CHAPTER 1 INTRODUCTION
1.3 Thesis Organization
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
1.3 Thesis Organization
The thesis is organized as follows: Chapter 1 gives an outline of the current study, raising several research questions as above. Chapter 2 introduces earlier research pertinent to our discussion, and reviews the theoretical frameworks used in this thesis.
Chapter 3 examines the nature of English clipping under an OT schema. Also, in Chapter 4, OT is used to account for the nature of English blending and offers cross-linguistic evidence. Chapter 5 concludes this study.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5
2.1 Optimality Theory
2.1.1 Basic Concepts of Classic OT
Optimality Theory (OT) is a constraint-based framework which is proposed by Prince and Smolensky (1993, 2004). Different from the traditional model of The Sound pattern of English (Chomsky and Halle, 1968), the OT framework does not
develop on the basis of ordered rules1. Rather, it advocates the use of a set of
constraints to govern the grammar of language. Different languages are characterized by different rankings of the constraints. The main idea of OT expresses that a surface output results from competition against possible candidates. The competition is fulfilled by means of three components under the framework of OT, including CONSTRAINT, GENERATOR, and EVALUATION. The three components are responsible for the operation of OT, as indicated in (1).
1 For more details about the development of phonology from 1970s to 1990s, see Archangeli (1997) and McCarthy (2008).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(1) OT schema
Input CONSTRAINT
GENERATOR CON 1 CON 2 CON 3
Candidate (a)
Candidate (b) CON 1 >> CON2 >> CON3 Output
Candidate (c) EVALUATION
:
Given an input in the OT schema (1), component GEN associates it with possible representations. These representations are referred to as candidates. CON contains a set of ranked constraints. The ranking is used to evaluate the possible candidates.
There is no intermediate stage for evaluation within the model. Candidates go through a parallel evaluation of the ranking of the constraints and the constraint ranking selects the most harmonic one as the optimal output.
In OT, constraints are universal, violable, and ranked language-particularly.
Specifically, the universality of the constraints means that the same set of constraints is used to construct grammar in all languages. These constraints are ranked in
different ways, depending on the nature of a given language. In addition, these
constraints can be violated. The violation of a lower-ranking constraint can be allowed in order to fulfill the requirement of a higher-ranking constraint. The optimal output is selected if there is no violation or only the minimal violations of the constraints or by the satisfaction of a higher-ranking constraint. A formal OT approach to the choice of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the surface output is generalized as follows:
(2) CONSTRAINT 1 >> {CONSTRAINT 2, CONSTRAINT 3} >> CONSTRAINT 4
/Input/ CON1 CON 2 CON3 CON4
Candidate (a) *Candidate (b) *!
Candidate (c) *!
Candidate (d) *!
Conventionally, the evaluation of OT is illustrated by tableaux, as shown in (2). The top left-most cell presents an input, which is longitudinally followed by its possible representations. The top row lists relevant constraints and forms a ranking. Constraint 1 ranks above Constraint 2 and Constraint 3, which dominate Constraint 4. The finger marker indicates the best choice of the evaluation. The solid line distinguishes the priority accorded to one constraint over another. The left-side constraint takes a higher priority than the right-side constraint if they are separated by a solid line. The dotted line between two constraints shows that the ranking is unknown. The asterisk denotes a violation and the exclamation mark represents a fatal violation that rules out a losing candidate. When a candidate is fatally removed, the cells of lower-ranking constraints are shaded. The shading of cells means that the constraints are irrelevant to the
competition.
As Tableau (2) shows, candidate (d) is first ruled out because it violates the highest-ranked constraint. Candidate (b) and (c) are equally less harmonic because
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
they violate the constraints at the same level of the ranking. Although candidate (a) also incurs a constraint violation of the lowest ranked constraint, it satisfies the three higher-ranking constraints and thus emerges as the optimal output.
2.1.2 Correspondence Theory
The description of OT provided above concentrates on the mapping of the relationship between a given input and its surface output by means of faithfulness constraints, which regulate the consistency between two strings. The use of
faithfulness constraints is extended by McCarthy and Prince (1994, 1995). They propose Correspondence Theory, using faithfulness constraints to examine the relationship between the reduplicant and its base. Benua (1995) refers to the relation between two outputs as Output-to-Output correspondence. The main idea of
Correspondence Theory is that correspondent elements, including input-to-output (IO) relation and base-to-reduplicant (BR) relation, should be governed by faithfulness constraints. Thus, the family of faithfulness constraints has to be more specific to account for IO relation and BR relation, as in (3)-(5) proposed by McCarthy and Prince (1995).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(3) The MAX Constraint Family (Specific Instantiations)
MAX-BR: Every segment of the base has a correspondent in the reduplicant.
(Reduplication is total.)
