i
Subcategorization and Optimality Theory: The Case of Spanish Diminutives
By
JASON ALLEN SMITH
B.A. (Western Washington University) 1997 B.A. (Eastern Washington University) 2004 M.A. (University of California, Davis) 2006
DISSERTATION
Submitted in partial satisfaction of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
Spanish
in the
OFFICE OF GRADUATE STUDIES
of the
UNIVERSITY OF CALIFORNIA
DAVIS
Committee in Charge
Professor Travis Bradley, Chair Professor Robert Blake Professor Orhan Orgun
Fall 2011
ii
Subcategorization and Optimality Theory: The Case of Spanish Diminutives
© 2011
by
Jason Allen Smith
iii
Abstract
Subcategorization and Optimality Theory: The Case of Spanish Diminutives
by
Jason Allen Smith
Doctor of Philosophy in Spanish
University of California, Davis
Professor Travis G. Bradley, Chair
In the present work I utilize the framework of Optimality Theory (Prince
& Smolensky 1993/2004) to account for the pattern of diminutive formation in
Spanish. I argue that Spanish diminutivization is an example of suppletive
allomorphy in which the -ito{o, a} and -cit{o, a} allomorphs are not related to one
another through either a process of segmental insertion or segmental deletion.
Crucially, I propose that both allomorphs are available to attach to all word types
in Spanish and that unattested output candidates, such as *casecita or *balconito,
are eliminated by faithfulness and markedness constraints that, with the exception
of RESPECT (Bonet 2006), are phonological in nature.
In my analysis I argue that Spanish substantives are divided into two
morphological classes for dimininutive formation. Class D1 includes all those
words that end in the unstressed canonical class markers /o/ and /a/. Words from
this class are lexically specified to select the –it{o, a} allomorph—a lexical bond
that is enforced by the RESPECT constraint. All other words that undergo
diminutivization belong to Class D2 and are not subject to any lexical
iv
specification. Therefore, allomorph selection in Class D2 words follows from the
interaction of phonological faithfulness and markedness constraints.
In addition to these arguments, I assert in this work that Spanish
diminutivization is always a process of suffixation in which the morphemes attach
to a stem that crucially does not contain the final vocalic element: casa~casita
[kas- + it{o, a}]. Contrary to analyses in support of infixation of -it- between the
root and the class marker morpheme, I argue that the class marker in diminutive
forms is a set {o, a} that is part of the inflexional possibilities of the diminutive
itself. Selection between /o/ and /a/ is determined by the interaction of three
constraints: 1) IDENT-VOWEL limits the final vowel in derived forms to the
inflectional capacity of the diminutive suffix, indicated by {o, a}; 2) REFLECT
mandates that derived forms reflect all segments present in the base form; and,
finally, 3) FEM = /a/ requires the selection of the final vowel /a/ for feminine
derived forms and prohibits its selection for non-feminine derived forms.
While this dissertation focuses primarily on an analysis of the diminutive
data for Sonoran Spanish in Northern Mexico presented in Crowhurst (1992), I
also demonstrate how my analysis accounts for the dialectal variation found in
Peninsular Spanish (Colina 2003), Paraguayan (Jaeggli 1978), and Nicaraguan
(Miranda 1999) through re-ranking of constraints.
v
Subcategorization and Optimality Theory: The Case of Spanish Diminutives
Table of Contents Chapter 1. Introduction to Spanish Diminutive Formation ……………………………………. 1 I. Introduction ………………………………………………………………………………………. 1 II. Spanish Diminutive Data ………………………………………………………........................... 2 III. Theoretical Issues: Underlying Assumptions and the Phonology-Morphology Interface ……... 6 Chapter 2. Previous Analyses of Spanish Diminutive Formation ……………………………. 10 I. Jaeggli (1978) …………………………………………………………………………………… 10 II. Prieto (1992) …………………………………………………………………………………… 12 III. Crowhurst (1992) ……………………………………………………………………………… 19 IV. Ambadiang (1997) …………………………………………………………………………….. 29 V. Elordieta and Carreira (1996) …………………………………………………………………... 31 VI. Miranda (1999) ………………………………………………………………………………… 38 VII. Colina (2003) ………………………………………………………………………………….. 44 VIII. Stephenson (2004) ……………………………………………………………………………. 50 Chapter 3. Class Marker Selection in Diminutive Formation …………………………………. 55 I. Introduction ……………………………………………………………………………………… 55 II. Review of Harris’ Word Classes ……………………………………………………………….. 56 III. Harris’ Word Classes and Diminutivization …………………………………………………… 62 IV. Underlying Assumptions ………………………………………………………………………. 79 Chapter 4. Morphological and Phonological Interactions in Diminutive Allomorph Selection ………………………………………………………………… 101 I. Introduction ……………………………………………………………………………………… 101 II. Analysis of Diminutive Class D1 Words in Sonoran …………………………………………… 102 III. Analysis of Diminutive Class D2 Words in Sonoran ………………………………………….. 113 IV. Additional Support for High-Ranking RESPECT Constraint …………………………………. 121 V. Dialectal Variation in Paraguayan and Nicaraguan Spanish …………………………………… 128 VI. Further Evidence for Constraint Rankings …………………………………………………….. 131 VII. Subcategorization Frames on Allomorphs: An Alternate View ……………………………… 138 Chapter 5. Concluding Remarks ………………………………………………………………… 141 I. Review of Underlying Assumptions …………………………………………………………….. 141 II. Review of OT Analysis ………………………………………………………………………… 145 III. Possible Objections and Future Work …………………………………………………………. 151 References ………………………………………………………………………………………… 154
vi
Acknowledgements My deepest gratitude goes to those wonderful people in my life who always believed in me much more than I believed in myself and were kind enough to remind me of it: My grandparents, Raymond and Eleanor Raines, who loved me unconditionally and gave me an example of how to be a good person—they know I am still trying. My 9th grade English teacher, Pat Roberts, who reminded me that learning should be fun. My dear friends, Brian Hagenbuch, Henry Murphy, Lynn and Carolee Pomeroy, and James Spears, who have given me so many laughs and thrown out so many lifelines. My first Spanish linguistics professor, Teresa Oteíza, who gave so generously of her time during the first year of her first position. I’ll be paying it forward for many years to come. Our Graduate Program Coordinator in the department, Kay Green, who saved many of us from the brink time and time again. Professor Robert Blake, who is a gentleman, a scholar, and one impressive educator. My dissertation advisor and friend, Travis Bradley, who introduced me to a whole new world and has helped me to find my place in it. My lovely wife, Iliana, who has never doubted me and has never let me doubt myself. Life just gets better and better with you. My parents, who had the faith to let me find my own way.
1
Chapter 1
Introduction to Spanish Diminutive Formation
I. Introduction
Several studies have attempted to account for the pattern of diminutive
formation in Modern Spanish within generative linguistics (Ambadiang 1997,
Colina 2003, Crowhurst 1992, Elordieta & Carreira 1996, Jaeggli 1978, Miranda
1999, Prieto 1992, Stephenson 2004, among others). While some researchers
have limited their analyses to a specific dialect (see Jaeggli 1978 for Paraguayan,
Miranda 1999 for Nicaraguan, and Crowhurst 1992 for Sonoran dialect of
Northern Mexico), still others have attempted to account for the data across
dialects, particularly through the framework of Optimality Theory (Colina 2003,
Elordieta & Carreira 1996, Stephenson 2004).
In this work I will present a new analysis of Spanish diminutivization
within the Optimality Theory framework (Prince & Smolensky 1993/2004) that is
greatly indebted to these previous accounts. For example, several of the
constraints that I utilize in my analysis are similar or identical to constraints
presented in the aforementioned OT analyses of diminutivization and, in some
cases, find their counterparts in the earlier rules-based accounts. That said, I
believe the present paper offers a more simplified approach to this complex
example of the interface between morphology and phonology, and, at the same
time, provides a more complete account of the various word types and dialectal
variation reported in the literature.
2
In Section II of this introductory chapter I present the data on Spanish
diminutives and highlight the variation among several dialects. Then, in Section
III, I discuss the major issues and underlying assumptions that must be addressed
with any account of diminutive formation and I describe briefly how my analysis
approaches these issues and outline my own underlying assumption about this
process. Finally, I end with a brief introduction to the preceding chapters.
II. Spanish Diminutive Data
The following data for Spanish diminutive formation is based on examples
presented in Jaeggli (1978) for Paraguayan, Crowhurst (1992) for Sonoran
(Northern Mexico), Miranda (1999) for Nicaraguan, and Colina (2003) for North
Central Peninsular.
1) Words that end in an unstressed /o/ or /a/ and do not contain a diphthong all
select the -it{o, a} diminutive:
-o/-a: -it{o, a} libro librito “book” casa casita “house”
2) Disyllabic words that end in an unstressed /e/ all select the -(e)cit{o,a}
diminutive:
-e: -(e)cit{o,a} clase clasecita “class” madre madrecita “mother”
3
3) Trisyllabic words or longer that end in an unstressed /e/ all select the -it{o, a}
diminutive:
-e: -it{o, a} chocolate chocolatito “chocolate”
comadre comadrita “godmother of one’s child”
4) Disyllabic words or longer that end in a consonant all select the -cit{o, a}
diminutive: 1
-C: -cit{o, a} pintor pintorcito “painter”
corazón corazoncito “heart”
5) All words ending in a stressed vowel select the -cit{o, a} diminutive:
-ˈV: -cit{o, a} té tecito “tea”
café cafecito “coffee”
matiné matinecito “matinee”
6) Monosyllabic words ending in a consonant select -cit{o, a} in the Nicaraguan
and Paraguayan dialects, but -ecit{o, a} in Sonoran and North Central Peninsular:
Paraguayan and Nicaraguan:
-cit{o, a} pan pancito “bread”
Sonoran and Peninsular:
-ecit{o, a} pan panecito 1 One exception to this generalization is found among the diminutive selection of some speakers when the final consonant is /l/. While some maintain the -citV pattern papelcito “little piece of paper”, it appears that many more speakers opt for -itV papelito.
4
7) Disyllabic words ending in the diphthong [jo] or [ja] select the -ecit{o, a}
diminutive in North Central Peninsular, Sonoran, and Paraguayan dialects, but the
-it{o, a} diminutive in Nicaraguan:
-[jo/ja]: -ecit{o, a} novio noviecito “boyfriend/groom”
Nicaraguan: -it{o, a} radio radito “radio”
8) Disyllabic words with an alternating diphthong (/o/ [we]; /e/ [je]) in the
initial syllable select -it{o, a} in Sonoran and Nicaraguan, but select -ecit{o, a} in
Peninsular:2
ue/ie.: -it{o, a) piedra piedrita “stone”
puerta puertita “door”
Peninsular: -ecit{o, a} piedra piedrecita
puerta puertecita
The term “alternating diphthong” refers to those pairs of Spanish words
that historically share the same root (e.g. pedr- piedra “stone,” pedroso “rocky;”
ten- tener “to have,” tienes “you have”) but in the development of Modern
Spanish the stressed vowel /o/ and /e/ underwent diphthongization: /o/ [ue];
/e/ [ie]. Some authors appeal to this alternation as an explanation for the
2 Jaeggli (1978) does not directly address alternating diphthongs in his analysis for Paraguayan and none are found in his examples.
5
dialectal variation among some disyllabic words with diphthongs in tonic position
(piedrita ~ piedrecita) and the lack of variation in others that do not alternate
(viudito ~ *viudecito “widower”).
Curiously, trisyllabic words with alternating diphthongs do not appear to
select the -ecit{o,a} in any of the dialects:
abuelo (~abolengo “ancestry, lineage”) abuelito “grandfather”
huérfano (~orfanato “orphanage”) huerfanito “orphan”
escuela (~escolar “school (adj.)”) escuelita “school (n.)”
To summarize, diminutive formation is consistent across these four
dialects for all words except monosyllabics ending in a consonant (6), disyllabics
ending in a diphthong [jo]/[ja] (7), and disyllabics containing an alternating
diphthong (8). As explained in detail in Chapter 4, I propose that these dialectal
variations are the result of different repair strategies to severe phonotactic
constraints (7) or, as in the case of (6) and (8), depend on varying sensitivities to
prosodic constraints. These variations are easily captured in Optimality Theory
through re-ranking of the relevant constraints along the constraint hierarchy.
6
III. Theoretical Issues: Underlying Assumptions and the Phonology-
Morphology Interface
The first major issue that must be addressed concerns the affixal status of
the diminutive morpheme(s). Some authors treat -it- as an infix and –(e)citV as a
suffix that attaches directly to the end of the word (Jaeggli1978, Colina 2003).
Others analyze the process as the case of one underlying morpheme—/ƟitV/ or
/sitV/—that then undergoes a process of deletion of the initial segment in
response to phonological pressures (Prieto 1992, Crowhurst 1992). A third
option, and the one that I adopt, analyzes both diminutive forms as suffixes that
attach to a stem that lacks the final vowel present in the base. Note that for words
like clase that end in unstressed /e/, I agree with the majority of authors in
analyzing the /e/ in the corresponding diminutive form as epenthetic.
Crucially, I view Spanish diminutivization as an example of suppletive
allomorphy (Paster 2006a provides a detailed discussion), in which one allomorph
is not phonologically derived from the other. Thus, -it{o, a} is not the result of
deletion of the fricative segment in -cit{o, a}, nor is the latter created from the
former through segmental insertion. Furthermore, and in accordance with the
tenets of Optimality Theory, I view both suffixes as available underlying to all
words. In this way, possible output candidates, such as *casecita and *balconito,
are eliminated by faithfulness and markedness constraints that, with one
exception, are phonological in nature.
This final point leads into a second important issue: the degree to which
morphology interacts with the phonology in this process. With the exception of
7
Jaeggli’s (1978) analysis, all the major papers on Spanish diminutivization allow
at least a minor role for morphology. In the case of Prieto (1992), for example,
she adopts Harris’ (1991b) form classes to explain the difference in diminutive
selection between the phonologically similar bases in casa~casita and
clase~clasecita. Crowhurst (1992) posits a morphological constraint on Terminal
Elements /-o/ and /-a/ appearing in non-final position, thus explaining the lack of
an output such as *casacita; this constraint will appear in various forms in the
later OT accounts as well. Colina takes the strongest stance on a role for
morphology stating that “diminutivization is primarily driven by morphological
factors” (2003:45). In addition to the Terminal Element restriction, Colina argues
that attachment of the diminutive morphemes is constrained and determined by
the morphological structure and class of the base.
This fundamental role for morphology found in almost every analysis of
the topic is due to the uneven distribution of the allomorphs in the data of all
dialects in the literature. Viewed strictly from the syllable-count standpoint,
monosyllabic words select -cit{o, a} exclusively, disyllabics select either -it{o, a}
or -cit{o, a}, and trisyllabics or longer also select either of the allomorphs. An
analysis of the data based only on word endings is just as unrevealing: diminutive
selection in words ending in unstressed /o/, /a/ are not limited to either allomorph
(with the exception of the Nicaraguan dialect), nor is there consistent selection in
those words ending in unstressed /e/. Only those words ending in a stressed
vowel and (to a lesser degree) those ending in a consonant exhibit consistent
allomorph selection, in this case -cit{o, a}.
8
Most analyses have divided the words, either implicitly or explicitly, into
varying classes to account for why prosodically similar words, such as the
disyllabics casa and clase, select different forms. In the analysis I lay out in the
following chapters, I adopt a similar strategy and provide a role for morphology
by creating two word classes for diminutive formation—Class D1 and D2. Where
I differ from previous analyses is that I propose that my first class, which consists
of those words that end in the unstressed canonical gender markers /o/ and /a/, is
lexically specified to select the -it{o, a} allomorph. Class D2 simply contains
those words that are not bound by this lexical specification and allomorph
selection is determined by a hierarchy of faithfulness and markedness constraints.
In Chapter 2 I review the major analyses of Spanish diminutive formation
over the past thirty five years and, where relevant, highlight possible
shortcomings in these approaches that my analysis overcomes.
Chapter 3 is dedicated to a discussion and analysis of class marker
selection in diminutive formation. In Section II, I review Harris’ (1991a, b) five
word form classes based on word endings for Spanish and then, in Section III,
demonstrate their inadequacy with respect to Spanish diminutives. I argue for a
merger of his Class I and II words into one (Class D1) for the purposes of
diminutivization and combine the remainder of his three classes into Class D2. In
Section IV I defend my assumptions that all diminutive morphemes are suffixes
that attach to a stem that lacks the final vocalic element present in the base.
Additionally, I present a novel OT analysis for class marker selection in Spanish
diminutive formation based on these assumptions.
9
In Chapter 4 I present my analysis of diminutive formation for the
Sonoran dialect presented in Crowhurst (1992). Adopting Bonet’s (2006)
approach to Spanish gender allomorphy, in Section II I argue that Class D1 words
are lexically specified to select the -it{o, a} suffix and will only select -(e)cit{o, a}
as a repair strategy to avoid severe phonotactic constraints. This lexical
specification is marked with a subcategorization frame /-it-/ on all members of the
diminutive class D1. Selection of -(e)cit{o, a} results in a violation of a high-
ranking constraint, RESPECT, that governs against selection of non-lexically
specified allomorphs. In Section III, I analyze Class D2 words in Sonoran and
propose that both suffixes are readily available to Class D2 words and that
selection is determined by faithfulness and markedness constraints that are
generally phonological in nature. In Section IV, I provide additional data and
support for the RESPECT constraint and also demonstrate how my analysis handles
the variation between the Sonoran and North Central Peninsular (Colina 2003)
dialects through constraint re-ranking. I also appeal to constraint re-ranking in
Section V to account for the dialectal variation in Paraguayan and Nicaraguan
Spanish. Section VI presents further evidence for my constraint rankings and
presents Hasse diagrams for the four dialects that I address. Finally, in Section
VII I argue that a possible alternative analysis in which the diminutive
morphemes themselves are lexically specified for certain word types (rather than
the other way around) is too cumbersome and misses several important
phonologically conditioned generalizations.
Chapter 5 concludes with a review of the major points of my analysis.
10
Chapter 2
Previous Analyses of Spanish Diminutive Formation
I. Jaeggli (1978)
In the first unified account within generative grammar, Jaeggli (1978)
analyzes diminutive formation in the Paraguayan dialect as a distinction between
infixation of the -it- allomorph and suffixation of -citV.3 Using a rule-based,
serial framework, he first establishes a syllabic domain for infixation in which it
occurs only in words that end in an unstressed [+back] vowel:4
1) a. pal + o “stick” pal + it + o palito b. cas + a “house” cas + it + a casita For all other syllabic domains, the suffix -cit + V attaches to the base, where V
will always agree with the gender of the base:
2) a. canción (f.) “song” canción + citV cancioncita b. café (m.) “coffee” café + citV cafecito While this analysis is particularly efficacious for words that end in a single
consonant (2a) or a stressed vowel (2b), Jaeggli must introduce a pair of
allomorphic rules, e-Insertion and s-Deletion, to account for the large group of
words that end in an unstressed -e:
3 Orthographic <c> refers to /s/ in this and other Latin American dialects and generally to /θ/ in Peninsular dialects. 4 While this should also include the high, back vowel /u/ as in tribu “tribe,” Jaeggli does not address such examples.
11
e-Insertion & s-Deletion Rules (Jaeggli 1978: 146, 150) 3) a. madre Base madr + e “mother” Suffixation madr + cit + V e-Insertion madr + ecit + V Output madrecita
b. comadre Base comadr + e “godmother” Suffixation comadr + cit + V s-Deletion comadr + it + V Output comadrita
The purpose of these two rules is to avoid an illicit consonant cluster, in
this case *[drs], that would otherwise surface following suffixation.5 It is not
clear from this analysis, however, what motivates the selection of one rule over
another. Why should e-Insertion be triggered with some words and s-Deletion
with others? Jaeggli’s acknowledgement that s-Deletion will occur only in non-
monosyllabic words, while descriptively correct, lacks explanatory power.
Furthermore, if the stated purpose of these rules is to break up disallowed
consonant clusters, then it is unclear why e-Insertion, for example, should occur
in words such as nube~nubecita “cloud” or naipe~naipecito “playing card” in
which the clusters [bs] and [ps] in unattested forms *nubcita/naipcito pose no
difficulties for speakers (cf. abstracción “abstraction,” lapso “lapse”).
This overall lack of motivation for the rules proposed in this analysis
constitutes its major drawback. The fact that infixation occurs with words that
end in an unstressed back vowel perhaps adequately reflects the data, but does
5 Jaeggli does not explain at length the status of word-final /e/ in (3a-b) except to mention later in the article that “suffixation destroys the final vowel of the base, if the base has a final vowel.” (emphasis added) This description of the facts, however, is too broad in that it would also include words with stressed-final vowels (e.g. café~cafecito) that are never deleted.
12
little to expand our understanding of why that might be. The word sofá, for
example, ends in a stressed back vowel and therefore must undergo suffixation
(i.e. sofacito), yet the reasons restricting an unattested, but phonotactically well-
formed, *sofita are left unexplained. Finally, Jaeggli’s analysis would require
considerable reworking to account for dialects that select for the -citV allomorph
in words with diphthongs ue/ie in the penultimate syllable despite ending in an
unstressed, back vowel: puerta “door” ~ puertecita, fiesta “party” ~ fiestecita.
II. Prieto (1992)
Prieto’s (1992) analysis of diminutive formation in Spanish argues for a
unified morpheme /-sit/ that maps onto stems or roots. Crucial to her analysis is
the proposal that prosodic structural—and therefore phonological—information is
available to the speaker prior to the application of the morphological process of
diminutivization.
Prieto follows Harris (1991a, b) in her adoption of his five form classes
that divide Spanish nouns, adjectives, and adverbs in accordance with their
subcategorization for inflectional suffixes, as shown in (4):
(4) Form Classes in Spanish (Harris 1991b)6
I. pas-o “step” II. pas-a “raisin” guap-o “handsome” guap-a “beautiful” dentr-o “inside” cerc-a “near, close”
6 See Chapter 3 for a detailed description and discussion of Harris’ and other accounts of Spanish class markers.
13
IIIA. jef-e “boss” IIIB. as “ace” verd-e “green” común “common” delant-e “in front” atrás “back, behind” IIIC. pas-e “pass”
inmun-e “immune” adred-e “on purpose”
IV. dos-is “dose” V. trib-u “tribe” vival-es “bon viveur” esnob “snob” lej-os “far”
Prieto also adopts Harris’ (1991b) Marker Spellout Rule in (5). As Prieto remarks
(1992:189), an advantage of this rule is that it stipulates that inflectional markers
[-o, -a] are lexically controlled rather than dependent on the gender of the root.
This obviates the need to exceptionally mark words such as problema which are
masculine but have a terminal element /-a/ generally associated with feminine
nouns.
(5) Marker Spellout Rule (Harris 1991b) Stem Extension: Ø V / [ [ …]d____]i d = uninflected word boundary i = inflected word boundary
a / [Class II]
V / ]d______ o / Elsewhere [i.e. Class I] Within Class III are those words with an epenthetic final /-e/ (IIIA), bare roots
(IIIB) and words with a non-epenthetic /-e/ (IIIC). The difference between IIIA
and IIIC lies in the fact that the words in IIIC need not have undergone word-final
14
epenthesis because they would end in one of the permissible word-final
consonants in Spanish / d s θ n l r /: pas-e “pass”, prol-e “progeny.” Therefore
he claims “the lexical entries of these words specify by brute force that they have
an unspecified vowel (manifested phonetically as the default vowel /e/) in stem-
final position” (Harris 1991b: 73). Class IV refers to those words that end in word
markers /-as/, /-es/, /-is/, /-os/ and /-us/. Class V words end in a consonant or /-i,
-u/.
Prieto argues that Diminutive Formation triggers stem extension in Class I
and II, which in her analysis includes the addition of a moraic unit. Class III
words, however, are lexically excepted from stem extension and, therefore,
receive no moraic unit upon Diminutive Formation. (The status of Class IV and
Classa V words in this process is unclear in Prieto’s analysis since she does not
directly address them in terms of diminutivization). The addition of the moraic
unit accounts for the difference between a Class III word like pintor/pintorcito
and a Class II, as in casa/casita since the morpheme /-sit/ will undergo stray
erasure of /s/ if the next available unit is vocalic. This distinction is illustrated in
(6a-b).7
7 I use the IPA throughout this paper and have adapted the symbols of the original authors to the IPA whenever necessary.
15
(6) a. casa ~ casita [kas-] [Class II] [ k a s ] [-sitV] µ µ Moraic unit through stem extension σ σ
Removed via ‘stray erasure’ [ k a s ] [ s i t V] µ µ σ σ [ k a s i t a ] µ µ µ σ σ σ (6) b. pintor ~ pintorcito
16
[ p i n t o r] [s i t V] µ µ σ σ “stray erasure” does not occur since no moraic unit has been added (i.e. No stem extension) [ p i n t o r] [s i t V] µ µ µ µ σ σ σ σ Prieto analyzes those words with final word marker /-e/ by stipulating a variation
of the Spellout Rule called the Default Vowel Rule in which /-e/ is added to the
root before diminutivization. This ordering in (7) would apply to words in Class
IIIA and IIIC.
