1.Introduction

Linguistic taboo is usually studied at the lexical-semantic level, due to the idea that the prohibition falls on a problematic concept (or denotatum, Allan and Burridge 2006Allan, Keith, and Kate Burridge 2006 Forbidden Words. New York: Cambridtge University Press. , 1). Semantic studies usually focus on the (euphemistic) source concepts used for particular taboo targets or on the relations between both (Crespo Fernández 2015Crespo-Fernández, Eliecer 2015 Sex in Language. Euphemistic and Dysphemistic Metaphors in Internet Forums. London: Bloomsbury.; Pizarro Pedraza 2013Pizarro Pedraza, Andrea 2013 Tabú y eufemismo en la ciudad de Madrid. Estudio sociolingüístico-cognitivo de los conceptos sexuales. Madrid: Universidad Complutense.; Gradecak-Erdeljic and Milic 2011Gradečak-Erdeljić, Tanja, and Goran Milić 2011 “Metonymy at the Crossroads: A Case of Euphemisms and Dysphemisms.” In Defining Metonymy in Cognitive Linguistics. Towards a Consensus View, edited by Réka Benczes, Antonio Barcelona, and Francisco José Ruiz de Mendoza Ibáñez, 147–166. Amsterdam/Philadelphia: John Benjamins. ; Warren 1992Warren, Beatrice 1992 “What Euphemisms tell us about the Interpretation of Words.” Studia Linguistica 46 (2): 128–72. ; Kany 1960Kany, Charles E. 1960 American-Spanish Euphemisms. Berkeley/Los Angeles: University of California Press.; among others). Nonetheless, when the taboo concept is a verb, its expression affects the choice of an onomasiological variant but also the construal of the event and the profiling of the participants. This is particularly relevant when the verb depicts an action.

Talking about a taboo action confronts the speakers with two interesting linguistic conundrums: how to construe the action conceptually and how to profile the participants (namely, the agent) of that action. Previously, we addressed the first conundrum through the study of the conceptual onomasiological variation (Geeraerts et al. 1994Geeraerts, Dirk, Stefan Grondelaers, and Peter Bakema 1994 The Structure of Lexical Variation. Berlin: Mouton De Gruyter. ) of the concepts abortion/to abort, construed through varied semantic strategies (direct vs. indirect construal, and particular conceptual bases, e.g. ‘to kill’ or ‘to decide’). The variation of these strategies provided evidence of a link between the construal of taboo concepts and conceptual, contextual and social variables (Pizarro Pedraza 2013Pizarro Pedraza, Andrea 2013 Tabú y eufemismo en la ciudad de Madrid. Estudio sociolingüístico-cognitivo de los conceptos sexuales. Madrid: Universidad Complutense., 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. , 2018Pizarro Pedraza, Andrea 2018 “Calling Things by Their Name: Exploring the Social Meanings in the Preference for Sexual (In)Direct Construals.” In Linguistic Taboo Revisited: Novel Insights from Cognitive Perspectives, ed. by Andrea Pizarro Pedraza, 245–268. Berlin: Mouton de Gruyter. ). Those results led us to claim that semantic preferences in the expression of concepts such as abortion/to abort had social meanings of indexical nature.

In this study, we focus on the second conundrum and investigate how the participants in the taboo action of ‘to abort’ are profiled, especially the agents. There is evidence in the literature that the expression of agentivity is related to the attribution of responsibility, in the sense that explicit agents can be held responsible for their actions. Since agency implies awareness and causality, the profiling of an agent as such can be face-threatening (Brown and Levinson 1987Brown, Penelope, and Stephen Levinson 1987 Politeness: Some Universals in Language Use. Cambridge: Cambridge University Press. ; Yamamoto 2006Yamamoto, Matsumi 2006 Agency and Impersonality: Their Linguistic and Cultural Manifestations. Amsterdam/Philadelphia: John Benjamins. ) and consequently considered dysphemistic (Crespo-Fernández 2013Crespo-Fernández, Eliecer 2013 “Words as Weapons for Mass Persuasion: Dysphemism in Churchill’s Wartime Speeches.” Text and Talk 33 (3): 311–30. ). These aspects are even more relevant for the expression of adverse situations where the attribution of responsibility implies blame (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ), such as stigmatised actions like abortion, situated at the intersection of the taboos of sexuality and death. Whether the abortion is understood as birth control or as a blameable crime against life (roughly pro-choice or pro-life discourses) will surely influence how the action is expressed.

In Spanish, agentivity can be mitigated through less agentive profilings in discourse, such as the use of passive constructions (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ), a syntactic resource for euphemism (Crespo Fernández 2007Crespo-Fernández, Eliecer 2007 El eufemismo y el disfemismo. Procesos de manipulación del tabú en el lenguaje literario inglés. Alicante: Universidad de Alicante., 45). However periphrastic passive constructions in oral language are quite scarce (De Cock 2014De Cock, Barbara 2014 Profiling Discourse Participants. Forms and Functions in Spanish Conversation and Debates. Amsterdam/Philadelphia: John Benjamins., 194), so how do speakers mitigate the possible face-threatening nature of the action ‘to abort’?

Examples (1–3) (from MadSex corpus, see 3.1) illustrate cases of ‘to abort’ that vary both at the onomasiological and at the profiling level. Onomasiologically, the action is construed by the direct concept abortar ‘to abort’ but also by other indirect construals such as matar ‘to kill’ and apañar ‘to fix’. In terms of participants, Examples (1–2) have the woman who aborts as the agent, whereas (3) profiles the woman as the patient (through a generic te ‘you’ as the direct object) and an indeterminate 3rd person plural as the agent of apañar ‘to fix’, which refers to whoever performs the medical procedure.

It seems that speakers have several options to express the action and its participant(s), and that these options could possibly be classified according to the conveyed degree of agentivity. An important part of the abortion debate is related to the (ir)responsibility of the person who aborts and the kind of event that an abortion is (Pizarro Pedraza 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. ). With that in mind, and in light of the examples, the goals of this paper are twofold: first, to research what linguistic resources are used to mitigate the face-threatening nature of the event to abort and its participant(s), and second, to interpret the meanings of the different patterns against the discourses about responsibility and abortion. Our study will be led by the following research questions:

1. How do speakers combine linguistic elements to encode different levels of agentivity when expressing taboo actions, such as ‘to abort’?

2. How do different levels of agentivity interplay with the conceptual choice for expressing ‘to abort’?

3. Are different levels of agentivity correlated with social variables of the speakers?

Our hypotheses are the following: 1. Speakers encode different levels of agentivity by modulating the verbal expression of ‘to abort’ at different syntactic levels (subject, verb, object) and additively; 2. The profiling will have some relation with the semantic choices for the construal of ‘to abort’, since at least part of the agentive profile depends on semantic requirements of the action verbs themselves; 3. Given our previous findings (Pizarro Pedraza 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. ), syntactic choices will be correlated with social features such as gender, age, or the ideology towards abortion, namely pro-abortion speakers will prefer more agentive profilings.

The paper has the following structure: Section 2 summarizes theoretical aspects related to the face-threatening nature of agentive construal, with a focus on abortar ‘to abort’. In Section 3, the analyses and results are explained. Section 4 provides an interpretation of the results. The conclusion is given in Section 5.

2.Agentivity and taboo

The definition of an action is since Aristotle related to the notion of ‘intentionality’ (Aristotle, De Anima; in Yamamoto 2006Yamamoto, Matsumi 2006 Agency and Impersonality: Their Linguistic and Cultural Manifestations. Amsterdam/Philadelphia: John Benjamins. ). Other relevant concepts are ‘awareness’, ‘causality’ and ‘responsibility’, often related to ‘blame’, a key notion to understand the face-threatening nature of agentivity (Yamamoto 2006Yamamoto, Matsumi 2006 Agency and Impersonality: Their Linguistic and Cultural Manifestations. Amsterdam/Philadelphia: John Benjamins. ).

In Linguistics, an action is a dynamic event with an agent that has control over it, which can be determined through different criteria (Halliday 1967Halliday, Michael Alexander Kirkwood 1967 “Notes on Transitivity and Theme in English: Part 2.” Journal of Linguistics 3 (2): 199–244. ; Lyons 1968Lyons, John 1968 Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. ; Gruber 1967Gruber, Jeffrey S. 1967 “Look and See.” Language 43 (4): 937–47. ; cited by Cruse 1973Cruse, D. A. 1973 “Some Thoughts on Agentivity.” Journal of Linguistics 9 (1): 11–23. ). Cruse (1973)Cruse, D. A. 1973 “Some Thoughts on Agentivity.” Journal of Linguistics 9 (1): 11–23. explains the ‘to do’-test, which requires that the sentence (for instance, John jumped) allows for a paraphrasis by a ‘to do’ rather than a ‘to happen’ sentence, such as What John did was that he jumped, rather than ?What happened to John was that he jumped. Interestingly, quite different verbs can pass the test, which leads Cruse to argue that four agentive features are brought about by it: volitive, effective, initiative and agentive features. This means that not only the prototypical agent (that does the action with her/his own force) will emerge from the test (henceforth, we will use ‘prototypical agent’ as proposed by Cruse). For Spanish verbs, Delbecque and Lamiroy (1999)Delbecque, Nicole, and Béatrice Lamiroy 1999 “La subordinación sustantiva: Las subordinadas enunciativas en los complementos verbales.” In Gramática Descriptiva de La Lengua Española (Vol. 2), edited by Ignacio Bosque and Violeta Demonte, 1965–2083. Madrid: Espasa-Calpe. have tested the (non)agentivity of a construction by adding the adverbial locutions a drede (‘on purpose’) or sin querer (‘without wanting to’), which underline the voluntariness and the non-voluntariness of the event, respectively.

Some resources contribute to a non-agentive profiling of an action or to the obscuring of the agents’ participation. In Spanish, these are the passive voice, reflexive passives or se-middles and the use of impersonal constructions such as se-impersonals, generic 2nd person singular, 3rd person plural, uno, etc. (Mendikoetxea s1999aMendikoetxea, Amaya 1999a “Construcciones con se: Medias, pasivas e impersonales.” In Gramática Descriptiva de la Lengua Española, edited by Ignacio Bosque, and Violeta Demonte, 1631–1722. Madrid: Espasa-Calpe.; Delbecque 2003Delbecque, Nicole 2003 “La variable expresión del agente en las construcciones pasivas.” Nueva Revista de Filología Hispánica LI (2): 373–416. ; Tolchinsky and Rosado 2005Tolchinsky, Liliana, and Elisa Rosado 2005 “The Effect of Literacy, Text Type, and Modality on the Use of Grammatical Means for Agency Alternation in Spanish.” Journal of Pragmatics 37 (2): 209–37. ; Gómez Torrego 1998Gómez Torrego, Leonardo 1998 La impersonalidad gramatical: Descripción y norma. Madrid: Arco Libros.). Another relevant aspect is the actual reference, that is, whether the speaker talks about a specific reference, with presupposed existence (Leonetti 1990Leonetti, Manuel 1990 El artículo y la referencia. Madrid: Taurus.) – including discourse participants – or s/he rather talks in general terms. The degree of factuality affects the choices related to the agent, which creates differences for instance in expository versus personal-narrative texts (Tolchinsky and Rosado 2005Tolchinsky, Liliana, and Elisa Rosado 2005 “The Effect of Literacy, Text Type, and Modality on the Use of Grammatical Means for Agency Alternation in Spanish.” Journal of Pragmatics 37 (2): 209–37. ). Moreover, the lexical semantics of the verb itself can also encode different degrees or a particular kind of agentivity (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ; De Cock and Michaud Maturana 2014De Cock, Barbara, and Daniel Michaud Maturana 2014 “La expresión de la agentividad en el Informe Rettig (Chile, 1991).” Revista Internacional de Linguistica Iberoamericana 23: 123–40.; Ávila and Gras 2014Ávila, Rubén, and Pedro Gras 2014 “ ‘No sin él’: Análisis crítico del discurso de las campañas de prevención del VIH dirigidas a hombres que tienen sexo con hombres en españa.” Discurso y Sociedad 8 (2): 137–81.). In a particular construal, agentivity is encoded at several of these linguistic levels simultaneously. Hypothetically then, the presence or absence of features, and a particular combination of them, will encode different degrees of agentivity.

Very few studies have focused on the comparison of the “agentive power” of those resources or their combinations. Berk-Seligson (1983)Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. considers agentive, dative and reflexive passive as a continuum with decreasing agentivity. De Cock and Michaud Maturana (2014De Cock, Barbara, and Daniel Michaud Maturana 2014 “La expresión de la agentividad en el Informe Rettig (Chile, 1991).” Revista Internacional de Linguistica Iberoamericana 23: 123–40., 2018De Cock, Barbara, and Daniel Michaud Maturana 2018 “Discursive Construction of Human Rights Violations: The Case of the Chilean Rettig Report.” Text & Talk 38(1): 1–28. ) analysed the representation of the agents of violent deaths in the Rettig Report on human rights abuses during Pinochet’s dictatorship in Chile. They determined five agentivity levels according to several linguistic features, such as the expression referring to the perpetrator, its syntactic function, the transitivity and the conceptual meaning of the verb. In both studies, the face-threatening nature of agentivity in negative actions was revealed in the variation of participants’ profiling: in Berk-Seligson’s examples, higher agentivity implied blame for the situations in her experiment (e.g. breaking a vase), and even more in the Rettig Report, where the avoidance of agentive profiling obscures the responsibility of the perpetrators in the death of the Pinochet regime’s victims.

2.1The case of taboo actions

We have seen that there is potential threat in agentivity, grounded on the fact of being responsible for an action and therefore potentially reprehensible. In verbs depicting actions related to taboo topics, such as sexuality, the face-threatening potential of agentivity meets the socially imposed threat of linguistic taboo. Linguistic taboos are concepts whose mention is highly regulated and even forbidden in particular contexts (Allan and Burridge 1991Allan, Keith, and Kate Burridge 1991 Euphemism and Dysphemism. Language Used as Shield and Weapon. New York/Oxford: Oxford University Press., 2006Allan, Keith, and Kate Burridge 2006 Forbidden Words. New York: Cambridtge University Press. ).

Sexual actions can be defined as dynamic events within the realm of sexuality, with agents that have control over them (such as sexual practices, seduction, reproduction or measures against). They are often interpersonal, which is conceptualized linguistically in the inclusion of more than one participant. Among the sexual actions included in Rodríguez González’s (2011Rodríguez González, Félix 2011 Diccionario del sexo y el erotismo. Madrid: Alianza., 1127–1150) Spanish lexicographical ontology of sexuality, this paper focuses on ‘to abort’.

2.2 Abortar ‘to abort’

‘To abort’ is particularly interesting for agentivity research for several reasons. Extra-linguistically, because the debate about abortion pivots partly on the concept of moral “responsibility” (Purcell et al. 2014Purcell, Carrie, Shona Hilton, and Lisa McDaid 2014 “The Stigmatisation of Abortion: A Qualitative Analysis of Print Media in Great Britain in 2010.” Culture, Health & Sexuality 16 (9): 1141–55. ), and responsibility is often related with agentive profiling. The attribution of that responsibility (or lack of it) will probably be encoded in different participant profilings in the event ‘to abort’. Moreover, abortion is, for some speakers, at the intersection of the taboos of sexuality and death. The balance towards one or the other is related to the speaker’s ideology about abortion, and whether it implies the notion of life or not (Janicki 2006Janicki, Karol 2006 Language Misconceived. Arguing for Applied Cognitive Sociolinguistics. Mahwah, New Jersey: Lawrence Erlbaum Associates.). For each case, responsibility has radically different interpretations: for the first, abortion is considered morally irresponsible, because the woman didn’t avoid pregnancy and she didn’t bear with the consequences, in the lighter version, or because she killed her future child, in the stronger version (Observatorio de Salud de la Mujer 2005Observatorio de Salud de la Mujer, O.S.M. 2005 “Estudio sociológico: Contexto de la interrupción voluntaria del embarazo en población adolescente y juventud temprana.” Madrid: Ministerio de Sanidad y Consumo.; Ramos 1982Ramos, Ramón 1982 “Informe-resumen de los resultados de una investigación sociológica sobre el aborto mediante discusiones de grupo.” Revista Española de Investigaciones Sociológicas 21 (enero-marzo): 243–54.). For the second, abortion is morally responsible because it avoids having an unwanted child or having a child in unfavourable circumstances (idem). In our previous works on online comments reacting to the current Spanish abortion law, those ideologies emerged on the onomasiological variation of the concept (Pizarro Pedraza 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. ).

Linguistically, the verb abortar ‘to abort’ is interesting in that it has a complex semantic structure that allows for syntactic variation (regardless of its figurative reading abortar un plan ‘to abort a plan’, out of our scope). First, abortar is ambiguous between two readings: abortar^I , to have a miscarriage, and abortar^II , to have an induced abortion (DLE). Where other languages have two different verbal constructions with different readings (Fr.: faire une fausse-couche / avorter; Dutch: een abortus hebben / aborteren), Spanish relies on the context in order to convey the (in)voluntary nature of the event, like in Examples (4) and (5). This also applies to the verbal expression tener un aborto (‘to have an abortion/miscarriage’). The subject of abortar (literal)/tener un aborto can only be a woman.

(4)

MAD12: […] podíamos haber sido nueve hermanos / nos quedamos en cuatro / porque mi madre abortaba /

‘we could have been nine siblings / we were just four / because my mother had miscarriages’

(5)

AA09: yo creo que los padres / no pueden obligar a a una hija / a absolutamente nada / ni a abortar / ni a tener un hijo /

‘I think that the parents / cannot force their daughter to absolutely anything / neither to abort / nor to have child /’

Our analysis will focus on the voluntary reading abortar^II since we are interested in seeing whether (and how) the ethical problematic present in the public debate is reflected in different participant profilings, according to the speakers’ ideologies.

The most common use of abortar^II is as an action verb with a woman agent as the subject and without an object (one-participant construction). It is indeed defined as intransitive by the DLE, although it can less often be used as transitive. According to Lemmens (1998)Lemmens, Maarten 1998 Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/Philadelphia: John Benjamins. , this is also the case of to abort in Modern English, which has evolved into the transitive paradigm from the formerly dominant ergative construction (with the possible alternation causative/non-causative constructions, promoted by the former dominant involuntary-abortion reading). In Lemmens’s interpretation, the woman is conceptualized as the actor and the fetus/child is under her control, as opposed to ergative constructions where the complement would be a co-participant in the event. In Spanish, this is also the case: when there is an effective transitive construction with abortar, the complement is also an affected entity and the woman is the actor. Nonetheless, strictly speaking, it is not the woman who performs the medical procedure to herself, but (probably) a doctor. In Spanish, though, ergative constructions such as the doctor aborted the woman (documented by Lemmens 1998Lemmens, Maarten 1998 Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/Philadelphia: John Benjamins. , 193, for English, although generally judged as inacceptable) are not possible.11.We confirmed this externally by searching for abortar + a, so as to trigger human direct objects equivalent to ‘The doctor aborted the woman’ (*El médico abortó a la mujer). No such constructions were found, only accusative sentences where the abortion is the object and the woman the recipient: practican un aborto a una mujer ‘[they] practice an abortion to a woman’ (retrieved from REAL ACADEMIA ESPAÑOLA: Banco de datos (CORPES XXI) [online]. Corpus del Español del Siglo XXI (CORPES). <http://​www​.rae​.es> [08-02-2017].) When there is an object, it is mostly the fetus or child, depending on the pro-choice or pro-life ideology, respectively (Lemmens 1998Lemmens, Maarten 1998 Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/Philadelphia: John Benjamins. , 213–214; Croft and Cruse 2004Croft, W., and A. Cruse 2004 Cognitive Linguistics. Cambridge: Cambridge University Press. , 19; Coulson 1992Coulson, Seana 1992 “Is Incest Best? The Role of Pragmatic Scales and Cultural Models in Abortion Rhetoric.” Center for Research in Language Newsletter 7 (2), accessed March 20, 2017, https://​crl​.ucsd​.edu​/newsletter​/7​-2​/Article1​.html.). For Lemmens, this kind of construction is the prototype of the abort cluster. We shall see whether that is also the case in our Spanish data.

This complex semantic structure and the variable features in the construction of the direct construal abortar possibly allow the speaker to convey different agentivity degrees of the woman who aborts. Interestingly, the concept can also be expressed through indirect construals, which multiply the possibilities for profiling agentivity. In the following, we will analyse how these semantic and syntactic resources interplay for particular pragmatic effects in our data.

3.Data and method

3.1Data

We base our analysis on MadSex (Pizarro Pedraza, forthcomingPizarro Pedraza, Andrea Submitted. “ MadSex: Collecting a spoken corpus of indirectly elicited sexual concepts.”), a self-collected corpus of 54 sociolinguistic interviews about sexuality in Madrid from 2011–2012 (pre-stratified by age, sex and education level). Most occurrences come from the answers to the questions about the Spanish abortion law22. Ley Orgánica 2/2010, de 3 de marzo, de salud sexual y reproductiva y de la interrupción voluntaria del embarazo. (‘Organic Law […] of Sexual and Reproductive Health and Voluntary Termination of Pregnancy’) Boletín Oficial del Estado, 04/03/2010, nº 55. In https://​www​.boe​.es​/buscar​/act​.php​?id​=BOE​-A​-2010​-3514. (crafted in 2010), but previous questions also triggered answers about abortion, such as those about the day-after pill.

We focus specifically on the action concept referring to the voluntary termination of pregnancy, ‘to abort’, both in personal or impersonal verb forms. We include the concept in its direct construal abortar ‘to abort’ or tener un aborto ‘to have an abortion’ and in all other indirect construals used to refer to that event, such as matar ‘to kill’ or decidir ‘to decide’. Every verbal expression substitutable by ‘to abort’ was considered a valid token. For each, we analyse semantic and syntactic features of the verb, the subject and the object and we measure the frequencies of their presence or absence. Although a larger data set would contribute to the statistical strength of the results, our analysis is based on the entirety of MadSex. As such, the results are representative of our corpus and are likely to hold for similar corpora. In total, the corpus contains 224 verbal expressions referring to the concept ‘to abort’ (including indirect construals).

3.2Analytical method

For each token, we extract the subject of the verb and the object, when present. For verbs, we are interested in observing the agentive or non-agentive nature chosen for expressing ‘to abort’, the presence of an object and the presence of a modal auxiliary. For the subjects, we analyse the morphological type, the entity of reference and the personal/impersonal referentiality. The objects are analysed in terms of the entity of reference. We will exemplify the analysis in the following.

3.2.1Verbs

The verbs are analysed according to the following aspects: first, we code for the presence or absence of an agentive construal, that is, whether they are action verbs that could pass Delbecque & Lamiroy’s (1999)Delbecque, Nicole, and Béatrice Lamiroy 1999 “La subordinación sustantiva: Las subordinadas enunciativas en los complementos verbales.” In Gramática Descriptiva de La Lengua Española (Vol. 2), edited by Ignacio Bosque and Violeta Demonte, 1965–2083. Madrid: Espasa-Calpe. a drede (‘on purpose’) test. In (6), we can add a drede to mata ‘kills’, because it is agentive, but not to muere ‘dies’. Semantically, choosing a non-agentive verbal concept to express an action reduces the agentivity by construing a different kind of event. Most of our tokens are agentive (n = 210/225).

(6)

MR13: eh… es una píldora… (…) mata al feto / o… el feto muere /

‘hm… it’s a pill… (…) it kills the fetus / or… the fetus dies’

Second, we code for presence (7) or absence (8) of a direct object. Most of our verbs could potentially be used with an explicit object, but they are very often used without it, especially when it is understood as prototypical (Mendikoetxea 1999bMendikoetxea, Amaya 1999b “Construcciones inacusativas y pasivas.” In Gramática descriptiva de la lengua española (vol. 2), edited by Violeta Demonte, and Ignacio Bosque, 1575–1630. Madrid: Espasa-Calpe.). This optional nature of the realization of an object is a locus for modulation by the speaker. Linking the action of the agent to its effect through the expression of the affected object, is a transfer that increases agentivity (Delancey 1984Delancey, Scott 1984 “Notes on Agentivity and Causation.” Studies in Language 8: 181–213. , 207). We have found 66 constructions with an object and 158 without (including transitive verbs without explicit object – absolute constructions – and intransitive verbs).

(7)

FR07: a… a el proceso de… que uno quiera interrumpir un embarazo no deseado / ¿no? /

‘to… to the process of… that one wants to interrupt an unwanted pregnancy / right? /’

(8)

PM02: a mí un- a mí alguien que… que aborte / porque porque el niño pueda llegar con malformaciones […] me parece perfecto /

‘to me a- to me someone who… who aborts / because because the child can come with malformations […] it seems perfect to me’

Third, due to their abundance, we code for the presence of modal elements in the verbal construction, both periphrastic and non-periphrastic (Gómez Torrego 1999Gómez Torrego, Leonardo 1999 “Los verbos auxiliares. Las perífrasis verbales de infinitivo.” In Gramática Descriptiva de La Lengua Española (Vol. 2), edited by Ignacio Bosque, and Violeta Demonte, 3323–90. Madrid: Espasa-Calpe.), such as poder (‘be able to’/‘may’), tener que (‘have to’), querer (‘want to’)… (n = 92), that imply mostly a deontic (related to volition) (9) or a dynamic (related to the circumstances) meaning (10) (Nuyts et al. 2010Nuyts, Jan, Pieter Byloo, and Janneke Diepeveen 2010 “On Deontic Modality, Directivity, and Mood: The Case of Dutch Mogen and Moeten.” Journal of Pragmatics 42 (1): 16–34. ) in our corpus. These meanings add a layer to the interpretation of the verb and the agentivity of the participants, since they modulate how the speaker represents the modal agent and patient (Verstraete 2005Verstraete, Jean Christophe 2005 “Scalar Quantity Implicatures and the Interpretation of Modality. Problems in the Deontic Domain.” Journal of Pragmatics 37 (9 SPEC. ISS.): 1401–18. , 1409).

(9)

CC01: también / no se entiende que la mujer / pueda interrumpir- o pueda tomar ese tipo de decisiones /

‘also / it’s not understood that the woman / can interrupt- or can take that kind of decisions’

(10)

JM08: que tenga que abortar y también es / en parte- bueno / en parte no / es súper chungo para ella / o sea que…

‘she has to abort and it’s also / in part- well / not in part / it’s super hard for her / so…’

3.2.2Subjects

The subjects include NPs (n = 38) and pronouns (n = 25), but mostly grammatical information deduced from the verbal morphemes of person and number (n = 102), since the realization of the subject is optional in Spanish. We also code for syntactic impersonality (11) (n = 59). When the verb is in impersonal form due to requirements of syntax (such as co-reference with the subject in the main clause), the information about the main subject is noted.

(11)

MR13: pienso que a- a lo mejor con 16 años / pues tampoco eres lo suficiente madura como para poder elegir… dar o quitar una vida /

‘I think that m- maybe when you’re sixteen / you’re also not mature enough so as to be able to chose… to give or take a life’

Then the subjects are coded according to the entity of reference. Not surprisingly, they refer mostly to the woman (n = 209) (cf. previous examples, 9–11 for instance), very rarely to the fetus/child (n = 2) (cf. Example 6), to the pregnancy (n = 1) (12), to the abortion itself (n = 2) (13), or to a third party: persons provoking the abortion (n = 7) (14) or the means for the abortion (n = 4) (15).

(12)

FC11: digamos que es un… chute de hormonas tremendo / para que… el proceso se interrumpa /

‘Let’s say that it’s a… huge hormonal shot / so that… the process stops’

(13)

CR10: yo ahora me pongo en el caso de que… en vez de un hijo tuviera una hija / y dentro de… tres años / eh… le pasara esto / pues me gustaría que… saberlo /

‘I imagine the case that… instead of a son I had a daughter / and in… three yeards / hm… this happened to her / I would like to… to know’

(14)

PJ07: o sea que el mismo que… que te está diciendo unas cosas / luego si querías / te provocaba el aborto igual /

‘So the same one that… that is telling you some things / then if you wanted / he provoked you the abortion anyway’

(15)

GR04: es un… es un… es como un chute de hormonas tremendo / que… […] te fulmina todo lo que tengas ahí /

‘it’s a… it’s a… it’s like a huge hormonal shot / which… […] fulminates everything you may have there /’

Finally we include another kind of referential information: whether the subject refers to a specific entity (whose existence is known by the speaker) (n = 33), a generic entity (n = 163) or an impersonal entity (n = 28) (16), which is hidden behind syntactically and semantically impersonal verbal constructions (infinitives, haber que-constructions, se sentences). We separate these cases from impersonal forms due to syntax because here the impersonal form is optional and therefore a matter of profiling.

(16)

PM02: disminuye el- o aumenta el plazo de… de tiempo / para que se pueda abortar sin ningún problema /

‘it decreases the- or it increases the period of … of time / for being able to abort without problems’

3.2.3Objects

When there is an object, we code for the entity of reference: whether it refers to the woman (n = 8), the child/fetus (n = 11), the pregnancy (n = 9) or the abortion (n = 39). We include in the latter category the concept of ‘decision’ (n = 12/39) (17), as we interpret that it refers metonymically to the abortion.

(17)

MLG12: la pregunta es / ¿una niña de 16 años (…) está realmente capacitada / para tomar una decisión así? /

‘the question is / a sixteen-year-old girl (…) is really prepared / to take a decision like that?’

4.Results

Table 1 summarizes the distribution of the former features in our data. Some are more frequent than others for each category: most of our construals are actions, without an object, the subjects refer mostly to a woman, and when there is an object, it mosly refers to the abortion itself. For clarity, presence of modals will be analized separately in Section 4.2.

4.1Levels of agentivity

The next step in our analysis is to observe whether the combination of the previous features can be classified in a scalar way, from the most agentive situation to the least. We propose a classification in three major levels of agentivity according primarily to a minimum of features (cf. Figure 1): their semantic nature as agentive verbs [+− action] and the referential focus on the woman as subject [+− subject = woman]. Given that the prototypical construction of ‘abortar’ in Spanish implies a woman agent (cf. 2.2), we distinguish the following levels: Level 1 Prototypical agents (actions with a woman agent), Level 2 Non-prototypical agents (actions with another agent) and Level 3 Non-agents (non-actions).

Figure 1.

Secondarily, our three levels can be subdivided according to the reference of the subject, presence of an object and its reference. They contribute to a more or less agentive profiling within the levels. We have been able to find examples of 10 combinations that we have classified in a decreasing order of agentivity.

Level 1.Prototypical agents (PA)

In our view, the combination of features will work additively: the most agentive situation is when all features that are considered to enhance agentivity are present, as represented in PA1 schema below.

PA1:

[+action, +subject =woman, +object=fetus/child]

We can find examples like the following:

(18)

GR04: lo que pasa es que abortas algo que no sabes que tienes

‘what happens is that you abort something you don’t know you have’

Note that, if the action verb is the direct abortar ‘to abort’, this schema coincides with Lemmens’s (1998)Lemmens, Maarten 1998 Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/Philadelphia: John Benjamins. prototype. Interestingly, some realizations of the schema are based on indirect construals (viz. with other onomasiological bases, e.g. to kill). Profiling someone as the agent of some actions, like to kill, enhances the face-threatening nature by explicitly situating it within a taboo conceptual area.

A modification on the maximally agentive schema of level 1 will be examples where the object is not the fetus/child (19–21).

PA2:

[+action, +subject =woman, +object≠fetus/child]

(19)

PJ07: pues- o te ibas a una clínica privada / o te ponías en manos de un carnicero /

‘so- either you went to a private clinic / or you put yourself in the hands of a butcher’

(20)

IP08: pero no te arriesgas… a hacerte una cosa tan importante y tan peligrosa … como es eso… sin comentarlo /

‘but you don’t expose yourself… to get such an important and dangerous thing … like that one is… done to yourself… without mentioning it’

(21)

MLG12: pero… ¿está realmente capacitada / para tomar una decisión así ? /

‘but… is she really prepared / to take such decision ?’

The absence of the fetus/child contributes greatly to a change in perspective: the woman becomes more central and her action has repercussions only on herself, either by explicitly profiling herself as object (19), or by profiling the procedure (20) or the decision (21). Again, only one of the examples is based on the direct construal abortar.

If the change in the object mitigates the agentivity, its absence will do so even more, since the effect of the action will not be specifically transferred to any entity.

PA3:

[+action, subject ≠ zero =woman, −object]

The version of the schema without an object is very frequent in our data (n = 69), mostly based on the direct construal (22), but also on other conceptual bases, either intransitive (23) or transitive without an object (24).

(22)

JL08: a lo mejor entra dentro de esas locuras que haces con dieciséis años / ¿no? / que abortes /

‘maybe it is falls within these crazy things you do when you’re sixteen / right? / that you abort /’

(23)

IC08: porque como tenían dinero / ellos se iban a Londres 33.Euphemistic reference to abortion through the concept ‘trip to London’ is based on the fact that women went to England to abort when it was still forbidden in Spain. / y no se enteraba nadie /

‘because they had money / they went to London / and nobody found out /’

(24)

IA13: no le dejaría abortar / me parece a mí que no tiene edad para decidir / creo / ¿eh? /

‘I wouldn’t let her abort / it seems to me that she doesn’t have the age to decide / I think / uh?’

Up until here, the examples still belong to a big group of agentive profilings (Level 1): the abortion is construed as an action with a woman agent. The next modifications have stronger effects on agentivity, and we consequently classify them as a different level.

Level 2. Non-prototypical agents (NPA)

Level 2 is constituted by examples that are agentive but that do not profile the woman as the agent, which mitigates the agentivity of abortar. This is done through two main ressources in our data, semantically, by referring to a different person as agent, or syntactically, by using an impersonal construction, which defocuses the agent altogether. In the first case, the woman can be profiled as the direct object (25) or as the indirect object (26):

NPA1:

[+action, +subject ≠ woman, +−object=woman]

(25)

PJ07: en el fin de semana te… te apañaban

‘in the week-end they… they fixed you ’

NPA2:

[+action, +subject ≠ woman, +object ≠ woman]

(26)

PJ07: o sea que el mismo que… que te está diciendo unas cosas / luego si querías / te provocaba el aborto igual /

‘So the same one that… that is telling you some things / then if you wanted / he provoked you the abortion anyway/’

In the cases of impersonal constructions, the woman can only be deduced by our extralinguistic knowledge, because her presence – and her role- is obscured (27).

NPA3:

[+action, −subject, +object ≠ woman]

(27)

FR07: me parece bien que haya la- que haya la posibilidad / de interrumpir… procesos de procreación no deseados /

‘it seems good to me that there is the- that there is the possibility / to terminate… unwanted procreation processes’

The least agentive case of the level would be impersonal constructions without object, where the action is presented as devoided of both agent and object (28).

NPA4:

[+action, −subject, −object]

(28)

IA13: Es un trauma de abortar

‘it is a trauma to abort’

Level 3Non-agents (NA)

The last level of agentivity in our classification, Level 3, groups examples which rely on a feature that differentiates them from the previous ones: they are based on non-agentive construals. Indeed, a small sample is construed with verbs that do not pass the a drede (‘on purpose’) test, expressing abortion as an event that is not an action. The woman can still be put in a central position, by being profiled as the experiencer:

NA1:

[−action, +subject = woman, +object ≠ woman]

(29)

RR06: los padres deben de comprender / que las hi- que sus hijos / pueden te- pueden- pueden tener / determinados problemas

‘parents need to understand / that the daugh- that their children / might ha- might- might have / certain problems’

Nonetheless, the possibility of not profiling the woman also exists, in cases where the focus is put on the death of the fetus:

NA2:

[−action, +subject = child/fetus, −object]

(30)

MR13: el feto muere

‘the fetus dies’

The theoretically least agentive profiling, when none of the features is present, only exists in the data with a woman profiled as the experiencer through an indirect object.

NA3:

[−action, +subject ≠ woman, −object]

(31)

CR10: yo ahora me pongo en el caso de que… en vez de un hijo tuviera una hija / y dentro de… tres años / eh… le pasara esto / pues me gustaría que… saberlo /

‘I imagine the case that… instead of a son I had a daughter / and in… three years / hm… this happened to her / I would like to… to know’

Table 2 shows the unbalanced distribution of the data within our scalar classification of agentivity: prototypical agents are much more frequent than the rest (77% of the data).

Table 2.Distribution of the data according to the levels and sublevels of agentivity: prototypical agents (PA), non-prototypical agents (NPA), non-agents (NA) (numbers and percentages)

Levels of agentivity	n	%
PA 1	8	3,6
PA 2	32	14,3
PA 3	132	58,9
Subtotal PA	172	76,8
NPA 1	4	1,8
NPA 2	7	3,1
NPA 3	7	3,1
NPA 4	22	9,8
Subtotal NPA	40	17,9
NA 1	9	4,0
NA 2	2	0,9
NA 3	1	0,4
Subtotal NA	12	5,4
Total	224	100

This is due specifically to the frequency of the subgroup PA3, that is, the agentive group without an object with a woman agent. If we understand that prototypicality is related to frequencies (Geeraerts et al. 1994Geeraerts, Dirk, Stefan Grondelaers, and Peter Bakema 1994 The Structure of Lexical Variation. Berlin: Mouton De Gruyter. ), this is the most central category in our data for abortar, and not the transitive version, as in Lemmens’s (1997)Lemmens, Maarten 1997 “The Transitive-Ergative Interplay and the Conception of the World: A Case Study.” In Lexical and Syntactical Constructions and the Construction of Meaning, ed. by Marjolijn Verspoor, Kee Dong Lee and Eve Sweetser, 363–82. Amsterdam: John Benjamins. analysis. The second most frequent level is PA2, with a woman agent and an object different to the child/fetus. The third most frequent is NP4, that is, actions without agent (properly impersonal contexts) and without object. This construction is the same as in PA3, our most frequent construal, without an explicit agent, and therefore we consider this group as an impersonal version of the prototype. Both subgroups without an object (PA3, NPA4) are the most frequent within their classes, which confirms the salience of the intransitive construction abortar.

The rest of the subgroups have frequencies lower than 5%. Within the non-agents, the group of constructions with object and with a woman as the subject (mostly, the experiencer) is the most frequent.

4.2Presence of modals

The presence of modals in the construction has an effect on agentivity, because they add a layer of meaning to the semantics of the verb. Since the same constructions can happen with or without modals, we haven’t included them as a subdividing criterion in 3.3, but rather as an added element.

In our interviews, where speakers are arguing about abortion mostly hypothetically, the modal meanings range from moral need or acceptability (deontic), to need or obligation related to the situation or to the participant (dynamic), and to volitional (intentionality) and boulomaic (preference) meanings (Nuyts et al. 2010Nuyts, Jan, Pieter Byloo, and Janneke Diepeveen 2010 “On Deontic Modality, Directivity, and Mood: The Case of Dutch Mogen and Moeten.” Journal of Pragmatics 42 (1): 16–34. ), but the most common are dynamic modals. These construe the abortion as an action that is allowed or needed due to the situation or the participant, displacing the focus from the action to its circumstances and mitigating agentivity. In our view, this focus displacement occurs in all the cases, therefore not primarily construing the event of the action itself, but the need, the possibility or the intention for it to happen. This displacement mitigates agentivity, even in cases of volitional modality where volition, an essential aspect of agentivity, is highlighted:

(32)

JB03: pues a mí me parece bien que ella tenga el poder de decidir- la última decisión de decir / “no / quiero abortar” /

‘I think it’s good that she has the power to decide- the ultimate decision to say / “no / I want to abort”/’

Around 40% of our data are used with a modal element (n = 87/224), although this varies across levels (Table 3). Arguably, besides the 98 constructions with a prototypical agent (Level 1) without a modal, the rest of the data mitigate agentivity, either by avoiding a prototypical agent or by using a modal, or both.

The most frequent of our levels, PA3, which profiles a relatively high level of agentivity, receives a modal in almost half of the cases. In fact, most of the modals are used with these constructions. This indicates an intention of nuance: speakers don’t just construe the action in its prototypical form (the woman aborted), but they highlight aspects of it, showing a more complex picture of the event, where circumstances such as necessity and possibility (had to/could) or intentions (wanted) of the agent are considered.

Table 3.Presence or absence of modal verbs per levels of agentivity (numbers and percentages)

	Modality	Modal		No modal
Levels		n	%	n	%
PA1		1	1,1	7	5,1
PA2		10	11,5	22	16,1
PA3		63	72,4	69	50,4
Subtotal PA		74	85,1	98	71,5
NPA1		–	0,0	4	2,9
NPA2		1	1,1	6	4,4
NPA3		1	1,1	6	4,4
NPA4		6	6,9	16	11,7
Subtotal NPA		8	9,2	32	23,4
NA1		4	4,6	5	3,6
NA2		1	1,1	1	0,7
NA3		–	0,0	1	0,7
Subtotal NA		5	5,7	7	5,1
Total		87	100	137	100

It is impossible to analyse in depth the role of modality for some of our other levels, since their occurrences are too few, but we shall nonetheless describe an interesting result that should be looked into with more data. The use of modals seems to increase within levels 1 and 2 (PA and NPA), so that the most agentive sublevels (PA1 and NPA1) have less modals than the second most agentive, which have in turn less modals than the least agentive groups. Apparently, when the intention is precisely to construe an agentive profile for the woman agent (están matando a sus hijos ‘they are killing their children’) or for another agent (te fulmina todo lo que tengas por ahí ‘it fulminates everything you may have there’), modal elements are infrequent, which is the case of PA1 and NPA1-2-3. On the contrary, speakers add more modals when they are already using other linguistic features that jointly reduce agentivity, such as absence of object (PA3) and/or impersonal constructions (NPA4). In other words, speakers reduce agentivity by combining linguistic means, which work additively, as suggested with our hypothesis for research question 1.

4.3Interaction with directness of construal

Our second research question was to explore the interaction between the levels of agentivity and the conceptual choices which influence the interpretation of agentivity (like to kill). In our previous studies on the conceptual onomasiological variation of abortion/to abort (Pizarro Pedraza 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. ), we distinguished between direct (abortar/tener un aborto) and indirect construals (other conceptual bases).

Direct construals are about 2/3 of our data (n = 146/224). Within the indirect construals, we also coded for the source concept (‘to terminate the pregnancy’, ‘to decide’, ‘to kill/to die’…). Table 4 shows the distribution of the different concepts according to the particular levels of agentivity.

Table 4.Distribution of source concepts per levels of agentivity

	to abort	to decide	to have a traumatic experience	to do something to oneself	to terminate the pregnancy	to go to London*	to kill/to die	to not be born	to have a surgery	to have a problem/solution	Total
PA1	1	–	–	–	–	–	5	–	2	–	8
PA2	–	16	–	7	4	1	–	–	3	1	32
PA3	119	5	–	1	1	2	–	–	4	–	132
NPA1	–	–	–	–	–	–	1	–	2	1	4
NPA2	1	–	–	2	2	–	–	–	2	–	7
NPA3	2	–	–	1	2	–	1	–	–	1	7
NPA4	19	–	–	1	1	1	–	–	–	–	22
NA1	4	–	2	–	–	–	–	–	–	3	9
NA2	–	–	–	–	–	–	1	1	–	–	2
NA3	–	–	–	–	–	–	–	–	–	1	1
Total	146	21	2	12	10	4	8	1	13	7	224

*In our previous coding, the agentive to kill and the non-agentive to die belonged to the same conceptual base, related to death. Here, however, the different profilings arise. With more data, they should be treated separately.

Some of these concepts are intrinsically agentive or non agentive, which is reflected by a higher correlation with levels of agentivity which require agents, both prototypical (the woman) and non-prototypical (someone else or impersonal constructions), like ‘to do something to oneself’, ‘to terminate the pregnancy’, ‘to have a surgery’, ‘to kill/to die’44.Cf. footnote in Table 4. and ‘to decide’. Interestingly, the latter only happens with PA, with a woman profiled as the agent of the decision. Source concepts that are processes rather than actions require participants that are not agents, like ‘to not be born’ and ‘to have a traumatic experience’.

The direct construals are mostly profiled as agentive intransitive constructions with a woman agent (PA3) (n = 119/146), our prototype, or with a lower level of agentivity, in its impersonal version (NPA4) (n = 19/146). The non-agentive cases are expressions such as pasar por un aborto ‘to go through an abortion’, which are experiences, rather than actions (the result of adding a drede ‘on purpose’ would be illogical). As stated in our second hypothesis, there is a coherent link between the speakers’ conceptual choices and the profiling of different levels of agentivity.

4.4Interaction with social variables

Finally, we wanted to explore whether we could find patterns in the profiling of agents of to abort related to the following social factors (research question 3): sex (man-woman), age group (20–34, 35–54, 54+), level of education (primary, secondary, university studies) and ideology towards abortion (pro-/anti-abortion). Previously, we had observed that there were some preferences at the semantic level (Pizarro Pedraza 2015Pizarro Pedraza, Andrea 2015 “Who Said ‘Abortion’? Semantic Variation and Ideology in Spanish Newspapers’ Online Discussions.” Australian Journal of Linguistics 35 (1): 53–75. ), similar to those found at the lexical level in referential (López Morales 2001López Morales, Humberto 2001 “Estratificación social del tabú lingüístico: El caso de Puerto Rico.” In Actas del I Congreso de la Asociación de Lingüística y Filología de América Latina (ALFAL) Región Noroeste de Europa, edited by Bob de Jonge. Estudios de Lingüística del Español 13, accessed March 20, 2017, http://​elies​.rediris​.es​/elies13​/lopez​.htm. or Cestero Mancera 2015Cestero Mancera, Ana María 2015 “La expresión del tabú: Estudio sociolingüístico.” Boletín de Filología 50 (1): 71–105. ) and non-referential taboo phenomena, such as swearing (Christie 2013Christie, Christine 2013 “The Relevance of Taboo Language: An Analysis of the Indexical Values of Swearwords.” Journal of Pragmatics 58: 152–69. ).

Chi squared and Fisher tests were performed in R. The p-values of the interaction of agentivity levels with the speakers’ sex (0.7566), age group (0.7999), level of education (0.9229), and ideology towards abortion (0.6198) are all non significant. The global distributions are followed by all levels of all groups: PA constructions are largely the most used, followed by NPA and finally NA. Therefore, against our third initial hypothesis, the preference for one of the levels of agentivity (as a measure of a preference for a certain profiling) is not related to our social categories.

Although this seems unexpected, a previous sociolinguistic study on agentive constructions in Costa Rican Spanish had also found the variable ‘sex’ non significant (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ), which was surprising for the author, since gender preferences were found for active/passive and transitive/intransitive alternations (such as Barron 1971Barron, Nancy 1971 “Sex-Typed Language: The Production of Grammatical Cases.” Acta Sociologica 14: 24–72. , cited in Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. , 159). For now, we can only conclude that prototypical agents seem to be the preferred option for all speakers regardless of their social profile, but an explanation for this result would require further empirical exploration.

5.Interpretation: Reducing agentivity as a euphemistic strategy

When speaking about the event of abortion, speakers construe three main agentivity degrees: a prototypical agent (the woman), a non-prototypical agent (not the woman), and a non-agent, with decreasing involvement from the woman. Semantic and syntactic resources are tightly interwoven.

A prototypical agent (PA) is the one who, according to the speakers, does the action in reality. This is reflected in our data by profiling a woman as the agent of the action verb. In this group, we find most of the examples of the intransitive direct construal abortar, which constitute also most of the data (our prototype). Although this can give the impression that our speakers prefer very agentive profilings, the use of modals is actually very frequent. Modals reduce agentivity, by putting the focus on something different than the action. Moreover, this level also includes indirect construals, such as decidir ‘to decide’ o hacerse algo ‘to do something to oneself’. In these, speakers retain the agentive nature of the event, but they construe the action of abortion through metonymy or hyperonymy, respectively. The stigmatised meanings of ‘to abort’ are then avoided by conceptualizing it as a (morally responsible) decision or euphemistically depriving it of meaning by using a vaguer expression (Grondelaers and Geeraerts 1998Grondelaers, Stefan, and Dirk Geeraerts 1998 “Vagueness as a Euphemistic Strategy.” In Speaking of Emotions: Conceptualisation and Expression, edited by Angeliki Athanasiadou, and Elżbieta Tabakowska, 357–74. Berlin/New York: Mouton de Gruyter. ), only interpretable in context. Therefore, the cases of direct construal in combination with agentive profiling without modals (e.g. la mujer aborta ‘the woman aborts’) are reduced to 61 out of 224. In the rest of the data agentivity is mitigated through different semantic and syntactic means.

The resource of profiling non-prototypical agents (NPA) (doctors, means of abortion, zero-agents) reduces the face-threatening nature of the action by not focusing on the woman, and therefore the responsibility, and the potential blame, is not attributed to her. These strategies have also been found in previous agentivity studies, such as the use of inanimate entities as subjects (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ) or the use of impersonal constructions (De Cock and Michaud Maturana 2014De Cock, Barbara, and Daniel Michaud Maturana 2014 “La expresión de la agentividad en el Informe Rettig (Chile, 1991).” Revista Internacional de Linguistica Iberoamericana 23: 123–40.). The use of modals in the NPA group underlines their mitigating role, since the impersonal constructions that have a deduced woman agent (NPA3–4) have more modals. For the others, the profiling of different agents seems to be less face-threatening and thus focusing on the circumstances is less needed.

The profiling of non-agents takes a step further: the type of event changes by construing an action as an experience or a process. The presence is very limited in our data, probably due to the situation and to topics of the interview. Since the focus is put on the debate about the legality of abortion, speakers discuss the possibility of aborting and the circumstances of a hypothetical woman, rather than personal narratives that could probably trigger experience verbs.

The previous summary answers our first research question by describing the different linguistic elements that mitigate agentivity and how they combine in three different levels. The answer to the second research question – how do these levels combine with conceptual choices – raises an apparently contradictory fact: the majority of the construals have some form of mitigation of agentivity, but at the same time, most of the data use the direct construal – abortar. The explanation to this apparent semantic-syntactic contradiction sheds light on what is truly taboo in abortar.

Since the majority of the tokens are based on the direct construal, we propose that the concept itself is not taboo. We have argued elsewhere that conceptual directness is related to the formality of the question topic (Pizarro Pedraza, forthcoming). It is likely that a tendency to orthophemism – the neutral option – (Allan and Burridge 2006Allan, Keith, and Kate Burridge 2006 Forbidden Words. New York: Cambridtge University Press. ) is in place when using the direct abortar, triggered by an intention of straightforwardness. Nevertheless, this intention does not go against a mitigated agentive profiling: in our data, non-mitigated abortar is not the preferred construal, as is shown by the insistence of our speakers in using some form of mitigation of agentivity (either through a lower level of agentivity or through the use of modals). Consequently, we can conclude that speakers are not euphemizing the concept, but the profiling of the action, because what they are avoiding is a maximally agentive expression of to abort. This linguistic preference arguably reflects cultural discourses about abortion, intentionality and responsibility.

Agentivity is related to responsibility and blame, when the action is considered negative (Berk-Seligson 1983Berk-Seligson, Susan 1983 “Sources of Variation in Spanish Verb Construction Usage: The Active, the Dative, and the Reflexive Passive.” Journal of Pragmatics 7 (2): 145–68. ). This is the case of abortar (Coulson 1992Coulson, Seana 1992 “Is Incest Best? The Role of Pragmatic Scales and Cultural Models in Abortion Rhetoric.” Center for Research in Language Newsletter 7 (2), accessed March 20, 2017, https://​crl​.ucsd​.edu​/newsletter​/7​-2​/Article1​.html.), still stigmatised in Spain despite being legal. The stigmatisation of abortion rests partly in its conceptualization as an undesired event, as opposed to (wanted) maternity, which is the desirable situation, an ideology that is deeply rooted in established discourses about women’s reproductive features (Kumar et al. 2009Kumar, Anuradha, Leila Hessini, and Ellen M. H. Mitchell 2009 “Conceptualising Abortion Stigma.” Culture, Health & Sexuality 11 (6): 625–39. ; Norris et al. 2011Norris, Allison, Danielle Bessett, Julia R. Steinberg, Megan L. Kavanaugh, and Davida Becker Silvia De Zordo 2011 “Abortion Stigma: A Reconceptualization of Constituents, Causes, and Consequences.” Women’s Health Issue 49–54. ). This implies that, to be morally acceptable, abortion has to be unwanted and problematic for the woman. This idea is present in some of our Examples (33). A fully agentive construal of the abortion event connoting straightforward intentionality and full control, clashes with those discourses about abortion and is, we argue, the true taboo of abortion.

Since our speakers are talking about abortion referring to hypothetical situations not about own actual experiences (with one exception), mitigation of agentivity is thus not covering responsibility of any discourse participant because they are not part of the action. This does not seem to reduce the need for euphemism, although only a comparison with personal narrations would confirm this possibility. For now, we can hypothesize that social tact and dignification, general functions of euphemism (Crespo Fernández 2007; Chamizo Domínguez and Sánchez Benedito 2000Chamizo Domínguez, Pedro J., and Francisco Sánchez Benedito 2000 Lo que nunca se aprendió en clase. Eufemismos y disfemismos en el lenguaje erótico inglés. Granada: Comares.), are at play: reducing agentivity is a way for our speakers to be tactful in their arguments about abortion, but also to somehow “dignify” the blunt agentive construal, in a way that is coherent with the cultural and ideological context of our interviews. Interestingly, going back to our third research question, all our speakers, regardless of age, sex, education or ideology about abortion, use our three levels of agentivity similarly. Whether these phenomena are impermeable to social variation (an unlikely conclusion, considering that pragmatic and discourse features are subject to variation) is a question that we should examine with more data. What we can provisionally deduce is that the topic of abortion is generally understood as highly delicate and it does not come without nuance, a fact that reflects the stigmatisation of the procedure despite its legality, especially when the woman is fully agentive.

6.Conclusion

In this paper, we have analysed 224 utterances referring to the action concept abortar, in a Spanish corpus of sociolinguistic interviews about sexuality. Our analysis shows that speakers have a clear preference for action verbs, for the explicit profiling of a woman agent and for constructions without an object, taken individually. Despite these isolated preferences, it is only in their combination that we can understand the strategies that speakers use to convey their meanings in discourse. Based on the presence or absence of action verbs and on the reference to a woman as the subject, we have proposed a scale with 3 major levels of agentivity: prototypical agents, non-prototypical agents and non-agents, subdivided in a more fine-grained classification according to the entity of reference of the subject (other than the woman) or impersonal constructions, the presence of an object, and the entity it refers to. Modals are used often, which we have interpreted as an added resource for mitigating agentivity. Furthermore, our scale interplays with directness of construal at the lexical semantic level, which is in part explained by the fact that some conceptual bases require agentive or non-agentive construals, but their relation is not categorical.

All resources considered, we can conclude that there is a fair amount of mitigation of agentivity in our data and just a few tokens that construe a maximally agentive situation that indeed convey blame (e.g. están matando a sus hijos ‘they are killing their children’). Although the context of the interview and the formality of the subject under discussion (the legality of abortion) favour an orthophemistic use of abortar instead of other ideological and affective construals, the agentivity of the direct construal is frequently mitigated. We believe this mitigation serves euphemistic purposes since its function is to avoid the taboo, which is not the concept of ‘to abort’ itself, but the agentive profiling of the woman who aborts. In our view, this is based on the conceptualization of abortion as a highly stigmatised action, which has to be unwanted in order to be morally acceptable.

Against our initial hypothesis, the distribution of our levels of agentivity is not significantly related with our social variables (sex, age, level of education, ideology), at least for the data at hand.

With this analysis, we have tried to contribute to a better understanding of linguistic taboo and how it affects discourse beyond lexical semantics. When dealing with problematic concepts, the choice of a construal reveals affective and ideological meanings that the speakers link to the target. The research of those conceptual links and their cultural implications is certainly relevant, if not urgent. Moreover, this study proves that taboos rely not only on concept-to-concept associations, but also on the articulation of those concepts within more complex conceptualizations that are reflected at the level of sentence construction. In that sense, the perceived moral unacceptability of a fully agentive woman who aborts, motivates the euphemistic mitigation of agentive profiling, not only through concept-to-concept choices, but also through referential and syntactic choices. This leads us to conclude that the effect of linguistic taboo has a broader span than is usually taken into account, since it falls on culturally based conceptualizations, and not only on individual concepts.