Combinatorial Productivity of Spanish Verbal Periphrases as an Indicator of Their Degree of Grammaticalization

Abstract

Studies on the constitution of the Spanish periphrastic system show that there is a great ease with which verbal periphrases admit different lexical items in the second verb slot as they go through their grammaticalization process. However, it has not been sufficiently explored whether the evolution of combinatorial patterns in near-synonymous periphrases follows similar grammaticalization paths. Adopting a constructionist, usage-based approach, we investigate the evolution of the so-called near-synonymous periphrases dejar de + inf and parar de + inf, as in Deja de/Para de gritar, ‘Stop shouting.’ More specifically, we discuss the semantic areas they cover, the functional distribution between the two throughout time, their evolution in terms of collostructional patterns, and their realized and potential productivity, paying special attention to the Aktionsart of the predicates in the Vinf slot. All tokens in the corpus were extracted from CORDE and analyzed in terms of morphosyntactic and semantic-pragmatic parameters, as well as contextual elements. We conduct a distinctive collexeme analysis to investigate which lexemes are strongly attracted or repelled by the Vinf slot in each construction. This analysis shows that the evolution of parar de + inf is not parallel to that of dejar de + inf and that there is a clear distribution of labor between the two constructions.

1. Introduction

The objective of this work is to prove that the differences that allow us to distinguish so-called near-synonymous periphrases are related, at least partially, with the predicate types that appear in the position of the verb in the non-finite form. This is particularly true about phasal periphrases, whose meaning determines the Aktionsart of the verb in the non-finite form. To test our hypothesis, we will address the evolutionary study of two aspectual periphrases, namely, dejar de + inf and parar de + inf, as in (1a–1b):

(1)	a.	Por	fin	dej-ó			de	llov-er
		for	end	leave-ind.pst.3sg			of	rain-inf
		‘At last, it stopped raining’
	b.	Cuando	par-e		de	llov-er,		tend-er-é	la	ropa
		when	stop-subj.pst.3sg		of	rain-inf		hang.out-fut-1sg	the	clothes
		‘When it stops raining, I’ll hang out the clothes’

The periphrasis in (1a) has been documented in Spanish since the first written manifestations of the language. The periphrasis in (1b), on the other hand, is only found in texts from the 16th century onward. The existence of dejar de + inf, a consolidated verbal construction in the language since the medieval period, raises the question of how the most recent periphrastic construction arose and how it was integrated into the paradigm of phasal periphrases. More specifically, we wonder whether parar de + inf is born as a syntactic copy of dejar de + inf and in its expansion in the language is thus subordinated to the values of dejar de + inf, or whether it follows a free model, which would confirm the constructivist hypothesis that syntactic evolution affects not only saturated constructions but also more abstract schemes (Hoffmann 2014, p. 161). Likewise, we will look into the changes in productivity of both periphrases throughout their history and what the implications of such changes are. Note that we use the concept of productivity in the sense of Barðdal (2008), that is, the capacity of a construction to gradually attract new lexical items.

To answer these questions, we will address the historical study of both periphrases from a corpus-based and collostructional perspective, which will allow us to measure the similarities and differences between the periphrases under study by investigating which lexemes are strongly attracted or repelled by the non-finite verb form (Vinf, in this case) slot in each construction. If the predicate types that appear as infinitives in the periphrasis parar de + inf replicate those that appear in dejar de + inf, we can argue that the paradigmatic attraction exerted by dejar de + inf on parar de + inf affects not only the meaning of the construction but also its combinatorics. In such a case, the analogic pressure exerted by dejar de + inf would affect the more saturated levels of the construction with parar. On the other hand, if differences relative to the predicate types are detected, this might suggest that the paradigmatic relations between near-synonymous periphrases suppose a distribution of values already from the first uses of the constructions. At this point, we want to point out that we use the term analogy as it is conceptualized in Barðdal (2008, pp. 2–3), that is, as “analogical extensions, i.e., extensions of a grammatical pattern based on only one model item, because of structural or semantic similarities between the two items.”

Before continuing, it is worth mentioning that the periphrastic character of dejar de + inf has not raised doubts and that it has been accepted as a member of the category of verbal periphrases by all the authors who have dealt with it (Yllera 1980; Gómez Torrego 1988, 1999; Gómez Manzano 1992; Olbertz 1998; Camus 2006a; RAE/ASALE 2009). On the other hand, the unanimity is not total with regards to the status of parar de + inf, for which links continue to be traced back to the lexical verb parar, ‘to stop.’ Thus, while some authors do not hesitate to consider that parar de + inf is a periphrasis (Gómez Torrego 1999; Camus 2006b; RAE/ASALE 2009; Aparicio Mera 2016), others argue that it only works as a periphrasis in contexts of negative modality (Fernández de Castro 1999, p. 267), where it expresses a continuative meaning with no traces of the sense of ‘stopping a movement’ (e.g., No para de molestar means ‘(s)he bothers all the time’). Finally, other authors exclude parar de + inf from the category of periphrases, regard it as a semi-auxiliary construction (Olbertz 1998, p. 114), or do not even include it in their studies (Gómez Manzano 1992).

In this paper, we are not going to discuss whether parar de + inf meets all the requirements that are considered defining of the paradigm of periphrases, since we start from a radial conception of grammatical categories, with some prototypical members (such as tener que + inf, ’to have to + inf’) and others that may not meet all the defining criteria (such as parar de + inf). We follow the approach set out in previous studies by Garachana (2017) and Rosemeyer and Garachana (2019) and argue that parar de + inf is a grammatical construction in the sense of construction grammar, i.e., it is a conventionalized pairing of form and meaning. Given that in practically all the contexts of use of parar de + inf, it is feasible to replace it with dejar de + inf, and given that in the cases in which said alternation is not possible, this is due to pragmatic or usage issues, we fail to see any reason for not contemplating parar de + inf as a non-prototypical periphrastic construction. However, we must highlight the semantic differences between the lexical meaning of parar and its employment as an auxiliary verb. Indeed, the lexical meaning of parar refers to the interruption of a movement; thus, parar el coche (lit. ‘to stop the car’) means ‘interrupting the movement of a car.’ On the contrary, the interruption that is predicated in the periphrasis, especially with certain verbs, does not refer to any movement but must be interpreted in terms of aspect. In this sense, ha parado de llover, ‘it has stopped raining’ does not imply the cessation of any movement but rather that the event expressed by the infinitive llover, ‘to rain’ is interrupted. That is, from the moment that an infinitive appears in the position of complement of the preposition, parar is reanalyzed as an auxiliary, and ensemble comes to mean interruptive aspect. Considering that parar in parar de llover ‘to stop raining,’ or even in parar de molestar ‘to stop bothering’ or parar de beber ‘to stop drinking,’ is equivalent to stopping a movement or activity, it implies confusing a semantic interpretation with an aspectual one: the construction parar de + inf is used to explain how the event designated by the infinitive develops. Certainly, a metaphorical relationship can be perceived between parar ‘to stop’ and parar de + inf ‘cessative egressive aspect,’ however, an absolute identification between the lexical verb and the auxiliary verb in the periphrasis is not feasible; only the persistence (in the sense of Hopper 1991) can be identified (for a more detailed discussion on this issue, cf. Garachana 2021).

The rest of the article is organized as follows: Section 2 briefly discusses the theoretical framework, corpus data, and methods used. Section 3 presents a general description of the semantics of “dejar de + inf” and “parar de + inf” in contemporary Spanish. Section 4 discusses the evolution of both periphrases under study throughout time. In this section, we also look at the realized and potential productivity of both periphrases, paying special attention to the Aktionsart of the predicates. Section 5 discusses the results of our qualitative and quantitative data analysis, and Section 6 summarizes our findings and advances some final conclusions.

