to leverage the possibility that prosodic words may be nested within larger prosodic words (Ito & Mester, 2009). This analysis is commonly used in the analysis of cliticization, which involves adjunction of a clitic to a prosodic word to create a higher-level (non-minimal) prosodic word including the clitic. For example, prepositions in English are cliticized to a following prosodic word, e.g., to + (ω London) → (ω to (ω London)). If RA is a prosodic word, prefixal material that precedes it may be left-adjoined to it in the same way. The forms in (22) show how the variant forms from (21) might be structured prosodically.


The variability can be understood partly in terms of a constraint *Adjoin, which penalizes syllables that are left-adjoined to RA. The question is then how any syllables intervening between RA and the root are incorporated into prosodic structure. One possibility is that they are not parsed into a word, and therefore incur violations of *Parse (σ‎). How each variant in (22) fares on *Adjoin and Parse(σ‎) is shown in (23).8


One proposal for dealing with variability in Optimality Theory is to allow conflicting constraints to be unranked in the grammar (e.g., Anttila, 1997). Conflicts are then resolved by imposing rankings as performance in real time demands. Ranking Parse (σ‎) over *Adjoin will result in choosing (23c) as the optimal candidate, while reversing the ranking will return (23a). This does not explain the possibility of (23b), which cannot be optimal on either ranking. Applying the logic of OT, there must be an additional constraint on which (23b) is superior to (23a). Notice that candidate (23a) has two consecutive violations of Parse (σ‎), which (23b) does not. Smolensky (1995, 1997) proposes that there are cases where two local violations of the same constraint may be assessed as worse than two non-local violations of the same constraint taken separately. Without some additional mechanism, however, Optimality Theory does not allow us to capture this state of affairs, since the domination relation is strict. Given two constraints, C1 and C2, if C1 dominates C2, then no amount of violation of C2 can imperil a candidate that satisfies C1. The mechanism Smolensky proposes to accommodate the intuition given the logic of OT is local self-conjunction. A constraint C may be locally self-conjoined to give a new constraint C2 that universally dominates C. Applying this idea to the case at hand, candidate (23a), which violates Parse (σ‎) twice in consecutive syllables, also violates the self-conjoined constraint Parse (σ‎)2, which dominates Parse (σ‎). Optimization of (23b) as a variant over (23a), therefore, requires that Parse (σ‎)2 outrank *Adjoin.9

5. Retrospect and Prospect

Morpheme ordering is largely syntactically and semantically natural, reflecting relations of scope or the order in which functional heads are merged in the syntactic representation. However, there are at least apparent exceptions, and in this article we have looked at some of the machinery linguists have proposed in order to account for this residue. Some of these mechanisms are morphological: deviant orderings may be stipulated by template, favored by processing considerations, or may result from local edits of the syntactic representation once syntactic operations have applied. We have also considered recent technologies developed in research on ‘narrow’ syntax (phrasal movement) that may provide traction in analyzing (apparently) deviant orderings, and considered the possibility that some deviant orderings may be the epiphenomenal effects of constraint interaction in the ‘narrow’ phonology. Further research in the years to come can be expected to clarify to what extent deviant orders can be analyzed as the result of these narrow syntactic and phonological mechanisms, as opposed to distinctly morphological ones.


I would like to thank Paul de Lacy and Peter Svenonius for discussion of the issues discussed in this articles, and two anonymous reviewers for comments. Any errors of fact or interpretation are my own responsibility.

Further Reading

  • Inkelas, S. (2016). Affix ordering in optimal construction morphology. In D. Siddiqi & H. Harley (Eds.), Morphological metatheory (pp. 479–511). Dordrecht, The Netherlands: John Benjamins.
  • Manova, S., & Aronoff, M. (2010). Modeling affix order. Morphology, 20, 109–131.
  • Rice, K. (2011). Principles of affix ordering: An overview. Word Structure, 4, 169–200.
  • Saarinen, P., & Hay, J. (2014). Affix ordering in derivation. In R. Lieber & P. Štekauer (Eds.), The Oxford handbook of derivational morphology (pp. 370–383). New York: Oxford University Press.


