Adding and subtracting alternation: resumption and prepositional phrase chopping in Spanish relative clauses

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ADDING AND SUBTRACTING ALTERNATION: RESUMPTION AND
PREPOSITIONAL PHRASE CHOPPING IN SPANISH RELATIVE CLAUSES

by

Álvaro Cerrón-Palomino López

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(LINGUISTICS, HISPANIC LINGUISTICS)

May 2010

Copyright 2010 Álvaro Cerrón-Palomino López
ii

Dedication
To my parents, Rodolfo Cerrón Palomino and María López Gutiérrez, to whom I owe all
that may be good in me.
iii

Acknowledgements
The writing and completion of this dissertation would not have been possible
without the support of numerous individuals. Therefore, I am sincerely grateful to all of
them: professors, colleagues, peers, friends, and family members.
To begin with, I would like to extend my most sincere gratitude to my advisor and
dissertation defense committee chair, Professor Carmen Silva-Corvalán. Her immense
knowledge and insightful feedback and suggestions have guided during my Ph.D. studies,
throughout the different stages of this dissertation, and in every aspect of my academic
life. Also, I would like to thank the members of my defense committee: Professor Mario
Saltarelli, who followed my work from its inception and provided me with valuable
insights from a formal perspective, and Professor Elaine Andersen, who showed interest
in my work from its initial stages and familiarized me with the psycholinguistic aspects
of relative clause processing.
I am indebted to the USC Linguistics Department faculty who were members of
different committees on my path towards my Ph.D. defense: John Hawkins, Edward
Finegan, Todd Haskell, and Elsi Kaiser. Also, I would like to thank Professor María
Luisa Zubizarreta, for her valuable comments and suggestions regarding my research.
Very special thanks go to Dr. Gayle Fiedler-Vierma, director of the Spanish
language program at USC. Not only was she supportive in every aspect a coordinator can
be, but she also mentored me in all the subjects of academic life, and proved to be a
wonderful friend.
iv

I am also grateful to the USC Department of Spanish and Portuguese‟s staff, in
particular, Marta Galvan and Amelia Acosta, for their constant support and friendship.
Likewise, I would like to thank the USC Department of French and Italian Faculty,
Francesca Italiano and Carol Hofmann; and the French and Italian staff, Patrick Irish and
Valentina Stoicescu for their help and constant support.
I am eternally indebted to Mónica Cabrera and Héctor Velásquez, whose
friendship I benefitted from since we studied and worked at the Pontifical Catholic
University of Peru. They were my Peruvian family in Los Angeles, and they were there
for me in times of hardship.
I am very appreciative of the insightful comments and feedback of my fellow
Hispanic Linguistics students and HILSA members: Asier Alcázar, Omar Beas, Rebeka
Campos-Astórkiza, Roberto Mayoral, Michal Temkin Martínez, Magdalena Pire-
Schmidt, Ana Sánchez Muñoz, Benjamin Parrell, Sergio Robles-Puente, Michael
Rushforth, Laura Tejada, and Erika Varis. Very special thanks go to Emily Hinch Nava,
who kindly read my manuscript and polished its style.
I would like to thank the staff in the Linguistics Department of USC, Karma
Dolma, and Frankie Hayduk, for all their help. Also, my special gratitude goes to Joyce
Perez, without whom my journey through the appalling world of graduate students‟
paperwork would have been a disaster.
Last but definitively not least, I would like to thank my family. In particular, I
thank my parents, Rodolfo Cerrón Palomino and María López Gutiérrez, for everything
in my life before, during and beyond this enterprise. I thank my beloved daughter, Sofia
v

K. Cerrón-Palomino, who unknowingly, is a permanent source of inspiration for me in
every step I take. I am also very grateful to my brother, Fernán Cerrón-Palomino, and his
family for their never-ending support. Finally, I would also like to thank Rosario
Huanaco Chile for her continuous support and encouragement since my tender years.

vi

Table of Contents

Dedication ii
Acknowledgements iii
List of Tables ix
List of Figures xii
Abstract xiii
Chapter 1: Spanish Relative Clause System and Variables
1.1 Introduction 1
1.2 Spanish Relative Clause Marker Paradigm 4
1.3 The Nature of que 7
1.4 Spanish Relative Clauses and Types 9
1.4.1 Restrictive and Non-Restrictive Relative Clauses 12
1.4.1.1 Semantic Difference 12
1.4.1.2 Structural/Syntactic Difference 13
1.4.1.3 Intonational Difference (Pause) 15
1.4.2 Coordinated Relative Clauses 16
1.4.3 Juxtaposed Relative Clauses 18
1.4.4 Complex Relative Clauses 20
1.4.5 Definite and Indefinite Noun Phrase Antecedents 22

Chapter 2: Previous Studies of Resumptive Pronouns and Prepositional
Phrase-Chopping
2.1 Theoretical Studies 24
2.1.1 The Accessibility Hierarchy and the Resumptive Pronoun
Hierarchy Prediction 24
2.1.2 Exceptions to the Accessibility Hierarchy 28
2.2 Quantitative Studies 32
2.2.1 Resumptive Pronouns in English, Yiddish, and Hebrew 32
2.2.2 Resumptive Pronouns and Prepositional Phrase-Chopping
in Portuguese 34
2.3 Studies of Spanish Resumptive Pronouns and Prepositional Phrase-
Chopping 38
2.3.1 A Pan-Hispanic Account 38
2.3.2 A Cross-Dialectal Study of Direct Object Resumptive
Pronouns 41
2.3.3 Resumptive Pronouns in Santa Cruz (Spain) 46
vii

2.3.4 Resumptive Pronouns in Santiago (Chile) 47
2.3.5 Resumptive Pronouns in Caracas (Venezuela) 50
2.3.6 Resumptive Pronouns in Gran Canaria (Spain) 51
2.3.7 Resumptive Pronouns in Lima (Perú) 52

Chapter 3: Adding and Subtracting in Spanish Relative Clauses
3.1 Methodology and Data 56
3.1.1 Data Set 1 57
3.1.2 Data Set 2 58
3.1.3 Data Set 3 59
3.1.4 Data Set 4 61
3.1.5 The Nature of the Data 61
3.1.5.1 Incomplete Sentences 63
3.2.5.2 Unclear Sentence Structures 66
3.2.5.3 Hard to Classify Tokens 70
3.1.5.4 Relative Clauses with No Overt Subordinator 71
3.2 Envelope of Variation 72
3.2.1 Headless Relative Clauses and Headless Antecedents 73
3.2.2 Passive Constructions 74
3.2.3 Indirect Objects 76
3.2.4 Genitives 78
3.2.5 Attributes 79
3.2.6 The Oblique Antecedent vez 81
3.2.7 Passive Reflexive Voice 82
3.2.8 Resumptive Noun Phrases 83
3.3 Hypotheses Common to Subject, Direct Object and Oblique Relative
Clauses 85
3.3.1 Syntactic Category of the Antecedent 86
3.3.2 Restrictiveness of the Relative Clause 86
3.3.3 Definiteness of the Antecedent 87
3.3.4 Previous Mention of the Antecedent 88
3.3.5 Distance between the Antecedent and the Complementizer
que 89
3.3.6 Animacy of the Antecedent 90
3.3.7 Size of the Relative Clause Domain 90
3.3.8 Weight of the Information Borne by the Relative Clause 92
3.3.9 Coordinated Relative Clauses 93
3.3.10 Juxtaposed Relative Clauses 94
3.4 Hypotheses Exclusive to Oblique Relative Clauses 94
3.4.1 Type of Preposition 95
3.4.2 Category Selecting the Prepositional Phrase 96
3.5 Coding of the Data 96

viii

Chapter 4: Results and Discussion
4.1 Resumptive Pronoun Presence/Absence Alternation 102
4.1.1 Results for Subject 102
4.1.2 Results for Direct Object 117
4.1.3 Results for Oblique 129
4.2 Prepositional Phrase-Chopping Presence/Absence Alternation 143
4.3 Discussion of the Results 154
4.4 Prepositional Phrase-Chopping and Resumptive Pronoun
Overall Comparison 158
4.5 Qualitative Analysis: Contrast and Subject Resumptive Pronouns 160

Chapter 5: Conclusions
5.1 Aims of the Dissertation 166
5.2 Summary of Findings 169
5.2.1 Conditioning Factors and Their Uniformity across
Syntactic Positions 169
5.2.2 Oblique Resumptive Pronouns and Prepositional
Phrase-Chopping 171
5.2.3 The Universality of Resumptive Pronoun Factors 172
5.3 Implications and Future Research 174

Bibliography 180

Appendix 184
ix

List of Tables

Table 1: Spanish Relative Clause Markers 5
Table 2: Spanish Definite and Indefinite NPs 22
Table 3: PP-Chopping, RP and Gap in BP RCs 36
Table 4: Description of the Speakers of Data Set 1 58
Table 5: Description of the Speakers of Data Set 2 59
Table 6: Subject – Restrictiveness 102
Table 7: Subject – Definiteness 103
Table 8: Subject – Antecedent Syntactic Category 105
Table 9: Subject – Weight 107
Table 10: Subject – Distance 108
Table 11: Subject – Domain Size 109
Table 12: Subject – Animacy 110
Table 13: Subject – Previous Mention 111
Table 14: Subject – Coordination 112
Table 15: Subject – Juxtaposition 113
Table 16: Subject RPs Goldvarb 116
Table 17: Object – Restrictiveness 117
Table 18: Object – Definiteness 118
Table 19: Object – Antecedent Syntactic Category 119
Table 20: Object – Weight 120
x

Table 21: Object – Distance 121
Table 22: Object – Domain Size 122
Table 23: Object – Animacy 123
Table 24: Object – Previous Mention 124
Table 25: Object – Coordination 125
Table 26: Object –Juxtaposition 126
Table 27: Object RPs Goldvarb 128
Table 28: Oblique – RP – Restrictiveness 132
Table 29: Oblique – RP – Definiteness 132
Table 30: Oblique – RP – Weight 133
Table 31: Oblique – RP– Distance 135
Table 32: Oblique – RP– Domain 136
Table 33: Oblique – RP – Animacy 137
Table 34: Oblique – RP – Previous Mention 138
Table 35: RP – Oblique – Coordination 139
Table 36: Oblique – RP – Selecting Category 140
Table 37: Oblique – RP –Preposition Type 141
Table 38: Oblique RPs Goldvarb 142
Table 39: Oblique – PP –Restrictiveness 144
Table 40: Oblique – PP –Definiteness 145
Table 41: Oblique – PP –Weight 146
Table 42: Oblique – PP –Distance 147
xi

Table 43: Oblique – PP – Domain 148
Table 44: Oblique – PP –Animacy 149
Table 45: Oblique – PP –Previous Mention 150
Table 46: Oblique – PP – Coordination 151
Table 47: Oblique – PP – Preposition Type 152
Table 48: Oblique – PP – Selecting Category 153
Table 49: Oblique PP-Chopping Goldvarb 154
Table 50: Effect of Conditioning Factors across the Three
Syntactic Positions 155

Table 51: Oblique RP / PP-Chopping Factor Significance
Comparison 159

Table 52: Languages Combining [-Case] Gaps with [+Case]
Resumptive Pronouns 184
xii

List of Figures

Figure 1: Subject RP Rising Pitch 162
Figure 2: Subject RP Lowering Pitch #1 163
Figure 3: Subject RP Lowering Pitch #2 164
xiii

Abstract

This dissertation is a variationist account of two non-standard relative clause (RC)
structures in Spanish: resumptive pronouns (RPs) and prepositional-phrase (PP)
chopping. Previous typological studies considered RP explanations based on difficulty of
processing (Hawkins, 1994), while Spanish-specific quantitative studies proposed a
number of factors regarding the relationship between the RC and its antecedent (Silva-
Corvalán, 1996; Bentivoglio, 2003; Samper et al., 2003). To my knowledge, there are no
previous studies of PP-chopping in Spanish RCs, and the only study of a Romance
language, Portuguese, suggests that RPs and PP-chopping stand in complementary
distribution (Tarallo, 1986). I argue that the factors favoring RP presence vary according
to the syntactic position relativized, and that oblique RPs and PP-chopping are processes
independent from each other.
Early in the development of variationist methodology, Labov (1972) had pointed
out the difficulties in studying syntactic variables; in particular, their low frequency,
compared to phonological variables. RP-containing RCs in Spanish are very infrequent:
they represent only 14.5 per cent of the total number of RCs in my data. In order to study
the occurrence of RPs in subject, direct object (DO), and oblique RCs separately and
extensively, I gathered data from two different sources: 1) twenty-six hour-long
sociolinguistic interviews of native speakers of the Lima (Peru) Spanish dialect; and 2)
data from twelve hours of sessions of the Peruvian Congress.
xiv

Since 70 per cent of the proposed factors did not equally affect the presence of
RPs in subject, DO, and oblique RCs, this study suggests that RP occurrence is favored
by different factors depending on the syntactic function relativized within the RC.
However, some factors are common to the three syntactic positions relativized, in
particular, sentential antecedent, a factor that was omitted in most of the previous studies.
In addition, the results show that there is a subject-specific type of RP that performs the
pragmatic function of contrast. With regard to PP-chopping in oblique RCs, the results
suggest that, unlike Brazilian Portuguese, this strategy is not in complementary
distribution with RPs, but rather obeys factors related to the characteristics of the RC
antecedents.
This study contributes to the understanding of non-standard RC structures in
Spanish, and supports the general variationist assumption (Labov, 1972) that linguistic
variation can hardly ever be explained by a single factor. Regarding RPs, this study
presents evidence against implicational universal proposals like Comrie‟s (1981) and
Hawkins‟s (1994), suggesting that language-specific constraints must not be overlooked.
With respect to PP-chopping, this is the first study of such a phenomenon in Spanish, and
it shows that PP-chopping is favored by different factors than those shown to be valid for
Portuguese, and that it is a structure completely independent from RP occurrence.
1

Chapter 1: Spanish Relative Clause System and Variables

1.1 Introduction
Unlike most languages, in Spanish, resumptive pronouns (RPs) are allowed within
relative clauses (RCs) corresponding to all syntactic positions. An RP is a pronoun copy
within an RC, co-indexed with the antecedent NP head. However, in Spanish, this
possibility is restricted to RCs introduced by complementizer que (that, which), as shown
in (1)-(5), which include examples of RPs within subject, direct object (DO), indirect
object (IO), oblique, and genitive RCs, respectively.

(1)
a. Tenía una amiga
i
que ella
i
en la universidad leía cartas, le gustaba la Ouija...
1
(I)
„(I) had a friend
i
that she
i
used to read tarot cards at school, she liked the Ouija [board]‟
b. Tenía una amiga que en la universidad leía cartas, le gustaba la Ouija...
(I) had a friend
i
who Ø
i
used to read tarot cards at school, she liked the Ouija [board]‟
(2)
a. ¿Te acuerdas del pata
i
que lo
i
cuelgan del helicóptero? (J)
„Do (you) remember the dude
i
that (they) hang him
i
from the helicopter?‟
b. ¿Te acuerdas del pata
i
que cuelgan Ø
i
del helicóptero?
„Do (you) remember the dude
i
that (they) hang Ø
i
from the helicopter?‟

1
I have adopted the following convention regarding the source of the examples used in this study: 1) I =
interviews, 2) J = jotted-down after the recording session, 3) C = Congress sessions, 4) Author X, year:
page(s) for quotations, and no source information for made-up examples.

2

(3)
a. La congresista electa también, Mirtha Ramos, va a hacer uso de la palabra, va a
referirse también a los temas
i
que ella le
i
[sic] va a dar la prioridad. (C)
„The also elected congresswoman, Mirtha Ramos, will speak, (she) will also refer to
the topics
i
that she will give priority to them
i
.‟
b. La congresista electa también, Mirtha Ramos, va a hacer uso de la palabra, va a
referirse también a los temas
i
a los que ella Ø
i
/ les
i
va a dar la prioridad.
„The also elected congresswoman, Mirtha Ramos, will speak, (she) will also refer to
the topics
i
to which she will give priority Ø
i
/ them
i
.
(4)
a. Es una persona
i
que yo sí confío en ella
i
. (J)
„She is a person
i
that do I trust her
i
.‟
b. Es una persona
i
en (la) que yo sí confío Ø
i
.
„She is a person
i
that do I trust Ø
i
.‟
(5)
a. Ivancito tiene una amiga
i
que su
i
mamá tiene pelos en la cara. (J)
„Little Ivan has a friend
i
that his
i
mother has hair on her face.‟
b. Ivancito tiene una amiga
i
cuya
i
mamá tiene pelos en la cara.
„Little Ivan has a friend
i
whose
i
mother has hair on her face.‟

As can be seen, in each and all sets of examples, the first utterance is the one
containing an RP, whereas the second structure is the gap-containing counterpart. In
3

Spanish, the gap-containing structure is considered to be the standard one, and the RP-
containing one a non-standard form.
Moreover, in the oblique position, Spanish not only has the aforementioned RP –
gap alternation as in (6), but also an RP – PP-chopping alternation, as seen in (7).

(6)
a. No es algo
i
que le paguen por eso
i
. (J)
„It is not something
i
that they pay you for that
i
.‟
b. No es algo
i
por lo que le paguen Ø
i
.
„It is not something
i
that they pay you for Ø
i
.‟
(7)
a. La novia
i
, con que tiene 14 años Ø
i
, todos son ludópatas. (J)
„The girlfriend, that (he) has 14 years, they are all ludopathic.‟
b. La novia
i
, que tiene 14 años con ella
i
, todos son ludópatas.
„The girlfriend
i
, that (he) has 14 years with her
i
, they are all ludopathic.‟

This study focuses on two phenomena regarding Spanish relative clauses (RCs).
First, it aims at explaining the factors that govern the alternation of RCs containing RPs
and gap-containing RCs as shown in (1)-(5). Secondly, the study also focuses on the
factors causing the alternation between gap-containing and RP-containing RCs, on the
one hand, and RCs containing PP-chopping, on the other hand, as shown in (6)-(7).
The dissertation is organized as follows: in the remainder of this chapter, I present
the structure of Spanish RCs. In Chapter 2, I present the previous studies that have been
4

conducted regarding RP presence and PP-chopping from a cross-linguistic and a Spanish-
specific perspective. In Chapter 3, I present the data sets, the envelope of variation, the
hypothesis proposed to explain the phenomena, and the coding convention used in this
study. In Chapter 4, I show the quantitative results, the qualitative analysis, and I discuss
their interpretation. Chapter 5 offers the conclusions reached after analyzing the data, and
suggestions for future studies.

1.2 Spanish Relative Clause Marker Paradigm
Spanish has a variety of words that introduce RCs. Traditionally, grammarians
have indistinctively called all of them relative pronouns (Alcina et al. 1991, Alarcos
1994, Brucart 1999). However, they vary considerably in their form and in the
grammatical features that they display.
As can be seen in Table 1, the relative system in Spanish varies in gender, number and
case-marking. Some of the categories introducing RCs are marked for gender, number
and case (el cual, cuyo, cuanto); some are marked for number and case (quien); and some
only for case (que).
5

Table 1. Spanish Relative Clause Markers
RC marker gender number morphological
case
syntactic
case
quien, quienes _ + _ +
el / la / lo cual, los / las cuales
2
+ + _ +
cuyo, cuya, cuyos, cuyas + + + _
cuanto, cuanta, cuantos, cuantas + + _ +
(el / la / lo / los / las) que
3
_ _ _ +

We present in (8)-(13) some examples of RCs introduced by these forms,
relativizing an IO, subject, genitive, subject, oblique and DO, respectively:

 quien, quienes (who, that)
(8)
Los espectadores
i
a quienes
i
gustó la obra aplaudieron mucho. (Alcina et al., 1994: 1083)
„The spectators
i
to whom
i
the play was pleasing applauded a lot.‟

2
cual must be preceded by an article, otherwise, it would render an ungrammatical sentence.
3
que can be preceded by an article, especially when there is a preposition signaling case, without a change
in the referential meaning of the sentence:
i) La ciudad en que vivió
ii) La ciudad en la que vivió
“The city in which he lived”. (Alarcos, 1994:108).

6

 el / la / lo cual, los / las cuales (which, that)
(9)
Fue alertado [...] [sic] el ingeniero
i
de la fábrica, el cual
i
llamó por teléfono. (Alarcos,
1994:107)
„They alerted the engineer
i
of the factory, who
i
called on the phone.‟

 cuyo, cuya, cuyos, cuyas (whose)
(10)
Hombre
i
delgado [...] [sic], cuyos
i
ojos tenían rebrillos fugaces de cuchillo. (Alarcos,
1994:100)
„Thin man
i
whose
i
eyes had fugacious knife-shining.‟

 cuanto, cuanta, cuantos, cuantas (whatever, all that)
(11)
Los invasores
i
, cuantos
i
entraron en la aldea, estaban rendidos. (Alcina et al., 1991:
1097)
„All the invaders
i
that
i
entered the village were tired.‟

 el / la / lo / los / las que
(12)
Los amigos
i
con los que
i
salí de excursión acaban de llegar. (Alcina et al., 1991:1024)
„The friends
i
with which
i
I went on excursion have just arrived.‟
7

 que
(13)
Una cierta senadora
i
que Luis llamó Ø
i
. (Suñer, 1998: 337)
„Certain female senator
i
that Luis called Ø
i
.‟

Interestingly, as stated in 1.1, RPs occur in Spanish only when the RC is
introduced by que
4
, in particular, when it is not preceded by an article. Hence, (13) is the
only example of all of the above than can have a counterpart including an RP, which is
presented in (14):

(14) Una cierta senadora
i
que Luis la
i
llamó. (Suñer, 1998: 337)
„Certain female senator
i
that Luis called her
i
.‟

In the following section, I will present the different positions in the literature with
regard to the nature of this particular RC-introducing category.

1.3 The Nature of que
Despite the fact that a traditional analysis had always considered que a relative
pronoun, Lope Blanch (Brucart, 1999: 404) suggested a “de-pronominalization” of the
word, and its conversion into a complementizer, given the loss of its pronominal features.

4
Although De Mello (1992) mentions 3.2 1% of RP occurrence in RCs introduced by subordinators other
than que, they have not been attested in my data.

8

His rationale was that the „old‟ relative pronoun que had merged with the
complementizer que into a single unit. Hence, when an RC is introduced by the former,
an RP may occur to signal the function that the relativized antecedent occupies within the
RC, because que by itself fails to do so. In consequence, the word que in (15) and in (16)
would be two instances of the same complementizer with different functions: introducing
an RC, and introducing a completive clause (CC), respectively.

(15) El hombre
i
que Ø
i
vio
5
. (Lavandera, 1984: 84)
“The man
i
who Ø
i
saw”.
(16) Dijo que vio. (Lavandera, 1984: 84)
„(He) said that (he) saw‟.

Along these lines, Lavandera (1984) suggests that there is only one que, a
complementizer, and what RCs and CCs share in common is that the subordinating
category que is a “transparent” element “crossed by a function between the verb of a
clause (either the main clause or the subordinate clause) and either an element of the
other clause or the other complete clause” (1984:87).
The aforementioned choices depend on whether the clause introduced by que is
next to an NP or next to a different constituent. In the former case, the result is that the
verb of the subordinate clause (an RC, as in 15) completes its structure with external

5
Note that in (15) a subject RP may occur: El hombre
i
que él
i
vio todo (the man that he saw everything),
whereas in Dijo que él vio todo (He said he saw everything) the pronoun is not resumptive.

9

information included in the main clause, i.e. the antecedent; in the latter (a CC, as in 16),
it is the verb of the main clause the one that completes its information with the whole
subordinate clause.
Though her study is limited to restrictive RCs, Suñer (1998) also agrees on the
fact that the que that introduces RCs is not a relative pronoun. She argues that the choice
between using a relative pronoun or complementizer que depends on the head of the CP:
according to the author, C includes the feature [pronominal]. When C is [+ pronominal],
there is a relative pronoun that moves to Spec-CP attracted by the positive value of the
feature. However, when the feature is [- pronominal], there is no relative pronoun, but the
complementizer stays in situ. It is precisely in these cases in which there are no
pronominal features, that an RP may show up, to convey the missing information inside
the RC.
Solana (1996) agrees with Suñer in the sense that within restrictive RCs, que is
the head of CP, whereas relative pronouns (quien, cual, cuyo, etc.) occupy a different
position in the structure: they are located in Spec-CP.

1.4 Spanish Relative Clauses and Types
Relative clauses (RCs) are subordinate clauses headed by a relative pronoun,
adjective, adverb, or complementizer (Brucart, 1999), which function as modifiers of an
element called „antecedent‟. In its internal structure, an RC is always headed by a
subordinating category which, in turn, is anaphoricaly linked to the antecedent, by
functioning as an argument or adjunct within the RC. So, in (17), the complementizer is
10

que and the antecedent is the NP el libro, and the relativized element functions as a DO
inside the RC.

(17)
a. El libro que Luis me regaló era muy interesante. (Brucart, 1999: 400)
„The book that Luis gave me was very interesting.‟
b. El libro que me regaló Luis era muy interesante. (Brucart, 1999: 400)
„The book that to me gave Luis was very interesting.‟

According to Brucart, Spanish RCs have often been referred to as a construction
similar to qu-movement in questions. However, subjects in qu-questions have a fixed
position, i.e. they are post-verbal, as in (18a). By contrast, in an RC, there is a choice of
having either a pre-verbal subject as in (17a) or a post-verbal subject, as in (17b). The
sentences in (18) are unacceptable
6
when the subject is placed in a pre-verbal position.

(18)
a. ¿Qué libro te regaló Luis? (Brucart, 1999: 400)
„What book to you gave Luis?”
b. *¿Qué libro Luis te regaló? (Brucart, 1999: 400)
„What book Luis gave you?‟

6
18b and 18c can be acceptable under certain pragmatic constraints, but they do not follow the common
order of questions in Spanish.

11

c. *¿Luis qué libro te regaló?
„Luis what book gave you‟?

Given that RCs‟ mood is generally assertive, the verbs inside them tend to be
indicative. However, there are also RCs with subjunctive verbs and infinitival forms, as
in (19).

(19)
a. La participante que logre reunir todos los requisitos obtendrá el premio mayor.
„The participant that succeeds in meeting all the requirements will get the main prize.
b. Rosa precisa de un verdadero amigo con quien hablar.
„Rosa needs a real friend to talk with.‟

(19a) is used to convey a non-specific value to the NP, implying „any participant‟.
The RC in (19b), introduced not by the complementizer que but by the relative pronoun
quien, is also used in a non-specific context, in addition to expressing a modal value of
possibility, easily paraphrased by (20).

(20) Rosa precisa de un verdadero amigo con el que pueda hablar.
„Rosa needs a real friend that she can talk with.‟

12

Spanish RCs have been classified into different types, which are restrictive and
non-restrictive, coordinated, juxtaposed, complex, and with definite and indefinite
antecedents.

1.4.1 Restrictive and Non-Restrictive Relative Clauses
The classification of RCs into restrictive (specifying) and non-restrictive
(explicative) RCs is based on semantic, structural and intonational differences between
them.

1.4.1.1 Semantic Difference
The semantic distinction between the two clause types is that restrictive RCs limit
the meaning of the antecedent NP (i.e. they define the antecedent NP) enabling the hearer
to identify the referent of the antecedent NP head, whereas non-restrictive RCs bear
parenthetical information (description, modification, explanation) that does not limit the
meaning of the NP (Finegan and Besnier, 1989:235).

