The resolution of ambiguous anaphora in english, spanish, and portuguese

This study aims to determine how (a) ambiguity resolution diﬀ ers in English, Spanish, and Portuguese; (b) L2 learners are inﬂ uenced by their L1 regarding ambiguity resolution; and (c) the order of the clauses, (d) the salience of the anaphor, or (e) the choice of verbs may aﬀ ect this process. A total of 181 people answered an online survey and selected the antecedent of the anaphor for 16 ambiguous sentences. The Position of Antecedent Strategy was tested for the three languages and the conclusion was that it applies to European and Brazilian Portuguese, but not for Spanish or English. While Spanish and English native speakers tend to interpret the subject of the subordinate clause as corresponding to the subject of the main clause, learners are inﬂ uenced by their L1 and seem uncertain especially on how to resolve pronominal cataphora. Finally, we


2022
Amanda Maraschin Bruscato, Jorge Baptista have concluded that semantics has a great infl uence on the resolution of ambiguous anaphora in these languages, and we suggest that it should be taken in consideration during the design of a study like this one.
Keywords: ambiguity; anaphora resolution; position of antecedent strategy.

Introduction
Ambiguity is a classic problem in anaphora resolution, and a considerable amount of literature has been published on this subject  Valenzuela et al., 2011). These studies tried to explain how L1 and L2 speakers solve ambiguity in diff erent languages, but to date there has been little agreement on what theory would better describe this phenomenon. This paper expands our previous research (Bruscato & Baptista, 2022a, 2022b and compares ambiguity resolution in the three European languages most spoken in the world (Eberhard et al., 2020): English, Spanish, and Portuguese. The objectives of this study are to determine: (a) how ambiguity resolution strategies diff er in English, Spanish, and Portuguese; (b) how L2 learners are infl uenced by their L1 regarding ambiguity resolution; and whether (c) the order of the main and subordinate clauses, (d) the salience of the anaphor, or (e) the lexical choice of verbs may aff ect this process.
Data from Portuguese and Brazilian speakers who study English or Spanish in a Brazilian or a Portuguese university are compared to data from native speakers of these languages 3 . They volunteered to answer an online questionnaire that tested their profi ciency level in the L2 and asked them to identify the subject of ambiguous sentences in their L1 and L2.
The paper has been organized in the following way: the second section provides a brief overview of the theoretical framework; the third section is concerned with the methodology used for the study; the fourth section presents and discusses the fi ndings of the research; fi nally, the fi fth section reports the conclusions.

Theoretical Framework
According to King and Lewis (2018, p. 1), in simple terms anaphora could be defined as "the phenomenon whereby the interpretation of an occurrence [the anaphor] of one expression depends on the interpretation of an occurrence of another [the antecedent]". The anaphor and its antecedent can be in the same or diff erent sentences, and the anaphor can appear before or after its antecedent in the discourse. If the anaphor comes after its antecedent, there is an example of forward 3. We would like to thank all the people who contributed to this study by answering or disseminating the survey, as well as the institution where it was conducted. We would also like to thank Professor Elena Valenzuela, for making available material from her research, in which we were partly inspired for this work. Jorge Baptista's work is supported by national funds through FCT, Fundação para a Ciência e a Tecnologia, under project UIDB/50021/2020.

2022
Amanda Maraschin Bruscato, Jorge Baptista anaphora or just anaphora; see example (1) 4 . Otherwise, there is an example of backwards anaphora or cataphora (2). In this study, we will use the term anaphora to talk about both, anaphora and cataphora, when it is not relevant to diff erentiate them.
(1) Johni ignored Mary while hei was on the phone.
(2) While hei was on the phone, Johni ignored Mary.
Examples (1) and (2) have pronominal anaphora since the anaphor is a pronoun. In Spanish or Portuguese, however, there is no need to use a pronoun in the subject position of the subordinate clause if the antecedent is also in the subject position of the main clause. These languages are called pro-drop languages, also known as null subject languages (Chomsky, 1981;Rizzi, 1982). They allow null or zero anaphora when the antecedent is grammatically or pragmatically inferable, as seen in (3). However, if the subject of the subordinate clause is diff erent from the one of the main clause, Spanish and Portuguese speakers use a personal pronoun, as seen in (4). The examples so far are not ambiguous, and it is not diffi cult to identify the anaphor's antecedent. But if we change the name Mary to Peter, as in (5), would people think that the anaphor's antecedent is the subject or the complement of the main clause? Or maybe another referent, altogether -someone not mentioned at all in the sentence? This is what the present study aims to fi nd out. In other words, the focus of this research will be on intrasentential, coreferential, pronominal and zero anaphora with nominal human antecedents.
Traditionally, it has been argued that there is a preference in English to correlate the pronoun in a subject position to the subject of the main clause (Chambers & Smyth, 1998;Crawley et al., 1990), as seen in (6), unless it is stressed. In this case, stressed pronouns in subject position would impose a diff erent (disjoint) interpretation from the subject of the main clause (Luján, 1986;Smyth, 1994), as seen in (7). This behavior would be analogous to what speakers of null subject languages do regarding the distinction between using zero anaphora (null pronoun) and pronominal anaphora (unstressed overt pronouns).  (4)). However, it is not clear if this theory would apply to the diff erences between unstressed and stressed pronouns in a non-null subject language like English.
In this paper, we will test PAS with English and Spanish learners and native speakers, as well as with Brazilian (BP) and European (EP) Portuguese native speakers. Many authors (Duarte, 2000;Holmberg et al., 2009;Kempchinsky, 1984) claim that Brazilian Portuguese, unlike European Portuguese, is one of the "partial null-subject languages, that is, languages which allow null subjects but under more restricted conditions than consistent null-subject languages" (Holmberg et al., 2009, p. 1). Thus, ambiguity resolution might diff er in these varieties.
Apart from the salience of the anaphor, the order of the clauses may also aff ect anaphora resolution. The hypothesis of the Active Search Mechanism (ASM), from Kazanina et al. (2007), suggests that, in cases of pronominal cataphora in English (as in (8)), speakers would choose the fi rst possible antecedent that appears after the anaphor. Usually, the fi rst possible antecedent is the subject of the main clause. However, few studies compare anaphora resolution in multiple languages, and Portuguese is not usually included in them. They also do not specify semantic criteria for the verb selection in the exercises. Our study, on the other hand, not only specifi es them, but investigates how diff erent combinations of verbs can aff ect anaphora resolution. Valenzuela et al. (2011), the resolution of zero and pronominal forward and backwards anaphora in Spanish and Portuguese will be compared to non-emphatic and emphatic pronominal forward and backwards anaphora in English. After our pilot-studies (Bruscato & Baptista, 2022a, 2022b), we have expanded our research to investigate how L1 and L2 speakers solve ambiguity in these languages.

Method
As it was stated in the Introduction, the purpose of this study is to determine: (a) how ambiguity resolution strategies diff er in English, Spanish, and Portuguese; (b) how L2 learners are infl uenced by their L1 regarding ambiguity resolution; and whether (c) the order of the main and subordinate clauses, (d) the salience of the anaphor, or (e) the lexical choice of verbs may aff ect this process.
A quantitative approach was employed to determine the factors that aff ect ambiguity resolution. A written questionnaire was designed based on Valenzuela et al. (2011) and, after approval from the ethics committee was obtained, it was sent online to undergraduate students from the language department at the Federal University of Rio Grande do Sul (in March 2020) and at University of Algarve (in October 2020). The questionnaire, available in Google forms, was also shared on the researchers' social media to reach more English and Spanish native speakers. Everyone's participation was anonymous and voluntary, and respondents expressed their agreement to take part in the research.
A total of 181 people from 18 to 54 years old (median of 20 years) have answered the survey. Almost 70% of the sample was female, and more than 90% of them studied at a university. There were 78 Portuguese and 73 Brazilian undergraduate students who were learning English (61 and 56, respectively) or Spanish (17 and 17) as a foreign language, as well as 20 English and 10 Spanish native speakers.
In the fi rst moment, Portuguese native speakers answered 20 questions from Cambridge or Cervantes' reading profi ciency tests, distributed between levels A2, B1, B2 and C1. Then, participants were asked to select the subject of 20 sentences in each language, of which 16 were ambiguous sentences. Example (9), below, illustrates the task: Half of the sentences had anaphora, and the other half had cataphora. Half of them had a salient anaphor (unstressed pronoun in Portuguese or Spanish; stressed pronoun in English), and the other half had a non-salient anaphor (ellipsis in Portuguese or Spanish; unstressed pronoun in English). Participants were instructed to consider pronouns written in capital letters as stressed 5 . Sentence (9), above, showed an example of stressed pronominal anaphora, while sentence (10), below, shows an example of unstressed pronominal cataphora in English. The examples above are part of the set of sentences designed to have a "neutral" combination of verbs, that is, where little, if any, semantic interference can be detected between the choice of the verbs and the reference resolution strategy adopted. Half of the sentences were designed in this way. The other half was designed to have "tendentious" pairs of verbs, that is, verbs that in that combination preferably point back either to the subject or to the complement of the verb in the main clause. Below, sentence (11) was designed to elicit a preference for resolving the anaphor as referring to the subject of the main clause; and in sentence (12) to the complement of the main clause. (11) Julia disrespected Paula when she betrayed her. (12) When she fi ned her, Laura bribed Lucy.
In the appendix of this paper, all sentences in the three languages are presented. In the next section, results are presented and discussed. Statistical analysis was performed using SPSS software (version 26). The small size of the dataset means that it is not possible to generalize conclusions. However, it contributes to test the syntactic and semantic hypothesis in the languages. 5. Although we only had written data, we wanted to know if learners would consider the opposition between stressed and unstressed pronouns in English as the opposition between unstressed pronouns and ellipsis in Portuguese. However, we recognise that the lack of spoken data is a limitation of this study.

Profi ciency test
As it was said before, Portuguese speakers answered a small reading profi ciency test in their target language before performing the task of selecting the subjects of the ambiguous sentences. The results are shown in Table 1. In general, Portuguese and Brazilian students presented similar results in this initial test, though participants from Portugal and Spanish learners achieved a slightly higher score than the others, as seen in Table 1. To identify whether there is a diff erence in anaphora resolution depending on the respondents' profi ciency level in the language, we have decided to divide the sample in two groups: intermediate and advanced learners of each language.
In the next section, we will analyse how Portuguese speakers resolve ambiguous anaphora in their native language. Then, we will analyse the results in Spanish and, fi nally, in English. The tables present the percentages of answers for the following situations: -intrasentential anaphora and cataphora; -anaphor: unstressed pronoun or ellipsis in Portuguese and Spanish, stressed or unstressed pronoun in English; -antecedent: subject, complement or other; -verb combination: neutral or biased.
The results related to the Position of Antecedent Strategy (PAS) were highlighted in bold. When we present the results for the biased sentences, we will mark the expected preference with an asterisk (*). Results are given in percentage of answers: for each cell in the tables, a pair of sentences was tested for all the participants in the sample.

Portuguese sentences
The Portuguese results were collected from 78 Portuguese and 73 Brazilian undergraduate students of languages. First, we will present the percentages of answers for the neutral sentences. Then, we will compare them with the biased ones. As we can see in Table 2, results from EP and BP speakers are very similar, though in cases of zero anaphora Brazilians tend to relate the ellipsis with the subject more than Portuguese people do (90% and 75% respectively). They all seem to follow the Position of Antecedent Strategy. Nonetheless, the strategy is not as strong in cases of pronominal cataphora as it is in cases of pronominal anaphora. The alternative other was only chosen when the anaphor was a pronoun, and it was chosen much more often in cases of cataphora than of anaphora. These results indicate that pronominal cataphora is more diffi cult to resolve than the other types of anaphora. If we compare the results from Tables 2 and 3, it becomes clear that participants' answers were infl uenced by the biased sentences. When the verb combination resulted in a preference corresponding to PAS (cases of anaphora), those percentages have increased. On the other hand, when the verb combination resulted in a preference opposite to PAS (cases of cataphora), their preferential choice has changed. Participants seem to be more infl uenced by semantics than by syntax when trying to resolve ambiguous anaphora, except for Brazilians, who still seem to be infl uenced by PAS when there is an ellipsis in subject position. As we have seen in Table 2, in Table 3 the alternative other was only chosen when the anaphor was a pronoun, and especially in cases of cataphora. This time, however, this option was chosen only by Portuguese students.

Spanish sentences
While the Portuguese questionnaire was completed only by native speakers, the Spanish one was completed by 10 native speakers, 17 learners from Portugal, and 17 learners from Brazil. Again, the results for the neutral sentences will be presented before the results for the biased ones. In our previous studies (Bruscato & Baptista, 2022a, 2022b), we had the participation of only 5 Spanish native speakers (all learners of Portuguese), and respondents seemed to follow PAS, except for cases of pronominal cataphora. Now that we have twice as many participants (not learners of Portuguese), we can see that the exception was not only in cases of pronominal cataphora, but also of pronominal anaphora. It seems that, although Spanish native speakers tend to interpret the ellipsis in subject position of the subordinate clause as corresponding to the subject of the main clause more than they do when the anaphor is a pronoun, they do not tend to interpret the unstressed pronoun in subject position of the subordinate clause as corresponding to the complement of the main clause. These results corroborate the fi ndings Spanish learners tend to interpret ellipsis in the subject position of the subordinate clause as corresponding to the subject of the main clause, especially Brazilians in cases of zero anaphora. However, participants did not show a clear preference for the antecedent of the pronoun. Again, the alternative other was only chosen in cases of pronominal cataphora, and only by native speakers and Portuguese advanced learners. In fact, the two groups of EP speakers presented opposite results for pronominal cataphora, and the intermediate learners were closer to native speakers than advanced learners. It is important to say that the size of each group was very diff erent, and the intermediate one only had 4 participants. If we compare Tables 4 and 5, we can see that participants' answers were infl uenced by the verb combination of the biased sentences, especially the natives' ones. When there was a semantic preference corresponding to PAS, those percentages have increased, except for intermediate learners, whose preferential choice for zero anaphora has decreased, although it was still the subject of the main clause. When the semantic preference was opposite to PAS, native speakers were more infl uenced by semantics. Although learners were also infl uenced by semantics, their preferential choice has only changed for pronominal cataphora (except for Portuguese intermediate learners, who did not show a preference -but, again, there were only 4 participants in this group). As we have seen with the Portuguese biased sentences, Portuguese speakers still seem to be infl uenced by PAS when there is an ellipsis in subject position. Thus, their choice to correspond the ellipsis in subject position of the subordinate clauses to the subject of the main clauses decreased in cases of zero cataphora, but it was still their preference.