  • Abdel-Massih, E. T. (1968). Tamazight verb structure: A generative approach. The Hague, The Netherlands: Mouton.
  • Anceaux, J. C. (1965). The Nimboran language: Phonology and morphology. The Hague, The Netherlands: Martinus Nijhoff.
  • Anderson, S. R. (1992). A-morphous morphology. Cambridge, UK: Cambridge University Press.
  • Anttila, A. (1997). Deriving variation from grammar: A study of Finnish genitives. In F. Hinskens, R. van Hout, & W. L. Wetzels (Eds.), Variation, change and phonological theory. Amsterdam: John Benjamins.
  • Aronoff, M., & Xu, Z. (2010). A realization optimality-theoretic approach to affix order. Morphology, 20, 381–411.
  • Baker, M. (1985). The mirror principle and morphosyntactic explanation. Linguistic Inquiry, 16, 373–415.
  • Boas, F. (1947). Kwa utl grammar with a glossary of the suffixes. Vol. 37 (Transactions of the American Philosophical Society. New Series). Philadelphia, PA: American Philosophical Society.
  • Bobaljik, J. (1994). What does adjacency do? MIT Working Papers in Linguistics, 22, 1–32.
  • Bobaljik, J. (2008). Where’s φ‎? Agreement as a postsyntactic operation. In D. Harbour, D. Adger, & S. Béjar (Eds.), Phi theory (pp. 295–328). Oxford, UK: Oxford University Press.
  • Bybee, J. (1985). Morphology: A study of the relation between meaning and form. Philadelphia, PA: John Benjamins.
  • Bye, P., & Svenonius, P. (2012). Exponence, phonology and non-concatenative morphology. In J. Trommer (Ed.), The morphology and phonology of exponence: The state of the art (pp. 427–495). Oxford, UK: Oxford University Press.
  • Caballero, G., & Kapatsinski, V. (2015). Perceptual functionality of morphological redundancy in Choguita Rarámuri. Language, Cognition and Neuroscience, 30(9), 1134–1143.
  • Carrier, J. (1979). The interaction of morphological and phonological rules in Tagalog: A study in the relationship between rule components in grammar (Unpublished doctoral dissertation). MIT, Cambridge, MA.
  • Chomsky, N. (1993). A minimalist program for linguistic theory. In K. Hale & S. J. Keyser (Eds.), The view from Building 20: Essays in honor of Sylvain Bromberge (pp. 1–52). The Cambridge, MA: MIT Press.
  • Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.
  • Chomsky, N. (2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: MIT Press.
  • Chomsky, N. (2002 [1957]). Syntactic Structures (2nd ed.). Berlin, Germany: Mouton de Gruyter.
  • Chomsky, N., & Halle, M. (1968). The sound pattern of English. Cambridge, MA: MIT Press.
  • Cinque, G. (1999). Adverbs and functional heads: A cross-linguistic perspective. Oxford, UK: Oxford University Press.
  • Cinque, G. (2005). Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry, 36(3), 315–332.
  • Cinque, G. (2009). The fundamental left-right asymmetry of natural languages. In S. Scalise, E. Magni, & A. Bisetto (Eds.), Universals of language today (pp. 165–184). Berlin, Germany: Springer.
  • Comrie, B. (1980). Morphology and word order reconstruction: Problems and prospects. In J. Fisiak (Ed.), Historical morphology (pp. 83–96). The Hague, The Netherlands: Mouton.
  • Crysmann, B., & Bonami, O. (2016). Variable morphotactics in information-based morphology. Journal of Linguistics, 52, 311–374.
  • Damoiseau, R., & Gesner, J-P. (1975). J’apprends le créole haïtien. Paris, France: Éditions Karthala.
  • Dimmendaal, G. J. (1983). The Turkana language. Dordrecht, The Netherlands: Foris.
  • Egerød, S. (1965). Verb inflection in Atayal. Lingua, 15, 251–282.
  • Embick, D. (2007). Linearization and local dislocations: Derivational mechanics and interactions. Linguistic Analysis, 33, 303–336.
  • Embick, D., & Noyer, R. (2001). Movement operations after syntax. Linguistic Inquiry, 32, 555–595.
  • French, K. M. (1988). Insights into Tagalog: Reduplication, infixation, and stress from nonlinear phonology. Arlington: Summer Institute of Linguistics/University of Texas at Arlington.
  • Givón, T. (1971). Historical syntax and synchronic morphology. In Papers from the Seventh Regional Meeting of the Chicago Linguistic Society (CLS 7), 394–415.
  • Greenberg, J. (1966). Some universals of grammar with reference to the order of meaningful elements. In Joseph Greenberg (Ed.), Universals of language (pp. 73–113). Cambridge, MA: MIT Press.
  • Hale, K. (2001). Navajo verb stem position and the bipartite structure of the Navajo conjunct sector. Linguistic Inquiry, 32(4), 678–693.
  • Hale, W. G., & Buck, C. D. (1966). A Latin grammar. Tuscaloosa: University of Alabama Press.
  • Halle, M., & Marantz, A. (1993). Distributed Morphology and the pieces of inflection. In K. Hale & S. J. Keyser (Eds.), The view from Building 20: Essays in honor of Sylvain Bromberger (pp. 111–176). Cambridge, MA: MIT Press.