17
(7) clase ~ clasecita Application of Default Vowel Rule [ k l a s V ] V = Default Vowel /e/ µ µ σ σ [ k l a s e ] [ s i t V] µ µ µ µ σ σ σ σ
To account for dialectal variation between words like pan
(pancito~panecito), Prieto (1992: 195) invokes a minimal word constraint (8) that
establishes a word template of two trochaic feet.
(8) Minimal Word Template Wd Foot Foot σ σ σ σ pa ne si to
18
The panecito dialect of Central Peninsular Spanish, then, requires satisfaction of
this constraint. Presumably speakers of pancito dialects (Bolivian, Paraguayan,
etc.) would not be sensitive to this constraint. However, Prieto does not specify
that certain word classes must adhere to this constraint while others do not. Given
the fact that she uses this constraint to partially explain forms such as
sabio~sabiecito “sage” (a Class I word) and puerta~puertecita “door” (a Class II
word), it is safe to assume that the minimal word constraint must be operational
across all word classes under her analysis.
Following this line of reasoning, in the relevant dialects the minimal word
constraint should also apply to other Class I and Class II words that never trigger
an epenthetic /e/ despite violating the template of two bisyllabic feet, such as libro
“book” or mesa “table.”
(9) libro li.br.ito *li.bre.si.to
mesa me.si.ta. *me.se.si.ta
It could be argued that Prieto does not intend the epenthetic /e/ found in puertecita
and sabiecita types to be considered a result of the same process of template
satisfaction as found with panecito, but rather that for independent reasons—
preservation of the diphthong in puertecita and avoidance of an illicit [ji] cluster
in the case of sabiecito—the epenthetic /e/ emerges during diminutivization and
has the added benefit, in some cases, of creating a new word that satisfies the
19
minimal word template. Nevertheless, it would still be necessary to explain why
a prosodic constraint should apply to one class of words, but not to others.
Aside from the issue of selective application of constraints, Prieto’s
analysis is not able to account convincingly for all of the data. As mentioned
above, she argues that epenthesis is triggered in words with alternating diphthongs
(e.g. piedra/pedroso “stone” / “rocky”, nuevo/novedad “new” / “newness,
novelty”), as a means of preserving that diphthong. For example, in some dialects
puerta is diminutivized as puertecita. However, she offers no explanation as to
why the epenthetic /e/ is instrumental in preserving the diphthong nor is there any
indication of why in dialects that favor puertita the diphthong remains despite the
lack of epenthesis. Clearly there is a generalization to be uncovered with respect
to the behavior of cases of alternating diphthongs; it is just not captured in
Prieto’s analysis.
III. Crowhurst (1992)
Crowhurst (1992) limits her analysis to the Sonoran dialect of Northern
Mexico. She posits a minimality constraint that triggers /-e/ epenthesis on the end
of a stem that does not contain a disyllabic foot prior to a process of suffixation of
the diminutive morpheme -(c)itV. This approach functions particularly well with
monosyllabic words such as pan.
20
(10) [ p a n ] + sito [ p a n e ] + sito [panesito]
F[σ σ] F[σ σ]
In her view a template of a disyllabic foot is mapped onto the stem. Lacking two
syllables to satisfy the template, the epenthetic /e/ is added, thus fulfilling the
template that will allow for proper suffixation.
With disyllabic words that end in /o/ or /a/ (Terminal Elements in
Crowhurst’s terminology), the disyllabic template is satisfied, but she must still
account for why the suffixed form is not *librosito.
(11) [l i b r o ] + sito *librosito
F[σ σ]
She argues that after satisfaction of the disyllabic template and before suffixation
the Terminal Element must be deleted. This move is motivated by a
morphological constraint that a Terminal Element must surface in final position.
(12) Following TE Deletion
[ l i b r ] + sito
21
Much as in the Prieto analysis in which she claims Stray Erasure of
syllabically unattached segments, Crowhurst must also account for the fact that
the /s/ does not surface in the suffixed form. She appeals to a second template
apart from the process of diminutive formation that demands that an emptied
moraic node must reattach to the nearest unattached segment capable of bearing a
mora (13).
(13) [l i b r ] + sito [l i b r ] + s i t o
µ µ µ µ µ
σ σ σ σ σ
This leaves Crowhurst with essentially the same issue as Prieto, and she deals
with it in a similar way. The reattachment occurs at the segment /i/ of the
diminutive suffix and she argues that this leaves the sibilant ‘trapped,’ which is
then erased from the surface structure. Crucial for this analysis, monosyllabic
words like pan lack a terminal element and therefore this second template
satisfaction is not relevant. Longer words lacking a Terminal Element, such as
those ending in a consonant (escritor) or a stressed vowel (café), not only satisfy
the disyllabic template and therefore lack /-e/ epenthesis as in panecito, but also
do not require reattachment of a mora since there is no TE to be deleted. In these
cases, simple attachment of the -citV morpheme applies, as in (14).
22
(14) [k a f e] + s i t o
µ µ µ µ
F[σ σ] σ σ
To account for diminutive forms such as burdelito and papelito that also
lack a Terminal Element but in which the /s/ fails to surface after suffixation,
Crowhurst appeals to the Obligatory Contour Principle (OCP) and states that
adjacent segments with the feature [+continuant] violate the OCP in Mexican
Spanish with the stipulation that the restriction is in effect only across morpheme
boundaries. The rightmost [+continuant] segment (i.e. /s/), then, is deleted.
(15) [ p a p e l ] s i t o [papel] ito
[+cont] [+cont]
To capture the contrast between the diminutive forms of disyllabics ending
in /-e/ (madrecita) and trisyllabics and longer ending in /-e/ (comadrita,
uniformito) which surface without the /-es-/, her analysis assumes that the final /e/
in both cases is epenthetic and therefore not present in the stem at the time of
satisfaction of the disyllabic foot template. The epenthetic /e/, it is claimed,
23
surfaces as a repair for consonants or consonant clusters that Spanish prohibits in
coda position.
In this way, the stems for madre and comadre at the moment of disyllabic
template mapping are as follows in (16).
(16) [m a d r ] [k o m a d r ]
µ µ µ
σ σ σ
While at this stage they both contain unsyllabifiable consonant clusters in the
coda position, the crucial distinction is that madr- does not satisfy the disyllabic
requirement, thus triggering /e/ epenthesis prior to suffixation as in pan/panecito
above. On the other hand, komadr- satisfies the template and suffixation will
apply in the same manner as libro/librito with erasure of the trapped /s/.
(17) [m a d r e] + sita [ k o m a d r ] s i t a
µ µ µ µ µ µ σ σ σ σ σ σ
With certain words that contain the diphthong [je] in the first syllable
(such as diente), Crowhurst notes an alternation in their diminutive form: both
24
[djentesito] and [djentito] are acceptable variations. She accounts for the two
forms by assuming the two different syllable structures for diente in (18).
(18) a. [djèn.te] b. [dì.en.te]
The first is underlyingly parsed as monosyllabic [dyent-] and therefore unable to
satisfy the minimality constraint, triggering the epenthetic /e/ and /sito/ as in
panecito. In the second, the surface glide is resyllabified as the nucleus of its own
syllable [di.en.ti.to] and in this way satisfies the disyllabic template and
diminutivizes as libro/librito above. However, her informants vary in their
acceptability of these alternating forms, which, she suggests, could indicate that
this alternation and the subsequent resyllabification in (18b) is no longer
productive, but rather residue of a permissible process at an earlier point in the
language. As she claims, this accounts for why it occurs with some words but not
others, (cf. piel “skin” [pyelesita], *[pi.elita]).
In a response to Crowhurst’s analysis, Harris (1994) suggests that
diminutive formation may owe more to morphological and lexical factors than
Crowhurst’s phonological approach acknowledges. One of Harris’ criticisms
concerns Crowhurst’s analysis of words that end in the -io/-ia diphthong:
ebrio/ebriecito, novia/noviecita. Since these words vacuously satisfy the
disyllabic template and therefore predict the suffixation of –itV, Crowhurst
appeals to a ‘well-formedness condition’ proposed by Harris in (19) that prevents
two tautosyllabic homorganic vocoids from occurring in the same syllable, thus
25
precluding the ill-formed [no.βji.ta]. Harris asserts that she has misunderstood his
constraint in that this restriction is only applicable in rhyme position (cf. Harris
1983: 17, 34).
(19) *-ij. *-uw. *-ji. *-wu.
He notes that diminutives such as arro.yi.to and ra.yi.ta (*ra.ye.ci.ta) do
not pose a problem despite the presence of tautosyllabic [+high, +front] vocoids.
However, neither Crowhurst or Harris draws attention to the generalization that
the final diphthong -io/-ia words that are addressed in Crowhurst are all preceded
by one or two consonants (cf. el pa.tjo ~ pa.tje.cito, el gar.fjo ~ gar.fje.cito)
whereas the two examples in the Harris article are not preceded by tautosyllabic
consonants. The relevant question, then, may not be whether homorganic vocoids
are permissible tautosyllabically (clearly they are), but rather if [.Cji] is an
allowable syllable. Indeed, Harris (1983) (also see Hualde 1992, Kaisse 1999)
analyzes pre-vocalic glides as part of the nucleus in syllables that contain another
segment available for onset position. Thus, the glide in ra.yi.ta is in onset
position, but part of the nucleus in e.brje.ci.to. One needs only to alter slightly
Crowhurst’s definition to state that tautosyllabic homorganic vocoids in rhyme
position are prohibited. This would account more precisely for the relevant data
presented in Crowhurst without impacting her argument for the ungrammaticality
of *[no.βji.ta].
26
Of related interest, Prieto (1992) notes variation on these types in
diminutive formation, such as radiecita and radita from radio. Miranda (1999)
reports that in Nicaraguan radita is the norm. The lack of the form *radyita offers
prima facie evidence in favor of such a constraint.8
Problematic, perhaps, are Crowhurst’s proposed repairs of this restriction
against homorganic vocoids. For disyllabic words that end in the diphthong [jo]
or [ja] (novio “boyfriend”, tapia “wall”), she proposes a repair in which the mora
that attaches to the high vowel [i] in the diminutive following Terminal Element
Deletion is subsequently delinked as a response to the aforementioned well-
formedness constraint. The empty mora then is filled by an epenthetic /e/ and
thus the entire diminutive allomorph can attach without stray deletion of the [s],
as shown with the disyllabic tapia in (20).
(20) Double-/i/ Repair I (Crowhurst 1992: 246)
F F
σ σ σ σ σ σ σ σ µ µ µ µ µ µ µ µ [t a p j ] s i t a [t a p j e] s i t a
This implies some ordering difficulties for her analysis that she does not
address. First, in the diminutive formation of a word such as chamaka “girl”, she
must order stray erasure of the unattached [s] following Terminal Element
8 Although Harris (1994) does cite an informant who accepts labiíto, apiíto and ampliíto, it is not clear from his transcriptions if these are tautosyllabic. This issue will be discussed in more detail in Chapter 3.
27
Deletion, but before syllabification of the suffix. Otherwise, the analysis would
predict *[cha.mak.si.ta]. With a word like tapia, the attachment of the mora to the
high vowel in the diminutive would in turn trigger stray erasure of the [s], as it
does in the chamaka example. Thus the [s] would be unable to surface and we
would be left with the unattested *tapieíta [ta.pje.i.ta].
Since longer than disyllabic words ending in these diphthongs generally
pattern differently than disyllabic words (cf. iglesia~iglesita “church”), Crowhurst
must introduce a second repair strategy to account for the distinct output. In this
case the mora remains attached to the high vocoid in the diminutive and the glide
is delinked from the syllable structure, leaving both the glide and the [s] to be
stray erased.
(21) Double-/i/ Repair II (1992: 247)
F F
σ σ σ σ σ σ σ µ µ µ µ µ µ µ [i g l e s j] s i t a [i g l e s] i t a
Again the question of ordering emerges since both repair strategies must
be taken into account during diminutive formation. If the repair in (20) is ordered
first, then there would be no need for the second repair and all words ending in
[jo] or [ja] would trigger epenthesis, predicting *iglesiecita. If the second repair
is ordered first, then the reverse would occur, with *tapita being the surface form.
28
One remedy would be to propose that the two repairs are simultaneously in play.
However, this would go against the very nature of a serial analysis such as this
one and would still require the introduction of some separate mechanism through
which the second repair strategy (21) applied only to longer bases and ignored
shorter ones and vice versa for the first repair (20).
Beyond the issue I have raised here, Harris (1994) also finds problems
with Crowhurst’s distinction between forms like madre/madrecita and
tigre/tigrito (*tigrecito) and her suggestion that the final /e/ in tigre qualifies as a
Terminal Element and is present in the stem when the disyllabic template is
satisfied, subsequently the /e/ in tigre would undergo Terminal Element Deletion.
In madre, however, Crowhurst analyzes the /e/ as epenthetic and therefore not
present in the stem at the moment of template satisfaction. Harris reiterates his
claim in (1991a) that no such distinction can be made and that all word-final /e/
segments are Terminal Elements. It is not clear, then, how the two above forms
should differ, except that tigrito does not fit into Crowhurst’s analysis.
Finally, Crowhurst’s invocation of the OCP to explain the
ungrammaticality of *arbolcito due to a constraint on adjacent [+continuant]
segments is countered by examples presented by Harris (1994): des + liar; compul
+ sión. In both, neither the heteromorphemic cluster /ls/ or the inverse /sl/ causes
any difficulty for speakers. Harris raises further doubts by noting that the status
of /l/ as [+continuant] is highly controversial.
Without fully developing an analysis of his own, Harris claims that
selection of -itV or -(e)citV cannot be accounted for phonologically, but rather by
29
morphology and that lexical specification of the roots results in an arbitrariness
that confounds attempts at systematic explication.9
One example that Harris provides to support his preferential treatment of
morphology over phonology comes in the form of an alternation between
llorona/lloroncita and corona/coronita. From a phonological standpoint there is
little to suggest that they should behave differently. However, in Harris’ account,
lloroncita is actually derived from llorón, which belongs to his Class III stems
that have a bare root and would therefore select the -citV morpheme. On the other
hand, corona belongs to Class II that includes all words with the Terminal
Element -a. In his view a strictly phonological analysis fails to account for this
distinction.
IV. Ambadiang (1997)
Ambadiang (1997) rejects the claims of previous analyses with respect to
the influence of prosody as well as the adoption of Harris’ declensional classes in
diminutive formation. One of Ambadiang’s critiques of these declensional
classes in the discussion of diminutives is the lack of a one-to-one correspondence
between his five classes and the two diminutive allomorphs that he assumes.
According to Harris, his form classes are based on shared word markers and not
by gender as is sometimes suggested. Thus, masculine mapa “map” (notable for
its canonically feminine word marker –a) is joined to Class II with words such as
hija “daughter” that exhibit no such incongruity between word marker and
9 As an example of such arbitrariness he notes some preference of fres[k]ecito by some speakers when fres[k]ito would be the expected form.
30
gender. However, Ambadiang argues that Harris’ discussion of the structural
differences between corona and llorona actually undermines his motivation for
his five classes based on word markers. He questions the need to establish an
exception for llorona-type words that should, by Harris’ own reasoning, be
included with mapa and hija in Class II based on their shared word marker -a.
According to Ambadiang, one could argue that chico/chica should belong to the
same class in spite of their different word markers, since they both share the same
diminutive allomorph (in their case -itV) in the same manner as
lloroncito/lloroncita.
Ambadiang proposes three morphological classes to correspond to three
diminutive allomorphs: 1) minimal roots take -ecitV; 2) radical morphemes take
-citV; 3) all others take -itV.
Minimal roots refer to those words that do not fulfill the following
‘canonical’ morphological structure of Spanish nouns and adjectives.
[ [ [ _________ ] root ] gen. ] num.
In accordance with this structure the last two segments of a fully inflected noun or
adjective would refer to gender and number respectively. While chicos/chicas
would satisfy this regular structure, soles “suns” (from sol) would not since -e-
makes no reference to gender.
Radical morphemes refer to a class of words that may be interpreted as
satisfying the above morphological structure. The entire reasoning behind the
31
manner in which a word such as rincón “corner” is said to fulfill this structure
will not be recounted here, but it will be mentioned that Ambadiang views some
words that end in accented vowels, /n/ or /r/ as having ‘saturated’ the above
structure while not making reference to gender.
V. Elordieta and Carreira (1996)
Elordieta and Carreira (1996) offer the first account of Spanish Diminutive
Formation within the framework of Optimality Theory (OT). In their analysis
they argue that –cit is the base form of the diminutive morpheme and that –it only
surfaces (with the first segment deleted) in order to satisfy certain prosodic and
faithfulness constraints. In particular, they appeal to a faithfulness constraint on
moraic structure, IDENT-IO (Stµ), to account for much of the distribution.
(22) IDENT-IO (Stµ) (Elordieta & Carreira 1996: 54)
Let S1 be a stem in the input representation, and S2 its correspondent in the output
representation. The association of S1 and S2 to moraic structure must be the same.
This constraint interacts with constraints on epenthesis (DEP-IO) and deletion
(MAX-IO) to predict the correct output in consonant-final words of two syllables
or more. In (23) candidate (a) is optimal because it maintains the final consonant
of the base [n] in coda position, whereas candidate (b) with the suffix –it would
32
result in [n] being in onset position and, consequently, a loss of its moraic status
and a fatal violation of IDENT-IO (STµ).
(23) balcón ~ balconcito “balcony”
balkón -cit
µ
IDENT-
IO(STµ)
DEP-IO MAX-IO
a. bal.kon.ci.to µ
b. bal.ko.ni.to *! *
c. bal.ko.n[e].ci.to *! *
Tableau (24) illustrates their analysis for words that end in class markers –o or –a,
such as cana “gray hair.” They introduce an alignment constraint, ALIGN-GM,
which states that the left edge of a gender marker must align to the right edge of
the derived word. This constraint eliminates candidate (d) and the possibility of
the –cit morpheme attaching directly to the base. Candidate (b) is eliminated
because attachment of the –cit morpheme adds a mora, violating IDENT-IO (Stµ).
There are two possible repair strategies to maintain the [n] in non-moraic onset
position: 1) add an epenthetic /e/ (candidate c) or 2) delete the first segment in the
diminutive morpheme. The crucial ranking of DEP-IO over MAX-IO reflects the
preference for the second option.
33
(24) cana ~ canita “gray hair” cana, -cit ALIGN-GM IDENT-IO
(STµ) DEP-IO MAX-IO
a. ka.ni.ta *
b. kan.si.ta µ
*!
c. ka.n[e].si.ta *!
d. ka.na.si.ta *!
For words that end in –e (e.g. calle “street”), they introduce an additional
constraint, GM = Class I-II, which states that a gender marker as understood in
ALIGN-GM must only be of Harris’ form class I or II (i.e. –o or –a). The sole
purpose of this constraint seems to be to eliminate outputs such as candidate (c) in
Tableau (25) (1996: 57).
(25) calle ~ callecita “street”
{[kaʝ]-e, -sit} GM = CLASS I-II
ALIGN-GM DEP-IO MAX-IO
a. ka.ʝe.si.ta * b. ka.ʝi.ta * *!* c. ka.ʝi.te * *!
This tableau introduces some confusion into their analysis because it implies two
separate definitions for gender markers (GM). If GMs refer only to [o] and [a],
then it is unclear why candidates (a) and (b) should violate ALIGN-GM. Since
they do and candidate (c) does not, we must assume that they want the GM in this
constraint to refer to any class marker. This confusion of terms taken together
with the oddness of a constraint that defines rather than constrains suggests a
more efficient approach in which GM = CLASS I-II is eliminated and where
34
“GM” in ALIGN-GM is defined more narrowly as referring to only [o] or [a]. This
adjustment would not fundamentally change their analysis and candidate (a) still
wins (26).
(26) calle ~ callecita
{[kaʝ]-e, -sit} ALIGN-GM DEP-IO MAX-IO
a. ka.ʝe.si.ta
b. ka.ʝi.ta *!*
c. ka.ʝi.te *! *
Both of these approaches, however, propose a role for morphology in their
analysis and return us to the question of the existence of a fundamental difference
between words that end in –o or –a and those that end in –e. This difference in
Diminutive Formation is undermined somewhat by the behavior of words of more
than two syllables that end in –e, such as comadre, whose diminutive is
comadrita.
In Tableau (27), their analysis would incorrectly predict candidate (b),
comadrecita, due to the MAX-IO violations for candidates (a) and (c).
(27) comadre ~ comadrita
{[komadr-] –e, -sit} GM = CLASS I-II
ALIGN-GM DEP-IO MAX-IO
a. ko.ma.dri.ta * *!*
b. ko.ma.dre.si.ta *
c. ko.ma.dre.i.ta * *!
Furthermore, for those words that end in diphthong –io/a, such as radio,
Elordieta and Carreira again appeal to IDENT-IO(STµ) to account for the
35
diminutive form radiecito. As Tableau (28) illustrates, they argue that the glide in
radio is non-moraic and only candidate (a) preserves the original moraic structure,
at the expense of a DEP-IO violation due to [e] epenthesis.
(28) radio ~ radiecito “radio” (1996: 56) radio, -sit µ σ
ALIGN-GM IDENT-IO (STµ)
DEP-IO MAX-IO
a. ra.dj[e].si.to µ σ
*
b. ra.di.í.to µ σ
*!
*
c. ra.di.sí.to µ σ
*!
While it is true that candidates (b) and (c) fundamentally change the status of the
glide to the head of its own syllable, epenthesis in candidate (a) is not the only
possible solution. Their analysis does not consider a candidate such as radito, in
which the glide and the first element of the diminutive is simply deleted. As
Tableau (29) indicates, this candidate (d) would in fact be optimal based on their
constraints and rankings. Since DEP-IO is ranked above MAX-IO without any
other intervening constraints, their rankings predict that, all things being equal,
epenthesis is always a less-favored repair strategy than deletion.
36
(29) radio ~ radiecito “radio” radio, -sit µ σ
ALIGN-GM IDENT-IO (STµ)
DEP-IO MAX-IO
a. ra.dj[e].si.to µ σ
*!
b. ra.di.í.to µ σ
*!
*
c. ra.di.si.to µ σ
*!
d. ra.di.to10 **
Evidence that deletion is a viable repair strategy in diminutive formation is
found in longer than disyllabic words that end in the diphthong –io/a, such as
dinosaurio “dinosaur.” These types are not addressed in their analysis, but
Tableau (30) shows that their constraint ranking does predict the attested
candidate (d) dinosaurito. However, as with the problems in their analysis for
disyllabic calle vs. longer comadre types, there is no mechanism for adequately
differentiating between the behavior of radio~radiecito and
dinosaurio~dinosaurito types—correctly predicting dinosaurito, but incorrectly
*radito.
On the surface the generalization seems to relate to the number of
syllables. In both cases the disyllabic bases maintain the –cit morpheme
10 I assume the high vowel here to be part of the diminutive morpheme. It is possible to consider the vowel to be part of the base, but IDENT-IO (STµ) would eliminate that candidate in the same manner as candidates (b) and (c), still leaving (d) as the optimal candidate.
37
(following the assumption of Elordieta & Carreira that there is just one
morpheme), while undergoing deletion of two segments in the longer bases.
(30) dinosaurio ~ dinosaurito “dinosaur” dinosawrjo, -sit µ σ
ALIGN-GM
IDENT-IO (STµ)
DEP-IO MAX-IO
a. di.no.saw.rj[e].si.to µ σ
*!
b. di.no.saw.ri.í.to µ σ
*!
*
c. di.no.saw.ri.si.to µ σ
*!
d. di.no.saw.ri.to **
While Elordieta & Carreira do not successfully account for the different
outcomes for these cases, they do integrate a constraint, DISYLLST/B, which refers
to the number of syllables in a word.
(31) DISYLLST/B: The base of suffixation must be disyllabic, or the stem must
be underlyingly disyllabic.11 (1996: 59)
11 The stated reason for this distinction between the base and the stem is to account for disyllabic words such as cana, in which the base would be monosyllabic can and therefore would trigger epenthesis under their analysis *canecita. Since this does not occur, they assume that the stem ca.na must be disyllabic for these words. It is not clear why the constraint does not simply state that the stem must be disyllabic. It would not seem to impact their basic analysis.
38
This constraint accounts for epenthetic [e] in monosyllabic types such as
pan~panecito.12
(32) pan ~ panecito “bread” (1996:58)
pan, -sit µ
DISYLL IDENT-IO(STµ)
DEP-IO MAX-IO
a. pa.n[e].-si.to
* *
b. pan.-si.to µ
*!
c. pa.n-i.to *! *
Candidates (b) and (c) both violate DISYLL since their pre-suffixal stems are
monosyllabic. Candidate (a) satisfies this constraint through epenthesis, but
comes at the price of a violation of IDENT-IO(STµ) because the nasal [n] that had a
mora in the base is now in non-moraic onset position of the following syllable.
