Professional Documents
Culture Documents
A thesis submitted in partial satisfaction of the requirements for the degree Master of Arts in Linguistics
by
Table of Contents
0. INTRODUCTION.....................................................................................................1 1. FARSI.........................................................................................................................2 2. THE EPENTHETIC VOWEL.................................................................................5 2.1 NOVEL FORMS ........................................................................................................7 3. ANALYSIS ..............................................................................................................14 3.1 SELECTING /E/ AS THE EPENTHETIC VOWEL ..........................................................15 3.2 FEATURE SPREADING AND BLOCKING BY CONSONANTS .......................................17 3.3 SPREADING OF [+ FRONT].....................................................................................19 3.4 SPREADING OF ROUND VOWELS............................................................................24 3.4.1 Consonant duration .....................................................................................24 3.4.2 Formalizing duration-based constraints......................................................25 3.4.3 Degree of rounding ......................................................................................29 Spreading of round vowels across [0DUR]:.......................................................30 Spreading of round rowels across [1 DUR]:.......................................................33 Spreading of round vowels across [2DUR]:.......................................................35 3.5 SPREADING OF OTHER FEATURES .........................................................................44 3.6 SPREADING AND EPENTHESIS A COMBINED STRATEGY.......................................45 4. ALTERNATIVE ANALYSES OF COPY EPENTHESIS..................................49 4.1 CORRESPONDENCE THEORY .................................................................................50 4.2 PLACE AND STRICTURE THE DETERMINING FACTORS .........................................51 5. DISCUSSION ..........................................................................................................52 REFERENCES............................................................................................................56
Table of Tables
TABLE 1.1 IO-CONTIGUITY-SC >> ANCHOR-L >> IO-CONTIGUITY ..............................................5 TABLE 2.1 THE LIST OF STIMULI AND PARTICIPANTS RESPONSES ..................................................8 TABLE 2.2 GENERALIZATION OF THE PARTICIPANTS RESPONSES ...........................................14 TABLE 3.1 - DEP (MF) >> DEP (UF) .......................................................................................16 TABLE 3.2 *C-LOW >> DEP (UF).........................................................................................19 TABLE 3.3 *NCOR-FRONT >> DEP (UF) ...............................................................................21 TABLE 3.4 - *NCOR-FRONT >> DEP (UF) ................................................................................22 TABLE 3.5 - *1DUR-LOW >> DEP (UF) >> *0DUR-LOW ..............................................................28 TABLE 3.6 DEP (UF) >> *0DUR-1RD ....................................................................................31 TABLE 3.7 DEP (UF) >> *0DUR-2RD ....................................................................................32 TABLE 3.8 - *1DUR-1RD >> DEP (UF) .....................................................................................34 TABLE 3.9 - *1DUR-1RD >> DEP (UF) >> *1DUR-2RD ................................................................34 TABLE 3.10 *2DUR-1RD >> DEP (UF) ..................................................................................36 TABLE 3.11 DEP (UF) >> *2DUR-2RD ..................................................................................37 TABLE 3.12 *2DUR-2RD & *NLAB-2RD >> DEP (UF) >> *1DUR-2RD .........................................38 TABLE 3.13 - *2DUR-2RD & *NLAB-2RD >> DEP (UF) >> *1DUR-2RD ..........................................39 TABLE 3.14 *NLAB-2RD >> DEP (UF) >> *1DUR-2RD .............................................................41 TABLE 3.15 *NLAB-2RD >> DEP (UF) >> *1DUR-2RD .............................................................42 TABLE 3.16 *2DUR-2RD & *NLAB-2RD >> DEP (UF) >> *1DUR-2RD, *NLAB-2RD........................42 TABLE 3.17 DEP (UF) >> *C-HI ..........................................................................................44 TABLE 3.18 - DEP (MF) >> 1DUR-1RD >> DEP (UF) ..................................................................46 TABLE 3.19 SP-ALL >> DEP(UF) .........................................................................................48