(21)
a. No sé dónde andarán los hombres que valen la pena. (I)
„(I) don‟t know where the men that are worthwhile may be‟.
b. No sé dónde estarán los hombres, que valen la pena.
„(I) don‟t know where the men, who are worthwhile, may be‟.
13

(22)
a. No somos tan instintivos como los animales que lo único que quieren es preservar la
especie, no?
„(We) are not as instinctive as the animals that only want to preserve their species,
right?‟
b. No somos tan instintivos como los animales, que lo único que quieren es preservar la
especie, no? (I)
„(We) are not as instinctive as the animals that only want to preserve their species,
right?‟

The semantic distinction between restrictive RCs and non-restrictive RCs can be
observed in the examples presented above. In (21a), the RC limits the meaning of the
class hombres, restricting it to a subset of that class: only those that are worthwhile. (21b)
would be its non-restrictive counterpart, and it means that men as a class are worthwhile.
In (22a), the RC restricts the meaning of the head animales, including only those that
want to preserve their species; in (22b), the non-restrictive RC is only ancillary
information added to the head, thus implying that all animals want to preserve their
species.

1.4.1.2 Structural/Syntactic Difference
Besides the semantic contrast between restrictive RCs and non-restrictive RCs,
presented above, there is also a second distinction between them: they have different
syntactic structures. Thus, restrictive RCs are modifiers of the NP head, whereas non-
14

restrictive RCs are adjuncts to the whole NP including the determiner. This difference
can be seen in the formal representations of (21 a-b), presented below in (23 a-b), adapted
from Ouhalla (1999).
(23)
a. Restrictive Relative Clause
NP
2
Spec N‟
2
los N Comp
5
hombres que valen la pena

b. Non-restrictive Relative Clause
NP
2
NP Comp
V 5
Spec N‟ que valen la pena

los N

hombres

In (23), the restrictive RC is a complement of the NP head hombres, and the
complex N‟ hombres que valen la pena in its entirety is modified by the determiner los,
rendering the meaning that only a subset of men are worthwhile. In contrast, in (23b), the
15

non-restrictive RC functions as an adjunct to the NP los hombres, in which case the
meaning is that the class composed of men is worthwhile in its entirety.

1.4.1.3 Intonational Difference (Pause)
The third criterion for distinguishing restrictive and non-restrictive RCs is the
acoustically perceptual difference between them: that there is a pause or an intonational
break (Gili Gaya, 1961: 57) between the antecedent NP and the non-restrictive RC. In
writing, this non-restrictive RC‟s feature is represented with commas, as in (21b) and
(22b), repeated here as (24a) and (24b), respectively:

(24)
a. No sé dónde estarán los hombres, que valen la pena.
„(I) don‟t know where the men, who are worthwhile, may be‟.
b. No somos tan instintivos como los animales, que lo único que quieren es preservar la
especie, no? (I)
„(We) are not as instinctive as the animals that only want to preserve their species,
right?‟

Pause, understood as a silence or lack of sound (Crystal, 2003), has been analyzed
by some to coincide with syntactic units. Also, they have been considered as markers of
the boundaries of phonic groups, which may coincide with short sentences or, if the
sentences are relatively long, they may coincide with subparts (not necessarily
constituents) of the sentence. This can be seen in (25).
16

(25) Aquellos cuentos de princesas encantadas / de tesoros ocultos y hadas protectoras //
entusiasmaban al auditorio infantil. (Gili Gaya, 1961: 56)
„Those tales of charming princesses / of hidden treasures and protective fairies //
filled the children‟s auditory with enthusiasm.‟

According to Gili Gaya, in (25) there are three phonetic groups delimited by silent
pauses: aquellos cuentos de princesas encantadas is the first one, de tesoros ocultos y
hada protectoras is the second one, and entusiasmaban al auditorio infantil is the third
one. However, there are only two basic syntactic constituents, roughly, subject and
predicate, indicated by the double slash.
It is precisely this type of pause that is found between some non-restrictive RCs and
their antecedents, according to Leal (1995: 8).

1.4.2 Coordinated Relative Clauses
Brucart (1999: 466-467) describes the characteristics of Spanish RCs
coordination. First, coordination can take place between RCs headed by different
subordinating elements. Thus, in (26) we can see that there are two RCs coordinated by
the conjunction y; the first one is headed by the relative pronoun cuyo, and the second
one by the complementizer que.

17

(26) Se trata de una persona cuyo padre se exilió en México y que regresó al país
recientemente.
(Brucart: 1999: 467)
„It is a person whose father was exiled in Mexico and recently returned to the country.‟

The second characteristic that Brucart mentions is that RCs corresponding to
different relativized positions can also be coordinated, as (27) shows. The first clause is a
subject RC, whereas the second one is a DO RC.

(27) Es un libro que se publicó el año pasado y que leí durante el verano. (Brucart: 1999:
467)
„It is a book that was published last year and that I read during summer.‟

The third characteristic is that whenever there are two coordinated RCs that
correspond to the same syntactic function, the subordinating element in the second one
can be dropped. In (28), there are two DO RCs coordinated by y, but the complementizer
que has been dropped due to the sameness of the relativized position.

(28) Aquella película que descalificaron en Venecia y luego premiaron en Cannes.
(Brucart: 1999: 467)
„That movie that they disqualified in Venice and then they awarded a prize in Cannes.‟

18

A fourth characteristic is that there has to be some semantic parallel between the
two coordinated RCs in order to drop the subordinator in the second one. Thus, Brucart
considers a sentence like (29) to have an unclear status, given that having being awarded
a prize in Cannes and having been seen at a London theater do not belong to the same
semantic field. In such case, he says that speakers would prefer a structure like (30),
without dropping the complementizer que in the second RC.

(29) ??Aquella película que premiaron en Cannes y vimos en un cine de Venecia.
(Brucart: 1999: 467)
„That movie that they awarded a prize in Cannes and we saw in a London theater.‟
(30) Aquella película que premiaron en Cannes y que vimos en un cine de Londres.
(Brucart: 1999: 467)
„That movie that they awarded a prize in Cannes and that we saw in a London theater.‟

It is worth mentioning that although (30) sounds more accurate than (29), the
latter cannot be ruled out as ungrammatical; what we have here is rather a distinction
between a standard coordination and a less standard coordination procedure.

1.4.3 Juxtaposed Relative Clauses
Although these structures are not mentioned in Brucart (1999), Alcina and Blecua
(1991), or Alarcos (1995), they are common in Spanish, and follow the same patterns
outlined for coordinated RCs in 1.4.2: juxtaposed RCs can be headed by different relative
markers, as in (31); RCs corresponding to different syntactic positions can also be
19

juxtaposed, as in (32); and when two RCs with the same function are juxtaposed, the
second subordinating element can be dropped, as in (33).

(31) Raúl tiene dos niños que son muy lindos, por los cuales es capaz de todo. (I)
„Raul has two boys that are cute, for whom he would do anything‟.
(32) Esas mujeres que las tienes en la casa, que están haciendo bulto. (I)
„Those women, that you have them at home, who are taking up space‟.
(33) César, que vino, vio y venció, es un personaje importante de la historia romana.
„Cesar, who came, saw, and conquered, is an important character in Roman history‟.

In (31) and (32), we see two examples of juxtaposed RCs. In (34), both RCs are
subject RCs that refer to the antecedent NP la siguiente pirámide; in (35), the first clause
is a DO RC, whereas the second one is a subject RC.

(34) la siguiente pirámide, que es la última, que es la del Sol. (I)
„the next pyramid, which is the last one, which is the Sun (Pyramid).‟
(35) una guitarra que me hizo él, que está ya arriba, colgada. (I)
„a guitar that he made me, that is already up there, hanging.‟

Although all the cases of juxtaposed RCs in my data are headed by the
complementizer que, it is perfectly possible to juxtapose an RC with another
subordinating element, such as de la cual, as in (36).

20

(36) una guitarra que me hizo él, que está ya arriba, colgada, de la cual tengo gratos
recuerdos.
„a guitar that he made me, that is already up there, hanging, of which I have nice
memories.‟

1.4.4 Complex Relative Clauses
Brucart (1999) defines complex RCs as the ones that include a sentential
boundary between the gap and the relative element. Sentences (37)-(40) are examples of
complex RCs.

(37) Un libro [que sería conveniente [que leyeras [_]]]. (Brucart, 1999:473)
A book that it would be convenient that you read.
„A book that it would be convenient for you to read.‟
(38) Un libro [que el profesor cree [que deberíamos [leer [_]]]].(Brucart, 1999:473)
A book that the professor thinks that we should read.
„A book that the professor thinks we should read.‟
(39) ¿Qué película nos recomendó Luis [que viéramos [_]? (Brucart, 1999:473)
What movie did Luis suggest that we watched?
„What movie did Luis suggest for us to watch?‟
(40) ¡Qué extraordinaria película nos recomendó Luis [que viéramos [_]! (Brucart,
1999:473)
What an extraordinary movie Luis suggested that we watched!
„What an extraordinary movie Luis suggested for us to watch!‟
21

As we can see in the previous examples, in Spanish there can be a number of
embeddings of clauses with inflected verbs that are different from English. Thus, in the
English translations of (37), (38) and (39), the verb in the RCs is an infinitive, as opposed
to the Spanish version, which includes inflected verbs. Also, in sentence (38), the
complementizer que cannot be deleted in Spanish, as opposed to English.
Irrespective of the degree of embedding in Spanish complex RCs, an RP is never
mandatory in complex clauses. In English, however, there is a special type of RP that
Prince (1990) calls „amnestying‟, because the structure rendered would be ungrammatical
without the RP. Some examples of these RCs are presented in (41) and (42).
(41) That asshole X, who I loathe and despise the ground he walks on, pointed out that…
(Prince 1990: 483)
„Ese desgraciado X
i
, que detesto y desprecio la tierra por la que Ø
i
camina, señaló
que….‟
(42) There are always guests who I am curious about what they are going to say. (Prince
1990: 482)
„Siempre hay invitados
i
que me pregunto qué Ø
i
van a decir.‟

In these last two English examples, the subject presence in the RC is mandatory.
In contrast, we can see that the RC verbs in their Spanish counterparts do not require an
overt (resumptive) subject, even in cases of multiple embedding in complex RCs.

22

1.4.5 Definite and Indefinite Noun Phrase Antecedents
The last classification to be mentioned divides RC NP antecedents into definite or
indefinite. In Spanish, as in English, definiteness is marked syntactically in NPs. Table 2
(adapted from Abbott, 2004: 123-124, 137) gives a summary of the Spanish categories
that introduce definite and indefinite NPs.

Table 2. Spanish Definite and Indefinite NPs
Definite Indefinite
 Personal pronouns: él, ella (he, she)
 Demonstrative pronouns / Nouns
preceded by demonstrative adjectives: este
camión, aquel (this truck, that one)
 NPs with determinate articles: el
niño (the boy)
 NPs with possessive adjectives /
possessive pronouns: mi hija, la tuya (my
daughter, yours)
 Proper names: Pedro, Raúl Alcázar
 NPs with a universal quantifier as
determiner: cada persona (each person)
 Nouns preceded by no articles:
problema
( problem)
 NPs with affirmative indefinite
adjectives: algún problema (some
problem)
 NPs with negative indefinite
adjectives: ningún problema (no problem)
 NPs with indeterminate articles: un
problema (a problem)
 Indefinite pronouns: alguien, nadie
(somebody, nobody)

23

The only criterion used in Spanish grammars when referring to definite or
indefinite NPs is this formal distinction: there are no pragmatic or semantic aspects taken
into consideration, but only the belonging to or exclusion from the aforementioned list
Following the previous table, in (43) we have examples of RCs that have definite
antecedents.

(43)
a. Ahí todo el mundo conocía a Cecilia Fonseca, pues, mi amiga que... te digo que vive
en la Alborada. (I)
„Everybody know there Cecilia Fonseca, my friend that… I have told you lives at the
Alborada Avenue.‟
b. Me ha llamado la hembrita de "Maelstorm", ese gru… ese fanzine que estaban
regalando. (I)
„The girl from “Maelstrom”, that grou…that fanzine that (they) were giving out,
called me.‟
c. El pata que salía a la calle, el reportero, ganaba más que cualquiera. (I)
„The dude that went to the streets, the reporter, made more money than anyone else.‟

By the same token, in (44) we have RCs whose antecedents are indefinite.

(44)
a. Ahí están unos materiales de una francesa que trabajó con el chipaya hace tiempo. (I)
„There are some materials of a French lady that worked on the Chipaya (language) long
time ago.‟
b. Tiene cosas que no me gustan. (I)
„(She) has things that I don‟t like.‟
c. No, no he escuchado a alguien que haya ido a la isla. (I)
„No, (I) haven‟t heard of anyone that has gone to the island.‟
24

Chapter 2: Previous Studies of Resumptive Pronouns and Prepositional Phrase-
Chopping

In the following section I will review some of the accounts that have been
proposed to explain the variation between RCs containing RPs, gaps, and PP-chopping.

2.1 Theoretical Studies

2.1.1 The Accessibility Hierarchy and the Resumptive Pronoun Hierarchy
Prediction
Hawkins (2004) suggests that the occurrence of RPs cross-linguistically is due to
the processing difficulty that relativized syntactic positions exhibit across the
Accessibility Hierarchy (Comrie, 1988).
(1)
Accessibilty Hierarchy (AH)
SU > DO > IO > OBLI > GEN

The AH states that if in a language one of the low positions on the scale (the ones
to the right) can be relativized, then all higher positions (the ones to the left) can also be
relativized. In language processing terms, a “high” position means that such a syntactic
function is easier to retrieve within the RC, whereas a “low” position means the opposite.
For example, in Spanish, all positions on the hierarchy are relativizable, a possibility that
is not shared by most languages in the world. In the particular case of RPs, the AH is
25

important for this study when it is complemented by the Resumptive Pronoun Hierarchy
Prediction in (14) (Hawkins, 2004: 186):
(2) Resumptive Pronoun Hierarchy Prediction (RPHP)
“If a resumptive pronoun is grammatical in position P on a complexity hierarchy
H, then the resumptive pronouns will be grammatical in all lower and more
complex positions that can be relativized at all.”

According to this view, in any language that allows RPs within RCs, the RP
hierarchy prediction should hold. This seems to be the case cross-linguistically, as shown
in Appendix 1. However, though they are a good starting point, the AH and the RPHO do
not suffice to explain the occurrence of RPs.
Along these lines, Hawkins proposed a means of measuring the “processing
weight” of the NP + RC compound by counting the words intervening between the
antecedent and the arguments of the RC needed to identify the function of the position
relativized.
The rationale underlying this proposal is that the larger the domain (i.e., the more
words that are needed to process the relativized function within the RC), the more
difficult the parsing of the utterance is. Therefore, he claims that RPs appear in order to
facilitate the cognitive task of processing such complex structures.
According to his account, depending on the occurrence of RPs, there are two
kinds of co-referential domains for the establishment of “processing weight”, each linked
26

to a lexical domain. These are a filler-gap domain, when there is no RP, and a head-
pronoun domain, when there is. The first of these is given in (3):
(3)
Filler-Gap Domain (FGD)
An FGD consists of the smallest set of terminal and nonterminal nodes dominated
by the mother of a filler and on a connected path that must be accessed for gap
identification and processing; for subcategorized gaps the path connects the filler
to a co-indexed subcategorizor and includes, or is intended to include, any
additional arguments of the subcategorizor on which the gap depends for its
processing; for nonsubcategorized gaps the path connects the filler to the head
category that constructs the mother node containing the co-indexed gap; all
constituency relations and coccurrence [sic] requirements holding between these
nodes belong in the description of the FGD. (Hawkins, 2004a: 175).

We have a Head-Pronoun Domain (HPD) when an RP shows up inside an RC.
The HPD‟s starting point is the head of the antecedent NP and it extends to the RP,
including all words intervening in between.
In addition, a Lexical Domain (LD) is the set of arguments of the subordinate
verb
7
. LD weight is calculated differently depending on whether there is an RP or a gap.
If there is an RP, its domain is local, in the sense that the RP is counted as an argument
within the RC. If there is a gap, the LD extends to the antecedent NP head outside the
RC.
These measuring calculations are illustrated in sentences (4) and (5) below. In (4),
the filler is que and the gap is the position from where the filler is considered to have
moved (in theories that propose movement in RC formation). There is co-referentiality

7
Occasionally, an LD may include adjuncts that are needed to process the structure of the RC.

27

between the filler que and the head patas, which is the antecedent of the RC. The FGD
goes from the antecedent‟s head, patas, to the verb he conocido, which is the clue word
needed to process the gap. This FGD consists of the counting of all the words contained
within the domain‟s boundaries. The LD for the verb he conocido starts with its object
argument patas and ends in the verb. By adding the words of the FGD and the LD, we
obtain the weight required for the processing of the RC.
(4)
unos patas
i
digamos de esa edad, mayores, que

yo ya he conocidoØ
i
ya.
--------------------------------------------------------------------------------
FGD: 1 2 3 4 5 6 7 8 9 10 11
-------------------------------------------------------------------------------
LD: he conocido 1 2 3 4 5 6 7 8 9 10 11 TOTAL = 22

Let us consider now the RC in (5), which instead of a gap includes an RP as a
local argument for the verb he conocido. In this case, given that there is no gap, the HPD
applies. The HPD begins with patas
i
and ends in los
i
. The LD for the verb he conocido
includes only the arguments within the RC.

(5)
unos patas
i
digamos de esa edad, mayores, que yo ya los
i
he conocido ya. (I)
-------------------------------------------------------------------
HPD: 1 2 3 4 5 6 7 8 9 10
-----------------------------
LD: he conocido 1 2 3 4 5 TOTAL = 15

The minimum domain for processing the relativized DO in (5) consists of 15
words. By contrast, when the RP is not present, as in (4), the domain increases
28

considerably, from 15 to 22. The RP strategy as in (5) is considered to be an easier means
of retrieving the function of relativized positions. What Hawkins calls “argumenthood
locality” or proximity of arguments to their verb, manifested in RPs, seems to be a
processing efficiency means in languages that have RCs either with no overt case-
marking for relative pronouns (morphological or syntactic), or RCs introduced by
complementizers, such as is the case of Spanish.
Hawkins‟s study is of particular interest for this study because it is a formal
account of the hypothesis that RPs surface due mainly to processing factors (Comrie,
1989:147). Moreover, it provides a basis for measuring the relative difficulty of
processing the antecedent of a RC, and it can be tested with a quantitative methodology.

2.1.2 Exceptions to the Accessibility Hierarchy
Joseph‟s (1983) study on modern Greek RCs is a criticism of Keenan and
Comrie‟s Accessibility Hierarchy account. His study contradicts the generalizations of
the Accessibility Hierarchy (AH) in three ways:
1. In Greek, RPs are found in each and every position on the scale, as opposed to the
almost inexistent presence of subject RPs attested by Keenan and Comrie.
2. The AH is not followed in Greek, where for some speakers it is the IO that
requires a mandatory RP.
3. Subject RPs show up not due to processing difficulty, but are rather due to
emphatic purposes. (p. 10)
29

The importance of Joseph‟s study is that Spanish has a very similar behavior to Greek
in all the three aforementioned characteristics. The first coincidence between
relativization in Greek and Spanish is that RPs occur in every syntactic position that each
language allows to relativize. Hence, Greek allows RPs in RCs in objects of comparison,
genitives, obliques, IOs, DOs, and subjects; and the same is true for Spanish except for
objects of comparison, which cannot be relativized.
Another striking coincidence is the one concerning IO RPs. Despite IOs not being
the hardest syntactic position to relativize, most Greek dialects demand an RP in this
position, whereas an RP is only optional in the lower positions of oblique, genitive, and
object of comparison. (6) shows an example of an IO RC with an RP. The subordinator is
the complementizer pu, and the RP is tu.

(6) kséro ton ánθropo
i
pu tu
i
édoses to vivlío. (Joseph, 1983 : 4)
(I) know the man that him (you) gave the book.
„I know the man that you gave the book to.‟

In Spanish, whose equivalent sentence would be (7), IO RPs are also mandatory
when the RC is introduced by the complementizer que
8
as stated above, but RPs are not
mandatory in the lower positions of oblique and genitive.

8
When the RC is introduced by quien (see 1.2, p. 4) , there is no need for an IO RP:
Conozco al hombre a quien diste el libro.
„(I) know the man to whom you gave the book‟.

30

(7) Conozco al hombre
i
que le
i
diste el libro.
(I) know the man that him (you) gave the book.
„I know the man that you gave the book to.‟
A subject RP, which would not be expected to show up according to the AH, due
to its being the easiest syntactic function to process when relativizing, is shown in (8),
which Joseph calls an emphatic pronoun.

(8) kséro ton ánθropo
i
pu móno aftós
i
méni s tin Aθína. (Joseph, 1983 : 4)
I know the man that only he lives in the Athens‟
„I know the man that is the only one that lives in Athens.‟

As we can see, the presence of the adverb móno signals the exclusiveness of the
referent that the property of living in Athens is applied to, ton ánθropo. Joseph also
mentions the fact that this subject RP, in contrast to the IO RP tu, is stressed, which is
another reason why he argues that it is an emphatic element. Just like in the case of IO
RPs, there is a very symmetrical counterpart for (8) in Spanish, as is shown in (9).

(9) Conozco al hombre
i
que sólo él
i
vive en Atenas.
I know the man that only he lives in Athens‟
„I know the man that is the only one that lives in Athens.
31

A fourth similarity between modern Greek and Spanish is that besides the
alternation of gap-containing RCs and RP-containing RCs headed by complementizers,
speakers of the former language also have a possibility of producing RCs headed by a
relative pronoun, o opíos (literally „the which‟). In (10) and (11), we can see some
examples presented by Joseph and their Spanish counterparts.

(10) kséro ton ánθropo
i
ton opíon
i
édoses to vivlío. (Joseph, 1983 : 4)
Conozco al hombre
i
al cual
i
diste el libro.
(I) know the man to which (you) gave the book
(11) sinándisa ton ánθropo
i
tu opíu
i
kséris to ónoma. (Joseph, 1983 : 5
Conocí al hombre
i
del cual
i
conoces el nombre
9
.
(I) met the man of the which (you) know the name

The last parallel feature between the modern Greek RC system and the Spanish
one is that, despite the fact that both of these languages offer the possibility of
introducing RCs headed either by an inflected relative pronoun or by an invariable
complementizer, the choice preferred by most speakers in casual speech is the latter.

9
Although this choice for a genitive RC is common in Spanish, and is considered more formal than its
counterpart with que and the possessive su, the more standard genitive RC construction is headed by the
genitive relative pronoun cuyo, as mentioned in 1.2.

32

In conclusion, although Joseph does not offer a quantitative study or a qualitative
explanation of why RPs occur in modern Greek RCs, besides his proposal that subject
RPs are emphatic, the facts he presents are very symmetrical to Spanish, and they suggest
that the occurrence of RPs in both languages cannot be explained exclusively in terms of
processing difficulty.

2.2 Quantitative Studies

2.2.1 Resumptive Pronouns in English, Yiddish, and Hebrew
One of the earliest quantitative studies on the occurrence of RPs within RCs was
Prince (1990) with Yiddish and English data. In accordance with the variationist
paradigm, she excluded from her analysis cases in which the occurrence of RPs was
mandatory, such as the case of English island-extraction sentences like (12).

(12) Apparently there are such things as bees
i
in the area which if you‟re stung by them
i
,
you die.
(Prince 1990: 483).

Prince‟s results showed that two factors are crucial in triggering RPs: restrictiveness
of the RC and definiteness of the antecedent NP. In order to account for these results, the
author argued for a hearer-processing explanation for the variation between the presence
and absence of RPs, which she calls the „file-card‟ account.
33

The starting point of Prince‟s account is that antecedent NPs and RCs are entities that
may belong either to the same files or to different files depending on the novelty
(definiteness) of the antecedent NP and the restrictiveness of the RC.
According to Prince, when there is an indefinite antecedent, which generally brings
forth a „brand-new‟ entity, the hearer must add a new file card to represent that entity or
construct
10
. On the contrary, a definite antecedent is either an entity already evoked in the
discourse, present in the hearer‟s knowledge store, or an entity that the hearer is assumed
to be able to infer on the basis of prior knowledge. Given its lack of novelty, no new file
needs to be created in such case. (Prince, 1990: 493)
With regard to their restrictiveness, RCs can be restrictive or non-restrictive.
According to Prince, the former belong in the same file as their antecedent NPs, and the
latter belong in different files.
Prince‟s claim is that non-restrictive RCs and indefinite restrictive RCs, both favoring
environments for triggering RPs in her data, form some sort of natural class: in both
cases, the NP head is retrieved from a different file-card than the RC. Non-restrictive RCs
are not necessary to determine the reference of the head NP and have been commonly
referred to as adjuncts: they simply add characteristics to the head; these characteristics
are in a different file card.
In Prince‟s view, indefinite restrictive RCs function more like non-restrictives, given
the „new‟ nature of their NP head. According to her, when RCs are attached to indefinite
heads, irrespective of their restrictive nature, they are retrieved by the hearer from a

10
Although Prince does not equate indefinite = new, definite = old, she does state that there is a tendency
of NPs to follow that trend.

34

separate file, just as in the case of non-restrictives. By contrast, definite restrictive RCs
do not demand the opening of a new file-card: the RC is information that must already be
on the file-card for the hearer.
Prince‟s study is important for a number of reasons. First, the file card hypothesis is,
to the best of my knowledge, the first processing account for the occurrence of RPs not
related to the AH. Second, the two RP-triggering factors involved in the account,
indefiniteness and non-restrictiveness, were proven to be relevant after a statistical
analysis. Finally, the contexts of RP variable occurrence were clearly distinguished and
isolated from the contexts in which RPs occur categorically.

2.2.2 Resumptive Pronouns and Prepositional Phrase-Chopping in Portuguese
Fernando Tarallo (1986) conducted a variationist study on the alternation of RP-
containing RCs, gap-containing RCs
11
, and RCs containing PP-chopping in spoken
Brazilian Portuguese. Although the gap-containing vs. RP alternation holds throughout
the entire syntactic scale, the three-way alternation occurs only when the relativized NP is
the object of a preposition, i.e. in indirect objects, obliques and genitives.
In (13a-b) we have the alternation between the RP-containing RC and the
standard gap-containing RC, and in (14a-b) we can see the alternation between the PP-
chopping RC and the gap-containing RC.

11
Tarallo does not use the term „gap-containg‟, but rather „piedpiping‟, which describes the syntactic
process rather than the final state of the standard RC.

35

(13a) Estava, lembra, con aquela sacolinha
i
, que você ia na faculdade com ela
i
. (Tarallo,
1986: 249)
I had, remember, that small bag, that you went to school with it.
(13b) Estava, lembra, con aquela sacolinha, com que / com a qual você ia na faculdade.
(Tarallo, 1986: 249)
(14a) Eu não ligo para piscina. É uma coisa que eu não ligo (para) (e)
i
. (Tarallo, 1986:
249)
I don‟t care for a pool. It is something that I don‟t care (for).
(14b) Eu não ligo para piscina. É uma coisa para a qual eu não ligo. (Tarallo, 1986: 249)
In order to explain this alternation between a standard variant and variants that
add forms (RPs) or subtract forms (PP-chopping), Tarallo assumed a functionalist
approach, according to which in order to preserve meaning, speakers should delay
processes that would entail the deletion of meaning-bearing forms.
Tarallo states that standard gap-containing RCs are losing ground to the non-
standard forms. Following the functional principle, the RP variant should be the strongest
candidate for replacing the dying standard form, rather than the PP-chopping variant.
However, as we can see in Table 3 (Adapted from Tarallo, 1986: 259), it is the
PP-chopping variant that is taking over most of the contexts.
36

Table 3. PP-Chopping, RP and Gap in BP RCs
Indirect
object
oblique genitive total %
gap
RP
PP-chopping
3
16
57
17
24
191
1
9
6
21
49
254
6.5%
15.2%
78.3%
Total 76 232 16 324 100%

In order to explain the unexpected results, Tarallo uses Poplack‟s (1979: 9-10),
formalization of the three possible outcomes of such a functionalist principle:
1. Phonological simplification may occur regardless of the grammatical status of the
segment at stake (the functional hypothesis is disconfirmed).
2. Morphological segments may be resistant to phonological weakening processes (the
functional hypothesis is confirmed).
3. There may be an interaction between morphological function and phonological
processes: the latter apply more frequently when there are no means of disambiguation
for the former. The results can be either marker deletion or retention (a weak form of the
functional hypothesis I confirmed).
Tarallo suggests that the third possible outcome, i.e. the weaker version of the
functionality principle, describes accurately the state of variation attested in spoken
Brazilian Portuguese, because the hypothesis cannot be either confirmed or disconfirmed
37

due to the primacy of PP-chopping over RPs, even in cases like (15), in which the loss of
phonic substance entails the presence of ambiguity.

