Who is Doing What: The
Recoverableness of the Subject Referent in Clauses with Lexically Null Subjects in Spanish
Shirley
A. Wright
University of Texas at
Arlington
Government and Binding theory has postulated pro as the non-overt pronominal
in languages in which the subject slot may be empty at the surface. Spanish is a pro-drop language which poses
the questions of why speakers choose to employ a lexically null subject even when 1) there
is a need to distinguish among several possible candidates for the semantic content of the
subject, 2) there is not sufficient agreement in the verbal affixes to recover the
subject, and 3) the lexically null subject is not the main focus of a discourse (and
therefore is not immediately suppliable by the reader).
This paper demonstrates how discourse analysis techniques used on Borges’s
short story “Abenjacán el Bojarí, Muerto en su Laberinto” can answer the
previous questions. Spanish speakers may
choose to delete surface subject pronouns throughout a discourse, but, guided by the
pragmatics of efficiency, they maintain subject continuity from independent clauses to the
following clauses in cases where the sentences are closest to the central theme of the
discourse.
1. Introduction.
In pro-drop languages speakers/writers have the choice of using an overt
pronominal or of having a lexically null subject. If language producers choose to use a
lexically null subject, then the issue of reconstructing the semantic content, i.e. the
referent, for the dropped subject pronoun becomes pertinent. In this paper I use a
discourse model of Jorge Luis Borges’s “Abenjacán el Bojarí, Muerto en su
Laberinto” to add to the body of knowledge previously established on this phenomenon
(Iatridou and Embick 1997, Levy 1999, Steele 1989, and many others). Specifically, I hypothesize some boundaries for
the recoverableness of the superficially deleted subject in Spanish.
An examination of the data shows that language producers choose to omit a subject
pronoun even when counter indicated by traditional explanations. While much work has been done to help describe the
use of subject pronominals versus pro, there remain some questions as to why
speakers of Spanish delete surface pronominals under various conditions.
First, conventional accounts specify that subject pronouns in pro-drop languages are overt when they are
emphasized (Haegeman 1994) or they are used contrastively or distinctively (Rigau 1985). While generally not stated outright, the corollary
to this position is that pro-drop
occurs when there is no emphasis, contrast, or distinction.
In other words, only when there is subject continuity from one statement to the
next (and therefore no need to emphasize, contrast, or distinguish among subjects), should
there be pro-drop. However, pro-drop
does occur even when there are multiple possibilities for the semantic content of the
subject. As Rigau (1985) points out, in some
cases the only way to recover the subject is to examine “the discourse context in
which the sentence appears.” Indeed, in
this paper I seek to answer the question of why speakers omit surface subjects when there
is no subject continuity from one clause to the next by utilizing discourse analysis
techniques.
Second, while Rizzi (1986) hypothesizes that pro-drop
is possible when the referent can be determined because the agreement system is rich, and
Huang (1984) believes that it is possible at the two extremes (either a rich or a
non-existent agreement system), the question remains, “How rich is rich?” As
Schroten (1986) points out, the inflectional system in Spanish is not so rich as to
provide for complete explicitness.1 Even
though there is ambiguity in the agreement system, especially as in the third person forms
in Spanish, pro-drop occurs such that the subject referent is not precisely
recoverable from the verbal affix. Why do
language producers omit subjects when the verb cannot guide the audience toward the
correct interpretation of the subject as any of a number of third person entities?
Third, Dimitriadis (1996) takes a different tack from the above and states that
language producers choose pro-drop with maximally prominent entities. In the case of “Abenjacán el Bojarí, Muerto
en su Laberinto,” the phenomenon occurs with all the characters (no matter how minor
a part they play in the plot) and many inanimate entities. Saying that pro-drop
occurs in a particular sentence because the entity is maximally prominent in that sentence
(or that entity is maximally prominent, for it undergoes pro-drop) only leads to
every entity’s being considered equally maximally prominent. While it is true that focus within a narrative can
(and does) shift so that one character is more the topic at one point and another
character more the topic at another, positing lightning quick changes of focus simply to
account for every subject’s ability to become lexically null seems a bit facile. This leaves the question: why do speakers delete
subject pronouns regardless of the prominence of the subject?