VI. Miranda (1999)
In an OT analysis of Nicaraguan, Miranda (1999) proposes two
allomorphs -ito/a and -cito/a and argues that the process is an output to output
correspondence relation in which “the standard of comparison...is not the
underlying representation, but an output form which has been footed, has received
stress, and has been assigned gender” (1999: 131). Selection of the diminutive is
determined through an evaluation of the base. Miranda identifies three main
12 As noted previously, some dialects exhibit a preference for pancito. Unfortunately, Elordieta & Carreira do not indicate to which dialect(s) their analysis refers. It is not clear how their analysis would account for dialectal variation such as the panecito/pancito distinction.
39
elements that are evaluated: (a) the size of the base; (b) the syllabic structure of
the last syllable; and (c) the characteristics of final vowels (134).
She analyzes the basic patterns of diminutivization through the following
ranking and constraints in (33).
(33) STEMMIN, MAX-HEAD, IDENT-OO (STµ) >> ALIGN-DIM
STEMMIN: STEMMIN = WORDMIN: The stem must at minimum be a prosodic
word, that is, a bimoraic foot (1999: 135).
MAX-HEAD: The prosodic head of the base must be present in the output
(1999:143).
IDENT-OO (STµ): Outputs must be faithful to the syllabic structure of the base
(1999: 137; cf. Elordieta & Carreira 1996: 54).
ALIGN-DIM: The left edge of the diminutive must align to the right edge of the
stem (1999: 132).
In the case of the constraint STEMMIN, it is worth examining in detail the
definitions for stem and base put forth by Miranda in order to better understand
the following discussion of her analysis. In a footnote, Miranda states that the
“[b]ase refers to the unaffixed surfaces forms. Stem is the part of the base
prosodically determined to which diminutive allomorphs attach” (1999: 101).
Therefore, by this definition the base of casa, for example, is ca.sa and the base
of clase is cla.se. It is not clear, however, what the stems of these two words
should be. Based on their respective diminutive forms, casita and clasecita, it
40
stands to reason that the stems are cas- in the case of casa and clase- in the case
of clase. However, if this were the case then none of the disyllabics that end in –o
or –a would satisfy STEMMIN and we might expect epenthesis to form a stem
such as *case- and a diminutive form of *casecita.
Every analysis must deal in some way with the fundamental difference
between casa types and clase types that share the same prosodic and phonological
structure, but exhibit distinct diminutive forms: casita and clasecita. For her part,
Miranda argues that for those words that end in unstressed –o or –a, the final
vowels are part of the stem when diminutives attach (1999: 103). Her evidence is
that non-canonical masculine words that end in an –a instead of an –o, such as
poema (m.) “poem,” or vice versa for feminine words ending in an –o, foto (f.)
“photo” tend to maintain the marked vowel in their diminutive forms:
poema~poemita, foto~fotito. While interesting, this does not in itself seem to
constitute sufficient evidence to postulate two separate processes. The fact that
Miranda offers no Tableaux for words that end in unstressed –o or –a confounds
the matter further, since we are left to guess how the various candidates for a
word like casa would fare against her constraints and rankings.
Apart from the incomplete nature of Miranda’s analysis, there is also
suspect treatment of disyllabics ending in –e and their lengthier counterparts, such
as comadre (Adopted from Miranda 1999: 136). Miranda illustrates two possible
analyses of disyllabics that end in –e. In (34), she assumes that the base-final [e]
is retained in the stem in candidate (a) but is not in candidate (b).
41
This tableau raises two questions about the violations for candidate (b). The first
concerns the evaluation of the stem. Since the clasita candidate violates
STEMMIN, we must assume that the evaluation of the stem occurs only after
affixation. Otherwise, the stem [klas-] would satisfy the bimoraic requirement to
fulfill STEMMIN. If evaluation occurs after affixation (which, incidentally, strips
the constraint of any meaning), then this invalidates the argument that the
unstressed –o or –a is present in the stem of words like casa. As Tableau (35)
illustrates, this would wrongly predict *casecita. For Miranda’s analysis to work,
we must assume that stem evaluation of casa types occurs before affixation, but
after affixation for clase types. Her analysis provides no theoretical basis for such
a move.
(34) clase ~ clasecita “class” Base: [klá.se] STEMMIN ALIGN-DIM
a. [kla.se.-sí.ta]
b. [kla.sí.ta] *! *
(35) casa ~ casita “house”
Base: [ká.sa] STEMMIN ALIGN-DIM
a. [ka.se.-sí.ta] *
b. [ka.sí.ta] *!
The second issue in (34) is related to the ALIGN-DIM constraint that
dictates that the left edge of the diminutive attach to the right edge of the stem. In
disyllabics like clase, this constraint is not a factor and seems to be introduced to
42
deal with longer words, like comadre. According to Miranda, there are two ways
to analyze the stem of comadre. In Tableau (36), the stem is analyzed as
maintaining the base-final /e/. This approach, however, wrongly predicts
candidate (a), *comadrecita.
(36) comadre ~ comadrita
Base: [ko.ma.dre] STEMMIN ALIGN-DIM
a. [ko.ma.dre.-sí.ta]
b. [ko.ma.drí.ta] *!
Therefore, she argues that the base-final /e/ must not be retained in the stem, as in
(37).
(37) comadre ~ comadrita
Base: [ko.ma.dr-e] STEMMIN ALIGN-DIM
a. [ko.ma.dr-[e].-sí.ta] *!
b. [ko.ma.dr-í.ta]
This explanation seems ad hoc and serves more to validate her constraints and
their rankings rather than a consistent analysis of the data. In order for this
analysis to function, we must establish at least three distinct processes:
A) Disyllabics ending in unstressed –o or –a:
casa Base: ka.sa
Stem: ka.sa
1) STEMMIN evaluation of candidates prior to affixation
43
2) Attachment of diminutive –it and “relocation” of stem-
final [a]
Optimal output: casita
B) Disyllabics ending in unstressed –e:
clase Base: kla.se
Stem: kla.se
1) Attachment of diminutive
2) Stem evaluation of candidates following affixation
Optimal output: clasecita
C) Trisyllabics or longer ending in unstressed –e:
comadre Base: ko.ma.dre
Stem: ko.ma.dr-
1) Attachment of diminutive
2) Stem evaluation of candidates following affixation
Optimal output: comadrita
Process A differs from B and C in that STEMMIN must evaluate the length of the
candidates’ stem prior to affixation in A, but after affixation in B and C. This is
antithetical to the principles of the framework of Optimality Theory in which she
44
is working in that all possible candidates must be evaluated at the same time.13
Finally, B and C establish two separate processes for words ending in unstressed
–e, in which base-final /e/ is retained in the stem in disyllabics, but dropped from
the stem in longer words in C.
VII. Colina (2003)
In contrast to the analyses in Elordieta and Carreira (1996) and Miranda
(1999), Colina (2003), in her analysis of Northern/Central Peninsular Spanish
diminutive formation, assumes that –e, -a, and –o are all Terminal Elements (TE)
and therefore subject to the same constraints. She proposes that -it is an infix that
attaches to words that end in a TE and -citV attaches to those without a TE.
However, in a word such as clase, the latter is exceptionally selected to form
clasecita due to the emergence of an unmarked prosodic structure of two binary
feet (clà.se) (cí.ta) and the segmental identity of final -e with that of the
epenthetic vowel -e in Spanish. Although the aforementioned potential candidate
for casa, *casa-cita, would share this preferred prosodic structure, it is eliminated
by a higher-ranking constraint against TEs in non-final position. Under this
analysis, clasecita, avoids this constraint because the TE /e/ is re-analyzed as
epenthetic.
38) *TE-: Terminal Elements shall not surface in non-final position.
13 It is possible to handle this issue in the Lexical Phonology model of OT, such as Stratal OT (Bermúdez Otero 2007) in which constraints can be reranked at different strata, but Miranda offers no indication that she is working from such assumptions.
45
-citV TO PrWd: Align left edge of -citV to the right edge of the
prosodic word.
DIM TO PRWD: Align the left edge of the diminutive to the right
edge of the prosodic word.
The ranking of these constraints *TE-, -citV TO PRWD >> DIM TO PRWD reflects
Colina’s view that diminutives want to attach at the word level and that –citV
would actually be the preferred allomorph were it not for the fact that so many
words end in a TE, which must always emerge in word-final position. This
comes at the expense of the alignment constraint DIM TO PRWD, which favors
attachment of the suffix to the prosodic word, as in canción/cancioncita.
Tableaux (39) and (40) illustrate this basic constraint ranking.
(39) vaso ~ vasito “(drinking) glass” bas-o + DIM *TE- -citV TO PRWD DIM TO PRWD
a. bas-it-o *
b. bas-sito *! *
c. bas-o] sito *!
Candidate (c) in (39) is eliminated because TE [o] is in non-final position,
a violation of highest ranking *TE-. Candidate (b) satisfies this constraint, but
fails to emerge as the winner due to the stronger preference for –citV to attach to a
prosodic word. Therefore, candidate (a) is the optimal choice.
46
(40) rincón ~ rinconcito “corner” rinkón + DIM *TE- -citV TO PRWD DIM TO PRWD
a. rinkón]-sito
b. rinkón]-it-
Colina (2003: 58) notes that candidate (b) in Tableau (40) would not be
eligible because she claims that the right edge of the -it- allomorph must align
with the TE of the base, but since candidate (b) lacks a TE it cannot be the
optimal candidate. The constraint Colina proposes in order to eliminate (40b) is
–it- TO VTE: “The right edge of the –it- allomorph must be aligned to the left edge
of a vocalic [+syllabic] TE of the base.” (2003: 59)
There are a couple of issues with her application of this constraint. First,
this constraint is never integrated into any of her tableaux nor is it clear where it
fits into her constraint rankings. Second, this constraint precludes the addition of
a candidate in (40), such as *rinconito, in which there is a TE [o] aligned to the
right edge of –it-. The legitimacy of such a candidate is supported by the many
disyllabic and longer words that end in /l/ (e.g. árbol) and lack a TE in the base,
but that select for the –it- allomorph (arbolito).
Another potential problem for the basic analysis in (39) and (40) concerns
disyllabic words ending in TE [e], such as clase/clasecita. The diminutive form
would seem to leave the TE -e in non word-final position, a violation of a high-
ranking constraint under Colina’s analysis. She resolves this issue by proposing
47
that the -e segment in clasecita is no longer considered a TE and has been
reanalyzed as an epenthetic -e, thus averting any violation of *TE-. This move, in
turn, incurs a violation of a low-ranking constraint LINEARITY-BSF in which the
“precedence structure” between the base and suffixed form must align.14
Colina also introduces a series of prosodic constraints that she
incorporates into PRWD-MINWD, which is a cover term to represent the effects of
the following four constraints in (41) (Colina 2003: 59).
(41) PARSE SYLL: All syllables must be parsed.
FTBIN: All feet must be binary.
ALIGN-L (FOOT, PRWD): Every foot stands in initial position in the
prosodic word.
ALIGN-R (FOOT, PRWD): Every foot stands in final position in the
prosodic word.
One issue with the PRWD-MINWD constraint (again, actually a cover term
for the constraints in (41)) is that it necessitates an additional layer of evaluation
beyond the two extant levels traditionally proposed in Optimality Theory: input-
to-output evaluation and output-to-output evaluation. For example, in the latter
the output candidate casita would be evaluated against the output form of its pre-
suffixal form casa. A candidate such as casecita, in which the segmental identity
of the word-final vowel [a] is changed to [e], is a violation of output-to-output
faithfulness. Colina’s constraint IDENT-SEG. PRWD formalizes this restriction. 14 See Colina (2003: 61-63) for a full discussion of this issue.
48
PRWD-MINWD, however, does not evaluate an output-to-output
correspondence, but rather the stem to which the diminutive attaches.
Additionally, it only evaluates those stems that form a prosodic word. For
example, candidates (a) and (b) in Tableau (42) vacuously satisfy PRWD-MINWD
because kas- is not a prosodic word.
(42) casa ~ casita “house”
*TE- IDENT. SEG.
PRWD
PRWDMINW
D
-citV TO
PRWD
DIM TO
PRWD
DEP-BSF
LINEARITY BSF
a. kas-it-a * b. kas-sita *! * c. [kas-a] sita *! d. [kas_1,2e2,1] sita *! *
Essentially, Colina establishes a constraint set (PRWD-MINWD) that is
extremely limited in terms of the candidates to which it applies. Colina merely
stipulates that this constraint set does not apply to non-prosodic words, but does
not explain why that should be so. Recall that PRWD-MINWD is not a constraint
in and of itself, but a cover term for four constraints that make no specific
reference to prosodic words. In addition, one would presume that these prosodic
constraints must be ranked in some way, yet it is unclear from the analysis how
they are to be ranked. Finally, as Stephenson notes (2004: 35-36), if we accept
that the constraint applies to the pre-suffixal stem, then it still must also apply to
the entire prosodic word.15 In which case, an unattested form for casa, *(cà.se)
(cí.ta), is more prosodically sound than the winner casita, which would violate
both PARSESYLL and ALIGN-L (FOOT, PRWD). 15 This again would force a cyclical application of the constraint set: applying first at the stem level and then at the word level.
49
Even if we accept that the prosodic structure of “prosodic words” can be
evaluated for markedness (i.e. PRWD-MINWD) prior to suffixation, there are still
puzzling inconsistencies in Colina’s use of the term “prosodic word.” In her
evaluation of pan~panecito in Tableau (43), the right hand brackets indicate the
boundary of the prosodic words. Since neither of the candidates violates DIM to
PRWD, it is clear that she assumes both [pan] and [pan-e] to be prosodic words.
This raises the question of what constitutes a prosodic word in her analysis. If it
is a word that actually exists in the lexicon, then [pan-e] could not be considered.
In a later tableau for the word clase (2003: 63), clas- is not considered a prosodic
word, but non-existent [clas-a] is. Therefore, we must assume that, with respect
to the constraint set PRWD-MINWD, by “prosodic word” Colina means not a
word, but a stem that has a particular prosodic shape. This runs counter to
prevailing definitions of “prosodic word.”16
(43) pan ~ panecito “bread”
*TE- IDENT. SEG.
PRWD
PRWD MINWD
-citV to
PRWD
DIM to PRWD
DEP-BSF
LINEAR-BSF
a. pan –cito *! b. pan-e] –cito *
Colina’s ranking of the constraints *TE >> –citV TO PRWD >> DIM TO
PRWD illustrate her argument that diminutive allomorphs prefer to attach at the
word level. The ranking –citV TO PRWD >> DIM TO PRWD reflects her claim that
–citV would actually be the preferred allomorph, if not for the fact that so many
16 Hall (1999:2) notes in his introduction to Studies on the Phonological Word that “almost all of the contributors operate under the assumption that the phonological generalizations that refer to the pword are ‘concrete’ in the sense that they refer to the surface level of representation and not to an abstract stage in the derivation.”
50
words end in a TE, which in her analysis requires selection of the –it- allomorph,
due to an undominated alignment constraint, -it- TO VTE: The right edge of the –it-
allomorph must be aligned to the left edge of a vocalic [+syllabic] TE of the base.
One possible objection to these constraints and their rankings concerns the
redundancy in positing two constraints that are basically identical, –citV TO PRWD
being more specific and DIM TO PRWD more general. Therefore, a candidate such
as (42b) cas-cita incurs a violation of both constraints for the same issue of non-
attachment to a prosodic word. In addition, the –it- allomorph never attaches to a
prosodic word (understood in Colina’s terms), so that it always violates DIM TO
PRWD. The sole purpose of this ranking seems to be to eliminate candidate like
(42b) cas-cita over the winner (42a) casita.17
VIII. Stephenson (2004)
Stephenson (2004) moves the discussion somewhat back to phonology
from morphology in her OT account of diminutives. She assigns bases to one of
two class systems based on their phonological and morphological properties.
Interestingly, these classes are not divided by whether they take one allomorph or
another, but by their sensitivity to a constraint on minimality. In particular, the
distinction between the two systems is determined by whether a class of bases
distinguishes between long and short in terms of allomorph selection. That is, if a
stem prior to suffixation has a minimality requirement of two syllables, then it
would belong to System II. Table 1 shows her division of the relevant data. 17 The issue Colina seems to have uncovered here concerns how to formalize the fact that, at least with some word types, -it- is the preferred allomorph. The analysis outlined in Chapters 3 and 4 of this work offers another approach.
51
Base Assignment to System I or II (Stephenson 2004: 30)18
In some dialects only short bases (one syllable before suffixation) ending in
unstressed -o/-a are not subject to a minimality requirement. In Stephenson’s
analysis, those words ending in a TE (unstressed -o,-a,-e) will have it deleted prior
to suffixation.
(44) casa: kas- + DIM clase: klas + e (epenthetic) + DIM
18 Stephenson does not include Prieto (1992) in this table or in her bibliography.
Peninsular Colina (2003)
Paraguayan Jaeggli (1978)
Sonoran Crowhurst (1992)
Nicaraguan Miranda (1999)
short o-/a-final bases: /libr-o/
System I System I System I
System I
long o-/a-final bases with short allomorphs: /el-o/ ~ /iel-o/
System II ??? System I System I
long o-/a-final bases: ventana, rad/jo/
System II System II System II System I
short e-final bases: /klas-e/
System II System II System II System II
long e-final bases: /komadr-e/
System II System II System II System II
C-final bases: /pan/, /ademán/
System II System II System II System II
52
This approach accounts for the thorny issue of why the pre-suffixed stem klas-
would require an epenthetic -e, but not cas-, since both are monosyllabic prior to
suffixation, under an analysis in which all TEs are initially deleted.
In terms of her OT analysis that accompanies this dual-system approach,
she assumes the following constraint ranking with the constraint definitions
provided below.
(45) a. panecito dialects:
PHON, *HIATUS, *TE (NON-FIN), MIN, IDENT [± SYLLABIC] » IDENT SYL-POS »
DEP-V » MAX-V
(45) b. pancito dialects:
PHON, *HIATUS, *TE (NON-FIN), IDENT SYL-POS, IDENT [±SYLLABIC] » MIN »
DEP-V » MAX-V
PHON: Refers to basic high-ranking phonotactic constraints (that rule out
such malformed items as *madrcita).
*HIATUS: No two adjacent vowels in a heterosyllabic relationship.
*TE (non-fin): Terminal Elements cannot appear in non-final word
position.
MIN: A stem must contain a minimum of two syllables prior to suffixation.
IDENT [± SYLLABIC]: The feature [syllabic] in the diminutive form of a
segment must correspond to its status [±] in the base form.
53
IDENT SYL-POS: A segment’s syllabic position in the diminutive form
must correspond to its position in the base form.
DEP-V: A vowel in the output must correspond to a vowel in the input.
MAX-V: A vowel in the input must correspond to a vowel in the output.
The crucial ranking distinction between pancito dialects (Paraguayan (Jaeggli
1978) and Nicaraguan (Miranda 1999)) and panecito dialects is between MIN and
IDENT-SYLPOS. Under Stephenson’s analysis pancito dialects would rank them
IDENT-SYLPOS » MIN due to the lack of an epenthetic –e that if present would
alter the syllable position of /n/ from coda to onset position. In panecito dialects,
on the other hand, the minimality requirement MIN would outrank the conflicting
constraint of input-output correspondence of a segment’s syllabic position. As
Colina (2003) has noted, one of the advantages of an OT approach is its ability to
account for this type of dialectal variation through re-ranking of the relevant
constraints.
Tableau (46) offers an example of Stephenson’s OT proposal for Spanish
diminutives (panecito dialects).
(46) mesa ~ mesita “table”
/mesa/ + DIM PHON *HIATUS
*TE (non-fin)
MIN IDENT [± SYL]
IDENT [Syl-Pos]
DEP-V MAX-V
a. mes-ita (*) * b. mesa-ita *! * c. mes-sita */ok *! * d. mese-ita *! * e. mesa-sita *!
54
In Tableau (46) the apparent fatal violation of MIN for the mesita candidate would
appear to disqualify it. Stephenson settles this matter by placing mesa-type words
(disyllabics with TE –o/-a) in System I. She suggests two ways of formalizing
this system. One, as above, is to have the MIN constraint apply only in System II,
which would leave mesita as the optimal candidate. Another option is to stipulate
that only the –ito/a allomorph is available to System I, as in Tableau (47).
(47) mesa ~ mesita “table”
/mesa/ + DIM PHON *HIATUS
*TE (non-fin)
MIN IDENT [± SYL]
IDENT [SYL-POS]
DEP-V MAX-V
a. mes-ita (*) * b. mesa-ita *! * c. mese-ita *! * *
One problem with this analysis, as Stephenson freely admits, concerns the dual-
system approach and how it might have arisen historically. She remarks that
some of the dialectal variation (as seen in Table 1) might be attributed to Sonoran
and Nicaraguan operating at a different stage in the language, one in which
System I still holds more ground. She suggests that a search of the historical
record may reveal a time when System I covered more of the domain of
diminutive formation than it does today.
55
Chapter 3
Class Marker Selection in Diminutive Formation
I. Introduction
One of the major difficulties that all of the aforementioned analyses of
diminutivization must grapple with is the incongruent behavior between disyllabic
words that end in unstressed final /o/ or /a/ and that invariably select the –it{o, a}
allomorph (pato ~ patito “duck”) and those disyllabics that end in unstressed /e/
and select –cit{o, a} (madre ~ madrecita “mother”). The most obvious possible
solutions to this quandary must be found either in the phonology, the morphology,
or some combination of the two.
Jaeggli’s (1978) derivational account appeals to purely phonological
information to account for this distinction in that an infix -it- will attach only to
unstressed [+ back] vowels (i.e. /o/ or /a/) and –cito/a attaches to words ending in
other vocalic elements. As noted in Chapter 2, this approach leaves us with
several descriptive and explanatory gaps.
Some analyses that have included a role for morphology (Crowhurst 1992,
Prieto 1992) have argued for a fundamental difference in terms of form class for
those words that end in unstressed /o/ or /a/ and those that end in /e/. In this
chapter, we will explore some of the literature on Spanish non-verbal word types
in order to determine to what degree (if any) such divisions play a role in a unified
analysis of Spanish diminutive formation.
In Section II, I review the development of Harris’ Spanish word form
classes. Section III considers the possibility of using Harris’ form classes to
56
delineate the selection of one diminutive allomorph over another. I argue that the
data do not support such an approach and propose dividing Spanish non-verbal
words into two classes for the purposes of diminutivization. Section IV presents
the underlying principles that I will assume about the process of diminutivization
as I move forward to my OT analysis of Spanish diminutive formation in Chapter
4. I argue that Spanish diminutivization is a stem-level process and that all
diminutive allomorphs are suffixes. Furthermore all diminutive allomorphs are
underlyingly available for selection by any word type.
II. Review of Harris’ Word Classes
Many authors have pointed out in the literature (Harris 1991a, Klein 1989,
and Roca 1989 to name a few), that there is sufficient contradictory evidence to
refute the traditional notion of a relationship between a Spanish noun’s final
vocalic element and its grammatical gender. While the assumption that word-
final /a/ is associated with the feature [feminine] and word-final /o/ with
[masculine] will serve as a reasonable rule of thumb, this method will fail the
learner of Spanish on some of the most commonly used words in the language: el
problema, el programa, el clima, la mano, etc. Worse yet, those substantives that
end in a consonant or in vowels other than /o/ or /a/ are even less reliable markers
of grammatical gender.19
Not surprisingly, this fact has led linguists to propose alternative
classifications of Spanish nouns and other non-verbal elements in the language
19 See Teschner & Russell (1984) for a helpful analysis of the statistical relationship between word-final segments and gender in Spanish substantives.
57
(adjectives, adverbs, etc.). Harris (1985) first sketched out a proposal for
classifying these words based on his concept of a “word-marker” and his claim
that “[they] make no syntactic or semantic contribution to the words they are a
part of; their only grammatical properties are their phonological form and their
distribution, which is unlike that of any other class of morphemes (34).” The
common denominator of the words in (1) is the presence of a final word-marker
that takes the phonological shape of an unstressed vowel /V/ or /V/ + [s]. Harris
argues for the peripheral status of these segments (in both a semantic and
phonological sense) due in large part to their absence in derivational morphology.
(1) herman[o] “brother” cas[a] “house” part[e] “part”
dos[is] “dosage” lej[os] “far” (2) A B C
herman[o] herman-dad *herman[o]dad “brother” “brotherhood”
dos[is] dos-ificar *dos[is]ificar “dosage” “to measure out” lej[os] lej-anía *lej[os]anía
“far” “distance” The derived words in column (2b) lack the word markers that appear in (2a)
suggesting that the morphological structure of the words in (2a) is the following:20
20 Harris actually analyzes this process through an autosegmental representation (1985: 40-41). Omission of these details has no bearing on the current discussion.