(15) É para entender mesmo, daí é para ele não encher mais o saco, porque tive, teve um
que a professoura falou, onde já se viu falar que um aluno não é homem. (Tarallo,
1986: 250)
„He should understand (it) really, so that he would stop bothering, because there was
one that the teacher said, how come you say that a student is not a man.‟

In (15), due to the fact that more than one preposition could have been deleted,
there is an ambiguity whether the teacher said to the boy, in which case the PPs deleted
would have been a/para/con, or about the boy, in which case the PP deleted would have
been sobre.
Although he does not give examples, Tarallo found that a subset of contexts that
favor the occurrence of RPs are “structures which are more loosely woven into
discourse” (p. 253), such as non-restrictive RCs, intervening material between the head
and the relative subordinator, the greater distance between the head of the antecedent NP
and the „would be‟ gap, and the position of the relative with regard to the matrix. Thus,
he suggests that RPs show up to tighten the relationship between the antecedent NP and
the RC. Also, the semantic features of the NP head proved to be conditioning factors for
RP occurrence: human, singular and indefinite heads being the precise values that favor
their presence.
38

PP-chopping, on the other hand, stands in complementary distribution with RPs.
Its occurrence is strongly favored by restrictive RCs, and disfavored when there is a
considerable distance between the head of the antecedent NP and the gap within the RC;
likewise, it is favored by non-human and definite heads.
The aforementioned facts, and in particular, the greater occurrence of RPs in non-
restrictive RCs, lead Tarallo to propose that RPs are a modifier level strategy, while PP-
chopping is a complement-level strategy. In other words, Tarallo suggests that RPs tend
to occur in RCs that are processed only after their antecedents have been already
processed. Given that the antecedent was processed before, the function of the antecedent
within the RC can be hard to retrieve, therefore an RP is used to clarify that function. On
the other hand, PP-chopping tends to occur when the complex antecedent + RC is
processed at the same time, and the antecedent‟s function inside the RC is clear early on,
so the PP can be chopped.

2.3 Studies of Spanish Resumptive Pronouns and Prepositional-Phrase Chopping

2.3.1 A Pan-Hispanic Account
Trujillo (1990) proposes a theoretical semantic analysis of RP occurrence in non-
restrictive Spanish RCs. According to him, the great difference between restrictive and
non-restrictive RCs is that in the former, the antecedent is a bi-functional element that has
two syntactic roles: one in regard to the main verb, and another one in regard to the
subordinate verb, its natural constituent. Conversely, in non-restrictive RCs, there is no
39

functional simultaneity of the antecedent, but rather an overt function in the main clause,
and an implicit reference to the antecedent inside the RC. The consequence is that
restrictive RCs are considered a whole unit together with their antecedent, as opposed to
non-restrictive RCs, which are loosely related to their antecedents, and form no single
unit with them.
Thus, Trujillo (1990: 32, 37) states that this loose relation between a non-
restrictive RC and its antecedent, in particular, the fact that the antecedent is not a direct
or natural constituent of the verb structure in the RC, creates the possibility for an
anaphoric pronoun (RP) to emerge inside the RC to cover the otherwise only implicit
function of the antecedent inside the RC, as in (16) and (17).

(16) Es el libro
i
, que me lo
i
recomendó el profesor. (Trujillo, 1990: 32)
„It the book, that the profesor recommended it to me‟.
(17) Hace caso a su papá
i
, que lo
i
admira mucho‟. (Trujillo, 1990: 37)
„He pays attention to his dad, that he admires him a lot‟.

Moreover, Trujillo states that not only non-restrictive RCs allow the presence of
RPs, but also some restrictive RCs. In (16) and (17), both antecedents, el libro and su
papá, are definite. In order for a restrictive RC to admit an RP, its antecedent has to be
indefinite, as in (18) and (19).
40

(18) Es un libro
i
que me lo
i
recomendó el profesor. (Trujillo, 1990: 30)
„It is a book that the professor recommended it to me‟.
(19) Tenía algunas novelas
i
que no las
i
había leído. (Trujillo, 1990: 30)
„I had some novels that I had not read them‟.

According to him, the reason why these restrictive RCs allow the occurrence of
RPs is that the indefinite nature of the antecedents is complemented by the RP. In other
words: the RP is not a real pronoun, but rather a sort of lexicalized [+ definite] feature
that makes (18) and (19) equivalent to (20) and (21).

(20) Es el libro que me recomendó el profesor. (Trujillo, 1990: 30)
„It is the book that the professor recommended to me‟.
(21) Tenía las novelas que no había leído. (Trujillo, 1990: 30)
„I had the novels that I had not read‟.

In general, Trujillo‟s analysis of the phenomenon seems coherent, although the
examples that he uses do not belong to real utterances, but are rather made up. Therefore,
there are other types of contexts for RPs to occur in spontaneous data, as we will see in
2.3.3, (sentential antecedents), and in 4.5 (contrastive RPs) that his proposal may not
account for.
41

2.3.2 A Cross-Dialectal Study of Direct Object Resumptive Pronouns
George de Mello (1992) conducted a quantitative study of RP-containing RCs in
Spanish, using the transcriptions of many hours of interviews corresponding to 10
capitals of Hispanic American countries, plus the Spanish city of Sevilla. His study,
though, is focused only in DO RPs, therefore no claims with regard to subject or oblique
RPs were made
12
.
The linguistic factors whose frequencies showed more favorable contexts for the
occurrence of RPs were non-restrictive clauses and indefinite antecedents.
Just like in Trujillo‟s study, non-restrictive RCs favored the occurrence of RPs,
triggering 74% of the total number of RP occurrences. By the same token, indefinite
antecedent NPs represent 48% of the total number of RP occurrences. Conversely,
another factor that De Mello proposed, the greater distance between que and the RP
proved not to be a statistically significant context for RP occurrence, triggering only 23%
of the cases of RP in the corpora. The only social factor that was favorable for RP
presence was high class, although De Mello does not offer exact numbers on the
frequency of this factor.
Besides his quantitative results, De Mello offers qualitative explanations as to
why RPs can occur beyond the aforementioned contexts: emphasis, clarification, passive
voice replacement, IO / DO sequence, coordinated RCs, and juxtaposed RCs.

12
De Mello‟s study is not limited to RP-containing RCs headed by que. It also includes RCs headed by el
cual, el que and quien, but these altogether amount only to 3.2% of the total of RP-containing RCs.

42

De Mello proposes that some DO RPs are used to convey an emphatic stress to
que, suggesting that que + RP functions like a compound, a sort of stressed relativizer, as
in (22) and (23).

(22) La única ciudad que yo realmente recuerdo algo es Burdeos
i
, que lo
i
asocio a lluvias.
(De Mello, 1992: 32)
„The only city that I really remember is Burdeos, that I associate it with rain‟.
(23) Llegó un camión, lleno de unos caños
i
enormes, que los
i
han tenido que bajar con un
guinche.
(De Mello, 1992: 32)
„There arrived a truck, full of huge faucets, that they had to unload them with a crane‟.

Another factor he proposes is clarification, in the sense that que can introduce an
RC, but it can also introduce a different type of subordinate clause, a CC, as seen in 1.3.
Hence, in a sentence like (24), he argues, if we were to take out the RP, the subordinate
clause could be interpreted as meaning „because I had not practiced it‟, and the RP
guarantees its being processed as an RC by the hearer.
43

(24) Yo fui a Panamá enviada para aprender lo que era la laminación
i
a mano, que no la
i

había practicado.
„I went to Panama sent to learn what lamination by hand was, that I had not practiced
it‟.

Another possible factor, according to De Mello, is the replacement of an
otherwise passive structure with an active one including an RP. Thus, the idea is that a
speaker who inadvertently started to produce an RC that would end up like (26), repairs it
by treating it as if it were left dislocation, by including the RP and maintaining the active
verb, as in (25).

(25) Es un arma
i
que la
i
manejan así. (De Mello, 1992: 44)
„It is a weapon that they handle it like this‟.
(26) Es un arma que es manejada así.
„It is a weapon that is handled like this‟.

Another context that De Mello finds suggestive for RP occurrence is when there
is an IO pronoun in the RC. In Spanish, the rule is that when you have two object
pronouns together, the first one is an IO pronoun, and the second is a DO pronoun.
Examples like (27) and (28) show this type of environment.
44

(27) Va a ser una cosa
i
que nadie se la
i
va a quitar a uno nunca. (De Mello, 1992: 45)
„It is going to be a thing that nobody can take it away from you‟.
(28) Era la vivencia de poseer una lengua, y a la vez, absorber la cultura
i
que te la
i
está
dando.
(De Mello, 1992: 45)
„It was the experience of having a language, and at the same time, of absorbing a
culture that they are giving it to you‟.

Another type of context that De Mello finds plausible to be an RP trigger,
although not statistically significant, is the second (or third, and so on) RC in a sequence
of coordination or juxtaposition. (29) and (30) are examples of RCs coordinated by y, and
in both cases we see that RPs show up in the second RC, presumably due to the difficulty
of keeping track of the relativized position in the second RC.

(29) Un amigo mío me ha prestado una cosa
i
que es muy divertida y que la
i
estoy
leyendo.
(De Mello, 1992: 41)
„A friend of mine has lent me a thing that is really funny and that I am reading it‟.
45

(30) Consigo un abogado
i
, que estaba ejerciendo y que lo
i
nombraron subdirector de X.
(De Mello, 1992: 41)
„I get a lawyer, who was practicing, and that they appointed him vice principal of X‟.

(31) and (32) are examples of juxtaposed RCs, which despite the fact that they do
not have a conjunction adding an overt connection (i.e. addition, contrast, concession,
etc.) between them, they are presented in a sequence, such that the relativized element in
the last RC of the sequence would seem to present a greater processing challenge than the
one in the first RC, therefore triggering an RP.

(31) Esto que ves acá es un libro
i
que tengo en perspectiva, que ya lo
i
tengo listo. (De
Mello, 1992: 41)
„This one that you see here is a book that I have in perspective, that I have finished
it‟.
(32) Ya me conseguí un libro
i
de historia que me regaló el doctor Espinosa, que él lo
i

hizo‟.
(De Mello, 1992: 41)
„I already got myself a history book that Dr. Espinosa gave me, that he wrote it‟.

In general, most of De Mello‟s hypotheses seem plausible. However, I don‟t see
any reason for considering a stressed relativizer in sentences (22) and (23) as opposed to,
46

for instance, (31) and (32). It would seem more plausible that the RP in (22) is used to
disambiguate the antecedent: lo signals a masculine antecedent, Burdeos, instead of the
feminine ciudad. Likewise, in (23), los refers to a plural antecedent, caños, and not to the
singular camión.

2.3.3 Resumptive Pronouns in Santa Cruz (Spain)
Juana Herrera Santana (1990) offers a sociolinguistic study of variation between
RP-containing and gap-containing RCs in the directed interviews of 14 speakers of Santa
Cruz, in the Canary Islands. Although she does not give details about the specific
grammatical constraints that favor the presence of RPs, her findings show that the
syntactic positions that favor the occurrence of RPs within RCs are DOs and subjects,
with 36% of occurrence each. Also, she notices that 66.6% of the subject RPs have a
sentential antecedent, a factor that we will return to in chapter 4. The sentential
antecedents are underlined in (33) and (34).

(33) … y luego cubierta con una capa de estuco o cal o mortero de cal
i
, que eso
i
no es
nada más ni ni nada menos que mantener la nevera… (Herrera Santana, 1990: 551)
„…and then, covered with a layer of stucco or lime or a lime mortar, which that is like
keeping the fridge…‟
47

(34) Pues engancha la rueda de otro vehículo
i
, que esto
i
es… que eso
i
es pa‟ uno reírse.
(Herrera Santana, 1990: 551)
„Well, it attaches to the wheel of another vehicle, which this is… which that is
something to laugh about.‟

Also, she points out that these subject RP-containing RCs are non-restrictive. As
we have seen in Prince‟s (2.2.1), Trujillo‟s (2.3.1), and De Mello‟s (2.3.2) studies, non-
restrictive RCs seem to be a a favorable context for RPs to occur in Spanish. With regard
to the social factors, Herrera Santana found that men (60%) produce the RP variant more
than women (40%), and that the more educated classes produce RP-containing structures
(64%) more than the less educated classes (36%). As we will see in the following pages,
her gender-related findings were not confirmed in any of the subsequent Spanish RP
sociolinguistic studies.

2.3.4 Resumptive Pronouns in Santiago (Chile)
Silva-Corvalán (1996) offers a more detailed quantitative study of RPs within
RCs introduced by the complementizer que in Spanish, concretely, in the Chilean variety
of Santiago, the capital. Her study was partially similar to Prince‟s, but like Tarallo‟s and
Herrera Santana‟s, it included different factors such as the syntactic function of the
relativized element, animacy of the antecedent, and the distance between the antecedent‟s
head and the complementizer que. Besides the aforementioned differences in the factors
proposed, Silva-Corvalán‟s study differs from Prince‟s (and is closer to Tarallo‟s and
48

Herrera Santana‟s) in that it was based on data gathered by a uniform interview
procedure, whereas Prince‟s included data extracted from textbooks, novels, data
recorded from TV and radio shows, and utterances that she had jotted down after hearing
a speaker produce them.
Silva-Corvalán‟s results were similar to Prince‟s with respect to the factors that
they coincided in studying: non-restrictive RCs and restrictive RCs with indefinite
antecedents favored the occurrence of RPs. However, she disagreed with Prince‟s view
according to which the two aforementioned constructions belong to the same natural
class: Silva-Corvalán‟s data showed examples like (35), where the restrictive RC doesn‟t
seem to form a single unit (i.e. belonging to the same file) with its definite head.

(35) No se atreven a meterse en proyectos nuevos. Sí, ese es el problema
i
que se está
topando ahora la minería Ø
i.
(Silva-Corvalán, 1996: 392)
„They don‟t dare start new projects. Yes, that‟s the problem
i
that the mining industry
is having to deal (with) Ø
i
‟

On the other hand, in (36), the restrictive RC defines the property of the indefinite NP
antecedent‟s head (not just „any couple‟, but a particular type of couple), suggesting that
this is in fact a single complex unit and, furthermore, no RP shows up.
49

(36) [Primero está el marido] … en condiciones de un matrimonio
i
que Ø
i
tiene
comunicación.
(Silva-Corvalán, 1996:392)
„[The husband is more important] … in the normal circumstances of a couple that
does communicate.‟

These counterexamples, among others, lead Silva-Corvalán to suggest that RP
occurrence is not only a matter of belonging to the same or different file, but also to the
processing „weight‟ or difficulty the RC poses.
With regard to the syntactic position of the relativized element, DO RCs proved to be
the most favorable context (13.7%) over obliques (5.5%) and subjects (2.3%). Silva-
Corvalán clearly points out that these results do not go along the lines of the Accessibility
Hierarchy presented in 2.1.1, because according to it, obliques are harder to process and
therefore should be a more favorable context for RPs than DOs. Also, Silva-Corvalán
excluded IO RCs because the occurrence of a pronoun within the RC is mandatory, in
another clear disagreement with the Accessibility Hierarchy.
In general, the VARBRUL results ranked DOs as the context with the greatest
predictive power (.89), followed by indefinite antecedent (.74), non-restrictive RCs (.65),
intervening material (.68) and animate head (.63).
Silva-Corvalán‟s study was later replicated by Bentivoglio (2003) for the Venezuelan
variety of Spanish, and Samper et al. (2004) for the Canary Island variety. These two
studies are presented in the following sections.
50

2.3.5 Resumptive Pronouns in Caracas (Venezuela)
In her study of RP/gap alternation in RCs in Caracas, Venezuela, Bentivoglio
(2003) followed Silva-Corvalán‟s methodology for the most part, for comparative
purposes, and it is restricted to RCs headed by que. Like in the previous study, she
excluded IO RCs because RP presence is mandatory, but she also excluded antecedents
containing temporary expressions like el día, semana, mes, año, etc. (day, week, month,
year) because, according to her, they are lexicalized and never trigger RPs
13
. Bentivoglio
also included social factors in her analysis, such as age, gender, and socioeconomic
status.
Her results show that, with regard to the syntactic position relativized, DO RCs
favor more RP presence with a .93 probability of occurrence, followed by oblique RCs,
with .529, and with subject RCs being a non favorable context, results which in general
coincide with Silva-Corvalán‟s.
Other factors that confirm Silva-Corvalán‟s results are non-restrictive RCs (.773),
and features related to the antecedent: indefinite antecedent (.655), animate antecedent
(.605). With regards to the social factors, the only one that proved to be significant was
socioeconomic status, with a .626 probability for the highest group and very low
probability for the middle and lower groups. Bentivoglio concludes that, being the

13
Although Bentivoglio excludes oblique RCs with temporal expressions from her analysis because they do
not trigger RPs, there does in fact exist variation between gap-containing RCs and PP-chopping-containing
RCs, as i-ii show:
i) el momento en que tú te cansas.
ii) el momento que tú te cansas.
The moment that you get tired.
Therefore, I did include them in the study of obliques.

51

highest socioeconomic group the one that uses the non-standard more frequently, it seems
that it is not a stigmatized variant.

2.3.6 Resumptive Pronouns in Gran Canaria (Spain)
In 2003, Samper, Hernández and Pérez replicated the aforementioned studies of
Silva-Corvalán and Bentivoglio in the Spanish of Las Palmas de Gran Canaria. Just like
in the previous studies, they excluded IO RCs from the analysis due to their obligatory
nature, and they also excluded NP antecedents corresponding to time expressions.
Like in the Chilean and Venezuelan Spanish studies, Samper et al. considered the
following linguistic factors: syntactic function of the relativized element, restrictiveness
of the RC, and antecedent NP features such as definiteness and animacy. Besides the
aforementioned factors, Samper et al. included two new factors that, together, are
somewhat equivalent to Hawkins‟s FGD and PHD
14
: one the one hand, they took into
account the distance between the antecedent and the complementizer que (à la Tarallo),
and the distance between the complementizer and the „would be‟ gap. As opposed to
Bentivoglio‟s study, Samper et al. did not include any social factors in their study.
The results obtained in this study are in general parallel to the ones in the Chilean and
Venezuelan studies, but there are also some differences. Although Samper et al. do not
offer probabilities but rather frequencies, with the effect that there cannot be a factor
ranking, the results that resemble those of Silva-Corvalán and Bentivoglio are the

14
See 2.1.1.

52

following: non-restrictive RCs are a significant factor in favoring RPs, just like indefinite
antecedents, and material intervening between the antecedent and que.
With regard to the differences, first, the results for the Gran Canaria have an almost
identical distribution of frequencies for DO and oblique RPs, as opposed to the greater
frequency of DO RPs over oblique RPs in Chile and Venezuela. Secondly, animate NPs
do not favor RPs in a statistically significant way when compared to inanimate NPs,
while the Chilean and Venezuelan studies show a clear preference for RPs top show up in
RCs with animate antecedents.
In regard to the factor that was proposed for the first time, RCs in which the distance
between que and the would-be gap is greater, it proved to be a very favorable context for
RP occurrence, although Samper et al. do not specify what the cut-off point is to be
considered a „less distance‟ or „more distance‟ variant.
2.3.6 Resumptive Pronouns in Lima (Perú)
Cerrón-Palomino‟s (2006) Peruvian Spanish study also replicated Silva-Corvalán‟s
including almost all of the factors she had proposed. The factors including DO RCs, non-
restrictive RCs and indefinite antecedents proved once again to be significant, but not the
animacy of the antecedent.
The novelty of this study was that Hawkins‟s (2004) domains account was tested for
the first time using a quantitative methodology. Hawkins‟s account wasn‟t supported by
the data, i.e., RPs did not show up more frequently in larger domains than in shorter
domains. However, given that the idea that RPs do in fact ease the processing of complex
RCs had been rooted in the literature, Cerrón-Palomino (2006) combined Hawkins‟s
53

„domain‟ concept, with a more systematic version of Silva Corvalán‟s „weight‟ factor.
The new factor, as opposed to Hawkins‟s domains presented in 2.1.1, included only
words inside the RC, i.e., starting with the complementizer que, and did not take into
account any word outside the RC. The variants of this variable were material on both
sides of the verb, material to the right of the verb, material to the left of the verb, and
light (no material, only the verb). The „weight‟ factor is similar to some extent to Samper
et al.‟s factor of distance between que and the „would be gap‟ / verb, but it also analyzes
the entirety of material within the RC.
The results showed that the new factor was statistically significant, with the following
pattern: material to the left is the most favorable context, followed by material on both
sides, material to the right and lastly, light RCs. In sum, from a processing point of view,
this study supported the idea that RPs occur in Spanish in order to facilitate the hearer‟s
parsing of RCs.
In Cerrón-Palomino (2006), I also suggested that DO RPs and subject RPs do not
obey to the same reasons, namely that DO RPs occur due to processing ease factors and
subject RPs are caused by the pragmatic function of contrast. Subject RCs are the easiest
position to process (Hawkins, 2004), and still Spanish shows RPs in such positions. My
hypothesis was that subject RPs are used not for processing ease, but rather as a means of
singling out a specific subject referent from a set of possible candidates, as (37) and (38)
show.

54

(37) Tengo otras recetas
i
ahí, que esas
i
sí son recetas argentinas. (J)
„(I) have other recipes there, that those are indeed Argentinean recipes.‟
(38) Alan García apunta a una generación
i
nueva, que definitivamente ellos
i
no saben
nada. (I)
„Alan García aims at a younger generation that they do not know anything.‟

In (37) there is a competing set of different recipes that are not really
Argentinean, and the only one authentically Argentinean is singled out by the RP esas.
Likewise, in (38), there is a competing set of different generations politicians aim at, and
the new generation is the only one that Alan García (Perú‟s current president) aims at, a
fact singled out by the RP ellos.
However, due to the scarcity of the data, a separate study of DO RPs and subject
RPs could not be undertaken. In the present study, with a considerable increase of the
sample size, I am able to test this hypothesis.
After reviewing a number of studies that differ not only in their approach
(theoretical or descriptive), but also in the scope of their explanations (cross-linguistic or
language specific), in their methodology (researcher‟s own examples or real speech
recordings), and in the contexts in which RPs occur (RC, island, etc.), it seems safe to
conclude that there does not seem to be a single explanation that can account for all the
contexts in which RPs show up. If we accept this, then we are facing a phenomenon that
obeys multiple causation and is not amenable to one particular factor.
However, there seem to be some Spanish RP-triggering factors that are consistent
across the studies. These factors are the non-restrictive nature of some RCs and the
55

indefiniteness of the antecedent NP. Also, although the ways they have been
implemented for each study varies, it does not seem inadequate to claim that there is a
processing factor involved, and it has to do with distance, whether it is between que and
the verb/gap or between the antecedent and que.
Another important conclusion is that all the varieties of Spanish studied do in fact
show the occurrence of RPs, so it is a Pan-Hispanic phenomenon, and we can assume that
it will be found in any Spanish variety once its study in undertaken.
All the factors that have been proven to show significance will be analyzed in the
present study, as well as those that rendered some significant results in some studies and
did not in others, in order to test their significance with a larger data set.

56

Chapter 3: Adding and Subtracting in Spanish Relative Clauses

3.1 Methodology and Data
The main methodology used is the variationist sociolinguistics model, inspired in
Cedergren and Sankoff (1974). On the one hand, this quantitative methodology consists
in using statistical programs that allow the researcher to find correlations between the
independent variables (factors) and the dependent variable (the alternation being studied).
On the other hand, the statistical programs can generate a ranking of the significant
factors that are statistically significant, in accordance to their predictive strength. Thus,
this methodology allows not only the identification of the factors involved in the
alternation studied, but it also makes predictions about what contexts are the most and
least favorable for a particular variant to occur.
The statistical programs utilized in this study are SPSS 11.5 for Windows for the
crosstabulation and identification of the significant factors, and Golvarb 2001 for
Windows for the multivariate factor analysis.
The data analyzed in this study represents the speech of Peruvian Spanish
speakers from the capital city Lima. 56.3% of the data was elicited following the
sociolinguistics interview patterns, while 43.7% was elicited by post- recording
elicitations and congress sessions. There are four different sources of data, and they were
gathered at different times. A detailed account of these data sets is presented below.
57

3.1.1. Data Set 1
The first data set corresponds to the study of the speech of the Peruvian educated
class, and it was elicited in 1978 by Rocío Caravedo (published later, in Caravedo, 1989)
following the sociolinguistic interview patterns presented in Labov (1972: 85-99). There
was a uniform questionnaire consisting of four basic topics: childhood, education, work,
and travell. Particular questions arose at different points, depending on the responses of
the speakers and on how much they would engage in a dialog-like rather than a strict
interview interaction.
There were 12 people interviewed, in sessions ranging from 45 minutes to an hour. The
data that I used are the transcriptions of those conversations, as the audio files are not
accessible.
All the speakers that participated in this study had completed their college studies,
and all the interviews were conducted on an individual basis. I do not have information of
where each of the interviews took place, nor do I have any information of how long each
interview lasted (some transcriptions are considerably longer than others). The total
number of tokens containing RCs in this data set is 585. The specifications of the
speakers of this set of data are in Table 4.
58

Table 4. Description of the Speakers of Data Set 1

Sex of
Subject

Age at the time of
the recording
1 Male 70
2 Male 59
3 Male 46
4 Male 43
5 Male 37
6 Male 29
7 Male 26
8 Female 66
9 Female 60
10 Female 40
11 Female 39
12 Female 25

I decided to include this data set, twenty years older than the data sets presented in
the following lines, given that, when analyzed separately from data set 2, the overall
patterns were the same, which suggests no change in progress was attested during that
time span. Therefore, no bias is introduced when joining set 1 with sets 2, 3 and 4.