  • Halle, M., & Marantz, A. (1994). Some key features of Distributed Morphology. In A. Carnie, H. Harley, & T. Bures (Eds.), MIT Working Papers in Linguistics (MITWPL) 21: Papers on Phonology and Morphology (pp. 275–288). Cambridge, MA: MIT Press.
  • Harley, H. (2013). Getting morphemes in order: Merger, affixation, and head movement. In L. Cheng & N. Corver (Eds.), Diagnosing syntax (pp. 44–74). Oxford, UK: Oxford University Press.
  • Harris, A. C. (2017). Multiple exponence. Oxford, UK: Oxford University Press.
  • Haspelmath, M. (1993). The diachronic externalization of inflection. Linguistics, 31, 279–310.
  • Haugen, J. D., & Siddiqi, D. (2013). Roots and the derivation. Linguistic Inquiry, 44(3), 493–517.
  • Hay, J. (2002). From speech perception to morphology: Affix-ordering revisited. Linguistics, 78(3), 527–555.
  • Hay, J. (2003). Causes and consequences of word structure. London, UK: Routledge.
  • Hay, J., & Plag, I. (2004). What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language and Linguistic Theory, 22, 565–596.
  • Hyman, L. M. (2003). Suffix ordering in Bantu: A morphocentric approach. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 2002 (pp. 245–281). Dordrecht, The Netherlands: Kluwer.
  • Inkelas, S. (1989). Prosodic constituency in the lexicon. New York: Garland Press
  • Inkelas, S. (1993). Nimboran position class morphology. Natural Language and Linguistic Theory, 11, 559–624.
  • Inkelas, S. (2016). Affix ordering in Optimal Construction Morphology. In D. Siddiqi & H. Harley (Eds.), Morphological metatheory (pp. 479–511). Amsterdam, The Netherlands: John Benjamins.
  • Inkelas, S., & Cho, Y. Y. (1993). Inalterability as prespecification. Language, 69, 529–574.
  • Ito, J., & Mester, A. (2009). The extended prosodic word. In B. Kabak & J. Grijzenhout (Eds.), Phonological domains: Universals and deviations (pp. 135–194). Berlin, Germany: Mouton de Gruyter.
  • Jacques, G. (2013). Harmonization and disharmonization of affix ordering and basic word order. Linguistic Typology, 17, 187–215.
  • Jenks, P., & Rose, S. (2015). Mobile object markers in Moro: The role of tone. Language, 91(2), 269–307.
  • Julien, M. (2002). Syntactic heads and word formation. Oxford, UK: Oxford University Press.
  • Kim, Y. (2010). Phonological and morphological conditions on affix order in Huave. Morphology, 20, 133–163.
  • Kim, Y. (2015). Mobile affixation within a modular approach to the morphology phonology interface. In S. Manova (Ed.), Affix ordering across languages and frameworks (pp. 111–123). Oxford, UK: Oxford University Press.
  • Kiparsky, P. (1982). From cyclic to lexical phonology. In H. van der Hulst & N. Smith (Eds.), The structure of phonological representations (Vol. 1, pp. 131–175). Dordrecht, The Netherlands: Foris.
  • Koopman, H. (2017). A note on Huave morpheme ordering: Local dislocation or generalized U20. In G. Sengupta, S. Sircar, G. Raman, & R. Balusu (Eds.), Perspectives on the architecture and acquisition of syntax: Essays in honour of R. Amritavalli (pp. 23–48). Singapore: Springer.
  • Korotkova, N., & Lander, Y. (2010). Deriving affix ordering in polysynthesis: Evidence from Adyghe. Morphology, 20(2), 299–319.
  • Larson, R. K. (1988). On the double object construction. Linguistic Inquiry, 19(3), 335–391.
  • Lewis, G. (2000). Turkish grammar (2nd ed.). Oxford, UK: Oxford University Press.
  • Lieber, R. (1988). Configurational and nonconfigurational morphology. In M. Everaert, A. Evers, R. Huybregts, & M. Trommelen (Eds.), Morphology and modularity: In honour of Henk Schultink (pp. 187–215). Dordrecht, The Netherlands: Foris.
  • Marantz, A. (1982). Re reduplication. Linguistic Inquiry, 13(3), 435–482.
  • McCarthy, J. J., & Prince, A. S. (2001 [1993]). Prosodic morphology: Constraint interaction and satisfaction. Revision of 1993 Ms. Rutgers Optimality Archive #482.
  • McDonough, J. (2000). Athabaskan redux: Against the position class as a morphological category. In W. U. Dressler, O. E. Pfeiffer, M. Pöchtrager, & J. R. Rennison (Eds.), Morphological analysis in comparison (pp. 155–178). Amsterdam, The Netherlands: John Benjamins.
  • Melnar, L. R. (2004). Caddo Verb morphology. Lincoln: University of Nebraska Press.
  • Mithun, M. (2000). The reordering of morphemes. In S. Gildea (Ed.), Reconstructing grammar: Comparative linguistics and grammaticalization (pp. 231–255). Amsterdam, The Netherlands: John Benjamins.
  • Muriungi, P. K. (2008). Phrasal movement inside Bantu verbs: Deriving affix scope and order in Kı̂ı̂tharaka (Unpublished doctoral dissertation). University of Tromsø, Norway.
  • Muysken, P. (1988). Affix order and interpretation: Quechua. In M. Everaert, A. Evers, R. Huybregts, & M. Trommelen (Eds.), Morphology and modularity: In honour of Henk Schultink (pp. 259–279). Dordrecht, The Netherlands: Foris.