58
(3) a. [[[herman]stem o]word marker ]word
b. [[[lej]stem os]word marker ]word
c. [[[dos]stem is]word marker ]word Under this analysis, derivations are formed through attachment of derivational
affixes to stems that crucially lack the word marker:21
(4) a. [[herman]stem dad]derived word b. [[lej]stem anía]derived word With the peripheral status of word markers established, in a later work Harris
(1991a: 33) then proposes a lexical hierarchy based on the regularity of the word
markers:
(5)
Regular Irregular
Inner Core Outer Core Residue [masc] [fem] [masc] [fem]
hij[o] hij[a] padre mujer problem[a] m. ‘son’ ‘daughter’ ‘father’ ‘woman’ ‘problem’
cedr[o] sidr[a] mar liebre lej[os] ‘cedar’ ‘cider’ ‘sea’ ‘hare’
trib[u] f. ‘tribe’
As shown in (5), he divides words into a canonical inner core in which masculine
and feminine words end in the expected word markers /o/ and /a/, respectively.
The outer core consists of those words that end in an epenthetic /e/ (e.g. padre, 21 Bermúdez-Otero (2007) advocates for an analysis in which the word marker (‘stem formative’ in his terminology) is present and then deleted by the phonology: [[[herman] o] dad]. I will address his analysis in more detail later in this chapter.
59
liebre) or those that lack a word marker (e.g. mujer, mar). Lowest on the
hierarchy are words that are lumped into a “residue,” which he classifies as a
series of subclasses noted for their irregularity (in the case of words such as
problema that end in /a/, but are masculine) and relative lack of productivity and
token frequency (such as those ending in word markers /u/ or /i/).
This rough classification system is refined into the five form classes in
Harris (1991b) already presented in Chapter 2, but reprinted in (6) for
convenience.
(6) I. pas-o “step” II. pas-a “raisin” guap-o “handsome” guap-a “beautiful” dentr-o “inside” cerc-a “near, close” IIIA. jef-e “boss” IIIA´. as “ace”
verd-e “green” común “common” delant-e “in front” atrás “back, behind (adv.)” IIIB. pas-e “pass”
inmun-e “immune” adred-e “on purpose”
IV. dos-is “dose” V. trib-u “tribe” vival-es “bon viveur” tax-i “taxi” lej-os “far” esnob “snob” In contrast with the earlier partition based on regularity and presented in (5),
Classes I and II now include all items that end in word marker /o/ or /a/,
respectively, regardless of the syntactic gender of the word. In Harris’ view, his
morphological form classes and syntactic gender are two separate, but related,
domains. Therefore, a word like casa belongs to Class II and is syntactically
60
[feminine] in the sense that it triggers feminine gender agreement in the relevant
modifiers: casa amarilla, “yellow house.” However a word such as mano ‘hand’,
which is syntactically [feminine] (i.e. mano amarilla, “yellow hand”), belongs to
morphological Class I due to its termination in /o/. Class IV includes those words
that end in word marker /Vs/; whereas Class V is reserved for those words, such
as xenonyms and those lexemes with non-canonical word markers /i/ and /u/, that
do not fit in the other four classes.
Harris groups together Classes I, II, and IV in the sense that the relevant
stems “are bound to the (semantically empty, syntactically nonfunctional,
phonologically noncyclic) suffixes /o/, /a/, and /Vs/, respectively (1991b: 81).”
Class III, however, includes those stems that require an epenthetic /e/ for
syllabification purposes (Class IIIA), lack a word marker (Class IIIA´), and those
that end in an [e] that is not readily attributable to syllabification issues (Class
IIIB).
According to Harris, the difference between words in Class IIIA and Class
IIIA´ is related to the acceptability of word-final codas in Spanish. Final coda
position is a more restrictive subset of allowable codas in general in Spanish.
Typically, complex codas are disallowed in word-internal position with the
exception of glide + consonant (veinte [beyn.te] “twenty”) or consonant + [s]
clusters (instante [ins.tan.te] “instant”); while in word-final position all complex
codas—with the exception of a limited number of [Cs#] types: biceps [bi.seps]
“biceps”—are forbidden and the allowable set of single segment codas is limited
to /d s θ n l r/. Thus, he proposes that the words in Class IIIA are those that would
61
otherwise end in an unacceptable coda without the addition of the epenthetic [e]:
*jef/jef[e], *verd/verd[e], *delant/delant[e]. Class IIIA´ corresponds to words that
end in the acceptable set of final codas /d s θ n l r/. In this way, Class IIIA and
IIIA´ are in complementary distribution.
However, Harris must address two inconsistencies in the data. The first is
the number of words in Class B that end in [e] but are preceded by consonants
that could be syllabified as word-final codas and, therefore, should not trigger
epenthesis. Harris argues that these exceptions must have an unspecified vowel in
their lexical entry that manifests as [e], the default vowel in Spanish (1989: 73):
inmune /inmun + V/ “immune.”
The second issue regards relatively recent borrowings into the language
that do not adhere to the final coda condition described above. For example, a
borrowing from English, esnob [esnob] “snob,” and one from French coñac
[koñak] “cognac” illustrate that final epenthesis is not activated even though they
end in the otherwise illicit final codas [b] and [k], respectively. Furthermore, they
do not form plurals in the typical manner of Spanish lexemes that end in a
consonant: esnobs/*esnobes, coñacs/*coñaques (cf. árbol ~ árboles/*árbols,
escritor ~ escritores/*escritors).
Harris accounts for these contrasts by assuming two separate grammars:
one for xenonyms and one for “domestic” substantives (as Harris labels them).22
22 This latter approach, of course, begs the question if and when a word like esnob would ‘graduate’ from the xenonymic grammar into the domestic grammar. Since many non-patrimonial Latinate words in Spanish were at some point xenonyms, one wonders if all such words remained for a time in the xenonymic grammar before being accepted as “Spanish.” Conjecturing into the future, if the plural form esnobs never transforms into esnobes or the singular esnob into esnobe, are these types of lexemes destined to remain forever in a xenonymic linguistic limbo?
62
Harris groups these words, along with words that end in /i/ or /u/, together into
Class V.
III. Harris’ Word Classes & Diminutivization
For the purposes of diminutivization, it is worth taking a moment to
evaluate these form classes to determine what, if any, correspondence can be
made between them and the selection of diminutive morphemes. For the purposes
of clarity, I have chosen the Sonoran data in Crowhurst (1992) as the basis for this
evaluation. Relevant dialectal variants are included and will be integrated in the
full analysis in Chapter 4.
On the face of it in (7), there seems to be little evidence to suggest a one-
to-one correspondence between the diminutive suffixes and Harris’ form classes.
The –it{o, a} suffix appears in Class I, Class II, and all the subsets of Class III.
Meanwhile, the –cit{o, a} suffix plays a prominent role in Class I and Class II,
and can be found in all the subsets of Class III.23 This lack of one-to-one
correspondence is not surprising since none of the previous analyses of Spanish
diminutives has proposed a strictly morphological account in which form class
alone dictates diminutive selection without any appeal to prosodic or phonological
restrictions.
(Interestingly, the word for “snob” in Portuguese has been borrowed into that language as esnobe with an apparent final epenthetic [e]). 23 Surprisingly, the diminutives for words that end in [l] in Class III A´ (e.g. árbol, canal) select the –it{o, a} suffix (arbolito, canalito) instead of –(e)cit{o, a}, generally selected by other consonant-final words. Previous analyses have either left the issue unsettled or proposed a condition/constraint on the heterosyllabic cluster [ls] (or [lθ] for Peninsular). I place the present work among the former group.
63
(7) Class I = -ito pasito “step”
guapito “handsome” Exceptions: novio ~ noviecito/*novito24 “boyfriend/groom” Dialectal: miedo ~ miedecito/miedito “fear” Class II = -ita casa ~ casita “house”
pelota ~ pelotita “ball” Exceptions: llorona ~ lloroncita “crybaby” ladrona ~ ladroncita, etc. “thief (f.) novia ~ noviecita “girlfriend/bride” Dialectal: puerta ~ puertecita/puertita, etc. “door”
Class III A = -cito/a (disyllabics)
madre ~ madrecita “mother” Exceptions: None. = -ito/a (longer than disyllabic)
uniforme ~ uniformito “uniform” Exceptions: Occasional. chocolate ~ chocolatecito/chocolatito Class III A´ = -(e)cito/a (monosyllabics) pan ~ panecito/pancito25 “bread”
24 Harris (1994: 186) notes that for many speakers the diminutive fresquecito of fresco “fresh” is as or more accepted than the expected fresquito. In a footnote, Harris elaborates that it “is a fact, however, that some classes of stems have fewer arbitrary lexicalised diminutives than others, and we would like an explanation for that” (189). I take those words that end in unstressed /o/ to be highly regular in their formation of diminutives, both inter- and intradialectally.
64
Exceptions: Words ending in [l] árbol ~ arbolito “tree” papel ~ papelito “paper” = -cito/a (disyllabics or longer)
balcón ~ balconcito “balcony” corazon ~ corazoncito “heart” Class III B = -cito/a (disyllabics) clase ~ clasecita “class” = -ito/a (longer than disyllabics) envase ~ envasito “container” Class IV Limited data available. virus ~ virusito/?viruscito/?viritus “virus” brindis ~ brindisito “toast” Socrates ~ Socratitos/?Socratesito “Socrates” lejos ~ lejitos “far” dosis ~ dosisita “dosage” Class V Limited data available. esnob ~ esnobito/esnobcito “snob” *esnobecito tribu ~ tribucita/?tribita “tribe”
Leaving aside for a moment the data available for Class IV and V, in the
following section I will discuss the data and issues for Classes I-III that must be
25 The difference between these two monosyllabic forms is dialectal: panecito for North Central Peninsular and Sonoran; pancito for Nicaraguan and Paraguayan. For present purposes it is crucial that *panito is not an attested option.
65
considered in order to arrive at an adequate assessment of both a specific role for
Harris’ form classes and for morphology in general. The conclusions reached in
this section will form the basis for the morphological assumptions that underpin
my analysis of diminutives in Chapter 4.
Class I The data reflect that –ito is the suffix most associated with Class I. In the
Sonoran dialect presented by Crowhurst (1992), the only exceptions to this
generalization to be found are disyllabic words that end in the diphthong [jo]:
(8) gafio ~ gafiecito “hook” patio ~ patiecito “patio” ebrio ~ ebriecito “tipsy” The data in (8) illustrates an inherent limitation of the –it{o, a} suffix. Normally,
as a vowel-initial suffix, it attaches with ease to a consonant-final stem, as in (9).
(9) chango “monkey” [ʧang] -it] -o] However, as discussed in Chapter 2, tautosyllabic homorganic vocoids are illicit
in Spanish when onset position is occupied: *[ji]Rhyme. Direct attachment of -it{o,
a} to the words in (8) would create precisely such an environment:
(10) *ga.fji.to *pa.tji.to *e.brji.to
66
Data presented in Harris (1994: 185) from a speaker of a Mexican (non-Sonoran)
dialect suggests that this phonotactic constraint might be resolved in a non-
systematic fashion.
(11) 1 2 3
ind[j]o *ind[i.í]to *ind[je]cito ind[i]to “Indian” sab[j]o *sab[i.í]to sab[je]cito *sab[i]to “wise” lab[j]o lab[i.í]to lab[je]cito ?lab[i]to “lip” ap[j]o ap[i.í]to ap[je]cito *ap[i]to “celery” ampl[j]o ampl[i.í]to *ampl[je]cito *ampl[i]to “ample” At least in the case of this individual speaker, there are three available repair
strategies in order to avoid the illicit syllable structure. In Column 1, the glide
found in the stem surfaces as the head of its own syllable, creating a hiatus. The
data in Column 2 is consistent with what is presented in Crowhurst and shows that
selection of the –(e)cit{o, a} suffix has the benefit of avoiding *[ji]Rhyme while at
the same time conserving segmental identity through maintenance of the glide.
Complete elimination of the stem-final segment—the repair strategy in Column
3—appears to be the least-favored of the three options, judged as fully acceptable
only for indio.
It is revealing to note the absence of another possible candidate,
*sab[i]cito, in which -cit{o, a} attaches to a stem in which the glide has lost
segmental identity and heads its own syllable. This negative evidence suggests
67
that attachment of –cit{o, a} to this class of words is only acceptable to the extent
that it resolves any conflict that would otherwise be created by attachment of
–it{o, a}. Whereas sab[je]cito maintains the non-syllabicity of the glide, the
hypothetical candidate *sab[i]cito, in a sense, brings nothing to the table. It
neither maintains the glide (as in Column 2) nor does it allow for selection of the
(presumably) preferred suffix of Class I (as in Column 1 and 3). While the data in
(11) at first blush does little but underscore the idiosyncratic nature of diminutive
formation for some speakers, closer examination strengthens the case for a
morphological preference of the –it{o, a} allomorph to Class I words. The
acceptability of competing forms such as api.ito and labi.ito suggests the
willingness of speakers to sacrifice segmental identity of the glide in order to
maintain the association between –it{o, a} and this class of words.
Further evidence that supports this notion that the idiosyncratic and
“fringe” nature of words such as sabio is due to the incompatibility of the glide
with the first segment of the suffix –it{o, a} can be found in an analysis of Class I
stems that end in a vowel (12a-b) or the back glide [w] (12c).26
(12) a. feo [[[fe-] it] o] fe.íto “ugly” b. tío [[[ti-] it] o] ti.íto “uncle” c. antigüo [[[antigw-] it] o] antigüito “ancient” In (12a), the attachment of the –cit{o, a} suffix to the vowel-final stem [fe-]
would afford a distinct phonological advantage over –it{o, a} attachment because
26 I present these examples for the purpose of illustration. Although they do not appear in Crowhurst, there is nothing in her data to indicate that they form their diminutives in an exceptional manner.
68
the unattested candidate *fecito avoids the creation of a dispreferred onsetless
syllable that we find in the actual output [fe.i.to]. However, the critical distinction
to keep in mind is between the notion of “dispreffered”—in the case of onsetless
syllables in Spanish—and that of “disallowed”—in the case of the never violated
constraint *[ji]Rhyme. As the data in (8) and (11) indicate, the –cit{o, a} suffix is
clearly available to speakers for Class I words, but we must conclude that
avoidance of an onsetless syllable is not a severe enough condition for its
selection.
Meanwhile, the diminutive form in (12b) is arguably more marked than
the form in (12a). Not only does –it{o, a} attachment to tío create an onsetless
syllable as in (12a), but the hiatus is of homorganic vowels [i.i]. Independent
evidence that Spanish employs mechanisms to avoid such a hiatus of homorganic
vowels can be found in the behavior of the definite article for certain [+ feminine]
nouns.27 For the overwhelming majority of feminine nouns in Spanish, the
definite article is la, as in la chica “the girl.” However, in cases in which the
noun begins with stressed /a/, the canonically masculine definite article, el, is
selected: el agua [el ágwa]; *la agua [la ágwa]. In the plural form, Spanish
reverts to the standard feminine definite article since the final /s/ breaks up the
adjacent homorganic vowels: las aguas [las ágwas] “the waters.” In words that
begin with an unstressed /a/ the phenomenon is not triggered and presumably the
two unstressed vowels coalesce: la almeja [lal.mé.xa] “the clam.” In the case of
el agua, then, Spanish permits selection of a marked definite article (el for
27 See Kikuchi (2001) for an OT analysis of this phenomenon couched in Correspondence Theory.
69
feminine nouns) in order to avoid a marked structure.28 As with the example of
noviecito, in which the –cit{o, a} allomorph is selected instead of expected –it{o,
a} in order to avoid *[ji]Rhyme, Spanish tolerates a degree of irregularity in the
service of the phonology.
Again, as in the case of feo, there are distinct advantages to an unattested
candidate *ticito [ti.si.to] for tío “uncle” in which the first element of the
diminutive suffix works to break up the adjacent homorganic vowels present in
tiíto [ti.í.to]. Any analysis of diminutivization that relies purely on phonological
information will have difficulty explaining how [ti.í.to] and [fe.í.to] are more
well-formed than their unattested counterparts *[ti.si.to] and *[fe.si.to].
Consequently, it is becoming clear that the evidence points to a tightly bound
morphological relationship between Class I words and the diminutive suffix –it{o,
a}. This relationship will be formalized in Chapter 4.
Finally, it might be argued that the glide-final stems are somehow
responsible for selection of the –cit{o, a} suffix. However, in (12c) the stem for
word antigüo ends in a back glide [w] that is not homorganic with the high front
vowel [i] in –it{o, a}. Since the cluster [gwi] is well formed and markedness is
not at stake, we would predict the standard selection of the -it{o, a} suffix for a
Class I word, which it indeed selects: antigüito [an.ti.gwi.to].
28 A similar situation occurs in Spanish with o [o] “or” becomes u [u] and y [i] “and” becomes e [e] when the following word begins with a homorganic stressed or unstressed vowel: uno u otro “one or another;” Tristano e Isolda.
70
Class II In terms of diminutive formation, there is little to distinguish the behavior
of Class II words from that of Class I words. The distribution of –it{o, a} and
–cit{o, a} is nearly identical within the two classes. However, there is one point
of contention with respect to the classification of the seemingly exceptional
diminutive types in (13a) versus the classification of those in (13b).
(13) a. ladrona ladroncita “thief (f.)” patrona patroncita “boss/landlady” b. corona coronita “crown” persona personita “person” The issue is whether a minimal pair such as corona ~ coronita / patrona ~
patroncita are members of the same word class or not. If all lexemes that end in
/a/ belong to Class II, then the answer is simple. Yet the exceptional behavior of
ladrona and patrona indicates that they are adhering to the diminutive pattern of
their masculine counterparts, which Harris groups into Class IIIAˈ (14).
(14) ladrón ladroncito “thief (m.)” patrón patroncito “boss/landlord” Harris reconciles this inconsistency precisely by appealing to the shared
morphological structure between the words in (13a) and (14). In his argument he
presents the contrast between Llorona ~ Lloronita and llorona ~ lloroncita
“crybaby”. The former is the name of a character in Mexican folklore and has no
71
masculine counterpart in the lexicon (*Lloron) and therefore is properly classified
as a Class II noun that selects the –it{o, a} suffix. The latter is the [+ feminine]
manifestation of llorón, which belongs to Class IIIAˈ and consists of the stem
llor- and the suffix –ón. Therefore, “in llorona (lower case)...the acccessibility of
the Class III suffix determines the choice of allomorph –citV, though llorona is
segmentally identical to Llorona (1994: 187).”
While this argument appears to strengthen Harris’ form class partition,
Ambadiang (1997: 108) is critical of this explanation and wonders why the same
appeal to internal structure could not be made for the pair niño/niña ~
niñito/niñita “boy/girl,” which Harris divides into Class I and Class II
respectively. If llorona is simply a [+ fem.] instantiation of the Class IIIAˈ noun
llorón, then it could be argued that niña is simply a [+ fem.] instantiation of niño.
Thus, niño/niña would both be placed in Class I.
Whether or not this issue is grounds for a radical reworking of Harris’
form classes is a debate that is beyond the scope of the present work. However, it
is clear from Harris’ approach to the Llorona/lloroncita issue that he implicitly
accepts the conclusion that the final word marker is not always the ultimate
arbiter of form class distribution. Thus, I propose that his Class IIIAˈ be
amended to include this subset of feminine nouns:
(15) Class IIIAˈ ladrón29 [+/- fem] escritor [+/- fem], etc. 29 It is not immediately clear whether the –ón in ladrón and patrón is related to the augmentative suffix –ón in llorón, which is derived from the verb llorar [[llor-] ar] “to cry.”
72
With this movement of the previously exceptional ladroncita types from Class II
into Class IIIAˈ, it is now clear that—at least with respect to the distribution of
the diminutive suffixes—Class I and II exhibit precisely the same behavior. From
the available evidence for Sonoran, all words in these two classes select the –it{o,
a} suffix (16a) except under the severe phonotactic conditions wrought by the
incompatibility between the stem-final glide and the initial segment of the
diminutive suffix, in which case –citV is (variably) selected (16b). The fact that
diminutive forms of these words maintain their respective word marker, /o/ or /a/,
is not germane to the selection of the suffix and therefore not a reasonable
criterion for partitioning them into two separate classes. Therefore, at least for the
purposes of diminutivization, I propose the merger of these into a single Class D1,
as shown in (17).
(16) a. chamaka chamakita “girl” molacho molachito “toothless” b. novio noviecito “boyfriend” novia noviecita “girlfriend” (17) Class I Class D1
Class II
73
Class III
Within the three subclasses that Harris establishes for Class III, the suffix
–(e)cit{o, a} is dominant in terms of its distribution. The –itV suffix is selected
only by trisyllabic or longer stems in Class IIIA (uniforme ~ uniformito
“uniform”) and Class IIIB (envase ~ envasito “container”). Trisyllabic or longer
stems in Class IIIA´ select the –cit{o, a} suffix (corazón ~ corazoncito “heart”).
That said, whatever ties that one might establish between –citV and Class III
words, it is clear that this relationship is considerably less constrained than the
bond between –it{o, a} and Class D1 words. As discussed above, the –cit{o, a}
suffix only manifests itself in Class D1 as a response to a severe phonotactic
constraint. On the other hand, the evidence does not support an analogous
situation between Class III and –(e)cit{o, a}. In fact, Prieto (1992: 173) indicates
that some speakers will use and accept both suffixal options for trisyllabic or
longer words ending in /e/ (18a). Crowhurst also indicates that some disyllabic
words in the Sonoran dialect, such as sangre “blood” and tigre “tiger,” select the
–itV suffix: sangrita, tigrito. Yet, no such variation has ever been attested for
Class D1 words (18b).
(18) a. chocolate chocolatito/chocolatecito “chocolate”
envase envasito/envasecito “container”
pupitre pupitrito/pupitrecito “(school) desk”
b. casa *casacita/*casecita “house”
74
libro *librocito/*librecito “book” With respect to members of the form Class IIIA´, that is those words that end in
any of the segments in the consonant set /d s θ n l r/, there is little consistency
with respect to their diminutive formation. All monosyllabics in the Sonoran
dialect select the –(e)cit{o, a} suffix (pez ~ pezecito “fish,” miel ~ mielecito
“honey”), while disyllabics or longer vary greatly. For example, /l/-final words
select –itV (papel ~ papelito “paper”), /n/- and /r/- exclusively select -citV (balcón
~ balconcito “balcony,” pintor ~ pintorcito “painter,” corazón ~ corazoncito
“heart”). Meanwhile, Crowhurst notes that speakers reject either possibility for
the /d/-final word pared ~ *paredcita/*paredita “wall”).
Given this state of affairs, it would be imprudent to propose any type of
morphological association between the words that make up Class III (or any of its
subsets) and either the -(e)cit{o, a} or –it{o, a} suffix. However, for the purposes
of notational clarity, I will assign them to Class D2. This designation is not meant
to imply any uniformity of the membership it contains, but merely to exclude it
from those words in Class D1 that I propose are subject to stringent
morphological restrictions.
Class IV & V
Crowhurst (1992) offers no examples of Class IV types in her article, but
Prieto (1992: 177) offers several from an unspecified dialect, which I have
divided into two groups based on the lack of change (19a) or change (19b) of the
word marker (-Vs, as Harris understands it) in the diminutive form.
75
(19a) a. -Vs = /-os/ or /-as/
lejos lejitos “far” Carlos Carlitos “Charles” Marcos Marquitos “Mark” Lucas Luquitas “Lucas” Honduras Honduritas “Honduras” b. -Vs = /-es/ or /-is/
Dolores Dolorsitas “woman’s name” Mercedes Merceditas “woman’s name” lunes lunecito “Monday” Gertrudis Gertruditas “woman’s name”
Upon examination of the data in (19a) it is tempting to analyze –it{o, a} as
an infix rather than a suffix, since the –os in Carlos is carried over into the
diminutive form Carlitos. However, there are several reasons to doubt such a
conclusion. The first concerns the data in (19b), particularly the names Mercedes
and Gertrudis. If –it is an infix, then there is no way to explain why the
diminutive forms of these words would not be *Mercedites and *Getruditis. Of
course these are proper nouns and it is reasonable to expect some erratic behavior,
yet the same logic would apply to those Class III nouns that variably accept –it{o,
a}. For example, an analysis that assumes –it is an infix would need to explain
why the diminutive for uniforme is uniformito and not *uniformite. One possible
76
explanation is to assume that –it is a suffix in some environments and an infix in
others; but this adds an unnecessary level of complexity when compared to an
approach that assumes across the board suffixation.
Under the assumption of universal suffixation, both diminutive suffixes
must remain faithful to those word endings that they are able to replicate: -/o/,
-/a/, -/os/, -/as/. The final -/es/ in Mercedes is not an available diminutive ending,
so the closest possible ending -/as/ is selected.
As has been noted elsewhere, diminutives seem to be unique among
affective affixes in that, with few exceptions, they maintain the marked /a/ word
marker for masculine nouns, such as el problema ~ problemita “problema,” and
the /o/ for feminine nouns, la modelo ~ modelito “(fashion) model.” In (20a) the
canonical feminine word marker /a/ of the masculine noun poema is maintained in
the diminutive form. Yet, in (20b) the masculine option of the augmentative –azV
is selected even though a form –aza is available that would maintain the /a/ word
marker in the base.30
(20) a. el poema el poemita “the poem” b. el poema el poemazo “the poem” This implies that the diminutive suffixes, although limited to either of the
canonical word markers, would seem to have the flexibility to accommodate any 30 Unfortunately, Crowhurst (1992) does not present poema/problema types in addressing the –ote/ota augmentative. An informal Google search, however, strongly suggests that problemota is more common than problemote. While this is not a completely analogous case to this –azo/aza augmentative since there is no –oto variant of the affix, the data in Crowhurst’s article makes it clear that –ote is the preferred suffix for masculine nouns. Additional searches have revealed the less frequent usage of unexpected poemito and poemaza. This fact does not alter the basic assertion that Spanish diminutive forms of poema/problema types are able to maintain the final vowel.