3.1.2 Data Set 2
The second set of data consists of transcriptions of approximately 17 hours
corresponding to personal interviews I conducted with 17 Peruvian native Spanish
speakers, from 2004 to 2007. The interviews follow the techniques utilized by Rocío
Caravedo when eliciting data set 1. In contrast with data set 1, data set 2 includes
speakers with a different level of education, because the study for which I gathered the
data entertained the hypothesis that the occurrence of RPs in RCs was greater in less
educated groups than higher educated groups, a hypothesis that was not confirmed
59

(Cerrón-Palomino, 2004). The total number of tokens in this data set containing RCs is
469, and the specifications of the speakers and recording settings are in Table 5.
Table 5. Description of the speakers of Data Set 2

Sex of
Subject

Age at
the time
of the
recording Education
Duration
of
Recording
Location of
Recording
RP tokens
outside
the
recording
1 Male 37 No college 45‟ House Yes
2 Male 36 No college 1h House Yes
3 Male 33 No college 45' Studio
4 Male 33 College 45‟ House Yes
5 Male 32 College 1h Studio
6 Male 31 No college 1h Studio Yes
7 Male 31 No college 1h House Yes
8 Male 29 College 1h Cafeteria Yes
9 Female 66 College 1h House Yes
10 Female 44 College 1h House Yes
11 Female 34 No college 1h House Yes
12 Female 33 No college 55' House Yes
13 Female 33 College 1h Office Yes
14 Female 31 College 45‟ House
15 Female 29 No college 50‟ House
16 Female 29 College 1h Cafeteria
17 Female 28 No college 45‟ Car

The last column in Table 5 is related to data set 3, in the sense that the same
speakers produced both sets of data. However, data set 2 corresponds to utterances that
were recorded acoustically, and data set 3 does not (see the following section).
3.1.3 Data Set 3
The third set of data, as stated above, was elicited from the same speakers that
produced data set 2, but after the recording session was over. RP and PP-chopping-
containing RCs in sets 1 and 2 were not as frequent during the interviews, due to the
60

observer‟s paradox (Labov 1972: 209) and the rarity of syntactic variables (Labov 1972:
204-205). Moreover, as it will be shown in chapter 4 (4.2), RPs do not tend to occur in
situations that may entail any type of formality, such as the sociolinguistic interview.
However, RP-containing and PP-chopping-containing RCs started showing up
after the recorder was turned off when the interview had ended, in accordance with one of
Labov‟s (1972: 88) findings: “The most frequent place for casual speech to emerge… is
at the end of the interview. It is perhaps most common when the interviewer has packed
away his equipment, and is standing with one hand on the door knob”.
I decided to jot down these RP and PP-chopping occurrences, to compensate for
their lack of presence during the sociolinguistic interviews. Even long after the interviews
had taken place, I would still write down RP and PP-chopping-containing sentences
uttered by the individuals that had originally participated in the sessions, whenever I had
the chance to speak to them. Thus, the jotted down data correspond mostly to RCs
including either RPs or PP-chopping, not to gap-containing RCs. The total number of
these casual RP-containing tokens is 231.
One shortcoming this set of data shows is its relative lack of context, when
compared to sets 1 and 2. Although I only included complete sentences in the data, the
sentences preceding the utterance at stake were not registered, which in consequence
makes most of the tokens in this set unfit for one of the factors in the following lines,
which is the previous mention of the RC antecedent NP, presented in 3.3.4.
Although the use of this data set could seem to have a biased impact on the
overall frequency of RP and PP-chopping occurrence, they represent only 10% of the
61

whole data. Moreover, David Sankoff (personal communication) agreed that the inclusion
of this data set would not distort the results.

3.1.4 Data Set 4
This data set corresponds to transcriptions of over 12 hours of debate of the
Peruvian Congress, which is composed of 120 members. The original source consists of
videos of Congress sessions shown in TV in 2007, aired by the Congress‟s own channel.
The inclusion of this new set of data is of great importance to this study, since it
has four advantages over the three other sets. First, it shows great contextual information,
which compensates for its lack in data set 3. Secondly, it includes sound files that sets 1
and 3 lack. Thirdly, these data lack the observer‟s paradox (Labov, 1972) problem that
individual interviews exhibit most of the time, since it consists of spontaneous
interactions between peers. It is worth pointing out here that the Congress members were
completely used to their sessions being recorded for the state television channel, since
this practice was in its fourth consecutive year. The fourth and last advantage that this
data set has over sets 1 and 2 is a consequence of the third advantage: it includes a
considerable number of RPs and PP-chopping because Congress sessions were not
considered formal contexts as the sociolinguistic interview was.

3.1.5 The Nature of the Data
The tokens that I used in this study reflect actual speech, and the data have been
transcribed as faithfully as possible. In other words, I have not cleaned or polished the
utterances in order for them to appear as a more standardized version than what was
62

actually produced by the speaker, thus including discourse markers as (1) or false starts
as (2).

(1) Para agregar, en todo caso, eh… esta responsabilidad y sanción cuando, en periodos
electorales, se excedan estos topes
i
que la misma Ley los
i
está estableciendo. (C)
„In order to add, in any case, uh… this responsibility and penalty when, in electoral
periods, these limits that the law is setting set are exceeded.‟
(2) Efectivamente se lee pues este... el informe que nos tiene que pasar a la Comisión de
Salud. (C)
„Effectively, um… the report that the Health Committee has to hand us will be read.‟
(3) Si incurrimos en este tipo de gasto, deje… dejamos de hacer más carreteras, más
viviendas, dejamos de atender a la educación, Ø que tenemos problemas…(C)
„If we incur in this type of expenses, we stop… we will stop building more freeways,
more houses, we will stop taking care of education, that we have problems‟.
(4) Si pudiera recoger otra… una disposición complementaria que quiero alcanzarle, Sra.
Denisse Chuquival… (C)
„If you could receive another… a complementary disposition that I want to hand you,
Mrs. Denisse Chuquival.‟

In (1) and (2) the presence of the discourse markers eh and este, respectively, do
not affect at all the structures being studied i.e., the RCs contained in these examples.
Likewise, the hesitation when uttering the first main verb in (3) or when uttering the
63

determiner of the antecedent‟s NP head in (4) has no effect on the RCs included in the
sentences.
However, there were some other utterances that were not included in this study
because of other reasons that made them inappropriate for the purposes of this study,
mainly because some of their characteristics made it impossible to analyze them through
the factors proposed. In this section, I will present the types of tokens that were excluded
from this study and the reason why they were excluded.

3.1.5.1 Incomplete Sentences
Given that the data used correspond to actual speech, there were many hesitations,
false starts, and structures that were not completed by the speakers. Some of these tokens,
like (5) and (6), do not affect the production of the RC that they include, therefore, they
were not taken out from the study.
In (5), there is a missing adjunct of the verb va a pasar, but it does not
compromise the antecedent of the following RC, el proceso electoral and its being the
DO of the subordinate verb va a tener, thus this token is suitable for coding in all the
factors proposed in this study. Likewise, in (6), despite all the speaker‟s hesitation and
incomplete structures, the RC que nos dé buenos resultados can be analyzed with no
problems, mainly because its antecedent, una comi… comisión eh.. eh... libre is
recoverable, so this token is fit to be subject to the analysis of all the proposed factors.

64

(5) Esto va a pasar al... después del proceso electoral que va a tener, donde ya se conoce
que los demócratas, que se oponen al Tratado de Libre Comercio, van a ser mayoría
en el Congreso. (C)
„This will go to the… after the electoral process that it will have, where we already
know that the Democrats, who oppose the Free Trade Agreement, will be the majority
in the Congress.‟
(6) y poniéndose a dos miembros interesados, de… de los mismos trabajadores, creo que
no… trabajadores y despedidos, creo que esto desviaría completamente la posibilidad
de tener una comi… comisión eh.. eh... libre y que tenga eh.. buena… buenos... eh...
que nos dé buenos resultados. (C)
„and putting two interested members out of… of the workers themselves, y think that
it would not… workers and those laid off, I think this would alter the course of
having a commi… a committee um… uh.. free and which has uh.. fine… good uh…
which can give us good results.‟

However, there were other incomplete structures that either posed a problem in
analyzing the factors proposed for the occurrence of RPs, or were impossible to assign a
structure by the researcher. For example, (7) shows an incomplete RC, which lacks the
DO announced by the transitive verb dice. Given this lack, this token could not be
classified for the factor weight (3.3.8 in this chapter). Another case of a missing DO is
(8), in which only the determiner el is uttered twice, but not the head of the NP that
would function as a DO. In (9), the speaker neither clearly states the antecedent of the
RC, nor is the DO of the verb hizo complete, because only the determiner este is
65

mentioned. There is uncertainty about the antecedent, in the sense that we do not know
whether it is the ministry or the minister, which means that this token cannot be coded for
animacy (3.3.6) The incomplete DO poses the same problem that (7) and (8) entail, i.e.
the impossibility to code this sentence for the factor weight.

(7) En segundo lugar, Sr. Presidente, yo creo que los parlamentarios deben pensar
claramente qué es lo que van a votar cuando aprueben una norma que dice… (C)
„Secondly, Mr. President, I think that the congressmen have to think clearly what it is
that they will vote on when they approve a regulation that states…‟
(8) El...pedido de la congresista Doris Sánchez, que sugiere el-el… y ha dado las
observaciones en el inciso N del artículo 10, en el 'desarrollo, proposición y emisión
de la formativa'. (C)
„Congresswoman Doris Sánchez‟s petition, that suggests the-the… and has provided
the observations in the development, proposition and issue of the assessment.‟
(9) A pesar de que todos los congresistas hicieron presente que se apoyaba el presupuesto
del Ministe… del minis que hizo este... el Ministro de Salud, en su exposición, de los
480 millones, faltaba hacer la votación correspondiente para poder eh… enviar esta
con un oficio, ¿no? (C)
„Despite that all the congressmen stated that they were supporting the budget for the
ministr… of the minis… that did this… the Health Minister, in his speech, of the 480
million, the corresponding voting to…send this official letter‟.
66

There were some sentences like (10), which were incomplete because they were
re-phrased or restarted at a point in which the RC had not been completed. Moreover,
given the abrupt restart of the last sentence, the last acoustic portion of the RC could
not be identified, because it could have been either the preposition con (with) or the
first syllable of any other lexical item such as consecutive or común that was
truncated due do the restart.

(10) Era el momento, la fecha que era el deadline para elegir la carrera de letras, que era
un trámite con – tenía que hacer un trámite yo. (I)
„It was the moment, the date that was the deadline to choose a Letters career, which
was a procedure with [?]… I had to go through a procedure.‟

Tokens like these were excluded from the study because their unfinished
structures do not fit the requirements of some conditioning factors proposed in this study.

3.1.5.2 Unclear Sentence Structures
There were some utterances that included structures that were hard to process, and
whose syntactic functions were either very difficult to retrieve or difficult to classify and
were thus removed from the data. In (11), for example, the relativized antecedent, único,
does not have a direct syntactic function in regard to the RC está. In other words, this
sentence‟s thematic structure could be rephrased, without using an RC, as está en
posición de discutir lo único („[he] is in charge of discussing the only thing.‟), in which
case it is evident that what has been relativized is the DO of a non-inflected verb
67

(discutir) that is a constituent of a higher ranked PP (en posición). Because it would have
been odd to classify this RC as a DO RC, especially when the subordinated verb is a
copulative verb, this token was excluded from the data.

(11) de manera que lo único que está en posición de discutir es si son cinco delegados de
trabajadores. (C)
„hence, the only thing that [he] is in charge of discussing is whether there are five
representatives of the workers‟.

Another possible analysis of (11) would be to consider lo único as the passive-
reflexive subject of the infinitive discutir, with the meaning „the only thing that can be
possibly discussed‟. However, in such a construction, the pronoun se would be expected
to be added to the infinitive, rendering discutirse instead of discutir.
In (12), the inclusion of the preposition in a la que todos estamos solicitando
makes it impossible to determine the syntactic function of the RC. If it were a DO RC, in
which case the meaning of the RC would be „the transparency that we all demand‟, then
the rule is not to include a preposition when the DO does not refer to a person. Precisely,
the antecedent that is relativized in this clause is transparencia, which does not refer to a
human being, however, the Spanish DO personal a is present. Assuming that it is not a
DO RC but rather an IO RC, in which case it would imply that we are demanding
something else from the transparency, the DO that would make explicit what it is that we
demand is not present.
68

(12) Creo que de eso se trata, de un… reordenar todo y llegar, pues, a una transparencia a
la que todos estamos solicitando. (C)
„I think that is what it is all about, about a… reorganizing everything and arriving,
thus, to a transparency [to] which we are all demanding.

In (13), given that the complementizer que also introduces subordinate clauses
that are not RCs, the clause que no se les ha repuesto is really hard to assign a syntactic
function. In principle, it would seem to be an RC in which the relativized element is
either quejas or Banco de Piura. Assuming the former possibility, the meaning would be
that the complaints had not been replaced, which does not make sense; assuming the
latter, it would mean that the Bank of Piura has not been reinstated, which doesn‟t make
sense either.

(13) Por ejemplo, yo tengo muchas quejas del Banco de Piura, que no se les ha repuesto,
incluso habiendo las vacantes y estando recibiendo personal nuevo. (C)
„For example, I have many complaints from the Bank of Piura, that they have not
been reinstated / replaced, even when there were vacancies and hiring new
personnel‟.

The last attempt at interpreting the relation between the aforementioned antecedent
and its RC would be to assume that the plural IO pronoun les refers to an implicit NP
inside a PP, quejas [de empleados] del Banco de Piura („complaints of employees of the
69

Bank of Piura‟), in which case the antecedent would not be quejas, but rather empleados,
as shown in (14).

(14) Por ejemplo, yo tengo muchas quejas [de empleados
i
] del Banco de Piura, que no se
les
i
ha repuesto, incluso habiendo las vacantes y estando recibiendo personal nuevo.
„For example, I have many complaints [from employees
i
] of the Bank of Piura, that
they have not reinstated / replaced them
i
, even when there were vacancies and
hiring new personnel‟.

However, given that I cannot be sure about what the speaker‟s intention was when
producing such an utterance, I decided to exclude it from the analysis.
In (15), we have another example of an unclear structure. In the second sentence, the NP
head estudio, present in the previous sentence, is implicit, and it would seem that it is the
antecedent of a subject RC. However, inside the RC, we have the plural verb fueron,
which seems to agree in number with cuatro años, the adjunct NP inside the RC. This
means either that there is no antecedent, and therefore no relation between the RC and
any component of the main clause, or that the clause at issue is not an RC, but rather a
different type of subordinate clause. Another possibility in interpreting (15) would be that
estudio is the implicit antecedent and subject of the verb fueron. The fact that some
copulative verbs can agree in number with their predicates could be an argument in favor
of this interpretation. In sum, the unclear structure of this sentence makes it unsuitable for
this study.
70

(15) Terminé ya el estudio del idioma en la Alianza Francesa. El [estudio] de quechua se
dio en la universidad, que fueron cuatro años. (I)
„I have already finalized the study of the language at the Alliance Française. The one
on Quechua took place in college, which were four years.‟

3.1.5.3 Hard to Classify Tokens
There were also some tokens that could not be coded, and as a consequence had to
be excluded from this study. For example, in (16) we have a coordinated IO RC which is
not common in two respects. On the one hand, it lacks the preposition (a contraction, a +
el = al) that its standard counterpart would include before the complementizer que i.e. al
que, in which case this could have been considered an instance of PP-chopping; on the
other hand, there is an almost categorical use of pronouns referring to IO antecedents as
will be seen in 3.2.3, by which pattern the non-standard but common counterpart to the
second RC in (16) would be que no le tienes confianza. In other words, this would be an
example of PP-chopping and RP-chopping, that I have no means of accounting for in the
coding for this study.

(16)¿Cómo decirle algo a alguien que tú lo quieres y Øque no tienes confianza? (J)
„How can you say something to someone that you love him and you don‟t have trust?‟

In (17) we have an oblique RC (introduced by the infinitive contar con, „to count
on‟) that could have been coded as a gap-containing RC in the sense that it lacks an
71

oblique RP. However, a genitive RP (see 3.2.4 for more details) is included inside the PP
that functions as an oblique, making it unfit to be coded in my study.

(17) Primero, para agradecer que el miércoles pasado asistieron a la Comisión eh…
primero, el Presidente, que fue muy importante contar con su asistencia, con su
presencia… (C)
„First, I want to thank that last Wednesday, the attendants to the Committee session
were um… first, the President, that it was really important to count on his attendance,
on his presence‟.

3.1.5.4 Relative Clauses with No Overt Subordinator
There were cases in which an RC did not have an explicit subordinator, i.e. RCs
that were not headed by a que
15
. This happens when they are coordinated RCs, generally
linked by the conjunction y, and the second RC lacks the presence of the complementizer
que, as in the italicized RCs in (18), (19) and (20).

(18) De tal manera que este Congreso pueda… pueda poner... pueda poner candados
firmes para que no saquen leyes que son declarativas y, finalmente, no se cumplen.
(C)
„So that this Congress is able… is able to put… is able to put firm restraints in order
not to produce laws that are declarative and, ultimately, are not executed.

15
See 1.4.2 for more details.

72

(19) Me pone nerviosa, muchas… muchas cosas que no podía hacer todavía y ya estaba
grande para hacerlas. (I)
„I became nervous, many… many things that I was not able to do yet, and I was old
enough to do them.‟
(20) Ahora, cosas que yo quisiera volver a hacer ahora, pero no tenía la mentalidad para
disfrutarlas. (I)
„Now, things that I would have liked to do over now, but I didn’t have the state of
mind to do them.‟

An overt complementizer is a requirement to be coded for the distance (3.3.5) and
the weight (3.3.8) factors, and because of this lack, the utterances above were discarded.
Even (18) and (19), which contain RPs, were not taken into consideration for the
coordination of RCs factor (3.3.9).

3.2 Envelope of Variation
A crucial part of a variationist study is to establish the contexts in which the
variable occurs: in order to determine what factors cause such variations, these variable
contexts have to be isolated and made explicit. But when isolating the contexts in which
there is variation, the contexts in which the alternation between variants does not occur
will be made evident. These contexts are called categorical contexts (Tagliamonte 2006:
86). In a categorical context, the variant of interest either appears 100 per cent of the time
or is absent 100 per cent of the time, which shows a clear case of lack of variation.
73

Likewise, there are other contexts that are not categorical, but are very similar, in
the sense that they show around 95 per cent of presence or absence of the variant of
interest. These contexts are called near categorical (Tagliamonte 2006: 87)
In this section, I will specify the types of tokens that were removed from the
quantitative analysis, because they appear in categorical or close to categorical contexts,
and therefore had to be excluded from the quantitative study.

3.2.1 Headless Relative Clauses and Headless Antecedents
By headless RCs I refer to those that do not have an overt NP functioning as an
antecedent, but rather a definite article. The article can be neuter as in (20) and (21),
feminine as in (22), or masculine as in (23) and (24).

(20) Acá parece que no se quiere entender lo que dice el dictamen. (C)
„Here it seems that nobody wants to understand what the ruling says‟.
(21) lo que ya está en las listas y lo que se plantearía ahora, que no llega a tres mil. (C)
„what is already in our lists and what would be proposed now, which does not even
make three thousand.‟
(22) La que llevaba el burro por delante eran (sic) unas mujeres vestidas todas de negro.
(I)
„The ones that followed the donkey were some women dressed all in black‟.
(23) Un español
i
que lo
i
acusaron de sexual harassment (...) El
i
que él
i
mismo contó. (J)
„A Spaniard that they accused of sexual harassment (...) He, the one who said (it)‟.
74

(24) El Duce
i
es el
i
que fuiste con él
i
a la fiesta. (J)
„El Duce is the (one) that you went to the party with him‟.

The reason to exclude this type of RC from the quantitative analysis is that it is a
semi-categorical context for the occurrence of gap-containing RCs. In fact, there were no
PP-chopping occurrences and only two cases of RP presence in such contexts, shown in
(23) and (24). In (23), we have a subject RP that is co-referential with the article of the
headless RC, the previously mentioned NP un español, and the RP lo present in the
preceding sentence. On the other hand, in (24), we have an oblique RP, which is co-
referential with the article of the headless RC and the subject of the main verb, el Duce.
There were no cases of oblique RPs in headless RCs.
An example like (21) includes another type of token that was excluded: an RC
whose antecedent consists of headless RCs. The coordinated headless RCs lo que ya está
en las listas and lo que se plantearía ahora are the antecedents of the RC que no llega a
tres mil. Very few tokens of this type showed up in the data, and their rate of RP or PP-
chopping presence was 0 per cent.

3.2.2 Passive Constructions
RCs with passive verbs were also taken out from the quantitative analysis of
subject RCs, because the absence of subject RPs in such contexts was categorical. With
regard to the analysis of DO RCs, for obvious reasons, it is impossible for DO RCs to be
generated by passive verbs, therefore no DO RPs could be produced in such a context.
75

Examples of subject RCs that include a passive verb and lack a subject RP are
presented in (25) – (27).

(25) Primero, el plazo necesario para dar el debate y para poder opinar sobre los puntos y
sobre el texto que sería propuesto. (C)
„First, the necessary deadline for the debate to take place and to express opinions
about the matters and about the text that would be proposed‟.
(26) Por lo tanto creo que, por un debate, que ha sido aprobado por unanimidad... (C)
„Therefore, I think that, due to a debate that has been approved unanimously…‟
(27) Sin embargo, hay casos de transfuguismo, travestismo político, que ni siquiera son
considerados por el Congreso ni menos sometidos a la justicia penal. (C)
„Nonetheless, there are cases of turncoats, political transvestites, that are not even
considered by the Congress and even less so subjected to penal justice‟.

Similarly to the case of passive subject RCs, oblique RPs did not show up with
passive verbs. However, 50 per cent
16
of oblique RCs with passive verbs did trigger PP-
chopping, and that is the reason why they were kept only for the oblique RC analysis.
In (28) we can see that the oblique RC that functions as an adjunct of cause for
the passive subject este dictamen is the standard one, i.e. there is no RP within the RC,
and there is no PP-chopping as the preposition por is heading the PP.

16
This does not necessarily mean that passive verbs are a favorable context for PP-chopping occurrences.
The total number of passive oblique RCs was 4, and 2 of them did show PP-chopping. The tokens are too
few for the type of verb to be considered a statistically significant factor.

76

(28) Por eso Presidente, yo le pido que revise la cuestión previa por la que fue sometido
este… este dictamen. (C)
„For that reason, President, I ask you to revise the previous question by which this
ruling was submitted‟.

In (29), we have a case in which the oblique RC triggers PP-chopping with a
passive verb. The standard PP-including version would have been el objetivo para [el]
que fueron creados.

(29) Por supuesto que debería el Estado preocuparse, pero no con impuestos antitécnicos
y anticonstitucionales, que lo único que han hecho es engrosar las planillas de
muchos de los ministerios, no han ido a… a… para el objetivo Ø que fueron
creados. (C)
„The Administration should certainly look after, but not with anti-technical and ant-
constitutional taxes, which only broaden the payrolls of many of the ministries,
without going for… for… to the target that they were created for‟.

3.2.3 Indirect Objects
All indirect object (IO) RCs were excluded from the quantitative analysis,
because the pronoun occurrence was categorical. (30) and (31) show occurrences of the
RP le.
77

(30) A veces puede ser una persona
i
, digamos, con conocimiento, pero que las dos cosas
realizadas al mismo tiempo no le
i
permiten darse íntegramente al… al niño, ¿no? (I)
„Sometimes it can be a person, say, with the knowledge, but that the two things
being done at the same time do not allow her to devote herself to the… the child,
right?‟
(31) No así tanto Río de Janeiro, que es una ciudad
i
muy agradable pero que le
i
falta
sabor. (I)
„Not really, Rio de Janeiro, which is a very nice city but that it lacks flavor.‟

However, this RC structure can be paralleled to the main clause pattern attested in
all Spanish dialects, in which the IO pronoun shows up in an almost mandatory fashion.
Thus, in (32) the pronoun les co-occurs with the full-fledged IO a ustedes, and in (33), le
precedes the IO al director.

(32) A veces les
i
hablan a ustedes
i
de las variaciones como la de Brahms y cosas por el
estilo.(I)
„Sometimes [they] talk to you about variations like Brahms‟s and the like‟.
(33) Yo me he instalado en mi casa y le
i
mandé un oficio al director
i
diciéndole que… lo
que necesito. (I)
„I have settled home and I sent the director a letter telling him what I need‟.
78

As can be seen, the object pronoun that shows up in IO RCs is the same one that
can be found in main clauses, and in both clause types it is mandatory
17
. It is thus hard to
conceive it as an RP, but it rather appears to be functioning as an IO agreement mark.

3.2.4 Genitives
In general, genitive RCs were very scarce, making it impossible to achieve
statistically significant results for the alternation of gap-containing genitive RCs and RP-
containing genitive RCs. Besides that, the typical genitive RC variation was between
sentences like (34) and (35). In (34) there is no que RC, but rather cuyo, a relative
pronoun presented in 1.2. The RP counterpart is introduced by the complementizer que
and it includes a possessive adjective that functions as the resumptive element, as can be
seen in (35).

(34) Juan Ramón Jiménez, cuyo Platero es una de las obras más importantes que hay en
la historia de la literatura española. (I)
„Juan Ramón Jiménez, whose Platero is one of the most important books there are in
the history of Spanish literature‟.

17
In Peruvian Spanish, utterances like i) and ii), which are the gap-containing counterparts of (32) and (33),
are unlikely.
i) A veces Ø hablan a ustedes de las variaciones como la de Brahms y cosas por el estilo.
„Sometimes [they] talk to you about variations like Brahms‟s and the like‟.
ii) Yo me he instalado en mi casa y Ø mandé un oficio al director diciéndole que… lo que necesito.
„I have settled home and I sent the director a letter telling him what I need‟.

79

(35) Efectivamente, como lo ha propuesto la… nuestra congresista
i
de Amazonas, que su
i

primera acción ha sido presentar un proyecto de ley para que se elimine los escasos
incentivos tributarios que goza una de las regiones más pobres del país, como es
Amazonas. (C)
„In fact, just like it has been proposed by our Amazonas congresswoman, that her first
action has been to present a law project in order to eliminate the scarce tax stimuli
that one of the poorest regions in the country, like Amazonas is, has‟.

However, there were sentences like (36) in which the possessive element was not
present, but rather a definite article. This poses a problem for coding: on the one hand, it
is not a strict gap-containing genitive RC, like (37) would be; on the other hand, it does
not contain an RP. For these reasons, genitive RCs were excluded from the quantitative
analysis.

(36) Estoy preparando este libro, que estoy en la corrección final. (I)
„I am preparing this book, that I am in the final correction‟.
(37) Estoy preparando este libro, en cuya corrección final estoy.
„I am preparing this book, in whose in final correction I am‟.

3.2.5 Attributes
Copulative verbs generally have a complement that refers to the subject, an
attribute. However, there are not that many instances in which an attribute is relativized.
In the data for this study, only a handful of attribute RCs were found, like (38) and (39).
80

(38) Por más Comisión de Justicia que seamos, somos legisladores. (C)
„It does not matter what a Justice Committee we are, we are legislators.‟

In (38), the attribute RC is gap-containing. However, in (39) we see an attribute
RC containing the invariable attribute pronoun lo
18
. In this example, it is repeating the
coordinated attribute una isla de paz, una isla de tranquilidad y una isla de respeto a los
Derechos Humanos, probably because its processing within the RC is hampered by the
presence of material intervening between que and the verb era (see 3.3.8).

(39) Se tenía la sensación de que en América Latina era una isla de paz, una isla de
tranquilidad y una isla de respeto a los Derechos Humanos, ¿no?, que efectivamente
no lo era, ¿no? (I)
„One had the sensation that Latin America was a peaceful island, a quiet island, and an
island that respected human rights, right?, which in fact it was not it.‟

Given that due to the scant occurrence of attribute RCs it would have been impossible
to arrive at statistically significant results, these few tokens were removed from the
quantification.

18
In Spanish, attributes can only be replaced by the pronoun lo, irrespective of the gender and number of
the NP functioning as an attribute.
i) Pedro es alto. Pedro lo es.
„Peter is tall. Peter is [it]‟.
ii) María es alta. María lo es.
„Mary is tall. Mary is [it]‟.
iii) Pedro y María son altos. Pedro y María lo son.
Peter and Mary are tall. Peter and Mary are [it]‟.