2. Theoretical and Methodological Framework

We investigate the evolution of the near-synonymous Spanish periphrases dejar de + inf and parar de + inf, adopting a constructionist, usage-based approach. Specifically, our work is framed in the models of Construction Grammar (as developed by Langacker 1987, 2003; Fillmore 1996; Croft 2000; Croft and Cruise 2004; Goldberg 2006; Hoffmann 2013, a.o.) and Diachronic Construction Grammar (e.g., Fried 2009; Traugott and Trousdale 2013; Barðdal et al. 2015).

To carry out this research, all documented instances of both periphrases in the diachronic corpus of the Real Academia Española (2007) CORDE (Corpus diacrónico del español) were extracted. This written corpus constitutes the largest historical data bank in Spanish, whose ecdotic reliability can be verified from the work carried out by Rodríguez Molina and Octavio de Toledo y Huerta (2017). CORDE covers a historical period that spans from the origins of the language until 1974 and contains texts of different textual typologies. We obtained a total of n = 308 tokens of valid data for the study of parar de + inf and n = 31,362 tokens of dejar de + inf. The datasets on which we base our study come from the documentary archive of the GRADIA group.

Each token in our datasets was analyzed in terms of morphosyntactic and lexicogrammatical parameters, such as predicate type and lexical aspect of the predicate; semantic-pragmatic parameters, such as meaning of the construction and communicative function; and contextual elements, such as, e.g., sentence modality and adverbial complements with scope over dejar de + inf and parar de + inf. Next, we conducted a collostructional analysis to investigate which lexemes are strongly attracted or repelled by the non-finite verb form slot in the construction, both per period and throughout time. This was done through a distinctive collexeme analysis (Gries and Stefanowitsch 2004), which is often used to study one slot in two (or more) similar constructions and is known to be useful when applied to alternating pairs of constructions. We found distinctive collexeme analysis appropriate to be applied to the analysis of diachronic corpus data, as it is a very useful tool for the study of grammaticalizing constructions (Hilpert 2006). The collostructional analysis performed on both periphrastic constructions focused on their non-finite verb slot (Vinf). Hence, the first step in the analysis was semi-automatically identifying the verbs used as infinitives with dejar de and with parar de. For each different type, the number of tokens was counted, including the hapax legomena, i.e., the items that appear only once in the corpus. The statistical significance of the association between the construction and the lexical item in the Vinf slot was calculated by means of a Fisher-Yates exact test, and the results are presented in the form of the negative base-10 logarithm of the p-value. Finally, we also looked at the realized and potential productivity (see Baayen 2009) of both periphrases, paying special attention to the Aktionsart of the attracted or repelled predicates.

3. The Semantics of dejar de + inf and parar de + inf in Contemporary Spanish

The periphrases dejar de + inf and parar de + inf are often described as phasal periphrases (Yllera 1980; Olbertz 1998; Fernández de Castro 1999; Camus 2006a, 2006b; RAE/ASALE 2009). They both express interruptive and continuative meanings. The first ones are exemplified in (2): in (2a), dinner is not over but is interrupted; similarly, in (2b), the interlocutor is asked to interrupt the movement that is causing the speaker to get irritated.

(2)	a.	Dej-ó		de	cenar	porque	lo	llam-ar-on
		leave-ind.pst.3sg		of	dine-inf	because	3sg.m.acc	call-ind.pst-3pl
		‘He stopped eating dinner because someone called him’
	b.	¡Par-a	de	mov-er-te,		por	favor!
		stop-imp	of	move-inf-pron		for	favour
		‘Stop moving, please!’

The interruptive value of these constructions, when the non-finite verb is an accomplishment or an achievement, equates them to negative implicative verbs since the action, when interrupted, does not take place anymore; e.g., dejar de cobrar un impuesto (lit. ‘to stop charging a tax’) means that the tax is no longer being collected. In this way, when the periphrasis falls under the scope of a negative polarity term, an affirmative interpretation can be activated. Thus, the ensemble ‘negative + interruptive periphrasis’ comes to signify the continuation of the action expressed by the verb in the infinitive. In (3a), the periphrases dejar/parar de + inf indicate that a person continues reading a book. The interruptive and continuative meanings require verbs expressing processes, accomplishments, and non-permanent states in the Vinf slot. When achievements or permanent states appear in the Vinf slot, the interruptive and continuative meanings add a nuance of iteration. Thus, in (3b), Sara is asked to stop jumping once and again, and in (3c), Juan thanks his mother repeatedly.

(3)	a.	No	dej-a/par-a		de	le-er	ese	libro
		neg	leave/stop-ind.prs.3sg		of	read-inf	that	book
		‘(S)he doesn’t stop reading that book’
	b.	¡Sara,	dej-a/par-a		de	salt-ar!
		Sara	leave/stop-imp		of	jump-inf
		‘Sara, stop jumping!’
	c.	Juan	no	par-ab-a/dej-ab-a		de	agradec-er	a	su	madre
		Juan	neg	leave/stop-ind.pst-3sg		of	thank-inf	to	his	mother
		‘Juan kept on thanking his mother’

These interruptive and continuative values allow us to characterize the periphrasis parar de + inf. However, dejar de + inf is not so closely linked to the expression of phasal values (see Section 4). In fact, the continuative values have always had a very low frequency of use in the history of the periphrasis, and the more frequent interruptive values come into direct competition with other meanings in which the periphrasis seems to have specialized, namely the assertive ones (4) and those that express the negation of the event designated by the infinitive (5). In assertive meanings, dejar de + inf is used to formulate attenuated assertions, and commissive and directive speech acts in a polite manner. That is, in these cases, the periphrasis is always negated and has either a neutral assertive value (4a), expresses a vehement assertion (4b), or constitutes an attenuated or polite formulation of a request (4c) or a suggestion (4d).

(4)	a.	no	dej-ó	de	advert-ir	que	ten-í-an		[…]
		neg	leave-ind.pst.3sg	of	notice-inf	comp	have-ind.pst-3pl
		‘She noticed that they had […]’ (Alegría, Perros hambrientos, 1939, CORDE)
	b.	no	dej-o	de	conoc-er	que	lo	que
		neg	leave-ind.prs.1sg	of	know-inf	comp	3sg.m.acc	comp
		pid-es		es		bueno
		ask.for-ind.prs.2sg		be.ind.prs.3sg		good
		‘I know that what you ask for is good’ (Menéndez Pelayo, Orígenes de la novela, 1905, CORDE)
	c.	Adiós,	no	dej-e-n		de	avis-ar
		goodbye	neg	leave-subj.prs-2pl		of	notice-inf
		‘Goodbye, please let us know.’ (Ignacio Aldecoa, El fulgor y la sangre, 1954, CORDE)
	d.	No	dej-e-s	de	ir	a	la	Villa	d’Este
		neg	leave-subj.prs-2sg	of	go.inf	to	the	Villa	d’Este
		‘Be sure to go to Villa d’Este.’ (Pedro Salinas, Correspondencia, 1951, CORDE)

Utterances in which dejar de + inf makes it possible to negate the event expressed by the infinitive are usually activated when the verb in the infinitive is an achievement or if the periphrasis is temporally delimited. In (5a), for instance, había dejado de aparecer is interpreted in the sense of ‘she no longer appeared.’ The sense of negation is also activated very often in periphrastic chains. In (5b), for example, no podemos dejar de advertir should be read as ‘we cannot not point out’ (for a detailed exposition of the values of dejar de + inf, cf. Garachana n.d.).

(5)	a.	hab-í-a		dej-ado	de	aparec-er	por	sus	habitaciones
		have-ind.pst.3sg		leave-ptcp	of	appear-inf	for	their	hedrooms
		‘(S)he had ceased to appear in their rooms’ (Martín Virgil, Los curas comunistas, 1968, CORDE)
	b.	no	pod-emos	dej-ar		de	advertir		que	[…]
		neg	can-ind.pst.3sg	leave-inf		of	notice-inf		comp
		‘We cannot fail to point out that […]’ (Malpica, El desarrollismo en el Perú, 1974, CORDE)

4. Qualitative and Quantitative Analysis of the Corpus Data

In this section, we will discuss the frequency of use and the realized and potential productivity of the two periphrases under analysis, both per period, i.e., independently of the other centuries, and throughout time, i.e., establishing a comparison among centuries. Then, we will present the results of the distinctive collexeme analysis of the data from CORDE, and we will interpret the results in relation to the Aktionsart of the lexemes that enter into the Vinf slot and that are significantly preferred by either one construction or the other. Finally, these outcomes will allow us to comment on the semantic evolution of dejar de + inf and parar de + inf in the history of Spanish.

4.1. Frequency of dejar de + inf and parar de + inf in CORDE

In addition to the semantic differences that we have discussed in the previous section, if we look at the absolute frequency of use of dejar de + inf and parar de + inf in the language throughout time, we observe significant divergences (see

Table 1. Tokens of dejar de + inf y parar de + inf in CORDE per century.

Construction	13th	14th	15th	16th	17th	18th	19th	20th
dejar_de + inf	349	219	1268	10,058	5363	2151	5290	6664
parar_de + inf	0	0	0	13	28	7	56	202

below).

In this section, we will present a comparison of the frequency of use of both periphrases throughout the history of Spanish, normalizing the frequency scores to a common base per million words (see Figure 1 below). The results show that, while dejar de + inf reaches very high frequencies of use throughout the entire history of Spanish, parar de + inf moves along with very low relative frequencies, which do not manage to exceed one appearance per million words until the 19th century.

If we look at Figure 1 below, we can see that already in the medieval period, dejar de + inf was used with a frequency that oscillates between 25 tokens per million words in the century in which its presence is lowest and 55 tokens per million words in the 15th century, which marks the moment of maximum use of the periphrasis in that period. Its frequency reaches 201.51 tokens per million words in the 16th century, and the 17th and 18th centuries stand at about 140 appearances per million words each (140.82 in the 17th and 148.68 in the 18th century). At the end of the so-called “First Modern Spanish” (ca. 1675–1825), a slight decrease in the frequency of the periphrasis is detected, which possibly reflects the incipient establishment of parar de + inf in the language.

The high frequency documented in the 16th century must be properly contextualized since it is distorted by the predilection that some authors show for this structure in this period. This is not the place to properly assess the disproportionate increase in the normalized frequency of dejar de + inf in the 16th century, but it is significant that among the group of ‘anonymous authorships,’ we have documented the periphrasis 546 times only in the category ‘long stories.’ Works by a known author in which the periphrasis is well documented include, in addition to long narratives, historiographical texts (which in the 16th century were stylistically close to long narratives) and religious texts. These findings suggest that the high frequency of use of the periphrasis in the 16th century may be due to stylistic preferences. We observed that individual authors such as Santa Teresa de Jesús (435 tokens), Fray Luis de Granada (306 tokens), Antonio de Torquemada (310 tokens), and Gonzalo Fernández de Oviedo (558 tokens) tend to use dejar de + inf much more frequently than other authors in the same or other time periods.

It could be assumed that the high frequency of use of periphrasis in the works of these authors could be due to the fact that the CORDE contains a greater number of texts by these particular authors than by others. However, if we compare, for example, the use of dejar de + inf in two contemporary authors such as Santa Teresa de Jesús and Miguel de Cervantes, we observe that this high frequency of use of the periphrasis is not related to the length of the texts by these authors. Indeed, Santa Teresa de Jesús uses dejar de + inf on 435 occasions, and, in CORDE, the texts by this author make up a total of 560,405 words, which means a frequency of 776 tokens per million words. Instead, Miguel de Cervantes resorts to dejar de + inf on 45 occasions, and in CORDE, his works represent a total of 1,449,640 words, which means a frequency of use of 31 tokens per million words. These data seem to prove that the use of periphrasis obeys the rhetorical tendencies of the 16th century. In fact, as we will point out in the discussion section, this periphrasis was progressively more and more used within specific constructions, such as auxiliary chains. Furthermore, after the 16th century, this periphrasis tended to be used with non-aspectual meanings for the expression of pragmatic-discursive meanings, particularly for mainly assertive, commissive, and directive speech acts. It is not surprising, given the communicative relevance of these types of speech acts, that at one point in the history of Spanish, writers used this construction profusely for stylistic purposes.

The normalized frequency of parar de + inf presents a construction that is poorly consolidated in the language, at least in the written language recorded in CORDE. The low frequency of use of parar de + inf is reflected in Figure 2, adapted from Garachana (2021), which highlights the modern nature of the construction.

According to the data contained in Figure 2, the progressive expansion of parar de + inf in Spanish started in the 16th century. During the Spanish Golden Age (16th–17th c.) and the First Modern Spanish (ca. 1675–1825), the frequency of parar de + inf did not reach one token per million words. The 19th century shows a slight rise in the use of construction, which should not have been unrelated to the fact that, at this time, some features of orality are emerging in the written language, a process possibly favored by the rise of literary realism and naturalism, which exalted the reflection of everyday language in texts. However, the first moment in which a significant change in the frequency of use of the construction is observed is in the last quarter of the 20th century, when the normalized frequency of parar de + inf reaches four tokens per million words. In the second half of the 20th century, we documented frequencies between 4 and 7 occurrences of parar de + inf per million words.

4.2. Productivity of the Constructions dejar de + inf and parar de + inf

The comparative study carried out required considering the syntactic productivity of dejar de + inf and parar de + inf. Specifically, we measure the potential of both constructions to attract new verbal types to the Vinf slot, i.e., we measure the extensibility of both constructions. This is a question that has already been addressed in existing publications on the history of the Spanish language (Rodríguez Molina 2004, 2007; Garachana 2016, 2022). In these works, it was pointed out that the key to the establishment of a periphrasis in the Spanish grammatical system lies in the gradual incorporation of new verbs into the verbal construction. We will resume this line of studies here and combine it with a focus on productivity in order to be able to properly quantify the extensibility of the verbal periphrases under study and their productivity at different moments in the history of Spanish.

If we look at the realized productivity of both periphrases, i.e., the number of word types of category C in a corpus of N tokens, we observe that parar de + inf is more productive than dejar de + inf, as

Table 2. Realized productivity of dejar de + inf and parar de + inf in the history of Spanish.

Construction	13th	14th	15th	16th	17th	18th	19th	20th
parar de + inf	0	0	0	0.92	0.85	1	0.48	0.37
dejar de + inf	0.23	0.29	0.23	0.07	0.11	0.22	0.13	0.12

shows. According to Figure 1 and Figure 2 above, dejar de + inf has a higher frequency of use than parar de + inf. However, that does not translate into greater productivity.