77
idiosyncrasies of the base form as long as they are in line with the inflectional
possibilities afforded by the diminutives. In the example of el uniforme ~
uniformito, the diminutive suffix cannot reproduce the base-final /e/ and so it
selects the canonical word marker /o/ for masculine nouns. In the case of lejos,
the diminutive form can easily replicate the final /os/ because of its similarity to a
canonical masculine, plural ending as in pat/os/ ~ patitos “ducks.” With Mercedes
~ Merceditas, the diminutive form reflects the base form to the degree possible by
maintaining the final /s/ and respecting the gender by selecting word marker /a/,
since /e/ is not available.31 The segmental similarity between [lex-os] and [pat-os]
and their identical diminutivization is sufficient justification to include the former
in Class D1 with the latter.
The limited data available on diminutives for Class IV (and the
inconsistencies therein), the wide range of word markers assumed by Harris for
this form class (-as, -es, -is, -os, -us), and the apparent resistance of some /s/-final
words to diminutivization all combine to create highly unfavorable conditions for
postulating that the class as a whole should be associated with one or another
diminutive suffix. The situation is even grimmer for Class V given that Harris
considers this a “catch-all bin” for xenonyms and other words that do not fit into
one of the other four classes. Therefore, I will include all members of Class IV
and Class V in Class D2.
Finally, in addition to these words, I include in Class D2 a group of words
that end in a stressed final vowel—exemplified in (21)—that Harris does not
31 This is similar to the conclusion of Ambadiang (1997) that the morphological structure of the word is analyzed as being saturated, as discussed in Chapter 2.
78
address directly. Since the final vowel of these words are maintained in
derivational morphology (as opposed to the situation as described in (2) at the
beginning of this chapter), they do not fit the definition of a word marker as
proposed by Harris. Thus, I assume that he intends them to be included in Class
V. Regardless of word length these words always take the –citV suffix.
(21) matiné matinecito “matinee” menú menucito “menu” In this section I have examined Harris’ five word classes in order to
evaluate to what degree a correspondence can be made between these classes and
their respective diminutive forms. I have argued that in the Sonoran dialect there
is a strong preference among Class I and Class II words for the –it{o, a}
morpheme. Selection of the –cit{o, a} morpheme by these words is only in
response to phonological constraints. Since Class I and Class II words exhibit
identical behavior with diminutivization, I combined them into a single group,
Class D1. All other word types are placed in Class D2.
In a sense, this division hearkens back to Harris’ earlier work on word
markers mentioned in Section II in which he proposes a division based on a
lexical hierarchy (1991a: 33). My proposal for diminutives, however, differs in
that I include in Class D1 all words that end in canonical unstressed /o/, /a/, /os/,
or /as/, regardless of their regularity; this includes words such as problema
“problem” that end in /a/, but are actually masculine, and those that end in /o/, but
are actually feminine, such as modelo “(fashion) model.” Although problema and
79
lejos “far” are arguably irregular in some sense, this issue does not play a role in
diminutivization. What matters is that they are marked with the canonical
unstressed vowels and therefore subject to the morphological association with
–it{o, a}.
While some word types that I have placed in Class D2 may exhibit a
strong tendency toward a particular diminutive morpheme, the key distinction that
I will develop in Chapter 4 is that these preferences are revealed to be
phonological in nature and not a result of a morphological bond of the type I
propose for Class D1.
IV. Underlying Assumptions Now that I have established the two word classes for diminutivization, in
this section I present my assumptions concerning the input. The chief issue
concerns whether diminutivization should be modeled as a word-level process
(22a) in which the terminal element is deleted by the phonology upon suffixation
(Bermúdez-Otero 2007) or if suffixation occurs at the level of a stem that lacks
the final vowel (22b).
(22) a. casa Input: [[kasa] DIM] casita b. casa Input: [[kas-] DIM] casita I argue that the deletion position illustrated in (22a) is insufficient to account for
the behavior of much diminutive formation and that the evidence points to an
analysis in which the final vowels are not present in this morphological process.
80
The persistence of the stressed final vowels in the diminutivized words in
(21) (menú~menucito) underscores again the apparent status of the terminal
element as a separate morphological entity in Class D1 words. As previously
stated, forms such as *casacita or *casaíta are unattested due to an apparent
restriction on terminal elements appearing in non-word-final position.32 This is
also true for derivational morphology in general, as seen in other morphological
processes (23).
(23) casa + ero/a casero/a *casa-ero/a “homemade” casa + ona casona *casa-ona “big house” There are several issues, however, in the suggestion of such a restriction that are
worth addressing with an eye toward the assumptions I will make for my
Optimality Theory analysis in the following chapter. The first concerns how to
analyze the potential output candidates for casa in (24).
(24) a. casacita b. casaíta c. casita
d. casecita As we know, candidate (c) is the winning candidate; and if our constraints and
their ranking are correct, then the tableau should illustrate the superiority of
candidate (c) over candidates (a), (b) and (d).
32 Colina (2003: 57) formalizes this restriction with the following constraint:
*TE-: No word markers (or terminal elements) in positions other than word final (as an alignment constraint: The right edge of a terminal element must be aligned with the right edge of a word).
Colina assumes that –it{o, a} is an infix.
81
At the most basic level—between a single lexical input and its
corresponding winning output—Optimality Theory (Prince & Smolensky
1993/2004) provides us with a rather straightforward framework for modeling the
relative costs involved in the violation of universal constraints. In Tableau (25),
candidate (a) removes the segment /l/ from the input and thus fatally violates the
constraint MAX, which legislates against such removal. Conversely, candidate (b)
fatally violates DEP through the addition of the vowel /e/; the improbable
candidate (d) violates both. Assuming no markedness constraints are factored
into the equation, candidate (c) is the optimal candidate through its most faithful
reflection of the input.
(25) INPUT: árbol MAX DEP a. árbo *! b. árbole *! c. árbol d. cárbo *! *
Modeling a morphological process such as diminutivization depends to a great
degree upon the assumptions that are made about the input. In the case of árbol
“tree” in (25), Spanish offers no independent evidence to suggest that the input
form be at variance with the output. With respect to diminutive formation,
however, the evidence available to us—the forms that speakers produce—
suggests between one and three diminutive allomorphs. On the one hand, we can
assume an underlying –cit{o, a} allomorph, in which the first segment [s] or [θ] is
deleted in some contexts, surfacing as –it{o, a}; in other contexts an epenthetic [e]
82
produces a third allomorph –ecit{o, a}. Other approaches might assume all three
allomorphs are available in the input or some combination of two out of the three.
Obviously, whatever assumptions are made about the input will impact the types
of constraints that are in play and the respective violations that each candidate
incurs.
The question of underlying assumptions must also be extended to the
lexeme to which the diminutive attaches. Bermúdez-Otero (2007), for example,
argues within a Stratal OT framework that diminutivization is a word-level
process in which the terminal element (“stem formatives” in his terminology) is
present in the input (26).
(26) casa “house” [[kas - a]stem it - a
Under this approach, a candidate such as casaíta is not eliminated through some
morphological constraint that requires final placement of /a/, but rather due to the
following alignment constraint:
ALIGN (suffix, onset): If an input vowel V is initial in a suffix attached to
a stem, then assign one violation mark for every segment intervening
between the output correspondent of V and the nearest preceding onset
segment. (Bermúdez-Otero 2007: 295).
Thus, the stem formative /a/ is essentially deleted.
83
(27) kas-a]stem it-a ALIGN (s, o) MAX-V a. ka.sa.í.ta *! b. ka.sí.ta *
While this is an attractive approach, it raises more questions than it answers. First
among them concerns an explanation of why consonant-initial –citV is not
available to attach to the stem [kas-a] and thus vacuously satisfy ALIGN (s, o).
Tableau (28) illustrates that a candidate (28c) casacita would actually be
predicted. If there is no morphological constraint against terminal elements/stem
formatives in non-final position as Bermúdez-Otero contends, then we must
presume that the grammar exacts a cost for their deletion; in effect, this is the
purpose of MAX-V in his analysis. Candidate (28c) is able to satisfy MAX-V and
maintain the stem formative. The fact that *casacita is not an attested form
among speakers remains unexplained in his analysis.
(28)
kas-a]stem it-a, sit-a ALIGN (s, o) MAX-V a. ka.sa.í.ta *! b. ka.sí.ta * c. ka.sa.sí.ta
As we have seen previously in this chapter with respect to the
diminutivization of certain diphthong-final words (patio~patiecito “patio”,
novia~noviecita “girlfriend”), Spanish reveals a willingness to select a diminutive
morpheme that (from a statistical standpoint) is not generally associated with a
class of words if such a selection produces desired effects—in this case,
avoidance of the *[ji]Rhyme constraint and maintenance of the glide in the
diminutive form.
84
These cases prove difficult to explain within an account that analyzes the
final vowel –o/a as a stem formative that is deleted in the phonology. Assuming
the underlying structure in (29a), it is incumbent on such an account to explain
why –cit{o, a} does not simply attach to the stem (29b). This unattested form
would afford certain advantages: non-deletion of the stem formative, avoidance of
the illicit *[ji]Rhyme structure, and preservation of the glide in the output.
(29) a. [nobj-a]Stem b. [nobj-a]Stem [–sit-a] = *[nobjasita] Instead, the deletion analysis must explain why noviecita is a more advantageous
form than *noviacita. I submit, however, as in (28), that the deletion analysis
actually predicts the latter form and would need to account for a rather peculiar
and far more complex process in order to arrive at the attested output, noviecita.
Recall that the deletion analysis as proposed by Bermúdez-Otero (2006) assumes
that diminutivization occurs at the word level and that the stem formative is
deleted by the phonology. In the case of novia (30a), the selected morpheme [–
sit-a] is not vowel initial and therefore will always vacuously satisfy the constraint
ALIGN (suffix, onset). This leaves the analysis with no justification for [a]
deletion. At any rate, the requirement for word-level attachment leaves us to
assume that either [a] transforms into [e] by some unexplained process (30ai) or
that [a] is deleted (unjustifiably) and then an epenthetic [e] is added (31aii).
On the other hand, the analysis in (30b), which assumes a never-violated
constraint on non-final placement of Terminal Elements, obviates these issues and
85
eliminates the complexity inherent in (30a). It requires only a well-documented
process of [e] epenthesis in order to arrive at the attested output form.
(30) a. Stem Formative Deletion Analysis
[nobj-astem] [–sit-a] i. [a] → [e]
ii. [a] → [ø]; [ø] → [e] [nobj-e] [–sit-a] [nobjesita]33 b. Morphological Constraint on Non-Final TE Analysis
[nobj-stem] [–sit-a] [ø] → [e] [nobj-e] [sita] [nobjesita]
In defense of the deletion analysis, Bermúdez-Otero (2007) is not meant to
be a full-fledged analysis of Spanish diminutive formation and many word types
are not addressed. That said, the diminutive analysis presented in the article is
offered as support for the phonological deletion of stem formatives in derivational
morphology and as evidence for a Stratal OT approach more generally.
Therefore, it is not unreasonable to extrapolate the approach across other word
types as I have done here.
33 This approach would create similar issues for puerta ~ puertecita types in Peninsular dialects.
86
With respect to those words that end in /e/, such as clase or madre, whose
diminutives are clasecita and madrecita, it might be asserted that the presence of
the [e] in the diminutives suggests a word-level process. These examples would
run counter to my claim that diminutivization is a stem-level process. So, if it is
assumed that the final /e/ in clase and madre is a terminal element on par with /o/
and /a/, then either the [e] present in the diminutives severely undermines the
morphological constraint approach in (30b) or it is not the same [e].34
Whether the concept to which we refer be called a Terminal Element,
Stem Formative, Class or Word Marker, in a sense, these terms are all shorthand
for that (those) segment(s) appearing at the end of the base form of a word, but in
a word derived from that base either does not appear at all (31a), does not appear
in the same position (31b) or in the same form (31c).
(31) a. Base: cabeza ‘head’ k a β e s a X Derived: cabezón ‘pig headed’ k a β e s _ o n ø b. Base: perro ‘dog’ p e r o X
34 Another potential issue with the proposal that diminutivization is a word-level process concerns plurals. One must assume that there is some constraint on plural markers appearing within morphological boundaries, but not for TEs/stem formatives/word markers, etc. i.e. [cas-as] –it-a or [cas-a] –it-a-s.
87
Derived: perrito ‘tiny dog’ pe r i t o X
c. Base: casa ‘house’ k a s a X
Derived: caserío ‘hamlet’ k a s e ɾ i o Y In (31a) the augmentative suffix –ón transforms the noun cabeza into the
masculine adjective cabezón. In this case, the base contains a terminal element
that is not present in the derived form.35 The fact that a phonotactically-sound
suffix that contains the canonical masculine terminal element /o/ is unattested,
such as *-ono (*cabezono), suggests that preservation of a slot for terminal
elements in derivational morphology does not trump faithfulness to the form of
the morpheme. Of more import for our current discussion, none of the derivations
in (31a-c) precisely reflects the structure of the base with respect to the final
vowel. Indeed, the debate between phonological deletion and morphological
restriction of the unstressed final vowel concerns why the segment does not
surface word-medially, but not whether it surfaces word-medially. In other
words, both approaches assume that terminal elements do not appear word-
medially, regardless of the motivation. Thus in the case of the unstressed final /e/
35 The feminine form of the augmentative –ona does manifest the Terminal Element /a/, as in cabezon/a/.
88
in clase and madre, there is no basis to believe that it is exceptional in this regard;
the word-medial [e] in clasecita and madrecita is not the same [e] as in the
respective bases, clas[e] and madr[e]. Any confusion arises simply because the
epenthetic vowel in Spanish happens to be phonologically identical to the final
unstressed vowel of these words. It is fortuitous, perhaps, that the optimal outputs
are clasecita and madrecita (for reasons that will be explained fully in Chapter 4)
and that they appear to preserve the structure of the base, but the medial /e/ is not
a carryover from the base. In a similar way, the medial /i/ in dosificar “to
measure out” is part of the verbal morpheme –ificar (cf. clase~clasificar
“class~classify,” diverso~diversificar “diverse~diversify”) and should not be
analyzed as deriving from the base, dosis, just as the medial /e/ in broncear “to
tan” and cobardear “to act cowardly” are part of –ear and do not come from
bronce “bronze” and cobarde “coward” (cf. gato~gatear “cat~climb like a cat,”
plata~platear “silver~plate with silver”).
Essentially, we must conclude that the [e] present in the diminutive form
is epenthetic and that the stems have the structure in (32).
(32) [klas-]Stem [madr-]Stem This is an attractive assumption because it establishes parity between –o, –a and
–e final words, in that the diminutive morphemes attach at the stem level in all
these cases.36
36 See Colina (2003b) for arguments in support of the Terminal Element status of word-final /e/.
89
Note that even if we assume that the final /e/ in the base is not a Terminal
Element but epenthetic, this would not change the assumption about the
underlying forms since the epenthetic [e] would be analyzed as falling outside the
domain of the stem, as in (33).
(33) [klas-]Stem [e] [madr-]Stem [e] To summarize, I argue that the input for words in both Class D1 and D2
do not contain the unstressed, final vocalic element. The difference between the
behavior of (34a-b) and (34c) is that the final vowel in café is stressed and,
therefore, analyzed by speakers as critically forming part of the stem. This is
essentially analogous to the case in (34d), balcón “balcony,” in which there is no
base-final vowel and the entire word acts as the stem.
(34) a. Base: casa Input: [[kas-]Stem]DIM casita b. Base: clase Input: [[klas-]Stem ]DIM clasecita Base: uniforme Input: [[uniform-]Stem]DIM uniformito c. Base: café Input: [[kafe-]Stem]DIM cafecito d. Base: balcón Input: [[balkon-]Stem]DIM balconcito With the input of the stems established, it is now necessary to discuss the
input of the diminutive morpheme. By strictly looking at the outputs, one can
postulate a maximum of six forms.
90
(35) a. -ito gato “cat” gatito -ita mesa “table” mesita
b. -cito balcón “balcony” balconcito -cita mujer “woman” mujercita
c. -ecito pan “bread” panecito -ecita flor “flower” florecita As argued previously in Section III, I consider –it{o, a} to be a suffix rather than
an infix. In the masculine word poema “poem,” the diminutive form poemita
retains the idiosyncratic word marker /a/ because it falls within the inflexional
possibilities of the morpheme. As a point of comparison, consider the
augmentative morpheme –ote/ota. In (36a), -ote is not a suffix for gatote and an
infix for grandote. Maintenance of word-final /e/ in the latter is only a reflection
of the inflectional possibilities of the suffix. Note that the feminine augmentative
of grande is grandota. Thus, an assumption of infixation for grandote would
require a separate process of suffixation for grandota and gatote. A far simpler
system assumes suffixation in all cases and analyzes the grandote example for
what it is: a coincidental correspondence between the base-final vowel grand[e]
and the augmentative-final vowel grandot[e].
(36) a. -ote gato “male cat” gatote grande “large (m.)” grandote b. -ota gata “female cat” gatota
91
grande “large (f.) grandota Of course as an adjective, grande is syntactically bound to respect the
gender of the noun it is modifying (although this fact is not visible, since final /e/
does not have a canonical gender correspondence). Nouns, on the other hand, are
under no such jurisdiction. The diminutive form of poema is free to replicate the
word-final vowel /a/ without additional consequences. In other words, poemita is
no more “irregular” than poema. Again, rather than assume infixation and its
inherent complexity, I argue that diminutive allomorphs are suffixes that
accommodate the base form to the degree possible. When there is a clash
between the base form and the diminutive suffix, the inflectional limitations of the
suffix will always trump faithfulness to the base form.
Since diminutive forms always end in either /o/ or /a/, in Tableaux (37)
and (38) I establish the available set of inflectional markers for diminutive
morphemes {o, a} as part of the input. This move allows for the exchange of a
morpheme-specific constraint IDENT-DIM (of dubious universal status) for the
universal constraint IDENT-VOWEL.
IDENT-VOWEL: Any vowel in the input must have an identical correspondent in the output.
IDENT-BASE-FINALV: The final vowel in the Base form of a word must have an identical correspondent in the Derived form of a word.
Crucial Ranking: IDENT-V >> IDENT-BASE
92
In Tableau (37), the above crucial ranking has the effect of eliminating candidate
(b) through a violation of IDENT-VOWEL since word-final /e/ is not part of the
input set {o, a} for the diminutive morpheme. Although candidate (a) violates
IDENT-BASE by selecting word-final /o/ instead of the base conforming /e/, this is
the best option based on the inflectional capacity of the diminutive.
(37) uniforme (m.)
uniform- + DIM {o, a}
IDENT-V IDENT-BASE
a. uniformito * b. uniformite *!
While this constraint ranking prohibits final /e/ in diminutive forms, it
does allow for maintenance of base-form idiosyncrasies that fall within the
possibilities of the diminutive morphemes. The non-canonical final vowel /a/ for
the masculine word poema in (38) is preserved in this analysis. Neither candidate
(a) or (b) incurs a violation of IDENT-V because their respective final vowels are
included in the diminutive set.37 Candidate (b) is judged less optimal since
candidate (a) better preserves the base form.
(38) poema (m.)
poem- + DIM {o, a}
IDENT-V IDENT-BASE
a. poemita b. poemito *!
The above analysis of poema essentially argues that the diminutive
morpheme is not strictly bound to canonical instantiations of gender. For
37 A third possible candidate poemite violates higher-ranked IDENT-VOWEL and has been omitted for purposes of clarity.
93
example this is in contrast to the phrase la modelo bonita “the beautiful fashion
model.” Even though the feminine head of the phrase modelo is marked by
idiosyncratic /o/, both the form of the modifier (bonita) and the definite article
(la) adhere to their feminine forms.
The analysis in (38), however, poses problems for /e/-final words as in
(37), reproduced in (39) with the additional candidate (c), *uniformita. Since {o,
a} is an unordered set and selection of either results in a violation of IDENT-BASE,
there is no mechanism for distinguishing between the optimal candidate (a),
uniformito, and candidate (c) that selects the final vowel /a/.
(39) uniforme (m.)
uniform- + DIM {o, a}
IDENT-V IDENT-BASE
a. uniformito * b. uniformite *! c. uniformita *
Clearly, when faced with a choice of final vowels, the diminutive morpheme
adheres to the canonical values of /o/ for masculine and /a/ for feminine. In fact,
if we take an inventory of the diminutive data, we find, with few exceptions, that
the derived form contains all segments present in the base form:
(40) casa: casita libro: librito madre: madrecita “house” “book” “mother”
padre: padrecito café: cafecito poema (m.): poemita “father” “coffee” “poem”
modelo (f.): modelito pan: panecito balcón: balconcito “fashion model” “bread” “balcony”
94
nov[j]a: nov[j]ecita ladrona: ladroncita “girlfriend” “(female) thief” menú: menucito “menú”
Exceptions:
uniforme: uniformito dinosaur[j]o: dinosaurito “uniform” “dinosaur”
From the standpoint of the recovery of the base, this pattern makes complete
sense. Within highly productive derivational processes, such as diminutivization,
functionality is dependent on the interlocutors’ ability to “agree” on the base form
of the derivation. The most suitable derivation is the one that is best able to
reflect the base form within the confines of the morphology. The existence of
certain exceptional types is just an indication that other variables are present (see
Chapter 4). I propose to formalize this concept with the following constraint:
REFLECT: Segments present in the Base form are reflected in the Derived
form.
Crucially, this constraint makes no reference to position fidelity or to origin of the
segments. For example, the diminutive poemita receives the final vowel /a/ from
the diminutive suffix, but satisfies REFLECT because it reflects the segment /a/
present in the base. As illustrated in Tableau (41), REFLECT makes the decision
between *poemito and poemita. Candidate (b) incurs one fatal violation of
95
REFLECT since the final vowel /a/ of the base is not reflected in the diminutive
form.
(41) poema (m.)
Base: poema poem- + DIM {o, a}
IDENT-V REFLECT
a. poemita b. poemito *! c. poemite *! *
Nevertheless, to this point the analysis is still unable to predict the final vowel of
/e/-final bases (42). While candidate (c) fatally violates IDENT-V by selecting a
final vowel not available to the diminutive morpheme, neither candidate (a) or (b)
violates REFLECT because the /e/ in the base is present in both, albeit not in word-
final position. Again, note that the constraint REFLECT makes no reference to the
origin of the /e/ in the diminutive form; whether it be epenthetic, part of the stem,
or part of the diminutive, its presence is enough to satisfy the constraint.
(42) madre (f.)
Base: madre madr- + DIM {o, a}
IDENT-V REFLECT
a. madrecita b. madrecito c. madrecite *!
In order to distinguish between candidates (a) and (b), I propose the following
constraint:
FEMININE = /a/: Feminine derived forms and only feminine derived forms
are marked by final vowel /a/.
96
Although the masculine of suffix types may end in /o/, /e/, or lack a final vowel
(llorar~llorón “to cry” ~ “crybaby”), this constraint speaks to the generalization
that their feminine counterpart often ends in /a/. In (43), *madrecito is now
eliminated and the correct candidate (a) is predicted.
(43) madre (f.)
Base: madre madr- + DIM {o, a}
IDENT-V REFLECT FEM = /a/
a. madrecita b. madrecito *! c. madrecite *! *
This approach also accounts for the masculine noun, poema, which ends in
unexpected /a/.
(44) poema (m.)
Base: poema poem- + DIM {o, a}
IDENT-V REFLECT FEM = /a/
a. poemita * b. poemito *! c. poemite *! *
FEM = /a/ must be crucially ranked below REFLECT in order to maintain the
correct output for idiosyncratic poema (44) and modelo (45).
(45) modelo (f.)
Base: modelo model- + DIM {o, a}
IDENT-V REFLECT FEM = /a/
a. modelita *! b. modelito * c. modelite *! * *
97
Finally, Tableau (46) illustrates the analysis for uniforme. Neither candidate (a)
or (b) satisfy REFLECT, so the decision is passed down to FEM = /a/. Since
uniforme is a masculine noun, candidate (a) fatally violates this constraint.
(46) uniforme (m.)
Base: uniforme uniform- + DIM {o, a}
IDENT-V REFLECT FEM = /a/
a. unformita * *! b. uniformito * c. uniformite *!
The legitimacy of such an approach is strengthened if it holds up in the
evaluation of similar morphological processes. Returning to the earlier example
of the augmentative –ote/ota, we find that this approach makes the correct
predictions. In the case of the masculine word gato “cat” in Tableau (47),
candidate (a), gatota, is eliminated by FEM = /a/, since gato is masculine.