In order to answer the question of why speakers and writers choose to delete
surface subjects in cases where there are multiple subjects from which possibly to choose,
the agreement system is not sufficient to specify one and only one possible referent, and
the subject dropped does not qualify as prominent, I have chosen to use discourse
analysis, specifically Longacre and Levinsohn (1978) and Longacre (1989). A careful examination of the structure of the
narrative along with an investigation tracking the referents for pro show that
while pro-drop transpires at all levels of the narrative, the subject referent is
most easily recoverable along the main plot line. Here
the pattern is frequently that the referent for pro is in the preceding independent
clause. In other parts of the story, Borges uses pro without regard for how it
might strain the memory (or logic) limitations of the reader to reconstruct the subject. In other words, pro-drop is a constant
phenomenon throughout the text, but in the sentences that are most important for the
reader to follow, Borges makes sure that the semantic content of the subject is close
enough that the reader can remember it.
Section 2 of this paper takes a preliminary view of the data. Section 3 shows that the referent for pro
is much more likely to reside in independent clauses than dependent ones. In section 4, I show that the recoverableness of
the referent for pro is governed by the salience of the section of the narrative in
which the pro-drop specimens appear. Conclusions follow.
2. The Data.2 Borges’s story, “Abenjacán el Bojarí,
Muerto en su Laberinto,” readily divides itself into two sections, the first being
the frame of the two characters (Unwin and Dunraven) who explore the labyrinth where the
mystery took place many years before, the second being the mystery itself of who killed
Abenjacán el Bojarí. In other words, the first section is the portion of the story told
by the narrator while the second is the dialogue spoken by the two characters as one of
them relates the mystery and the other of them solves it.
Within the first section of the story, there are 24 examples of pro-drop. Of
these, only 11 (approximately 46%) show no
change in subject from the closest preceding subject. For example,
(1) Unwind le respondió,
como si prod pensara
en voz alta...
Unwind to him responded, as
if prod were thinking out loud...
Unwin answered him, as if he were thinking out loud...
Far more common is the
phenomenon that the reader must be able to coindex pro with a subject several
clauses away:
(2)
Dunravena ...pensó que
la solución del misterio siempre es inferior al misterio.
Dunravena...thought that the solution of the mystery always is inferior to the mystery.
Dunraven...thought that the solution of the mystery is always inferior to the
mystery.
El misterio participa de lo sobrenatural...; la solución del juego de manos.
The mystery participates in the
supernatural...; the solution of the slight of hand.
Mystery belongs to the supernatural...; its solution to a slight of hand.
proa Dijo, para aplazar
lo inevitable:
proa said,
to postpone the inevitable:
He said, to postpone the inevitable:,
or even coindex pro
with two separate subjects from separate clauses to form a plural subject:
(3)
Dunravena dijo que.... Unwind recordó.... Hacia la
medianoche proa+d descubrieron...
Dunravena said that....Unwind remembered....Towards midnight proa+d
discovered...
Dunraven said that.... Unwin remembered....Towards midnight they discovered....
The second section of the story shows even less consistency of subject from one
clause to the next. Of 108 cases of pro-drop,
only 36 (or approximately 33%) have subject referents held over from the closest preceding
subject. Once again, the norm is for the reader to have to remember two or more clauses
back in order to discern the subject.
(4)
Luego Zaidg deshizo
las tres caras con una
piedra.
Then Zaidg destroyed
the three faces with a rock.
Then Zaid destroyed the three faces with a rock.
prog Tuvo que obrar así;
un solo muerto
con la
cara deshecha hubiera
prog had to work thus; a single dead man with the face
destroyed would have
He had to do that; one dead man with his face destroyed would have
sugerido un problema de identidad, la fiera,
el negro, y el rey
formulaban una serie
suggested a problem of identity,
the beast, the black man and the king formed a
series
posed an identity problem; the beast, the black man and the king formed a series
y dados los
dos términos iniciales, todos
postularían
el último.
and given the two terms
initial, everyone would
postulate the last.
and given the first two terms, everyone would guess the last one.
No es raro que
lo dominara el temor cuando prog
habló con Allaby;
not is strange that him dominated the fear when prog spoke with
Allaby;
It isn’t strange that fear consumed him when he spoke with Allaby;
Clearly, one cannot adequately describe pro-drop as taking place in Spanish
simply when there is holdover of a subject from clause to the next. Spanish requires a
more sophisticated model.
3. Pro-drop
examined under the Longacre Levinsohn (1978) discourse analysis model. While
Longacre and Levinsohn’s (1978) methodology for charting texts is intended primarily
to help analyze lesser known languages, the basic tenets set forth are also appropriate to
help analyze a not completely understood facet (for instance, pro-drop) of a well
known language (Spanish). Splitting each
sentence of a discourse into its varied parts (introducers, dependent clauses, independent
clauses, post-verbal dependent clauses), identifying the syntactic elements of each
(subject, predicate, etc.), tracking the movement in location of the action, all these aspects of the Longacre and Levinsohn model
are applicable to practically any study.
In the case of Borges’s story, a narrative discourse, charting the story shows
a marked improvement in being able to predict when pro-drop will occur. While in
the previous section we saw overall that pro-drop occurs 37% of the time that the
subject carries over from one clause to the next, using the Longacre and Levinsohn
methodology we see that pro-drop occurs 76 out 144 times (approximately 53%) when
there is subject carryover from the nearest preceding independent clause. In other words, in order to determine the referent
for a dropped pro, it is better to disregard any preceding dependent clauses and
look for the last independent clause.
(5)
Unwind creía
que no le había
interesado la historia de la muerte del Bojarí
Unwind believed that no him had
interested the story of the death of
Bojarí
Unwin believed that the story of Bojari’s death had not interested him
pero prod se
despertó con la convicción
de haberla descifrado.
but prod
himself woke with
the conviction of having it solved.
but he woke with the conviction of having solved it.
In example 5 it is clear that ignoring the that clause simplifies the
sentence structure such that the referent for pro is easily seen. Analysis of all the cases of pro-drop shows
that for the purpose of finding the referent of pro, it is better to disregard
dependent noun and adverbial clauses. Or, to
approach the matter from another point of view, the subject of a dependent clause rarely
(only once in this Borges story) carries over to the next clause as a case of pro-drop.
However improved our analysis by excluding dependent clauses, the problem remains
that there is a very large percentage of pro-drops whose referents are not
immediately obvious.
4. Pro-Drop
and Longacre’s Etic Bands of Salience (1989).
Longacre (1989) proposes that some parts of a narrative text are more salient than
others. The storyline (or eventline) is more
essential than the setting or evaluations (commentary by the author), for instance. By breaking Borges’s story down into its
bands of salience, we see another part of the puzzle as to when pro-drop takes
place. Because of the duality of the action
in “Abenjacán el Bojarí,” I believe it is reasonable to assume two storylines
(one for the two men who explore the labyrinth and one for Abenjacán el Bojarí, who is
murdered), two settings, two sets of evaluations, etc.
The most salient of the etic bands is the storyline, marked by the preterite in
this story. There are 16 specimens of pro-drop
in the first storyline, and of them 13 (81%) have pro coindexed with the subject in
the last preceding independent clause. In
the second storyline, the numbers are 32 of the 44 cases, or approximately 73 % (see
tables 1 and 2). In other words, where it is more important for the reader to be able to
follow the plot, the referent for pro is more easily accessed (see table 3 for
overall rates).
In addition to the storyline, backgrounded actions and activity in the second story
have a high rate of pro’s being coindexed with the subject of the preceding
independent clause. (In the first story there
is only one example of backgrounded action and none of backgrounded activity.) Whereas in the case of backgrounded action in the
second story the low number of occurrences of pro (only 2) might result in the
rate’s being coincidentally high (100%), backgrounded activity’s high rate (10
out of 11, or 91%) is due to its salience. Backgrounded
activity, marked by the imperfect and present historical at peak, includes habitual
actions and states without which important plot points would not take place. For instance, without the slave’s repeated
trips to the docks to talk with crews from his homeland, word would never have reached the
real Abenjacán of where his imposter was currently living.
While backgrounded action is significant to the overall advance of the plot, it is
surprising that its rate of easy recoverableness of referents for pro is higher
than the rates occurring in the storyline.
Flashbacks in this story (marked by the pluperfect) often serve to advance the plot
line; indeed, in a murder mystery much of the eventline is only revealed to the reader by
way of the detective’s (whether official, or in this case, unofficial, in the form of
Unwin’s character) reconstructing how the murder transpires. Thus, it is not surprising to see a percentage of pros
(57% and 39%) coindexed with the subject in the previous independent clause that falls
somewhere between the percentage for the storylines and most of the other etic bands.
For other etic bands of salience, pro-drop quite frequently takes place with
pro’s being coindexed with a subject two or more independent clauses back.
Here we see the corollary of findings for the storyline, with the opposite polarization. That is, in the areas of the narrative that are
less important for the reader to follow, pro-drop occurs with less regard for how
difficult it might be for the reader to reconstruct the referent. A particularly striking example of this is the
cohesive band,3 of which there are 19 specimens overall and for which the
percentage of pros’ being coindexed with the subject in the last independent clause
is 0%.
Etic Bands of Salience |
Number of Pro-drops |
Number of Pros
Coindexed with the Subject in the Last Preceding Independent Clause |
Percentage |
Storyline |
16 |
13 |
81% |
Backgrounded Actions
(events) |
1 |
0 |
0% |
Flashback |
7 |
4 |
57% |
Setting |
1 |
0 |
0% |
Irrealis |
1 |
0 |
0% |
Evaluations |
4 |
2 |
50% |
Cohesive |
18 |
0 |
0% |
Table 1. Pro-Drop for the First Story, that of
Unwin and Dunraven
Etic Bands of Salience |
Number of Pro-drops |
Number of Pros
Coindexed with the Subject in the Last Preceding Independent Clause |
Percentage |
Storyline |
44 |
32 |
73% |
Backgrounded Actions
(events) |
2 |
2 |
100% |
Backgrounded Activity
(durative) |
11 |
10 |
91% |
Flashback |
23 |
9 |
39% |
Irrealis |
6 |
2 |
33% |
Evaluations |
9 |
2 |
22% |
Cohesive |
1 |
0 |
0% |
Table 2. Pro-Drop for the Second Story, that of
Abenjacán el Bojarí
Etic Bands of Salience |
Number of Pro-drops |
Number of Pros
Coindexed with the Subject in the Last Preceding Independent Clause |
Percentage |
Storyline |
60 |
45 |
75% |
Backgrounded Actions
(events) |
3 |
2 |
67% |
Backgrounded Activity
(durative) |
11 |
10 |
91% |
Flashback |
30 |
13 |
43% |
Setting |
1 |
0 |
0% |
Irrealis |
7 |
2 |
29% |
Evaluations |
13 |
4 |
31% |
Cohesive |
19 |
0 |
0% |
Table 3. Pro-Drop for the Entire Discourse
5. Conclusions. In section 3, it became clear that
the semantic content of a lexically null subject is far more likely to refer to an entity
in a previous independent clause than in a previous dependent clause. This is, perhaps, not surprising considering that
dependent clauses carry less of the burden of the communication of information. Through their very nature dependent clauses may be
deleted, and any data in them lost. Therefore,
from a purely pragmatic point of view, it is more sensible to encode information that will
be needed in later utterances in independent clauses.
Section 4 showed that the recoverableness of the referent for pro depends on
in which etic band of salience the occurrence of pro-drop appears. Overall in the storyline the referent for pro
surfaces in the last independent clause 75% of the time, and in backgrounded durative
activity the percentage is 91%. While it is
reasonable that both these etic bands have high percentages, it is unexpected that
backgrounded durative activity be the higher of the two.
I would like to see more work done in this area to determine whether this result is
peculiar to this particular narrative. At the
opposite end of the spectrum for the etic bands, the cohesive band has a zero percentage
rate overall. In other words, the clauses
that simply serve to orient the reader toward the organization of the text do not have
easily accessed antecedents.
The examination of this Spanish text shows that the semantic content of the subject
for declarations exhibiting pro-drop is reconstructible by inspecting the
independent clauses in the portion of the discourse that are closest to the main point. The overall pattern, then, for the appearance of
the referent for pro is that it shows up in clauses that have the greatest import
in terms of communicating information. That
is, pro-drop occurs throughout the text, but where it is most important for the
reader to be able to remember and access the referent, the referent appears close by. If language producers deem a particular utterance
to be less vital to the global scheme of the text, they will simply supply a verb without
a surface subject and let their audience catch the meaning as best they can.
In fact, the choice of using a lexically null subject in languages that allow that
option may well boil down to a type of efficiency. It
is quicker and easier to omit surface subjects, and so language producers do so at will. However, to communicate effectively, language
producers must also take into account their audience’s ability to follow the
discourse. To ensure absolute and complete understanding, speakers/writers could choose to
always use subject pronouns. A balance
between the two extremes (of never or of always using subject pronouns) can be formed by
making sure that the referents for deleted subject pronouns are readily accessible if it
is crucial that the utterance be understood. Thus,
producers employ pro-drop regularly. If
they feel that a sentence is only on the periphery of their main point, they might not
expend the effort to be sure that there is subject continuity along with the pro-drop. Correspondingly, when speakers/writers express
their central idea, they provide subject continuity along with the pro-drop to be
sure that their listeners/readers can follow the discourse.
As with any study, this one is not absolute and definitive; questions remain. I see two main areas of investigation waiting: 1)
within Spanish and 2) universally. For the
first area, analyses with larger corpuses of data in Spanish would serve to better rank
the salience level of each etic band and possibly explain the unanticipatedly high rates
for coindexation of pro with immediately preceding subjects in backgrounded action
and activity. Also, although the percentages
of coindexation for etic bands such as the storyline are high, there remains some
detritus, for instance, the last 25% for which coindexation takes place at a distance. In addition, I would like to see analyses on types
of discourse other than narrative, i.e. procedural, behavioral, etc. In the second area, that of universality, I would
like to see whether studies in other pro-drop languages show a balance of lexically
null subjects and their referents based on some type of efficiency principle. Do other languages habitually pro-drop
across all etic bands? Do they provide ready
attainability of referents only in the most salient bands?
Are the referents concentrated in the more information dense syntactic structures,
such as independent clauses?
In this study I have analyzed the recoverableness of the referent for dropped
subjects and advanced a hypothesis for the reason behind the same. I hope others will continue the study, perhaps
finding better and more complete answers to the questions that remain.
Notes
1 Schroten
mentions the ambiguity of such forms as trabajaba, which could have four different
subjects: yo, él, ella, and usted.
2 For the
purposes of this study, it is unnecessary to consider expletive pro, cases in which
the verb cannot take subjects. The data here,
therefore, do not count those examples.
3 It is
noteworthy, perhaps, that many of these examples, such as, “pro Empezaré
por la mayor mentira...” “[I] will start will the biggest lie...” have
verbal affixes which are not as ambiguous as the third person forms.
References
Borges, Jorge Luis. 1994.
Abenjacán el Bojarí, Muerto en su Laberinto. Obras Completas 1923-1949, vol. 1, 600-606.
Buenos Aires: Emecé Editores.
Dimitriadis, Alexis.1997. When pro-drop languages don’t:
Overt Pronominal subjects and pragmatic inference. Chicago Linguistic Society 32.
Haegeman, Liliane. 1994. Introduction to Government and
Binding Theory. Oxford, UK: Blackwell.
Huang, J. 1984. On the
distribution and reference of empty pronouns. Linguistic Inquiry 15.531-74.
Iatridou, Sabine and David Embick. 1997. Apropos pro. Language 73.1.
58-78.
Levy, Ruth. 1999. De Jorge Luis
Borges a Walt Whitman: Por qué Borges sí escribe el implícito “yo.”
Sincronía Spring 1999.
Longacre, Robert E. and Stephen H. Levinsohn. 1978. Field analysis of
discourse. Current Trends in Textlinguistics. New York: Walter de Gruyter.
Longacre, Robert E. 1989. Two
hypotheses regarding text generation and analysis. Discourse Processes 12.413-60.
Rigau, Gemma.1986. Some remarks
on the nature of strong pronouns in null-subject languages.
Generative Studies in Spanish Syntax, ed. By Ivonne Bordelois, Heles Contreras, and
Karen Zagona, 143-163. Dordrecht: Foris.
Rizzi, L.1986. Null objects in Italian and the theory of pro. Linguistic
Inquiry 17.501-58.
Schroten, Jan.1986. Dos
aproximaciones a la sintaxis de las oraciones finitas sin sujeto léxico. Aproximaciones a la sintaxis del español:
Estudios sintácticos del español y el progreso de la teoría lingüística, ed. by
Magdalena García Pinto and Mario A. Rojas, 246-301. Barcelona: Puvill Libros.
Steele, Susan. 1989. Subject Values. Language 65.3. 537-578.
Return to Sincronía General Index
Return to Sincronía Winter 2000