0. Introduction Loanwords are often modified to conform to the phonological patterns of the native language. Some of these changes include phoneme substitution, stress adjustment, and epenthesis. With respect to epenthesis, there are languages that epenthesize a copy vowel (e.g., /tri//tiri/), as well as languages that epenthesize an invariant vowel (e.g., /tri//tri/). However, there are also languages that show a split between copy and default epenthesis. This split is often contextually conditioned. This paper examines a case of split epenthesis in Farsi and offers an analysis within the framework of Optimality Theory. This theory employs a notion of constraint dominance and a mechanism for selecting the optimal output with respect to a set of ranked constraints (for a detailed understanding of the theory, see Prince & Smolensky 1993). Epenthesis in Farsi arises in loanwords. Since no word in Farsi may begin with two consonants, initial consonant clusters in loanwords are broken up by vowel epenthesis. Sometimes /e/ is inserted (e.g., Florida [feloida]), and sometimes the inserted vowel is a copy of the following vowel (e.g., the French borrowing bross brush [boos]). I will propose that the cases of copying are the result of the vocalic features spreading from one vowel to another. The data exhibit a case of consonant participation in the spreading of vocalic features. If a consonants features are compatible with the vocalic features that are spreading (in a sense to be explained below), the inserted vowel is a copy of the
following vowel (i.e., the vowels share their features). When a consonants features are not compatible with the feature(s) being spread, the default vowel /e/ is inserted. I will argue that there are three main factors that predict the compatibility, with respect to spreading, of the consonants features with the vowels features. First, this compatibility depends on the consonants place of articulation and the vocalic features that are spreading (e.g., labial obstruents and round vowels are compatible, whereas non-labial obstruents are not compatible with round vowels). Second, the rounded vowels must be distinguished from each other based on their degree of rounding. Third, a consonants participation in the spreading is influenced by its duration. This paper is organized as follows. In the next section, section 1, I will offer some background on Farsi. In section 2, I will discuss the existing epenthesis pattern in loanwords, as well as data elicited from native speakers on novel forms, in order to test epenthesis patterns. The analysis is given in section 3. Finally, in section 4, I will compare my account with other proposed analyses, followed by a general discussion in section 5. 1. Farsi Farsi has a six-vowel system, with three lax vowels (/e/, //, /o/) and three tense vowels (/i/, /u/, //). The distinction between the two sets of vowels is sometimes assumed to be a difference in length, so that lax vowels are short and tense vowels are long (Comrie 526). // is rarely borrowed; instead, // in foreign words
surfaces as // (e.g., pass is borrowed as [ps] and not as [ps]). The other vowels in the inventory do appear regularly in loanwords. Complex codas are common in Farsi. Some examples are: (1) xtm qb qtl sobh end/funeral grave murder morning
On the other hand, complex onsets are absent. Indeed, the absence of complex onsets is the motivation for vowel epenthesis in loanwords; inserting a vowel breaks up the initial consonant cluster. Some examples are given in (2): (2) ski [eski] Florida [feloida] plastic [pelstik] bross (Fr.) [boos] freezer [fiize] blouse (Fr.) [buluz] There are two issues to be addressed with respect to the epenthesis observed in forms such as above. The first issue is to determine the location of the epenthetic vowel. This means predicting whether a [CCV] sequence results in [eCCV] (e.g., /ski/[eski]), or results in [CeCV] (e.g. /florida/[feloid]. The patterns observed are as follows: Prothesis - The clusters involved in prothesis are /s/+consonant (SC) clusters. In these cases, it is consistently observed that the epenthetic vowel is located before the /s/:
(3)
SC clusters /slv/ Slav [eslv], *[selv] /spot/ sport [espot], *[sepot] /stejk/ steak [estejk], *[setejk] /stop/ stop/ [estop], *[setop] Anaptyxis In all other clusters the vowel is anaptyctic; the epenthetic vowel is
located between the two members of the cluster: (4) Non-SC clusters /tfik/ traffic [tefik], *[etfik] /kom/ chrome [koom], *[okom], *[ekom] /pes/ press [pees], *[epres] /plstik/ plastic [pelstik], *[eplstik] There is cross-linguistic evidence for associating anaptyxis-prothesis asymmetries with the nature of the consonants involved in the process. In fact, it has been suggested that the reason for the existence of such asymmetries is to maximize auditory similarity between the input and the output (Fleischhacker 1999). In an Optimality theoretic analysis, one option for predicting the location of the epenthetic vowel is to specify that IO-CONTIGUITY (McCarthy & Prince 1995) must be split between SC clusters and the general contiguity constraint. (5) IO-CONTIGUITY: Segments that are contiguous in the input must be contiguous in the output. (This constraint penalizes cases of wordinternal epenthesis and deletion.)
(6)
IO-CONTIGUITY-SC: Sibilants that are adjacent to a consonant, C1, in the input must be adjacent to C1 in the output.
Furthermore, the more general IO-CONTIGUITY is outranked by ANCHOR-L, as defined below: (7) ANCHOR-L: a segment C1 at the left edge of the input has a correspondent segment C1 at the left edge of the output. Such a ranking will derive the observed asymmetry, as the following table shows:
Table 1.1 IO-CONTIGUITY-SC >> ANCHOR-L >> IO-CONTIGUITY
IO-CONTIGUITY-SC *!
ANCHOR-L
IO-CONTIGUITY * *
*!
However, this paper is not primarily concerned with the location of the epenthetic vowel, but rather the factors and constraints that determine which vowel is epenthesized. 2. The choice of epenthetic vowel The epenthetic vowel is always either /e/, or a copy of the following vowel. The quality of this copy vowel, I will claim, is the result of feature spreading from the subsequent vowel. Variation arises because the spreading is blocked in some cases; in these, feature insertion occurs, and the result is /e/.
