1. Introduction

In Spanish, and seemingly universally in human language, there exist certain particles that fulfill various functions of guiding discourse that are, accordingly, called discourse markers (Martín Zorraquino & Portolés, 1999; Portolés, 2001). These markers carry out various roles such as reformulation (explication and correction), mitigation, and augmentation of the discourse, among many others. Examples (1-4) give instances of these uses.

(1) Reformulation – explication
Ella es obstetra, es decir, médica para mujeres embarazadas.
‘She is an obstetrician, es decir, a doctor for pregnant women.’

(2) Reformulation – correction
No comí el chocolate… más bien, tomé dos o tres pedazos…
‘I did not eat the chocolate… más bien, I took two or three pieces…’

(3) Mitigation
A: Cómo te gustó la película?
B: Pues, no sé, no fue mi favorita.
‘A: How did you like the movie?
B: Pues, no sé, it wasn’t my favorite.’

(4) Augmentation
Limpié mi dormitorio. Además, saqué la basura.
‘I cleaned my room. Además, I took out the trash.’

Both (1) and (2) are reformulative markers, but with different functions. In explicative contexts as in (1), the information before the marker is equivalent to what follows it, simply restated in other turns. This contrasts with the corrective function in (2) in which the information following the marker contradicts what precedes it. Markers with a mitigating function, such as those in (3), reflect the desire of the speaker to reduce the impact of the statement by not committing to it completely. Finally, augmentative markers such as in (4) reflect the speaker’s desire to add to the previously expressed information.

Even though most discourse markers (alternatively DMs) are commonly composed of words that exist elsewhere in the language, they do not carry the same lexical-semantic weight in these contexts. That is to say, for example, that the word bueno used as a discursive marker does not convey the adjectival meaning of ‘good’, e.g. el jamón está bueno ‘the ham is good’. Rather, a speaker would use bueno to indicate agreement or acknowledgment of the statement made by their interlocutor – in this case, bueno is understood as a communicative device and not an adjective, and the pragmatic weight in the discourse is the acknowledgment of the interlocutor by the speaker. Such is the case with discourse markers evolved from lexical words: their semantic meaning is largely bleached, but their communicative function – pragmatic weight, in other words – is still readily understandable.

Less commonly studied are hesitation markers, a subset of discourse markers whose primary role is to affect the structure of the conversation instead of the content. Canonically, the speaker that employs a hesitation marker (alternatively HM, in this paper) during their turn communicates intent to pause, hesitate, or reconsider for any reason. Consider the following examples (5-6) in which speakers use two different markers to carry out the same discursive function:

(5) I: también participé y/ recibimos golpes / eh/ hemos recibido de todo/ humillaciones/ toletazos […]
‘I: I also participated and/ we received strikes/ eh/ we received all types of/ humiliation/ strikes with clubs […]’
(CSCM,[1] interview 72)

(6) Inf. B.- Pues sí. ¿Ps para qué vamos a pedir más de lo que no tenemos… de querer… este… de lo que no hay?
‘Inf. B. – Well, yes. Well, for what reason are we going to ask for what we don’t have… to want… este… what isn’t there?’
(HP, interview 20)

In example (5), we see that the particle eh carries no semantic information; its function is purely pragmatic. The speaker uses eh to indicate that, even though there is a pause or break for whatever reason, her turn is not over and that she wishes to continue. The hearer perceives eh within the speaker’s turn and intuitively knows to wait. The marker este in example (6) serves the same purpose as eh in (5). There are other discourse particles that have this same function such as mm and ah; moreover, other markers that have different main functions such as pues ‘well’, bueno ‘good’, and o sea ‘that is’, are also employed as HMs in order to fill pauses and maintain turns. These functions of pausing and mediating turns are essential to the structure and coherence of discourse (Blas Arroyo, 1995; Briz Gómez, 1993; Swerts, 1998).

Hesitation markers tend to be stigmatized as signals of poor speaking skills. Through the years, these markers have been labeled as muletillas ‘crutches’ (Christl, 1996), bordones/latiguillos ‘clichés’ (Stoesslein, 2014), or disfluencias ‘disfluencies’, each one of which carries a negative connotation. This negative impression of said markers, of being empty or unnecessary elements, arises from schoolteachers who are of the opinion that their use is a manifestation of poor management or command of the language (Soler Arechalde, 2006).

(7) E: (risa)/ pero no usas muchas muletillas fíjate
I: ¿no?/ ¿qué muletillas?
P: bueno como ‘y y y’
E: o el [‘tons’]
I: [este]
E: este
I: este
E: [bueno]
I: [no uso] el ‘este’/ [ese sí]
E: [¿no verdad?]
I: te lo corriges dando clase/ o sea no puedes
E: ándale
I: estar con ‘este este’/ eso sí lo corriges
‘E: (laughter)/ but you don’t use many crutches look
I: no?/ what crutches?
P: well, like ‘y y y
E: or [‘tons’]
I: [este]
E: este
I: este
E: [bueno]
I: [I don’t say] 'este’/ [that yes]
E: [no, really?]
I: you correct that out when teaching/ that is, you can’t
E: ándale
I: be there with ‘este este’/ yes you correct that out’
(CSCM, entrevista 35)

In example (7) here, we see a teacher who admits to avoiding este, for example, while she teaches. It is understood from her tone that these types of markers are inappropriate in her role as an educator. This stigma carried by HMs has caused them to be understudied in linguistic research in favor of other markers such as the aforementioned pues and o sea, as well as claro ‘of course (lit. ‘clear’)’, entonces ‘so, then’, etc., whose semantic import to the discourse is more readily apparent. But the truth about HMs is that, even though they communicate no semantic information whatsoever, their pragmatic and discursive importance is immeasurable, especially in terms of discursive cooperation.

2. Hesitation markers: overview and previous studies

Even as HMs are now beginning to be more frequently studied for their importance to the fields of sociolinguistics and pragmatics, there has traditionally been a lack of attention paid to these particles. For example, Blas Arroyo (1995) and Soler Arechalde (2006) emphasize the lack of research attention paid to the marker, noting that it has traditionally been considered empty language and “evidence of a poor and inadequate grasp of the oral language” (Soler Arechalde, 2006). Furthermore, she notes that she is not completely in agreement with those accounts, among them Martín Zorraquino and Portolés (1999), that claim that the only use of HMs is of hesitation or turn-keeping. In this section, I detail some of the most relevant previous studies to HM use, function, and behavior.

2.1. Hesitation markers este and eh: uses and behaviors

Even though the discursive function of hesitation is universal, the world’s languages have their own distinctly codified markers to serve this function (Blondet, 2001). Speakers of American English most frequently use um [ɐːm] and uh [əː ~ ɯː] (see Clark & Fox Tree, 2002 for their analysis of these markers). French speakers tend to use the vowels [əː] and [œː] to indicate hesitation. More pertinently, Spanish speakers prefer sounds such as eh [eː], em [eːm], and este [esteː ~ ehteː] as manifestations of discursive hesitancy (Erker & Bruso, 2017). The interesting aspect of hesitation is the frequent use of neutral vowels in these markers and, additionally, the tendency to elongate the marker with level or falling intonation (Martínez et al., 2004), which communicates to the hearer that the speaker intends to maintain his or her turn after the pause.

2.1.1. Eh as a marker

Blas Arroyo (1995) shows the uses of eh as a functional marker which is integral to inter-speaker communication. In his chapter he focuses on two distinct functions of eh as a marker: (1) as an interjection at the end of an utterance in the form of a “tag question”, and (2) as a facilitator of cohesion within the turn as a signal of reinforcement, of thematic contrast, or as a guide to the hearer’s understanding. Blas Arroyo observes that it is mistaken to claim that the use of eh is only associated with speakers of low education, given that the diversity in use of the marker extends to a variety of communicative contests, as well as social situations. He proposes at the end that the frequency of use of eh as a marker varies inversely as the familiarity of the interlocutors; that is, that usage of eh decreases as speaker familiarity increases.

Continuing with the use of eh as a marker, Roggia (2012) investigates the frequency and functions of this marker in the speech of the Dominican Republic. He describes eh as invariable, extrasyntactic, and conversational, as is the case of other DMs. Roggia points out that eh is a polyfunctional marker: " […] as prototypically having a hesitation function and peripherally having a reformulation function, functions shared by other information structuring and reformulating DMs" (2012, p. 1796). While in agreement with Blas Arroyo concerning the discursive utility of eh, Roggia also finds that the frequency of eh in Dominican Spanish is sensitive to social factors such as sex, social class, and age.

2.1.2. Este as a marker

Also relevant to the current study is the HM este with has also gone understudied in traditional scholarship. Reyes Trigos (2002) describes the presence and the sociopragmatic uses of este in narratives. The marker lacks semantic import; ostensibly it was discursivized[2] from the demonstrative este ‘this’[3] and now serves no referential purpose whatsoever. Nevertheless, it figures prominently in the development and structuring of discourse. Reyes Trigos mentions the recursivity of the marker, meaning that it can appear both at the onset and in the development of the narrative. Where este occurs in a story, it signals the beginning of the next stage of the narrative, such as a new discursive frame or theme, the building of said frame, and the beginning of the end of the story. The most interesting aspect of Reyes Trigos’ work is that she sets aside the metadiscursive hesitation function of este in favor of its discourse-structuring functions.

In a comparative study of discourse markers in the Spanish of Caracas, Galué (2002) compares the relative frequencies of various markers such as claro, la verdad ‘the truth (is)’, mira ‘look’, fíjate ‘check it out’, okey, etc. alongside este. She classifies este as a conversational metadiscursive marker, indicating the taking or keeping of the turn by the speaker. During each turn-keeping instance, the speaker processes information before speaking; this use concurs with Roggia’s (2012) account of the pragmatic functions of eh, indicating that these markers are capable of sharing the same discursive space.

2.2. Structural and social factors conditioning HM production

The primary studies of discursive and sociopragmatic variation in HM production upon which the current study is based are those of Soler Arechalde (2006, 2012), Soler Arechalde and Serrano (2010), and Graham (2013, 2018). Toward the question of sociopragmatic variation between specific HMs, Graham (2013) conducts a preliminary study of the use of such markers in the speech of San Juan (Puerto Rico). More recently, Graham (2018) conducted a study of the differences of use of este and eh within four metropolitan areas of Latin America, including Mexico City.[4] The commonality among the studies by these authors is the focus paid specifically to Mexico City based on the social and discursive factors listed here.

2.2.1. Position of the marker within the turn

Soler Arechalde (2006, 2012) introduces other functions of este – which in Graham (2013) are almost equally applicable to both este and eh – that are recognizable depending on the position of the marker within the turn. The formation function is ascribed to the marker in turn-initial position, reformulation in turn-internal position, and topic-closing in turn-final position. Focusing on the prototypical function of eh, Roggia finds the same: that the nature of the hesitation changes depending on the position of the marker within the turn or utterance. If it occurs at the beginning, it indicates cognitive organization of the discourse on the part of the speaker. Within the turn, it signals to the hearer that the speaker wishes to maintain his or her turn while he or she is organizing thoughts, a characteristic shared with other markers such as pues or o sea (Cortés Rodríguez, 1991; Félix-Brasdefer, 2006; Schwenter, 1996). When comparing eh and este production based on utterance position, Graham (2018) finds that speakers across the study tend to prefer using este over eh in turn-medial position.

2.2.2. Sex

In her more recent study of contextual uses of este, Soler Arechalde (2012) contrasts the relative frequencies of este according to various social and discursive variables. Working solely with a small set of interviews from the HC corpus,[5] Soler Arechalde researches the difference in prevalence of este based on the communicative situation, the sex of the participants, the relationship between the speaker and other participants, the topic of discussion, and the time at which the discourse occurs. Regarding personal differences, she finds that women tend to avoid este in contexts of greater formality, showing consistency with her earlier study (Soler Arechalde, 2006). It was found in Graham (2013, 2018) that women tended toward este more frequently than men, irrespective of other factors.

2.2.3. Register, formality, or educational attainment

In searching for details of social stratification in the use of the marker, Soler Arechalde finds that the vast majority of instances of este occur in popular speech versus in higher registers. She observes that discursive contexts – i.e. the type of discourse being examined, such as spontaneous conversation versus guided interviews – condition the occurrence of este more frequently than social differences between participants.[6] Notable about Soler Arechalde’s results is the intersection of sex and register: men tend to predominate in the frequency of este in higher registers, while women favor it more in popular speech. This reflects an observation by Silva Corvalán (2001) that women tend to be more conscientious of their language in formal situations, preferring to avoid less prestigious forms such as HMs (and este in particular). This intersection was not analyzed in Graham (2013, 2018); however, the latter study did concur with Soler Arechalde in that that este was strongly associated with less education on the part of the speaker.

2.3. The current study: hypotheses

As a continuation of the study of social effects on the choice of HMs by Mexico City speakers, the current study takes advantage of the availability of sociolinguistic corpora from the past and the present. The goal of this type of study is to identify trends in discourse marker usage in two distinct time periods, and, furthermore, to discover how much the use of said markers has changed across eras. This study continues with the factors under consideration in Graham (2013, 2018) of age, sex, and educational attainment of the speakers; a new factor of time period (past versus present) is introduced, within which the social factors will also be analyzed.

As such, the hypotheses of the current study are as follow:

  1. Este will be the most frequently produced HM across the study.

  2. Marker production will differ proportionally across the two time periods under study, causing time period to exert a significant influence, though passive, on marker choice.

  3. Sex will be a significant factor in HM choice. Following the studies by Soler Arechalde (2006, 2012), Soler Arechalde and Serrano (2010), and Graham (2018), women will show the greater preference for este than men. These proportions will differ in the two time periods under study.

  4. Educational attainment will be significant as a factor in HM production. Speakers of lower formal education will produce more such markers than more highly-educated speakers, and, furthermore, they will show a stronger tendency toward este as a marker. These tendencies will not differ in strength between the past and the present.

  5. Age will be a significant factor in HM choice. Younger speakers will be the most likely population to favor este as a marker. This tendency will not differ between time periods.

As a departure from Graham (2013, 2018), this study focuses solely on societal/social factors determining variation in marker selection, keeping the tradition of the previous work on este by Soler Arechalde (2006, 2012). These previous works, specifically the 2006 study, found that este was overwhelmingly encountered within a speaker’s turn, much more frequently than in turn-initial or turn-final position. This distribution coincides with the finding for Mexico City speakers in Graham (2018), despite the time elapsed between the publishing of each corpus. At this juncture, it is not the expectation that the discursive/pragmatic functions of este would have evolved over time in a manner observable in the corpus data, and therefore the factor of position within turn will not be considered. The analysis of this factor may be considered in future studies.

3. Methodology of the current study

3.1. Corpora

In the current study, the data have been compiled from two different corpora. To analyze the speech of the past, the corpus El habla de la ciudad de México (Lope Blanch, 1971, 1976; Serrano Morales, 2014) is consulted. These three sections of the corpus consist of sociolinguistic interviews conducted between 1967 and 1975. Each interview was tagged according to the speaker’s characteristics, specifically sex, age, and educational level. With the division between the learnèd norm (Habla culta; Lope Blanch, 1971; Serrano Morales, 2014) and popular speech (Lope Blanch, 1976), this facilitates the distinction between highly- and lower-educated speakers. All the speaker data available from this corpus are utilized, which amount to 37 interviews from Habla popular (HP) and 56 from the two sections of Habla culta (HC). Because of this division between HC and HP, we can distinguish between the sociolect of the higher-educated speakers and that of the lower-educated population.

The analysis of more modern speech incorporates a section of the Corpus sociolingüístico de la Ciudad de México (CSCM, Martín Butragueño & Lastra, 2011–2015), which was performed as a component of the Proyecto para el estudio sociolingüístico del español de España y de América (PRESEEA). This corpus consists of interviews conducted between 1997 and 2007. The section of CSCM as a PRESEEA component contains 108 transcriptions, and each interview is tagged according to sex, generation (age), and educational level.

The principal benefit of these corpora is that, as they contain decidedly sociolinguistic interviews, each and every phonetic/phonological production by the participants is registered in the transcriptions. That is, instead of ignoring sounds that are not purely lexical, such as for example metadiscursive markers such as mm, ah, eh, etc., these productions are faithfully transcribed. In the CSCM, this information is shown even more explicitly by the inclusion of phonetic variations such as lengthened vowels. This inclusion makes sociopragmatic study of HMs possible, contrasted with literary works within which these markers never appear.

3.2. Token mining and the dependent variable

Because the transcribed interviews are in electronic formats (either plain text files or PDFs), it would be more efficient to use computer software to carry out each phase of the project. The token search is conducted using AntConc (Anthony, 2015), which contains tools to delimit searches in order to return all the desired occurrences. Since this study aims to compare the effects of social factors on HM production, it makes sense that the dependent variable should be the speaker’s choice of este or eh.

In order to get every possible occurrence of the markers in questions, including phonetic productions with elongated vowels (such as esteeee) and intrusive final consonants (such as ehm), it is necessary to use specific search strings in AntConc. The strings are composed accordingly:

  • este* (for all predictable instances of este)

  • *eh* (for all predictable instances of eh)

The asterisks function as wildcards to capture any variant of the marker as transcribed in the interviews, including lengthening (eeh, ehhh, esteee, etc.). Obviously, the wildcard causes other words to appear in the results. Therefore, before the next step in the data analysis, it is necessary to exclude various sequences:

  • Demonstrative este (este país ‘this country’)

  • Tag question eh (¿… eh?)

  • Repetitions of the same marker (eh eh eh or este este este) after the first in the sequence[7]

  • Markers produced by non-participants, e.g. by the interviewers or other bystanders

After removing invalid hits, there were 13,151 usable instances of este and eh to be analyzed.

In the process of coding tokens for variable rules analysis (VARBRUL), it is necessary to mark the dependent variable and each independent factor with a lone character associated with a distinct category. The syntax of the program used to conduct the VARBRUL recognizes each place in the token string as a factor and each character as a distinct factor level, and the strings are by default case-sensitive. Each instance of este was coded as ‘E’ and each instance of eh was coded as ‘e’.