Meanwhile, as Baayen (2009, p. 905) points out, there are aspects of productivity that are not well represented by a category’s realized productivity, as a high realized productivity “does not imply that its […] potential productivity will be high as well.” Thus, we also calculated the potential productivity of both periphrases per century. The potential productivity of a rule “is estimated by its hapax legomena in the corpus divided by the total number of its tokens N(C) in the corpus: P = V(1, C, N)/N(C)” (Baayen 2009, p. 902) and expresses the growth rate of the lexemes of the category C itself. The results obtained, presented in

Table 3. Potential productivity of dejar de + inf and parar de + inf in the history of Spanish.

Construction	13th	14th	15th	16th	17th	18th	19th	20th
parar de + inf	0	0	0	0.84	0.75	1	0.39	0.22
dejar de + inf	0.12	0.19	0.12	0.03	0.04	0.12	0.06	0.05

below, show that the potential productivity of the periphrases under study reveals an inversely proportional relationship with their normalized frequency. Furthermore, this finding supports the claim that the potential productivity is inversely proportional to the degree of grammaticalization of grammatical constructions (Bybee 2003; Bybee and Torres-Cacoullos 2009; Torres Cacoullos 2012; Aaron 2010; Copple 2011).

The grammaticalization of a verbal periphrasis implies, among other changes, the progressive elimination of restrictions that affect verbs that may appear in the position of the non-finite verb form (vid. Garachana 2017). That is, the more grammaticalized a construction is, the higher the number of types that can function in such a position, so that the possibility that new verb forms appear in this position will eventually be increasingly limited. This is precisely what can be seen in Table 3: The potential productivity of parar de + inf is much higher than that of dejar de + inf. However, with the course of the centuries, we also observe that this potential productivity becomes gradually more and more limited, which is symptomatic of the progressive consolidation of parar de + inf in the Spanish grammar. Even so, the potential productivity of parar de + inf was still higher than that of dejar de + inf in the first centuries in which this periphrasis was used in the language. From the perspective of productivity, these data, which highlight the weakest settlement and fixation of parar de + inf in the language, offer us valuable information to respond to the discussions that remain in grammatical studies in relation to deciding whether parar de + inf is a verbal periphrasis or a ‘semi-periphrastic construction.’ The restrictions that are detected in relation to the verb types that may appear as Vinf connect directly with the potential productivity of the periphrasis. The grammatical fixation of dejar de + inf does not raise doubts since the lexical restrictions that affect the verbs that function as infinitives are very limited. However, that means that its potential productivity is lower since, in its expansion as verbal periphrasis in Spanish, it has already reached most of the verb types. Parar de + inf, on the other hand, still shows a greater capacity to admit new verb types precisely because it has advanced less in its grammaticalization process.

Besides, we also observed that, if we look at the tokens of both constructions together, out of the 1655 different types documented, only 65 of them are shared by the two constructions, which represents an extremely low proportion. This suggests that the two constructions present a very limited overlap in their Vinf slots, not only in terms of semantic fields but also in terms of specific lexemes. To disentangle this issue, we now move on to the structural study.

4.3. Distinctive Collexeme Analysis

The quantitative analysis of the CORDE data includes a distinctive collexeme analysis. In this subsection, we will present and discuss the results of such an analysis. We looked at the dataset from two perspectives: (i) ignoring diachrony, we were able to check which lexemes (Vinf) are attracted by either construction in general terms (see Section 4.3.1); and (ii) dividing the dataset into periods, we could check which lexemes are attracted by one or the other construction in each period, and, in turn, we could review the attraction of each of these lexemes in all periods (see Section 4.3.2). A methodological sidenote is in order at this point. The values provided are based on the code proposed by Levshina (2015) and are presented as a base-10 logarithm of the p-value of the Fisher-Yates exact test. The value is positive when the occurrences are higher than expected, i.e., there is a preference for dejar de, and negative when they are lower than expected, i.e., there is a preference for parar de. Absolute values higher than 1.3 are classified as instances of ‘categorical preference’ and they are indicated with blue lines on the graphs presented in this subsection. For example, a graph showing verbs with preferences in the 20th century shows in blue the trajectory through the centuries of those verbs with <−1.3 or >1.3 in the 20th century, regardless of their values in the other centuries. Let us now present the findings of the quantitative analysis from the two above-mentioned perspectives.

4.3.1. In General

Table 4. n of lexemes that prefer one or the other construction, or none, in the CORDE.

Preference	Quantity
dejar de	5
none	1594
parar de	64

shows the n of lexemes (Vinf) that prefer one or the other construction (collostructional value ‘collStr’ is higher than 1.3 or lower than −1.3), or none, throughout the centuries. We can observe that most of the lexemes do not prefer either variant.

Next, we can observe the lexemes that prefer dejar de + inf and their collStr value in

Table 5. Lexemes that prefer dejar de + INF and collStr value.

Lexeme	collStr
ser ‘to be’	24,231
tener ‘to have’	4342
hacer ‘to do’	3514
haber ‘to have’	1782
ver ‘to see’	1609

.

It is interesting to notice that the lexemes that prefer dejar de + inf in the dataset are mostly states, contrary to what has been claimed in the literature (cf. Olbertz 1998; Camus 2006a). Deja de ser tan (insistente), for instance, means ‘give up this habit or way of being (insistent).’ In this sense, we can argue that dejar de + inf is often used to characterize the entity, as in No deja de ser paradógico, which means Es paradógico, ‘It’s paradoxical’ or No dejas de tener razón, which means Tienes razón, ‘You’re right.’

Now,

Table 6. Lexemes that prefer parar de + inf and collStr value.

Lexeme	collStr	Lexeme	collStr	Lexeme	collStr	Lexeme	collStr
contar ‘to tell’	−45,805	quejarse ‘to complaint’	−3034	refregar ‘to scrub’	−2012	vocear ‘to holler’	−1713
remover ‘to stir’	−17,019	subir ‘to go up’	−2855	retroceder ‘to move back’	−2012	trabajar ‘to work’	−1649
llover ‘to rain’	−9892	echar ‘to throw’	−2336	roznar ‘to bray’	−2012	agitar ‘to shake’	−1539
hablar ‘to speak’	−8533	mover ‘to move’	−2207	trajinar ‘to bustle’	−2012	regañar ‘to scold’	−1539
hervir ‘to boil’	−8040	bailar ‘to dance’	−2131	trapear ‘to mop’	−2012	revolotear ‘to flutter’	−1539
revolver ‘to stir’	−6187	acumular ‘to accumulate’	−2012	tronar ‘to thunder’	−2012	roltar ‘to let go’	−1539
beber ‘to drink’	−4876	aporracear ‘to beat’	−2012	zaherir ‘to taunt’	−2012	trinar ‘to trill’	−1539
correr ‘to run’	−4301	chancearse ‘to jest’	−2012	clamar ‘to cry out’	−1986	vomitar ‘to throw up’	−1539
cocer ‘to cook’	−4135	comerciar ‘to trade’	−2012	cantar ‘to sing’	−1933	cegar ‘to blind’	−1416
bullir ‘to boil’	−3551	contrastar ‘to contrast’	−2012	derrocar ‘to overthrow’	−1713	exclamar ‘to exclaim’	−1416
provocar ‘to provoke’	−3551	dejar ‘to leave’	−2012	frotar ‘to rub’	−1713	tragar ‘to swallow’	−1416
rezongar ‘to grumble’	−3551	disparatar ‘to talk nonsense’	−2012	insultar ‘to insult’	−1713	reír ‘to laugh’	−1405
nevar ‘to snow’	−3253	levar ‘to carry’	−2012	lisonjear ‘to flatter’	−1713	dar ‘to give’	−1404
gritar ‘to shout’	−3252	parlotear ‘to chatter’	−2012	mecer ‘to rock’	−1713	acrecentar ‘to increase’	−1322
tocar ‘to touch’	−3110	rebullir ‘to start moving’	−2012	preparar ‘to prepare’	−1713	fregar ‘to scrub’	−1322
protestar ‘to protest’	−3034	rechinar ‘to grind’	−2012	retorcer ‘to twist’	−1713	toser ‘to cough’	−1322

presents the lexemes that prefer parar de + inf and their collostructional value. Note that, since the values are calculated on the basis of dejar de, we consider as lexemes that prefer parar de those that demonstrate a significant rejection for dejar de. Therefore, the values in Table 6 are negative.

With parar de, there are 145 types documented in total, and 64 of them show a significant attraction to the construction. Contar, ‘to tell,’ the lexeme most attracted to parar de, appears with dejar de in only 324 instances, which are few if you consider the total number of tokens of dejar de + inf in the corpus. Among the 64 types that show a significant attraction to parar de, we find communication verbs such as tell, speak, provoke, grumble, and yell; predicates related to cooking, such as boil and cook; meteorological verbs, such as rain and snow; and physical contact verbs, such as remove, stir, and touch. Some of these are processes and do not imply completion or culmination. Many of these lexemes denote an activity or movement -they are not stative- and they generally take an animate subject.

4.3.2. Per Period

We applied the same procedure to samples from specific periods. In order to do so, we grouped together the data until the 16th century because before that century, there were no documented examples of the periphrasis parar de + inf in the CORDE. We also grouped together the 17th and 18th centuries to get a more balanced normalized frequency per period. The nineteenth century already marked a noticeable increase in the frequency of use of parar de + inf, which, in turn, increased considerably in the 20th century. These groupings are motivated by the behavior of the periphrasis parar de + inf in the corpus rather than by external factors. The n of lexemes with preferences for either variant or none of them in each period is shown in

Table 7. Lexemes that prefer parar de + inf, dejar de + inf or none per period.

Period	None	parar de	dejar de
until–16th	845	7	0
17th–18th	809	16	1
19th	716	18	1
20th	808	35	3

. Throughout time, both periphrases seem to ritualize certain types. And with a higher degree of grammaticalization, each seems to prefer more predicates over time.

As can be observed in Table 7, no lexeme shows a significant preference for dejar de before the 17th century. From this century on, ser ‘to be’ becomes the lexeme most closely linked to dejar de + inf. We also see that tener ‘to have’ and existir ‘to exist’ are only documented with dejar de in the 20th century. These three lexemes are very frequent with dejar de, and they never occur with parar de (see

Table 8. Lexemes that prefer dejar de per period.

Lexeme	collStr	Frequency	Period
ser ‘to be’	2375	1158	17th–18th
	6283	1247	19th
	24,910	1682	20th
tener ‘to have’	3414	276	20th
existir ‘to exist’	1821	156	20th

and

Table 9. Lexemes that are only documented with parar de in particular periods.

Period	Lexeme	Frequency	collStr
until–16th	bullir ‘to boil’, levar ‘to carry’, lisonjear ‘to flatter’	1	−2900
17th–18th	aporracear ‘to beat’, contrastar ‘to contrast’, derrocar ‘to overthrow’, rechinar ‘to grind’, tronar ‘to thunder’, vocear ‘to holler’	1	−2334
19th	disparatar ‘to talk nonsense’, mecer ‘to rock’, refregar ‘to scrub’, regañar ‘to scold’, tragar ‘to swallow’, zaherir ‘to taunt’	1	−1980
20th	acumular ‘to accumulate’, cegar ‘to blind’, chancearse ‘to jest’, comerciar ‘to trade’, dejar ‘to leave’, insultar ‘to insult’, maldecir ‘to curse’, parlotear ‘to chatter’, penar ‘to suffer’, preparar ‘to prepare’, rebullir ‘to start moving’, retroceder ‘to move back’, roznar ‘to bray’, soltar ‘to release’, trajinar ‘to bustle’, trapear ‘to mop’	1	−1531
20th	nevar ‘to snow’, provocar ‘to provoke’, quejarse ‘to complaint’, rezongar ‘to grumble’	2	−3065

).

In the same way, we can list the verbs that prefer parar de, which are, in general, infrequent in the Vinf position of either construction. If one looks at the absolute frequency in the corpus, they tend to occur more in the context of dejar de, just because the construction itself is much more frequent. Since there are 68 lexemes that prefer parar de over dejar de in at least one period, we will present the data in parts. First, Table 9 shows the lexemes that only occur in the construction parar de + inf in a certain period and are not documented with dejar de + inf. Except for four lexemes in the 20th century subcorpus, which occur twice with parar de, the others are only documented once. Throughout time, we have observed a gradual increase in the use of communication verbs with negative connotations, which are marked in bold here.

Table 10. Lexemes that only occur with parar de once in certain periods but prefer it.

Period	Lexeme	Frequency with dejar de	collStr
until–16th	acrecentar ‘to increase’	2	−2423
	revolver ‘to stir’	3	−2298
	tañer ‘to toll’	14	−1727
17th–18th	danzar ‘to dance’, recoger ‘to pick up’	1	−2034
	moler ‘to grind’, porfiar ‘to persist’	3	−1735
	agradecer ‘to thank’, combater ‘to fight’	6	−1495
	tentar ‘to tempt’, volar ‘to fly’	7	−1438
19th	bullir ‘to boil’, exclamar ‘to exclaim’, hervir ‘to boil’, subir ‘to go up’, vomitar ‘to throw up’	1	−1681
	añadir ‘to add’	2	−1507
	practicar ‘to practice’	3	−1384

shows the lexemes that, statistically, prefer parar de in a certain period, although we only documented one occurrence of them in the construction. The frequency column indicates the number of occurrences of dejar de in that period, for comparison.

Finally,

Table 11. Lexemes that prefer parar de in certain periods and occur more than once.

Period	Lexeme	Frequency with parar de vs. dejar de	collStr
until–16th	trabajar ‘to work’	2 vs. 52	−2690
17th–18th	hablar ‘to speak’	3 vs. 44	−2890
	beber ‘to drink’	2 vs. 11	−2802
19th	contar ‘to tell’	19 vs. 13	−30,602
	correr ‘to run’	5 vs. 17	−5617
	echar ‘to throw’	4 vs. 19	−4084
	clamar ‘cry out’	2 vs. 1	−3493
	hablar ‘to speak’	4 vs. 60	−2366
20th	contar ‘to tell’	35 vs. 12	−44,326
	remover ‘to stir’	10 vs. 3	−12,986
	hablar ‘to speak’	18 vs. 85	−9108
	llover ‘to rain’	10 vs. 23	−7700
	hervir ‘to boil’	5 vs. 5	−5329
	revolver ‘to stir’	4 vs. 5	−4088
	dar ‘to give’	11 vs. 94	−3618
	cocer ‘to cook’	3 vs. 4	−3094
	beber ‘to drink’	5 vs. 31	−2424
	protestar ‘to protest’	2 vs. 2	−2304
	gritar ‘to shout’	3 vs. 10	−2239
	decir ‘to say’	5 vs. 37	−2133
	subir ‘to go up’	2 vs. 3	−2090
	tocar ‘to touch’	5 vs. 55	−1509
	bailar ‘to dance’	3 vs. 22	−1446

shows the lexemes that, statistically, prefer parar de in a certain period, occur in that construction more than once, and also occur with dejar de. The frequency column indicates the occurrences with parar de vs. those with dejar de.