First, all SC clusters have epenthetic /e/, as in (3) above. In non-SC clusters, the second member of the cluster in the existing loanwords is either /l/ or //. In these cases, if the cluster is followed by a high vowel, then there is copy epenthesis. Examples are given in (8): (8) freezer [fiize] plisse (Fr.) [pilise] groupe (Fr.) [uup] flute [fulut] If the cluster is followed by a low vowel, then /e/ is inserted (e.g., traffic /tfik/[tefik]; plan /pln/[peln];). If the cluster is followed by a mid vowel, then there is copy epenthesis if the second member is // , and default epenthesis if the second member is /l/. Examples are given in (9) and (10): (9) //: Copy Epenthesis chrome [koom] preuve (Fr.) [poov] bronze (Fr.) [boonz] (10) /l/: Default Epenthesis Florida [feloid] flottasion (Fr.) [felotsijon] blonde I(Fr.) [belond] It must be noted that, in general, words with copy epenthesis typically have two variants. One variant, containing the default epenthetic vowel /e/, reflects a
careful, formal pronunciation. The variant containing the copy epenthetic vowel represents the colloquial, informal speech. The existing loanwords with initial consonant clusters are mainly borrowed from French, English, and Russian. I consulted a large corpus (about 2000 words) of Farsi European loanwords (Muhammad-Khan 1983) and found no exceptions to the mentioned patterns. However, the set of consonant clusters found in existing loanwords is limited. In order to discover any further factors that influence feature spreading in the data, I designed and tested a set of novel forms.
encouraged to try using the words in a sentence and produce the sentence out loud (e.g. I have to go to the pharmacy because I need ptustrophin.). The list of the visual stimuli and the participants responses is provided below in table 2.1. The first column contains the stimuli. I included some existing words, which are marked by (). The table provided here is alphabetized by the loanword, but the list for the speakers was randomized. Some novel words were associated with a made-up meaning in order to help participants use them in a sentence, and those meanings are listed in the second column. The third column contains the responses that participants offered. If an additional response was available (either offered by another participant or suggested by the experimenter and accepted by the participants) it is listed in the fourth column.1
Table 2.1 The list of stimuli and participants responses
Gloss
Given response
boou (B) buruu (A,C,D) diink (A,B,C,D) felsk (A,B,C) felks (D) feloid (A,B,C,D) fulut (A,B,C,D)
Also possible
buruu2 (B) borou (A,C,D)
As mentioned earlier, for all the words with copy epenthesis, there is always a formal variant in which /e/ is epenthesized. That form is only specified when it was the form volunteered by a participant. Otherwise, it is left out of the table because it is always accepted as a formal form. This variation might be a result of the influence of Farsi orthography. The vowel following // is written in Farsi and is ambiguous between /o/ and /u/. This means that for some speakers the epenthetic vowel is followed by a mid round vowel, and for some it is followed by a high back vowel, hence the variation in the quality of the epenthetic vowel.
2
frizer froyd klineks mute plastik prajd ski smoking sport standard switch trafik twist bmaktik bnesky bzinpit dmuse drup fsunius ftansie fzerdi qmildan ktinora ktopski kvuski lkimnik lramik lristiyus lrobto lsiknel
fiize (A,C,D) feize (B) foojd (A,C,D) feojd (B) kilineks (A,B,C,D) mijut (A,B,C,D) pelstik (A,B,C,D) pejd (A,B,C,D) eski (A,B,C,D) esmokin (A,B,C,D) espot (A,B,C,D) estndd (A,B,C,D) suvit (A,C,D) suit (A,B)
traffic
Bulgarian car
bemktik (A,B,C,D) beneski (A,B,C,D) bizinpit (A,B,C) bezinpit (B) dumuze (A,B,C,D) duup (A,B,C,D)
European city
kitino (A,B,C,D) ketopski (A,B,C,D) kovuski (A,B,C) kuvuski (D) lekimnik (A,B,D) elkimnik (C) lemik (A,B,C) lmik (D) liistijus (A,B,C,D) loobto (A,B,C,D) kovuski (D) kuvuski (A,B,C)
Swedish candy
lisiknel (A,C,D)