3.3. Independent factors and their codes

In the following subsections, the distinct independent factors and their values are detailed. The significance level of each factor will be measured, both on an overall basis and within each particular time period. It bears mentioning that each interview from CSCM involves only one participant (besides the interviewer), and so the coding is straightforward. Many of the HC/HP transcriptions also involve a single informant, though in contrast with CSCM, there are a number of interviews involving two informants. In order to correctly code each marker occurrence in HC/HP, it was necessary to pay closer attention to who was speaking at each moment, particularly in the interviews that involved speakers of either different ages, sexes, or educational levels.

3.3.1. Time period

In general, the most important independent factor in this study is whether the marker in question was produced in the older interviews – from HC/HP – or in the more recent ones – from CSCM.

(8) I: hace mucho que no voy a la facultad/ este/ te recargabas ahí/ te/ te acostabas en el pasto/
‘I: it’s been a long time since I’ve gone to the building/ este/ you could recharge there/ you/ you could lie back on the grass/’
(CSCM, interview 24, R)

(9) Inf. A. -y ¿qué es, precisamente, la actual… este… juventud? Es una rebeldía hacia esa situación de… de guerras nada más. Realmente eso es.
‘Inf. A. -and what is, precisely, the current… este… youth? It is a rebellion toward this situation of… of wars, that’s all. Really that’s what it is.’
(HC, interview 21, P)

Instances from the past are coded as ‘P’, while instances from the present are coded as ‘R’.

3.3.2. Age/generation

Following the pattern of previous studies (Graham, 2013, 2018; Soler Arechalde, 2006), the participant’s age is one of the factors under consideration. The interviews from CSCM appear with age already tagged in the title of each transcription. Following the PRESEEA coding system, participants are identified according to generation instead of exact age: speakers from 18-34 are of the younger generation ‘1’, those aged 35-54 are of the middle-age or “adult” generation ‘2’, and those 55 and older are of the “elder” generation ‘3’.

(10) Enc.- ¿Dónde es la Liga?
Inf. B.- La Liga está en Congreso número treinta, aquí, en Tlalpan.
Enc.- Aquí en Tlalpan.
Inf B.- Sí. Y este… los martes es la junta. La junta consiste en… en recordar lo… este… acontecido en la… en la junta anterior.
‘Int.- Where is the Liga?
Inf. B.- The Liga is on Congress number 30, here in Tlalpan.
Int.- Here in Tlalpan.
Inf. B.- Yes. And este… the meeting is on Tuesdays. The meeting consists of… of recording the… este… business of the… of the previous meeting.’
(HP, interview 16 – younger generation 1)

(11) Enc.- Pero se amparaban con la pluma.
Inf.- Con la pluma, sí; escribían… preciosamente los dos. Fueron grandes. Del padre Méndez… este… ¿cómo se llama?… Gabriel, yo recuerdo el elogio fúnebre que le hizo monseñor Octaviano Valdés… este… en el seminario conciliar de México, en sus honras fúnebres.
‘Int.- But they exercised their rights with the pen.
Inf.- With the pen, yes; they wrote… preciously, the two of them. They were great. From Father Méndez… este… what’s his name?… Gabriel, I remember the somber eulogy that Monsignor Octaviano Valdés delivered of him… este… in the council seminar of Mexico, in his funerary honors.’
(HC, interview 4 – middle-age generation 2)

(12) I: porque íbamos a llevar/ eh/ al santísimo que son este/ las hostias consagradas
E: mm
I: y bueno pues para nosotros es muy importante y/ mm// muy importante/ ¿no?/ este/ lo del santísimo/ y la gente bien contenta también/ participaron/ ehm/ pues eso fue lo de/ Michoacán/ y ahorita estoy esperando el otro paseo (risa)
‘I: because we were going to take/ eh/ the most holy which are este/ the holy hosts
E: mm
I: and yes well for us it is very important and/ mm// very important/ right?/ este/ the most holy/ and the very content people/ participated/ ehm/ well that was what it was in Michoacán/ and right now I’m waiting for the next outing (laughter)’
(CSCM, interview 70 – elder generation 3)

The interviews from HC/HP are not explicitly tagged in this manner. However, thanks to the biographical information available at the beginning of each file, which includes the ages of all the participants, it is possible to categorize the speakers into the three generations to which they belong and maintain consistency with the PRESEEA norms.

3.3.3. Educational attainment

For this study, participants are divided along this line: those with college/university degrees are coded as ‘h’ (for higher education), and those who did not surpass high school are coded as ‘d’ (for diploma).

(13) I: había ahí/ toda clase de verduras/ eh ganado no/ toda clase de verduras vendían ahí
‘I: there there were/ all types of greens/ eh not cattle/ they sold all types of greens there’
(CSCM, interview 108 – high school or less d)

(14) I: el pensar que la mujer había nacido para para/ ser ama de casa/ para ser madre de familia/ buena esposa/ madre abnegada/ y este/ esposa la compañera de su esposo/ en en/ en el hogar y en este/ digamos una cosa muy idealizada […]
'I: the thought that women had been born to to/ be a housewife/ to be a mother to a family/ good wife/ a self-sacrificing mother/ and este/ wife and partner to her husband/ in in/ in the house and in este/ let’s say it’s a very idealized thing […]/
(HC, interview 6 – higher education h)

Again, the CSCM transcriptions are tagged according to the educational level of each participant. The set of transcriptions is divided into sections based on the education attained by the participant: without a diploma, with no more than a diploma, and, at the highest level, with a university degree. The HC/HP corpora do not carry these tags within the transcriptions; instead, they tell what the career of each participant is. However, the division between HC and HP leads to the assumption that the HC speakers have a higher educational attainment than those of HP. Therefore, the fact that there are three levels of education specified in CSCM but only two between HC/HP appears to be a complication. There is a clear demarcation, though, between those who have university degrees and those who do not.

3.3.4. Sex

In both corpora – CSCM and HC/HP – interviews are marked according to the sex of each participant. In this study, ‘M’ indicates male or man, while ‘F’ indicates female or woman.