Candidate (b), gatoto, is eliminated by IDENT-V because the suffix-final [o] is not
part of the available set of final vowels available to this augmentative suffix {e,
a}. Candidate (c) is the winner and the only candidate to satisfy all of the
constraints. Note that all three candidates satisfy REFLECT because the initial
vowel of the augmentative suffix reflects the final vowel of the masculine word
gat/o/.
(47) gato (m.) Base: gato gat- + AUG {e, a}
IDENT-V REFLECT FEM = /a/
a. gatota *! b. gatoto *! c. gatote
98
In Tableau (48), REFLECT is violated by both candidates (b) and (c) since
the final /a/ in gata is not present in the augmented form, a fatal violation in the
case of gatote (c). Again the winning candidate (a), gatota, is the only candidate
able to satisfy all of the constraints.
(48) gata (f.)
Base: gata gat- + AUG {e, a}
IDENT-V REFLECT FEM = /a/
a. gatota b. gatoto *! * * c. gatote *! *
Tableaux (49) and (50) illustrate that this constraint ranking also
successfully accounts for the augmentatives of /e/-final words.
(49) padre (m.) Base: padre padr- + AUG {e, a}
IDENT-V REFLECT FEM = /a/
a. padresota *! b. padresoto *! c. padresote
All candidates for both padre and madre satisfy REFLECT because the final /e/ is
present in all of the diminutive forms. Thus, the selection is left to IDENT-V and
FEM = /a/.
(50) madre (f.) Base: madre madr- + AUG {e, a}
IDENT-V REFLECT FEM = /a/
a. madresota b. madresoto *! * c. madresote *!
99
From the above discussion, it is now clear that the number of diminutive
morphemes can be reduced to three: -it {o, a}, -cit {o, a}, -ecit {o, a}. In Chapter
4, I argue that the [e] is epenthetic, thus reducing the number of morphemes to
two: -it {o, a} and –(e)cit {o, a}. Given their collective presence in both Class D1
and D2 words, it seems prudent to assume that they are available to attach to all
stems and must be evaluated accordingly in the output. Therefore, I will continue
to utilize the shorthand DIM {o, a} to denote all of the possible inputs of the
diminutive morpheme.
This chapter has argued for the following:
1) For the purposes of diminutivization, Spanish non-verbal words are divided
into two classes. Class D1 includes all non-verbal words that end in unstressed
/o/ or /a/. Class D2 includes all other words.
2) All diminutive morphemes are analyzed as suffixes. 3) Diminutive suffixes attach to stems that crucially do not contain the final
vocalic element from the base.
4) The final vocalic element in the output of diminutives is determined by the
interaction of three constraints: a) IDENT-VOWEL limits the final vowel in derived
forms to the inflectional capacity of the diminutive suffix, indicated by {o, a}; b)
REFLECT mandates that derived forms reflect all segments present in the base
form; and, finally, c) FEM = /a/ requires the selection of the final vowel /a/ for
100
feminine derived forms and prohibits its selection for non-feminine derived
forms.
5) All diminutive suffixes are present in the input, as well as an unordered set of
both final vocalic elements: DIM {o, a}.
101
Chapter 4
Morphological and Phonological Interactions in Diminutive Allomorph Selection
I. Introduction
As illustrated in Chapter 3, I assume the ranking IDENT-V >> REFLECT >> FEM =
/a/ is responsible for selection of the final vocalic element in diminutive
formation. In Tableau (1), casita is the winning candidate because it reflects the
final vocalic element /a/ from the base, casa.
(1) casa ~ casita
/kas- + DIM {o, a}/
IDENT-V REFLECT FEM = /a/
a. kasit-a b. kasit-o *! * c. kasit-e *! *
In the following sections, I present an analysis that accounts for the selection of
the diminutive allomorphs themselves. The fact that the –(e)cit{o, a} and the
–it{o, a} allomorphs are both present in Classes D1 and D2 precludes the
possibility of a purely morphological approach in which one allomorph is
associated exclusively with one class. However, I will argue for a role for
morphology within Class D1, based on a novel application of morphological
subcategorization frames (Bonet et al 2007, Bonet 2006, Paster 2005, 2006).
I propose placing a subcategorization frame on Class D1 words that
indicates that they are idiosyncratically marked to select for a particular
diminutive morpheme, specifically –it{o, a}.
102
(2) Class D1/it/ casa /kasit/ “house” palo /palit/ “stick” problema /problemit/ “problem” modelo /modelit/ “model” No other subcategorization frames are required and all other patterns are
determined by the ranking of phonological constraints, resulting in no loss of
generalizations.
As with Bonet’s analysis of class marker allomorphy, the constraint
RESPECT will be violated whenever a member of Class D1 chooses a diminutive
other than –it{o, a}. However, I leave the diminutive allomorphs as an unordered
set and, as a result, obviate the need for the PRIORITY constraint. Instead,
RESPECT interacts with universal faithfulness and markedness constraints to select
the correct output. Crucially, the present analysis places limitations not on the
–it{o, a} allomorph itself, but rather on Class D1, which exhibits a strong (non-
optimizing) preference for selection of –it{o, a}. In this way, -it{o, a} is not
constrained and, indeed, does attach to words outside of Class D1, for example
uniforme ~ uniformito.
II. Analysis of Diminutive Class D1 Words in Sonoran
As stated in the conclusion to the previous chapter, one of the principle
assumptions in my analysis is that all diminutive allomorphs are present
underlyingly. Thus, DIM represents the unordered set {it (o, a), cit (o, a)}. I view
the process as an example of suppletive allomorphy (see Paster 2006 for
103
discussion) in which one form is not related through phonological processes to the
other. In other words, -it{o, a} is not the result of deletion of the first segment of
–cit{o, a} and, conversely, -cit{o, a} is not an augmented form of –it{o, a}.
Therefore, beginning with the Class D1 word, casa, it is necessary to determine
what forces are involved in the selection between the two candidates casita and
*casecita.
I will adopt Bonet’s (2006) approach to Spanish gender allomorphy, in
which she established an ordered set of class markers for masculine and feminine
nouns based on markedness.
(3) Masculine: {o > e, ø > a} Feminine: {a > e, ø > o} For masculine nouns, /o/ is the canonical gender marker and is ordered highest, /e/
and words without a gender marker (ø) are equally more marked, and the class
marker /a/ is analyzed as most marked and ordered last. Feminine nouns are
similarly ranked, with /o/ and /a/ inverted. Those words that do not end in the
canonical marker for their respective gender receive a subcategorization frame
that indicates this idiosyncratic selection.
(4) Masculine: nombre /nombre/ “name” pan /panø/ “bread” problema /problema/ “problem”
104
Feminine: noche /noche/ “night” flor /florø/ “flower” modelo /modelo/ “model” Bonet appeals to two constraints, RESPECT (Bonet et al 2007) and PRIORITY
(Bonet et al 2003, 2005; Mascaró 2007), to account for the allomorphy (Bonet
2006: 327).
RESPECT: Respect idiosyncratic lexical specifications. In an example from (4), /nombre/, RESPECT requires that the idiosyncratic lexical
information present in the subcategorization frame /e/ be present in the output.
PRIORITY: Respect lexical priority (ordering) of allomorphs. Based on the ranking for masculine words {o > e, ø > a}, PRIORITY states that
those words that end in /e/ or ø will incur one violation mark, and those that end
in /a/ will incur two violations. The constraint ranking RESPECT >> PRIORITY
ensures that idiosyncratic markings are present in the surface form.
(5) nombre (masc.)
/nombre + {o > e, ø, > a}/
RESPECT PRIORITY
a. nombro *! b. nombre * c. nombr *! * d. nombra *! **
Candidate (5a) satisfies PRIORITY through selection of the highest ordered marker
/o/, but at the expense of a fatal violation of RESPECT since nombre is lexically
105
specified to end in /e/. Candidates (5c) and (5d) fatally violate RESPECT for the
same reason, in addition to one and two PRIORITY violations respectively for
selection of markers ordered below /o/. Candidate (5b) also incurs one violation
of PRIORITY, but is able to satisfy highest ranked RESPECT via selection of the
lexeme’s idiosyncratic marker /e/.
While I do not adopt this analysis to account for surfacing class markers in
diminutive forms (as stated in Chapter 3, I view the “class markers” as
inflectional endings that are part of the diminutive allomorph), I do propose that a
lexical specification approach is able to capture the non-optimizing nature of
certain diminutives. Specifically, I propose that Class D1 words (those ending in
canonical unstressed /o/ or /a/) are lexically specified to select –it{o, a}.
(6) Class D1/it/ casa /kasit/ “house” palo /palit/ “stick” problema /problemit/ “problem” modelo /modelit/ “model”
The ranking P >> M in Prosodic Morphology (McCarthy and Prince
1993a, b) argues that some phonological constraint P must outrank some
morphological constraint M; otherwise, one must abandon the influence of
prosody (and phonology more generally) on allomorphy. In view of their
constraint PARSE-SYLL (McCarthy & Prince 1993a, b), [ka.se.si.ta] is clearly the
least-marked of the two candidates.
PARSE-SYLL: All syllables must be parsed into feet.
106
(7) casa ~ casita
/kas- + DIM {o, a}/
PARSE-SYLL
a. ka (sí.ta) *! b. (kà.se) (sí.ta)
The first syllable of candidate (7a) is left unparsed, a violation of PARSE-SYLL,
and unattested *casecita is wrongly predicted. From a purely phonological
standpoint, there is no reason to believe that the form *[ka.se.si.ta] is less well
formed than [ka.si.ta]; on the contrary, the former is, at least, prosodically better
formed than the latter.
My use of the subcategorization frame /it/ and the constraint RESPECT
provide a mechanism to isolate the optimal candidate. The crucial ranking of
RESPECT >> PARSE-SYLL in Tableau (8) illustrates the fact that it is more critical
to respect the lexical specification that Class D1 words select –it{o, a} than to
satisfy the prosodic constraint that dictates parsing of all syllables. Again, there
seems to be no phonological motivation for selection of casita over casecita,
therefore some other mechanism must account for the intuitively clear preference
for Class D1 words to select –it{o, a}.
(8) casa ~ casita
/kasit + DIM {o, a}/
RESPECT PARSE-SYLL
a. ka (sí.ta) * b. (kà.se) (sí.ta) *!
One alternative to lexical specification is to establish morpheme-specific
constraints. This is the approach that Colina (2003) proposes for Spanish
diminutives; but, as Bonet et al (2007: 915) point out, this opens up the possibility
107
of positing as many constraints as there are morphemes in a language. Of course,
it would be difficult to argue for the universal status of constraints that refer to
specific morphemes in specific languages.
As I mentioned in my discussion in Chapter 3, examination of
Crowhurst’s Sonoran data indicates that Class D1 words select for the –it{o, a}
allomorph except when such a selection would create an insurmountable
phonological conflict. Such a conflict arises with words whose stems end in a
front glide [j], such as nov[j]-a “girlfriend” and pat[j]-o “patio;” the resultant
homorganicity underscores the one limitation of the –it{o, a} allomorph. A never
violated OCP constraint in Spanish on tautosyllabic homorganic vocoids in rhyme
position (Harris 1983) (i.e. *.Cji. or *.Cwu.) overrides the requirement to respect
the lexical specification on these words to select –it{o, a}.
(9) nov[j]- + -it{o, a} = *[no. ßjí. ta] pat[j]- + -it{o, a} = *[pa.tjí.to] However, the constraints introduced to this point are insufficient to account for
the correct output. While *.Cji. eliminates candidate (10a), both candidates (10c)
and (10d) satisfy the constraint ranking better than the actual output in (10b), in
which (nò.ßjé) (sí.ta), fatally violates high-ranking RESPECT, since novia is a
Class D1 word lexically specified to select the –it{o, a} allomorph.
108
(10) novia ~ noviecita /nobj-it + DIM {o, a}/
*.Cji. RESPECT PARSE-SYLL DEP-V
a. no (ßjí.ta) *! * b. (nò.ßje) (sí.ta) *! * c. no (ßí.ta) *! d. (nò.ßi) (í.ta)
Candidates (10b-10d) reflect three possible repair strategies to avoid the Cji
structure. Candidate (10b) selects the –ecit{o, a} allomorph, which allows it to
maximally reflect the base through maintenance of the glide in the diminutive
form. Candidate (10c) satisfies the constraint by simply deleting the glide and
selecting –it{o, a}. Meanwhile, candidate (10d) avoids the illicit structure through
conversion of the glide into a syllabic vocoid [i].
In Tableau (11) I introduce MAX, an input-to-output faithfulness
constraint, IDENT (SYLL) an output-to-output faithfulness constraint (also used in
Colina’s 2003 and Stephenson’s 2004 accounts of Spanish diminutives), and
ONSET, a markedness constraint on syllable structure:
MAX: Every segment S1 present in the input must have a corresponding segment
S2 in the output.
IDENT (SYLL): Non-syllabic vocoids in the input must remain non-syllabic in the
output. Syllabic vocoids in the input must remain syllabic in the output.
ONSET: All syllables have an onset.
109
(11) novia ~ noviecita /nobj-it + DIM {o, a}/
RESPECT MAX IDENT-SYLL
ONSET PARSE SYLL
DEP-V
a. (nò.ßje) (sí.ta) * * b. no (ßí.ta) * * c. (nò.ßi) (í.ta) * *
I claim that the top three constraints are crucially unranked as well as the bottom
three constraints: RESPECT, MAX, IDENT-SYLL >> ONSET, PARSE SYLL, DEP-V.
The state of affairs in Tableau (11), in which each candidate appears to be
equally flawed, is not terribly surprising given the type of variation that Harris
(1994: 185) reports for the diminutive forms of disyllabic words ending in the
diphthong [jo] (discussed in Chapter 3).
It is instructive to test the current constraint ranking on the behavior of
longer than disyllabic words that end in the troublesome diphthong, particularly
since no variation has been reported for them. In Tableau (12), the current
constraint ranking predicts the correct output for iglesia. Although candidate
(12a), *iglesiecita, preserves the glide through selection of the –(e)cit{o, a}
allomorph, it comes at the cost of creating an unparsed syllable. Compare this
result with that of noviecita in (11a). In this case, selection of –(e)cit{o, a}
affords the advantage of forming a prosodically unmarked word of two trochaic
feet, (`σ σ) (σ´ σ) ~ (nò.βje) (sí.to). Interestingly, the winning candidate in (12)
shares the same prosodic structure (ì.gle) (sí.ta).
(12) iglesia ~ iglesita “church”
/iglesj-it + DIM {o, a}/
RESPECT MAX IDENT-SYLL
ONSET PARSE SYLL
DEP-V
a. (i.gle) sje (sí.ta) * * *(!) *(!) b. (i.gle) (sí.ta) * * c. (i.gle) si (í.ta) * **(!) *(!)
110
The loss of both the glide and the segment in the winning candidate (12b) is
somewhat surprising given the data reported by Harris in (13). Of course, one
must take care given the small sample size, but it is worth noting that the least
accepted repair strategy (both glide and segment lost) for these disyllabic words
is, conversely, the chosen strategy for longer than disyllabic words (iglesita).
(13)
Lexeme Glide loss (segment retained)
Glide maintained Glide loss (segment loss)
ind[j]o ind[i]to “Indian” sab[j]o sab[je]cito “wise” lab[j]o lab[i.í]to lab[je]cito ?lab[i]to “lip” ap[j]o ap[i.í]to ap[je]cito “celery”
ampl[j]o ampl[i.í]to “ample” This does make intuitive sense in that segmental loss should have a greater impact
on shorter words than longer words in terms of recovery of the base. For
example, if for the word patio “patio” speakers were to choose the segmental loss
repair strategy, resulting in *patito instead of attested patiecito, then there would
exist an unnecessary merger with the diminutive for pato “duck,” patito. On the
other hand, segmental loss between pairs such as iglesia ~ iglesita and dinosaurio
~ dinosaurito are less costly and the respective diminutives unlikely to elicit
confusion as to the base to which they refer.
More to the point, however, I concur with Colina (2003) that the tendency
of certain diminutive forms to emerge as two trochaic feet suggests a process of
the Emergence of the Unmarked (TETU). The divergent repair strategies of
iglesia and the disyllabics in (13) funnel into this prosodic structure. In this way,
111
the incompatibility of the stem-final glide [j] and the diminutive suffix initial [i]
provides the environment for an unmarked prosodic structure to emerge as
speakers settle on repair strategies. I formalize this process through two
constraints: ALIGN-FT-L and ALIGN-FT-R (McCarthy and Prince 1993).
ALIGN-FT-L (FT, L, PRWD, L): Align the left edge of a foot with the left edge of
the prosodic word.
ALIGN-FT-R (FT, R, PRWD, R): Align the right edge of a foot with the right edge
of the prosodic word.
Given that the process of TETU expresses a preference for an unmarked structure
that is essentially latent in the grammar and only emerges under certain
circumstances, the above constraints must be low ranking. Finally, I integrate a
markedness constraint on identical adjacent segments: OCP (Colina 2009).
OCP: No identical adjacent segments. (14) novia ~ noviecita
/nobj-it + DIM {o, a}/
RES
PECT
MA
X
IDEN
T-SY
LL
ON
SET
PARS
E SY
LL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a.(nò.ßje) (sí.ta) * * b. no (ßí.ta) * * *! c. (nò.ßi) (í.ta) * * *!
The free variation present in (13) is admittedly difficult to formalize in a
framework such as OT, but it does serve to justify the peripheral status of the
OCP constraint. It also provides evidence for crucial unranking of the top three
112
constraints. As will become clear in the presentation of the analysis of other
words types, these three constraints do not interact to decide the optimal candidate
except within the context of the stem-final [j] words, precisely where we
encounter variation and multiple repair strategies. For example, crucially ranking
RESPECT above MAX and IDENT-SYLL would only serve to eliminate entirely the
selection of –cit{o, a} for noviecita and similar words (a move not justified by
their appearance in the data), but would have absolutely no impact on evaluation
of all other word types.
The clash between the homorganic vocoids can only be resolved through
the selection of one of three violations on faithfulness. A close comparison of the
tableaux for iglesia and novia (15) reveals that the number of syllables in their
respective stems inadvertently plays a role in their evaluation. The fact that
iglesia has one more syllable than novia ensures that the glide deletion repair
strategy for iglesia (15) results in all syllables parsed, but does not have the same
result for novia (15). The area shaded gray for iglesia indicates that no further
evaluation of candidates is relevant, whereas the matter is not resolved for novia.
(15) iglesia ~ iglesita / novia ~ noviecita
/iglesj-it + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (ì.gle) sje (sí.ta) * * *(!) *(!) b. (ì.gle) (sí.ta) * * c. (ì.gle) si (í.ta) * **(!) *(!) * /nobj-it + DIM {o, a}/ d.(nò.ßje) (sí.ta) * * e. no (ßí.ta) * * *! f. (nò.ßi) (í.ta) * * *!
113
III. Analysis of Diminutive Class D2 Words in Sonoran
Tableau (16) compares the diminutive formation of the minimal pair
madre / comadre. Since Class D2 words are not lexically bound to any particular
allomorph, they are free to choose the option that results in the least-marked
structure. As with the novia/iglesia pair, we see that the winning candidates
converge into the unmarked structure of two trochaic feet: (mà.dre) (sí.ta) and
(kò.ma) (drí.ta). In both cases, the eliminated candidates are phonologically
sound (certainly madrita (16a) should fare no worse than casita) and are only
slightly more marked than their optimal counterparts. This winning pair provides
the clearest evidence that /e/-final bases have no allomorphic preference and that
both –it{o, a} and –(e)cit{o, a} are available underlying options.
(16) madre~madrecita / comadre~comadrita
/madr- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. ma (drí.ta) * *! b. (mà.dre) (sí.ta) * /komadr- + DIM {o, a}/ c. (kò.ma) (drí.ta) d. ko (mà.dre) (sí.ta) *(!) *(!) * e. (kò.ma) dre (sí.ta) *(!) *(!)
The next type to be addressed includes those words that end in a stressed
final vowel (e.g. menú, café, matiné). These words differ from previous types we
have seen because the final vowel carries the main stress. In an analysis of
Spanish truncated forms, Piñeros (2000a, b) argues convincingly that truncated
forms are sensitive to the foot structure of the source form. Of most relevance for
the purposes of our current discussion is his assertion that vocalic segments in the
114
prosodic head of the source form (the stressed syllable) are maximized in the
truncated form.
(17) Spanish hypocoristic exemplars (Piñeros 2000a: 14) a. Final syllable stress [be.a.(trís)]: [(tí.ča)] Beatríz [se.ßas. (tján)]: [(ča.no)] Sebastián
b. Penultimate syllable stress [kar. (ló.ta)]: [(ló.ta)] Carlota [ar. (mán.do)]: [(mán.do)] Armando
c. Antepenultimate stress
[i. (pó.li). to]: [(pó.lo)] Hipólito
[(mé.li). ða]: [(mé.la)] Mélida
While it is clear from the examples in (17a-c) that not all segments in the prosodic
head are consistently and faithfully reproduced in the truncated form (see Piñeros
2000a for full discussion of this issue; also see Alderete 1995 for prosodic head
correspondence), the examples also make evident that the stressed vowel is
always faithfully maintained in hypocoristics. This is formalized through the
faithfulness constraint HEAD-MAX (Piñeros 2000a: 14).
HEAD-MAX: Every segment contained in a prosodic head S1 must have a
correspondent in S2.
115
A key difference between truncation and diminutivization is that in Spanish
hypocoristics the stressed vowel in the source form is also the stressed vowel in
the truncated form, [ar.mán.do] : [mán.do], but this is not the case in
diminutivization because main stress is always carried by the diminutive
allomorph itself. Crucially, HEAD-MAX legislates only that the segments be
maintained in the input without respect to faithfulness to their prosodic role. With
respect to diminutivization of Spanish words ending in stressed vowels, this
amounts to complete Base-Stem identity.
(18) Base Stem me.nú [me.nu]- ka.fé [ka.fe]- ma.ti.né [ma.ti.ne]- In this way, HEAD-MAX is essentially a subset of already high ranked MAX that
prohibits segmental deletion in the stem; therefore MAX will continue to apply to
any deletion from the input. Candidates (a) and (b) in Tableau (19) fatally violate
this constraint. Although candidate (b) conforms to the unmarked prosodic
structure of exactly two trochaic feet, it does so at the expense of deletion of a
stem segment. Candidates (c) and (d) also conform to the structure, but without
segmental deletion. Notice that in this analysis the winning candidate (d) selects
–cit{o, a} not out of some morphological predisposition on the part of either the
allomorph or the word type, but because the alternative selected by candidate (c),
-it{o, a}, results in an onsetless syllable.
116
(19) menú ~ menucito
/menu- + DIM {o, a}/
MAX
ONSET PARSE-SYLL DEP-V
a. me (ní.to) *! * b. (mè.ne) (sí.to) *! * c. (mè.nu) (í.to) *! d. (mè.nu) (sí.to) e. (mè.nu) e (sí.to) *(!) *(!) *(!)
Tableau (20) underscores this near parity in terms of the candidates’ optimality. (20) menú ~ menucito
/menu- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (mè.nu) (í.to) *! b. (mè.nu) (sí.to)
In the case of trisyllabic matiné, high-ranking MAX disallows deletion of
stem-final [e] and, thus, prevents the diminutive form from conforming to the
unmarked prosodic structure. Unlike with iglesia, the phonological conditions are
not severe enough to justify such deletion. Candidates (21a) and (21b) are
prosodically unmarked and satisfy all low-ranking constraints, but it comes at too
high a price. Again, the ONSET constraint alone is responsible for the decision
between the –it{o, a} and –cit{o, a} options in candidates (21c) and (21d).
117
(21) matiné ~ matinecito /matine- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (mà.ti) (ní.to) *! b. (mà.tin) (sí.to) *! c. (mà.ti) ne (sí.to) * d. (mà.ti) ne (í.to) *(!) *(!)
Under this analysis, it now becomes clear why the final /e/ in words such as
madre and comadre cannot be analyzed as forming part of the stem. If we
compare the behavior of trisyllabic /e/-final matiné with trisyllabic /e/-final
comadre, we see that such an assumption could never predict comadrita. High-
ranking MAX would eliminate candidate (22d). If stem-final vowels were
expendable (suggesting a much lower ranking for MAX), then we would expect
that the prosodically unmarked candidate (22a), *matinito, would emerge as a
winning candidate. Therefore, as argued in Chapter 3, segmental similarity
between the final /e/ in madre and its diminutive form madrecita is coincidental
and not indicative of any attachment preference on the part of –cit{o, a}.
(22) /matine- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (mà.ti) (ní.to) *! b. (mà.ti) ne (sí.to) * c. (mà.ti) ne (í.to) *(!) *(!) /komadre- + DIM {o, a}/ d. (kò.ma) (drí.ta) *! e. (kò.ma) dre (sí.ta) * f. (kò.ma) dre (ít.a) *(!) *(!)