MAX-IO: Every segment of the input has a correspondent in the output.
(No phonological deletion.)
(4) The DEP Constraint Family (Specific Instantiations)
DEP-BR: Every segment of the reduplicant has a correspondent in the base.
(Prohibits fixed default segmentism in the reduplicant.)
DEP-IO: Every segment of the output has a correspondent in the input.
(Prohibits phonological epenthesis.)
(5) The IDENT(F) Constraint Family (Specific Instantiations)
IDENT–BR(F): Reduplicant correspondents of a base [(F] segment are also [(F].
IDENT–IO(F): Output correspondents of an input [(F] segment are also [(F].
Constraints MAX-IO, DEP-IO, and IDENT-IO regulate the correspondence between the input and the output. They prohibit the deletion, addition, or change of a given input. Constraints MAX-BR, DEP-BR, and IDENT-BR guard the correspondence between outputs (base/ reduplicant). They protect the base from being deleted, inserted, or changing. Thanks to the Correspondence model, the relation between surface forms can be examined by OT schema.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.1.3 The schema of Output-to-Output Correspondence
Like the analysis of reduplication, Benua (1995) employs Correspondence Theory to investigate truncation, presenting a schema of output-to-output correspondence in (6).
(6) Truncation schema (Benua 1995) BT-Identity
Base - - - Truncated Form IO-FAITH
Input
The schema illustrates the mapping relation between the truncated form and its base.
It characterizes truncation as two types of correspondent relations: an input-to-base mapping (IO-correspondence) and an output-to-output mapping (OO-correspondence) between the base and the truncated form. To maintain BT-Identity, the constraints MAX-BT, DEP-BT, and IDENT-BT are relevant to govern the truncatory process.
2.1.4 Anchoring
Anchoring is another kind of faithfulness constraint which regulates the identity on the edges of two correspondent strings. It demands that the two strings must share
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the same constituent at the designed periphery, proposed by McCarthy and Prince (1995), as shown in (7):
(7) {RIGHT, LEFT}-ANCHOR(S1 , S2)(McCarthy & Prince, 1995)
Any element at the designated periphery of S1 has a correspondent at the designated periphery of S2
Let Edge(X, {L, R}) = the element standing at the Edge = L,R of X.
RIGHT-ANCHOR . If x=Edge(S1 , R) and y=Edge(S2 , R) then xRy.
LEFT-ANCHOR. Likewise, mutatis mutandis.
Left-Anchor requires the completeness of the constituent at the left edge of two independent relevant outputs. Right-Anchor demands that the constituent at the right edge of the output is identical to that of the base. Previous studies show that anchoring constraints often have an effect on reduplication. Marantz (1982) suggests that there is a tendency for locality between the reduplicant and the base: a reduplicant is placed as close as possible to its correspondent in the base. Thus, prefixing reduplicants
correspond to the initial part of the source and suffixing ones correspond to the final part of the source. McCarthy and Prince (1995) point out that the tendency for locality is motivated by anchoring constraints. The ranking L-Anchor >> R-Anchor accounts for prefixing reduplication while R-Anchor >> L-Anchor explains suffixing
reduplication. Nelson (2003) reveals an asymmetric relation between Left-Anchor and Right-Anchor and concludes that only left edge anchoring constraints are necessary.
Lunden (2004) notes that both anchoring and alignment constraints are needed to account for the few cases which do not follow the tendency of locality. In this thesis,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
anchoring constraints are employed to account for the preservation of the constituent of the source.
2.1.5 Cophonology Theory
In OT, there is a single fixed ranking of the constraints for each language. That is, constraints cannot be re-ranked in a language. If the ranking of the constraints is reordered, it composes the grammar of another language. However, on the
cophonology approach (Inkela and Zoll, 2007; Anttila and Cho, 1998; Orgun, 1996;
Ito and Mester, 1993, 1999), the phonology of a language can consist of co-existing distinct subgrammars if the language shows different phonological patterns. Anttila (1997, 2002) proposes the partial ordering model to account for cophonological
categories. Under this model, the grammar of a language is not completely determined.
Certain unspecified constraints are ranked in different ways depending on the nature of the subgroup of the grammar, as formalized in (8).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(8) The formulation of a grammar lattice (Anttila, 2002)
{ABC, ACB, BAC, BCA, CAB, CBA}
(Universal grammar) A>>B
{ABC, ACB, CAB}
(Language L)
A>>B A>>B
A>>C C>>B
{ABC, ACB} {ACB, CAB}
(Subgrammar 1) (Subgrammar2)
A>>B A>>B A>>B
A>>C C>>B C>>B
B>>C A>>C C>>A
{ABC} {ACB} {CAB}
(Subgrammar 3) (Subgrammar 4) (Subgrammar 5)
Given three constraints, there are six rankings to consist of the grammar. There is no further designated ranking in the total orderings (ABC, ACB, BAC, BCA, CAB, and CBA). When a ranking is designated (A>>B), the grammar that makes up Language L becomes more specific. From a cophonological perspective, constraints can be
reordered within a language. This thesis examines English clipping and blending types in light of cophonologies of morphological grammar.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2 Clipping as Word-Formation
Adams (1973) has discussed the situation in which clipped words are used, finding that they are commonly expressed in less formal situations than their source.
The informal usage can reflect the speaker’s familiar attitude toward the listener or to the object to which reference is being made. Adams observes, however, that no clear phonological rules govern the way in which the source is cut. Bauer (1983) agrees that the clipped form is shorter than its source while it is not known how many syllables of the source are retained. He doubts whether clipping is phonologically predicted. Minkova and Stockwell (2001) point out that the existence of clipped words can save time and space; however, they do not explain the grammar of clipping in more detail. These previous studies take a number of English clipped examples but they do not provide a systematic account for the operation of clipping, especially in
the area of phonology.
2.3 Blending as Word-Formation
In English, blending is another important mechanism to create new lexemes. In fact, it is a productive morphological device that exists in many languages of the world (Cannon 1986, Minkova and Stockwell 2001, Kemmer 2003, Gries 2004).
Cannon (1986) reviews various terms about and definitions of blending, finding that blends have been referred to as haplology, jumbles, lapses, telescopes, and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
portmanteau words. In the literature, the definitions of a blend are not clearly created
in a systematic way. The central focus of the studies above lies in a discussion of the terminology, definition, popularity, and usage of blends based on quantitative
grounds.
The following literature is concerned with the discussion of internal structure of the blend. Kelly (1998) assumes that the linguistic determinants that govern the structure of blends contain a coordinate structure and an orthographic factor. In spite of his assumption, many other studies have claimed that the structure of blends can be explained more accurately by reference to their occurrence as a phonological
phenomenon. Algeo (1977) compares blending with other word formations, suggesting that the creation of blends is predictable from the partial omission or deletion of overlapping of sounds, as in (9).
(9) The structure of blends (Algeo 1977)
a. They combine two or more forms, in which respect they are like compounds (hothouse) or affixed derivatives (hopeful).
b. Some of them omit part of one source form, or parts of both, in which respect they are like stump words (exam).
The observations in (9) do not describe how these sources are truncated in a
systematic way. Berman (1989) specifies the development of the blend in Hebrew as a
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
fusion of stem + stem combinations. This finding may be language-particular. For instance, the English blend smog from smoke and fog is an exception to the
combination of two stems. Kemmer (2003) presents a schema-based investigation of lexical blends, showing the possible range of form-meaning mapping between the blend and its sources. The schema-based is a constraint-based approach, providing an integrated and cognitive account for the conceptual structure of lexical blends. Gries (2004) discusses the contribution of the source in blend formation, finding that recognizability of the source and the similarities between the source and the blend have an effect on forming a blend.
Nevertheless, Kemmer and Gries’s studies do not clearly explain how the blend shares highly phonological similarities with its sources.
2.4 Bat-El (1996)
Bat-E1 (1996) argues that the sequence of the sources in Hebrew blends is not unordered, as given in (10). (The underlined boldface stands for the preserved part of the source.)
‧
(10) Hebrew blends (adapted from Bat-E1, 1996)
Sources Optimal blend Ill-formed blend
mitnaxalim, mexablim mitnaxablim
mitna(xalim)+(me)xablim
*mexablim
mexab(lim)+(mitnaxa)lim demokrat, diktator demoktator *dikrat
demok(rat) + (dik)tator dik(tator) + (demok)rat paxman, meiman paxmeiman
pax(man)+meiman
*meiman
mei(man)+(pax)man In (10), there are two sources for each blend. The left edge of the first source corresponds to the left edge of the blend. The right edge of the second source corresponds to the right edge of the blend. For the blend to be harmonic, the first source may consist of ‘mitnaxalim’, ‘demokrat’, or ‘paxman’ and the second source may consist of ‘mexablim’, ‘diktator’, or ‘meiman’. Bat-E1 proposes that the order of the sources is determined by constraint interaction rather than the function of the specific constraint which demands a sequence of the source elements, as shown below.
(11) Constraints on the number of syllables
σMAX: Every syllable in both elements of the base must have a correspondent in
the blend. (It would be violated as if syllables were deleted.)
σDEP: Every syllable in the blend must have a correspondent in both elements of
the base. (It would be violated as if syllables were added.)
‧
(12) adapted from Bat-E1, 1996
/ demokrat, diktator / σMAX σDEP
Tableau (12) considers two possible sequences of the sources. MAX and DEP pose restrictions on the number of the syllables in the blend rather than on the order of the sources. Hence, Bat-E1 proposes that the sources are not ordered randomly.
Furthermore, it is not necessary to use a specific constraint that requires the order of the sources. It can be indirectly determined by the interaction of independently motivated constraints. On the basis of Bat-E1’s proposal, I arrange the order of the sources inferred from the structure of blended outputs in English. Specifically, the left edge of the word is preserved in the blend is regarded as the first source, and the right edge of the word is preserved in the blend is regarded as the second source.
In Bat-E1’s study, σMAX and σDEP refer to syllable nodes instead of to the segmental content of the corresponding syllables. σMAX requires that the blend must not be smaller than the longest source. σDEP requires that the blend must not be larger than the longest source. Since σMAX outranks σDEP, the length of the blended output is never shorter than that of the longest source, as shown in (12) above. The
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
property of the size of Hebrew blends is different from that of English blends. In English, the syllabic length of the blend is identical to that of the second source. For the template, I use the family of IDENT constraints or a specific markedness
constraint to account for the size of the blend. MAX and ANCHOR are used to require the segmental preservation of the source.
2.5 Lappe (2008)
Lappe (2008) examines the structure of English truncated patterns on the basis of three levels of description. The first level is word structure, where truncated forms can be described in terms of the number of syllables, stress assignment, and the part of the base which is retained. The second level is syllable structure, where truncated forms are described in terms of syllabic constituents such as onset, nucleus, and coda. The third is segmental level, which imposes restrictions on the kind of segments that make up the output. Lappe assumes that truncatory patterns differ from one another
according to the degree of constraints on different levels of structure. Each truncatory process is subcategrized in terms of the ranking of markedness and faithfulness constraints. Thus, Lappe shows that the difference between monosyllabic and disyllabic clippings is decided by the competition of the various specific MAX constraints.
This paper divides English monosyllabic and disyllabic clipped data into several
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
types and proposes ANCHOR constraints instead of MAX constraints to account for different types of clipped outputs. One reason for the adoption of ANCHOR
constraints is that MAX is often used to regulate the segmental relation between the base and truncated output. It disallows the deletion of segments for the sake of completeness of the base. With respect to the positional relation between two strings, MAX constraints do not mention the directionality and locality. Further, the family of MAX constraints cannot rule out a possible candidate with the insertion since they are not concerned with the integrity of the truncated output. ANCHOR, like MAX, is a type of faithfulness constraint. Furthermore, it imposes a positional correspondence between two strings. Specifically, the truncated output must share constituents with the base on the same edge.
2.6 Summary
In brief, clipping and blending have been surveyed in recent years. Many studies claim that they should be part of the grammar of language. Nevertheless, few
researchers have provided a systematic account for the internal structure of the morphological processes from a phonological perspective. The use of the family of MAX constraints in Lappe’s study (2008) is not enough to account of the positional faithfulness between a blend and its sources. In addition, variations in clipping and blending have not been characterized in more detail. Following this discussion, I
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
propose the application of Cophonology which makes it possible to analyze the variations. Chapter 3 and Chapter 4 will demonstrate how OT and Cophonology account for the internal structure of English clipping and blending, respectively.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
This chapter examines the segmental preservation of English clipping from morphological and phonological perspectives. The analysis of the process is based on Correspondence Theory. The relation between the clipped output and its source is considered Output-to-Output correspondence. Under the truncation model (Benua 1995), the mapping relation between a clipped word and its source can be treated as follows:
(13) Clipping: laboratory lab
SC-Identity (OO-FAITH)
Base(Source):
[]
- - - Clipping:
[]
IO-FAITH
Input:
//
In the model of (13), Input-to-Output correspondence (IO-Faith) relates the input to the source, and Output-to-Output correspondence (OO-Faith) relates the source to the clipped form. The clipped output derives from the preservation of the segment of the source. The SC-identity is regulated by the constraints in the following discussion.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Chapter 3 is organized as follows: Section 3.1 introduces clipping types that preserve the morpheme of the source, and provides an OT account for the different types. Section 3.2 introduces clipping types that retain the phonological constituents of the source, and also takes an OT approach to account for the derivation of the different types. Section 3.3 presents ranking contradictions under classic OT, and employs a cophonological theory to solve inconsistencies. Section 3.4 summarizes
Chapter 3 is organized as follows: Section 3.1 introduces clipping types that preserve the morpheme of the source, and provides an OT account for the different types. Section 3.2 introduces clipping types that retain the phonological constituents of the source, and also takes an OT approach to account for the derivation of the different types. Section 3.3 presents ranking contradictions under classic OT, and employs a cophonological theory to solve inconsistencies. Section 3.4 summarizes