English sentences
Finally, the English questionnaire was completed by 20 native speakers, 61 learners from Portugal, and 56 learners from Brazil. As in the other sections, the results for the neutral sentences will be presented before the results for the biased ones. As we have reported in previous studies (Bruscato & Baptista, 2022a, 2022b, English native speakers do not seem to be infl uenced by the order of the clauses or the salience of the anaphor. They tend to interpret the anaphor as corresponding to the subject of the main clause. Learners, on the other hand, do not show a clear preferential interpretation. They seem uncertain in most of the cases, except for Brazilian advanced students, who seem to interpret the stressed pronoun in subject position of the subordinate clause as corresponding to the complement of the main clause; and Portuguese students, who seem to interpret non-emphatic pronominal cataphora as corresponding to the subject of the main clause. Finally, when we compare Tables 6 and 7, it becomes clear that participants are again infl uenced by the biased sentences. However, although learners are infl uenced by semantics regarding non-emphatic pronominal cataphora, they still seem uncertain about the best alternative for the antecedent of the pronoun. After the percentages presented in the tables were obtained from SPSS, multiple t-tests were conducted to determine what may have aff ected the choice of the antecedent in the pairs of sentences. The tests have shown that Portuguese speakers' results were infl uenced at the 0.001 signifi cance level by the salience of the anaphors in the three languages when there was no semantic opposition to the Position of Antecedent Strategy. Although Spanish and English native speakers were not infl uenced by the order of the clauses or by the salience of the anaphors, participants from all languages were infl uenced by the semantic relation between the verbs (p < 0.03). Students from diff erent Portuguese varieties and with distinct profi ciency levels did not signifi cantly diff er from one another.

Conclusions
This study expanded previous research (Bruscato & Baptista, 2022a, 2022b) and aimed to determine: (a) how ambiguity resolution strategies diff er in English, Spanish, and Portuguese; (b) how L2 learners are infl uenced by their L1 regarding ambiguity resolution; and whether (c) the order of the main and subordinate clauses, (d) the salience of the anaphor, or (e) the lexical choice of verbs may aff ect this process.
A total of 181 people answered an online survey and selected the antecedent of the anaphor for 16 ambiguous sentences. We tested the Position of Antecedent Strategy (PAS) for the three languages and compared the infl uence of semantics and syntax on the resolution of ambiguous anaphora.
We have concluded that Portuguese native speakers seem to follow the Position of Antecedent Strategy to resolve ambiguous anaphora and that Brazilian speakers tend to interpret zero anaphora as corresponding to the subject of the main clause more than Europeans do. For participants from both varieties, however, it seems that pronominal cataphora is more diffi cult to resolve than the other types of anaphora. Although most of the participants only chose the subject or the complement of the main clause as the antecedent of the anaphor, some of them chose another referent (which was not previously stated in the sentence) as the antecedent in cases of pronominal cataphora, as it had been observed before in Italian by Sorace and Filiaci (2006).

The Position of Antecedent Strategy was not confi rmed with
Spanish nor with English native speakers. Both native groups tend to interpret the subject of the subordinate clause as corresponding to the subject of the main clause, though Spanish native speakers have a stronger preference to establish that correspondence in cases of zero anaphora rather than of pronominal anaphora.
When learning a foreign language, Portuguese native speakers seem to be infl uenced by their L1 anaphora resolution strategy (PAS) regardless of their level of profi ciency in the L2. In Spanish, Brazilians again tend to interpret zero anaphora as corresponding to the subject of the main clause more than Europeans do. However, learners have shown some uncertainty to choose the antecedent of the pronoun in Spanish (especially in cases of cataphora) and in English.
Finally, the results from the "neutral" and "biased" sentences were quite diff erent. Although Portuguese speakers were still infl uenced by PAS, especially when the anaphor was an ellipsis in Portuguese or Spanish, we have concluded that the lexical choice of verbs and the meaning relation between them has a great infl uence on the task resolution of ambiguous anaphora. This aspect of the study has received little attention (if any) in previous studies. In view of these results, we posit that the semantics of the verb combination should also be considered when producing ambiguous sentences for anaphora resolution tasks.