118
Further evidence in this dialect that the [e] is epenthetic and not part of the stem is
found in careful observation of the diminutivization of consonant-final bases, as
in (23). In the case of monosyllabic pan, the [e] is absent in the base, but surfaces
in the diminutive—a clear case of epenthesis.
(23) pan panecito “bread” balcón balconcito “balcony” corazón corazoncito “heart” Not coincidentally, the epenthetic [e] has the effect of creating the same
unmarked structure (`σ σ) (σ´ σ) that we have seen in many of the winning
candidates. I assume that [pan-] is an analogous stem to [madr-] and that the
grammar treats them in the same way. There is nothing inherently “wrong” with
the non-epenthetic candidates; it is simply that the grammar is such that
epenthesis is a relatively low cost procedure if the benefit means a less marked
prosodic structure. In Tableau (24), candidates (c) and (e) win out because all
syllables are parsed and the resulting unmarked structure of two trochaic feet
emerges.
(24) /pan- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-
SYLL
O
NSE
T
PAR
SE
SYLL
D
EP-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. pa (ní.to) * *! b. pan (sí.to) * *! c. (pà.ne) (sí.to) * madr- + DIM {o, a} d. ma (drí.ta) * *! e. (mà.dre) (sí.ta) *
119
As we have seen before, epenthesis is not inherent to any particular word type.
Tableau (25) evaluates the disyllabic, consonant-final word balcón. Simply being
consonant-final is not sufficient to trigger epenthesis. The epenthetic candidate
(25c) actually creates a more marked structure with one syllable left unparsed.
(25) balcón ~ balconcito
/balkon- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-
SYLL
O
NSE
T
PAR
SE
SYLL
D
EP-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (bàl.ko) (ní.to) b. (bàl.kon) (sí.to) c. (bàl.ko) ne (sí.to) *(!) *(!)
However, the current constraint hierarchy is insufficient to distinguish between
candidates (25a) and (25b). In terms of segmental preservation, the two
candidates are equally faithful to the corresponding base balcón and, therefore,
neither violates MAX or DEP-V. In addition, both have the unmarked prosodic
structure with all syllables parsed. What does distinguish them concerns the
syllabic role of the stem-final consonant [n]. Candidate (25b) is able to maintain
the coda position of the base correspondent because the –cit{o, a} allomorph is
consonant initial. Conversely, stem-final [n] in candidate (25a) must be
resyllabified as an onset to the vowel-initial suffix –it{o, a}. This syllabic role
correspondence is governed by the constraint STRUCTURAL ROLE (McCarthy and
Prince 1993a:141; for use with Spanish diminutive analyses see Agüero-Bautista
1998, Elordieta & Carreira 1996, Miranda 1999, and Stephenson 2004).
STRUCROLE: Corresponding elements have identical syllabic roles.
120
(26) balcón ~ balconcito
/balkon- + DIM {o, a}/
RES
PECT
MA
X
IDEN
T-SY
LL
ON
SET
PARS
E SY
LL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
STRU
C R
OLE
a. (bàl.ko) (ní.to) *! b. (bàl.kon) (sí.to) c. (bàl.ko) ne (sí.to) *(!) *(!)
STRUCROLE is fatally violated by (26a) and balconcito emerges the winner.
This constraint has important implications for trisyllabic consonant-final
words, such as corazón. Whereas the disyllabic balcón has an inherent prosodic
advantage in that attachment of either of the disyllabic diminutive allomorphs
results in all syllables parsed and an unmarked prosodic structure, this is not the
case for trisyllabics. Unless an epenthetic [e] is added to –cit{o, a}, neither of the
allomorphs creates the unmarked structure.
(27) corazón ~ corazoncito
/korason- + DIM {o, a}/
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
STR
UC
R
OLE
a. (kò.ra) so (ní.to) * *! b. (kò.ra) son (sí.to) * c.(ko.ra) (so.ne) (sí.to) * *!
Both candidates (27a) and (27b) leave one syllable unparsed, but the equal
ranking of DEP-V and PARSE-SYLL moves the decision farther down the
constraint ranking. If DEP-V were crucially ranked below PARSE-SYLL, then we
would expect that an epenthetic [e] to always be added in order to have all
121
syllables parsed regardless of the impact lower on the constraint ranking.
Conversely, if DEP-V were ranked above PARSE-SYLL, then an epenthetic [e]
would never be inserted to improve prosodic structure (wrongly predicting
*madrita, etc.). Candidate (27c) satisfies PARSE-SYLL through epenthesis, but
this causes a detrimental effect by changing the structural role of [n] without a
concomitant improvement in prosodic structure (i.e. the ALIGN constraints).
In the case of pan (28), the epenthetic candidate, panecito, has a positive
impact on prosodic structure, but satisfies markedness at the expense of
faithfulness to the syllabic role of the final segment. This competition provides a
crucial argument for the ranking of STRUCROLE below ALIGN-FT-L in panecito
dialects.38
(28) pan ~ panecito
/pan- + DIM {o, a}/
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
STR
UC
R
OLE
a. pa (ní.to) * *! * b. pan (sí.to) * *! c. (pà.ne) (sí.to) * *
IV. Additional Support for High-Ranking RESPECT Constraint
Given its high ranking and elevated importance in my analysis, it is worth
presenting additional data that support the claim that in Sonoran only never
violated constraints (e.g. *Cji) are sufficient justification for violation of RESPECT.
38 As we will see later in this chapter, constraint re-ranking will account for pancito dialects.
122
Crowhurst’s data set provides an interesting contrast between two
disyllabic word types (1992: 242): /e/-final words with an initial alternating
diphthong (29a) and /o/- or /a/-final words with an initial alternating diphthong
(29b).
(29) a. diente > dientecito, dientito “tooth” liendre > liendrecito, liendrito “nit” b. miedo > miedito, *miedecito “fear” cielo > cielito, *cielecito “sky” In set (29a), Crowhurst reports free variation between the two allomorphs, yet in
(29b) selection of the –cit{o, a} allomorph is unacceptable. The common feature
shared by the two sets is the diphthong [je] in the first syllable. Spanish exhibits
an alternation grounded in a historical process in which certain stressed vowels
(/o/ and /e/) diphthongized while etymologically related words with differing
stress patterns did not (see Hualde 2005 chapters 5 and 12 for discussion).
(30) [djén.te] “tooth” [den.tál] “dental” [ljén.dre] “nit” [len.dró.so] “nit-ridden” [mjé.do] “fear” [me.dró.so] “fearful” [sjé.lo] “sky” [se.lés.te] “celestial” In the data from the peninsular dialect analyzed by Colina (2003), disyllabic
words with a diphthong (alternating or not) select the –cit{o, a} allomorph without
123
regard to whether the base ends in /e/ or canonical /o/ or /a/39 (e.g. suave ~
suavecito “soft”, piedra~piedrecita “stone”). One group of words that Colina
does not address, however, is trisyllabic words with a historical alternating
diphthong (31).
(31) abuelo > abuelito, *abuelecito “grandfather” huérfano > huerfanito, *huerfanecito “orphan” escuela > escuelita, *escuelecita “school” abuelo alternates with the word abolengo “ancestry, lineage;” huérfano with
orfanato “orphanage;” and escuela with escolar “school (adj.).” The unattested
selection of –cit{o, a} for these words suggests that simply containing an
alternating diphthong is not a sufficient conditioning factor for –cit{o, a}
selection. The difference between the disyllabics in (30) and the trisyllabics in
(31) is that selection of the –cit{o, a} in the former results in the prosodically
unmarked structure of two trochaic feet, whereas in the former it does not,
*(à.ßwe) le (sí.to).
Colina (2003: 71-72) argues that the selection of –cit{o, a} for these words
in Peninsular responds principally to a pair of constraints referring to stress,
PARSE-DIPHTHONG and FTFORM (TROCHEE).
PARSE (DIP): Diphthongs must be parsed by feet.
FTFORM (TROC): Feet are trochaic; align the left edge of a foot with the
left edge of its head (a stressed syllable).
39 Dialectal variation will be addressed in the relevant section later in this chapter.
124
Without reviewing Colina’s entire constraint set and ranking, Tableau (32)
sketches how these high-ranking constraints eliminate certain candidates. The
winning candidate, piedrecita, satisfies both PARSE-DIPHTHONG and FTFORM
(TROCHEE).
(32) piedra~piedrecita “stone”
PAR
SE
DIP
HTH
ON
G
FTFO
RM
(T
RO
CH
EE)
a. (pje.drí) ta *! b. pje (drí.ta) *! c. (pjè.dre) (θí.ta)
I adopt these constraints in order to account for this variation, which is the
principle distinction between the Sonoran and Peninsular dialects. While I
maintain the constraint FTFORM (TROCHEE) as high-ranking (above even
RESPECT), I propose that PARSE-DIPHTHONG is lower ranking in Peninsular and on
par with RESPECT, essentially canceling each other out. Candidate (33b), then, is
eliminated by ALIGN-FT-L; the claim being that candidate selection is based on
the adherence of candidate (33c) to the unmarked prosodic structure of two
trochaic feet.40
40 An alternative approach entails removing the subcategorization frame from words that contain diphthongs in this Peninsular dialect. This move would mean that all candidates would vacuously satisfy RESPECT and piedrecita would still emerge the winner, with PARSE-DIPHTHONG as the deciding constraint.
125
(33) piedra ~ piedrecita (Peninsular) /pjedr-it + DIM {o, a}/
FTFO
RM
(TRO
CHEE
)
RES
PECT
PARS
E D
IPH
THO
NG
MA
X
IDEN
T-SY
LL
ON
SET
PARS
E SY
LL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
a. (pje.drí) ta *! * * b. pje (drí.ta) * * *! c. (pjè.dre) (θí.ta) * *
Meanwhile, in the Sonoran dialect (34), I argue that PARSE-DIPHTHONG is
ranked lower than RESPECT.41 Fatal violation of RESPECT removes the possibility
of –cit{o, a} selection and, therefore, piedrita (34b) is the winning candidate. I
submit that from the Peninsular to the Sonoran dialect there is an increasingly
prominent role for RESPECT; in other words, the lexical bond between –it{o, a}
and Class D1 words has grown stronger in Sonoran.
(34) piedra ~ piedrita (Sonoran)
/pjedr-it + DIM {o, a}/
FTFO
RM
(TRO
CHEE
)
RES
PECT
MA
X
IDEN
T SY
LL
PAR
SE
DIP
HTH
ON
G
ON
SET
PARS
E S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
a. (pje.drí) ta *! * * b. pje (drí.ta) * * * c. (pjè.dre) (sí.ta) *! *
The other issue that is raised by the examples in (29) concerns the variation
reported for /e/-final disyllabic words that contain the diphthong [je] in the first
syllable. Crowhurst (1992: 244) notes that there is inter-speaker variation in the
acceptability of these types, with one speaker, for example, finding both dientito
41 It is quite possible that the Peninsular dialect is more conservative in terms of diminutive formation. In this case dialectal variation could be the result of a diachronic process in which once high ranked constraint(s) gradually move down the constraint hierarchy until they are essentially “deactivated.”
126
and dientecito acceptable yet claiming only clientecita as an option, and yet
another speaker accepts clientita but rejects dientito.
This report by Crowhurst again speaks to the level of variation across
dialects and across speakers, and perhaps suggests a process of lexical diffusion in
which the behavior of certain diminutive words begins to impact that of others.42
(35) Stage A Stage B
miedecito → miedito/miedecito dientecito dientecito (North-Cent. Penin.)
Stage C Stage D → miedito → miedito dientecito dientecito/dientito (Sonoran) Stage E
→ miedito dientito
(????)
For example, as in (35), the North Central Peninsular dialect reflects a more
conservative stage (A) in which parsing of diphthongs matters more than the
morphological bond between Class D1 diminutives and –it{o, a}. Whereas
Sonoran represents a later stage (D) where this bond has usurped the requirement
that diphthongs be parsed in Class D1 words and is even beginning to exert an
influence on Class D2 words.
42 Stephenson (2004: 39) makes a similar suggestion based in her two-system account of diminutive formation. However, she proposes that Peninsular is the innovative dialect in this respect, whereas my sketch in (35) assumes it is conservative. I echo her sentiment that investigation of the historical record is needed.
127
If this proposal has merit, then it might be formalized in the following manner.
As the bond between –ito{o, a} and Class D1 strengthens (i.e. RESPECT), it comes
at the cost of a progressively lower ranking for PARSE-DIPHTHONG in the
grammar. Consequently, /o/ and /a/ final words with a diphthong in penultimate
syllable position increasingly select –it{o, a}. This in turn reduces the overall
number of word types and tokens that maintain the unmarked prosodic structure
(mjè.de) (sí.to) → mje (dí.to). The high token frequency of /o/ and /a/-final words
that select –it{o, a} eventually absorb the diphthong containing words into their
ranks. The resultant lower ranking of PARSE-DIPHTHONG begins to have an
impact on /e/-final words that contain diphthongs in the penultimate syllable.
Although Crowhurst does not address diminutive variation among other
diphthongs, the fact that the variation presented occurs among words that all share
the same diphthong [je] indicates that the proper conditions exist for lexical
diffusion through analogical association.
Consider the speaker in Crowhurst’s article that only accepts
diente~dientito (*dientecito) and cliente~clientecita (*clientita43). It is difficult to
imagine a phonologically principled explanation as to why this minimal pair
should be treated in such a disparate manner. Therefore, it is reasonable to
assume that for this speaker there is a subcategorization frame for diente that
lexically specifies selection of –it{o, a}, as with the Class D1 words. The speaker
would maintain no such frame for cliente.
43 This distribution is also a bit puzzling, because the possibility exists that the speaker could parse the base as the common variant containing the canonical feminine final vowel /a/ in la clienta rather than la cliente, thus opening the door for selection of the –it{o, a} alternative.
128
In Tableau (36), candidate (b) violates RESPECT, which essentially
removes the possibility of the lower ranked constraints influencing the matter.
The lack of the subcategorization frame for cliente allows for the prosodically
unmarked candidate to emerge as the winner.
(36) diente ~ dientito / cliente ~ clientecita
/djent-it + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T SY
LL
PAR
SE
DIP
HTH
ON
G
ON
SET
PAR
SE
SY
LL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
a. djen (tí.to) * * * b. (djèn. te) (sí.to) *! * /kljent- + DIM {o, a}/ c. kljen (tí.ta) *! * * d. (kljèn.te) (sí.ta) *
V. Dialectal Variation in Paraguayan and Nicaraguan Spanish
Dialectal variation: pan ~ panecito/pancito
In contrast to the reported data for the Peninsular and Sonoran dialects,
both Jaeggli (1978) for Paraguayan and Miranda (1999) for Nicaraguan report
that consonant-final monosyllabic words select –cit{o, a} rather than the
epenthetic variant –ecit{o, a}. For convenience, Tableau (37) presents the
analysis for pan ~ panecito in Sonoran from (28) above. The critical ranking is
ALIGN-FT-L >> STRUCROLE.
129
(37) pan ~ panecito (Sonoran) /pan- + DIM {o, a}/
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
STR
UC
R
OLE
a. pa (ní.to) * *! * b. pan (sí.to) * *! c. (pà.ne) (sí.to) * *
The pancito variant in Nicaraguan and Paraguayan can easily be captured by a
simple re-ranking of these two constraints: STRUCROLE >> ALIGN-FT-L.
Selection of consonant initial –cit{o, a} by candidate (38b) allows the base-final
[n] to maintain its coda position and thus satisfy higher ranking STRUCROLE.
Since STRUCROLE is now ranked above ALIGN-FT-L, candidate (38c) affords no
advantage by adhering to the unmarked prosodic structure. The epenthetic [e]
forces the final [n] into onset position and a fatal violation of STRUCROLE.
(38) pan ~ pancito (Paraguayan, Nicaraguan)
/pan- + DIM {o, a}/
ON
SET
PAR
SE S
YLL
DEP
-V
STR
UC
R
OL
E
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. pa (ní.to) * *! * b. pan (sí.to) * * c. (pà.ne) (sí.to) * *!
Disyllabic or longer consonant-final words in the Paraguayan and Nicaraguan
dialects form their diminutives as with the Sonoran dialect. Re-ranking of
STRUCROLE to capture the monosyllabic pattern has no impact on optimal
candidate selection for these lengthier words; the correct candidates are predicted
in (39).
130
(39) balcón ~ balconcito / corazón ~ corazoncito (Paraguayan, Nicaraguan) /balkon- + DIM {o, a}/
ON
SET
PAR
SE S
YLL
DEP
-V
STR
UC
R
OLE
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (bàl.ko) (ní.to) *! b. (bàl.kon) (sí.to) c. (bàl.ko) ne (sí.to) *(!) *(!) * /korason- + DIM {o, a}/ a. (kò.ra) so (ní.to) * *! b. (kò.ra) son (sí.to) * c. (ko.ra) (so.ne) (sí.to) * *!
Dialectal variation: tap[j]a ~ tapita Pattern in Nicaraguan
The current constraint ranking is unable to account for an additional
variant pattern in Nicaraguan as reported by Miranda (1999). In this dialect, all
words (including disyllabics) that end in the diphthong [jo] or [ja] select the –it{o,
a} allomorph. The constraint ranking established for Sonoran wrongly predicts
*tapiecita (40a) in Nicaraguan.
(40) tapia ~ tapita “mud wall” (Nicaraguan)
/tapj-it + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a.(tà.pje) (sí.ta) * * b. ta (pí.ta) * * *! c. (tà.pi) (í.ta) * * *!
As with the pancito/panecito variation, the present account requires only minimal
constraint re-ranking to account for the Nicaraguan pattern. In this case, I
propose that the constraint MAX is now ranked below RESPECT and IDENT-SYLL.
This captures the fact that speakers in this dialect always drop the glide in order to
131
repair the phonotactic incompatibility that arises when –it{o, a} attempts to attach
to [j]-final stems.
(41) tapia ~ tapita “mud wall” (Nicaraguan)
tapj-it + DIM {o, a}/ R
ESPE
CT
IDEN
T-SY
LL
MA
X
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a.(tà.pje) (sí.ta) *! * b. ta (pí.ta) * * * c. (tà.pi) (í.ta) *! * *
This re-ranking has no impact on longer [jo]- [ja]-final words, since the repair
strategy of glide deletion is consistently applied in Nicaraguan: iglesia ~ iglesita.
VI. Further Evidence for Constraint Rankings
ONSET
In all dialects thus far discussed, ONSET is always ranked below the first
group of constraints.
RESPECT, MAX, IDENT-SYLL >> ONSET (Sonoran, Peninsular, Paraguayan) RESPECT, IDENT-SYLL >> MAX >> ONSET (Nicaraguan)
In this section, I justify this crucial ranking through analysis of a series of words
that have vowel final stems.
(42) Base Stem Dim. feo fe-it [fe.íto] “ugly” correo corre-it [ko.re.í.to] “mail”
132
tía ti-it [ti.íta] “aunt” té te- [te.síto] “tea” In Tableau (43), the winning candidate (43a), fe(í.to), contains an onsetless
syllable because all of the available repair strategies violate higher-ranked
constraints. Selection of consonant initial –cit{o, a} would satisfy ONSET, but
since feo is a part of Class D1, candidates (b), (c), and (f) all incur violations of
RESPECT. The solution of candidates (c) and (d) entails deletion of stem-final
[e]—the right bracket for candidate (c) indicating that the [e] is epenthetic and not
part of the stem. This has the added benefit of resulting in a single trochaic foot;
however, it violates the high-ranking constraint on segmental deletion, MAX.
Finally, candidate (e) avoids hiatus through transformation of the initial /i/ into a
coda position glide, a violation of IDENT-SYLL.
(43) feo ~ feíto “ugly”
/fe-it + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. fe (íto) * * * b. fe (sí.to) *! * * c. f]e (sí.to) *(!) *(!) * * * d. (fí.to) *! e. (féj.to) *! f. (fè.e) (sí.to) *! * * *
Although the grammar is forced to tolerate hiatus creation within Class D1 as the
lesser of several evils, this is not the case for Class D2 words. The monosyllabic
word té “tea” in Tableau (44) is able to avoid hiatus through selection of –cit{o,
a}, since RESPECT is not at issue for this diminutive class.
133
(44) té ~ tecito “tea”
/te- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. te (íto) *(!) *(!) * b. te (sí.to) * * c. t]e (sí.to) *! * * * d. (tí.to) *! e. (téj.to) *! f. (tè.e) (sí.to) *(!) *(!) *
Constraints and their rankings for the various dialects are presented on the
following pages using Hasse diagrams.
134
Constraints and rankings for Sonoran dialect:
*#Cji
FT-FORM(TROCHEE)
RESPECT, MAX, IDENT-SYLL
ONSET, PARSE-SYLL, DEP-V IDENT-V
PARSE-DIP REFLECT
ALIGN-FT-L, ALIGN-FT-R, OCP FEM = /a/
STRUCROLE
135
Constraints and rankings for North-Central Peninsular dialect (variation from Sonoran in bold): *#Cji
FT-FORM(TROCHEE)
RESPECT, MAX, IDENT-SYLL, PARSE-DIP
ONSET, PARSE-SYLL, DEP-V IDENT-V
ALIGN-FT-L, ALIGN-FT-R, OCP REFLECT
STRUCROLE FEM = /a/
136
Constraints and rankings for Paraguayan (variation from Sonoran in bold): *#Cji
FT-FORM(TROCHEE)
RESPECT, MAX, IDENT-SYLL
ONSET, PARSE-SYLL, DEP-V IDENT-V
PARSE-DIP REFLECT
STRUCROLE FEM = /a/
ALIGN-FT-L, ALIGN-FT-R, OCP
137
Constraints and rankings of Nicaraguan dialect (variation from Sonoran in bold): *#Cji
FT-FORM(TROCHEE)
RESPECT, IDENT-SYLL
MAX
ONSET, PARSE-SYLL, DEP-V
IDENT-V
PARSE-DIP REFLECT
STRUCROLE FEM = /a/
ALIGN-FT-L, ALIGN-FT-R, OCP
138
VII. Subcategorization Frames on Allomorphs: An Alternate View
While this analysis proposes a subcategorization frame for Class D1
words, a more common approach in the literature has been to attach a
subcategorization frame to an allomorph itself. Therefore, it is useful to examine
briefly how such an approach might work for Spanish diminutives. I argue that
Spanish diminutive formation does not easily lend itself to such an analysis due to
the uneven distribution of the allomorphs.
Several authors in the literature have argued for the use of
subcategorization frames to account for certain morphological processes. Booij
(1998), for example, argues for the use of subcategorization to account for non-
optimizing allomorphy. We have already seen in this chapter how Bonet (2006)
proposes the use of subcategorization frames to account for Spanish class marker
allomorphy.
Meanwhile, Paster (2006), in a cross-linguistic survey, has noted that
many cases of suppletive allomorphy are not phonologically optimizing and
argues that the P >> M model, by definition, has no mechanism to account for
such cases. She argues that associating subcategorization frames to allomorphs
that specify the phonologically conditioned context in which they appear is a
straightforward approach to apparent shortcomings of the P >> M model in cases
of non-optimizing allomorphy.
139
In one example, she examines the distribution of nominalizing suffixes in
Nakanai, an Austronesian language (2006: 212-213; her Nakanai examples are
from Johnson 1980: 177-178). In Nakanai, the nominalizing affix -il- is found
when it can both form part of the first syllable of the word and be adjacent to the
main stress of the word. The -la form occurs in all other contexts.
(45) au “steer” il-au “steering” peho “die” p-il-eho “death” go-ilo “go in” goilo-la “entrance” vi-kue “fight (v.)” vikue-la44 “fight (n.)” Paster establishes the following two subcategorization frames for these
allomorphs (214).
(46) Nominal construction A [ilnominalizing prefix [V, σ´]verb stem]noun
Nominal construction B (“elsewhere”) [ [ ]verb stem lanominalizing suffix]noun
Essentially the first subcategorization frame lexically specifies the phonological
conditions under which -il- will attach to a verb stem. The second frame indicates
that la will nominalize in all other contexts in which the conditions for -il-
attachment are unmet.
Unfortunately, this approach seems to work best when there is a clear
distribution between the two allomorphs. This is not the case for the Spanish
diminutive allomorphs. As (47) illustrates, a full account of the coverage of the
44 Paster indicates that stress in Nakanai falls on the penultimate syllable. Thus, I take each vowel to be the head of its own syllable. vikue must be syllabified vi.ku.e without a final diphthong.
140
two allomorphs would require a minimum of five separate subcategorization
frames, while missing several key phonologically conditioned generalizations in
the process.
(47) Diminutive allomorph –it{o, a}
[ [Class D1 non-[j] final]non-verbal stemit{o, a}diminutive suffix]diminutive
[ [3+ σ /e/-base]non-verbal stemit{o, a}diminutive suffix]diminutive
Diminutive allomorph –(e)cit{o, a}
[ [1 σ Bare stem]non-verbal stemecit{o, a}diminutive suffix]diminutive
[ [2+ σ Bare stem]non-verbal stemcito{o, a}diminutive suffix]diminutive
[ [2 σ /e/-base]non-verbal stemecit{o, a}diminutive suffix]diminutive
Rather than abandon subcategorization as a means to account for non-
optimizing allomorphy, in this chapter I have proposed placing a frame on Class
D1 words instead of several frames on each allomorph. In this way, words in
Class D1 are idiosyncratically marked to select for a particular diminutive
allomorph, specifically –it{o, a}. No other subcategorization frames are required
and all other patterns are determined by the ranking of phonological constraints,
resulting in no loss of generalizations.