lesiknel (B) lvampilin lvoski mboldin mlandrin mloztikian mronst mrust mtumir mzarsky nforons nlestrophine nlismits nlustic nmaritin nmisious ntarisin ntindon psondik ptandiz ptustrofin rlumine rmitidin rmostism rmuspi shnabel snoopy sshomis tfanilin tfindor tkulster Type of chemical Type of chemical A motorcycle brand Type of medication Type of medication Type of chemical Toothpaste brand A Russian name A philosopher Cereal brand Type of medication Type of medication levmpilin (A,B,C) lvmpilin (D) levoski (A,B,C,D) meboldin (A,B,C,D) melndiin (A,C,D) melndein (B) meloztikjn (A,B,C,D) meonest (A,B,C,D) meust (A,B,C,D) metumi (A,B,C,D) mezski (A,B,C,D) nefoon (A,B,C,D) nelesteofin nilismitz (A,C,D) nelismitz (B) nolustik (A,C,D) nelustik (B) nemitin (A,B,C,D) nemizijus (A,B,C,D) netisin (A,B,C,D) nitindon (A,C,D) netindon (B) pesondik (A,B,C,D) petndiz (A,B,C,D) petusteofin (A,B,C,D) olumine (A,C,D) elumine (B) emitidin (A,C,D) emetidin (B) ulumine (A,B,C,D) nulustik (A,B,C,D) * lvmpilin (A,B,C)
Some school of thought emostizm (A,B,C,D) umuspi (A,B,C,D) enbel (A,B,C,D) esnupi (A,B,C,D) seomiz (A,B,C,D) tefnilin (A,B,C,D) tefindo (A,B,C,D) tekulste (A,B,C,D)
10
telondik (A,B,C,D) tenois (A,B,C,D) tepois (A,B,C,D) vumuzi (B,C,D) vemuzi (A) venoditz (A,B,C,D) uvunti (A,B,C,D) zeunsken (A,B,C,D)
Not all the responses were unanimous; nevertheless, variation was rare. For example, there is only one case, involving // (lramik), in which a response offered by one participant ([lramik]) was rejected by other participants.3 A majority of the variation is due to participant Bs tendency to epenthesize /e/ more frequently than other speakers, and other speakers judging the varied form as stilted. Another case of variation arises in forms in which the initial cluster is followed by /u/. In cases in which spreading is allowed, the epenthetic vowel can sometimes surface as /o/ (specifically, in [kovuski] ~ [kuvuski], [nolustik] ~ [nulustik], and [rolumine] ~ [rulumine]). These cases are plausibly the result of hypercorrection. Colloquial Farsi has a process of height harmony whereby [oCu] sequences surface as [uCu] in casual speech (e.g., /doud/ greeting surfaces as colloquial [duud]). This means that speakers are aware of abundant cases of [uCu] which surface as [oCu] in formal pronunciation. It is possible that in an elicitation situation speakers use the more formal form of speech. This is further confirmed by the fact that, in all such cases in
This case will be addressed in section 3.4.2 where the constraint in (25) and its implications will be discussed.
3
11
The patterns depicted above conform to the existing patterns in loanwords that were discussed in section 2. Table 2.2 below captures the generalizations that were observed in participants responses. /e/ designates default epenthesis and copy refers to copy epenthesis. The cases that are missing from the data are left blank:
Table 2.2 Generalization of the participants responses Following Vowel Cluster u i o
/Cb/ /Cp/ /Ct/ /Cd/ /Ck/ /C/ /Cq/ /Cm/ /Cn/ /Cf/ /Cv/ /Cs/ /C/ /Cz/ /C/ /Cl/ /C/
The above table shows the pattern in the data. The analysis offered in the following section makes predictions regarding the blank cells in the table. I am currently planning experiments to that will address these cases. 3. Analysis The basic assumption of this analysis is that a vowel is inserted to break up an illegal consonant cluster. The preferred mechanism by which the epenthetic vowel is realized is feature spreading: this avoids violating faithfulness constraints of the DEP (FEATURE) type. Whenever feature sharing is blocked, the epenthetic vowel is /e/.
14
Nevertheless, the insertion of /e/ violates a faithfulness constraint against insertion. The constraints violated by the insertion of /e/ are defined such that they penalize the insertion of the following features: [-LOW], [-HI], [+FRONT], and [-ROUND]. I will collapse the following constraints into DEP (UNMARKED FEATURE): 1)
DEP (-LOW); 2) DEP (-HI); 3) DEP (-ROUND), and 4) DEP (+FRONT).
(13)
DEP (UF): A surface instance of [+FRONT], [-ROUND], [-HI], AND [LOW] must have an identical underlying correspondent.
15
The tableau shows how /e/ is preferred to /u/. Other vowels also lose to /e/, since //, //, // and // have 1,2,2, and 3 violations of DEP (MF) respectively. In other words, /e/ is the only vowel that does not violate DEP (MF).
F The focus of this paper is mainly the phenomenon of V-to-V contiguity. The constraints proposed in the analysis are such that they penalize candidates that have certain autosegmental representations, as illustrated below: (16) * FC FV: A [+FC] consonant cannot be associated with [+FV] (where FC is a consonantal feature and FV is a vocalic feature). *FC X FV Some examples of the constraints of the above type are the following, where FC refers to some feature of a consonant: