Measuring the Degree of Near-Synonymy of Spanish Verbs of Putting

Measuring the degree of near-synonymy of Spanish verbs of putting:

Amultivariable corpus analysis of poner and meter

Abstract

The present article proposes a corpus-based study of two near-synonymous verbs in Spanish: the (locative)verbs of putting poner and meter. Starting from the universal principle of linguistic economy, the paper aims to empirically identify potentially influencing variables that determine the native speaker’s choice between the two verbs. It is investigated whether and to what extent the verb choice is governed by a set of variables related to the nature of the placement movement itself and the characteristics of the participants taking part in the event. The difference between the near-synonyms is shown to bemainly determined by the direction of the locative movement, the semantic nature of the participants (animacy, concreteness), the reflexiveness of the event, as well as the cognitive construal of the locative event (the possibility of a container-reading).

Key words: (locative) verbs of putting,Spanish, near-synonymy, logistic regression analysis

1. Introduction: Near-synonymous (locative) verbs of putting

Many languagesdisplay a host of morphologically or semantically cognate lexemes and expressions which seem to be synonymous at first sight, but which, looking closer, present clearly divergent uses. From anintra-linguistic viewpoint, most, if not all, seemingly synonymous expressions turn out not to beperfect counterparts, presenting (subtle) differences in terms of meanings, frequencies orsyntactic, semantic and pragmatic configurations (Enghels & Jansegers 2013, Enghels & Roegiest in press, FagardMardale 2012, Soares Da Silva 2012 among others).

This paperzooms in on linguistic parallels and differencesbetween the verbs of putting (henceforth putVs)poner and meterin European Spanish, in their purely locative use.[1]PutVs express a change of location of an entity from one place to another, and typically present a syntactic configuration with three arguments in the V valency: [S + V + DO + Locative Complement]. Following Talmy (1985: 62, 1991: 488), the prototypical — i.e. cognitively most prominent — event of putting involves a human agentive participant (which will be called herethe Causer, first participant or P1), who moves an inanimate physical object, namely the Figure(the second participant involved or P2), towards a location in space, called the Ground. This is shown in Figure 1.

Figure 1.The prototypical event of putting

However, when the Vs are used in a reflexive construction, the Causer P1 and Figure P2 are co-referential and make reference to the same entity (e.g. Me pongo contra la pared ‘I put myself against the wall’; Me metoen mi cuarto a estudiar ‘I put myself in my room to study’):

Figure 2. Reflexive event of putting

In their locative use, poner and meterseem to be synonyms,and can appear in semantically similar contexts (1a)–(1b). Even in one and the same particular context both verbs can co-occur with the same displaced Figure (2):

(1a) Yanohayqueiraponerseenla

Yetnothave-PRS.3SG that gotoput-INF.REFLinthe cola[2]

line.

‘It is not necessary anymore to go stand in line.’ (lit. ‘to put oneself in the line’)

(1b) Alregreso, podrámeterseenlacola.

At.ART.SGreturn,can-FUT.3SGput-INF.REFLintheline

‘At his return, he could go stand in line.’ (lit. ‘to put himself in the line’)

(2) Paranoaplastarlastodavíadelicadasramasde

Tonotflatten-INFthestilldelicatebranchesof

losarbolitosqueacababadecomprar, enlugarde

thesaplingsthatend-PST.3SGofbuy-INF, inplaceof

meterlasdosmacetasenelmaletero,las

put-INFthetwoflowerpotsintheboot,them

pusoenelinteriordelcoche, entre los

put-PST.3SGintheinsideof.ART.SGcar between the

asientosdelanteros y el de atrás

seats front and the of back.

‘In order not to flatten the still delicate branches of the saplings he just bought, instead of putting the two flowerpots into the boot, he put them in the inside of the car,between the front seats and the ones in the back.’

Indeed these examples seem tosuggest that from an intra-linguistic viewpointboth near-synonymous Vs are freely interchangeable and may alternate in the same syntactic-semantic contexts. However, a quick analysis of a set of minimal pairs shows that in certain contexts onlyone of both Vs is highly frequent, while the other one causesproblems of acceptabilityor does not occur at all. Consider the following cases:

-meterse en el ascensor (853 000 hits)[3] vs. ponerse en el ascensor(5 hits), ‘to putoneself in theelevator’;

-poner los platos sobre la mesa (6700 hits) vs. *meter los platos sobre la mesa (no hits),‘to puttheplatesonthetable’.

Hence, the verbs are no perfect synonyms and their syntactic-semantic distribution is not identical.

The observed differences for the V pair were studied by Cifuentes (1996, 2004). The author argues that the fundamental difference between poner and meterresides in the direction of the placement movement (Cifuentes2004: 81). In this view, both Vs belong to two different subcategories of the locative Vs: (1) meter is classified as a directional V with a particular preference for movements with an internally oriented directionality, that is, objects are placed inside a space; (2) poner is defined as a more static positional V (cf. also Cifuentes 1996: 103,139). Although Cifuentes’ theory provides aplausibleexplanation for the clear preference for meterin the examples above expressing internal direction (e.g. meterseen el ascensor), two issuesremain unanswered:[4]

(a)Cifuentes’ analysis ismainly based on introspectivereflection, illustrated withwell-selected but often isolated examples. So, up to now, there is a lack of thorough empirical evidence for the semantic definitions and subtleties proposed for the putVs. The present articleaims precisely to check these hypotheses against empirical data from an extensive corpus of contemporary Spanish examples. The corpus analysisthus seeks tooperationalize the cognitive approach of Cifuentes.

(b)Is the internalvs. neutral directionality of the locative event (respectively leading to the use of meter vs. poner), as proposed by Cifuentes,the only possible source of influence for the choice between poner and meteror are there other intervening factors related to the semantics of the main constituents of the event of putting itself?

The outline of this paper is as follows. In Section 2, we present the data set and methodology on which the analysis is based. Section 3 introducesa series of hypotheses and research questions on variables that potentially determine the choicebetweenponerandmeter, in order to assess their influence on the choice between both putVs. Section 4 proposes a multivariable logistic regression analysisin order to test the proposed hypotheses. These results are compared and discussed in the last section of thepaper (Section 5).

1. Towards a more advanced statistical approach in semantics

2.1A multidimensional empirical method of analysis

The aforementioned research objectives situate this article within the domain of lexical studies from the perspective of cognitive semantics. Indeed, one of the central areas of research within this domain has been the investigation of polysemy and (near-)synonymy (e.g. the papers collected in GlynnFischer 2010). Recently, special attention has been paid to the methodologies by which semantics can be studied on the most accurate and verifiable basis.It is well known that linguistics has experienced a significant shift from intuition-based approaches towards the use of corpora and empirical methods. Not only have corpora proven useful for quantitative studies of morphosyntactic aspects of languages, but also within the field of semantics several authors have recently argued for the necessity of pursuing more empirical methods (Geeraerts 2010; Gries 2006; Janda 2013, among many others). Hence, the methodology of this article will also be corpus-based, and the degree of equivalence between the near-synonyms will be examined on the basis of a comparable corpus.

Usage-based models of language call for multidimensional empirical methods of analysis, as in most of the cases language phenomena are influenced by several linguistic variables at a time. Linguistic choices are indeed rarely the result of only one influencing variable, and therefore bivariate statistics,often applied in corpus studies, are not always the best way to describe them. Bivariate analyses do not take into consideration the combined effect of different influencing variables together, nor do they give insight into possible interactions between these variables. It is also hard to tell which effects are stronger or weaker than others, as effect measures (e.g. Cramer’s V) often lie in the same order of magnitude. We therefore propose amultivariablelogistic regression analysis in order to study the effect of a series of variables on the choice for poner and meter in present-day peninsular Spanish.

Logistic regression is a standard approach to analyze dichotomous outcomes of a so-called response-variable (in this case the V choice meter vs. poner, where meter is called the success outcome), often applied in corpus-linguistic research. For all possible occurrences, the resulting statistical model allows to predict the outcome (meter/poner) in function of a series of influencing variables (called predictor variables). The analysis provides insight into how strong each independent predictor variable determines the choice of one of both putVs in the presence of the other potentially influencing variables, and enables to detect possible interactions between two different predictor variables. The output of a logistic regression analysis is a series of estimated coefficients for each variable as part of a mathematical equation which predicts the outcome (i.e. whether meter or poner will be chosen) on the basis of all the predictor variables together (see Table 3 in the Results section). The further the coefficients are away from zero, the stronger the effect. In that sense, high (in absolute terms) coefficients with a statistically significant effect indicate a strong influence and coefficients close to 0 indicate a weaker influence.In our analysis below, positive coefficients indicate an increased chance for the success outcome (meter), whereas negative coefficients indicate an effect in the sense of the other outcome (poner). As these coefficients are not straightforward to interpret (they do not indicate a linear relation between the predictor variables and the outcome, but a logistic relation with respect to the response variable), they are more easily interpreted in the form of effect plots (see Figure 3below).[5]

2.2 Data

In order to study the choice between poner and meter, we collecteda random sample of 667 locative uses of the two putVsin modern Spanish fiction texts.[6] This samplewas extracted from a larger corpus of 2000 random examples of poner and meter, in threecontexts of usage: the locative use as studied in this article, a transfer use (e.g. El profesornos ha metidotantosdeberes ‘The teacher gave us so much homework’) anda pseudo-copulative use (e.g. Me pongonervioso ‘I get nervous’).The data were extracted from the online database Corpus de Referencia del Español Actual (CREA).[7]Since this is a non-lemmatized corpus, we have collected the data by searching for specific verb forms, namely all regular and irregular verb stems. For poner, we searched through the querypong* o pone* o puse o pusi* o pondr* o pusoopuest* o ponía* o ponie*, and for meter, we searched for meto o metí o metie* o mete* o meti* o metía* o meta*, where the o stands for ‘or’. CREA automatically gives the results of each verb in a random way, with all verb forms mixed up randomly.[8]

For the analysis in this paper, only the examples of the locative use were selected. Table 1 shows thatthere are many more locative examples of meter than of poner(401 vs. 266): meter seems to be more restricted to the expression of placement alone, and does not extend its use to other semantic domains to the same extent as poner(cf. Comer et al.: in press).

Table 1. Frequenciesof the two Vs in our corpus

Spanish
Poner / Meter
266 / 401
667

Moreover, only cases with three explicit arguments (Causer, Figure and Ground)wereretained, including reflexive uses of the Vs, so as to determine the impact of the semantic nature of all participants on the V choice (cf. Section 3.2).[9]As a result, examples with an implicit locative complement (3), and cases of passive voice, where the Causerremains unspecified,are excluded(4):[10]

(3)Fuealcementerioaponerunasflores.

Go-PST.3SGto.ART.SGchurchyardtoput-INFsomeflowers

‘He went to the churchyard to put some flowers.’

(4)Estámetidoenungrupoconsusamigos.

Be-AUXput-PTCP.3SGinagroupwithhisfriends

‘He is part of a group together with his friends.’ (lit.‘He is put in a group with his friends.’)

The corpus contains both concrete events of putting (e.g. poner el libroen la mesa ‘to put the book on the table’) and metaphoricallocation events (e.g. poner a alguienpor la nubes‘to praise someone’, lit. ‘to put someone in the clouds’).

3. Overview of potentially influencing variables

This section sets forth five groups of predictor variables that potentially influence the choice of one of both near-synonymous Vs. These variables present both event-related characteristics and inherent properties of the participants. In Section 3.1, we formulate two working hypotheses to be investigated in the corpus, both based on Cifuentes’ theory. In order to maximally broaden the scope of the empirical analysis, Section 3.2 adds three variables, leading towards three additional research questions, which take into account possible deviations of the prototypical putting event. The hypotheses and research questions presented hereafterwill be exemplified and verifiedin Section 4 and 5.

3.1 Empirical verification of Cifuentes’ hypothesis

3.1.1 Directionality of the movement

As mentioned in Section 1, Cifuentes (1996: 103, 139) claims that the main difference between poner and meter in their locative use is the directionality of the putting event: meter prefers an illative movement (i.e. placement of the Figure inside of the Ground), whereas poner is neutral with respect to the direction of the movement. In order to test this hypothesis and to gain further insight into the directionalities preferred by each of the two Vs, their locative complements are classified according to the direction of the placement event.

For the variable Direction,[11] with respect to the directionality conveyed by the preposition, three categories are distinguished in the corpus. Whenever the Figure is placed inside a container, the direction is marked as illative or internal. These illative movementsclearly contrast with a category that includes vertically (i.e. locating the Figure above or below the Ground) andhorizontally oriented changes of location (i.e. directed towards a location in front of, at the back of, next to the Ground, or in between two different Grounds).A third category includesother types of directionality (e.g. exterior movements or cases where no specific horizontal, vertical or internal direction can be distinguished due to semantic vagueness of the preposition: e.g. hacia, which designates a vague ‘towards’), as well as events without an explicit directionality (e.g. when the Ground is expressed by a semantically neutral adverb, as in I put the book here, or by an interrogative pronoun, as in Where did you put the book?).[12]This leads to the following working hypothesis:

Hypothesis (1): meter most frequently combines with prepositions that mark illative movement, whereas poner does not impose restrictions as to preposition selection.

The directionality of the movement will thus be a first variable to be examined in the corpus.

3.1.2Dimension of the Ground

Cifuentes (1996, 2004) also examines to what extent the directionality difference has an impact on the semantics of the Ground and, more particularly, on its dimension. On the one hand, the often internally directed orientation found with meter would make this V particularly suitable with large-sized Grounds (e.g. a house).Poner,on the other hand, would be less inclined to combine with a Ground of large dimensions: it expresses a simple positioning event, be it vertical, horizontal or interior. In contrast, according to Cifuentes (1996: 140–141) smaller-sized Grounds (e.g. aneye) combine both with poner and meter.

In order to empirically verify this hypothesis, all instances of poner and meterare tagged as to the dimension of their Ground for the variable Dimension_Ground.The human body is taken as a reference point: human beings and every entity withthe same size or larger than the human body are considered to be large Grounds (e.g. geographical places, streets, buildings, cars and rooms, as well as clearly spacious areas such as the sea).In contrast, entities with a smaller dimension compared to the human body (e.g. bags, birdcages, body parts) are marked as small Grounds. A third category (not applicable) is provided for events in which the dimension of the Ground is hard to classify in terms of large or small, i.e. metaphorical placements with a non-material abstract Ground (e.g.troubles), events where the Ground is expressed by an interrogative pronoun (e.g. where?), an adverb (e.g. here), or a semantically vague location place (e.g. location). On this basis,the following hypothesis is raised:

Hypothesis (2): meteroccurs with both large-sized and small-sized Grounds, whereas poner preferably combines with small-sized Grounds.

3.2 In search of other factors of influence

Cifuentes(1996, 2004) mainly examined occurrences of meter and poner in their prototypical locative meaning, that is, when referring to a literal non-reflexive placement event including a human instigator and an inanimate and concrete Figure and Ground. However, putting events can deviate from this basic definition in numerous ways. It is thus interesting to investigate to what extent the selection between poner and meter is influenced by possible semantic extensions from the prototype. More particularly we want to examine to what extent one V, in casumeter, could have specialized as a prototypically locative V, whereas ponerwould be more frequently used to denote non-prototypical locative events. This hypothesis is in line with the results of a previous study (Comer et al., in press) which showed that ponerallows for much more semantic extensions thanmeter. Consequently,metercan be defined as the locative V par excellence in Spanish.These semantic extensions can be related to the animacy and concreteness of the participants (3.2.1), the co-referentiality of the Causer with the Figure (3.2.2), and the literal or metaphorical interpretation of the putting event (3.2.3).

3.2.1Animacy and concreteness of the participants

In the first place, given that this study aims to investigate whether both lexemes poner and meter can be correlated with different event types and different patterns of location, it makes sense to take into account the semantics of the participants implied in the locative event. In fact, Talmy(2000a: 434-435) affirmed the impact of animacy on linguistic patterns of force dynamics in language and cognition, and specifically also with locative events (Talmy 2000b: 301, 319).As described in Section 1, the prototypical locative event consists of a human agentive Causer, moving an inanimate physical Figure towards a concrete Ground or location in space. However, various semantic extensions from this prototypical configuration are possible. The Causer, for instance, is not always a human being, but can be an inanimate abstract entity (e.g. a lucky chance put him in my life). Nor is the moved Figure necessarily an inanimate object: it can also be an animate entity (e.g. to put someone in jail). The Ground is not always a concrete location, but can also be abstract (e.g. to put yourself in someone’s life). Given the diverse semantic nature of the participants, the question arises whether the animate or inanimate and concrete or abstract nature of the three arguments in the locative event have an influence on the choice between poner and meter.