81

3.2.6 The Oblique Antecedent vez
This antecedent, meaning time, is very frequent in Spanish oblique RCs.
However, an RC version with this antecedent, including the PP (vez en que, „time in
which‟), is very infrequent, as is an RP-containing version. The variant that is almost
categorical is the one with PP-chopping, and it is now considered the standard version.
In (40) we find the only example containing the PP-inclusive version, although it
is a truncated RC, and the speaker immediately produces a new and complete RC with
the pluralized veces antecedent and its almost mandatory PP-chopping. It seems that vez
is being re-interpreted as an adverbial and that its nominal functions are inexistent for the
speakers. Likewise, example (41) contains another instance of a PP-chopping RC with
vez preceded by the article.

(40) Esto me hace recordar la… la… vez en que he dictado… las veces Øque he dictado
clases con micro, con micro en la universidad. (I)
„This reminds me the… the… time at which I have lectured… the times I have lectured
using a microphone in college‟.
(41) La vez Øque estuve más tiempo fue unos… cuatro meses. (I)
„The time that I stayed longer was around fourth months.‟

In (42) and (43), the antecedent at issue is preceded by the ordinal adjectives
primera and última, which add an extra semantic meaning to the noun, but the effect is
the same: it seems to be interpreted as an adverbial.
82

(42) La primera vez Øque vine, a mí este país no me gustó. (I)
„The first time that I came, I did not like this country‟.
(43) La última vez Øque he salido ha sido con motivo de la preparación del libro
Ayacucho, la libertad de América. (I)
„The last time that I have gone out has been when I had to prepare the book Ayacucho,
America‟s Liberty‟.

Finally, in (44) the noun is preceded by the distributive adjective cada, and the
result is exactly the same: the PP is chopped. This almost categorical lack of variation
was the cause for the exclusion of oblique RCs anteceded by vez from the analysis.
(44) Cada vez Øque yo vuelvo a un tipo de lectura especial, termino en… soy un poco
monotemático‟. (I)
„Every time that I return to a special type of reading, I end up in… I am very
monothematic‟.

3.2.7 Passive Reflexive Voice
The Spanish passive reflexive voice, which is constructed with the pronoun se, is
a structure that is active in form, but passive in meaning. In 97.7% of RCs with a passive
reflexive voice, neither RPs nor PP-chopping cases showed up. Gap-containing RCs like
(45) and (46) had an almost categorical occurrence.
83

(45) los programas que se van a hacer el próximo año por persona, por profesión, por tipo
de cargo. (C)
„the programs that will be done next year by person, by profession, by type of
position‟.
(46) Y a eso iba, tengo un trabajo que se va a publicar. (I)
„And I was getting there, I have an article that will be published‟.

There was a single case of a subject contrastive RP shown in (47), and a single
case of PP-chopping shown in (48) in the entire set of passive reflexive RCs. Therefore,
this type of clauses was excluded from the quantitative analysis.

(47) Todo costó $1500, más sus $150 de exámenes
i
, que eso
i
sí se pagó en partes. (J)
„Everything cost $1500, plus some $150 in exams, which that was paid in parts‟.
(48) Veamos esto con toda la transparencia y la legalidad Øque se deben ver estas cosas
de trabajadores que tiene la patria. (C)
„Let‟s approach this with the transparency and legality that these matters of workers
that our nation has must be seen [with].‟

3.2.8 Resumptive Noun Phrases
Along with RPs, there are resumptive noun phrases (RNPs) that appear in Spanish
RCs. However, they do not even represent 1 per cent of the data, so their inclusion in this
study would not give any statistically significant result, despite that, prima facie, one can
notice that all of them are non-restrictive RCs. For illustrative purposes, I present all of
84

them in the following lines. (49) – (53) are examples of subject RCs, and (54) – (56) are
DO RCs.

(49) Él es el hermano menor
i
, que el hermano menor
i
ni estaría en esa época. (J)
„He is the youngest brother, which the youngest brother was probably not even there
at that time‟.
(50) Había hablado con la otra chica
i
, que la otra chica
i
es doble cara. (J)
„I had talked to the other girl, which the other girl [gym trainer] is a hypocrit‟.
(51) Todos los grupos le tienen envidia a Mortem, excepto Disinter
i
, que esos huevones
i

sí nos tienen respeto. (J)
„All the [metal] bands envy Mortem, except for Disinter, which those dudes do in fact
respect us‟.
(52) Me escribo con K
i
, que el hombre
i
no se puede ver con M. (J)
„I am in touch with K, which the fellow does not get along with M at all‟.
(53) Entonces apareció una chica
i
, que esa chica
i
la
i
conozco del gimnasio. (J)
„Then a girl appeared, which that girl I know her from the gym‟.
(54) Yo he llevado más bien, am- a-allí a un grupito, ¿no? que es a, bueno, a Mary Ann
i
,
que sí la
i
conoces a Mary Ann Dixon
i
, ¿no? (I)
„I have taken there frien- a little group, right? Which is, well, Mary Ann, that you do
know [her] Mary Ann Dixon, right?
(55) “¡Quédate con tu plata!”, su plata
i
maravillosa, que sí tiene plata
i
, ¿ya? (I)
„“Keep your Money!”, his marvelous money, which he does have money, o.k.?‟
85

As can be seen in (49), (50) and (55), there are cases in which the RNP does not
only refer to the relativized element, but it rather repeats it. On the other hand, in (53) and
(54) these RNPs show up together with a DO RP.
Subject RNPs (49), (50), and (51) can be explained by performing a contrastive
function of singling out the antecedent from a set of competing candidates (siblings, gym
trainer girls, metal bands, respectively), a hypothesis that I will develop in more detail in
4.5. The subject RNP in (52) can be explained by the presence of material between que
and the verb (see 3.3.8).
DO RNPs in (53) and (54) could also be explained in terms of contrast: the RNP
singles out the antecedent from a set of candidates, and the extra RP occurs due to the
loose connection between the antecedent and its non-restrictive RC (see 3.3.3). The DO
RNP in (55) is not accompanied by an extra RP, but the fact that it is co-occurring with
the emphatic adverb sí (see 4.5) could suggest that it is contrastive as well.
19

3.3 Hypotheses Common to Subject, Direct Object, and Oblique Relative Clauses
There are a number of hypotheses to consider as possible explanations for the
presence of RPs; they are presented here as conditioning factors for the occurrence of
RPs. Most of these factors have already been shown to be quantitatively relevant. We will
test Hawkins‟ RP processing hypothesis for Spanish, and we will also test Prince‟s file-
card hypothesis. We will also present new hypotheses such as the internal RC weight in
3.3.8, and coordination and juxtaposition of RCs in 3.3.9 and 3.3.10, respectively.

19
Since the context this utterance (55) is the advantages and defects of her ex-boyfriend, it could be that the
RNP is used in a contrastive way to state that money, opposed to faithfulness and love, is an advantage that
her former boyfriend had.

86

The hypotheses presented here are common for RPs in all the three syntactic
positions studied, subject, DO, and obliques. In 3.4, I will present additional hypotheses
that pertain exclusively to the oblique position, based mainly in Tarallo (1986), and what
the predictions are for PP-chopping occurrence.

3.3.1 Syntactic Category of the Antecedent
Herrera Santana (1990) had already mentioned that sentential antecedents would
trigger neuter RPs like eso or esto. This hypothesis states that sentential antecedents like
(56) will trigger more RPs than nominal antecedents like (57), because it is harder to keep
track of a complex antecedent such as a sentence, and speakers use the neuter RP as a
means of reminding that the antecedent is not one of the nouns preceding the RC, but
rather an entire sentence.

(56) Por lo menos estoy tranquilo
i
, que eso
i
es lo que importa. (J)
„At least I am calm, which that is what matters.‟
(57) ¿Cómo voy a pagar $1200
i
, que esa
i
sería mi contribución? (J)
„How am I going to pay $1200, which that would be my contribution?‟

3.3.2 Restrictiveness of the Relative Clause
Based on a traditional analysis of syntactic “independence”, RCs have been
classified into restrictive and non-restrictive. The former, exemplified by (58), specify or
limit the meaning of their antecedents to the characteristics contained in the RC, whereas
87

the latter (59) bear characteristics or properties that are independent from the
interpretation of their antecedent; they are an addition of information.
As we saw in chapter 2, Prince‟s (1990) study showed that non-restrictive clauses
were more likely to trigger RPs than restrictive ones. According to her analysis, non-
restrictives represent information that is activated separately from the NP files as opposed
to restrictives, which are attached to their antecedent. Prince found that non-restrictives
favor more the occurrence of RPs. Silva-Corvalán (1994) and Cerrón-Palomino (1996)
found the same tendency in Spanish RCs.

(58) Es un arma
i
que la
i
puedes utilizar muy bien para poder vender tu producto. (J)
„It is a weapon that you can use it in order to sell your product‟
(59) También tengo tus recibos
i
, que te los
i
llevo. (J)
„I also have your bills, that I will take them to you.‟

3.3.3 Definiteness of the Antecedent
As presented in chapter 1 and discussed in chapter 2, some antecedents are
indefinite i.e., not specified, and some antecedents are definite in the sense that they are
specified or individualized in the discourse.
Although both of the examples given have RPs, this hypothesis, also presented by
Prince (1990), states that indefinite antecedents as in (60) are a more favorable context
for the manifestation of RPs than definite antecedents, as in (61). The former belong to
different paradigms (in the Saussureian term) than the RCs that modify them, because
these RCs are not necessary to understand the meaning of the antecedent. Hence, they are
88

not retrieved simultaneously in language processing, but rather taken from different files.
The RP‟s function, then, is to bridge the gap between them and facilitate the processing
of the compound antecedent + RC.

(60) una construcción
i
a medias, que nosotros la
i
llamábamos “la construc”. (I)
„A half-way building, that we called it „the buildin‟.‟
(61) Mira, tengo estos dos nombres
i
que Peter los
i
puede conocer. (I)
„Look, I have these two names that Peter may know them.‟

3.3.4 Previous Mention of the Antecedent
This is the first time that this hypothesis has been proposed. The underlying
rationale is that antecedents that have been mentioned before, as in (62), are present in
the speakers‟ and hearer‟s mind, but the ones that are introduced for the first time, as in
(63) and (64), may or may not. In order to bring attention on the latter or in order to
reinforce their reference, the speaker may use an RP to signal the antecedent‟s function.
If this hypothesis is right, the prediction is that RCs with antecedents that do not have
prior mention will be considerably more frequent than those with previously mentioned
antecedents.

(62) Vamos mañana al restaurante que querías ir.
„Let‟s go tomorrow to the restaurant that you wanted to go‟.
89

(63) El texto
i
que sería propuesto, que no lo
i
conocemos, y segundo, el planteamiento a
definir la conveniencia o no del tema de un dictamen de la Comisión de Justicia. (C)
„The text that would be proposed, which we do not know it, and secondly, the
proposal to define the adequacy or inadequacy of a topic of a ruling of the Justice
Committee‟.
(64) Yo estoy de acuerdo con el cuarto intermedio, pero, Presidente, hay el dictamen
i
en
minoría, que lo
i
comparten muchos congresistas que están aquí presentes. (C)
„I agree with the fourth intermediate, but, President, there is the minority ruling, that
many congressmen present here share it‟.

3.3.5 Distance between the Antecedent and the Complementizer que
The antecedent can be immediately before que, in which case it is labeled
“adjacent”, as in (65), or it can have any intervening material in the middle and in that
case, it is labeled “non-adjacent”, as in (66). It is expected that the further the antecedent,
the more likely it will be to trigger RPs, because the cognitive task needed to link the
antecedent to its argument / adjunct place within the RC will be greater.

(65) Es un pisapapeles
i
que me lo
i
regaló mi tía Julita. (J)
„It is a paperweight that my aunt Julia gave it to me‟.
(66) La letra
i
la hizo Daniel, que no la
i
entiendo. (J)
„The lyrics Daniel wrote, which I don‟t understand them.
90

3.3.6 Animacy of the Antecedent
This variable was first taken into consideration by Silva-Corvalán (1996). It did
show significant results in her study, and the hypothesis is that RPs occur mostly when
the antecedent is animate, as in (67), as opposed to inanimate, as in (68).

(67) Tienes un grupo de gente
i
que la
i
conoces un huevo de años. (I)
„You have a group of people that you know them for a long time‟.
(68) Tengo que tomar esta pepa
i
que se supone que tenía que tomarla
i
. (I)
„I have to take this pill that I was supposed to take it‟.

The rationale behind this hypothesis is that animate antecedents tend to be more
cognitively salient than inanimate ones, so the speakers attempt to bring attention to the
former by repeating their syntactic function within the RC with a pronoun.

3.3.7 Size of the Relative Clause Domain
This hypothesis follows Hawkins 2004a, and takes into account the size or weight
of the FGD (in gap-containing RCs) and the HPD (in RP-containing RCs). The rationale
is that larger domains will favor the presence of RPs more than smaller ones, because the
latter do not require a significant processing effort to retrieve the function of the
relativized element. The domains were taken in a gradient scale according to the sum of
their LD and HPD/FGD, so we have domains of 5, 6, 7, up to 38. Examples (69) and (70)
show large HPDs with a count of 19 and 17, respectively, that contain an RP that would
91

make it easier to process the relative element‟s function within the RC. Then they were
labeled as „small‟ (5-12). „medium (12-24), and large (25 and up).

(69) Tengo que leer por lo menos unas dos horas, porque hay muchas cosas
i
que las
instructoras de todas maneras te lo
i
dicen, pero son difíciles de recordar. (I)
„I have to read at least for two hours, because there are many thing that the instructors
will definitely tell you them, but are hard to remember‟.
(70) Por lo pronto, me están dando vacaciones
i
adelantadas que no saben si me las
i
van a
poder cobrar. (J)
„For the time being, they are giving me a vacation in advance that they don‟t know if
they will be able to charge me them‟.

On the contrary, in (71) the FGD is 6, and in (72) it is 8, very small domains that
require little effort in processing the antecedent‟s syntactic function within the RC, so no
RP shows up.

(71) Peleándome con la gente allí en al aeropuerto, para que me embarquen en el- en el
primer vuelo que salga. (I)
„Arguing with people there in the airport, so that they would let me board in the first
flight that takes off‟.
(72) Tatiana era la única que se levantaba y no saludaba a mi mamá. (I)
„Tatiana was the only one that used to get up and didn‟t say hello to my mother‟.
92

3.3.8 Weight of the Information Borne by the Relative Clause
The first to propose „weight‟ as a conditioning factor for RPs in Spanish RCs was
Silva-Corvalán (1996: 393), who defined it as “the weight of the information conveyed
by the antecedent and by the RC, and … the weight of the cognitive task required to
„create or activate files‟ for this information”. Here I operationalize the term „weight‟ as
follows: “light” RCs contain only “bare” simple and composed verbs and an RP when it
occurs, as in (73) and (74); “material on the left” RCs include a verb plus material other
than RPs (adverbs, NPs, other pronouns, etc.) on the left of the verb, as in (75) and (76);
“material on the right” includes material on the right of the verb, as in (77); and “heavy”
RCs are those that include material on both sides of the verb, as in (78).
The hypothesis is that the lighter the RC, the least it will favor the presence of
RPs, whereas heavier RCs will favor them, because RPs will make the processing of the
RC easier.

(73) ¿Sabes por qué nadie compró el otro
i
que lo
i
vendían? (J)
„Do you know why nobody bought the other one that they were selling it?‟
(74) Dos chistes
i
de mi papá, que los
i
ha leído. (I)
„Two jokes of my dad, that he has read them‟.
(75) También tengo tus recibos
i
, que te los
i
llevo. (J)
„I also have your bills, that I will take them to you‟.
(76) Esa cámara
i
, que yo ni siquiera la
i
he pedido. (J)
„That camera, that I have not even asked for it‟.
93

(77) Encontré una chiquita
i
linda que la
i
devoraba con los ojos. (J)
„I found a cute girl that I devoured her with my eyes‟.
(78) Es un pisapapeles
i
que me lo
i
regaló mi tía Julita. (J)
„It is a paperweight that my aunt Julia gave it to me‟.

3.3.9 Coordinated Relative Clauses
This hypothesis states that in a sequence of coordinated RCs, the second (or third
one) should pose more processing difficulties than the first one. The speaker‟s strategy
could be to use an RP to facilitate the identification of the antecedent‟s syntactic function
within the RC, as (79) may suggest. The prediction is that, when there are RCs standing
in this coordination relationship, the first RC is unlikely to have an RP as opposed to the
following RCs, which are harder to process. Another prediction is that strategies like the
one in (79) will be more frequent than gap-containing choices like the one in (80).

(79) un curso
i
, digamos, pequeño sobre técnicas pedagógicas modernas, pero que…
digamos, me… me agradó bastante, y que eso
i
sí me ha servido para… la vida
práctica, ¿no? (I)
„a –say– short course on modern pedagogical techniques, but that –say– I… I liked a
lot, and which that has indeed been useful for everyday life, right?‟
(80) un buen porcentaje que no se interesa y que pasa por la universidad y… y no deja
huella… ni… ni la universidad deja huella en él. (I)
„a significant percentage that are not interested and that they go trough college and…
they don‟t leave their mark nor… nor does college leave its mark on them.
94

3.3.10 Juxtaposed Relative Clauses
This hypothesis is similar to the previous one in two respects: it is the first time
that it is studied in a quantitative way, and it also refers to a sequence of RCs, only that in
this case there is no overt coordinating conjunction.
The predictions are the same as the ones expected for 3.3.9: in a sequence of
juxtaposed RCs, like the ones in (81), the second (or third) RC will be more likely to
include an RP than the first RC, and tokens like (82) should be more frequent than tokens
like (84).

(81) O sea, hacer cambios
i
sociales, que ha habido en el país, que hay que reconocerlos
i
,
importantes. (I)
„I mean, to make social changes, that there have been on the country, that we have to
recognize them‟.
(82) un trabajo que se va a publicar, que es un poco la reivindicación del… de la acepción
de selva que mi montaña tiene. (I)
„a study that will be published, which is somewhat the recognition of the acceptance
of the concept of jungle that my mountains have‟.

3.4 Hypotheses Exclusive to Oblique Relative Clauses
In the following, we will present the hypotheses that pertain only to PP-chopping
and to oblique RPs. It is worth mentioning that all of the previous hypothesis that predict
the occurrence of RPs are opposite to the ones that predict PP-chopping. If Tarallo (1986)
is correct, then PP-chopping and oblique RPs stand in complementary distribution, and
95

the contexts that they should appear should be reversed. Thus, restrictive RCs, nominal,
definite and inanimate antecedents, etc., should cause PP-chopping to occur.

3.4.1 Type of Preposition
Alcina and Blecua (1991) classify Spanish prepositions according to their lexical
meaning. „Full‟ prepositions are the ones that have a prominent lexical meaning, such as
ante (before), bajo (under), contra (against), desde (from), entre (between), hacia
(toward), hasta (until), para (for, to), por (for, to), según (according to), sin (without),
sobre (over), and tras (after). In contrast, „empty‟ prepositions are the ones whose lexical
meaning is not clear or blurry, like a (to, for), con (with), de (of), and en (in, on, at).
The hypothesis that I propose is that full prepositions will trigger more RPs, as in
(83) and (84), whereas empty prepositions will be more prone to be chopped, as in (85)
and (86).

(83) Los noisers, pe, los noisers
i
que para ellos
i
no hay diferencia. (J)
„The noisers, the noisers that for them there is no difference.‟
(84) Hay muchos problemas
i
que nadie se preocupa por eso
i
. (J)
„There are many problems that nobody cares about it.‟
(85) Hay una revista a la que yo me afilié, que se llama American Baby. (I)
„There is a magazine that I subscribed to, which is called American Baby.‟
(86) Yo lo pongo en colegios en que hay pocos niños. (I)
„I put him in schools in which there are few children.‟
96

3.4.2 Category Selecting the Prepositional Phrase
My hypothesis is that RCs selected by a preposition rather than by a verb are a
more favorable context for RPs to occur, due to their relative independence with regard to
the verb, as in (87). On the other hand, PP-chopping should occur more when the
preposition deleted introduces a verb complement rather than an adjunct, as exemplified
in (88).

(87) Es el tipo de chica
i
que tú quieres hacer algo con ella
i
; y ella, no. (J)
„She is the type of girl that you want to do something with with her, and she does
not.‟
(88) Está viviendo con el chico con el que se casó. (J)
She is living with the guy that she married with
„She is living with the guy that she married.‟

3.5 Coding of the Data
Two statistical programs were used to examine the behavior of the dependent
variable and the conditioning factors described above: SPSS 11.0 and Goldvarb 2001.
The former was used to cross-tabulate the dependent variable by the variables in each
factor group and to calculate chi-square. Crosstabs showed which variants of the
independent variable favor or disfavor the presence of RPs in the data.
Goldvarb was used to perform a multivariate analysis that handles all independent
variables at the same time. This program allows two different functions that SPSS does
not offer: first, it yields a constraint ranking of factors within a factor group, and
97

secondly, it gives an interfactorial relative strength, obtained by subtracting the lowest
factor weight from the highest factor weight in each factor group. (Tagliamonte, 2006:
239, 242)
In order to fulfill the programs‟ requirements, the hypotheses were coded as
variables (factor groups) with different values as follows:

Subject and Direct Object RCs
I. Dependent variable
1. RC strategy
RP p
Gap a
II. Independent Variables
Common to the three syntactic functions
2. Antecedent
Sentential s
Nominal n
3. Restrictiveness
Restrictive r
Non-restrictive n
4. Definiteness
Definite Antecedent d
Indefinite Antecedent i
98

5. Previous mention
Previously mentioned m
No prior mention n
6. Distance
Adjacent a
Non-adjacent n
7. Animacy
Animate a
Inanimate i
8. Domain
Small s
Medium m
Large l
9. Weight
Light l
Material to the left i
Material to the right d
Heavy b
10. Coordination
Coordinated RC y
Non-coordinated RC n
99

11. Juxtaposition
Juxtaposed RC j
Non-juxtaposed RC n
Oblique RCs
I. Dependent variable
1. RC strategy (RP variation)
RP p
Gap + PP-chopping a

2. RC strategy (PP-chopping variation)
PP-chopping c
Gap + RP a
II. Independent variables
3. Restrictiveness
Restrictive r
Non-restrictive n
4. Definiteness
Definite Antecedent d
Indefinite Antecedent i
5. Previous mention
Previously mentioned m
No prior mention n
100

6. Distance
Adjacent a
Non-adjacent n
7. Animacy
Animate a
Inanimate i
8. Domain
Small s
Medium m
Large l
9. Weight
Light l
Material on the left i
Material on the right d
Heavy b
10. Coordination
Coordinated RC y
Non-coordinated RC n
11. Juxtaposition
Juxtaposed RC j
Non-juxtaposed RC n
101

12. Preposition type
Full f
Empty e
13. Category selecting the RC
Verb c
Preposition m

102

Chapter 4: Results and Discussion

4.1 Resumptive Pronoun Presence/Absence Alternation
In this section, I will present the results for the RP vs. gap alternation in subject
and DO RCs, and the alternation RP vs. gap + PP-chopping in oblique RCs.

4.1.1 Results for Subject
The results of the hypotheses mentioned in 3.3 are presented in this section, in
particular, for RCs that relativize a subject.
Table 6. Subject – Restrictiveness

non-
restrictive restrictive total
Gap

Count 474 691 1165
% 92.6% 99.0% 96.3%
RP

Count 38 7 45
% 7.4% 1.0% 3.7%
total Count 512 698 1210
% 100.0% 100.0% 100.0%

p = .000
As Table 6 shows, the hypothesis that non-restrictive RCs favor the occurrence of
RPs is supported by the data. In fact, non-restrictive RCs (7.4%) favor the occurrence of
subject RPs over restrictive RCs (1%), with a statistically significant difference (p =
.000). These results are compatible with all the previous studies reviewed in chapter 2,
103

which suggest that the „loose link‟ between a non-restrictive RC and its antecedent favors
the presence of RPs. In other words, the adjunct status of the non-restrictive RC, as
opposed to the complement status of the restrictive RC, favors the occurrence of an RP to
make the relation between the two tighter.
Table 7. Subject – Definiteness

definite indefinite total
Gap Count 551 604 1155
% 97.2% 97.4% 97.3%
RP Count 16 16 32
% 2.8% 2.6% 2.7%
Total Count 567 620 1187
% 100.0% 100.0% 100.0%

p = .798
In order to determine the relation between RP occurrence and definiteness, 23
tokens that could not be classified as either definite or indefinite were removed. These
tokens are the ones that have a sentential rather than a nominal antecedent, and they will
be discussed in Table 3.
The results Table 7 shows imply that the hypothesis of indefinite antecedents
favoring RP use is not supported by the data. The definite nature of nominal antecedents
plays no apparent role in triggering subject RPs, as it does not make almost any
difference whether the antecedent is indefinite as in (1)-(2), or definite, as in (3)-(5) (the
104

percentages are 2.6% vs. 2.8%, respectively). The results are not statistically significant
(p = .798).

(1) Tengo un amigo
i
que él
i
es un arquitecto. (J)
„I have a friend that he is an architect‟.
(2) Él se había comprado una carabina
i
de aire, que le gustaba mucho eso
i
. (I)
„He had bought an air carbine, that he liked that a lot‟.
(3) A la otra chica le paga $20 por 25 minutos, y a mí
i
, que yo
i
sí estuve entrenando –
hasta tres grupos- no me dio nada. (J)
„She pays the other girl $20 for 25 minutes, and she did not give me, that I was indeed
training up to 3 groups, anything‟.
(4) Ahora, pero tú ya has visto cómo soy yo
i
también, que yo
i
sí...no puedo estar quieta.
(I)
„Now, you have already seen how I am, that I definitely cannot be still‟.
(5) La vi con Frank
i
, que él
i
estaba que se moría de miedo. (J)
„I saw her with Frank, that he was really scared‟.

These results are not on par with the findings in previous studies, which suggest
that the indeterminate nature of indefinite antecedents is „compensated‟ for by using an
RP. However, such studies included RPs as a whole, including DO RPs, a syntactic
position that includes the majority of RPs. As we will see later in this chapter, DO RPs
seem to be favored by indefinite antecedents.
105

Table 8. Subject – Antecedent Syntactic Category

sentential
antecedent
nominal
antecedent total
Gap Count 10 1155 1165
% 43.5% 97.3% 96.3%
RP Count 13 32 45
% 56.5% 2.7% 3.7%
total Count 23 1187 1210
% 100.0% 100.0% 100.0%

p = .000
The next factor to be analyzed is the nature of the antecedent, i.e. sentential
antecedents opposed to nominal antecedents. The results are presented in Table 8. There
is a very strong tendency of speakers to produce RPs when referring to an antecedent that
is a type of IP (56.5%): gerunds as in (6), main clauses as in (7), subordinate clauses as in
(8), and infinitives, as in (9).
By contrast, subject RPs show up only 2.7% of the total of occurrences when they
have nominal antecedents.

(6) Gastando en comida
i
, que eso
i
fue lo que te comenté la vez pasada. (J)
„Spending in food
i
, which that
i
is what I told you the other day‟.
(7) Llego el sábado en la mañana
i
, que eso
i
sería lo ideal, ¿no? (J)
„I arrive Saturday morning
i
, which that
i
would be ideal, right?‟
106

(8) otras organizaciones que están procurando su desafiliación
i
, que esto
i
es una vía
crucis. (C)
„other organizations that are working on theur unaffiliation, which that is a via crucis‟.
(9) Voy a estar hasta armar el nuevo semestre
i
, que eso
i
es en junio, así. (J)
„I will be there until organizing the new semester, which that is in June‟.