Figure 3 below presents the level of attraction of lexemes with significant attraction to any of the two constructions over time. Each line represents a lexeme that prefers either parar de or dejar de in any period. In order to show the evolution of the lines more clearly, we show each century, from the 17th century onwards, separately. If a lexeme is positioned between the dotted lines, the preference is not significant. The most striking lexemes, with extraordinary preference for either construction, are indicated with the corresponding label: ser ‘to be’ in the case of dejar de, and contar ‘to tell’ in the case of parar de. Interestingly enough, the combination of parar with the predicate contar is so frequent that it has ended up being lexicalized in the locution y para/pare de contar (lit. ‘and stop of to tell/count’) used to put an end to an enumeration or narration.

As these two extraordinary lexemes dominate the graph in Figure 3, Figure 4 and Figure 5, focus on smaller values, highlighting and labeling lexemes that prefer either one construction or the other in the first and last periods, i.e., until–16th century and the 20th century, respectively, so as to portray the evolution of the preferences.

Figure 4 is focused on the 16th century, a key period for the history of the two periphrases under study since it is the moment in which parar de + inf is defined as a construction in Spanish. At this time, as is to be expected given its recent introduction into the language, this periphrasis shows a particular affinity for certain verbal forms that express processes or achievements, such as tañer ‘to toll,’ revolver ‘to stir,’ acrecentar ‘to increase,’ trabajar ‘to work,’ llevar ‘to carry,’ lisonjear ‘to flatter’ and bullir ‘to boil.’ By contrast, the most well-established construction in the language, dejar de + inf, shows no particular constructional preference.

Figure 5 is focused on the use of dejar de + inf and parar de + inf in the 20th century. Unlike what we saw in Figure 4, Figure 5 shows a marked tendency for the first of these periphrases to be used with states (specifically, with three verbs: ser ‘to be,’ tener ‘to have,’ existir ‘to exist’). In turn, parar de + inf shows greater constructional diversity, which is reflected in the dense network of combinations made up by the different verbs that appear in the Vinf slot and that show a statistically significant attraction to the construction. Notice that the label that goes with the upper line corresponds to ser ‘to be,’ which is strongly attracted to dejar de + inf, and the one that goes with the lower line corresponds to contar ‘to tell,’ which is strongly attracted to parar de + inf. These are the two extraordinary lexemes that dominate the graph in Figure 3 above.

Once again, the collostructional study per period offers data that coincides with those resulting from the analysis of the Aktionsart of the verbs that appear in the Vinf slot of the periphrases dejar de + inf and parar de + inf. In addition, these results allow for drawing relevant conclusions on the semantic evolution of both periphrases. As can be seen in Figure 6, if we pay attention to the distribution of verb types by centuries, we observe that processes and accomplishments dominate almost exclusively throughout all evolutionary stages of parar de + inf.

Unlike parar de + inf, dejar de + inf shows a tendency to combine, always exclusively, with processes and accomplishments up to the 15th century, where we observe a clear rise of achievements, as Figure 7 shows. From the 16th century onward, achievements and states exceeded processes and accomplishments, with a growing disproportion from the 17th century onward. This situation continued until contemporary times, despite the fact that, in the 20th century, a recovery of the processes was documented.

The collostructional analysis, together with the study of the Aktionsart of the verbs that appear in the infinitive position, allowed us to draw interesting conclusions about the emergence of parar de + inf in Spanish as well as about the possible analogical influence that dejar de + inf could have exerted. The interruptive meaning of dejar de + inf, the most common in the medieval period, could have attracted the construction with parar, which might also have experienced the influence of other interruptive and continuative periphrases such as cesar de + inf. However, the results offered by the collostructional analysis and the analysis of the Aktionsart of the predicates in the Vinf slot show that this influence had to occur at a schematic level since the combinatorial patterns of parar de + inf present it as a construction with its own identity, which does not replicate the model of the most frequent interruptive periphrasis in Spanish. If the influence had occurred at the most saturated level of the construction, parar de + inf would have replicated the collostructional patterns of dejar de + inf, and no significant differences would be detected between the two periphrases regarding the type of verb that appears in the non-finite position.

On the collostructional side, we observe that already in the 16th century, at the moment when parar de + inf emerges in Spanish, there are lexemes that are only documented with this construction (Table 9) or that prefer it (Table 10 and Table 11). Therefore, we can claim that the emergence of parar de + inf in Spanish does not follow the collostructional patterns of dejar de + inf. On top of that, in the 16th century, this construction did not follow the model of dejar de + inf in terms of the Aktionsart of the verbs occupying the Vinf slot. As shown in Figure 8, even at this time in the history of Spanish, both periphrases show clearly differentiated patterns.

The preference of parar de + inf for processes and accomplishments, attested from its earliest documentation, makes it possible to claim that parar de + inf only competes with dejar de + inf in their uses as phasal periphrases, given that the interruptive and continuative meanings have always characterized the periphrasis; cf. Figure 9, adapted from Garachana (2021).

Interestingly, after the introduction of parar de + inf in the paradigm of periphrases, a significant change in the use of dejar de + inf occurs, as shown in Figure 10. In fact, starting in the 16th century, we observe the collapse of the interruptive meanings that, during the medieval period, had been the most characteristic of this periphrasis. From the 16th century on, the periphrasis was mainly used with assertive values or to deny the event expressed by the infinitive. This semantic evolution is linked to the regression of atelic predicates in the Vinf slot and allows us to advance the hypothesis that, since there is a new piece in the system of interruptive periphrases, which was added to the already consolidated cesar de + inf (see Yllera 1980 for more information), this would allow dejar de + inf to expand towards other areas of meaning, linked to stylistic and discursive needs. In the same way, the recovery of the interruptive meanings of the periphrasis in the 20th century is related to the increase in processes and can be explained by an analogical influence on the part of parar de + inf at the moment in which this periphrasis is consolidated in the language.

5. Discussion

The findings of the analysis of the productivity of both periphrastic constructions, as well as those that emerge from the collostructional analysis carried out, show that dejar de + inf and parar de + inf behave in particular ways that do not coincide with the descriptions contained in the existing literature on these periphrases. As was stated in the introduction, grammars readily admit the periphrastic character of dejar de + inf, while parar de + inf is usually subject to interpretations that oscillate between those who accept its status as a periphrasis and those who consider that it functions as a semi-auxiliary verb. However, the analysis of the use of these periphrases based on an extensive corpus forces us to review these statements.

The data show that the realized and potential productivity of parar de + inf is higher than that of dejar de + inf. In fact, the realized productivity of dejar de + inf is always lower than that of parar de + inf. And if we compare the contemporary realized productivity of parar de + inf (0.37) with that which dejar de + inf presented in the medieval period (0.23 in the 13th–15th centuries; 0.29 in the 14th century), that of parar de + inf is higher. Similarly, the potential productivity of parar de + inf in the 20th century (0.22) remains above that of dejar de + inf in the medieval period (0.12 in the 13th–14th centuries; 0.19 in the 14th century), when its use in the language must still have been recent and an expansion of the periphrasis could be expected.

As noted earlier (cf. Section 4.2), we could argue that this behavior is due to a more advanced stage of grammaticalization of dejar de + inf, which already in the medieval period would have advanced more in the language than parar de + inf. Indeed, the most grammaticalized periphrases have reached higher levels of lexical expansion than those that are not yet established in the language; an incipient grammaticalization determines that the possibilities of expansion turn out to be greater. Once a periphrasis is fixed for the expression of temporal, aspectual, or modal values, it usually presents few restrictions regarding the verb types that can appear in the infinitive position. If they exist, such restrictions have to do with the meaning that the periphrasis expresses. However, in the case of dejar de + inf, perhaps this lower productivity may be due to other grammatical and discursive motivations, especially if we consider that, according to Garachana (n.d.), this periphrasis shows few signs of grammaticalization throughout the history of Spanish, and its use seems to be closely linked to particular stylistic and rhetorical motivations from early dates.

Interestingly, Yllera (1980) already highlights it as the most frequent of the medieval interruptive periphrases (Yllera 1980, p. 194) and points out its scarce employment with continuative value (Yllera 1980, p. 206). This behavior separates dejar de + inf from the rest of the medieval interruptive periphrases mentioned in Yllera (1980), since these were mostly used in negative modality sentences to denote the continuity of the action expressed by the infinitive. In addition, unlike those other periphrases, dejar de + inf could be used to indicate the negation of the event expressed by the infinitive (Yllera 1980, p. 215) and to formulate more or less forceful assertions.

In this way, dejar de + inf was already outlined in the Middle Ages as a periphrasis with its own defining characteristics, which do not fit what is indicated in the bibliography on the subject. Along the same lines, its preference for combining with infinitives that express either achievements or states is contrary to what is expected of an interruptive (or continuative) periphrasis. Certainly, often the use of verbs that express achievements or states in the infinitive position of dejar de + inf allows adding a notion of repetition or iteration of the event (see Section 3); however, in most cases the construction expresses an assertion or negates the event expressed by the infinitive. To this propensity to combine with verbs that do not express actions that can be interrupted or continued, we must add the fact that our collostructional analysis clearly indicates that the lexemes that prefer dejar de + inf are above all states (specifically, these are the verb forms ser ‘to be,’ tener ‘to have,’ haber, ‘to have’). Among the accomplishments and the processes, only hacer ‘to do’ and ver ‘to see’ show a tendency to be used in the periphrasis.

It is interesting that the link between dejar de + inf and concrete verb types began in the 17th century—remember that, prior to that century, no verb form seemed to show a particular preference for this periphrasis—, since, in principle, a grammaticalized periphrasis should have a somewhat homogeneous behavior with all kinds of verb forms. However, starting in the 17th century, dejar de + inf began to specialize on a few verb types, which gave it a certain formulaic quality, atypical of a verbal periphrasis. On the other hand, parar de + inf shows that the verbs that prefer it are much more varied (vid. Table 6), and this group of verbs even includes meteorological verbs, which is usually considered a proof of the grammaticalization of verbal periphrases (Garachana 2017, p. 43). Furthermore, if we observe the collostructional preferences of parar de + inf per period, we observe that the number of verb forms increases throughout time. That is, parar de + inf shows greater collostructional flexibility than dejar de + inf.

Thus, the collostructional behavior of parar de + inf and dejar de + inf offers contrasting results regarding the very consideration of these constructions as verbal periphrases. The results of our analysis show that the form that best fits the definition of verbal periphrasis is, paradoxically, the construction that has given rise to the greatest debate about its periphrastic or non-periphrastic character and status. Indeed, parar de + inf has higher realized and potential productivity than dejar de + inf and manifests fewer restrictions that affect the types that can combine with it. On the other hand, dejar de + inf appears to be a more rigid construction than parar de + inf, with a certain tendency towards routinization with particular verb types.

At this point, it is worth asking why dejar de + inf tended towards routinization starting precisely in the 17th century. The answer probably lies in rhetorical and stylistic preferences. Precisely, this was the moment in which there was a higher use of periphrases within auxiliary chains, for instance, ‘No puedo dejar de admitir’, ‘I must (lit. cannot stop) admitting’. As Garachana (n.d.) shows, in this century the use of periphrastic chains in CORDE amounts to 18.50% of the total number of uses of verbal periphrases, and the percentages for subsequent centuries are: 13.18% in the 18th century, 12.05% in the 19th century, and 9.55% in the 20th century. Significantly, within these auxiliary chains, the most frequent verb is ser, ‘to be,’ which, in the 17th century, appeared in 13.66% of the periphrastic chains. This percentage rises to 21.27% and 21.83%, respectively, in the 18th and 19th centuries, and then drops to 13.33% in the 20th century (Garachana n.d.). These findings show that in a significant number of cases, the use of the periphrasis is associated with a more complex construction in which the verb ser, ‘to be’ occupies a prominent place. Now, this chained construction does not express aspectual meanings but negates the meaning of the event expressed by the infinitive.

These values are usually not the most frequent ones for verbal periphrases in Spanish, which are usually associated with the expression of temporal, aspectual, or modal meanings. Moreover, dejar de + inf is frequently used for the expression of assertive, commissive, and directive speech acts. At this point, it is worth asking whether dejar de + inf can really be considered, in all its meanings, a representative member of the category of verbal periphrases. Its restrictions related to its productivity, the strong affinity it shows for only a few verb forms, and its propensity to be used in auxiliary chains place it on the margins of the category of periphrasis. Only its uses as an interruptive and continuative periphrasis allow its inclusion in the category as a full member. The rest of its values force us to reflect on the type of category to which it should be assigned.

This issue goes beyond the discussion about the grammaticalization of verbal constructions since it affects the values they express. The most recent debates on this matter (see García Fernández et al. 2006; Carrasco Gutiérrez 2008; García Fernández and Carrasco Gutiérrez 2008) propose characterizing the constructions acabar + ger, acabar por + inf, alcanzar a + inf, comenzar + ger, comenzar por + inf, empezar + ger, empezar por + inf, llegar a + inf, pasar a + inf, terminar + ger, terminar por + inf, and venir a + inf as discourse markers. The decisive argument of these authors, counterargued by Olbertz (2007), rests on the fact that these periphrases do not express temporal, aspectual, or modal meanings, as is characteristic of the category, but rather, on the contrary, function as discourse structuring devices. Contrarily, Olbertz (2007) emphasizes the formal features of these periphrases to defend their maintenance within the category. In our opinion, both Carrasco and García Fernández, as well as Olbertz, are correct. On the one hand, the aforementioned periphrases work with values that are more typical of information structuring devices than those of verbal periphrases. On the other hand, the grammatical functioning of these structures is characteristic of verbal periphrases. Thus, if we also consider the characterization of dejar de + inf advanced here, it is necessary to review the very concept of periphrasis and to qualify the types of meanings that can be expressed by these grammatical constructions.

Considering all of the above, the evolution of these two periphrases in Spanish shows a completed process of constructionalization since their first documentation. The process, however, seems to have advanced further in the case of dejar de + inf, which has reached a stage in which it has surpassed the expression of aspectual meanings to encompass pragmatic meanings linked to the expression of assertive, commissive, and directive speech acts. The evolution of parar de + inf would have been stopped in the phasal (interruptive or continuative) meanings.

As

Table 12. The constructionalization process of dejar de + inf.

	Lexicon		Grammar		Pragmatics
Dejar de+ INF	STAGE 1 Etymon	STAGE 2 Bridging context	STAGE 3 Interruptive Verbal Periphrasis	STAGE 4 Continuative Verbal periphrasis	STAGE 5 Assertive and negative values
Semantics/pragmatics	dejar = ‘place sth. somewhere’ Dejó los libros sobre la mesa	Old Spanish: ‘interrupt (stop) a story at a point’ Dexa aquí la estoria de fablar	‘to interrupt an event’ Dejó de sufrir	‘to maintain an event’ No dejó de fumar ni un momento	‘To say or do something’ No dejes de ir No puedo dejar de admitir
Syntax	suject [+animate] + dejar + DO + locative complement	Suject [+/− animate] + dejar + locative complement (ø/a/de) + PP (de/a INF)	dejar de + INF	NEG dejar de + INF	AFF/NEG dejar de + INF Auxiliary verb + dejar de + INF
		S XIII & XIV. INF that mainly express ‘telling a story’ (decir ‘to say’, contar ‘to tell’, hablar ‘to speak’)	Other types start consolidating (processes, accomplishments, non-permanent states)	Processes, accomplishments, non-permanent states	inf = achievement or state Often, a temporal expression delimits the duration of the infinitive. Dejar de + inf is often preceded by another auxiliary verb (auxiliary chains)

shows, the constructionalization of dejar de + inf starts from the lexical transitive verb dejar, ‘to leave.’ This verb subcategorizes a direct object and a locative complement and means the action of depositing an object at a certain location, as in Dejar algo en un sitio, ‘place something somewhere’ (STAGE 1). Although there are documentations of the use of dejar as an auxiliary verb already in the 13th century, the frequent use of the locution Dexa aquí la estoria de fablar en X, ‘the narration of a certain matter is interrupted’ leads Garachana (n.d.) to propose the hypothesis that the periphrasis originates from such a locution, in which the verb dejar has a metaphorical meaning. Indeed, in Dexa aquí la estoria de fablar en ello, dexa, ‘present indicative of the verb dejar’ means ‘the abandonment of a story at a certain point in the narrative.’ It is, then, a medieval structure for changing the topic. This locution would have functioned as a bridging context: in it, dejar no longer requires an animate subject, and it often subcategorizes a deictic adverb (aquí ‘here’) that indicates the place in the story where it is abandoned, and what is deposited (or abandoned) metaphorically in that place is no longer an object (STAGE 2). From these bridging contexts, in which the interruption of a story is expressed, the interruptive aspectual meaning of the construction would have been reached (STAGE 3), where dejar functions as an auxiliary verb, combined mainly with processes and accomplishments in the Vinf slot, and the structure dejar de + inf is already fully constructionalized.

However, the evolution did not stop at this point, since when using dejar de + inf in contexts of negative modality, it acquires a continuative meaning (STAGE 4). Later, the periphrasis will be used to express assertions and the negation of the event expressed by the infinitive, usually in the formulation of assertive, commissive, or directive speech acts. The availability of these values presupposes a formal change in the construction, as they are triggered when the Vinf slot is occupied by states and achievements, often in auxiliary chains. In addition, often the event expressed by the verb is temporally delimited by a temporal expression (STAGE 5). Table 12 summarizes this constructionalization process, which starts in the lexicon and concludes in pragmatics.

The constructionalization of parar de + inf is represented in

Table 13. The constructionalization process of parar de + inf.

	Lexicon		Grammar
Parar de+ INF	STAGE 1 Etymon	STAGE 2 Supporting construction	STAGE 3 Interruptive Verbal periphrasis	STAGE 3–4 Continuative Verbal periphrasis
Semantics/pragmatics	parar ‘put in a position’ > ‘to stop physically’ Los comeres delant gelos paravan (Cid) Paravas delant al campeador	Dejar de + INF	‘to interrupt an event’ (nuance of annoyance; directive speech acts) Para de moverte	‘to maintain an event’ (nuance of annoyance) No paran de malestar
Syntax	parar (tr. V.) + OD + locative complement parar (intr. V.) + locative complement		parar de + INF	NEG parar de + INF

below. The starting point is found in the verb parar, a form that in the medieval period expressed both locative values (parar meant ‘to place something in a position’) and intransitive senses (parar = to stop) (Corominas and Pascual 1991: s.v. parar). In its locative senses, parar was close to dejar, which could have contributed to the use of parar as an auxiliary verb. It has been proven that in Spanish, it is common for the lexical synonymy between two verbs—one of which also functions as an auxiliary verb—to originate processes of analogical attraction that conclude with the auxiliary use of the other verb (Garachana 2011; Garachana and Rosemeyer 2011). This process would have been activated by the intransitive sense of parar, which expresses the interruption of a movement and, therefore, is linked to the interruptive aspectual sense that the periphrasis expresses in certain contexts. As in the case of dejar de + inf, the use of parar de + inf in negative modality contexts would motivate the development of continuative meanings.

6. Conclusions

The main contribution of the preceding discussion lies in having verified our hypothesis that the introduction of parar de + inf in the paradigm of verbal periphrases did not imply a copy of existing models but rather that the periphrasis followed its own constructional pattern. In addition, we were able to confirm that this introduction of parar de + inf into the system of verbal periphrases had collateral effects on the use of dejar de + inf, which, in turn, came to be used with values less linked to phasal meanings, which had been its defining values during the medieval period.

The conclusions that can be drawn for the study of parar de + inf and dejar de + inf are highly relevant and connect with findings of previous productivity research that has focused on quantitative frequency measures (see Hilpert 2013; Perek 2020). The study of their normalized frequency of use and of their realized and potential productivity, as well as the collostructional analysis carried out, allow us to claim that parar de + inf is consolidating in the language as a productive periphrasis, which attracts an ever-growing group of verb forms in the non-finite verb slot. In addition, its meaning is associated with the aspectual values that are typical of interruptive and continuative periphrases. On the other hand, dejar de + inf from the medieval period onwards appears as an increasingly less productive and more rigid construction than parar de + inf, with a certain tendency towards routinization with particular verb types. In addition, it is gradually being less and less used with interruptive aspectual values (continuatives were never especially frequent), and it is emerging as a construction specialized in the expression of values that are not typical of verbal periphrases. As in the case of the periphrases analyzed by García Fernández and Carrasco Gutiérrez (2008), we find, under a periphrastic form, a construction capable of expressing pragmatic meanings associated with a specific stylistic and communicative will. Thus, despite the fact that parar de + inf is the form that has aroused the greatest reluctance in the literature on periphrases when it comes to including it as a member of the category, an in-depth study based on a large corpus reveals that this construction fits the characteristics of the category even better than dejar de + inf does.

Still, some open questions remain. We wonder about the impact that the type of corpus used has had when it comes to describing the evolution and frequency of the periphrases. The use of parar de is associated with the expression of directive speech acts as well as of annoyance, as in ¡Para de moverte ya! ‘Stop moving already!’. We could assume that the low token frequency of parar de we observed may be partially explained by the kind of corpus that we used, which was only composed of written documents, which tend to avoid situations of communicative proximity. This is obviously only relevant for the centuries for which we have oral data available; however, it would be interesting to look into more recent oral corpora to check whether the observed distribution holds. This disadvantage can be remedied by complementing methodological approaches by, for instance, carrying out experiments such as acceptability rating or sentence completion tasks in order to investigate the extensibility of a construction in more detail. In future studies, we would also like to focus more on the sematic productivity of parar de + inf and dejar de + inf by conducting a fine-grained analysis of the semantic characteristics of the non-finite verbs in these verbal periphrases following the ADESSE classification (see García Miguel et al. 2010). Finally, we are planning on looking more closely at the kind of attraction that near-synonymous constructions exert to explain how the Spanish periphrastic system became structured to prevent constructions from developing the exact same meaning and how it is being recomposed throughout time.