141
Chapter 5
Concluding Remarks
In this final chapter I review the principal elements of my analysis of
Spanish diminutivization. In the first section, I briefly review my underlying
assumptions that diminutivization is a stem-level process in which both
allomorphs are present in the input, but that Class D1 words are lexically-
specified to select the -it{o, a} allomorph. Then, in Section II, I outline the major
elements of my analysis within the framework of Optimality Theory, highlighting
the interaction of the RESPECT constraint with Class D1 words and the
phonological constraints that account for the allomorphic distribution in Class D2
words. Finally, I conclude this work with a brief discussion of a possible
objection of the subcategorization frame and lexical specification approach key to
this analysis and suggest possible avenues for investigation in the historical record
to support my claims.
I. Review of Underlying Assumptions
In Chapter 3 I reviewed Harris’ five word form classes (1991a, b) and
compared them to the data for diminutivization in Sonoran Spanish. While
Harris’ analysis may offer a compelling division of non-verbal words in the
language in general, I argue that, at least in terms of diminutivization, the
distribution of -it{o, a} and -cit{o, a} reveals the limits of the utility of his classes.
Within his first three classes (in which the most abundant and consistent data can
142
be found), both diminutive forms occur with regularity. However, those words
that end in unstressed /o/ and /a/, Harris’ Classes I and II respectively, form their
diminutives in identical fashion.45 Thus, I proposed a merger of these two classes
for the purposes of diminutivization into a single Class D1. In Sonoran, this class
selects -it{o, a} without respect to prosodic shape or more general phonological
factors. The one exception to this generalization occurs with disyllabic words
ending in the diphthongs -[jo] or –[ja] that select -cit{o, a} as a repair strategy to
avoid a never-violated constraint on homorganic vocoids in rhyme position. In
the example of novio “groom, boyfriend,” the phonotactically impossible
candidate *nov[ji]to reveals the one limitation of the -it{o, a} allomorph for this
class of words. That is, its initial segment is a high front vowel [i] that makes it
incompatible with those stems that end in a high front glide [j]. Words from
Harris’ other form classes and those words ending in a stressed vowel (e.g. café)
were placed into Class D2.
The distinction between Class D1 and Class D2 is that those words in the
first are lexically specified to select for the -it{o, a} allomorph, whereas in D2,
diminutive selection is dictated strictly by phonological constraints. That is, the
difference between the behavior of Class D2 words madre~madrecita and
comadre~comadrita is not a case of lexical specification in one but not the other,
but rather that they are funnelling into an unmarked prosodic structure of two
trochaic feet.
45 The one exception, as discussed in Chapter 3, are the ladrona~ladroncita types, which I argue in Section III rightly belong in Class IIIA' along with their masculine counterparts (e.g. ladrón~ladroncito).
143
Due to the strict morphological constraint on Class D1 words, this
emergence of the unmarked only becomes apparent when comparing the behavior
of disyllabic and trisyllabic words that end in the aforementioned diphthongs. In
the case of novio, phonology trumps morphology (P >> M) and -cit{o, a} is
selected to avoid the illicit syllable structure with the added benefit of adhering to
the preferred prosodic structure of two trochaic feet [nò.ßje.sí.to]. Due to its extra
syllable, the trisyllabic iglesia is able to conform to its lexical specification and
the prosodic structure, (ì.gle) (sí.ta). However, with the stem of the disyllabic
casa, [kas-], there are no insurmountable phonological barriers to the selection of
the lexically specified -it{o, a} allomorph, even though the diminutive form casita
reflects a more marked prosodic structure in which one syllable is left unparsed
[ka.(sí.ta)].
Notwithstanding the lexical specification requirements on Class D1, it is
clear that both diminutive allomorphs are present in the input and are available
options to both classes. Spanish diminutivization is a case of suppletive
allomorphy in which one allomorph is not derived through segmental epenthesis
or deletion from the other. I have also argued that all words undergo a process of
suffixation in which the diminutive attaches to a stem that does not contain the
final unstressed vocalic element.46 Those proponents of an analysis of infixation
of -it- are left to explain the transformation of /e/ final words into /o/ or /a/ final
words, such as uniforme or comadre into uniformito and comadrita.
46 As noted previously, the final stressed vocalic element of words such as menú are crucially part of the stem to which the diminutive allomorph attaches.
144
In my analysis the final vocalic element comes not from the base, but is a
part of the diminutive suffix and is represented by the unordered set {o, a}. For
words that end in unstressed /o/ or /a/, the inflexional marker of the diminutive
suffix is constrained by REFLECT, which demands that derived forms of words
contain all segments present in the base form. The input-output constraint IDENT-
VOWEL is ranked above REFLECT and ensures that the final vocalic segment is
faithful to the inflexional possibilities of the diminutive suffix in the set {o, a}. In
other words, high-ranking IDENT-VOWEL ensures that a diminutive-final /e/ would
never be a possible outcome. At the same time, the ranking IDENT-VOWEL >>
REFLECT allows non-canonical words, such as masculine poema that ends in
typically feminine /a/, to preserve their idiosyncratic nature in the diminutive
form, poemita, because /a/ is a part of the inflexional set {o, a}.
Finally, I have argued that the diminutive suffixes attach to a stem that
does not contain the final vocalic element (unless it is stressed). The fact that the
diminutive madrecita appears to maintain its base-final /e/ in this form is simply a
fortuitous outcome due to the segmental similarity between word-final /e/ in
madre and the epenthetic /e/ in Spanish. In the same way, the /e/ in the verb
cobardear “to act cowardly” is a part of the verbal suffix –ear and does not come
from the final /e/ in the noun cobarde “coward.”47
In sum, Spanish diminutivization is a stem-level process of suffixation in
which both suffixes (represented by DIM), with their identical set of inflexional
markers {o, a}, are present in the input. Selection is constrained principally by
47 This is not to say that the final /e/ in the base cobarde does not exert some influence in the selection of the verbal suffix –ear instead of an alternative –ar. The constraint RESPECT could play a role in this selection.
145
phonological factors except in Class D1 words that are lexically specified to select
-it{o, a} and do so unless violation of higher ranking phonological constraints
prevent such selection, as will be reviewed in more detail in the following section.
II. Review of OT Analysis
One of the main goals of the present work has been to establish the
relationship between the morphology and the phonology in this highly productive
process. The fundamental difference in diminutivization between the prosodically
and morphologically identical words casa and madre48—selection of -it{o, a} in
the former and -cit{o, a} in the latter—forces us to consider one of various
solutions. The first, as discussed at the end of the previous chapter, follows
suggestions that subcategorization frames placed on allomorphs can account for
non-optimizing allomorphy (Booij 1998, Paster 2006a, b). Applied to
diminutivization, the assumption here would be that selection of the -it{o, a}
allomorph by disyllabic words ending in unstressed /o/ or /a/ results in a non-
optimal prosodic shape; for example, casita [ka.(sí.ta)] leaves one syllable
unparsed and does not fit into the unmarked pattern of two trochaic feet that we
have seen emerge with so many other examples: madre~(mà.dre) (sí.ta),
comadre~ (kò.ma) (drí.ta), iglesia (ì.gle) (sí.ta), chamaka~(ʧà.ma) (kí.ta), etc.
This approach, however, places too much complexity on the morphology without
dispensing completely with the phonology. No less than three subcategorization
frames would need to be posited for -cit{o, a} and no fewer than two for -it{o, a},
and, at the same time, since the subcategorization frames refer to phonological 48 Recall that I assume that unstressed final /o/, /a/, and /e/ are morphological equals (i.e. all three are class markers and not epenthetic in the case of /e/) and that they are all absent from the stem upon suffixation.
146
information in the base, the allomorphs would still need to “see” the phonological
structure of the words they attach to. This seems far too convoluted an analysis to
account for what amounts to the slightly marked case of one word type.
Another approach to the casita/madrecita issue assumes that there is a
fundamental difference in the behavior of the two allomorphs: -it{o, a} is an infix
and -cit{o, a} is a suffix. Putting aside the already discussed limitations of the
-it{o, a} as infix model, this reasoning also adds an unnecessary level of
complexity in that it creates two separate systems in a situation in which, given a
set of assumptions, it is possible to maintain a unified system.
The principal assumption that my analysis parts from is that what is
idiosyncratic or “different” is a particular class of words, and not the diminutive
allomorphs themselves. The fact that casa selects for -it{o, a} despite the
availability of a more prosodically-optimizing option (*casecita would conform
to the unmarked prosodic structure of two trochaic feet) is the result of Class D1
words being lexically specified for that suffix, which is due, presumably, to some
restriction yet to be unearthed in the history of the language. Put another way,
rather than defining the diminutive allomorphs by their behavior—that is, whether
they attach to one type of word or another—the data suggest that the words
themselves should be defined by their sensitivity to optimizing phonological
constraints (Class D2) or not (Class D1). It is not that the diminutive allomorphs
will only attach to certain words, but that certain words will only select a certain
allomorph.
147
I have formalized this distinction through the placement of the
subcategorization frame it on Class D1 words (e.g. casa /kasit/) and its interaction
with the high ranking constraint RESPECT that mandates adherence to the
idiosyncratic lexical specification of the -it{o, a} allomorph for these words. A
candidate such as *casecita satisfies the markedness constraints on prosodic
shape, such as PARSE-SYLLABLE, ALIGN-FOOT-LEFT and ALIGN-FOOT-RIGHT, but
they are ranked below RESPECT and therefore are not sufficient to “override” the
lexical specification for -it{o, a}.
In contrast to the Class D1 example (casa), the word madre belongs to
Class D2, a group of words that is unified only by its lack of lexical specification
for the -it{o, a} allomorph; a group of words, therefore, that is free to select
whichever allomorph allows for best satisfaction of the relevant markedness and
faithfulness constraints. The winning candidate, madrecita ~ (mà.dre) (sí.ta),
with its prosodic shape of two trochaic feet, satisfies PARSE-SYLLABLE, ALIGN-
FOOT-LEFT, and ALIGN-FOOT-RIGHT, *madrita ~ ma(drí.ta) does not. Tableau (1)
illustrates the contrast between lexically-specified Class D1, casa, and Class D2,
madre.
(1) casa~casita vs. madre~madrecita
/kas-it + DIM {o, a}/ R
ESPE
CT
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
a. ka (sí.ta) * * b. (kà.se) (sí.ta) *! * /madr- + DIM {o, a}/ c. ma (drí.ta) * *! d. (mà.dre) (sí.ta) *
148
In Sonoran, the only Class D1 words that violate the constraint RESPECT
are those disyllabics that end in the diphthong [jo] or [ja] (e.g. novio) and whose
stems therefore end in the front glide [j]. Selection of -it{o, a}, then, would result
in an illicit syllable structure of two homorganic vocoids in rhyme position *.Cji.
Selection of noviecito over *nov[ji]to reflects the constraint ranking *.Cji >>
RESPECT. In all other cases, RESPECT dominates lower-ranking markedness
constraints.
(2) novio~noviecito
/nobj-it + DIM {o, a}/
*.Cji. RESPECT
a. no (ßjí.ta) *! b. (nò.ßje) (sí.ta) *
As Colina (2003) has noted, the concept of the Emergence of the
Unmarked (TETU) is a key element in the understanding of Spanish Diminutive
Formation, and it plays an important role within my analysis of the data as well.
As it must certainly be, given the number of word types that, following
diminutivization, result in a prosodic structure of two trochaic feet:
Class D1 words Disyllabic ending in [jo] or [ja]: novio “groom” (nò.ßje) (sí.to) Trisyllabics ending in [jo] or [ja]: iglesia “church” (ì.gle) (sí.ta) Trisyllabics ending in /o/ or /a/: ventana “window” (bèn.ta) (ní.ta) Class D2 words
149
Monosyllabics:49 pan “bread” (pà.ne) (sí.to) Disyllabics ending in /e/: madre “mother” (mà.dre) (sí.ta) Disyllabics ending in stressed V: café “coffee” (kà.fe) (sí.to) Disyllabics ending in C: balcón “balcony” (bàl.kon) (sí.to) Trisyllabics ending in /e/: comadre “godmother” (kò.ma) (drí.ta)
The only exceptions in Sonoran are disyllabics that end in unstressed /o/ or /a/ and
those words that are four syllables or longer, which would not be able to conform
to this prosodic structure without deletion of several segments, incurring multiple
violations of MAX, the constraint on segmental deletion. The ranking of PARSE-
SYLLABLE, ALIGN-FOOT-LEFT and ALIGN-FOOT-RIGHT below MAX predicts that
we will not find segmental deletion in order to satisfy the unmarked structure of
two trochaic feet; and with one exception, this prediction is born out. In the case
of trisyllabic iglesia, we find, as mentioned, that the stem-final glide [j] creates an
illicit syllable structure if the lexically specified -it{o, a} allomorph is selected.
The two repair strategies of glide deletion or selection of the -(e)cit{o, a}
allomorph incur violations of equally high-ranking MAX in the former strategy
and RESPECT in the latter. Thus, candidate evaluation moves down the constraint
hierarchy and selection is based on conformity to unmarked structures. This
process is best illustrated by the minimal pairs of nov[ja] and igles[ja] as well as
madre and comadre in Tableau (3) and (4) in which the disyllabic and trisyllabic
words choose different allomorphs to funnel into the two trochaic feet structure.
49 With the exception of the Nicaraguan and Paraguayan dialects, which do not add an epenthetic [e]: pancito.
150
(3) novia~noviecita, iglesia~iglesita
/iglesj-it + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (ì.gle) sje (sí.ta) * * *(!) *(!) b. (ì.gle) (sí.ta) * * c. (ì.gle) si (í.ta) * **(!) *(!) * /nobj-it + DIM {o, a}/ d.(nò.ßje) (sí.ta) * * e. no (ßí.ta) * * *! f. (nò.ßi) (í.ta) * * *!
(4) madre~madrecita, comadre~comadrita
/madr- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. ma (drí.ta) * *! b. (mà.dre) (sí.ta) * /komadr- + DIM {o, a}/ c. (kò.ma) (drí.ta) d. ko (mà.dre) (sí.ta) *(!) *(!) * e. (kò.ma) dre (sí.ta) *(!) *(!)
In the case of disyllabics ending in a stressed vowel or a consonant, both
allomorphs adhere to the unmarked prosodic structure—(kà.fe) (sí.to) ~ *(kà.fe)
(í.to) and (bàl.kon) (sí.to) ~ (bàl.kon) (í.to)—and selection is determined by
additional constraints on onsetless syllables (ONSET) and faithfulness to
segmental structural roles in the base (STRUC ROLE) respectively.
(5) café~cafecito
151
/kafe- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
a. (kà.fe) (í.to) *! b. (kà.fe) (sí.to)
(6) balcón~balconcito
/balkon- + DIM {o, a}/ R
ESPE
CT
MA
X
IDEN
T-SY
LL
ON
SET
PAR
SE S
YLL
DEP
-V
ALI
GN
- FT
-L
ALI
GN
- FT
-R
OC
P
STR
UC
R
OLE
a. (bàl.ko) (ní.to) *! b. (bàl.kon) (sí.to)
III. Possible Objections and Future Work One of the principal objections to the analysis of Spanish diminutives that
I outline in this work, I suspect, will concern the use of a subcategorization frame
on a class of words and the corresponding RESPECT constraint that ensures
adherence to this idiosyncratic relationship. In a sense, one could argue that this
is notational shorthand for the lack of any explanation altogether; in other words,
the reason why Class D1 words select for the -it{o, a} allomorph is simply
because they do. As unsatisfying a response as this may be, it does allow for all
other diminutivization to be captured by established faithfulness and markedness
constraints within a unified system of suffixation to the stem. In other words, the
greatest number of word types is explained through the phonology and the
phonology alone. This approach also obviates the need for an abundance of
allomorph-specific constraints that are necessarily language-specific and of
doubtful universal status.
152
In addition, morphological idiosyncrasies that are unable to be captured by
conventional phonological explanations abound in the language. As just one
example, the masculine word poema “poem,” which ends in the canonically
feminine word marker /a/, must be lexically-specified to end in /a/; which is to say
that it ends in /a/ because it does. This is just as unsatisfactory an explanation, but
no less accurate.
Of course, one could explore the historical record and posit a reasonable
theory concerning the adaptation of Greek borrowings into the Spanish language,
but from a synchronic point of view this knowledge will not change the lexical
specification analysis. In the same vein, a thorough examination of
diminutivization in Latin and throughout the history of the Spanish language
could reveal the origins of the lexical bond between Class D1 words and the -it{o,
a} allomorph and strengthen the current work’s claims, but it would still leave us
with essentially the same non-phonological synchronic analysis.
Since the present analysis hinges squarely on lexical specification,
evidence from such a diachronic investigation would be of great support to this
approach. For example, one wonders if there was a correspondence between
diminutive selection in Latin and the declensional class of a word. If so, does the
diminutive allomorph distribution in modern Spanish reflect such a relationship?
Additionally, it remains to be seen whether the Peninsular dialect is
conservative in its selection of the -cit{o, a} allomorph for disyllabics ending in
unstressed /o/ or /a/ (e.g. puerta~puertecita) or if this is an innovation. This
information has implications for whether the lexical bond between Class D1
153
words and the -it{o, a} is being strengthened or weakened diachronically. In the
same way, does the preference for -it{o, a} in Nicaraguan for disyllabics ending in
the diphthong [jo] or [ja]—radio~radito—reflect a strengthening of the lexical
bond or a weakening of the two trochaic feet structure?
Spanish diminutivization is a complex morphological process with
several variables in play. Its high synchronic productivity makes it a fruitful
proving ground for theories on the morphology-phonology interface. If the
present work’s predictions are correct, there is still much work to be done to
understand how diachronic processes exert an influence on this interface.
154
References Aguero-Bautista, Calixto. 1998. “Cyclic and Identity Effects in Spanish
Diminutives and Augmentatives. Unpublished Phonology Generals paper, MIT.
Alderete, John. 1995. “Faithfulness to prosodic heads.” Rutgers Optimality
Archive, Report No. 94. Ambadiang, Théophile. 1996. “La formación de diminutivos en español:
¿fonología o morfología? LEA 18.2: 175-211. Ambadiang, Théophile. 1997. “Las bases morfológicas de la formación de
diminutivos en español.” Verba 24: 99-132. Bermúdez-Otero, Ricardo 2007. “Morphological structure and phonological
domains in Spanish denominal derivation.” In Fernando Martínez-Gil & Sonia Colina (eds), Optimality-theoretic studies in Spanish phonology. Amsterdam: John Benjamins: 278-311.
Bonet, Eulàlia; Lloret, Maria-Rosa; and Joan Mascaró. 2003. “Phonology-
Morphology Conflicts in Gender Allomorphy: A Unified Approach.” Handout of a talk given at Glow 26 (Lund, April 9, 2003).
Bonet, Eulàlia. 2004. “Morph Insertion and Allomorphy in Optimality Theory.”
IJES 4.2: 73-104. Bonet, Eulàlia and Maria-Rosa Lloret. 2005. “Against Serial Evaluation in
Optimality Theory.” Lingua 115.9: 1303-1323. Bonet, Eulàlia. 2006. “Gender Allomorphy and Epenthesis in Spanish.” In
Fernando Martínez-Gil and Sonia Colina (eds.), Optimality-Theoretic Studies in Spanish Phonology. Philadelphia: John Benjamins Pub. 312-338.
Bonet, Eulàlia; Lloret, Maria-Rosa; and Joan Mascaró. 2007. “Allomorph
selection and lexical preferences: Two case studies.” Lingua 117: 903-927.
155
Booij, Geert. 1998. “Phonological Output Constraints in Morphology.” In
Wolfgang Kehrein and Richard Wiese (eds.), Phonology and Morphology of the Germanic Languages. Tübingen: Niemeyer. 143-163.
Colina, Sonia. 2003a. “Diminutives in Spanish: A Morpho-Phonological
Account.” Southwest Journal of Linguistics 22.2: 45-88. Colina, Sonia. 2003b. “The Status of Word-Final [e] in Spanish.” Southwest
Journal of Linguistics 22: 87-107. Colina, Sonia. 2009. Spanish Phonology: A Syllabic Perspective. Washington, DC:
Georgetown University Press. Crowhurst, Megan J. 1992. “Diminutives and augmentatives in Mexican Spanish:
a prosodic analysis.” Phonology 9: 221-253. Elordieta, G. and María M. Carreira. 1996. “An Optimality Theoretic Analysis of
Spanish Diminutives.” CLS 32: 49-60. Hall, Alan T. and U. Kleinhenz (eds.). 1999. Studies on the Phonological Word.
Amsterdam: John Benjamins. Harris, James W. 1983. Syllable Structure and Stress in Spanish. Linguistic
Inquiry Monograph 8. Cambridge, MA: MIT Press. Harris, James W. 1985. “Spanish Word Markers.” In Frank H. Nuessel (ed.),
Current Issues in Spanish Phonology and Morphology. Bloomington: Indiana University Linguistics Club. 34-54.
Harris, James W. 1991a. “The Exponence of Gender in Spanish.” LI 22: 27-62. Harris, James W. 1991b. “The Form Classes of Spanish Substantives.” Yearbook
of Morphology 2: 65-88. Harris, James W. 1994. “The OCP, Prosodic Morphology and Sonoran Spanish
Diminutives: a reply to Crowhurst.” Phonology 11: 179-190. Hualde, José I. 1992. “On Spanish Syllabification.” In Campos, H. and F.
Martínez Gil (eds.), Current Studies in Spanish Linguistics: Washington, D.C.: Georgetown University Press. 475-493.
Hualde, José. 2005. The Sounds of Spanish. Cambridge: Cambridge University
156
Press. Jaeggli, Osvaldo. 1978. “Spanish Diminutives.” In F.H. Nuessel (ed.), Contemporary Studies in Romance Languages. Bloomington, Indiana: Indiana University Linguistics Club. 142-
155. Johnston, Raymond Leslie. 1980. Nakanai of New Britain: The Grammar of an Oceanic Language. Canberra: Australian National University. Kaisse, Ellen M. and J. Harris. 1999. “Palatal Vowels, Glides and Consonants in
Argentinian Spanish.” Phonology 16: 117-190. Kikuchi, Seiichiro. 2001. “Spanish Definite Article Allomorphy: A
Correspondence Approach.” Phonological Studies 4: 49-56. Klein, Philip W. 1989. “Spanish ‘Gender’ Vowels and Lexical Representation.”
Hispanic Linguistics 3: 147-162. Mascaró, Joan. 2007. “External allmorphy and lexical representation.” Linguistic
Inquiry 38: 715-735. McCarthy, John J. and Alan S. Prince. 1993a. “Prosodic Morphology I:
Constraint Interaction and Satisfaction.” Report No. RuCCS-TR-3. New Brunswick, NJ: Rutgers University Center for Cognitive Sciences.
McCarthy, John J. and Alan S. Prince. 1993b. “Generalized Alignment.” In Geert
Booij and Jaap van Maarle (eds.), Yearbook of Morphology. Dordrecht: Kluwer. 79-153.
Miranda Miranda, Inés. 1999. An Optimality Theoretic Analysis of Nicaraguan
Spanish Diminutivization: Results of a Field Survey. Ph.D. dissertation, University of Washington.
Paster, Mary. 2005. “Pulaar verbal extensions and phonologically driven affix
order.” Yearbook of Morphology 2005: 155-199. Paster, Mary. 2006. “Phonological Conditions on Affixation.” Berkeley, CA:
University of California dissertation. Piñeros, Carlos-Eduardo. 2000a. “Prosodic and Segmental Unmarkedness in
Spanish Truncation.” Linguistics 38.1: 63-98. Piñeros, Carlos-Eduardo. 2000b. “Foot-sensitive word minimization in Spanish.”
157
Probus 12.2: 291-324. Prieto, Pilar. 1992. “Morphophonology of the Spanish Diminutive Formation: A Case for Prosodic Sensitivity.” Hispanic Linguistics 5.1-2: 169-205. Prince, Alan and P. Smolensky. 1993/2004. “Optimality Theory: Constraint
Interaction in Generative Grammar.” RuCCS Technical Report 2. Rutgers University, Piscateway, NJ: Rutgers University Center for Cognitive Science. Revised version 2004 published by Blackwell.
Roca, Iggy. 1989. “The Organization of Grammatical Gender.” Transactions of the
Philological Society 87: 1-32. Stephenson, Tamina. 2004. “Declensional-type Classes in Derivational
Morphology: Spanish Diminutives Revisited.” Unpublished Phonology/Morphology generals paper, MIT.
Teschner, R. V. and W. M. Russell. 1984. “The Gender Patterns of Spanish
Nouns: An Inverse Dictionary Analysis.” Hispanic Linguistics 1: 115-132.
Top Related