Table 9 shows the results for the weight factor, i.e. whether or not there is
existence of material inside the RC, and when there is, where it is distributed. Contrary to
the findings in Cerrón-Palomino (2006), the hypothesis according to which the presence
of material to the left of the verb triggers RPs is not confirmed. There is no significant
difference between the position of the material, be it at both sides of the verb (5.2%), at
the left (3.8%) or at the right (3.4%).
It seems clear, though, that light RCs, the ones containing no material within the
RC besides the verb, do not favor at all (0%) the occurrence of RPs in subject RCs. (10)
and (11) show examples of this type of RCs.

(10) O sea, el único del problema que queda es un hermano, pues.
„I mean, the only problem that remains is a sibling‟.
(11) En esa época eran poquísimas las mujeres que estudiaron, ¿no?
„In those days there were very few women who studied, right?‟

107

Table 9. Subject – Weight

p = .322
In order to test the hypothesis according to which non-adjacent antecedent NP
heads favor RP occurrence more than adjacent NP heads, the 23 tokens with a sentential
antecedent were excluded. This was done because the verbal element would have had to
be coded as the head of the antecedent, which is arguable.
The results of the data free of sentential antecedents did not support the
aforementioned hypothesis. Table 10 shows that there is almost no difference of RP
occurrence between RCs with a non-adjacent antecedent (2.8%) and an adjacent
antecedent (2.6%), with no statistical significance (p = .841).
both
sides
right
side left side light total
Gap

Count 255 792 76 42 1165
% 94.8% 96.6% 96.2% 100.0% 96.3%
RP

Count 14 28 3 0 45
% 5.2% 3.4% 3.8% .0% 3.7%
Total Count 269 820 79 42 1210
% 100.0% 100.0% 100.0% 100.0% 100.0%
108

Table 10. Subject – Distance

adjacent
non-
adjacent total
Gap

Count
706 449 1155
% 97.4% 97.2% 97.3%
RP

Count 19 13 32
% 2.6% 2.8% 2.7%
Total Count 725 462 1187
% 100.0% 100.0% 100.0%

p = .841

The hypothesis stating that RCs with medium and large domains produce more
RPs than RCs with small domains demanded the removal of the sentential antecedents,
because no head could be determined from a sentence-like antecedent. This hypothesis
was not supported by the data. Table 11 shows that if anything, the pattern was reversed:
3.2% of RCs with small domains trigger RPs, followed by 1.7% of RCs with medium
domains, and 0% of RCs with large domains, but the results are not statistically
significant.
The domains results for subject are not that unexpected, since Hawkins‟s theory
also states that subject RCs are, in general, the easiest to process. Therefore, the greater
percentage of RPs in small domain RCs is likely to be due to different factors, rather than
to the difficulty in processing the function of the antecedent within the subject RC.
109

Table 11. Subject – Domain Size

large medium small total
Gap Count 64 233 859 1156
% 100.0% 98.3% 96.8% 97.3%
RP Count 0 4 28 32
% .0% 1.7% 3.2% 2.7%
total Count 64 237 887 1188
% 100.0% 100.0% 100.0% 100.0%

p = .182

In order to test the hypothesis that animate antecedents favor the occurrence of
RPs more than inanimate antecedents, the sentential antecedents were removed, because
due to their nature, they could not be assigned either an animate or an inanimate value,
which are properties of NPs. Table 12 shows that this hypothesis was confirmed by the
data. In fact, 4.4 per cent of animate antecedents trigger RPs, and only 1.5 per cent of
inanimate antecedents favor RP occurrence within their corresponding RCs.
110

Table 12. Subject – Animacy

animate inanimate total
Gap Count 483 672 1155
% 95.6% 98.5% 97.3%
RP Count 22 10 32
% 4.4% 1.5% 2.7%
total Count 505 682 1187
% 100.0% 100.0% 100.0%

p = .002
In order to test the hypothesis according to which RPs show up more frequently
when they refer to a not previously mentioned antecedent, 18 tokens of data set 2 were
excluded. This data set differs from the rest in that the context in which these tokens were
jotted down is in most cases insufficient to determine whether there was a prior mention
of the antecedent or not. Because there were no means of determining whether the
antecedents had been mentioned before or not, I excluded them from the cross-
tabulations.
Table 13 shows that the results do not confirm this hypothesis. The frequency of
RPs whose antecedents have a prior mention (2%) is slightly smaller than the one of RPs
whose antecedents have been mentioned for the first time (2.7%), but with no statistical
significance (p = .515).

111

Table 13. Subject – Previous Mention

mentioned
not
mentioned total
Gap

Count 293 869 1162
% 98.0% 97.3% 97.5%
RP

Count 6 24 30
% 2.0% 2.7% 2.5%
total Count 299 893 1192
% 100.0% 100.0% 100.0%

p = .515
Table 14 shows the results for the hypothesis that states that RPs have a greater
tendency to occur in the second (or third) RC of a coordinated group of RCs. This
hypothesis was not supported by the data. The slight advantage shown for RPs in
coordinated RCs (4.1%), when compared to the frequency of RPs occurring in non-
coordinated RCs (3.7%), is not statistically significant (p = .891).
112

Table 14. Subject – Coordination

p = .891

The hypothesis according to which the second (or third) RC in a succession of
juxtaposed RCs favors RP presence was not confirmed by the data either. Table 15 shows
that rather the tendency is reverted, i.e. there is a greater –though slight– frequency of
RPs in RCs that are not juxtaposed (3.8%) than there is in RCs that are juxtaposed
(2.3%), with no statistical significance (p = .479)

not
coordinated coordinated total
Gap

Count 1118 47 1165
% 96.3% 95.9% 96.3%
RP

Count 43 2 45
% 3.7% 4.1% 3.7%
total Count 1161 49 1210
% 100.0% 100.0% 100.0%
113

Table 15. Subject – Juxtaposition

p = .479

The cross-tabulations of the proposed conditioning hypotheses give us the
following factors as statistically significant for the occurrence of RPs within subject RCs:
non-restrictive RCs, sentential antecedents, and animate antecedents.
In order to determine the strength of each factor, and establish their predictive
values, I used the multivariate program Goldvarb 2001. However, more than a few
problems arose in doing so, due to the heterogeneous nature of the different data sets
compiled. Some of the difficulties were due to the nature of the data, and some were
common problems in multivariate analyses.
First, the uneven nature of the data made the comparison of factors difficult. RCs
including sentential antecedents could not be compared to RCs including nominal
antecedents, because the former are not accountable for by factor groups such as
definiteness, animacy of the antecedent, or distance. Therefore, the 23 tokens with
sentential antecedents were not included in the multivariate analysis, despite the fact that
precisely a sentential antecedent turned out to have a strong statistical significance in the

juxtaposed
not
juxtaposed total
Gap

Count 84 1081 1165
% 97.7% 96.2% 96.3%
RP

Count 2 43 45
% 2.3% 3.8% 3.7%
total Count 86 1124 1210
% 100.0% 100.0% 100.0%
114

cross-tabulations. The exclusion of the aforementioned tokens also entailed the exclusion
of the factor group „syntactic category of the antecedent‟.
A second difficulty related to the particular data of this study was that the factor
regarding prior mention of the antecedent entailed the exclusion of 18 jotted-down (not
acoustically recorded) tokens whose mentioning status was impossible to determine. But
given the fact that reducing the number of tokens rendered having very few tokens in
some of the factor cells, I decided to exclude the factor group „previous mention of the
antecedent‟ from the multivariate analysis.
There is a very common difficulty in multivariate analysis that was not absent in
my study: the presence of knockouts. A knockout is either a 100 per cent occurrence or a
0 per cent occurrence of one factor within a factor group. Once a knockout factor occurs,
the researcher has two possibilities: to either recode and collapse two categories (when
there is a valid linguistic reason to do so), or to exclude that particular factor from the
multivariate analysis.
I had two knockout factors in my study. The first one was due to the absence of
RPs in RCs with no material inside the RC besides the verb, and it had to be excluded or,
more precisely, „blocked‟ by the statistical program in order for the analysis to run its
course. The second knockout was the factor „large domain‟ in the domain size group
factor, with its subsequent „blocking‟ by the program.
In sum, the factor groups entered in the Goldvarb 2001 program were
restrictiveness of the RC, definiteness of the antecedent, weight of the RC, distance from
the antecedent head, domain size, animacy of the antecedent, coordinated RC, and
juxtaposed RC.
115

In order to calculate the strength of the factors that are significant in the
multivariate analysis results, the lowest value of a factor in the factor group has to be
subtracted from the highest value of a factor in the same group. The values obtained per
factor group are compared then, and the higher the number obtained for the factor group,
the stronger the factor is.
The results are presented in Table 16. They show that a non-restrictive RC is the
factor that most strongly favors the occurrence of a subject RC, followed by the existence
of material at both sides of the verb within the RC. The following factor is small domain,
which is completely contrary to the original hypothesis. Finally, the last conditioning
factor in the ranking is an animate antecedent. The group factors „definiteness of the
antecedent NP‟, „distance between the antecedent and the complementizer que‟,
„coordination of RCs‟, and „juxtaposition of RCs‟ were disregarded by the program, and
they had not been statistically significant in the cross-tabulations results either.

116

Table 16. Subject RPs Goldvarb
Input 0.013

Restrictiveness of the RC
non-restrictive RC 0.781
restrictive RC 0.289
relative strength 0.492

Weight borne by the RC
material at both sides 0.736
material to the left 0.634
material to the right 0.403
relative strength 0.360

Domain size
small domain 0.566
medium domain 0.271
relative strength 0.295

Animacy of the antecedent
animate 0.667
inanimate 0.374
relative strength 0.293

Log likelihood = -121.925 Significance = 0.008

117

4.1.2 Results for Direct Object
The results for the hypotheses for DO RP occurrence presented in 3.3 are shown
in the following lines.
Table 17. Object – Restrictiveness

non-
restrictive restrictive total
Gap Count 44 424 468
% 28.4% 77.1% 66.4%
RP

Count 111 126 237
% 71.6% 22.9% 33.6%
total Count 155 550 705
% 100.0% 100.0% 100.0%

p = .000
The hypothesis according to which non-restrictive RCs favor the occurrence of
RPs was confirmed for the DO position. As can be seen in Table 17, non-restrictives
show a greater frequency of RP occurrence (71.6%) when compared to restrictives
(22.9%), with a statistically significant p value (.000).
Likewise, the hypothesis that predicted RPs to be favored by indefinite NP
antecedents was confirmed. However, in order to include only NP antecedents, the 5
instances of sentential antecedents were removed from this analysis.
The results are presented in Table 18, and RPs show up 48.9 per cent of the time
in RCs with indefinite antecedents, while they show up only 22.6 per cent of the time in
RCs with definite antecedents, with a statistically significant difference (p = .000).
118

Table 18. Object – Definiteness

definite indefinite total
Gap

Count 325 143 468
% 77.4% 51.1% 66.9%
RP Count 95 137 232
% 22.6% 48.9% 33.1%
total Count 420 280 700
% 100.0% 100.0% 100.0%

p = .000

Table 19 shows the results corresponding to the hypothesis that sentential
antecedents like (12) and (13) favor the occurrence of RPs. This hypothesis was
supported by the data, given that DO RPs do appear 100 per cent of the time in RCs
whose antecedent is of a sentential nature, while they occur only 33.1 per cent of the time
in RCs whose antecedent is of a nominal nature. The results convey a statistical
significance (p = .002).

(12) Ya solamente para escribir
i
no más, que eso
i
ya lo
i
hace ya sola.
„Only for writing, which that (she) already does it by herself.
(13) Así mismo, se propone la Norma, Sr. Presidente, para llenar ciertos vacíos
i
, se señ…
y que no lo
i
hiciera la ley 27,803.
„Likewise, Mr. President, this regulation is proposed to fill some gaps, it was s… and
which the 27,803 Law did not do it‟.
119

Table 19. Object –Antecedent Syntactic Category

sentential
antecedent
nominal
antecedent total
Gap Count 0 468 468
% .0% 66.9% 66.4%
RP Count 5 232 237
% 100.0% 33.1% 33.6%
total Count 5 700 705
% 100.0% 100.0% 100.0%

p = .002
Table 20 shows the results for the weight factor. As the hypothesis predicted,
there is a greater frequency of RPs in RCs that have material to the left of the verb
(44.2%) and a smaller frequency in RCs that do not have any material inside the RC
except for the complementizer que and the verb (22%), a difference that is statistically
significant (p = .000). Material on both sides of the verb (40.3%) is very close to material
to the left (44.2%) as a favorable context for RPs, presumably because they also include
material to the left, whereas material to the right (28.6%) is closer to RCs with no
material inside.
120

Table 20. Object – Weight
both
sides
right
side left side light total
Gap Count 92 185 92 99 468
% 59.7% 71.4% 55.8% 78.0% 66.4%
RP

Count 62 74 73 28 237
% 40.3% 28.6% 44.2% 22.0% 33.6%
total Count 154 259 165 127 705
% 100.0% 100.0% 100.0% 100.0% 100.0%

p = .000

In order to test the distance hypothesis, the five sentential antecedents were
removed to avoid determining which was the head that the complementizer que would be
adjacent or not to. The statistically significant (p = .025) results are presented in Table 21,
with 39.3% of non-adjacent antecedents favoring RP presence versus a 30.6% of adjacent
antecedents showing RPs within their RCs.
121

Table 21. Object – Distance

adjacent
non
adjacent total
Gap Count 343 125 468
% 69.4% 60.7% 66.9%
RP Count 151 81 232
% 30.6% 39.3% 33.1%
total Count 494 206 700
% 100.0% 100.0% 100.0%

p = .025

In Table 22, we can see that the hypothesis that states that RCs with medium and
large domains would cause RPs to show up in the DO position was not supported by the
data. Small domains (32.9%) favor RPs more than large domains (22.2%), with a slight
more frequency of medium domain RCs triggering RPs (35.7%). However, the results are
not statistically significant (p = .672).

122

Table 22. Object – Domain Size

large medium small total
Gap

Count 7 63 398 468
% 77.8% 64.3% 67.1% 66.9%
RP Count 2 35 195 232
% 22.2% 35.7% 32.9% 33.1%
total Count 9 98 593 700
% 100.0% 100.0% 100.0% 100.0%

p = .672

The hypothesis that animate antecedents favor the occurrence of RPs over
inanimate antecedents was confirmed for the DO position. In order to do so, just like for
subject RCs, only NP antecedents were considered, and the 5 sentential antecedents were
removed from the analysis. Table 23 shows that 62.1% of RCs that correspond to animate
antecedents include an RP, whereas only 28.6% of RCs with inanimate antecedents favor
the occurrence of RPs, with a high statistical significance (p = .000).
123

Table 23. Object – Animacy

animate inanimate total
Gap Count 36 432 468
% 37.9% 71.4% 66.9%
RP Count 59 173 232
% 62.1% 28.6% 33.1%
total Count 95 605 700
% 100.0% 100.0% 100.0%

p = .000
In order to test the hypothesis according to which antecedents not mentioned
previously favor RPs over antecedents mentioned before, 76 tokens for which no context
was available to determine the aforementioned difference, were excluded. The hypothesis
was not confirmed with the data. As we can see in Table 24, the pattern was opposite to
the expected one: 38.7 per cent of the previously mentioned antecedents favor RPs,
whereas 20.8 per cent of first-mention antecedents favor RP occurrence.

124

Table 24. Object – Previous Mention

mentioned
not
mentioned total
Gap

Count 106 361 467
% 61.3% 79.2% 74.2%
RP

Count 67 95 162
% 38.7% 20.8% 25.8%
total Count 173 456 629
% 100.0% 100.0% 100.0%

p = .000

The hypothesis that the second or third RC in a set of coordinated RCs favor RPs
over RCs that are not coordinated was confirmed for the DO position by the data. The
results in Table 25 corroborate this hypothesis, as 61.5% of coordinated RCs do include
an RP, whereas only 33.1% of non-coordinated RCs do not exhibit an RC within them,
with a statistically significant value (p = .031).
125

Table 25. Object – Coordination

not
coordinated coordinated total
Gap

Count 463 5 468
% 66.9% 38.5% 66.4%
RP

Count 229 8 237
% 33.1% 61.5% 33.6%
total Count 692 13 705
% 100.0% 100.0% 100.0%

p = .031
In Table 26, we can see that the hypothesis according to which the second or third
RC in a set of juxtaposed RCs will favor the occurrence of an RP. In fact, 81.3% of
juxtaposed RCs present an RP, while only 32.5% of non-juxtaposed RCs favor RPs.
126

Table 26 – Object –Juxtaposition

p = .000
In order to rank the factors that were found statistically significant in the cross-
tabulation results, I used Goldvarb 2001. In the DO position, the five tokens with
sentential antecedents categorically include RPs, which gives a knockout factor that
prevents the program from executing the multivariate analysis. Therefore, just like in the
case of the subject Goldvarb analysis, I had to remove the RCs with sentential
antecedents before entering the data in the program.
Besides the aforementioned modification, and following the pattern used in the
case of the subject RP analysis, I did not include the factor regarding prior mention of the
antecedent. Just in the case of subject RPs, more than a few DO RP-containing tokens
corresponded to the data that I gathered outside the recorded interview‟s time span, a fact
that made it impossible to code them as either previously mentioned or not mentioned
previously.

juxtaposed
not
juxtaposed

total
Gap

Count 3 465 468
18.8% 67.5% 66.4%
RP

Count 13 224 237
% 81.3% 32.5% 33.6%
total Count 16 689 705
% 100.0% 100.0% 100.0%
127

Thus, the group factors included in the Goldvarb program were the same ones
used for the subject case: restrictiveness of the RC, definiteness, weight, distance,
animacy, domain size, coordination, and juxtaposition, and the results are in Table 27.
128

Table 27. Object RPs Goldvarb
Input 0.292

Restrictiveness of the RC
non-restrictive 0.882
restrictive 0.366
relative strength 0.516

Domain size
small domain 0.526
medium domain 0.409
large domain 0.060
relative strength 0.466

Juxtaposition of RCs
juxtaposed 0.896
not juxtaposed 0.487
relative strength 0.409

Coordination of RCs
coordinated 0.886
not coordinated 0.490
relative strength 0.396

Animacy of the antecedent
animate antecedent 0.743
inanimate antecedent 0.459
relative strength 0.284

Definiteness of the antecedent
indefinite 0.667
definite 0.386
relative strength 0.281

Weight borne by the RC
material to the left 0.617
material at both sides 0.529
no material 0.446
material to the right 0.436
relative strength 0.181

Distance between antecedent and complementizer que
adjacent 0.539
non-adjacent 0.407
relative strength 0.132

Log likelihood = -328.933 Significance: 0.036

129

No factor group was eliminated in the best stepping up and down. By subtracting
the highest value minus the lowest value of the factors in the factor group, we arrive at
the following ranking of the relative strength of the factors: non-restrictive RCs, small
domains, juxtaposed RCs, coordinated RCs, animate antecedents, indefinite antecedents,
material to the left of the verb, and finally, adjacent RCs, a result that not only contradicts
the hypothesis, but the results in the crosstabs.

4.1.3 Results for Obliques
In what follows, the results for the proposed factors for oblique RP occurrence,
advanced in 3.3, are presented. One must bear in mind that unlike the case of subject and
DO RCs, which included an opposition between RPs and gaps, in the oblique case the
opposition is between RPs and gaps + PP-chopping.
Unlike the case of subject and DO RCs, the occurrence of sentential antecedents in
oblique RCs was extremely reduced: it had only one token, presented in (14), which is
the reason why I didn‟t include the group factor „syntactic category of the antecedent‟ in
the quantitative analysis of oblique RCs.

(14) Tratar de cambiar al otro
i
, que todos hemos estado en eso
i
.
„Trying to change the others, which all of us have been there‟.

In Table 19, we can see that the hypothesis stating that non-restrictive RCs favor the
occurrence of RPs was confirmed. The percentage of non-restrictive RCs containing RPs
like (15) is greater than the percentage of restrictive RCs triggering RPs, as in (16): 21.4
130

per cent vs. 8.6 per cent, respectively, and these results are statistically significant (p =
.032).

(15) Había una chiquilla
i
, Cristina, que todo el mundo estaba detrás de ella
i
.
„There was a girl, Cristina, that everybody was after her‟.
(16) Es el… lo que quieren también los este… los jub... pensionistas
i
que hemos tenido
reuniones con ellos
i
.
„This is what... what the uh… reti... the pensioners that we have had meetings with
them, want‟.
131

Table 28. Oblique – RP – Restrictiveness

non-
restrictive restrictive total
No RP Count 33 96 129
% 78.6% 91.4% 87.8%
RP Count 9 9 18
% 21.4% 8.6% 12.2%
total Count 42 105 147
% 100.0% 100.0% 100.0%

p = .032

On the contrary, the hypothesis according to which RCs with indefinite NP
antecedents favor RPs was not confirmed by the data. In fact, we can see in Table 29 that
although the tendency of RPs to occur more in RCs with indefinite antecedents is still
maintained (17.5%) when compared to RCs with definite antecedents (8.9%), these
results are not statistically significant (p = .119).
132

Table 29. Oblique – RP – Definiteness

definite indefinite total
No RP Count 82 47 129
% 91.1% 82.5% 87.8%
RP Count 8 10 18
% 8.9% 17.5% 12.2%
total Count 90 57 147
% 100.0% 100.0% 100.0%

p = .119

The weight hypothesis, which stated that RCs that include material between the
complementizer que and the verb (material to left of the verb) are more prone to
triggering RPs was not confirmed for obliques. In Table 30, we can see that the tendency
of frequencies supports the hypothesis: RCs with material to the left trigger RPs 25% of
the time, followed from far away by RCs with material to the left and right of the verb
(13.6%), no material at all (5.3%), and material to the right (4.9%). These results were
not, however, statistically significant (p = .064).

133

Table 30. Oblique – RP – Weight

both sides right left light total
No RP Count 51 39 21 18 129
% 86.4% 95.1% 75.0% 94.7% 87.8%
RP Count 8 2 7 1 18
% 13.6% 4.9% 25.0% 5.3% 12.2%
total Count 59 41 28 19 147
% 100.0% 100.0% 100.0% 100.0% 100.0%

p = .064

The approach to the next factor for the case of the oblique position was slightly
different than it was for the subject and DO syntactic functions. For subject and DO RCs
we considered „adjacent‟ the absence of material between the antecedent NP head and
que, and „non-adjacent‟ the presence of any word between the antecedent NP head and
que.
However, standard oblique RCs include at least one word, i.e. a preposition
between the NP head and que, as con in (3). I decided to code „adjacent‟ those NPs that
didn‟t have a modifier between the NP head and que, irrespective of there being a
preposition between the antecedent NP head and que or not. Therefore, (3) and (4) were
coded both as adjacent. Only NP antecedent heads with modifiers, like de Perú in (5)
were coded „non-adjacent‟.
134

(3) O sea, lo que estamos haciendo es simplemente ordenar tres defectos con que salió la
ley.
„In other words, what we are doing is simply organizing three defects that the Law
came out with.‟
(4) Por un lado está la pleitista
i
, que ya me peleé con ella
i
.
„On the one hand, there was the argumentative one
i
, that I already had an argument
with her
i
.‟
(5) Sí, hay una flaca
i
de Perú que sí, todavía me comunico con ella
i
.
„Yes, there is a girl
i
from Peru that indeed, I am still in touch with her
i
.‟

Table 31 shows the results for the hypothesis which suggests that RPs show up
more frequently in RCs whose antecedent NP head is not adjacent to the complementizer
que, with which the RC begins. The results did not support this hypothesis for the oblique
position: 13.9 per cent of RCs with adjacent NP antecedent heads exhibited the presence
of RPS, whereas 10.3 per cent of RCs with non-adjacent NP antecedents caused RPs to
appear, with no statistical significance (p = .503). In other words, the vicinity of the
antecedent NP head to the complementizer que, appears to favor RP presence inside an
RC.

135

Table 31. Oblique – RP– Distance

adjacent
non-
adjacent total
No RP Count 68 61 129
% 86.1% 89.7% 87.8%
RP Count 11 7 18
% 13.9% 10.3% 12.2%
total Count 79 68 147
% 100.0% 100.0% 100.0%

p = .503

Table 32 shows that the hypothesis according to which RCs with medium and
large domains would favor more RPs than RCs with small domains was not supported by
the data. Moreover, oblique RCs with large domains have a complete lack of RP
presence, when compared to small (13.1%) and medium (12.5%) domains, but the results
are not statistically significant (p = .596).
136

Table 32. Oblique – RP– Domain

large medium small total
No RP

Count 7 49 73 129
% 100.0% 87.5% 86.9% 87.8%
RP Count 0 7 11 18
% .0% 12.5% 13.1% 12.2%
total Count 7 56 84 147
% 100.0% 100.0% 100.0% 100.0%

p = .596
In Table 33, we can see that the hypothesis according to which animate NP
antecedents trigger more RPs than inanimate antecedents was also confirmed for the
oblique data. 40 per cent of the RCs with an animate antecedent, whereas only 3.6 per
cent of RCs with inanimate antecedents show the occurrence of RPs, with statistically
significant results (p = .000).
137

Table 33. Oblique – RP – Animacy

animate inanimate total
No RP Count 21 108 129
% 60.0% 96.4% 87.8%
RP Count 14 4 18
% 40.0% 3.6% 12.2%
total Count 35 112 147
% 100.0% 100.0% 100.0%

p = .000

Unlike the case of DO RCs and similarly to subject RCs, the hypothesis stating that
antecedents that were mentioned for the first time would cause more RPs to occur was
not supported by the data, and the results are presented in table 34. The tendency stated
by the hypothesis is reversed, with a 15.6 per cent of RCs with previously mentioned
antecedents showing RPs compared to a 6 per cent of RCs with „brand-new‟ antecedents
triggering RPs, but the results are not statistically significant (p = .086).

138

Table 34. Oblique – RP – Previous Mention

mentioned
not
mentioned total
No RP Count 27 94 121
% 84.4% 94.0% 91.7%
RP Count 5 6 11
% 15.6% 6.0% 8.3%
total Count 32 100 132
% 100.0% 100.0% 100.0%

p = .086

Another hypothesis that was supported by the data was the one according to
which the second or subsequent RC in a set of coordinated RCs is very likely to trigger
an RP
20
. In Table 35, we can see that the hypothesis was confirmed: 50% of coordinated
RCs show RPs, whereas 11.2% of non-coordinated RCs show the presence of RPs. The
correlation is statistically significant (p = .020).

20
However, given that the number of coordinated oblique RCs is only 4, I would need a larger sample to
confirm this correlation.