(15) X: ay no sí es bien [bonito]
I: [no] no es cierto pero/ […] los que viven cerca del desierto
X: mh
I: ellos/ este/ tienen menos posibilidad de/ de construir una mejor casa/ y tienen todo muy árido
‘X: ay no yes it’s very [pretty]
I: [no] it’s not true but/ but/ […] those who live near the desert
X: mh
I: they/ este/ have less possibility of/ of building a better house/ and they have everything very dry’
(HC, interview 9 – women[8] F)

(16) I: [hasta] que llegan a pesar hasta siete/ [siete kilos seis kilos]
F: [¡ihh/ siete kilos!]
I: ¡no! / pues unas gallinotas/ pero gallinotas // eh/ llegó a criar unos gallos/ pero así gallotes
‘I: [until] they grow to weigh up to seven/ [seven kilos six kilos]
F: [ihh!/ seven kilos!]
I: no! / well some big hens/ but hens // eh/ came to raise some roosters/ but big roosters’
(CSCM, interview 37 – man M)

The CSCM interviews correspond with a single participant, and so tagging according to sex is relatively easy. The HC/HP files are not tagged in this manner, but again the biographical information at the beginning of each file gives the necessary details for appropriate coding.

All told, the distribution of participants is as follows, in tables 1 and 2. We begin with the HC/HP informants:

Table 1.Distribution of participants from HC/HP
Education Habla Culta Habla Popular
Age Sex Women Men Women Men
Elder speakers 15 6 5 7
Middle-age speakers 16 12 7 10
Younger speakers 19 20 9 6
TOTAL = 132 50 38 21 23

The number of participants differs from the number of transcribed interviews due to the fact that, in some of the interviews, there were two informants. This is not the case of the CSCM transcriptions, the participant distribution of which appears below:

Table 2.Distribution of CSCM participants
Education Higher education High school education Lesser education
↓ Age Sex Women Men Women Men Women Men
Elder speakers 6 6 6 6 6 6
Middle-age speakers 6 6 6 6 6 6
Younger speakers 6 6 6 6 6 6
TOTAL = 108 18 18 18 18 18 18

All told, the characters representing the dependent variable of marker choice and the four independent factors are concatenated, following the symbol ‘(’, in a token recognizable by the computer program GoldVarb (2005), which will be used to conduct the VARBRUL analysis. For example, a sequence such as (EP2dF represents a production of este by a middle-aged woman with no education past high school from the past, as in the following example:

(17) Inf. -La niña grande ya tiene diez años.
Enc.- Mm.
Inf.- Y es mala ella. Luego le digo: “Mira, este…- le digo a m’hija, le digo- tú no te vayas a meter allá a su casa d’ella, ni cosa ninguna”. Porque ai, como viven todas allí donde vive m’hijo el otro… y yo vivo atrás de la vecindá donde ellos viven…
‘Inf. -The big girl is already ten years old.
Int.- Mm.
Inf.- And she is bad. So I’ll tell you: “Look, este… - I tell my daughter, I tell her- don’t you go into her house, not for a thing.” Because there, how everyone lives over where my other son lives… and I live behind the neighborhood where they live…’
(HP, interview 30)

After entering the token file into GoldVarb, we received various details about the frequencies of marker production and significance of each factor. These results are analyzed in the next section.

4. Results

A total of 13,151 usable occurrences of hesitational este and eh were retrieved from the two corpora. Throughout this section, we will see the connection between the frequencies of use of each marker and how these frequencies, in association with the defined independent factors, differ between the two time periods under study. In the following subsections the factors will be analyzed for significance, first in the past and then in the present, culminating in an analysis of the significance of time period regarding marker usage.

4.1. Marker usage in the past

There are 3438 instances of the relevant HMs from the past corpora, 2461 of them being este (71.6%) and 977 of them being eh (28.4%). Let us examine how these frequencies differ with respect to the three independent factors under consideration.

4.1.1. Age

Looking at the percentages of este and eh associated with the generations in HC/HP, the difference in use among the age groups is evident.

Table 3.Frequencies in the past with respect to age
Age este eh Total N % Markers/ speaker
Younger speakers N 1377 622 1999 58.1 37.02
% 68.9 31.1
Middle-age speakers N 611 263 874 25.4 19.42
% 69.9 30.1
Elder speakers N 473 92 565 16.4 17.12
% 83.7 16.3
Total N 2461 977 3438 26.05
% 71.6 28.4

First, all three groups produced este more frequently than eh. One can see the differences in marker production in general among the generations. Younger speakers produced these markers, without distinction, more frequently than the elder two generations combined.

Regarding marker choice, the proportion of este to eh is nearly equal between younger and middle-age speakers, with a frequency of just under 70%. Interestingly and unexpectedly, the elder speakers produced este at the highest rate, with a frequency of 83.7%.

4.1.2. Educational attainment

With this factor, we can see a large gap between the frequencies of use of the markers.

Table 4.Frequencies in the past with respect to educational attainment
Educational attainment este eh Total N % Markers/ speaker
High school or less N 1363 64 1427 41.5 32.43
% 95.5 4.5
Higher education N 1098 913 2011 58.5 22.85
% 54.6 45.4
Total N 2461 977 3438 26.05
% 71.6 28.4

The participants with higher education produced the two markers fairly evenly, with este produced at a 54.6% rate. This outcome is very distinct from that of the lower-education group. Said group demonstrated an extremely strong preference for este, producing this marker at a rate of 95.5%. It is also noteworthy that the lower-education group produced the HMs in question at a rate higher than average, while the higher-education group did so at a rate lower than average.

4.1.3. Sex

The raw frequencies of marker usage between men and women in this study reveal distinct tendencies and appear to confirm the observations of Soler Arechalde (2006).

Women produced the higher number of HMs (56.5% of total). In accordance with the comparisons of previous studies by Soler Arechalde (2006) and Graham (2013, 2018), we see that women also demonstrate a relatively higher frequency of este in contrast to men.

4.1.4. Preliminary statistical analysis of independent factors

At this point we employ the VARBRUL analysis in order to measure the significance of each factor. The three factor groups – factors plus their levels – are compared to each other, taking into account the relative and absolute frequencies of each, in isolation and in conjunction with the other factors, in order to calculate relative factor weights and tendencies toward one marker or the other (table 6 below) shows the factors and their associated factor weights as calculated by GoldVarb.

Table 5.Frequencies in the past with respect to sex
Sex este eh Total N % Markers/ speaker
Men N 984 511 1495 43.5 24.51
% 65.8 34.2
Women N 1477 466 1943 56.5 27.37
% 76 24
Total N 2461 977 3438 26.05
% 71.6 28.4
Table 6.Factor weights for past corpus
Group Factor Weight App/Total
Education High school or less 0.84 0.96
Higher education 0.24 0.55
Range = 60
Sex Women 0.55 0.76
Men 0.44 0.66
Range = 11
Age Elder speakers [0.54] 0.84
Middle-age speakers [0.49] 0.70
Younger speakers [0.49] 0.69

Total Chi-square = 52.1850
Chi-square/cell = 4.3488
Log likelihood = -1631.388

Factor weight values of >0.50 in this VARBRUL indicate a preference for este, while values <0.50 indicate a dispreference of este (and a preference for eh). The closer to the extremes of 1.00 and 0.00, the stronger the preference in either direction. The factor weight calculations show that, for now, there are two significant factor groups: educational attainment and sex. The calculation of range, which indicates the numerical distance between the highest-weighted factor level and the lowest, is a representation of strength of that factor. In this case, the factor of educational attainment is the strongest, with sex, though significant, being weaker.

The VARBRUL calculations determined that age was not a significant factor in this study. Such a determination normally indicates that the relative frequencies are not appreciably affected by differing factor levels. However, recall from table 5 (and repeated in the “Apps./Total” column of table 8) that the elder group demonstrated a considerably higher frequency of este usage than the other two generations, yet the factor weights do not reflect this higher tendency. When a factor group is determined not to be significant despite such a discrepancy in outcomes, it is necessary to consider the possibility of interaction between factor groups. If factors interact, this means that they are not completely independent from one another. The most straightforward manner to determine which factors are interacting is to cross-tabulate two factors and observe the distributions of outcomes. In the interest of space, I will only show below the cross-tabulation of age and sex.

The cross-tabulation of age and sex (as shown in table 7) makes the picture clear. In terms of sex, we see that elder men far outpace the overall frequency for men (92% este versus 66% overall average), while younger men (60%) are slightly below the overall frequency. Regarding women, there is not as strong an effect; elder women slightly outpace the women’s average (80% este versus 76% overall), and middle-age women are slightly behind.

Table 7.Age and sex cross-tabulated in the past corpus
↓ Sex Age → Elder % Middle-age % Younger % Total %
Men este 158 92 257 68 569 60 984 66
eh 13 8 119 32 379 40 511 34
171 376 948 1495
Women este 315 80 354 71 808 77 1477 76
eh 79 20 144 29 243 23 466 24
394 498 1051 1943
este 473 84 611 70 1377 69 2461 72
eh 92 16 263 30 622 31 977 28
Total 565 874 1999 3438

After identifying the interaction between age and sex, the next step is to recode and perform the VARBRUL analysis again. The factor group of education remains unchanged, and the intersection of age and sex becomes our new factor group, with six levels: older men, middle-age men, younger men, older women, middle-age women, and younger women. Below is the new factor weight table:

Table 8.Factor weights with interaction considered
Group Factor Weight App/Total
Education High school or less 0.84 0.96
Higher education 0.24 0.55
Range = 60
Age + sex Older men 0.74 0.92
Younger women 0.58 0.77
Middle-aged women 0.50 0.71
Older women 0.49 0.8
Middle-aged men 0.47 0.68
Younger men 0.39 0.6
Range = 35

Total Chi-square = 18.5825
Chi-square/cell = 1.5485
Log likelihood = -1613.687

With the interaction accounted for, we see that both the educational attainment and the intersection of age and sex are significant. The factor weights associated with the levels of age + sex correspond with the frequencies in table 7. With the intersection as a factor group, now we can correctly identify how age affects marker production: it is dependent upon sex and is thus not truly independent. Older men show a favoring effect toward este, while younger men, the group least disposed to este, are represented by a factor weight closer to eh. Note that the factor weights for education are identical to the previous analysis: 0.84 for high school or less, and 0.24 for higher education.

4.2. Marker usage in the present

There are 9713 instances of este and eh as HMs from the present-day corpus, 7352 of este (75.7%) and 2360 of eh (24.3%). In the same manner as the analysis of the past corpus, let us examine the frequencies according to age, education, and sex and evaluate each factor for significance.

4.2.1. Age

Upon examination of the frequencies of este/eh usage among the three generations, we can observe important differences among the groups.

Table 9.Frequencies in the present with respect to age
Age este eh Total N % Markers/
Younger speakers N 2014 434 2448 25.2 68
% 82.3 17.7
Middle-age speakers N 2712 992 3704 38.1 102.89
% 73.2 26.8
Elder speakers N 2626 935 3561 36.7 98.92
% 73.7 26.3
Total N 7352 2361 9713 89.94
% 75.7 24.3

Each generation produces este more frequently than eh; middle-age and elder speakers actually produce more HMs overall than younger speakers. The most notable aspect of this distribution of este versus eh is the relative frequency among younger speakers: 82.3% of markers produced by them are este, while the two groups of more advanced age produce the marker at a rate closer to that of the population as a whole.

4.2.2. Educational attainment

Again, both groups produce este more frequently than eh, with the lower-education group producing 67% more markers total than the higher-education group – though, on a per-speaker basis, the higher-education group produces more markers overall.

Table 10.Frequencies in the present with respect to educational attainment
Educational attainment este eh Total N % Markers/ speaker
Higher education N 2434 1207 3641 37.5 101.14
% 66.8 33.2
High school or less N 4918 1154 6072 62.5 84.33
% 81 19
Total N 7352 2361 9713 89.94
% 75.7 24.3

We can see the considerable divergence in percentages between both groups. The group with the university education produces este at a rate of nearly 9 percentage points lower than the average of the entire population. Moreover, the proportion of este by the lower-education group is higher than the overall frequency by 5.3 percentage points.

4.2.3. Sex

As in the past, the women of the present use more HMs than men.

Table 11.Frequencies in the present with respect to sex
Sex este eh Total N % Markers/ speaker
Men N 2992 1146 4138 42.6 76.63
% 72.3 27.7
Women N 4360 1215 5575 57.4 103.24
% 78.2 21.8
Total N 7352 2361 9713 89.94
% 75.7 24.3

Women continue to produce HMs more frequently than men, comprising 57.4% of the total. We see that the proportions of este-eh between men and women are closer in the present than in the past, though women still predominantly produce este more frequently than women.

4.2.4. Statistical analysis of independent factors

As with the study of markers in the past, GoldVarb calculates the weights of each factor level and determines significance of each factor group. Below we see the factor weights as calculated.

Table 12.Factor weights from the present
Group Factor Weight App/Total
Educational attainment High school or less 0.57 0.81
Higher education 0.39 0.67
Range = 18
Age Younger speakers 0.59 0.82
Middle-age speakers 0.47 0.73
Elder speakers 0.47 0.74
Range = 12
Sex Women 0.54 0.78
Men 0.45 0.72
Range = 9

The factor group of educational attainment is the strongest according to the analysis, with a range of 18 between the two factor levels. We see that those with less education slightly favor este, while those with higher education tend not to prefer este as a marker (relatively), implicating a greater preference for eh. With respect to age, middle-age and elder speakers show a very slight preference toward eh as shown by the factor weights of just under 0.50, while younger speakers favor este. Finally, the weakest indicator is sex, with women only slightly favoring este and men only slightly favoring eh. The VARBRUL analysis determined all three factors to be significant, and, as a departure from the quantitative analysis of the speech of the past, there is no measurable interaction between any of the factors.

4.3. Significance of marker choice between the past and the present

Up to this point, we have discussed the significance of social factors – age, sex, education – with respect to their influences on the choice between este and eh in hesitation. Having seen the results, the question remains: is the period of time a significant factor in the preference for one marker over another? It is known that este is the most frequently produced HM during both the past and the present, and, furthermore, that the percentage of este usage in the present is slightly higher than that of the past (see table 1). Is this discrepancy statistically significant?

Table 13.Factor weights of time period
Group Factor Weight Apps./Total
Time period Present [0.50] 0.76
Past [0.49] 0.72

The factor weights of each level are nearly equal. Because of this, the VARBRUL analysis calculates, predictably, that the time period during which the interviews took place is not a significant indicator of marker choice or production.

5. Discussion of results

To summarize the results presented in the previous section, we revisit the stated hypotheses. Are the hypotheses of age, education, and sex as significant factors supported in HM choice of the speech of the past and of the present in Mexico City? And are these significances different between the two time periods?

5.1. Age – este as a generational marker, or trend in sociopragmatic change?

Regarding age as a factor, we find that it is a significant indicator in the present, with younger speakers producing significantly higher proportions of este compared to middle-age and elder speakers. This sharply contrasts with the past, during which the generation most likely to produce este was the elder generation, once again an unexpected finding. Recall that, in the past corpora, age was not completely independent as a factor; we encounter an interaction with sex.

Another notable finding regarding age across time was the difference in rates of HM production overall between the past and the present.

Table 14.Age across time
Age Marker Time period
Past % Present % %
Younger speakers este 1377 69 2014 82 3391 76
eh 622 31 434 18 1056 24
Total 1999 2448 4447
Middle-age speakers este 611 70 2712 73 3323 73
eh 263 30 992 27 1255 27
Total 874 3704 4578
Elder speakers este 473 84 2626 74 3099 75
eh 92 16 935 26 1027 25
Total 565 3561 4126

The elder generation of the past produced a total of 565 markers, which was the fewest across all generations past and present. We see that the middle-age group produced 874 HMs, whereas the younger generation produced 1999 HMs, more than their elders combined. This contrasts considerably with the speakers from the present corpus in which the younger generation produces HMs less frequently than their predecessors.

In both time periods the younger generation appears to be an indicator of a new trend. The upshot of HM production by the younger generation in the HC/HP corpora portends an increase in popularity of, or a decrease in stigma surrounding, HMs. By contrast, the lower occurrence of HMs in general by the younger speakers in the CSCM corpus indicates a decrease in popularity of these markers. Interestingly, the rate of variation in HM production in the middle-age generation remains relatively flat from the past to the present.

5.2. Educational attainment – greater acceptance of este across time?

According to various studies of these markers in modern or current usage (studies which have already been mentioned in this paper), speakers with higher levels of education tend to produce este less frequently on a percentual basis, contrasted with lower-education speakers. This sociolinguistic characteristic is observed in both the past and the present, though the proportions differ:

Table 15.Education across time
Educational attainment Marker Time period
Past % Present % %
Higher education este 1098 55 2434 67 3532 62
eh 913 45 1207 33 2120 38
Total 2011 3641 5652
High school or less este 1363 96 4918 81 6281 84
eh 64 4 1154 19 1218 16
Total 1427 6072 7499

We can see that, for the lower-education group, the ratio of este to eh in the present is not as wide as it was in the past. Even though this group produces a lower proportion of este in the present than in the past, nevertheless, it is evident that este maintains a considerably strong association with the speech of the less educated. In the data from the past, the higher-education group was more balanced in their choice of HM, with only a slight tendency toward este. This tendency becomes stronger in the present, with este becoming twice as frequently produced as eh. It is possible to describe this proportional increase as a decrease in stigma, though small, around este by those with greater educational attainment.

5.3. Sex – marker of identity?

In the same manner as educational attainment, one can observe notable differences in the frequencies of HM production by men and women. There is always a greater frequency of este produced by either group, but do these proportions differ across the years?

Table 16.Sex across time
Sex Marker Time period
Past % Present % Total N %
Women este 1477 76 4360 78 5837 78
eh 466 24 1215 22 1681 22
Total 1943 5575 7518
Men este 984 66 2992 72 3976 71
eh 511 34 1146 28 1657 29
Total 1495 4138 5633

The production patterns by women do not change much across the years; women produce este 76% of the time in the past compared to 78% in the present. We can see, though, a greater increase in use by men across the same time period: 66% este production in the past compared to 72% in the present. If there is a conclusion to be drawn by these statistics, it is that it appears that the relationship between sex and este usage has weakened slightly over the years.

5.4. Sex and education combined – implications for the value of formality between men and women?

An observation that appears to contradict the findings of Soler Arechalde (2006, 2012), perhaps for the fact that in the current study two HMs are compared as opposed to analyzing just one, is the rate of occurrences of este by both sexes. First, I present the data from HC/HP of the intersection of sex and education:

Table 17.Sex with respect to education in the past
Sex Marker Education (past)
High school or less % Higher education % Total N %
Men este 520 92 464 50 984 66
eh 44 8 467 50 511 34
Total 564 931 1495
Women este 843 98 634 59 1477 76
eh 20 2 446 41 466 24
Total 863 1080 1943

Nearly three fifths of HMs produced by women with higher education were este, demonstrating their tendency toward that marker. Men with higher education, however, did not clearly display a preference for one marker or the other. These outcomes stand in stark contrast to the men and women with less education, both of which show a nearly categorical preference for este. In both groups, we find that the women’s preference is still stronger toward este than that of men.

Now, let us consider the same intersection of data, but using CSCM:

Table 18.Sex with respect to education in the present
Sex Marker Education (present)
High school or less % Higher education % Total N %
Men este 2023 77 969 64 2992 72
eh 605 23 541 36 1146 28
Total 2628 1510 4138
Women este 2895 84 1465 69 4360 78
eh 549 16 666 31 1215 22
Total 3444 2131 5575

We can see that, contrasted with the past, women and men with higher educational attainment do not differ as much as the same groups in the past in their use of este versus eh, with a difference of only 5 percentage points between the sexes. Furthermore, we also see that both men and women, compared to the past corpus, display a clear preference for este that either used to be weaker (in the group of women) or that did not exist before (in the group of men). Within each sex group, the differences in frequency of este use, based on education, are still considerable, but the discrepancy between the two educational levels is less pronounced in the present than in the past. Therefore, we can consider the possibility that, sociolinguistically, Mexico City speakers in the present do not tend to avoid lower-prestige forms as frequently in more formal contexts, be they men or women.

5.5. Sex and age – reduced interaction?

The variable rules analysis found that the factors of sex and age were not wholly independent in the HC/HP corpus. Rather, there was an interaction effect between these two factors that had to be analyzed. No such effect was found in the present corpus however. What does this outcome mean in context?

Recall from table 7 (which I will not reproduce here for reasons of space) that the group which produced este at the highest proportion was the elder men, and that the younger men were the group least likely to use este. The other four factor levels (middle-aged men, and all three generations of women) did not exhibit a strong preference in either direction. This trend, though, is not retained in the modern corpus, as table 19 shows.

Table 19.Sex with respect to age in the present
↓ Sex Age → Elder % Middle-age % Younger % Total %
Men este 1089 69 1042 72 861 77 2992 72
eh 480 31 407 28 259 23 1146 28
Total 1569 1449 1120 4138
Women este 1537 77 1670 74 1153 87 4360 78
eh 455 23 585 26 175 13 1215 22
Total 1992 2255 1328 5575
Total este 2626 74 2712 73 2014 82 7352 76
eh 935 26 992 27 434 18 2361 24
Total 3561 3704 2448 9713

In contrast to the past corpus, elder men are the least likely to choose este over eh, whereas younger men have come to be the second-most likely group to make this same choice, alongside elder women. Thus, we see a return to this preference in the youngest speaker group. An interesting consideration is that the elder group in the CSCM corpus may largely corresponds to the younger group in the HC/HP corpus, age-wise. The youngest elder in CSCM – around 55 years old, according to convention – would have been 20 at the oldest when HC/HP was compiled. A possibility to consider is that este was simply not as popular among that group speakers of that generation compared to others. As a result, the speakers’ speech patterns remained the same as they aged.

6. Conclusions and directions for future study

It has been demonstrated that the same social factors affecting the variation between hesitation markers este and eh in the speech of Mexico City in the past remain active in the speech of the present. Within each period of interest, the effects of certain social factors on the tendency toward each marker have varied, indicating changes in the social associations and attitudes toward each marker. It has also been demonstrated that the factors are not all independent of one another. There is a significant dependency of age upon sex in Mexico City Spanish in the past, one which does not persist in the present.

The discussion around HM usage in Section 1 of this paper revealed a social stigma associated to these forms. Blas Arroyo (1995) and Soler Arechalde (2006) mention that they are traditionally regarded as improper or imperfect speech, with no communicative value, and conversations within the corpus data, specifically those of CSCM, reflect the presence of this attitude among certain speakers in Mexico City. As alluded to in the subsections discussing educational attainment and its effects on marker choice, the results herein suggest that the association between este and poor education has weakened significantly in the time between the compilation of these corpora. It is quite possible that the attitudes surrounding este, and HMs in general, have become more relaxed in the modern day. A study of linguistic attitudes surrounding HMs would be a welcome extension of this discussion.

Another possible future direction along this line of research may be a more thorough investigation of the discursive or pragmatic variation between the past and the present, focusing on the myriad of uses of the markers and if these nuances are maintained across time. It is also the hope of the author that the discussion around HM variation in the Hispanophone world would extend to other countries and cities, to include those areas in which other markers are used to indicate hesitation.

  1. “CSCM” = Corpus sociolingüístico de la ciudad de México. Hereafter, “HC” = Habla Culta y “HP” = Habla Popular. These three corpora will be introduced further in Section 3.

  2. Ocampo (2006) defines discursivization as, essentially, a process paralleling grammaticalization in which a form evolves from more lexical to more discursive/pragmatic.

  3. This is debatable given the lack of contextual fit (as remarked by Martínez et al., 2004). The true origins of este as a HM are reserved for future research.

  4. Alongside Mexico City, the other areas under study were San Juan (PR), Montevideo (UY), and Medellín (CO).

  5. Due to the lack of variability in the types of discourse from the HP informants, Soler Arechalde does not quantitatively analyze this corpus.

  6. Notable in these results from Soler Arechalde (2012) is the fact that, in a previous study by Soler Arechalde and Serrano (2010), neither sex nor the type of interview had any effect on the frequency of este. The authors attribute this to a low number of occurrences.

  7. If a speaker produces a cluster of both markers under study, such as este eh or eh este, both are included.

  8. Interview 9 from HC was one of those with two participants I and X. Both of these participants, though, were women, thus not complicating the coding.