  • Nordlinger, R. (2010). Verbal morphology in Murrinh-Patha: Evidence for templates. Morphology, 20, 321–341.
  • Noyer, R. (1997). Features, positions and affixes in autonomous morphological structure. New York: Garland Press.
  • Paster, M. (2009). Explaining phonological conditions on affixation: Evidence from suppletive allomorphy and affix ordering. Word Structure, 2, 18–47.
  • Perrott, D. V. (1957). Teach yourself Swahili. London, UK: Hodder & Stoughton.
  • Plank, F. (1985). On the reapplication of morphological rules after phonological rules and other resolutions of functional conflicts between morphology and phonology. Linguistics, 23, 45–82.
  • Pollock, J.-Y. (1989). Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry, 20(3), 365–424.
  • Prince, A. S., & Smolensky, P. (2004 [1993]). Optimality Theory: Constraint interaction in generative grammar. Rev. and annotated ed. Malden, MA: Blackwell.
  • Rice, K. (1989). A grammar of slave. Berlin, Germany: Mouton de Gruyter.
  • Rice, K. (2000). Morpheme order and semantic scope: Word formation in the Athapaskan verb. Cambridge, UK: Cambridge University Press.
  • Rice, K. (2011). Principles of affix ordering: An overview. Word Structure, 4, 169–200.
  • Rizzi, L. (1997). The fine structure of the left periphery. In L. Haegeman (Ed.), Elements of grammar: A handbook of generative syntax (pp. 281–337). The Hague, The Netherlands: Kluwer.
  • Roberts, I. (2011). Head movement and the minimalist program. In C. Boeckx (Ed.), The Oxford handbook of linguistic minimalism. Oxford, UK: Oxford University Press.
  • Ryan, K. M. (2010). Variable affix order: Grammar and learning. Language, 86(4), 758–791.
  • Saba Kirchner, J. (2010). Minimal reduplication (Unpublished doctoral dissertation). University of California, Santa Cruz.
  • Saba Kirchner, J. (2013). Minimal reduplication and reduplicative exponence. Morphology, 23, 227–243.
  • Schachter, P., & Otanes, F. (1972). Tagalog reference grammar. Berkeley: University of California Press.
  • Selkirk, E. (2011). The syntax-phonology interface. In J. Goldsmith, J. Riggle, & A. C. L. Yu (eds.), The handbook of phonological theory (Rev. and updated 2nd ed.) (pp. 435–484). London, UK: Wiley-Blackwell.
  • Siddiqi, D. (2010). Distributed Morphology. Language and Linguistics Compass, 4(7), 524–542.
  • Siegel, D. C. (1974). Topics in English morphology. (Unpublished doctoral dissertation). MIT, Cambridge, MA.
  • Simpson, J., & Withgott, M. (1986). Pronominal clitic clusters and templates. In H. Borer (Ed.), Syntax and semantics 19: The syntax of pronominal clitics (pp. 149–174). New York: Academic Press.
  • Smolensky, P. (1997, May). Constraint conjunction II. Handout from talk given at Johns Hopkins Optimality Workshop, Baltimore.
  • Spencer, A. (1991). Morphological theory: An introduction to word structure in generative grammar. Oxford, UK: Blackwell.
  • Stairs, E. F., & Hollenbach, B. E. (1981). Gramática huave. In G. A. S. Kreger & E. F. Stairs (Eds.), Diccionario huave de San Mateo del Mar (pp. 283–391). Mexico City: SIL.
  • Stump, G. T. (1989). A note on Breton pluralization and the elsewhere condition. Natural Language and Linguistic Theory, 7(2), 261–273.
  • Stump, G. T. (1993). Position classes and morphological theory. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 1992 (pp. 129–180). Dordrecht, The Netherlands: Kluwer.
  • Stump, G. T. (1997). Template morphology and inflectional morphology. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 1996 (pp. 217–241). Amsterdam, The Netherlands: Kluwer.
  • Svenonius, P. (2008). The position of adjectives and other phrasal modifiers in the decomposition of DP. In L. McNally & C. Kennedy (Eds.), Adjectives and adverbs: Syntax, semantics, and discourse (pp. 16–42). Oxford, UK: Oxford University Press.
  • Travis, L. (1984). Parameters and effects of word order variation. (Unpublished doctoral dissertation). MIT, Cambridge, MA.
  • Trommer, J. (2003). The interaction of morphology and syntax in affix order. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 2002 (pp. 283–324). Dordrecht, The Netherlands: Kluwer.
  • Trommer, J., & Zimmermann, E. (2014). Generalized mora affixation and quantity manipulating morphology. Phonology, 31(3), 463–510.
  • Van den Berg, R. ((1989). A grammar of the Muna language. Dordrecht, The Netherlands: Foris.
  • Werner, H. (1997). Die ketische Sprache. Wiesbaden, Germany: Harrassowitz Verlag.
  • Wiering, E., & Wiering. M. (1994). The Doyayo language: Selected studies. Arlington: Summer Institute of Linguistics/University of Texas at Arlington.
  • Williams, E. (1981). On the notions “lexically related” and “head of a word.”’ Linguistic Inquiry, 12, 245–274.
  • Zimmermann, E. (2017). Morphological length and prosodically defective morphemes. Oxford, UK: Oxford University Press.


  • 1. Position templates were first use in describing the morphology of Athabaskan, although they have been posited for a number of different languages. Other position class systems include Nimboran (nir; isolate; Papua, Indonesia; Anceaux, 1965; Inkelas, 1993), Caddo (cad; Caddoan; Oklahoma, USA; Melnar, 2004), and the Siberian language Ket (ket; Yeniseian; Krasnoyarsk Krai, Russia; Werner, 1997).

  • 2. In this case, the difference correlates with a level difference in LPM, since the Latinate suffixes are typically stem-level, and the Germanic, word-level.

  • 3. Notice that the epenthetic vowel [e] does not alter the status of the affix as ‘small’.

  • 4. Recent research bearing on this issue is Caballero and Kapatsinski (2015), who find psycholinguistic evidence that multiple exponence in Choguita Rarámuri (tac; a.k.a. Tarahumara; Uto-Aztecan; Chihuahua, Mexico) enhances exponence of the category by boosting the signal-to-noise ratio.

  • 5. Gen houses phonological processes, including the building of prosodic structures, inserting, deleting, and altering the featural specifications of segments, spreading and delinking of autosegmental features, and so on.

  • 6. I suppress a 14th possibility here, which involves subextraction of VP from an AspP in TnsP.

  • 7. Abbreviations: sm = subject marker; cl = noun class; rtc = root clause.

  • 8. There are other possibilities for dealing with the material between RA and the root, including enclitic attachment to the prosodic word headed by RA, or procliticization to the prosodic word headed by the root. Since these structures would also violate *Adjoin, we would have to introduce positionally sensitive *Adjoin constraints. Instead of *Parse (σ‎), we would have constraints that militated against right-adjunction (enclisis) specifically, and/or adjunction inside the ‘compound’ prosodic word. I adopt *Parse (σ‎) here for the sake of argument.

  • 9. The theory of constraint conjunction in principle allows any constraint in the n-th power. Parse (σ‎)3, which bans consecutive violations of Parse (σ‎)2 (and dominates Parse (σ‎)2), would rule out a fourth candidate not shown in (23). This is (má.)ωma.ʔi.pàg.(lí.nis)ω, where RA appears to attach outside the nonactual prefix. Parse (σ‎)3 is not the constraint responsible for preventing this pattern from surfacing, however. RA may only float ‘inward’ (rightward from the nonactual prefix) for reasons that we cannot pursue here.