139

Table 35. RP – Oblique – Coordination

not
coordinated coordinated total
No RP Count
127 2 129
%
88.8% 50.0% 87.8%
RP Count
16 2 18
%
11.2% 50.0% 12.2%
total Count
143 4 147
%
100.0% 100.0% 100.0%
p = .020

The hypothesis that the second and subsequent RCs in a juxtaposed RCs row
would trigger RPs more than an RC not in a juxtaposed context was not even tested
because there were no instances of juxtaposed oblique RCs. This does not entail
necessarily that the hypothesis is wrong; it simply means that there were no data to test
this hypothesis.
The following hypothesis concerning oblique RCs to be tested was one particular
to this syntactic position: the one stating that oblique RCs selected by a preposition rather
than by a verb are a more favorable context for the occurrence RPs, due to the relative
independence of the preposition-selected constituent (adjunct) as in (6), compared to the
dependency of the verb-selected constituent (complement), as in (7). The results are
presented in Table 36, and the tendency goes contrary to the hypothesis, with 20.7% of
verb-selected oblique RCs containing RPs, as opposed to 10.2% of preposition-selected
oblique RCs that contain RPs. However, the results are not statistically significant (p =
.122).
140

(6) No es algo
i
que le paguen por eso
i
. (J)
„It is not something that (they) pay him for it‟.
(7) Por un lado está la pleitista
i
, que ya me peleé con ella
i
. (I)
„On the one hand, there is the argumentative one, that (I) have already argued with
her‟.
Table 36. Oblique – RP – Selecting Category

verb
selected
preposition
selected total
No RP Count 23 106 129
% 79.3% 89.8% 87.8%
RP Count 6 12 18
% 20.7% 10.2% 12.2%
total Count 29 118 147
% 100.0% 100.0% 100.0%

p = .122
In Table 37, we can see that the hypothesis stating that full prepositions (para, por, sobre
in the data studied) would favor the presence of RPs, is supported by the data. As we can
see, 42.9 per cent of the oblique RCs introduced by full prepositions include RPs, and
only 9 per cent of the ones introduced by empty prepositions show RPs within them, with
statistically significant results (p = .000).

141

Table 37. Oblique – RP – Preposition Type
empty
preposition
full
preposition total
No RP Count 121 8 129
% 91.0% 57.1% 87.8%
RP Count 12 6 18
% 9.0% 42.9% 12.2%
total Count 133 14 147
% 100.0% 100.0% 100.0%

p = .000

In order to weight the factors‟ strength, I used Goldvarb 2001 for oblique RCs as
well. Given that there was only one instance of a sentential antecedent that would give a
knockout, it was excluded not only from the cross-tabulation analysis, but also from the
multivariate analysis. Also, given that there were no instances of juxtaposed RCs in the
oblique position, this factor group was not included in the statistical program. The factor
groups included when running the program were „restrictiveness of the RC‟, „definiteness
of the antecedent‟, „weight borne by the RC‟, „distance between the NP head and the
complementizer que‟, „domain size‟, „animacy of the antecedent‟, „previous mention of
the antecedent‟, „coordination of RCs‟, „category selecting the PP‟, and „preposition
type‟.
The results of this multivariate run are presented in Table 38. The strongest factor
is an animate antecedent, followed by full prepositions, indefinite antecedents, non-
142

restrictive RCs, and finally, adjacent RCs, a result that contradicts the hypothesis that the
further the RC is from its antecedent, the more likely it is to trigger an RP. The factor
groups that were eliminated by the program were weight of the RC, size of the domain,
coordinated RCs, and the selecting category of the PP.
Table 38. Oblique RPs Goldvarb
Input 0.025
Animacy of the antecedent
animate 0.961
inanimate 0.268
relative strength 0.693

Preposition type
full 0.945
empty 0.420
relative strength 0.525

Restrictiveness of the RC
non-restrictive 0.827
restrictive 0.349
relative strength 0.478

Definiteness of the antecedent
indefinite 0.778
definite 0.311
relative strength 0.467

Distance between antecedent and complementizer que
adjacent 0.666
non-adjacent 0.309
relative strength 0.357

Log likelihood = -28.170 Significance = 0.129

143

4.2 Prepositional Phrase Chopping Presence/Absence
The coding regarding RCs including PP-chopping was similar to that of oblique
RPs and different from subject and DO: the RCs that did have PP-chopping were coded
as „PP-chopping‟, as (8), and all the rest, including RP-containing RCs as (9), and gap-
containing RCs as (10), were coded as „no PP-chopping‟. In the following lines, the
results for the hypotheses formulated for this alternation are presented.

(8) Ese fue el fin de semana largo Øque tenía que trabajar. (I)
„That was the long weekend that he had to work‟.
(9) Los noisers, pe, los noisers
i
,que para ellos
i
no hay diferencia. (J)
„The noisers, yes, the noisers that there it makes no difference for them‟.
(10) Habían [sic] bastantes árboles a los que subíamos. (I)
„There were a lot of trees [to] which we would climb‟.

Table 39 shows that the hypothesis according to which restrictive RCs would
favor the occurrence of PP-chopping due to the tight relation the former have with their
antecedents was not confirmed. The percentages go along the lines of the hypothesis, i.e.
37.5 per cent of restrictive RCs exhibit PP-chopping, whereas 27.9 per cent of non-
restrictives show PP-chopping. However, these results are not statistically significant (p =
.266).
144

Table 39. Oblique PP –Restrictiveness

non-
restrictive restrictive total
No PP-
ch
Count 31 65 96
% 72.1% 62.5% 65.3%
PP-ch Count 12 39 51
% 27.9% 37.5% 34.7%
total Count 43 104 147
% 100.0% 100.0% 100.0%

p = .266
In Table 40, the results for the hypothesis suggesting that definite
antecedents would trigger more RPs than indefinites due to was not confirmed.
34.1% per cent of RCs with definite antecedents and 35.7 per cent of RCs with
indefinite antecedents show PP-chopping, a fact that rendered a non significant
outcome (p = .838).
145

Table 40 – Oblique PP –Definiteness

definite indefinite total
No PP-
ch
Count 60 36 96
% 65.9% 64.3% 65.3%
PP-ch Count 31 20 51
% 34.1% 35.7% 34.7%
total Count 91 56 147
% 100.0% 100.0% 100.0%

p = .838
The hypothesis that states that light RCs, i.e. the ones with no material inside
except for the verb would cause PP-chopping to take place was not supported by the data,
as seen in Table 41. Moreover, the pattern is similar to the one achieved for RPs, with
RCs with material to the left showing PP-chopping 40.7% of the time, and RCs with no
material showing the phenomenon only 15.8% of the time, with no statistical significance
(p = .212).
146

Table 41. Oblique PP –Weight

both
sides right left light total
No PP-
ch
Count 39 25 16 16 96
% 67.2% 58.1% 59.3% 84.2% 65.3%
PP-ch Count 19 18 11 3 51
% 32.8% 41.9% 40.7% 15.8% 34.7%
total Count 58 43 27 19 147
% 100.0% 100.0% 100.0% 100.0% 100.0%

p = .212
In table 42, we can see that the hypothesis according to which adjacent RCs trigger PP-
chopping more frequently than non-adjacent RCs was confirmed by our data. In fact,
73.3% of adjacent RCs favor the occurrence of PP-chopping, while only 17.6% of non-
adjacent RCs cause the PP to be deleted. The statistical significance is high (p = .000).

147

Table 42 – Oblique PP –Distance

adjacent
non-
adjacent total
No PP-
ch
Count 12
84
96
% 26.7% 82.4% 65.3%
PP-ch Count 33 18 51
% 73.3% 17.6% 34.7%
total Count 45 102 147
% 100.0% 100.0% 100.0%

p = .000
The hypothesis that stated that smaller domains would cause PP-chopping to
occur, due to the ease of processing that the former entail, was confirmed by the data. In
Table 43, we can see that oblique RCs with small domains show PP-chopping occurrence
46.4% of the time, as opposed to large domains (28.6%) and medium domains (16.1%).
These results are statistically significant (p = .001).
148

Table 43. Oblique PP – Domain

large medium small total
No PP-ch Count 5 47 45 97
% 71.4% 83.9% 53.6% 66.0%
PP-ch Count 2 9 39 50
% 28.6% 16.1% 46.4% 34.0%
total Count 7 56 84 147
% 100.0% 100.0% 100.0% 100.0%

p = .001

The results for the animacy factor are presented in Table 44. This hypothesis
states that inanimate antecedents cause PP-chopping to occur, due to reduced cognitive
prominence, and in fact, 39.3% of RCs with inanimate antecedents include PP-chopping,
compared to the 20% of RCs with an animate antecedent, which is a statistically
significant result (p = .036).
149

Table 44. Oblique PP – Animacy

animate inanimate total
No PP-
ch
Count 28 68 96
% 80.0% 60.7% 65.3%
PP-ch Count 7 44 51
% 20.0% 39.3% 34.7%
total Count 35 112 147
% 100.0% 100.0% 100.0%

p = .036
The hypothesis stating that PP-chopping would take place for RCs with
antecedents that had already been mentioned due to their ease of „recoverability‟ in
processing terms was supported by the data. Table 45 shows that 46.9% of RCs with
previously mentioned antecedents undergo PP-chopping, whereas only 27% of RCs
whose antecedents were mentioned for the first time exhibit PP-chopping, with
statistically significant results (p = .036).
150

Table 45. Oblique PP – Previous Mention

p = .036
The hypothesis according to which PPs would be deleted in RCs that are not
coordinated was not supported by the data, as seen in Table 46. Moreover, the tendency is
precisely the opposite, with only 34.3% of non-coordinated RCs triggering PP-chopping,
as opposed to 50% of coordinated RCs attesting PP deletion. However, the results are not
statistically significant (p = .514), therefore the reversed pattern is not confirmed either.

mention
no
mention total
No PP-
ch
Count 17 73 90
% 53.1% 73.0% 68.2%
PP-ch Count 15 27 42
% 46.9% 27.0% 31.8%
total Count 32 100 132
% 100.0% 100.0% 100.0%
151

Table 46. Oblique PP – Coordination

not
coordinated coordinated total
No PP-
ch
Count
94 2 96
%
65.7% 50.0% 65.3%
PP-ch Count 49 2 51
% 34.3% 50.0% 34.7%
total Count 143 4 147
% 100.0% 100.0% 100.0%

p = .514
The next hypothesis, according to which empty prepositions (a, con, de, en in the
present data set) favor the chopping of the PP, was not supported by the data, as Table 47
shows. The difference between the occurrence of PP-chopping with empty prepositions
(34.8%) and full prepositions (26.7%) follows the pattern proposed in the hypothesis, but
the results are not statistically significant (p = .526).
152

Table 47. Oblique PP – Preposition Type
empty
preposition
full
preposition total
No PP-
ch
Count
86 11 97
% 65.2% 73.3% 66.0%
PP-ch Count
46 4 50
% 34.8% 26.7% 34.0%
total Count 132 15 147
% 100.0% 100.0% 100.0%

p = .526
Table 48 shows the results for the hypothesis that PPs occur more when the RC
they contain is selected by a verb rather than by a preposition. As we can see, although
the tendency goes along the lines with this hypothesis, with 43.3% of the complement
PPs being chopped and 32.8% of the modifier PPs being deleted, the results are not
statistically significant (p = .265).
153

Table 48. Oblique PP – Selecting Category

verb preposition total
No PP-
ch
Count 17 79 96
% 56.7% 67.5% 65.3%
PP-ch Count 13 38 51
% 43.3% 32.5% 34.7%
total Count 30 117 147
% 100.0% 100.0% 100.0%

p = .265
The next step in the variable analysis was to enter the data in Goldvarb 2001 to
obtain the weight of the factors. Just like in the case of oblique RPs, the only factor group
not entered in the program was „juxtaposition of RCs‟, because there were no instances of
juxtaposed RCs in the oblique position. Therefore, the factor groups entered in the
program were the same ones that were used for the case of oblique RP presence/absence:
„restrictiveness of the RC‟, „definiteness of the antecedent‟, „weight borne by the RC‟,
„distance between the NP head and the complementizer que‟, „domain size‟, „animacy of
the antecedent‟, „previous mention of the antecedent‟, „coordination of RCs‟, „category
selecting the PP‟, and „preposition type‟.
The results are in Table 49, and they show that the hierarchy of factors favoring
PP-chopping occurrence is the following: previously mentioned antecedent, small
domain, and inanimate antecedents.

154

Table 49. Oblique PP-Chopping Goldvarb
Input 0.282

Previous mention of the antecedent
mentioned 0.708
not mentioned 0.430
relative strength 0.369

Domain Size
small 0.644
large 0.387
medium 0.303
relative strength 0.341

Animacy of the antecedent
inanimate 0.586
animate 0.247
relative strength 0.339

Log likelihood = -79.351 Significance = 0.078

4.3 Discussion of the Results
The most evident fact that the data shows is that the factors proposed are not
equally significant in favoring RP presence throughout the three syntactic positions
(subject, DO and oblique). In Table 50, we see the effect of the proposed factors for each
position. A check means that the hypothesis was confirmed, while an „x‟ means that the
hypothesis was not confirmed, either because the numbers were not statistically
significant or because another factor within the factor group turned out to be significant
contrary to the expectation/hypothesis
21
.

21
The factors included here are the ones that these three positions have in common. For obvious reasons,
the factors that were proposed exclusively for oblique RPs, i.e, hierarchy of the PP and preposition type are
not included here.

155

Table 50. Effect of conditioning factors across the three syntactic positions
Conditioning factor

Subject RP DO RP Oblique RP
Sentential antecedent   
Non-restrictive RC   
Animate antecedent   
Indefinite antecedent   
Material to the left of
the verb
  
Coordinated RC   
Juxtaposed RC   
Not previously
mentioned
antecedent
  
Large domain   
Non-adjacent RC   

The summary of the data shows that a sentential antecedent is significant for all
three positions. In subject RCs, 13 out of 23 RCs with sentential antecedents (56.5%)
showed RPs; in the case of DOs, 5 out of 5 (100%) triggered RPs; and in the oblique
position, one out of one (100%) included an RP. Although this factor was not included in
the multivariate analyses as explained 4.1.1, 4.1.2, and 4.1.3, it seems clear that it
definitely plays a role in the occurrence of RPs in Spanish RCs.
A non-restrictive RC proved to be a strong conditioning factor for all positions, in
accordance with the previous studies in Spanish RPs. And an animate antecedent also
156

plays an important role in triggering RPs, along the lines of Silva-Corvalán (1996). These
two factors, together with a sentential antecedent, are the only ones that are significant
across the three syntactic positions.
An indefinite antecedent does not appear to favor the presence of subject RPs, but
it does have significance in producing DO and oblique RPs. Material between que and
the verb proved to be only significant in causing DO RPs to occur, but were non-
significant for oblique RPs. In the case of subject RPs, although the crosstabs gave non-
significant results for this factor, the multivariate program did show that „material on both
sides of the verb‟ was significant, against the original hypothesis that stated that only
material to left of the verb would favor RP-occurrence. Although material on both sides
is a different factor, it does include the expected factor, material between que and the
verb, and further research should be done to solve this question. What is clear is that
„light‟ RCs with no material besides the verb do not favor RPs almost categorically.
The factor regarding the presence of RPs in the second or third RC in a
coordination sequence was not confirmed for subject. Probably, the inflection agreement
with the verb is used to keep track of an antecedent that is shared with previous RCs. This
factor was confirmed for DO RCs, in which the RP itself would function as a DO
agreement, as it has been said to happen even in main clauses (Franco, 1993). In the case
of obliques, although the cross-tabs rendered significant results for RPs in coordinated
RCs, the multivariate program eliminated the factor group, probably because there were
143 instances of non-coordinated RCs, and only 4 instances of coordinated RCs. More
tokens of coordinated oblique RCs would be needed in order to have a more definitive
answer.
157

A similar case is the one regarding the second or third RC in a series of
juxtaposed RCs. This factor was not significant for the subject position, as opposed to the
DO position, in which it turned out to be statistically significant. In the oblique position,
there were no instances of juxtaposed RCs at all, so the factor was eliminated before
running the multivariate program, as explained above. If we compare this case with the
case of coordinated oblique RCs, we see that there is a tendency of not aligning oblique
RCs in a sequence, a fact that may it even harder to test both hypotheses with definitive
results.
Similarly to the three previous cases, the factor „not previously mentioned
antecedent‟ did not show statistical significance in either the subject or oblique positions,
but it was statistically significant for the DO position in the cross-tabulations, with a
pattern contradicting the hypothesis. One possibility is that there is a confound factor
interacting with „previously mentioned antecedent‟, i.e. the fact that even in main clauses
there tend to be DO pronouns after the lexical DO has already been mentioned.
The factor labeled as „large domain‟ was not significant in any of the three
positions. Moreover, it gave a knockout result in the subject position and it had to be
removed from the quantitative analysis. The striking fact is that the factor labeled „small
domain‟ was significant for the subject and DO positions in the multivariate analysis.
This significance underlines the fact that in the data set studied there seems to be no
relation at all between an RP and a large domain of the RC, since RPs occur much more
frequently in RCs with small domains.
Finally, the factor „non-adjacent RC‟ was not significant for the subject position.
In the case of DO RCs, the crosstabs results do show statistical significance for this
158

factor, but the multivariate results reversed the trend, i.e. the factor „adjacent RC‟ turned
out to be significant. Similarly, for the oblique position, although the cross-tabulations
show no statistical significance, the probabilistic analysis renders the factor „adjacent‟ as
likely to predict the occurrence of RPs. However, in general, these results of a lack of a
significant relation between the greater distance of an RC from its antecedent go along
the lines of RCs with large domains not triggering RPs. In other words, a greater distance
between the antecedent‟s head and que, and a greater number of words intervening
between the antecedent‟s head and the verb, seem not to make the recoverability of the
antecedent‟s function inside the RC more difficult.

4.4 PP-Chopping and RP Overall Comparison
A comparison between RP presence and PP-chopping in the oblique position is
presented in Table 51. The general hypothesis, based on Tarallo (1986), was that the two
types of construction would stand in complementary distribution, meaning that they
would obey opposite factors within a single factor group. In terms of the convention
adopted for this table, a complementary distribution would entail having a check on both
structure‟s column, because each of the factors in these columns are opposites within a
factor group.
However, the pattern does not seem to be the case in Spanish. Besides the clear
contrast between RP presence with animate antecedents and PP-chopping with inanimate
antecedents, there is no complementary distribution pattern. What seems to happen here
is that each syntactic process obeys its own set of factors, irrespective of what factors
operate on the other one.
159

Table 51. Oblique RP / PP-Chopping Factor Significance Comparison
Conditioning
Factor
RP Conditioning
Factor
PP-chopping
Non-restrictive
RC
 Restrictive RC 
Animate
antecedent
 Inanimate
antecedent

Indefinite
antecedent
 Definite
antecedent

Material to the
left of the verb
 No material 
Coordinated RC  Non-coordinated
RC

Juxtaposed RC  Non-juxtaposed
RC

Not previously
mentioned
antecedent
 Previously
mentioned
antecedent

Large domain  Small domain 
Non-adjacent
RC
 Adjacent RC 
Full preposition  Empty
preposition

Verb-selected  Preposition-
selected


With regard to the factor group „domain‟, in particular the factor „small domain‟,
it seems that small domains do favor PP-chopping in a consistent way. Although the
related factor „adjacent antecedent‟ was eliminated in the binomial stepping up and down
for PP-chopping, it was a highly significant factor in the crosstabs analysis (Table 42, p =
.000). More data are needed to confirm this hypothesis, but the fact that PP-chopping
occurs more in small domains and obeys a much more limited set of factors than RPs
may suggest that it is unlikely to have confound factors interacting, and that a small
domain is in fact a favorable environment for PP-chopping to take place, since the
processing of small domains should be intrinsically easy.
160

4.5 Qualitative Analysis: Contrast and Subject RPs
When analyzing the subject data in more detail, I found an interesting type of RP
that was not attested in the other positions. Thus, sentences like (11) - (17) share a
common feature that sets them apart when compared to subject RPs that have a sentential
antecedent: the former are used in a contrastive way.

(11) A la otra chica le paga $20 por 25 minutos, y a mí
i
, que yo
i
sí estuve entrenando
hasta tres grupos, no me dio nada. (J)
„(She) pays the other girl $20 for 25 minutes, and to me
i
, that I
i
in fact was training up
to three groups, (she) gave nothing.‟
(12) La otra vez estuve hablando con una amiga
i
que ella
i
sí no tiene problemas. (I)
„The other day (I) was talking with a friend
i
that she
i
indeed has no problems‟
(13) Me recetaron unas pastillas
i
, que esas
i
sí son buenas. (J)
„(They) prescribed me some pills
i
that those
i
indeed are good‟
(14) Ahora, pero tú ya has visto cómo soy yo
i
también, que yo
i
sí...no puedo estar quieta.
(I)
„But you have seen how I
i
am, also, that I
i
definitely…cannot be still/quiet‟.
(15) Me quedo con la otra gemela
i
que ella
i
está sola. (I)
„(I) choose the other [identical] twin
i
, that she
i
is single.‟
(16) Bueno, yo no voy a estar, pero le digo a Laura
i
, que ella
i
va a estar acá. (J)
„Well I am not going to be here, but I will tell Laura
i
, that she
i
will be here‟.
(17) Mejor habla con Álvaro
i
, que él es… que él
i
hace eso. (J)
„You‟d better talk to Alvaro
i
, that he is... that he
i
does that‟.
161

I understand contrast as a pragmatic function characterized by two main
properties: a) the existence of a set of competing alternatives to the contrastive element,
and b) stress in the contrastive element (Molnár, 2001). In the following lines, I will
demonstrate that the aforementioned subject RPs meet both requirements.
The first requirement I will concentrate on is the set of competing candidates.
Sentences (11) – (14) have the extra aid of the emphatic adverb sí. Thus in (11), there is a
clear contrast between yo and la otra chica, in which case the property „to work with up
to three groups‟ applies only to yo, and excludes la otra chica. By the same token, in
(12), the speaker restricts the absence of problems to her friend, as opposed to herself,
who, almost immediately before, was talking about being in the middle of an economic
and sentimental crisis. In (13), the speaker was talking about different painkillers, and the
lack of effectiveness that most have, but then she was prescribed a type of painkillers
which, in contrast with all the others spoken of, does work. Finally, in (14), the speaker is
contrasting the interviewer‟s stillness or lack of activity with her own, by excluding him
from the characteristic of being resttless.
In sentences (15) – (17), although there is no express excluding adverb as sí, the
contextual information completes the idea of a set of candidates that is excluded. Hence
in (15), the property of being single only applies to one of the two identical twins,
because the other one is engaged. In (16), the speaker excludes herself from the property
of being home, which in this case is restricted to her sister, Laura. Finally, in (17), the
speaker excludes herself from the expertise in grilling meat, which is restricted to her
husband, Alvaro.
162

Regarding the stress requirement, the criterion I used to identify stress is a rise in
pitch in the RP, based on Ladefoged (1975: 225), who states that “a much more important
indication of stress is the rise in pitch that usually occurs.” In order to prove this, I used
the phonetics program Praat to obtain the pitch contour of these contrastive RPs.
Thus, Figure 1, corresponding to (9), repeated below as (18), shows how there is a
pitch rise corresponding to the subject RP yo, in accordance with my hypothesis.
(18) Ahora, pero tú ya has visto cómo soy yo
i
también, que yo
i
sí...no puedo estar quieta.
„But you have seen how I
i
am, also, that I
i
definitely…cannot be still/quiet‟.

Figure 1. Subject RP Rising Pitch
Time (s)
3.95239 5.84829
Pitch (Hz)
75
500
3.95239 5.84829
queyosicontexto
que yo sí no
puedo estar quieta quieta

However, upon analyzing the pitch contour of other RCs containing subject RPs, I
did not find the same clear rising pitch. When we look at Figure 2, corresponding to (19),
it is true that the pitch in the RP él is higher than in cuando nos fuimos and than in con él.
163

Then again, when we compare él to the immediately previous context, de campamento,
there seems to be a lowering in pitch in the RP él.
(19) Había un tipo que... se volvía loco.... o sea, no loco, ¿no? pero era tan tonto... era
un tipo
i
tonto de la base... que, cuando nos fuimos de campamento, él
i
... con el
profesor, con toda la gente, se fue a la montaña a decir que iba a hablar con no sé qué
espíritu.
„There was a guy that… went crazy… I mean, not crazy, right? But he was such a
fool… (he) was a pumpkin head guy
i
that, when we went camping, he
i…
with the
professor, with everybody, went to the mountain to say that (he) was going to speak to
God-knows-what spirit‟.

Figure 2 – Subject RP Lowering Pitch #1
Time (s)
5.82989 9.05855
Pitch (Hz)
75
500
5.82989 9.05855
01tonto_wav_file
que cuando nos fuimos
de cam pa mento el con el profesor

In this case, given that there is no contrastive effect, there is no pitch rise in él. In
addition, when we observe the pitch contour of utterance (20), in Figure 3, we can see
that there is a lowering on the subject RP ella instead of a pitch rising, and it also seems
164

that the subject RP is cliticized to the complementizer que, rendering a sequence like
[kéja].

(20) Tenía una amiga
i
que ella
i
... en la universidad leía cartas, le gustaba la Ouija...
„(I) had a friend
i
that she
i
… during college did card reading, she liked the Ouija
[board].

Figure 3 – Subject RP Lowering Pitch #2
Time (s)
11.7179 13.3358
Pitch (Hz)
75
500
11.7179 13.3358
ouija
una amiga una amiga que
ella ella
en la universidad

It must be pointed out that I have not been able to analyze a considerable number
of subject RPs with the Praat software. The reason is that, as stated in the data description
in 3.1, most of my subject RPs do not belong to recorded tapes, but rather to jotted down
data, on the one hand, and to 1978 transcriptions of acoustic files, on the other.
However, the pitch contours of the previous utterances suggest that there are two
types of subject RPs: „contrastive‟ as in (18), and real „resumptive‟ pronouns, as in (19) -
165

(20). Although it seems to be a clear determining factor, contrast was never mentioned
before in any study regarding RPs. It was difficult to carry a quantitative analysis of
contrastive vs. non-contrastive subject RCs, due to the fact that some of the jotted-down
sentences did not have a context out of which the set of competing candidates could be
retrieved.

166

Chapter 5. Conclusions

The previous chapters of this dissertation have presented the results of a
variationist corpus study conducted to explore the factors that favor the preference for
two non-standard RC production structures in Spanish: RPs and PP-chopping. In this
final chapter I will 1) remind the reader of my initial research questions and the structure
of the dissertation; 2) summarize the major findings encountered in the study; and 3)
discuss the implications of the results obtained by this research not only regarding the
understanding of RC production in Spanish, but also with respect to universal typological
accounts of RC production.
5.1 Aims of the Dissertation
The main goal of this dissertation has been to investigate the factors that favor the
occurrence of two non-standard constructions in Spanish RCs introduced by the
complementizer que. These structures are, on the one hand, RCs containing RPs, a
pronominal element that repeats the function of the relativized element inside the RC; on
the other hand, oblique RCs in which the preposition that introduces the oblique adjunct
is chopped off. A second goal, clearly dependent on the first one, has been to determine
whether the factors favoring RP occurrence are the same across all the syntactic positions
relativized. A third goal of this study has been to find out whether RPs and PP-chopping
stand in complementary distribution in the oblique syntactic position. The fourth goal of
this study has been to explore whether the factors favoring RP occurrence can be
predicted by universal principles or are language-specific.
167

In order to explain the occurrence of such structures, I described the Spanish RC
marker paradigm with particular reference to the complementizer que, I listed and
illustrated the general types of RCs allowed in Spanish, and presented the two non-
standard RC structures in contrast with their corresponding standard gap-containing
counterparts (Chapter 1).
The literature reviewed in this dissertation (Chapter 2) approached the occurrence
of RPs either from a typological universal point of view according to which such
pronouns are used within RCs to facilitate the processing of an otherwise difficult
antecedent + RC complex (Comrie, 1981; Hawkins, 1994), or from a language-specific
perspective according to which there is a set of concurring factors that favor the
occurrence of RPs within RCs (Prince, 1990; Silva-Corvalán, 1996; Bentivoglio, 2003).
Regarding oblique RCs with chopped PPs, the only attested study that I reviewed
(Tarallo, 1986) approached the occurrence of this construction from a language-specific
quantitative point of view, which suggests that RCs with PP-chopping stand in a
complementary distribution with RP-containing RCs.
Based on the principles of variationist studies (Labov, 1972), I used data
consisting of transcriptions of sociolinguistic interviews of twenty-six native speakers of
the Peruvian Spanish variety of Lima, the capital of Peru (Chapter 3). However, due to
the scarcity of syntactic variables and the Observer‟s Paradox (Labov, 1972), additional
data sets were required. These were, on the one hand, jotted-down utterances of RCs
containing RPs or PP-chopping, which the aforementioned speakers of the interviews
produced after the interview was over. On the other hand, I added data from
approximately twelve hours of sessions of the Peruvian Congress. As a result, the total
168

number of RCs analyzed in this study is 2062, a number that allowed me to study these
phenomena in great detail.
Following the sociolinguistic methodology, I proposed conditioning factor groups
for the occurrence of RPs in three different syntactic positions: subject, DO, and oblique.
The proposed RP-favoring factors within each group are sentential antecedents, non-
restrictive RCs, indefinite antecedents, antecedents mentioned for the first time,
antecedents not adjacent to que, animate antecedents, large RC domains, RCs with
material between que and the verb, coordinated RCs and juxtaposed RCs. Following
Tarallo (1975), I assumed that the opposite values in each factor group would favor PP-
chopping, if it were the case that such construction were in complementary distribution
with RP-occurrence. Also, I proposed two PP-chopping exclusive conditioning factors:
empty prepositions and oblique RCs introduced by a verb. All of these factors were
coded and entered in two statistical programs: SPSS 11.5 and Goldvarb 2001. SPSS 11.5
was used for descriptive purposes of the data, i.e., to identify the distribution of the
correlations between the dependent variable and the independent variables (i.e. factors)
and the statistical significance of such correlations; Goldvarb 2001 was used for a
multivariate analysis, in order to rank the factors according to their predictive and
probabilistic strength.
Finally, I presented and interpreted the results of the statistical analyses, and also
offered a qualitative analysis of cases that were not explained quantitatively, in particular,
subject RPs that are apparently used to convey a contrastive function, for which purpose I
presented acoustic evidence (Chapter 4).
169

5.2 Summary of Findings
In this section, I review the results obtained in this dissertation with regard to the
research questions posed in chapter 1.

