SPEAKER SELF-PROFILING THROUGH DISCURSIVE INDEXATION AND SYNTACTIC ENCODING IN SPANISH RADIO TALK *

Syntactic and discursive choices in context constitute resources for the interactional profiling of the direct participants. This study analyzes the frequencies with which speakers index themselves, as well as the syntactic functions they tend to accord themselves when doing so, in a corpus of Peninsular Spanish radio discourse featuring a variety of textual genres and speaker socioprofessional identities. The analysis is restricted to indexations of the singular first person in central syntactic functions, i.e. those with the capacity to establish agreement with the verb. A dichotomy is proposed between subject and (accusative or dative) object self-encoding, based on the different morphological means through which verbal agreement is carried out in this language, namely verbal endings and clitics. Both the statistical patterning of variation and the discursive-pragmatic motivations of particular choices are subsequently examined. The selection of a specific syntactic function for the encoding of the speaker is found to serve communicative goals related to the textual genre and to the kinds of socio-professional identities speakers intend to develop within it. Significant correlations are obtained between higher percentages of self-encoding as subject and of discursive self-indexation altogether, although speakers presenting themselves as political representatives diverge from this tendency for particular communicative reasons. The results are interpreted as being parallel to a discursive-cognitive continuum between subjectivity and objectivity that underlies speaker interactional self-profiling and discourse construction.


Introduction
Linguistic choices can hardly be explained without considering the broader, multimodal contexts where they are carried out; grammar and discourse are mutually constitutive (Ariel 2009;Mithun 2015).The insistence of formal linguistic paradigms on the need for a strict separation of analytic levels has been superseded by functional and cognitive perspectives.Any choice made at formal levels is at the same time the choice to communicate a particular meaning within a context, linguistic constructions being inherently meaningful (Goldberg 2006;Langacker 2009: ch. 1).This can affect any aspect of the configuration of clauses, not least the allocation of syntactic functions within them; a given grammatical configuration is a particular way to conceptualize the events of the world.Besides, it concerns the very choice of which elements are discursively indexed, as well as the relative degree of attention they receive.
The present study will approach discursive self-indexation and syntactic self-encoding in Spanish as strategies for the presentation of speakers and the achievement of communicative goals in particular contexts.Every time people produce some self-referring linguistic element, they are making themselves present in discourse.Besides, they can accord themselves a particular syntacticsemantic role, thus shape their own way of participation in the events described.In the canonical event model (Langacker 2008: 357), an agent acts on a patient and provokes some change of state in it.The prototypical linguistic manifestation of this kind of event in many languages is an active declarative clause where the agent-the most salient participant-is encoded as the subject, and the patient as the accusative object.It follows that, by encoding some participant as subject, the speaker makes it come under the focus of attention and thus enhances its salience (see García 2009: 52-54 for the specific case of Spanish).Furthermore, other syntactic features such as the expression vs. omission of elements, as well as their preverbal vs. postverbal placement in languages allowing such choices, also decisively interact with the structure of events and the relative salience of the participants within them (Delbecque 2005;Aijón Oliva & Serrano 2013;Serrano 2014Serrano , 2017)).
We will analyze some ways speakers in Peninsular Spanish radio programs manage self-indexation in discourse and choose between syntactic selfencoding as clause subject vs. as (accusative or dative) object.In order to delimit the scope of research, the analysis will be restricted to singular firstperson forms.There is of course a much wider variety of grammatical resources that can be used for self-expression, including the plural first person, the singular second one with a speaker-inclusive reference, third-person indefinite uno/una 'one', as well as different impersonal constructions, all of which would merit further consideration (see Posio 2016 for a study where several of these constructions are jointly addressed).
Expressions that refer to the direct participants of discourse have several simultaneous functions: they transmit information, conceptualize the relationship between the coparticipants, and contribute to the structuring of the interaction.These are all manifestations of the complex phenomenon of interactional profiling, which can be understood as the way discourse participants and the relationships among them are represented (De Cock 2014: 1-4).As will be observed, syntactic and discursive phenomena like the ones under study are used as devices for both the configuration of discourse and the management of personal identities and interpersonal relationships.
This also suggests the existence of intrinsic links between quantitative patterns of variation and particular strategies of contextual choice.The ultimate goal of research should be to achieve a theoretical explanation of such links.In this sense, here we will propose that strategies for self-profiling by means of linguistic choices are parallel to the discursive-cognitive continuum between subjectivity and objectivity.The way speakers profile themselves is undetachable from the construction of more subjective vs. more objective instances of discourse.The opposed poles of the continuum are directly connected with the particular grammatical variants embodying them across discourse-as manifested in the very terms subject and object-just as with 'subjective' vs. 'objective' interpretations of speakers and their utterances in the more everyday sense of the terms (see further Section 7).
The remainder of this paper is organized as follows.In Section 2 we will begin by presenting an overview of subject and object encoding in Spanish, taking verbal agreement as a basic grammatical feature whose communicative and cognitive relevance justifies the distinction between central and peripheral syntactic functions in this language.This is subsequently applied to the singular first person, which constitutes the main interest of the present study.Section 3 describes the corpus analyzed and the textual genres and types of speaker contextual identities featured within it, as well as the methodology to be employed in the analysis.The latter is developed across Sections 4 to 6, taking into account the statistical variation according to the factors considered as well as its most significant discursive-pragmatic projections.In Section 7, the quantitative and qualitative data obtained are jointly discussed as manifestations of a general continuum between cognitive subjectivity and objectivity in speaker self-profiling.Finally, Section 8 summarizes the main findings of the study and comments on some directions for further research.

Subject and object encoding in Spanish. The singular first person
Natural languages show a preference for the elements placed higher on salience hierarchies to be morphologically indexed in the verbal nuclei of clauses, as well as to impose their morphological marks on other elements (Company Company 2001: 7).In Spanish, agreement is a pervasive phenomenon whereby some connection between two or more elements of discourse is made manifest through morphological means.It happens at the phrasal, syntactic and discursive levels.In (1) below, all words within the NP las gatas blancas share the female gender morpheme -a, as well as the plural number morpheme -s; redundant as this might seem, the failure of any word to agree with the others would be considered ungrammatical.In (2) we can observe the coincidence of person and number morphemes between the subject and the verbal nucleus, as well as that of gender and number morphemes between the subject and the predicate object.In fact, all words in the utterance would need to be plural-inflected.Across a broader discursive context, it is the recurrence of such coreferential deictic-anaphoric elements that will make it possible for the audience to keep track of referents.
(1) The capability of establishing agreement with the verb-i.e. to be indexed through coreferential morphemes bound to the verbal root-is the main feature characterizing central syntactic functions as against peripheral ones in Spanish (García-Miguel 2015: 207) and appears to be a crucial feature in the configuration of clauses and the cognitive construction of the events they describe.Following the criterion of agreement, central functions would be those usually labelled as subject, accusative or direct object, and dative or indirect object.These must be understood as the grammaticalized ways of expression of the central participants in an event.However, verbal agreement is carried out by different means with subjects vs. objects.
Agreement between subject NPs and verbal endings is categorical in Spanish, thanks to a rich inflectional paradigm offering different forms for most grammatical persons and tenses, which in turn usually makes the lexical or pronominal expression of the subject not compulsory: (yo) voy 'I go', (tú) irás 'you (sing.)will go', (ellos) fueron 'they went', etc.The preceding statement on the categoricity of subject agreement is, however, somewhat circular given that, if agreement fails, the clause will be described as lacking a subject, i. e. as impersonal.In some contexts where variation in usage is observed, such as (3a) vs. non-standard (3b), the NP disturbios 'riots' will respectively be analyzed as accusative object and as subject, on the grounds of verb agreement itself.
(3) a. Hub-o disturbios en las calles There be 3RD.SING.PAST riots on the streets b.Hub-ieron disturbios en las calles There be 3RD.PL.PAST riots on the streets 'There were riots on the streets'.
In other words, it is actually coreferentiality with verbal morphemes that characterizes the syntactic function of subject.This nevertheless entails a preference for many other non-categorical traits: agentive semantic roles, preverbal placement or omission in the clause, referential animacy and definiteness, as well as a higher degree of cognitive salience altogether (Aijón Oliva & Serrano 2013: 73), all of which defines a prototype of subject that actual elements in discourse will approach to different degrees.It is easy to observe that the inanimate, indefinite disturbios in (3b) is scarcely prototypical for a subject, which does not prevent some speakers from according it this syntactic function in the absence of any other element with higher salience within the event.
On the other hand, Spanish accusative and dative objects can also establish agreement with the verb, in this case through verbal clitics (Enrique-Arias 1993;Franco 2000;Aijón Oliva 2006: ch. 4).These are a set of unstressed monosyllabic morphemes resulting from the evolution of Latin accusative and dative personal pronouns, and which are placed at the left of the verbal root in most contexts (e.g.Lo [Acc clitic] tengo 'I have it'), apparently in order to avoid the accumulation of inflectional morphemes at the right (Enrique-Arias 1993: 43, Rini 1995: 189).Their formulation is categorical whenever the object is a tonic pronoun, which includes all instances of first and second persons, as well as pronominal third ones.However, with lexical NPs clitic agreement is variable and subject to contextual features, such as their preverbal vs. postverbal placement and the degree of animacy and specificity of their referents.In (4), the singular first-person clitic me, coreferential with the tonic object a mí, is mandatory irrespective of the syntactic function, explicit formulation and placement of the latter.On the other hand, in (5) it can be observed that, whereas the pronominal dative object (a) ellos necessarily agrees with the verb through the dative clitic les, agreement fails with the lexical, inanimate accusative la noticia.However, the placement of the latter NP before the verb, entailing an increase in its topicality, would force agreement with both objects simultaneously, at least with unmarked declarative intonation, as shown by the formulation of the clitic la in ( 6 The relative complexity of Spanish verbal morphology-which is notorious in the accumulation of grammatical information within the verbal complex se la di in the preceding example-usually makes it possible to elide both the subject and the agreeing objects, as long as the discursive context makes verbal endings and clitics sufficient for their respective identification.This is specially evident with first-and second-person agreement morphemes, since they can only have the direct participants as referents, which justifies the usually higher rates of omission with these two persons as against the third one (see e.g.Dahl 2000: 64;Posio 2012: 340).
It is also necessary to acknowledge that the description of Spanish clitics as bound agreement morphemes is not unanimously accepted.These elements have undergone a long process of grammaticalization from early stages of the language that is not yet completed, and many descriptive studies and grammars still characterize them as (unstressed) object pronouns.In turn, cases of cooccurrence between the clitic and its coreferential object within the same clause are often termed clitic doubling (e.g.Vázquez & García 2012;Belloro 2015) and approached as a sort of redundancy whereby a referent is indexed through two different syntactic constituents-a construction that actually has no apparent equivalents across Spanish grammar.This is a complex matter that cannot be extensively discussed within the limits of the present study.However, a view of clitics as belonging in the domain of morphology rather than that of syntax helps solve the problems allegedly posed by clitic doubling constructions.If verbal inflections indexing the subject can hardly be described as 'expressed subjects' or as 'redundant' with subject pronouns or NPs, there is little justification to treat clitics in a different fashion.It actually seems more coherent to investigate the motivations for the discursive formulation of those pronouns whose referents are already indexed in the verbal nucleus through agreement morphemes.
As for the choice between the encoding of participants as clause subjects vs. objects, it has previously been addressed in some studies of syntactic variation, most prominently in those analyzing the semantic differences between passive and active constructions (e.g.Fernández 2007), as exemplified in (7a, b).The passive is known as a resource for the foregrounding of a semantic patient, together with the backgrounding of the agent, by inverting the prototypical distribution of syntactic functions.  Some functional and semantic analogies between Spanish constructions with an experiencer subject and those with an object playing the same semantic role are underlined by De Cock (2014: 148): "With cognition verbs taking a cogniser object, the proportion of cogniser obliques in comparison with clitic pronouns is similar to that of subject pronouns in comparison with verb inflection".It should be evident that the variants in (8), just as those in (7) above, can hardly be considered synonymous or interchangeable.The fact that in each case speakers opt for encoding themselves either as subject or as object will have important repercussions on how their role in the event is perceived, e.g.their degree of autonomy or of responsibility towards the facts described.This choice is in fact carried out in every single clause where the person speaking indexes him/herself in a central syntactic function.This suggests the usefulness of adopting a broad, comprehensive view of syntactic variation and choice.Limiting the analysis to contexts where the speaker can supposedly choose 'freely', aside from being highly questionable-it is difficult to think of any empirical justification for the claim that a speaker could have said something different from what he/she did say; see Sankoff (1988: 154)-would also block the possibility to reach a global explanation of the syntactic-discursive phenomena under study.All linguistic choices are meaningful in some way, regardless of whether they would qualify as instances of 'variation' in a more traditional sense (see e.g.Chambers 2003: 17-19 for such a notion of variation).This is an important principle for the most recent approaches to syntactic variation, whereby it is viewed as above all a way of constructing cognitive meaning (see Serrano 2016, as well as Aijón Oliva & Serrano 2013: ch. 1).
For the same reasons, in the sort of investigation proposed here it is preferable to examine all contexts where the speaker encodes him/herself as a verb-agreeing element, and analyze the patterns of choice involved in each case.Therefore, all clauses with verb forms showing first-person singular agreement-be it through verbal endings or clitics-will be taken into account.On the other hand, we will not analyze other possible ways of speaker selfindexation, such as possessives or non-agreeing objects like obliques and adjuncts, that is, peripheral syntactic functions that entail lower degrees of referent salience.
Also, even if it would be interesting to break down first-person objects into accusative (direct) and dative (indirect) ones, we will treat both functional categories jointly, as against the subject.This procedure is justified by the fact that, in the case of the first and second persons, clitics are not marked for case (Me [acc clitic] vio 'He saw me' / Me [dat clitic] dio una entrada 'He gave me a ticket'), nor are the corresponding tonic pronouns (A mí [acc] me vio / A mí [dat] me dio una entrada).Whereas in ditransitive constructions it is usually easy to formally and semantically distinguish an accusative and a dative, single objects often show intermediate features between the two prototypes. 3The distinction could perhaps be made by observing the typical behavior of each verb with third-person objects, whose clitics are marked for accusative vs. dative case; however, even in such contexts there is a fairly wide range of variability (Fernández-Ordóñez 1999).All this suggests that the most significant choice is that between subject and object as basic functional categories.In fact, the analysis will show that this choice can yield interesting communicative effects in a variety of syntactic and discursive contexts.

Corpus and methodology
The study is based on the radio subsection of the Corpus de Lenguaje de los Medios de Comunicación de Salamanca (MEDIASA). 4It comprises a total 151,995 words, resulting from the transcription of local radio programs broadcast by stations from the central Peninsular Spanish town of Salamanca during the years 2003 and 2004.The corpus is structured into five different textual genres, while its participants are classified into four main types of socioprofessional identity.First, here is a brief presentation of the main features of each of the genres.a) News reports (18,155 words).These quite short programs (between 5 and 15 minutes) usually consist in the aloud reading of written news texts.They are expected to be informational and ideologically aseptic; however, the inclusion of recorded clips of interviews, as well as of occasional comments by the broadcaster, often causes a notorious change in communicative style.
b) Talk magazines (62,483 words).They could also be termed generalist magazines, as opposed to the more sectorial sports and music ones.In fact, magazines combine materials typical of all other kinds of programs, but at the same time are recognizable as a genre in their own right.This is partly due to the unifying role played by the personality of their anchors.The shows are broadcast during the central hours of the day and have a predominantly female, middle-aged or mature audience as their target.
c) Sports programs (35,226 words).With football usually taking most of the airing time, they are conducted in a relatively casual tone and blend information with opinion; also frequently featured are interviews to players, coaches and other people from the sector.Most broadcasters in these programs are men, and it is easy to assume a predominantly male audience as well.
d) Music programs (20,901 words).In some stations the daily music program fills the whole morning and noon interval, even if speech itself does not take a large share of the airing time.These are freely structured magazines, progressing through successions of songs and comments on them, but also including brief news and culture reports, advertisements and contests.Most of these broadcasts are aimed at a predominantly young or middle-aged audience.
e) Commercials (15,230 words).Unlike the rest of the genres considered, commercials cannot be classified as programs, being just short texts that are inserted within programs or in the interludes between them.Their obvious goal is to persuade customers to acquire a product or a service, or even to practice some civic attitude.Depending on the topic discussed and on the targeted audience, they can adopt rather different stylistic orientations.
It is easy to infer that radio genres are often quite heterogeneous formats where very different kinds of interactions can take place.The taxonomy proposed is based on direct observation and takes into account the ways radio contents are offered to the audience, rather than the thematic and interactional features of particular sequences.Being strongly related to communicative goals, genres are prime candidates for the creation of discursive patterns for linguistic forms (Ariel 2008: 62), which justifies taking them into consideration as a potential source of variation in syntactic self-encoding.Their unequal word counts in the corpus aim at loosely reflecting the fact that their importance in the program schedules of local radio stations is also unequal.However, the calculation of normalized frequencies will help neutralize such differences when necessary.
As for the four basic socio-professional speaker identities considered, they can be labelled and described as follows: a) Journalists (83,957 words).The professionals of mass communication hold a dominant position in media discourse; they have the power to determine how, when and to what extent speakers from other groups are allowed to participate.However, this power is countered by the social exposure inherent to their job-they work in front of a presumably large and mostly unknown audience that will judge their professional performance.They are often obliged to display an image of efficiency, educatedness or neutrality which is likely to be reflected on their expressive choices.
b) Public figures (40,951 words).This is a more heterogeneous group, comprising speakers not professionally devoted to radio communication but who are expressly invited to participate in it.These include intellectuals, artists, entrepreneurs, sportspeople, advertisers, representatives for companies and associations, etc.Just like journalists, they are often guided by the purpose of displaying a personal image of professionalism and efficiency; however, they also enjoy a higher degree of expressive freedom.c) Politicians (18,533 words).This characterization is applied to speakers presenting themselves as political-party or trade-union representatives.The decision to detach them from the previous group is based on the peculiarities of their interactional behavior-with an explicit and commonly recognized orientation to the functions of argumentation and persuasion-as well as on the particular social profile that defines them.d) Anonymous individuals (8,554 words).Finally, this fourth group comprises all speakers who only take part in media interactions circumstantially: callers to radio programs where time is allowed for citizen participation, or passers-by being surveyed by reporters.Their occupations and social affiliations are in principle not publicly relevant, and sometimes their names are not mentioned-still, the term anonymous is used here as a mere conventional label.
These categories should be understood as contextual identities (Kerbrat-Orecchioni 2005: 157), i.e. as compounds of self-presentation features that make people qualify for playing particular interactional roles, which will in turn be characterized by sets of communicative rights and obligations.The same speaker might adopt different social identities depending on the textual genre, or even oscillate between two or more of them within the same interaction or across time.While journalists have traditionally been expected to be impartial and objective, there is a growing demand for them to adopt a critical stance towards politicians' utterances (see Patrona 2011).An anonymous individual calling to a program can reveal him/herself to be a specialist on some topic under discussion, assuming the identity features of a public figure; a journalist can express concern about an everyday matter as an anonymous caller would, and so on.However, in most cases the identities appear to be pre-assigned from the beginning of the interaction, and even possible changes will tend to be interpreted in the light of the 'primary' identities-a political speech uttered by a journalist will probably sound less 'political' than if pronounced by a politician proper, and more as information or as some sort of public service.
Socio-professional identities such as the ones considered can also entail a strong notion of groupness (Edwards 2009: 25-27), particularly when speakers appear as representatives of news corporations, political parties, sports teams or any other sort of human association.However, whereas such traits of identity will favor the inclusion of the speaker into a wider 'we', this investigation is primarily concerned with the domain of the 'I', which means that individual psychological and affective traits can prove just as relevant.
In the following sections, after briefly surveying the general figures of discursive self-indexation and subject vs. object encoding in the corpus (Section 4), we will quantitatively and qualitatively analyze these choices across the genres (Section 5) and speaker identities (Section 6) represented in the corpus.More specifically, we will first calculate the different percentages of subject vs. object encoding as well as the normalized frequencies (per each 10,000 words) of speaker self-indexation, in order to ascertain whether there are significantly diverging tendencies according to the contextual factors considered. 5If this is the case, we will subsequently attempt qualitative interpretations of the data by observing the contexts where particular choices are usually made, trying to ellucidate the basic discursive-pragmatic purposes they serve in such contexts.This way it will be possible to explain how the self-indexation of speakers, as well as the syntactic functions they accord themselves in each case, interact with the broader discursive contexts established by textual genres and with the management of speaker identities within them.

General data
There are a total 1763 tokens of first-person subject or object agreement morphemes across the corpus.It is much more usual for speakers to encode themselves as clause subjects: 1456 instances of singular first-person verbal endings as against just 307 of clitics.This would appear to be a grammatical reflection of a well-known cross-linguistic tendency: "Most utterances in discourse are egocentric, that is, the situation or event depicted in the utterance is presented from the point of view of the speaker" (Siewierska 2004: 201).Egocentrism promotes a dominance of clauses where the speaker is constructed as the main participant of the event. 6This, in turn, makes it advisable for our subsequent analysis to pay closer attention to patterns of object encoding, as it appears to be a less prototypical choice for the first person and should therefore be endowed with the power to generate more significant pragmatic effects.
However, this will of course also depend on what is made more expectable by each context.
Table 1 displays the total token numbers and percentages of first-person subject vs. object encoding.Also, the column on the right side shows the total amount and normalized frequency of self-indexations of the speaker in central functions.As can be seen, there are an average 116 tokens per 10,000 words.Frequency normalization will be particularly useful when analyzing and comparing textual genres as well as socio-professional identities, since their respective word counts are, as noted in the preceding section, rather disparate.

Speaker self-encoding across textual genres
Let us now consider the possible interactions between the self-profiling phenomena under study and the five radio genres distinguished.Table 2 shows the percentage of subject vs. object encoding in each of them.Again, the column on the right displays the total numbers of self-indexations and their normalized frequencies per 10,000 words (see the total word count of each genre in Section 3).The differences between subject and object encoding across genres can be considered statistically significant (χ 2 = 31.444;p < 0.01).It is interesting to observe that the three genres with the lowest normalized frequencies of speaker indexation-news reports, music programs and commercials-are at the same time those with the highest percentages of self-encoding as object, all of them exceeding 20%.Conversely, the genres that appear to be more speaker-oriented (talk magazines and sports programs) surpass the general percentage of subject encoding (82.6%).This suggests that the indexation of speakers in discourse and their encoding as clause subjects, though not being assimilable phenomena, may have analogous cognitive and functional motivations that need to be explained.Let us now turn to the examination of the contextual conditions under which different genres promote speaker indexation, and the syntactic function that is preferred in each case.
Both news reports and commercials are highly pre-planned and scarcely interactional genres where the discursive presence of the speaker is not often expectable, which explains their low frequencies of first-person indexation overall.In news discourse, broadcasters mostly limit themselves to reading narrative texts featuring third-person referents.When additional comments are uttered, most often with a discourse-structuring function, the plural first person is the most usual choice.The speaker avoids the manifestation of his/her own presence in discourse, rather opting for a plural that suggests inclusion in a wider group (see also Bull & Fetzer 2006: 14-15), apparently the staff working at the radio station.The only contexts showing some frequency of speaker indexation are recorded clips where interviewed speakers talk on their own behalf.In such cases, object encoding is often associated with narrative sequences where speakers recount deeds of others that caused some damage or discomfort to them, as in example ( 9).The encoding of the speaker as object is coherent with his/her role as the patient of external actions.
(9) empezó insultándome: y: después que me iba a hacer lo mismo que a mi amiga: / entonces le dije / que: / que qué amiga • me dijo que la que: / habían encontrao en el porta:l <Inf-SE-180603-14:20> 'He started by insulting me, and then he said he was doing to do the same to me as he had done to my friend.I asked what friend, and he told me it was the one that had been found [dead] at the entrance'.
In commercials, the singular first person is also much less common than the plural one, due to the fact that the intended addresser is usually a human group-a corporation, a political party, etc.-that goes beyond the specific speaker.However, there are texts written from the viewpoint of one (often fictitious) person recounting his/her experience as a customer or as a worker.Other occasional pieces are constructed in the form of fictional dialogues.Unlike in news reports, here we can find object encoding with someone whose semantic role is not that of patient, but rather that of experiencer, with verbal lexemes like gustar 'to please' or interesar 'to interest' (example 10).However, the number of object tokens is altogether very low.It is more frequent for the singular first person to appear as the clause subject, asserting ideas and expectations that should in turn be fulfilled by the advertiser (11).(11) frío / lluvia / viento nieve hielo / este invierno / quiero confiar en mi coche <Anu-On-141204-15:10> 'Cold, rain, wind, snow, ice...This winter I want to be able to trust my car'.
The three remaining genres (talk magazines and sports and music programs) are rather less scripted and more interactional than the previous ones.They share some orientation towards the adoption of subjective stances on the part of speakers, which results in an abundance of constructions with verbs of cognition and opinion encoding the human experiencer as object.However, there are striking differences between music programs and the other two genres.The former are singled out by the highest rate of object encoding (32.3%), clearly exceeding even that of news reports; they also have a relatively low frequency of speaker indexation altogether.The observation of examples suggests that broadcasters tend to take on a subordinate role in relation to their audience, as a means to obtain the latter's approval or cooperation.In these cases, the singular or plural second person will appear as clause subject, and the first person as object.The usual pragmatic functions are those of asking hearers to do something, like calling or emailing the radio station (example 12), or simply of requesting their attention (13).In the first example we can also observe how the syntactic structure is immediately reversed, placing the speaker in the position of the subject and the audience in that of the object.(12) tú me cantas una canción de Camela y: yo t-/ a cambio: / te doy / esa entrada: <Mus-Di-251104-13:15> 'You sing me a song by Camela, and I will give you the ticket in exchange'.
(13) ¡déjame pasar un ratito en tus oí:dos porfi:!<Mus-Ci-151204-13:10> "Let me stay for a little while in your ears, please!" Interestingly, the salience of the direct participants makes it possible for them to agree with the verb even when they are in principle not selected by the argument structure of verbs, and thus do not play a defined semantic role within the event.Such cases have been termed ethical datives (RAE 2009: §35.7).They can be considered discursive rather than syntactic choices in the sense that they do not construct syntactic functions proper, even if they adopt the form of functional agreement through clitics.The speaker can include him/herself within a clause through a (dative) clitic me; the fact that this is not a syntactic constituent is shown, for example, by the impossibility to formulate the coreferential phrase a mí.Neither irse 'to go away' in (14) nor madrugar 'to get up early' in (15) prefigure any central objects in their argument structure.In these cases, the motivation for self-indexation is clearly a discursive-pragmatic one and has important cognitive repercussions: it helps signal the speaker's involvement in an event where his/her presence would not a priori be expected.It thus endows the utterances with higher subjectivity.Across the corpus, the phenomenon is most often found in music programs and across the more interactional sequences; in both of the transcribed examples, the subject is a singular second-person one denoting any possible listener of the program.Finally, talk magazines and sports programs have the highest percentages of speaker self-encoding as subject.This is explained, first, by their stronger dialogic and argumentative nature.In these programs, a more marked tendency can be perceived on the part of speakers (especially all those not displaying a 'journalistic' identity; see Section 6 below) to talk on their own behalf and assume personal responsibility for what they say.This is why, even if we can again find constructions where a cognizer/experiencer is encoded as object, it is more usual to have these roles accorded to the subject, with verbs like creer 'to believe, to think' ( 16) and 'to think' ( 17).
( 16) yo creo que: la Unión Deportiva Salamanca necesita / además de dinero otras cosas / aportar idea:s aportar alguna solució:n / y <sic> intentar sobre todo: empujar todos pa-ra el mismo carro / y no cada uno pa uno porque eso yo creo que es malo <Dep-Co-080104-14:35> 'I think that the Unión Deportiva Salamanca needs many things besides money-it needs ideas, solutions, and above all it needs for us all to work hand in hand instead of having each one seek their own benefit, which I think is no good for the club'.
(17) pienso / que tienen que tener un poco de cuidado y la persona que sea responsable del Cuerpo de Bomberos / preparar a esas personas para los incidentes / no para que ellos preparen / más incidente <Var-Co-050204-13:05> 'I think they need to be more careful, and whoever is in charge of the Fire Brigade should train their staff to solve incidents, not to cause even more incidents than there were'.
As pointed out, variation between pronoun omission and expression is also a relevant matter of syntactic variation that cannot be fully addressed here.The stronger pragmatic weight entailed by overt subject pronouns (Davidson 1996, Stewart 2003), whose presence highlights the personal scope of the content, appears to be related to the higher rates of this syntactic function in genres where speakers are expected to convey their own personal views.
In spite of this all, alternation between subject and object encoding is also occasionally found across sports programs and talk magazines.In (18), a change of perspective is observed when the speaker, after expressing a personal stance as a subject, recounts some particular case where it seems more advantageous for her to appear as an experiencer or even a patient.( 18) yo también quiero de alguna forma animar a las mujeres anim:ar a las mujeres / a que / e: / ENtren dentro del mundo del trabajo / porque: / e: sí que se ha dado: alguna circustancia / que a m:í si quieres un poco: me ha dolido <Var-On-281204-13:20> 'Also, I want to somehow encourage women to enter the labor market, because there have indeed been some situations that, to put it this way, have hurt me myself'.
The preceding discussion has addressed the relationship between semantic, pragmatic and situational factors on one hand and the quantitative patterns of speaker self-indexation and syntactic encoding on the other across the five radio genres considered.The same kind of analysis will now be applied to the basic types of contextual identity displayed by participants in radio communication.

Speaker self-encoding and contextual identities
As already noted, speaker identities are often difficult to detach from the interactional contexts where they are enacted.A given textual genre will often predefine the kind of role each speaker is expected to play; this is quite obvious in a public domain such as radio communication.All the observations on genres made across the previous section are thus relevant to the understanding of variability across the speaker groups proposed.The statistical results are displayed on Table 3. Statistical tests indicate that the differences in subject vs. object encoding across speaker identities are again significant (χ 2 = 18.655; p < 0.01), and it must be observed that the normalized frequencies of self-indexation are even more disparate than those obtained in the comparison of textual genres.Despite all this, some figures in the table suggest a somewhat more complex situation according to the contextual factor under study.Again, the speakers with a stronger tendency to self-indexation in discourse-in this case public figures and anonymous individuals-also achieve higher percentages of subject encoding.However, journalists, with their comparatively very low rate of selfindexation, are not the group less inclined to encode themselves as subjects; this place corresponds to politicians, whose results can indeed be surprising and will need some more detailed qualitative examination.
In the preceding section we explained the high percentages of first-person subjects in talk magazines and sports programs, as well as their considerable rates of first-person indexation altogether, as primarily reflecting their interactional, sometimes contentious nature.It must now be added that these are the genres where the participation of public figures and anonymous individuals is more usual.People assuming such kinds of contextual identity are generally expected to express their own stances on a given topic.There is anyway a 6point difference in subject-encoding percentage between both groups, which suggests that public figures are the ones most clearly attributed the right-or, from a different perspective, imposed the duty-to speak for themselves in radio interactions.Across the following stretch we can see how a public figure-a fairly well-known local sports entrepreneur-stresses his own commitment to his club: (19) voy a poner toda la carne en el asador / porque el tema tire para adelante / voy a decir lo que yo pienso / y: le guste o no le guste a loh compañeros que estén allí yo: / quiero decirte / que por encima de todo está el Salamanca: y que si yo: lo que digo no le gusta y me tengo que ir / con to el dolor de mi corazón: me iré pa mi casa tranquilamente pero si estoy allí / voy a aportar y a decir lo que yo pienso <Dep-Co-080104-14:35> 'I am going to do my utmost effort to get things going.I'm going to say what I think.Whether the folks over there like it or not, I want to tell you that for me the Salamanca club is above all things.If they don't like what I have to say and I should be forced to leave, well, I will go home with great sorrow and that's all.However, if I remain in the team, I'm going to make my contribution and say what I think'.
By contrast, anonymous individuals are not attributed a particular professional status in the context; they only take part in media interactions incidentally and most often on their own initiative, by calling radio stations in order to come on air and report or discuss some fact.Lack of self-assurance in public situations can sometimes make it advantageous for these speakers to reduce their degree of personal involvement by encoding themselves as objects.Also, they are often not expected to offer anything on their part, but rather to ask things from others (e.g. the radio station, the authorities).In (20), the speaker expresses her stance through psychological verbs selecting the experiencer as object, such as gustar 'to like, lit. to please' and preocupar 'to worry', even if there are also some cases of subject encoding-which is anyway the choice clearly preferred by this group, as in the case of public figures.
(20) a mí me gustaba mucho el edificio / TOdo lo que-/ pero / yo prefiero las trenta y tantas o <sic>los</sic> / cuarenta y tantas familias / que no lo sé esactamente / que se han ido a la calle / que tienen trenta y tantos años y cuarenta y tantos años / qu(e) han estao ta-/ trabajando toda una vida ahí / y que no han con:tao / con ellos / para nada / los han echao a la calle indignamente / indignamente y sin ningún motivo / eso es lo que me preocupa a mí: realmente <Var-SE-211204-13:50> 'That building pleased me, of course, but I indeed prefer the thirty-something or forty-something families-I don't exactly know-, thirty-something or forty-something year-old people that have been given the sack after working there all their lives.They haven't been taken into account at all.They've been thrown away and deprived of their dignity for no reason.That is what really worries me'.
As also shown in Table 3, it is journalists and politicians that achieve percentages of object encoding above the average.The explanation seems to be more straightforward in the first case than in the second one.Journalists are by far the group with the fewest self-indexations across discourse (72.9 per 10,000 words); their usual dedication to informative rather than argumentative tasks, as well as their tendency to display an image of neutrality, result in the avoidance of the personal involvement suggested by self-indexation in most interactional contexts (Patrona 2011: 159).When they do index themselves, they do it as objects in slightly more than 20% of the cases.One of the areas where object encoding is recurrent is the expression of epistemic modality.The construction (a mí) me parece 'it seems to me' ( 21) is often used to modalize a personal opinion instead of other possible ones encoding the speaker as subject.(21) un centro del campo / que: recuPE:ra / que trabaja y que luego tuvo capacidá para distribuir / caso / de Royerio que a mí me parece que hizo un partidazo <Dep-Co-221104-14:35> 'The midfield players get many balls back, they work hard and they have the ability to organize the play.That's the case with Rogério, who, it seems to me, played a great game'.
As already pointed out when discussing music programs in Section 5, another important source of object self-encoding for broadcasters is found in contexts where they request something from another participant.This often consists in just an encouragement of their intervention through imperative forms such as dime and dígame 'tell me', which due to their highly grammaticalized status do not run the risk of being interpreted as impositions, but rather as phatic devices.In ( 22) we can contrast the way of self-encoding chosen by the anonymous caller with the one adopted by the broadcaster: the former is expected to express his/her own view, the latter to accept it.Finally, as pointed out, the behavior of politicians is apparently more difficult to assess.Having a relatively high frequency of discursive selfindexation (134.9 tokens per 10,000 words), they are nonetheless the group with the lowest percentage of subject self-encoding.The participation of these speakers in radio broadcasts is usually restricted to talk magazines and to recorded clips embedded in news reports.Obviously, they are normally requested to express their views on controversial matters.Whereas in many cases they resort to a first-person plural, indexing either their political team or the entire community they intend to represent (Bull & Fetzer 2006;Aijón Oliva 2013: 591-592), it seems almost inevitable, especially in the course of spoken interaction, that they should often end up by adopting a singular first-person viewpoint.In such cases, they show some tendency to encode themselves as objects when exposing personal stances, as in ( 23), where the speaker, after conveying a negative opinion with yo creo 'I think', presents his alternative proposal-which we can suspect is equally clearcut to him-as more of a suggestion with me gustaría 'I'd like, lit. it would please me'.(23) no sé: / pero yo lo:-/ la situación de Salamanca / yo creo que se está luchando demasiado por los archivos / o sea a mí me gustaría que se luchara igual / por los jóvenes que tienen que emigrar fuera / y no por unos papeles <Var-Pu-281204-12:40> 'I don't know, but as for the current situation of Salamanca, I think there is too much hype about the [Civil War] archives.I mean, I'd like [lit. it would please me] to see so much effort spent on all the young people who are being forced to emigrate, instead of fighting for just some paper sheets'.
The well-known decline in public trust in politicians across Western societies has resulted in an increase-as well as in some social acceptation-of aggressive, insulting attitudes towards this group on the part of audiences and even on that of journalists (Kampf & Daskal 2011).This, in turn, might well be promoting some orientation of politicians towards less imposing and more selfeffacing strategies, manifested in our case through higher rates of object self-encoding, and more evidently, as pointed out, in their preference for the plural first person, most often denoting the party or association they belong to. 7In short, they show some tendency to avoid responsibility for their own words, so that they cannot be held accountable for what they say neither by the audience nor by their adversaries (see also example 27 below).
Anyway, subject encoding is still recurrent in these speakers whenever a strong claim is being made, as in (24); the speaker also draws a sharp contrast between yo and the immediately preceding second-person subject usted.The case of ( 25) is also an interesting one: the same speaker uses two nearlysynonymous constructions within a relatively short stretch, where he respectively appears as the object and the (reflexive) subject of the verb sorprender 'to surprise '. 8   (24)  en el caso: / al que usté se está refiriendo / yo le tengo que señala:r que para el: equipo de gobierno municipal / nos parece absolutamente justificada la lectura de la Sala de lo Social-/ de la sentencia de la Sala de lo Social del Tribunal Superior de Justicia <Inf-Pu-021204-13:55> 'In the case you are referring to, I must point it out to you that the city council sees the ruling of the Social Section of the Higher Court of Justice as absolutely justified'.
[…] I am growing really surprised with such proposals'.
Many other contexts could further illustrate the potential of subject vs. object encoding for the establishment and progressive management of speaker identities.In the examples below we can compare two constructions that are often used for the expression of desires, as well as for the modalization of deontic speech acts: quiero 'I want' and the already mentioned me gustaría 'I'd like, lit. it would please me'.Whereas the first one selects the human experiencer as the subject, the second one downplays his/her own responsibility through object self-encoding.Neat differences are also established by further choices such as the verbal tense: present vs. conditional, the latter being a wellknown resource for modalization (RAE 2009: §23.15), in this case presenting the desire as much more hypothetical.In general, public figures that have been expressly requested to participate will feel entitled to use forms like quiero, particularly if such choices can help reinforce a personal commitment that will enhance their professional image (example 26; see also 19 above).On the other hand, in (27) we have a local politician that has recently come into office.When asked about his immediate projects, he repeatedly uses me gustaría, 7 Interestingly, politicians produce 407 tokens of plural first-person indexations across the corpus, of which no less than 362 (88.9%) are of subject encoding, in marked contrast to their behavior in singular contexts.It would no doubt be of great interest to further analyze the choice among grammatical persons within each discursive genre and contextual identity. 8Reflexivity, with its internalization of events within the subject (García-Miguel 2003: 74), results in the latter presenting a functionally and cognitively intermediate status between agentive subjects and objects.It is a relevant syntactic strategy in Spanish that clearly merits further attention than it can be accorded here.
suggesting an intention to avoid formal commitment.
(26) quiero amp-/ amplia:r / e: dar mejor servicio todavía de lo que lo hacemos que lo hacemos bastante bien / al tema de la celuliti:s <Var-On-080104-13:35> 'I want to expand [the clinic] and offer an even better service than we already do-and it is a fairly good one-to people suffering from cellulitis.' ( 27) me gustaría que estuviera redactado / para ir sometiéndolo a todo: estudio de todos aquellos que tengan que ver / dos reglamentos de cómo deben / circuLAR / todos los documentos <...> la cartelería de los bajos comerciales / en donde me gustaría / llevar a cabo / una reglamentación / que ya existe / y si / pudiera ser / financiar en: parte / esa reforma de la cartelería <Var-On-080104-13:10> 'It would please me to have the drafts of two sets of regulations establishing the way documents should circulate, in order to submit them to the scrutiny of all those involved.As for the signage of ground-floor shops, it would please me to be able to enforce the already existing regulation and, if possible, to partially finance the reform of signs.'

Discussion: Subjectivity vs. objectivity in speaker self-profiling
The preceding analysis makes it possible to infer that speakers can follow two basic directions in self-profiling, which we will explain as being respectively associated with the discursive-cognitive notions of subjectivity and objectivity.On the one hand, speakers can tend to highlight their own presence through recurrent self-indexation and self-encoding as clause subjects, turning themselves into the main participants of events and thus reinforcing the argumentative orientation of discourse.Subjectivity entails the tendency to construct and interpret discourse as focused on the direct participants, and particularly the speaker (Hopper & Traugott 2003: 3;Croft & Cruse 2004: 62-63).
On the other hand, the suppression of indexical traces of the direct participants tends to endow discourse with a higher degree of objectivity (Albentosa & Moya 2000;Farrar & Jones 2002: 6).When speakers do index themselves, they can nonetheless reduce their own salience by choosing an objectual role, thus situating discourse at some intermediate point along the continuum.Even with syntactic constructions that do not contemplate the presence of an object, there is often the possibility for speakers to add a first-person clitic-a so-called ethical dative-in order to signal self-involvement (see examples 14 and 15 above), this being a typical resource of conversational discourse.
Each strategy will be felt as more or less appropriate for the display of a particular kind of identity according to the context.We have found that it is journalists that more assiduously turn to objectivizing choices, particularly in their relative tendency to avoid self-indexation altogether, while public figures as well as anonymous individuals often highlight their own responsibility for the content.Politicians are a particularly interesting group in that they do not display the usual correlation between frequency of self-indexation and frequency of subject self-encoding.This makes it possible to suspect that different linguistic features can help enhance different facets of subjectivity, and shows that quantification is scarcely useful without qualitative interpretation.The typical speech of politicians is obviously quite 'personal' and argumentative, which results in frequent self-indexation; however, they often opt for encoding themselves as objects, thus reducing their own liability for what they say.This way, a personal stance (e.g. one of curiosity or surprise, as in the following example) can be interpreted as more of a perception of an external fact.(28) por las llamadas que hemos tenido / por la: atención que han prestado los medios de comunicación a esta campaña / tanto / que a mí incluso como concejal: me ha llamado un poco la atención <Var-On-281204-13:15> 'So many calls we have received, so much attention has been paid by the media to this campaign that, well, it has even been striking to me, as a city councilor'.
The tendency towards subjectivity vs. objectivity will also be conditioned by the interactional demands of the communicative channel and the textual genre.Oral communication has been shown to be altogether more subjective than writing, this difference being manifested in a variety of linguistic traits, among them the very frequency of participant indexations (see e.g.Vázquez Rozas & García-Miguel 2006).In the case of our study, it has been observed that the more conversational and less pre-planned radio genres, namely talk magazines and sports programs, are in fact those with higher rates of the first person in central syntactic functions.However, we have found that genres also significantly differ as regards the tendency of speakers to encode themselves as subjects, which represents a further step towards subjectivity.Constructions encoding a human cognizer/experiencer as an object-with gustar, parecer, etc.-help downplay subjectivity, since they situate the speaker in a less salient syntactic position.However, even in these contexts other features of salience prototypically associated with subjects generally remain intact (e.g.preverbal placement as well as agreement with the verb, in this case through clitics), all of which underlines the fact that these are hardly prototypical syntactic objects.In (29), the object a mí is a clearly topical one, appearing even before the thirdperson resumptive pronoun that acts as the subject.
(29) una canción que nos llama la atención: de un:-/ de un anuncio vamoh corriendo a asomarnos a ver qué:-/ de qué:-/ de qué va el anuncio ¿no? / y a mí eso / casualmente me ocurre cuando: / e: aparece el:-el: anuncio de los coches <Mus-Ci-230903-17:10> 'A song in a commercial calls our attention, and we rush to find out what the commercial is about, right?And to me this happens, as a matter of fact, whenever that commercial with the cars comes up'.
In sum, discursive self-indexation and subject vs. object self-encoding are just two among a wide array of choices whereby speakers carry out subtle modulations in order to situate their discourse, together with their own contextual identity, at some point along the discursive-cognitive continuum between subjectivity and objectivity.Also, it should be evident that the model proposed here only offers a partial view of the complexity of participant profiling and discourse configuration.Even if subjectivity and objectivity can be considered the basic poles of the continuum, further analysis should also take into account the intermediate dimension of intersubjectivity.The latter should prove particularly relevant for the explanation of how the plural first person-denoting any possible group of beings the speaker constructs him/herself as part of-as well as the singular and plural second personsdenoting any possible addressee or audience-are indexed in discourse and constructed in cognition (Aijón Oliva 2013: 586-587;Ädel 2014).

Conclusions and prospects
The present study has analyzed speaker self-indexation and syntactic selfencoding within the clause as resources for the configuration of discourse and the management of contextual identities in radio communication; in other words, as strategies for interactional profiling.The inherent meanings of discursive and syntactic choices-the power of self-indexation to signal the speaker's participation in an event; the higher salience entailed by subject encoding as against object encoding-are undetachable from their quantitative usage patterns and their potential to generate particular pragmatic effects in communicative situations.It has been shown that different textual genres and speaker contextual identities do not follow identical patterns regarding the choices considered and the meanings attached to them, and that this is explainable through qualitative observation.
However, it is obvious that further attention should be paid to particular constructions, in order to fully understand the cognitive motivations and pragmatic repercussions of their use.Even if the differences between epistemic modal constructions such as creo 'I think' and me parece 'It seems to me' may be subtle, they can hardly be considered equivalent, and speakers in radio communication do not randomly choose between them.The reason is partly that constructions selecting the speaker as subject entail a higher degree of responsibility on his/her part, which may be perceived as adequate or not depending on the set of rights and obligations derived from the kind of identity assumed.As pointed out, reflexive or 'middle' constructions could be approached as yet another syntactic choice with its own discursive and cognitive implications.
From the point of view of linguistic phenomena, it will also be necessary to more specifically address the variable expression and placement of both subject and object pronouns and lexical NPs (for the specific case of subject pronouns, see Aijón Oliva & Serrano 2013 and references therein) and to consider other syntactic functions, namely peripheral or non-agreeing objects, as well as deictic-anaphoric elements such as possessives.It is only this way that it will be possible to accomplish a complete picture of participant interactional profiling and its cognitive correlates.
Perhaps more importantly, it must be recognized that the consideration of speaker self-encoding within the clause and across discourse only provides a partial view of the complex ways in which grammar interacts with discourse and cognition.The analysis should be extended to the distribution of central and peripheral syntactic functions among the speaker, the audience and other possible referents.Speakers, by choosing to encode themselves as subjects and their addressees as objects or vice versa, will be managing interactional relationships and (im)politeness (see some examples in De Cock 2014: 75-76, 148-149).The fact that syntactic choice can have repercussions at so many linguistic and extralinguistic levels of meaning offers a wide and promising field for future research.
SING.CL seem 3RD.SING.PRES that be 3RD.SING.PRES well 'It seems to me it's OK'.

Table 1 .
General data

Table 2 .
Speaker self-encoding across radio genres

Table 3 .
Speaker self-encoding and socio-professional identities