5.2.1 Conditioning Factors and Their Uniformity across Syntactic Positions
First, I will review the results regarding the factors that favor RP-occurrence. To
begin with, out of the ten factors analyzed, only three turned out to be common to
subject, DO, and oblique RPs: non-restrictive RCs, sentential antecedents, and animate
antecedents. Such a small number of coincidental factors suggests that speakers have
different motivations when they produce RPs in each of the syntactic positions
relativized.
Besides the three aforementioned common factors, subject RPs are favored by
„heavy‟ RCs that have material at both sides of the verb and RCs including material
between que and the verb. My qualitative analysis shows that there is yet another
motivation for certain subject RPs, which is the pragmatic function of contrast, ignored in
all the previous studies of this phenomenon. DO RPs are favored not only by the three
previously mentioned common factors, but also by juxtaposed and coordinated RCs,
indefinite antecedents, „heavy‟ RCs with material between que and the verb, and material
at both sides of the verb. Oblique RCs are favored, besides the three common factors, by
semantically full prepositions and indefinite antecedents.
There were two proposed factors that did not show any significant correlation
with the occurrence of RPs in subject, DO, and oblique positions: first-mention
antecedents and RCs not adjacent to their respective antecedents. In the first case, the
170

hypothesis was contradicted by DO RPs, which showed up more frequently with
previously mentioned antecedents, a pattern not consistent with subject and oblique RCs.
In the second case, no significant results were attested for subject or oblique RPs;
however, DO RPs did show up with non-adjacent antecedents in the data, but the
probabilistic prediction showed precisely the contrary results: that DO RPs are more
likely to occur in RCs with adjacent antecedents than in RCs with non-adjacent
antecedents.
The oblique-only proposed factor stating that PPs subcategorized by a preposition
rather than PPs subcategorized by a verb favor RPs was not supported by the data. In
other words, it does not seem to make any difference whether the oblique PP is adjunct
(preposition-selected) or a complement (verb-selected): oblique RPs will occur due to
other factors.
In sum, returning to the first and second research questions regarding RP
occurrence, different types of factors, such as processing, pragmatic, semantic and
morpho-syntactic constraints are found in the production of RPs in Spanish RCs. No
single factor can predict the occurrence of RPs by itself. Moreover, these factors do not
operate uniformly in all the positions relativized: different syntactic positions are favored
by different sets of conditioning contexts.
Secondly, I will review the results regarding PP-chopping in oblique RCs. This
construction is favored by small domains, previously mentioned antecedents, and
inanimate antecedents. It appears that PP-chopping takes place when there is low
memory load required to retrieve the function of the relativized element within the RC.
On the one hand, the few words needed to parse the complex antecedent + RC make it
171

easy for the speaker to keep track of the relativized element, therefore the chopping of the
PP. On the other hand, a previously mentioned antecedent, which poses no memory
difficulty in retrieving its reference, can be chopped along with the preposition without
jeopardizing its referential content. Inanimate antecedents, who are less likely to function
as subjects inside an RC, and therefore pose no ambiguity in the parsing of the antecedent
+ RC complex, can have the preposition chopped without causing an extra processing
task for the speaker.
In contrast, Tarallo‟s (1986) results for Brazilian Portuguese, in which PP-
chopping takes place in RCs that are tightly connected to their antecedents, such as
restrictive RCs, RCs adjacent to their respective antecedents, were not confirmed in my
study of Limeño Spanish.

5.2.2 Oblique Resumptive Pronouns and Prepositional Phrase-Chopping
Regarding the third question of this dissertation, my results show that RPs and
PP-chopping in oblique RCs do not stand in complementary distribution, contrary to what
Tarallo (1986) found in the case of Brazilian Portuguese. Out of the twelve different
conditioning factor groups proposed, there was only one in which the two opposite
factors in the factor group favored RPs and PP-chopping respectively. This was the case
of animate antecedents favoring RPs and inanimate antecedents favoring PP-chopping. I
will return to the case of animacy in 5.3.
The other factor groups that proved to be in complementary distribution in
Tarallo‟s study, such as restrictiveness of the clause and distance between the antecedent
and the complementizer que yielded very different results in my study. First, although
172

non-restrictive RCs were a favoring context for RPs to occur, restrictive RCs were not a
statistically significant favoring context for PP-chopping to take place. Secondly,
contrary to Tarallo‟s findings, RCs adjacent to their antecedents were a statistically
significant favoring context for RPs to occur, whereas the distance between the
antecedent and que seems to play no role in the presence of PP-chopping.
In my study, oblique RPs are favored by semantically full prepositions and
indefinite antecedents. PP-chopping is not favored by either factor‟s counterparts: neither
semantically empty prepositions nor definite antecedents proved to be a fruitful context
for PP-chopping to take place. By the same token, PP-chopping is favored by antecedents
that have been mentioned previously, but first-mention antecedents do not seem to be a
conditioning factor of oblique RP presence.
In general, it seems clear that in Spanish oblique RCs, the RP and PP-chopping
strategies are two independent phenomena, each of which obeys its own set of
constraints. Tarallo‟s (1986) findings for Brazilian Portuguese, according to which RPs
and PP-chopping are in complementary distribution, were not attested in Spanish, and are
likely to be language-specific.

5.2.3 The Universality of Resumptive Pronoun Factors
It does not seem that RPs can be predicted by universal conditioning factors. In
particular, Comrie‟s (1981) Accessibility Hierarchy (AH) and Hawkins‟s (2004)
Resumptive Pronoun Hierarchy Prediction (RPHP), the conjoint processing account with
universal predictions for the occurrence of RPs, were not supported by the Spanish data.
173

This account was contradicted in two ways: DO RPs are much more frequent than
oblique RPs, and subject RPs do show up in Spanish in a fairly consistent way. First, the
AH states that oblique RCs are harder to process than DO RCs, and the RPHP predicts
that due to that difficulty in processing, RPs should occur more frequently in oblique RCs
than in DO RCs. In Spanish, each syntactic position relativized obeys different sets of
factors, and DO RCs are a favorable context for RPs to occur. In addition, it is possible
that the antecedents of DO RCs are being analyzed as fronted DOs. Since DO pronouns
in main clauses are mandatory with a preverbal DO, this pattern could be extending to
DO RCs, in which case they could become obligatory at some point. This could help
explain the greater frequency of DO RPs than of oblique RPs in Spanish RCs.
Secondly, the AH states that subject RCs are the easiest ones to process, and the
RPHP predicts no RP occurrence in this syntactic position. However, Spanish, a null-
subject language that does not require overt subjects even in main clauses, allows subject
RPs to occur in RCs. This inconsistency can be explained as follows: the AH and RPHP
propose a universal cause for RP occurrence based exclusively on the processing
difficulty of the different syntactic positions relativized; however, in Spanish subject
RCs, RPs show up due to factors that are different in nature. The factors favoring subject
RPs are pragmatic (contrast), syntactic (non-restrictive RCs, sentential antecedents), and
not only related to processing difficulty (heavy RCs).
Although my study does not support the account stating that RPs in RCs are
universally predictable due to processing facilitation, there is one factor that shows
resemblance to studies in other languages. Just like Prince‟s (1990) and Tarallo‟s (1986)
studies, this dissertation shows that one of the most favorable contexts for RPs to occur,
174

which has been found in English, Yiddish, Spanish and Portuguese, is non-restrictive
RCs. Whether one embraces Prince‟s „file card‟ theory or Tarallo‟s „loose connection‟
explanation, it seems clear that speakers of these languages tend to employ different
strategies when producing RCs that are an adjunct to their respective heads (non-
restrictive), than when producing RCs that are complements of their heads (restrictive
RCs). Although I make no claims of the universality of this factor, it is the only cross-
linguistic factor that has been attested in studies with spoken data.
In general, a detailed study of Spanish RPs in subject, DO and oblique RCs has
shown that there is no single factor that explains RP occurrence in all positions, and how
different factors are operating in each position. Although more in-depth cross-linguistic
studies of RPs in RCs could shed more light on the universality of their causes, my study
shows no evidence that the factors favoring RPs in Spanish can be reduced to universal
principles. What seems clear is that any universal account based exclusively on the
processing difficulty of RCs is unable to capture the occurrence of RPs in Spanish RCs.

5.3 Implications and Future Research
The general implication in RP occurrence that this dissertation has provided is
that the case of Spanish cannot be explained by a single processing efficiency account,
and that such account‟s universality is questioned by my data and results.
My results challenge Hawkins‟s (2004) Domain Minimization processing
hypothesis, since RPs are more commonly found in domains that are small (i.e. easier to
process) than in domains that are large (harder to process). This suggests two
possibilities: a) RPs in Spanish are not used to ease the processing of RCs at all, or b) the
175

notion of „domain‟ (Hawkins, 2004), should be reformulated perhaps to include only
elements „internal‟ to the RC. The results showing that adjacent antecedents (i.e. no
„external‟ elements between the antecedent and que) may be a more felicitous context for
RPs to occur in DO and oblique positions support the second possibility. Another result
supporting the possibility that the notion of domain may be reformulated in terms of what
is inside the RC is the one related to the group factor I labeled „weight‟. In subject and
DO relativizations,
22
RCs with material between que and the verb and material at both
sides of the verb are favorable contexts for RPs to occur. This may suggest that RPs
occur not to ease the retrieval of a distant antecedent, but rather to clarify the function of
the relativized element inside a large RC, in which such function could be hampered by
the presence of material between que and the verb.
Also, my results present evidence against the AH (Comrie, 1981) and the RPHP
(Hawkins, 2004). As mentioned in 5.2.3, RPs in Spanish behave differently from what
the AH and RPHP would predict. In other words, if one follows the aforementioned
theory in a strict sense, oblique RCs would not be harder to process than DO RCs in
Spanish, because the latter trigger more RPs than the former. However, a more feasible
explanation is that there are other factors, not only processing ones, intervening in the
production of RPs within RCs, which cause DO RCs to be a more favorable context for
RPs to show up. By the same token, contrary to the AH and RPHP prediction, subject
RCs would be a somewhat hard to process context in Spanish, given that RPs occur in

22
In oblique RCs the same tendency was found, although the results were not statistically significant (p =
.064). Perhaps a larger sample of oblique RCs could show significant results.

176

this syntactic position. However, it seems more appropriate to acknowledge that there are
other factors of different nature involved in the production of subject RPs.
In general, my results are compatible with the ones presented in Silva-Corvalán
(1996). This is true in three respects. First, she proposed the occurrence of RPs in Spanish
RCs to be an instance of multiple causation, and my study firmly supports that statement.
Secondly, the common factors that were analyzed quantitatively in Silva-Corvalán‟s
study and mine showed basically the same results. Finally, the qualitative hypothesis that
Silva-Corvalán suggested in her study, according to which „heavy‟ RCs would be harder
to process, therefore favoring the occurrence of RPs, was supported in this dissertation in
my quantitative analysis.
In a more specific way, one of the factors common to subject, DO and oblique
RC, Animacy, shows support for Comrie‟s (1981) Animacy Hierarchy. In general, one of
the concepts involved in the Animacy Hierarchy is that NPs with an animate referent tend
to exhibit a richer morpho-syntactic system than NPs with an inanimate referent.
Thus, in declension languages like Latin, Classical Greek, and Sanskrit, neuter
nouns (generally referring to inanimate entities) show a common morpheme for the plural
of the nominative and accusative cases, whereas masculine and feminine nouns
(generally referring to animate entities) had separate morphemes for each of these cases
(Ramat et al., 1995). By the same token, in Classical Greek, neuter plural subjects agree
with a singular verb, whereas masculine or feminine plural subjects agree with a plural
verb (Berenguer Amenós, 1969). Also, in Latin passive constructions, masculine and
feminine ablative agents are headed by the preposition ab/a, whereas neuter agents are
177

not headed by any preposition (Valentí Fiol, 1999). In Russian, animate accusatives have
a special ending that inanimate accusatives lack (Comrie, 1981).
The fact that in Spanish, RCs with animate antecedents show a greater frequency
of RPs in subject, DO and oblique positions goes along the lines of the cross-linguistic
tendency for a richer morphological marking of animate entities. Moreover, oblique RCs
with inanimate antecedents favor the occurrence of PP-chopping, a fact that also supports
the idea that animate NPs tend to have more sophisticated morpho-syntactic correlations
than inanimate NPs.
There are four topics that I consider worth exploring in future research, in order to
complement the findings in my dissertation. The first topic is the syntactic function of the
antecedent in the main clause. Given that the antecedent of an RC is an element with two
functions, i.e. one in the main clause and one in the subordinate clause, it could be that
there are some factors regarding the former function that play a role in the occurrence of
RPs or PP-chopping. Indeed, the Real Academia de la Lengua Española suggests that
when the antecedent of an oblique RC is, in turn, included in an oblique, if both
prepositions are the same, the preposition of the subordinate clause may be deleted. The
role of such main clause functions should be studied from a quantitative perspective, in
order to establish or discard a relationship with the occurrence of any of the non-standard
strategies presented in this dissertation.
The second topic I think would be worthwhile studying is whether or not contrast
and sentential antecedents in subject RCs have a similar effect in other languages, such as
English. My impressionistic a priori view, based on simple observation, is that English
does have such contrastive subject RPs, even in RCs headed by the relative pronoun
178

which. Also, I have noticed that some sentential antecedents in RCs headed by that
include it as a subject RP. If this pre-scientific impression can be backed by a quantitative
and qualitative study of English RCs, it would mean that there are at least, three Indo
European languages (Spanish, Greek, English) in which some subject RPs are
contrastive, and two languages (Spanish, English) who have subject RPs that follow a
sentential antecedent. Perhaps studies on other non-Indo European languages would show
that these are factors shared cross-linguistically by languages that belong to different
linguistic families.
The third topic that I consider worth studying is whether or not the Spanish main
clause structure of fronted DOs and their almost obligatory pronoun copy is being
extended to the antecedent + RC structure, therefore triggering an RP that mimics the
almost mandatory DO pronoun of the main clause. I think such a study should be
approached from a psycholinguistics perspective, through an experiment that includes
similar subjects, DOs, etc. but in which some structures are main clauses with fronted
DOs, and some are DO RCs. In the case of the latter, all the favoring contexts should be
controlled for, in order to avoid confound factors.
The fourth and last topic that I would pursue in further research is to do more in-
depth study on the effect of animacy in the processing of RCs. In Dutch, DO RCs with an
inanimate DO and an animate subject, like „the rock that the hikers climbed…‟, were
easier to parse (shorter time) than DO RCs with an animate subject and an animate DO,
like „the professor that the students have met…‟ in an experiment by Mak, Vonk and
Schriefers (2006). If this were true, in a language that allows RPs like Spanish, the
occurrence of DO RPs in RCs with animate antecedents would be a means of
179

disambiguating one of the two possible animate subjects, by marking it with an RP inside
the RC. A psycholinguistics experiment such as the aforementioned one could be carried
out in Spanish, to determine whether there is such a greater difficulty in processing DO
RCs with animate subjects and animate DO antecedents.
To conclude, this dissertation has provided an in-depth study about the production
of RPs and PP-chopping in Spanish RCs. Because of the amount of data gathered, it was
possible to study RPs separately in subject, DO, and oblique RCs for the first time. The
results show that Spanish RP production is rather a very complex phenomenon, and that
it is not likely to be explained by a single factor, nor does it seem to be accountable for by
universal principles. Nonetheless, some of the factors found for the first time in this
dissertation that contribute to RP presence in Spanish may be also found in other
languages. Although the attesting of these common factors may not be necessarily
universal, these factors may be taken into consideration in future studies of RC
production including RPs.

180

Bibliography

Abbott, Barbara. 2004. “Definiteness and Indefiniteness”. In Horn, Lawrence and
Gregory Ward (eds.), The Handbook of Pragmatics, 122-149, Oxford, Blackwell
Publishing.

Alarcos Llorach, Emilio. 1994. Gramática de la Lengua Española. Madrid: Espasa-
Calpe.

Alcina, Juan, and Blecua, J.M. 1991. Gramática española. Barcelona: Ariel.

Bentivoglio, Paola. 1987. Los sujetos pronominales de primera persona en el habla de
Caracas. Caracas: Universidad Central de Venezuela.

Bentivoglio, Paola. 2003. “Las construcciones „de retoma‟ en las cláusulas relativas: un
análisis variacionista”. In Francisco Moreno Fernández, José Antonio Samper
Padilla, María Vaquero, María Luz Gutiérrez Araus, César Hernández Alonso,
and Francisco Gimeno Menéndez (eds.), Lengua, variación y contexto: Estudios
dedicados a Humberto López-Morales. Madrid: Arco Libros, 507-520.

Berenguer Amenós, Jaime. 1969. Gramática griega. Barcelona: Bosch, Casa Editorial
S.A.

Biber, Douglas and Edward Finegan. 1994. “Register and social dialect variation: an
integrated approach”. In Douglas Biber and Edward Finegan (eds.) Sociolinguistic
Perspectives on Register. New York and Oxford: Oxford University Press, 315-
347.

Brucart, José María. 1999. “La estructura del sintagma nominal: Las oraciones de
relativo”. In: Bosque, Ignacio and Demonte, Violeta (eds.), Gramática descriptiva
de la lengua española , vol. 1.: Sintaxis básica de las clases de palabras. Madrid:
Espasa, 395-523.

Caravedo, Rocío. 1989. Español de Lima. Materiales para el estudio del habla culta.
Lima: Pontificia Universidad Católica del Perú.

Cerrón-Palomino, Álvaro. 2006. “Pronombres de retoma en cláusulas relativas del
castellano peruano: un fenómeno de causación múltiple”. In Lexis XXIX (2), 231-
258.
Comrie, Bernard. 1981. Language universals and linguistic typology. Chicago:
University of Chicago Press.
Coseriu, Eugenio. 1991. Principios de Semántica Estructural. Madrid: Gredos.

181

De Mello, George. 1992. “Duplicación del pronombre relativo de objeto directo en el
español hablado culto de once ciudades”. Lexis XVI (1), 23-52.

Franco, Jon. 1993. On Object Agreement in Spanish. Doctoral dissertation, University of
Southern California.

García, Erika. 1975. The Role of Theory in Linguistic Analysis: the Spanish Pronouns
System. Amsterdam: North-Holland.

Givón, Talmy. 1976. “Topic, Pronoun and Grammatical Agreement”. In Charles N. Li
(ed) Subject and Topic. New York: Academics Press, 149-188.

Hawkins, John. 2004. Efficiency and Complexity in Grammars. Oxford: Oxford
University Press.

Joseph, Brian. 1983. “Relativization in Modern Greek. Another Look at the Accessibility
Hierarchy Constraints”. Lingua 60. Leiden: North Holland Publishing, 1-24.

Labov, William. 1972. Sociolinguistic Patterns. Philadelphia: University of
Pennsylvania Press.

McCloskey, James. 2006. “Resumption”. In The Blackwell Companion to Syntax, Martin
Everaert and Henk van Riemsdijk, (eds.), Blackwell Publishing, 94–117.

Mak, Willem, Wietske Vonk, and Herbert Schriefers. 2006. “Processing Relative Clauses
in Dutch: When Rocks Crush Hikers”. Journal of Memory & Language 54, 466-
490.

Molnár, Valéria. 2001. “Contrast from a Contrastive Perspective”. In I. Krujff-
Korbayová and M. Steedman (eds.) Information Structure, Discourse Structure
and Discourse Semantics, ESSLLI2001 Workshop Proceedings, 99-114. Helsinki:
The University of Helsinki Press.

Paredes Silva, Vera Lúcia. 1993. “Subject Omission and Functional Compensation:
Evidence from Written Brazilian Portuguese”. Language Variation and Change 5,
35-49.

Prince, Ellen. 1990. “Syntax and Discourse: A look at Resumptive Pronouns”. In Hall,
K., et al (eds.), Proceedings of the Sixteenth Berkeley Linguistics Society Annual
Meeting, 482-97.

Ramat, Paolo and Anna Giacalone de Ramat. 1995. Las lenguas indoeuropeas. Madrid:
Cátedra.

182

Samper Padilla, José Antonio, Clara Hernández Cabrera, and Otilia Pérez Gil. 2004. Las
construcciones de retoma en las cláusulas relativas en el español de Las Palmas
de Gran Canaria. Manuscript.

Schwenter, Scott. 2002. “The notion of Contrast in the Study of Spanish Subject Personal
Pronouns”. Workshop on Subject Pronoun Expression in Spanish, CUNY
Graduate Center.

Silva-Corvalán, Carmen. 1994. Language Contact and Change. Spanish in Los Angeles.
New York: Oxford University Press.

Silva-Corvalán, Carmen. 1996. “Resumptive Pronouns: a Discourse Explanation”. In
Claudia Parodi, Carlos Quicoli, Mario Saltarelli, and María Luisa Zubizarreta.
(eds.), Aspects of Romance Linguistics. Washington D.C.: Georgetown University
Press, 383-395.

Silva-Corvalán, Carmen. 2001. Sociolingüística y pragmática del español. Washington
D.C.: Georgetown University Press.

Silva-Corvalán, Carmen. 2003. “Otra mirada a la expresión del sujeto como variable
sintáctica”. In Francisco Moreno Fernández, José Antonio Samper Padilla, María
Vaquero, María Luz Gutiérrez Araus, César Hernández Alonso, and Francisco
Gimeno Menéndez (eds.), Lengua, variación y contexto: Estudios dedicados a
Humberto López-Morales. Madrid: Arco Libros, 849-860.

Solana, Zulema. 1996. “Relativas explicativas en la gramática infantil”. Signo y Seña 5,
357-377.

Suñer, Margarita. 1998. “Resumptive Restrictive Relatives: A Crosslinguistic
Perspective”. Language 74, 335-364.

Suñer, Margarita. 1988. “The Role of Agreement in Clitic-Doubled Constructions”. In
Natural Language & Linguistic Theory 6 (3), 391-434.

Tarallo, Fernando. 1986. “Functional and Structural Properties in a Variable Syntax”. In
David Sankoff (ed.), Diversity and Diachrony. Ottawa: John Benjamins, 239-248.

Trujillo, Ramón. 1990. “Sobre la supuesta despronominalización del relativo.” In
Estudios de Lingüística 6, 23-46.

Umbach, C. 2004. “On the Notion of Contrast in Information Structure and
Discourse Structure”. Journal of Semantics 21 (2), 155-175.

183

Umbach, C. 2001. “Contrast and Contrastive Topic”. In I. Krujiff-Korbayová and M.
Steedman (Eds.) Information Structure, Discourse Structure and Discourse
Semantics.

Utts, Jessica. 2005. Seeing Through Statistics. Kentucky: Cengage Learning.

Valentí Fiol, Eduardo. 1999. Gramática de la lengua latina. Morfología y nociones de
sintaxis. Barcelona: Bosch, Casa Editorial S.A.
184

Appendix
Table 52. Languages Combining [-Case] Gaps with [+Case] Resumptive Pronouns

SU DO IO/OBL GEN
Aoban gap pro pro pro
Arabic gap pro pro pro
Gilbertese gap pro pro pro
Kera gap pro pro pro
Chinese (Peking) gap gap/pro pro pro
Genoese gap gap/pro pro pro
Hebrew gap gap/pro pro pro
Persian gap gap/pro pro pro
Tongan gap gap/pro pro pro
Fulani gap gap pro pro
Greek gap gap pro pro
Welsh gap gap pro pro
Zurich German gap gap pro pro
Toba Batak gap * pro pro
Hausa gap gap gap/pro pro
Shona gap gap gap/pro pro
Minang-Kabau gap * */pro pro
Korean gap gap gap pro
Roviana gap gap gap pro
Turkish gap gap gap pro
Yoruba gap gap 0 pro
Malay gap gap gap pro
Javanese gap * * pro
Japanese gap gap gap pro

Gaps = 24[100%] 17[65%] 6[25%] 1[4%]
Pros = 0 [0%] 9[35%] 18[75%] 24[96%]

Key: gap = [-Case] strategy
pro = copy pronoun retained (as a subinstance of [-Case])
* = obligatory passivization to a higher position prior to relativization
0 = position does not exist as such
RP = relative pronoun plus gap (as a subinstance of [+Case])

Abstract (if available)

Abstract This dissertation is a variationist account of two non-standard relative clause (RC) structures in Spanish: resumptive pronouns (RPs) and prepositional-phrase (PP) chopping. Previous typological studies considered RP explanations based on difficulty of processing (Hawkins, 1994), while Spanish-specific quantitative studies proposed a number of factors regarding the relationship between the RC and its antecedent (Silva-Corvalán, 1996

Linked assets

University of Southern California Dissertations and Theses

Asset Metadata

Creator Cerrón-Palomino López, Alvaro (author)

Core Title Adding and subtracting alternation: resumption and prepositional phrase chopping in Spanish relative clauses

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Linguistics (Hispanic Linguistics)

Publication Date 02/02/2010

Defense Date 11/17/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag accessibility hierarchy,animacy,contrast,definiteness,domain minimization,multiple causation,OAI-PMH Harvest,prepositional phrase chopping,processing,relative clauses,resumptive pronouns,sociolinguistics,Spanish,syntactic variation

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Silva-Corvalán, Carmen (committee chair), Andersen, Elaine (committee member), Saltarelli, Mario (committee member)

Creator Email acerropa@asu.edu,cerronpa@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2830

Unique identifier UC1452988

Identifier etd-CerronPalomino-3458 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-294368 (legacy record id),usctheses-m2830 (legacy record id)

Legacy Identifier etd-CerronPalomino-3458.pdf

Dmrecord 294368

Document Type Dissertation

Rights Cerrón-Palomino López, Alvaro

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu