Light Verbs and Syntactic Analyzability in the History of the Galician Language

Abstract

This contribution studies the behavior of the four main general light verbs (LVs) in the history of Galician (dar, facer and haber/ter). The research is structured around the following three fundamental axes: first, we study the evolution, comparison with equivalent full verbs and the morphosyntactic behavior of 26 different LV constructions (with examples that the literature identifies with different degrees of fixation) from medieval to contemporary Galician, all of which form a corpus with 8728 occurrences. Next, we discuss the results of a survey distributed to 162 respondents, which allows an assessment of these LVs from several perspectives, especially syntactic. Finally, we offer an original proposal to measure the degree of syntactic analyzability, based on the quantified review of the various parameters analyzed (of which we also provide a scale, applied synchronically and diachronically) and the results in the specific survey question. We call it Syntactic Analyzability Index (SAI) and, thanks to this index, we obtain an objective scale that projects each example on a gradation that explains the greater or lesser distance or proximity of a construction with LV from freely combined elements or from the most fixed of phrasemes.

For Teresa Bermúdez Montes, in memoriam, always.

1. Introduction

There are many structures in the Galician language that would admit, especially from a historical perspective, an analysis of their variation in terms of syntactic analyzability (see Langacker, 1987, pp. 292–298; Bybee, 2010, pp. 44–56 or Kardela, 2021), but few will probably be as rewarding (because of structural variation, the sheer number of examples, the complexity of the units involved, and obviously because of the richness that results from being able to analyze them across different periods of the history of the language) as the structure we analyze in the present study: constructions with light verbs (LVs) in Galician. Within the Galician linguistic tradition, several studies were recently conducted on LVs, focusing on different aspects. Among them, the following two major structural subtypes have been studied: locutions or phrasemes (Álvarez de la Granja, 2003, 2025; Soto Arias, 2022) and collocations (Paz Roca, 2004; Alvariño et al., 2007; Álvarez de la Granja, 2010; Cotelo, 2017; Martínez Vicente, 2023, 2025). Portuguese grammars devote specific sections to LVs, such as those by Gonçalves and Raposo (2013, §28.2.4), Castilho (2014, §10.2.1.3.3) or Perini (2016, §16).

In descriptions of these structures, in addition to emphasizing their multifaceted formation, “idioms occupy a region in a multidimensional lexical space, characterized by a number of distinct properties: semantic, syntactic, poetic, discursive, and rhetorical” (Nunberg et al., 1994, p. 492), most of the literature endorses the recognition of different construction types with the basic structure LV + noun. In addition to the unconstrained discourse with “construções de verbo pleno […] + nome objeto directo” [“constructions with full verb […] + direct object noun”] (Neves, 2002, p. 235), the following two main subtypes are identified: fixed structures in which the verb and noun form “em conjunto um predicado” [“a predicate as a whole”], traditionally referred to as “locução verbal” [“verbal phrase”], and the LV cases, located between the latter and the unconstrained noun phrases, which can either approximate or distance themselves from them “conforme o grau de ‘gramaticalização’ a que tenham chegado” [“according to the degree of grammaticalization they have reached”] (Neves, 2002, pp. 235–236).

Corpas Pastor (1998, pp. 41–42), using two criteria, “enunciado e fixación” [“utterance and fixation”], identifies the following three sets: collocations (fixed in the norm), locutions (fixed in the system), and phraseological utterances (fixed in speech, the only ones that constitute speech acts).

Alonso Ramos, from the perspective of their meaning, emphasizes the need to distinguish three groups in the sequences of “verbo-nombre tratadas a menudo como locuciones verbales” [“verb-noun commonly treated as verbal phrases”] (Alonso Ramos, 2004, p. 52): the first consists of sequences that are “indivisibles semánticamente” [“semantically indivisable”], the second includes verbs with a more relational meaning, such as “verbos de realización” [“performative verbs”], “verbos fasales” [“phase verbs”], and “verbos causativos” [“causative verbs”] (Alonso Ramos, 2004, p. 52, n. 15); and finally, the third group consists of LVs, verbs that provide “un significado general que produce parte del significado de la clase a la que estos nombres pertenecen” [“a general meaning that produces part of the meaning of the class these nouns belong to”] (Alonso Ramos, 2004, p. 53). For this author, only the first group contains “frasemas completos (o locuciones verbales)” [“complete phrasemes (or verbal phrases)”] (Alonso Ramos, 2004, p. 52), while the other two are analyzable and constitute subtypes within LV constructions. However, the boundaries between them are fuzzy (cf. Section 5).

For the characterization of Galician phraseological units, Álvarez de la Granja (2003, pp. 11–24), who also distinguishes between locutions and phraseological utterances, uses the following four defining features: component fixation, the formation process, conceptual and/or functional unity, and the capacity “to create texts” and relate “to a given situation”. This author thoroughly studies verbal locutions and, within them, establishes the following two major subgroups: rigid ones (with component and syntactic fixation) (Álvarez de la Granja, 2003, pp. 91–126) and flexible ones, which she further subdivides into those that “encerran unha metáfora transparente” [“contain a transparent metaphor”] or conventional transparent metaphor, with a high degree of figurativeness, which can explain their syntactic flexibility (Álvarez de la Granja, 2003, pp. 127–146), and those with a “semantically analyzable” character, which can “xustificar determinados comportamentos no discurso” [“justify their behavior in discourse”] (Álvarez de la Granja, 2003, p. 127).

Soto Arias argues that phraseological units are characterized by their multiverbal nature, fixation, and idiomaticity (Soto Arias, 2022, p. 166). To classify word combinations, Martínez Vicente (2023, p. 180; 2025, pp. 4–5) uses a lexicosemantic criterion based on component fixation, establishing the following two major groups (besides unconstrained combinations and lexical solidarities, with no lexical unit): collocations with LV and phraseological units. The latter exhibit high fixation, lexical unity, and are idiomatic in the form “A + B = C” (e.g., non dar golpe “to do nothing”). Collocations with LV, which have lexical unity and are semi-idiomatic in the form “A + B = B”, are divided in turn into the following three subgroups: less metaphorical (no total restriction, with a concrete “B”, e.g., dar verniz “to varnish”), prototypical (allows number inflection, and “B” is typically abstract, e.g., dar bruídos “to make noises”), and more metaphorical (high fixation, but lower than that of phraseological units, with “B” having a figurative sense, e.g., dar a lata “to be a nuisance”).

To distinguish between collocational light verb constructions (CLVs) and locutional light verb constructions (LLVs), Barrios Rodríguez concludes that both types “no pueden distinguirse de un modo claro con estos criterios semánticos” [“cannot be clearly distinguished with these semantic criteria”] and, in order to differentiate them, it is necessary to resort to “rasgos distintivos estrictamente sintácticos” [“strictly syntactic linguistic traits”] (Barrios Rodríguez, 2015, p. 13). Various syntactic features or tests have been used, with different levels of systematization, in many studies, such as Álvarez de la Granja (2003), Alonso Ramos (2004), Lorente (2006) or Martínez Vicente (2023), always with a synchronic orientation. The present research is based on the intensional and diachronic study of a series of syntactic parameters, projected onto some LV constructions of various natures.

Among the most common general light verbs (see Bustos Plaza, 2006; Koike, 2011, p. 4; Martínez Vicente, 2023, pp. 39–40; 2025, p. 6) are dar, facer “to do/make”, and, at different moments of the history of the language, haber (medieval aver) and ter (see, for example, the 59 verbal locutions with dar, 62 with facer, and 42 with ter in Álvarez de la Granja, 2003, pp. 233–235, 237–239 and 252–253, respectively)1. Some of the constructions formed by them as LVs are the focus of this research.

2. Methodology

For practical purposes we will distinguish between LLVs (light verbs combined with units that the literature identifies as maximum fixation, that is, locutions or phrasemes—for the definition of the concept of prototypical fixation, see Berlage, 2025, p. 159) and CLVs (general light verbs in collocational use). When we refer to them all we will apply the label LV.

2.1. The Corpus

The corpus of our study consists of a total of 8728 examples distributed as follows: 1357 with dar, 2790 with facer and 4581 with haber/ter (335 of the former and 4246 of the latter). As

Table 1. Distribution of corpus examples.

	CLV/LLV 1	Number of Examples
Dar + [noun] (1357)	Dar a lata [LLV]	35
	Dar aviso [CLV]	114
	Dar berros [CLV]	243
	Dar brados [CLV]	15
	Dar cabazas [LLV]	20
	Dar couces [CLV]	165
	Dar gloria [LLV]	97
	Dar golpes [CLV]	193
	Non dar golpe [LLV]	9
	Dar opinión [CLV]	25
	Dar paseos [CLV]	227
	Dar patadas [CLV]	135
	Dar testemuño [CLV]	79
Facer + [noun] (2790)	Facer acenos [CLV]	320
	Facer burla [CLV]	229
	(Non) facer caso [LLV]	1608
	Facer memoria [LLV]	99
	Facer mención [CLV]	157
	Facer oración [CLV]	150
	Facer visita [CLV]	227
Haber/Ter + [noun] (4581)	Haber/ter medo 2 [CLV]	282/2738
	Haber/ter vergonza 3 [CLV]	53/307
	Ter alerxia [CLV]	25
	(Non) ter idea [LLV]	156
	Ter lugar [LLV]	996
	Ter tose [CLV]	24
		Total: 8728

¹ All LLV examples are collected in Álvarez de la Granja (2003). We always refer to each LV construction with a citation form (dar paseos “to go for walks”, for example), but in establishing the corpus we include all the examples detected regardless of their variation (such as dar un paseo “go for a walk”, dar o paseo “go for the walk”, dar un paseíño “go for a short walk”…). ² In medieval Galician 98.9% of the examples are of aver medo (“Meus fillos, nõ ajades medo” [“My children, do not be afraid”] MS [I, 1]/186), and teer medo accounts for only 1.1% (“et que nõ te[u]esse medo nẽ uergonça” [“and that he has no fear or shame”] TC 1/587). The last examples that we documented with the verb haber in p1, with possessive sense, are from the early 20th century, from 1922 and 1933 (“¡Convosco xa non hei medo!” [“With you I am no longer afraid!”] Bouza Brey: Nao/1933, we must bear in mind that Nao senlleira is Bouza’s first poetic book, framed within neo-troubadourism, with many medieval reminiscences), all other examples in contemporary Galician are of ter medo or of existential haber, in the third person singular (“Na rectoral de Asados non había medo aos ladróns” [“In the rectory of Asados there was no fear of thieves”] Nós 73/1930). The latter is very common in negative contexts: on average, it occurs three times more frequently in negative contexts than in affirmative ones, especially in the 19th century (cf. López Taboada and Soto Arias, 2020, s. v. medo). The 20 examples of ter medo por [alguén] “to worry about [someone]” were not taken into account. ³ In addition to those of the verb, all historical variants of the noun are included: vergonça, vergonna, vergonza, vergoña… All the medieval examples are with aver, except for one example with teer in which it is used with medo (cf. supra), and all the contemporary examples with haber are existential.

shows, with the verb dar 13 different nominal combinations were analyzed, 7 with the verb facer and 6 with the verbs haber/ter, for a total of 26. We made sure that, in all cases, there were structures usually analyzed in the literature as one of two possible types (cf. Álvarez de la Granja, 2003; Bosque, 2006; López Taboada & Soto Arias, 2020):

(a) Collocations with light verb (CLV), a total of 18: dar “give” + [aviso “notice”, berros “shouts”, brados “shouts”, couces “kicks”, golpes “blows”, opinión “opinion”, paseos “walks”, patadas “kicks”, testemuño “testimony”], facer “do” + [acenos “gestures”, burla “mockery”, mención “mention”, oración “prayer”, visita “visit”], haber/ter “have” + [medo “fear”, vergonza “shame”] and ter “have” + [alerxia “allergy”, tose “cough”].

(b) Locutions or phrasemes (LLV), 8 in total: dar a lata ”to be a nuisance”, dar cabazas ”to rebuff a claim”, dar gloria ”to like very much”, non dar golpe ”not to lift a finger”, (non) facer caso ”(not) to pay attention”, facer memoria ”to remember”; (non) ter idea ”have no idea” and ter lugar ”to occur, happen”.

All the examples are interpreted syntactically as cases of verbs (predicates) with a functional structure with direct object (DO) (cf. Álvarez de la Granja, 2003, p. 28 and Soto Arias, 2022, p. 166 for LLV, and Gómez Clemente, 2024, p. 60 for dar testemuño).

On average, there are generally more than 335 occurrences for each item shown in Table 1, with a greater presence of examples per structure with haber/ter (916), which falls to 398 occurrences with facer and decreases to 104 with dar. If we compare the averages between those identified as locutions and the LV constructions, we can see that the former rise to almost 378 and the latter drop to 3172.

The examples come from Galician texts dated between the 13th and 21st centuries. For their extraction we used as main corpora the Xelmírez for the medieval period3 and the TILG for texts from the beginning of the 17th century to the present. Complementarily, we used the Gondomar corpus for middle Galician (16th–18th c.) and the CORGA to give some examples of contemporary Galician from the last fifty years in specific cases (where there are few examples in TILG: dar cabazas and ter alerxia). With all of them we performed a manual review of the many thousands of examples returned by these corpora, until reaching the total figure shown in Table 1 of 8728.

The chronological periods we used are: medieval Galician (13th–15th c.), middle Galician (16th–18th c.) and contemporary Galician, divided into 19th c., early 20th c. (1900–1936), late 20th c. (1937–1999) and 21st c.

Table 2. Chronological distribution of examples from the corpus.

	13th–15th c.	16th–18th c.	19th c.	Early 20th c.	Late 20th c.	21st c.
Dar + [noun] 1357 (67)	182	5	183 (15)	223	615 (47)	149 (5)
Facer + [noun] 2790 (106)	146	1	347 (13)	523 (1)	1421 (70)	352 (22)
Haber/ter + [noun] 4581 (312)	214/3	/8	27 (1)/489 (43)	47/790 (2)	40 (2)/2375 (226)	7 (1)/581 (37)
Total: 8728 (485)	545	14	1046 (72)	1583 (3)	4351 (345)	1089 (65)

shows the distribution by historical periods of all the examples.

Beyond the chronological factor, we also took into account diamesic variation and, in Table 2, the figures given in brackets for contemporary Galician indicate the number of cases drawn from spoken language. Regarding the clustering of examples irrespective of any other consideration, the largest number of occurrences are documented in contemporary Galician, especially in the last two thirds of the 20th century (with almost half of the total), followed by the first third of the 20th century (below one fifth) and, with 500 fewer examples, the 21st and 19th centuries (with approximately 12% each), and finally medieval Galician has half of the latter percentage (6%). All of the above are relatively high figures; the only truly very low figure is that documented in the middle period, where there is a notorious scarcity of texts (0.2%)4.

Regarding the weight of the samples drawn from spoken language in the different groups, it is 6.8% with haber/ter, 4.9% with dar and 3.8% with facer (the difference between the number of written and oral examples is explained by the imbalance in TILG itself, where the weight of orality is 4.7%). A slight preference for the former (haber/ter) is detected, while the latter two only differ from each other by just over one percentage point. If we compare the examples taken from spoken language between VSL and VSC, we can see that with the verbs dar and facer the differences remain very small (0.7% higher with VSC with dar and 1.5% higher with VSL with facer), while with (haber/)ter the cases of VSC are much higher, 8.7% versus 0.9% for VSL, clearly weighed down by the fact that, among the 996 cases of ter lugar, there is only one example taken from oral language (it should be borne in mind that ter lugar was interpreted as a Gallicism, cf. Corominas & Pascual, 1991–1997, s. v. lugar, which we detected in Galician for the first time in 1876).

2.2. The Survey

A written survey with several specific questions5 was given to 162 informants. In the present study, we used three of these questions, which will be analyzed in Section 3 and Section 4. The general profiles of the participants are as follows:

(a) By age, the majority belong to the 19–29 age range (53.7%), followed by the group under 18 (27.8%), adults between 30 and 59 (10.5%) and, finally, informants of 60 years or older (8%).

(b) Mother tongue: the majority claim to have both Galician and Spanish (41.4%) as mother tongues, with 22.2% having Galician as their initial language and 35.8% Spanish.

(c) The average self-assessment of command of Galician is 7.53 (out of 10). Only two people give themselves a failing mark (4 or less), the majority score between 7 and 8 (50.6%), with 24.1% between 9 and 10 and 19.1% between 5 and 6.

(d) The ratio between cities (46) and small towns (96) is favorable to the latter in a 2 to 1 relation. Most are from the province of Pontevedra (60.5%), with even results between those from A Coruña (12.3%) and Ourense (11.1%), and, at some distance, Lugo (3.7%).

(e) The family language is Galician or more Galician for 27.8%, Galician and Spanish for 25.9% and only or more Spanish for 45.7%. The language used with friends is Galician or more Galician for 21%, Galician and Spanish for 23.5% and only or more Spanish for 54.3%.

(f) Most of our participants are university students (69.1%), but we also included secondary education students (15.4%). Among the respondents we also counted 14 specialists in Galician philology (see Kardela, 2021, p. 223).

3. Data Analysis

In this section we offer (i) a contextualization of historical usage in Galician of the structures with (C/L)LV, of the nouns that serve as their basis, and of the verbal synonyms that share lexemes with those nouns, as well as (ii) an internal analysis of the constructions with a general LV over time.

The first piece of data we want to find out is the proportion with which the nouns identified at the beginning of Section 2.1 combined, over the centuries, with the LVs analyzed, that is, of all the attestations of a noun, what proportion of them are the examples studied here (cf.

Table 3. Percentage that the examples of LV + [noun] represent in comparison to the total occurrences of those nouns in all usage contexts.

%	13th–15th c.	16th–18th c.	19th c.	Early 20th c.	Late 20th c.	21st c.	Average
Dar + [noun]	6.7	6.8	7.9	6.9	7.7	6.5	7.3
Facer + [noun]	27	3.2	14.9	13.7	8.4	7.1	9.7
Haber/Ter + [noun]	5.8	17.8	18.9	13.4	12.2	9.5	11.8
Total	7.8	9.3	14.1	11.9	9.7	8.1	10.1

):

Several trends are immediately noticeable:

(a) The remarkable evenness shown by the nouns combining with dar, which is the lowest percentage at 7.3%, but it proves to be remarkably stable, since across all periods the greatest fluctuation is 1.2 percentage points, including the medieval period.

(b) Secondly, it is worth commenting on the nouns that combine with facer as LV, with an average yield of 9.7%, but with the following two clear situations: higher yield in the medieval period (mainly justified by the cases of fazer oraçõ—many—and fazer mençon—most examples with the noun appear in this construction), and a progressive decrease from the 19th century to the present (with a proportion reduced to half of that recorded in that century).

(c) With haber/ter, which presents the highest average of all the verbs at 11.8%, a decrease is also documented in the 21st century to almost half of the cases recorded in the 19th century, but in the medieval period it is, among the studied verbs, aver that presents a smaller proportion of the nouns with LVs (weighed down by the abundance of the noun logar/lugar, but not in combination with aver or teer).

Therefore, we can conclude, generally speaking, that the proportions these structures represent among all occurrences of the nouns in our structures are relatively low. In Figure 1 we show the same distribution, but distinguishing between CLV and LLV.

At first glance, the much lower values of LLV stand out, with a mean across the three verbs of 5.3%, compared to CLV examples that almost quadruple that figure and stand at 20% on average. In the medieval period, the weight of CLV examples with the verbs fazer and aver is really high (more than 50% in the former and more than 37% in the latter). The cases with locutions yield the highest peaks mainly in the 19th century, although always with relatively low figures, and they all stabilize in the second half of the 20th century and in the 21st century, with much greater productivity of CLVs, which roughly multiply those of LLVs by three (facer), by more than four (ter) and by five (dar).

Table 4. In oral discourse, proportion of cases of LV + [noun] in comparison to the total occurrences of those nouns in all contexts of use.

%	19th c.	20th c. 1	21st c.	Average
Dar + [noun]	26.3 (0/50) 2	10.5 (2.7/18.1)	11.6 (7.7/13.3)	12.3 (2.7/21)
Facer + [noun]	18.3 (20.8/23.1)	20.1 (17.3/31.9)	20.4 (17.8/33.3)	19.9 (17.8/31)
Ter 3 + [noun]	24.3 (0.9/60.6)	23.9 (2/44.6)	16.2 (0/44.7)	22.6 (1.4/46.4)
Total	24 (5.9/53.5)	19.7 (6.6/35.8)	16.8 (6.7/36.1)	19.8 (6.6/37.8)

¹ We added up the entire corpus from the 20th century because the records from the beginning of that century are very scarce: 181 names, of which only 3 are in constructions with LV. ² The first figure is the general average with all names; the figures in brackets are the examples of LLV, on the left, and those of CLV, on the right. ³ Only four cases of haber were documented, all existential.

analyzes the same aspect but exclusively with examples taken from spoken discourse.

The first noticeable thing about oral discourse, compared to the results shown in Table 3, is that the structures with LV represent a substantially larger proportion of the total uses of the nouns: 5 percentage points with dar and 10 with facer and ter. The overall mean of the three verbs rises from 10.2% to 19.8%, that is, almost 1 in 5 examples of these nouns is documented, in spoken discourse, in LV structures. But a clarification is in order here, since there is a difference between the two major types of examples: while with LLV the overall mean is 5.3%, the mean in spoken discourse is 6.6% (just over one point higher), with CLV the overall mean is 20%, whereas in oral discourse it rises to 37.8%, which is nearly twice as high.

Analyzing the oral data from a chronological perspective, except with facer, with an overall balanced result in contemporary Galician, with the other verbs, a decline in LV examples is noticeable from the 19th century, more pronounced with dar and much smaller with ter. If we distinguish between LLV and CLV (with means of 6.6% and 37.8%, respectively), we find that in the former the proportion is very low with the verb dar and, above all, with ter (probably as a result of the negligible presence of ter lugar in spoken discourse: of the 661 recorded examples of the noun lugar, only one of them is an LLV construction, that is, 0.15%), and much higher with facer (more than 6 and 12 times, respectively)—thanks, fundamentally, to (non) facer caso with 20.4%. For their part, with CLV the figures with dar (21%) and facer (31%) are remarkable, but they increase considerably with ter (with more than 46%—on account of the weight of ter medo, at 48.4%). In view of these results, it seems clear that in spoken language there is a preference for these structures with LV, especially CLV.

3.1. LV Constructions vs. Simple Verb

In this section we study the interrelationship between the constructions with LV and the equivalent full verbs. The lexically equivalent verbs6 to the structures LV + [noun] that we review here were the following: avisar “to warn” (dar aviso), berrar “to shout” (dar berros), bradar “to shout” (dar brados), golpear “to strike” (dar golpes), opinar “to opine” (dar opinión), pasear “to walk” (dar paseos), patear “to kick” (dar patadas), coucear “to kick” (dar couces), (a)testemuñar “to testify” (dar testemuño); acenar “to make gestures” (facer acenos), burlar “to mock” (facer burla), lembrar (<lat. MEMORARE) “to remember” (facer memoria), mencionar “to mention” (facer mención), orar “to pray” (facer oración), visitar “to visit” (facer visita); temer “to fear” (haber/ter medo), avergonzarse “to be ashamed” (haber/ter vergonza) and tusir “to cough” (ter tose) (cf. Section 3.1.2 to see in detail the contextual distribution of some of these pairs). All CLVs are represented except ter alerxia.

3.1.1. The Corpus Data

It is very common in the literature to highlight the fact that many CLV constructions coexist in the language with specific verbs that have equivalent meanings. With regard to Galician collocations with the verb dar, Martínez Vicente points out that only 8.87% of them do not have a corresponding verbal form (Martínez Vicente, 2023, p. 159). According, for example, to Mateus et al. (2003, p. 312), the existence of equivalent simple verbs would constitute one of the manifestations of the “processo de esvaziamento lexical [of LV] a que alguns autores chamam gramaticalização” [”process of lexical emptying [of LV] that some scholars call grammaticalization”]. Berlage defends the idea that “the simple verb and the CP [Composite Predicate] compete with each other in semantic space” (Berlage, 2025, p. 175). On the other hand, Barrios Rodríguez concludes that differences between LV constructions and simple verbs do exist, and these “no solo son sintácticas” [”are not merely syntactic”], but constructions with LV allow for the construction of a much richer message “desde el punto de vista comunicativo: la colocación permite enumerar, valorar, focalizar, etc.” [”from a communicative perspective: collocations allow for enumeration, evaluation, focusing, etc.”] (Barrios Rodríguez, 2015, p. 21) and “gracias a las colocaciones se amplían nuestras posibilidades expresivas” [”thanks to collocations, our expressive possibilities are expanded”] (Barrios Rodríguez, 2015, p. 22).

Figure 2 shows the evolution in preference, or lack thereof, for LV constructions versus the corresponding equivalent verbs (all examples were counted for the calculations). In the medieval period a great preference for LV with dar is noticeable (almost three quarters of the total), aver presents a balanced distribution and, finally, the LV structure decreases with facer to about two fifths of the cases7. The subsequent evolution marks a considerable and progressive decrease in LV cases (versus the equivalent verbs), especially with dar, but also with facer, until reaching relatively low percentages, around one tenth of the examples. The opposite path is followed by aver and ter: at least with the LVs studied here, it was very productive with aver in medieval Galician and continues to be so, in comparison with the equivalent verbs, in contemporary Galician with ter(/haber). This high percentage is explained by the large number of examples of ter medo and its preference over temer (the average distribution between them is 61.4–38.6%, respectively). Something similar happens with ter vergonza, with LV representing 44.3% versus 55.7% of avergonzarse. In the case of ter tose, the situation is diametrically opposite, as it only accounts for 6.1% of the examples vis-à-vis tusir8 (see Section 3.1.2).

Neves, commenting on the fact that the LV construction is more common than the corresponding ones with full verbs, argues that “é mais conforme à fala coloquial” [”it is more common in colloquial language”] (Neves, 2002, p. 227). The analysis of oral discourse in contemporary Galician confirms, indeed, the greater relative presence of LVs vis-à-vis their equivalent verbs, compared to the results shown in Figure 2: regardless of textual types (Figure 2), the global mean of LVs versus equivalent simple verbs is 24.9%, while in oral discourse this mean rises to 42.2%, that is, over 2 out of 5 examples are LVs (and a bit less than three are equivalent simple verbs). Figure 3 details the evolution of each of the three verbs examined here as follows: dar remains the least productive LV in this respect (with a mean of 11.7%), almost the same as that obtained for the whole set of texts (Figure 2). With facer as LV in oral discourse there is a clear increase compared to the results shown above, since now the mean is 51%, more than 30 percentage points higher than the figures shown in Figure 2. The LVs analyzed with ter remain the most productive, with an average percentage of 85.8% (more than 20 points higher than the results documented for the whole set of texts without type distinction). With these two verbs an increase in use can also be observed in the transition from the 19th century to the 20th century, along with a stabilization of their use, with only a slight decrease in the 21st century.

Table 5. Dating of the first attestations of LVs, equivalent full verbs analyzed and nouns.

	LV Constructions	Simple Verbs	LV Constructions	Nouns
Deverbal derivatives	dar aviso	avisar 14th c. (1603) 1	14th c. (1861)	14th c. (1861)
	dar brados	bradar 13th c. (1898)	13th c. (1925)	13th c. (1898)
	dar golpes	golp(e)ar 13th c. (1889)	13th c. (1671)	13th c. (1671)
	dar paseos	pasear 1812 2	1842	1746
	facer visita	visitar 13th c. (1807)	1858	14th c. [visitaçon]/1810 [visita]
	dar berros	berrar 1697	1746	1746
Different Latin etymologies	facer memoria	lembrar 13th c. (1746)	14th c. (1858)	13th c. (1594)
	aver/ter medo	temer 13th c. (1697)	13th c. (1922)/14th c. (1746)	13th c. (1671)
	ter tose	tusir 13th c. (1843)	1899	13th c. (1770)
	dar opinión	opinar 14th c. (1877)	1888	15th c. (1813)
	facer oración	orar 13th c. (1777)	13th c. (1813)	13th c. (1603)
Denominal derivatives	facer mención	mencionar 1891	14th c. (1861)	14th c. 1861
	aver/ter vergonza	avergonzarse 13th c. (1828)	13th c./14th c. (1807)	13th c. (1807)
	dar couces	coucear 1888	13th c. (1746)	13th c. (1697)
	dar testemuño	(a)testemuñar 14th c. (1899)	1846 3	13th c. (1836)
	facer (a)cenos	(a)cenar 1858	1886	1886
	facer burla	burlar 13th c. (1810)	1807	13th c. (1807)
Common etymology	dar patadas	patear 1807	1877	1840

¹ The date of the first attestation we have recorded after the medieval period is given in brackets. The use of boldface highlights the oldest documented date, between the equivalent full verb and the LV. ² We discard a first example from the year 1528 (and some later ones) because the notarial document where it appears is written in Spanish. ³ In the medieval examples that we found of dar testemuño, the noun always has the value of “notarial instrument”, that is, it is a material object: “que me de ende un testemuyo ou mays se mester feser firmado de seu nome” [“that he give me a written testimony of this, or more if necessary, signed with his name”] (PN 1357, CDMO 1740b/271).

groups the studied constructions according to their etymological origin (cf. Corominas & Pascual, 1991–1997 and González Seoane et al., 2006–2022) and we verify, from the consulted corpora, the earliest attestations of both LV examples and the equivalent simple verbs or the nouns that combine with those LVs.

With forms of regressive derivation, that is, deverbal derivation, half of the examples (3) are attested for the first time in the 13th or 14th centuries, both the simple verbs and the LV constructions. When this evenness does not occur, verbs are always documented first, often with a difference in a few years. With examples originated from different Latin etyma, we attested verbs earlier in three cases and simultaneous first attestations in the 13th century in two. In denominal derivation cases, there is more variation: on two occasions the first attestation of the LVs is older (facer mención and dar couces), in one example there is simultaneity in the 13th century, and on three occasions the verbs are documented first ((a)testemuñar, (a)cenar e burlar). For the only example with a common etymon, we document the verb earlier (patear).

In any case, the data studied in this section allow us to confirm that, although many verbs are older, the LV constructions are in many instances as old as them, and as an analytic construction, they are especially productive in medieval Galician. Over the centuries a general increase in preference for the full verbal unit is detected, but with specific behaviors depending on the LVs involved (and, obviously, on the concrete examples): the decrease is more pronounced with dar and somewhat less with facer; meanwhile, ter presents a specific behavior with a great productivity of the LV structure, but essentially conditioned by the enormous frequency of some examples (especially ter medo versus temer). In oral discourse, at least for the examples studied, LVs with dar are, proportionally compared to full verbs, the least productive, which is not the case with facer and, above all, with ter9.

3.1.2. LV or Simple Verb in Survey Answers?

Complementary to the previous results derived from the study of our corpus, one of the questions from our survey—question (I) of Appendix A—allows us to see the extent to which our informants opt for an LV construction or for the simple equivalent verb. Figure 4 shows the results for the examples constructed with the verb dar.

In the examples of dar berros (QI.1 and C1) and dar aviso (QI.8 and C8), both in the survey and in the corpus, the choices with the simple verbs, berrar and avisar, are clearly predominant. With dar patadas (QI.2, QI.7, and C2) the respondents of our survey clearly prefer, in 4 out of 5 responses, (2A) and (3A), instead of the simple verb, regardless of whether the syntactic construction is absolute, (2A) and (2B), or not, (3A) and (3B). In the corpus, it is interesting to note that the majority of constructions with the ESV present an absolute use, 88.5%, such as (2B), while constructions with LV and the plural noun (patadas) are only absolute uses in 26.2%, such as (2A) (with the singular noun, dar unha patada, absolute uses only account for 9.5%). Regarding the alternation in the corpus between the ESV and the constructions with LV (with patada in the singular, LVsg., or with patadas in the plural, LVpl.) the proportion is, in absolute use, ESV 82.5%/LVsg. 5.3%/LVpl. 12.2%, while in transitive constructions—prepositional or not, (3A) and (3B)—the distribution is ESV 11.1%/LVsg. 53.2%/LVpl. 35.7%.

(1)	A.	Dando berros non solucionas nada.
	B.	Berrando non solucionas nada.
		“Complaining doesn’t solve anything”
(2)	A.	Comezou a dar patadas co cabreo que colleu.
	B.	Comezou a patear co cabreo que colleu.
		“He started kicking with the anger he felt”
(3)	A.	Deulle patadas na barriga.
	B.	Pateoulle a barriga.
		“He kicked him in the stomach”
(4)	A.	Tardaron moito en dar aviso do furacán tropical.
	B.	Tardaron moito en avisar do furacán tropical.
		“They took a long time to warn about the tropical hurricane”

Figure 5 analyzes the results of LVs constructed with the verb ter:

As Figure 5 shows, the survey informants responded evenly to the questions (5A) and (6A), opting predominantly for the full verb option in three fifths of the responses. In the corpus this tendency increases and the simple verbal form, tusir, is chosen in more than 93% of the cases (85.4% if we consider only the 21st century). This disproportion is justified by the high number of examples of the verb tusir in constructions without any complement or adjunct (approximately 70% of the total), while this type of structure in the LV only accounts for 37.5%—that is, in most examples the construction with LV is accompanied by some complement or quantifier (cf. Barrios Rodríguez, 2015, pp. 21–22). In questions QI.5 and QI.9 of the survey the responses differ, since with the first example, (7A), the LV choice is the preferred one (with almost 60% of the total), while with the second example, (8A), two out of three chose the equivalent simple verb. In this case, the situation in the corpus is quite even, with 44.3% for LV and 55.7% choosing the simple verb (if we consider only the 21st century the percentage of LV falls to 22.5%).

(5)	A.	Tiven tose durante toda a noite.
	B.	Tusín durante toda a noite.
		“I coughed all night”
(6)	A.	Onte pola noite tiveches moita tose.
	B.	Onte pola noite tusiches moito.
		“Last night, you coughed a lot”
(7)	A.	Díxolle: “Non debes de ter vergonza”.
	B.	Díxolle: “Non debes de avergonzarte”.
		“He told him, ‘You should not be ashamed’”
(8)	A.	Ó final acabou tendo vergonza do seu pasado.
	B.	Ó final acabou avergonzándose do seu pasado.
		“In the end, he ended up feeling ashamed of his past”

Finally, in Figure 6 we reproduce the data of the constructions with the verb facer:

The preference for burlar, (9B), is quite even, both in the survey and in the corpus, with two quarters of the examples; the results for facer acenos are quite close, with more than 61% in the survey and eleven points less in the corpus (and it falls somewhat more if we focus only on the 21st century, since the share of LVs drops to 42%). Finally, with facer unha visita there is a clear divergence between the majority preference for the simple verb in the corpus (90%), (11B), and the near technical tie in the survey. For a better understanding of the results of the survey, it is necessary to note that the direct object of the visit was human and, in these cases, the weight of the construction with LV in the corpus is really important: in a fifth of the examples in the corpus it is not mentioned who or what is being visited and, when it is made explicit, this constituent is mostly human with 58%. We reviewed 250 randomly chosen examples of the verb visitar “to visit” (from the 19th, 20th and 21st centuries) and found that the DO (always realized) is inanimate in 54% of the cases and human in 46%.

(9)	A.	Fixeron burla da súa voz como cantante.
	B.	Burláronse da súa voz como cantante.
		“They mocked his voice as a singer”
(10)	A.	Facíanos acenos coa man para que fosemos.
	B.	Acenábanos coa man para que fosemos.
		“They waved to us with their hand so we would go”
(11)	A.	Con tres anos de atraso, por fin, púidenche facer unha visita!
	B.	Con tres anos de atraso, por fin, púidente visitar!
		“With three years of delay, finally, I was able to visit!”

3.2. LV in Galician: Morphological-Syntactic Flexibility or Rigidity?

Just as LLV structures are associated with great formal rigidity, several authors highlight “a possibilidade de as construções com verbos-suporte apresentarem flexibilidade morfossintática” [“the possibility that constructions with support verbs exhibit morphosyntactic flexibility”, our translation] (Chishman & Abreu, 2014, p. 166). In fact, Neves (2002, p. 222 et seq.) cites, among others, the following reasons for choosing an LV construction, on account of its greater syntactic versatility in comparison with the simple verb: (i) the presence of an adjective modifying the “nome complemento” [“complement noun”] of the LV, (ii) the determination of that noun (with reflexive possessives or quantifiers), (iii) the relativization of the LV noun (o medo que el ten “the fear he has”). And, finally, a reason of a different nature, (iv) the use of the LV structure allows easier omission of a “complemento de especificação” (Neves, 2002, p. 226), a specifier complement that can also be made explicit “se textualmente necessário” [“if it is textually required”] (Neves, 2002, p. 226).

In the following sections we study the behavior of the LLV and CLV constructions selected in our corpus, based on the analysis of the following parameters:

−

At the level of the noun: possibility of numerical variation and appreciative derivational variation (basically, with the noun in a diminutive form).

−

Inside the nominal unit: presence of determiner and existence of some modifier of the noun (adjectival, prepositional, and indefinite/quantifiers; relatives).

−

In the relative position verb/noun: preposing or postposing of the noun to the verb and presence or absence of contiguity between both.

−

In the syntactic structure: of the noun, with the presence of a noun complement governed by the noun (second actant of the noun for Alonso Ramos, 2004) and of the verb, with the use of the function of indirect object (lexical and/or clitic).

−

Possibility that the noun is a member of a coordination scheme.

−

Relativization of the noun.

−

In addition, we also analyze the presence of negation.

3.2.1. Numerical Variation

Most commonly in LV constructions in Galician the noun appears in the singular (cf. Figure 7). It is the preferred option in seven LLVs (except with dar cabazas—(12a) in pl. and (12b) in sg.) and is the most frequent in CLVs (13 versus 4).

However, what we want to emphasize especially is the existence of numerical variation, that is, the extent to which LV constructions tend to be monopolized by one choice of grammatical number, whether singular or plural. The overall means in our LLV corpus are: dar 97.3% and 100% for facer and haber/ter; with CLV they are: dar 71.8%, facer 82.7%—13a) in sg. and (13b) in pl.—and haber/ter 97.8%—(14a) in sg. and (14b) in pl. The global means are 98.6% and 84.1% for LLV and CLV, respectively. Therefore, this shows (i) that variation is minimal with LLV and 14 points higher for CLV and (ii) that the behavior of the studied verbs is not identical, since dar is the only one that shows variation with LLV and the one that varies the most in CLV, followed by facer and, lastly, haber/ter (between these extremes, in CLV there is a 26-point difference).

(12)	a.	Nantronte deulle cabazas a un siñorito mui repinicado unha costureira da vila (OTM II,8/1883)10
		“The day before yesterday turned down a very refined young man a seamstress from the town”
	b.	— Comigo soio, ¿e darás cabaza aos demais? (Alonso: Contos/1955)
		— “Only with me, and you will reject the others?”
(13)	a.	Fixen aceno de marchar (López López: Caderno/2001)
		“I made a gesture to leave”
	b.	Fai acenos coa man, chamando (Rodríguez Elías: Sobriño/1930)
		“He makes gestures with his hand, signaling”
(14)	a.	Acaso a tía terá alerxia? (65 2011/González Costa: Filla)
		“Could it be that the aunt has an allergy?”
	b.	temos alerxias (193 2007/Regueira: L’affiche)
		“we have allergies”

3.2.2. Presence of Determiner

In the Galician LV constructions analyzed here, it is not surprising that with LLV most nouns appear without a definite/indefinite article, except with dar a lata. In the case of CLV, the distribution is much more even, since in ten cases the option without a determiner is the most frequent one, while the opposite is the case in eight of them (cf. Figure 8).

As with the previous parameter, the main objective of our study is to analyze the level of fixation of these structures and, therefore, regardless of the preferred choice of each (the overall mean of cases without a determiner is 84.2%, while that of cases that predominantly present a determiner is 71.4%), we focus on the percentages of the preferred choices. Thus, if we focus on LLV, the proportions with the verbs studied here are: dar 92.6%, facer 99.4% and haber/ter 91.2%—(15a) with det. (and adjectival modifier) and (15b) without det. For its part, with CLV the distribution is the following: dar 70.2%—(16a) without det. and (16b) with det.—, facer 71.9%—(17a) with det. and (17b) without det.—and haber/ter 76%. Again, we note a clear divergence between the scarce variation with LLV (94.6% on average) versus that documented with CLV (76%), 18 points apart.

(15)	a.	Ninguén tiña a menor idea do seu parecido a ti (Veiga: Cinema/2005)
		“No one had the slightest idea of his resemblance to you”
	b.	cando non teñen idea do que foi o franquismo (Tempos 4/1997)
		“when they have no idea of what Francoism was”
(16)	a.	dando berros de trunfo (Carré Aldao: Rayolas/1898)
		“shouting triumphantly”
	b.	rapaz que dira o berro trunfal (Fernández González: Piruleiro/1925)
		“the boy who gave the triumphant shout”
(17)	a.	para facer as visitas proxectadas (Martínez Oca: Náufragos/1995)
		“to carry out the planned visits”
	b.	só facía visitas para mudar a venda unha vez por semana (Paz: Follas/2005)
		“he only made visits to change the window once a week”

Based on the eight examples of LV with medieval documentation collected in Figure 18 (cf. Section 5.1), we examine the use of the determiner across the centuries. The average percentage of determiner presence is 14.1% in the medieval centuries, 17.9% in the 19th century, and 18.8% in the 20th and 21st centuries. As expected, the medieval period exhibits the lowest proportion; however, the differences are relatively small, remaining within a range of less than five percentage points. In fact, some of these LVs show a significant presence of the determiner in the medieval period (fazer oraçõ “to pray” 48.2%, dar couçes “to kick” 44.4% or dar colpes “to hit” 19.9%), but the average is lowered by the 0% recorded for dar testesmoyo “testify”, fazer mençon “to mention”, and aver uergonça “to be ashamed”.

3.2.3. Presence of Modifiers

As we indicated above, we checked the occurrence of adjectival, prepositional, indefinite/quantifier and relative modifiers in the examples. This is a relevant criterion, and Barrios Rodríguez recalls that constructions with LV, as opposed to their equivalent simple verbs, open up a broader range of possibilities, mainly thanks to adjectives (Barrios Rodríguez, 2015, p. 22). In Figure 9 we show the relative weight that the sum of all of them represents in relation to the total occurrences of each example.

This time there are only two examples, both LLV, with no records of modifiers: non dar golpe and ter lugar. The mean presence of noun modifiers with LLV is as follows: dar 9.2%, facer 11.5% and haber/ter 11.8%; for their part, with CLV the proportions are 27.3%, 22% and 26.4%, respectively. Again, we find that LLV constructions are considerably less likely to take modifiers in comparison with the CLV examples analyzed. The global mean is 11.8% for LLV and just over two times as much (25.2%) for CLV.

The most common modifiers are quantifiers, (18a) and (18b), which account for slightly less than half of the total (see Alba-Salas, 2020, p. 26). They are followed by adjectives (19a) and (19b), with two fifths of the total, and, with much lower percentages, indefinites (20), relatives (21), and prepositional modifiers (22).

(18)	a.	e deulle moitas couces (Gallego: Ourense/2001)
		“and he kicked him several times”
	b.	fixeron pouco caso das súas ordes (Reigosa: Fuxidos/1989)
		“they paid little attention to his orders”
(19)	a.	ouuo grã uergonça (CT 1/328/14th c.)
		“he was greatly ashamed”
	b.	facendo acenos negativos coa cabeza (Rei Ballesteros: Loaira/1992)
		“making negative gestures with his head”
(20)		Et quiseralle dar outro golpe (TC 1/427/14th c.)
		“and he wanted to strike him again”
(21)		e dárennos avisos que precisamos (Risco: Cultura espiritual/1962)
		“and they gave us the warnings we needed”
(22)		que o vello lle fora facer unha visita de rutina ao médico (Trigo: Extintos/1998)
		“the old man had gone to make a routine visit to the doctor”

3.2.4. Preposing or Postposing of the Noun to the LV

In their “Correlation of grammaticality parameters”, Traugott and Trousdale propose “syntagmatic variability” as a sixth parameter, measured according to the degree of “fixation” of a sequence: this will be weakly grammaticalized if an item “can be shifted around freely” and will be strongly grammaticalized if an item “occupies [a] fixed slot” (Traugott & Trousdale, 2013, p. 101). In most cases in our corpus, the nouns that make up LV constructions appear postposed to the verb, which is unsurprising because it is the expected and less marked placement. On some occasions this is the only option documented in the corpus (dar brados, dar opinión, ter alerxia and ter tose), but it is important to note that with most of them, always in very low proportions, some cases of preposing of the noun to the verb are also attested, examples (23–26) (cf. Figure 10).

The extremely high percentage with facer oración can be explained because more than three quarters of the preposing examples we gathered are drawn from medieval Galician and, specifically, almost all of them from the Cantigas de Santa María (23). The mean proportions with LLV on this occasion are only slightly lower than those recorded in CLV for facer (2.9% versus 3.1%, respectively) and haber/ter (1.4% versus 2.2%, respectively) and are even somewhat higher with dar 8.5% (LLV) versus 3.5% (CLV), which is justified by the existence of some preposing that is proportionally very high because the overall number of examples is very small (this is the case with dar a lata (24), dar cabazas (25) and dar golpes (26a), with this latter, most contemporary examples involve a participle (26b)).

(23)		Pois ssa oraçon fezeron (CSM 165/553/13th c.)
		“After they had made their prayer”
(24)		¡Que lata me vai a dar este tío! (Labarta: Língoa/1897)
		“This guy is going to pester me!”
(25)		e cabazas non lle deu ningunha delas (OTM II, 28/1884)
		“and none of them rejected him”
(26)	a.	Coyrmaão, quẽ uos este golpe deu (TC 1/752/14th c.)
		“Cousin, the one who gave you this blow”
	b.	Aló pola meia noite unhos fortes golpes dados na porta nosa (Carballal: Tesouro/1919)
		“Around midnight, some strong knocks on our door”

3.2.5. Contiguity Between the Noun and the Verb

Another necessary parameter when assessing greater or lesser structural fixation is the presence or absence of contiguity between the LV and the noun (we checked this regardless of their relative position, preposed or postposed). When there is no other element between the verb and the noun, or if there is only a clitic, we interpret them as cases of contiguity. If we detect the presence of other elements between them, we distinguish between (i) homostructural separation, if the elements belong to the phrasal unit of the noun (determiners or modifiers), and (ii) heterostructural or effective separation, if the elements are non-nominal, namely intercalation of non, especially in preposing (27), tonic personal pronouns such as eu or ti (28), adverbs such as entón, tamén, aly, pronominal adverbs such as y, prepositional complements of the verb such as cõ elles, delo, por el…, and some members of the coordination scheme other than the LV noun can also be intercalated (29).

(27)	cortexounas, e cabazas non lle deu ningunha delas (OTM II,28/1884)
	“he courted them, and none of them refused”
(28)	¡Non lles teño eu medo (Feixoo: Contenda/1671)
	“I am not afraid of them”
(29)	As mulleres déronme pancadas e patadas (Miranda: Morning/1998)
	“The women gave me punches and kicks”

As Figure 11 shows, the highest proportion of cases of contiguity occur mainly with LLV: non dar golpe, ter lugar, dar gloria, facer memoria and (non) facer caso. Indeed, the mean proportion of contiguity with LLV is 76.3%, almost 42 points above the mean of CLV (34.4%). Certainly, there are LLV examples such as dar a lata where the most common situation is the presence of the determiner (besides some clitic) between the verb and the noun, and other CLV examples such as ter medo where contiguity is quite high. In any case, it is worth highlighting the small weight, in general, of heterostructural separation (the truly important one, cf. Section 5.1): 8.8% on average in CLV (10.9% dar (30), 8.6% facer and 6.9% haber/ter) and less than half that figure, 4.1%, in LLV (a percentage basically corresponding to dar with 8.8%, since facer with 2.1%, (31), and haber/ter with 1.3% are far from that figure). Homostructural separation on average is also twice as high in CLV (40.6%) as in LLV (19.7%).

(30)	dando cada cal súa opinión sobr’o asunto (Vázquez Taboada: Noites/1888)
	“each one giving their opinion on the matter”
(31)	¡Conque faga vosté agora caso de certas faladurías! (Carré Aldao: Forxa/1919)
	“So, now you’re going to take notice of certain gossip!”

3.2.6. Presence of a Noun Complement Governed by the Nouns (CGN) and the Indirect Object (IO)

As we indicated above, it is necessary to describe the behavior of LV constructions in Galician regarding the presence or absence of a noun complement governed by the different nouns11 (non faga burla [de min] “don’t mock [me]”, for example). But in order to obtain a picture as thorough as possible, it is also necessary to investigate how many of these LV constructions make an IO explicit because, regarding ter medo, Alonso Ramos points out that “su segundo actante puede ser desplazado como tercer actante sintáctico del verbo con la preposición a […]; o puede continuar como dependiente sintáctico de miedo” [“its second actant can be displaced as a third syntactic actant of the verb with the preposition a […]; or it may continue as a syntactic dependent of fear”] (Alonso Ramos, 2004, p. 66) (ter medo da noite/terlle medo á noite “be afraid of the night”). Figure 12 shows the results for our Galician LV corpus of both the complement governed by the noun and the IO.

Regarding CGNs, the first thing worth pointing out is that there are some LV examples of which we did not detect any cases—there are five occurrences with LLV (dar a lata, dar cabazas, dar gloria, non dar golpe and ter lugar), and two with CLV (facer visita and ter tose). In any case, as Figure 12 shows, the average presence of CGNs is high, both with LLV (it is true that it does not appear in any of the 4 examples with dar, but the mean with facer is 32.2% and with ter is 38.5%, concentrated in (non) ter idea12), since it reaches an overall mean of 23.6%, and with CLV, which has an overall mean of 30.3%, distributed as follows: 38% for facer, 36.1% for dar and 16.7% for haber/ter. Certainly, the examples of dar testemuño (32), facer mención and (non) ter idea (33) stand out for the large number of CGNs with which they are constructed, but there are many other constructions, beyond dar couces (with CGN in slightly more than half of the occurrences), both LLV ((non) facer caso or facer memoria, for example) and CLV (dar golpes, dar paseos, dar patadas, facer acenos (34), haber/ter medo or haber/ter vergonza (35)), oscillating between 20% and 40%. The CGN may be phrasal, (32) and (35), or clausal, (33) and (34).

(32)	xa dás testemuña da manda! (Gómez Vilavella: Historias/1992)
	“you already bear witness to the legacy!”
(33)	Teño idea de qu’había un por eiquí (Rodríguez Elías: Pano/1930)
	“I have an idea that there was one around here”
(34)	Fíxolle entón un aceno de agardar (Frías: Esperanzas/1997)
	“He then made a gesture for to wait”
(35)	Non teñas de min vergoña (Risco: Porco/1928)
	“Don’t be ashamed of me”

For its part, the IO presents a balance in average use both with LLV (33.3%) and with CLV (30.2%). With the former typology facer has the highest mean at 56.9%, dar reaches 42% and there is not a single case with ter. With the latter, CLV, facer continues to be the most productive at 38%, followed again by dar at 36.1%, while ter stands at 16.7%.

If we analyze the complementary distribution between CGN and IO, we see that in most cases where no example of IO is recorded (dar brados, dar paseos, facer memoria, ter idea or haber/ter vergonza), there is a significant presence of CGN. There are only two cases in which we do not detect any example of either CGN or IO, which happen to be two LLV constructions: non dar golpe and ter lugar. At other times, a null or scarce presence of CGN is complemented by a significant presence of IO: dar a lata, dar cabazas (it is the only LV construction that has IO in all its occurrences), ter alerxia and, above all, facer visita, which does not have any examples with CGN, but has IO in almost 4 out of 5 occurrences (36). On occasion there is a balance between both functions, as in dar aviso, dar berros, dar couces, facer acenos, facer oración and haber/ter medo. With regard to the last example, we analyze the distribution between CGN and IC: in medieval Galician the distribution of these constructive options of CLV with aver medo is 94.1% with CGN (37) and only 5.9% with IO. We also document a single example of ter medo with CGN in the Middle Ages (there are only two with this LV). In subsequent centuries, the proportions between ter medo with CGN or with IO are, respectively: 50–50% in the middle period, 48.6–51.4% in the 19th c., 55.7% (38a)–44.3% (38b) in the 20th c. and 60.4–39.6% in the 21st c. Existential examples with haber are documented from the 19th century exclusively with CGN, not with IO. From the perspective of all IO examples, only 1% of them are of aver medo (39) in medieval Galician; the rest, 99%, are occurrences of ter medo since the 18th century.

(36)		Porparadevos, amigos, a faguer unha visita a iste labrego que vive co seu amor a Galicia (Galiciano 93/1886)
		“Get ready, friends, to pay a visit to this peasant who lives with his love for Galicia”
(37)		et quen quer que el cataua avia del medo (MS [I, 1]/133/15th c.)
		“and anyone who looked at him was afraid of him”
(38)	a.	— «Seica tés medo do lobo (Otero Pedrayo: Señores/1928)
		“Maybe you are afraid of the wolf”
	b.	non lle tes medo ao lobo (Cao: Idilios/1983)
		“you are not afraid of the wolf”
(39)		ou nõ ouuessen medo ao prínçepe que au$\acute{\mathrm{j}}$ã (CT 1/303/14th c.)
		“or they were not afraid of the prince they had”

Table 6. Percentage of IO by unit types and LV.

%	Indirect Object
	LLV			CLV
	Lexical	Clitic	Reduplication	Lexical	Clitic	Reduplication
Dar…	5	31	7	5.3	20.5	6
Facer…	5.8	42.4	8.7	9	18.1	5.7
Haber/ter…	--	--	--	11.8	5.1	6.1

shows the overall distribution of IO in LV constructions according to the type of unit: an atonic personal pronoun (clitic) only (40), a lexical form only (36), or with doubling (both) (41). We can see that there are no major differences between LLV and CLV types, except for the inexistence in the former of cases with ter (not so with the latter) and for a certain preference in LLV for the clitic form of IO.

(40)	o home da eixada (…) déralles aviso (Alonso Ríos: Afranio/1979)
	“the man with the hoe (…) gave them notice”
(41)	E deulle grandes berros ao conde Nuno Méndiz (Cabana: Morte/1996)
	“And he shouted loudly at Count Nuno Méndiz”

3.2.7. Appreciative Derivation, Coordinative Schemes and Relativization

Near the end of this section, we conclude by discussing three parameters that we also analyzed and that take into account (i) the possible derivation of the noun in LV constructions, (ii) that the noun is a member of a multi-member coordination scheme and (iii) the existence of examples of relativization of the type (42). Figure 13 represents the following three values: the first two in percentage (left axis) and the third one in absolute numbers (right axis).

(42)	golpes que deron na parede (Q.II 14/Appendix A)
	“knocks they gave on the wall”

Regarding the first parameter, we verify that none of the nouns in LLV constructions present derivation. In the CLV cases the mean derivation rates obtained are very low: 1.5% with dar, 0.9% with facer and 2.1% with haber/ter. The most productive cases are dar paseos (43a), ter tose (43b) and, at quite a distance, facer visita and dar patadas.

Coordination schemes in LLV constructions are documented only with facer and account on average for a scant 0.6%. The means with CLV are somewhat higher: 4.9% with dar, 3% with facer and 10.9% with haber/ter. As Figure 13 shows, always with modest figures, coordination schemes are present above all with ter alerxia, dar brados, ter tose and haber/ter vergonza. With lower frequency, dar patadas, facer oración (44a), dar golpes and dar couces (44b) are also documented (note how some nouns allow derivation more readily in CLV structures, especially diminutives).

Finally, regarding relativization, it is important to note that there are only six examples for which we did not document cases in our corpus, half of them in LLV: non dar golpe, dar aviso, dar brados, facer memoria, ter lugar and ter alerxia. Although in absolute terms haber/ter medo and facer visita (45a) are very close, the percentage represented by the latter is considerably higher (by 31 percentage points). Apart from the aforementioned examples, only facer burla (45b) and (non) facer caso surpass ten examples.

(43)	a.	Deu un paseíño polo sobrado (Casado: Trancos/1970)
		“He took a little walk through the attic”
	b.	e ten unha tosiña que coido que lle irían ben os aires da mar (Reimóndez: Médico/1992)
		“and he has a little cough that I think would benefit from the sea breeze”
(44)	a.	Et o ome boo (…) esteue treze dias (…) fazẽdo esmolda et oraçõ (MS [I, 1]/213/15th c.)
		“And the good man (…) spent thirteen days (…) begging and praying”
	b.	empezou a dar couces e cabezazos contra a porta (Santamarina: Pinocchio/1987)
		“he started kicking and banging his head against the door”
(45)	a.	poidades atender na visita que agora faredes á Biblioteca América (Villares: Idea/1996)
		“you may attend during the visit you will make now to the América Library”
	b.	queixándose enfurecida da burla que fixera dela (Valladares: Contos/1888)
		“complaining angrily about the mockery made of her”

3.2.8. Negative Polarity

Although we do not include it among the parameters of analysis of our scale of morphosyntactic fixation (see Section 5.1), it is essential to review the behavior of negation in our Galician LV examples. Thus, Álvarez de la Granja indicates that negative polarity with non “dá lugar a un novo esquema” [“gives rise to a new pattern”] (Álvarez de la Granja, 2003, p. 57, n. 96) and, at the level of the study of phraseological units, its importance is highlighted by Baránov and Dobrovol’skij, who devote a specific chapter to it (Baránov & Dobrovol’skij, 2009, pp. 281–317).

First, of the total 8728 occurrences in our corpus, 28.9% display negative polarity. If we discriminate between CLV and LLV, we find that in the former negations drop to 21.4% and rise to 43% in the latter. When the light verb is dar, the average percentage of negations is 12%, which rises slightly with facer to 15.2% and increases clearly with haber/ter to 28.3%. The only LV examples for which we did not detect any negation are dar brados, dar gloria and ter tose. There is only one case in which all examples occur in a negative context, non dar golpe (which is not the negative counterpart of the CLV dar golpes: this, non dar golpes, was recorded in 5.4% of the examples). Two other LV constructions show very high rates of negation: (non) ter idea with 84.6% and (non) facer caso with 67.8%. The next three most productive in negative contexts are: haber/ter vergonza (35.3%), haber/ter medo (32.9%) and dar a lata (22.9%). In the range, above or below, of 10% are facer mención (14.6%), ter alerxia (12%), facer memoria (9.1%), facer burla (8.7%) e dar aviso (7.9%). The rest are below 5%.

Finally, it is necessary to analyze specifically the historical evolution of (non) ter idea. The earliest examples, in the 19th century, display an equal distribution of polarity: ter idea 50%/non ter idea 50%. However, the latter increase over time: 37.5% (46a)/62.5% (46b) (early 20th c.), 15.2%/84.8% (late 20th c.) and 10.6%/89.4% (21st c.), respectively. Regarding negative polarity, it is important to note that over time the incorporation of other intensifiers of negation has increased: (i) non + [(ii) nin/(iii) a máis remota, menor…/(iv) nin + a máis remota, menor…]: the only example attested in the 19th century does not have any intensifier, 80%/0/0/20% (47) (early 20th c.), 40.5%/34.5%/17.9%/7.1% (late 20th c.) and 19.1%/38.1%/33.3%/9.5% (21st c.), respectively. A contributing factor to the decrease in affirmative polarity was probably the concurrence of idea with the acceptation of “intention” (cf. González González, 2012), which we excluded from our study, but which we have already detected since 1928 (48).

(46)	a.	Algo máis teño ideia de ll’haber dito que vostede esquence (Dieste: Fiestra (1)/1927)
		“I have some idea of what I might have told you that you forget”
	b.	Namentras os nosos labregos non teñan idea do que é a instrucción (Lugrís: Ponte/1903)
		“While our peasants have no idea of what training is”
(47)		e dende logo, da España non tén nen a máis pequena ideia (Risco: Mitteleuropa/1934)
		“and certainly he doesn’t have the slightest idea about Spain”
(48)		e o xenro tivo idea de coller unha miga de goberno (Otero Pedrayo: Señores/1928)
		“and the son-in-law had an idea of achieving a piece of the government”

4. Survey Analysis

The data on which the results of this section are based are drawn from the survey described in Section 2.2, whose questions are reproduced in Appendix A.

4.1. The Degree of Fixation of LV Constructions

Question III explicitly asked about the perceived fixation in different LV + [noun] structures: a numerical scale was used where 0 implied free combination and the inexistence of fixation between verb and noun, and 10 meant the maximum degree of fixation. The results are shown in Figure 14.

First, there is no unanimity with regard to any construction, neither at the top (no LV achieves an average of 9–10), nor at the bottom (there are no verbs with an average of 0–2). The two constructions perceived as the most fixed are dar a lata (8.01) and non dar golpe (6.98), followed by terlles ganas and facer memoria in the six-point range, and (non) facer caso (5.65) and non ter idea (5.59). Note that, out of the total of sixteen examples, the six constructions identified as LLV in the literature indeed occupy the first six positions in the perception of the most fixed structures. It is true that the gap separating them from the next items—dar testemuño (5.35), facer burla (5.12) and facer oración (5.09)—is not large, which attests to the gradual nature of these processes. In the four-point band there are the following four constructions associated with CLV: facer acenos, dar aviso, dar ouveos and ter vergonza. Finally, the last three LV examples, which obtained the lowest overall scores and were thus interpreted as the least fixed are: dar golpes, ter alerxia and ter tose (3.88 for the first two and 3.72 for the last). In any case, it is worth emphasizing, as we indicated above, that the score obtained does not imply that any of them are unattested in free combinations (which would have required an average score close to 0).

Alba-Salas acknowledges that the “diferencias de aceptabilidad” [“differences in acceptability”] that he detects with the combinatory possibilities in some examples “son de esperar dada la naturaleza gradual y escalar de la fijación sintáctica” [“are to be expected given the gradual and scalar nature of syntactic fixation”] (Alba-Salas, 2020, p. 26). For this central focus, it is important to examine the coefficient of variation13 of these data on the fixation of LV constructions:

−

The constructions with the lowest coefficient of variation are dar a lata (32.6%), terlles ganas (42.6%), non dar golpe (43.9%), facer memoria (45%) and dar testemuño (49.7%). The first of these, an LLV, is by far the one that shows the smallest coefficient of variation; that is, the data are very tightly clustered around the mean, which is the highest mean in the appraisal of fixation among all the proposed constructions. The next three are all LLVs, and the only CLV is dar testemuño, which among them presents the highest figure, very close to 50%.

−

In an intermediate band of the coefficient of variation, around 50–60%, we document facer oración (53.2%), facer burla (53.9%), (non) facer caso (55.7%), facer acenos (57%), non ter idea (59.8%) and dar aviso (60.3%). Here, the dispersion of the data relative to their means is somewhat greater than in the previous examples. Among them there are two LLVs ((non) facer caso and non ter idea); note the concentration in this band of four of the five LV constructions with the verb facer.

−

Finally, the five LV constructions with the highest coefficient of variation are dar ouveos (66.1%), dar golpes (69.7%), ter vergonza (71.3%), ter tose (82%) and ter alerxia (87.9%). For these items, especially for the latter, the dispersion of the data is quite significant, since for many informants the fixation is maximal (10 respondents directly chose “10” and 16 respondents chose “9”), while for many others the degree of fixation is relatively low or very low (21 answered “0” and 28 chose “1”).

To better understand how the scores of participants were distributed, in Figure 15 we project them grouped into five blocks.

The highest marking percentages in the 9–10 block are concentrated in LLV cases, above all non dar golpe, (non) facer caso, terlles ganas, facer memoria and dar a lata, and almost always at a considerable distance from the other groups. The block with the lowest scores, 0–2, is most prominent in the following five CLVs: dar ouveos, ter tose, ter alerxia, dar golpes and ter vergonza. The remaining blocks, except for the 5–6 block in some specific constructions (facer burla, facer oración and dar testemuño), display medium or low values and are not generally the most salient across the different structures.

4.2. Assignment of Syntactic Function

Syntactically, in the LV constructions the weight of the pattern with direct object (DO) is very significant: the “V + DO” pattern accounts for 45.65% in Álvarez de la Granja (2003, p. 28), and the “VERB + DO (+IO and +Adjunct)” pattern accounts for 30% in Soto Arias (2022, p. 166). In general, all examples are syntactically interpreted as cases of verbs (predicates) with a functional structure involving a DO at the locutional (cf. Álvarez de la Granja, 2003, p. 28 and Soto Arias, 2022, p. 166), collocational (Gómez Clemente, 2024, p. 60 for dar testemuño; “substantivo (obxecto)” [“name (object)”], Corpas Pastor, 1998, p. 42, “complemento regido” [“governed object”] or Lorente, 2006, p. 103), or both levels. Barrios Rodríguez (2015, p. 15) refers to this as “Verb-DO”. Melendo, for Spanish, specifically uses the example dar calabazas “to rebuff a claim” as one that allows analyzing the noun as the DO of the verb dar (Melendo, 1965, p. 14).

In the survey we included an explicit question to check the extent to which informants continue to maintain the syntactic analyzability of the nominal unit in LVs as a direct object (DO). There are 13 examples that try to cover a varied spectrum of the studied LVs (both CLV and LLV) and their syntactic constructions, after verifying their representativeness in the corpus and their treatment in the literature (and within the time frame available to carry out the survey). In this question we included a control item, the first one, with a prototypical DO (Fixeron unha casa para toda a familia “they made a house for the whole family”), which the overwhelming majority of respondents (96%14) identified as such. The overall mean across the thirteen LV examples is 56.9% of respondents interpret these units as DO, versus 43.1% who do not view them as DO. In constructions identified in the literature as LLV, they are analyzed as DO in 47.7% of cases, nine percentage points below the mean, whereas in CLV constructions the reading as DO rises to 67.6%. The overall results are shown in Figure 16.

If we group the examples by the support verb, the means we obtain for identification of the syntactic function as DO are as follows: dar 62.1%, facer 57.7% and ter 48.4%. The means for the first two are higher—above the general mean—and stand apart from ter, which is the only one that does not reach even 50% interpretation as DO (which is not surprising, given that it is a prototypically stative verb). If we distinguish between construction types, we obtain the following distribution: LLV with dar 60%, facer 50% and ter 26.9%; CLV with dar 64.1%, facer 73.1% and ter 70%. With the verb dar (LLV) we have the highest proportion of syntactic reading as DO, at a very short distance from the percentage it displays with CLV, the lowest of the three verbs; facer, for its part, falls with LLV to half the interpretation as DO and is the one that presents the highest figure with CLV (23 points more than with LLV); finally, the verb ter shows the greatest contrast between the reading as DO with LLV—the lowest of the three—and with CLV, more than 43 points higher than the former.

Another aspect we do not want to overlook is the abstract or non-abstract nature of the noun forming part of the LV construction. If we group the five abstract examples (ter vergonza, (non) facer caso, facer memoria, (non) ter idea and ter lugar)15, the mean identification of the noun as DO drops to 42.3%, and rises to 77.5% for non-abstract nouns (acenos “gestures”, aviso with the meaning “texto, palabras ou sinais con que se avisa” [“text, words, or signs used to warn”], (cf. González González, 2012), s. v. aviso (cf. Alonso Ramos, 2004, pp. 153–154), tose “cough” and berros “shouts”), that is, more than a 35-percentage-point difference between them (this is consistent with Hopper and Thompson’s transitivity hierarchy and the “individuation of O” (Hopper & Thompson, 1980, p. 253)), which is lower with abstract nouns). In the subset formed by dar cabazas, dar a lata and non dar golpe (concrete nouns in constructions with abstract sense) the identification of these three examples as DO is 59.7%, halfway between the two previous sets. Furthermore, it should be highlighted that ter lugar is the only example that presents a non-prototypical, inanimate and postposed subject—that is, occupying the usual object position (cf. Section 2.1): “Mañá terán lugar todas as reunións” “All the meetings will take place tomorrow”—which also weighs on the very low functional-identification figures recorded for this construction (see, for medieval Galician, Rodríguez Guerra, 2004, pp. 173–180).

The comparison between those examples with and without a determiner is also significant: the proportion with which nouns with a determiner are identified as DO is 76.8%, which drops to 45.7% with nouns without a determiner (these data again agree with Hopper and Thompson’s (1980, p. 253) scale of transitivity, where a defined object is more individuated and, consequently, more affected). If we combine the presence of the determiner with the noun number, we see that when the noun is in the singular, the determinate examples reach 79% of identifications as DO, while those with the noun in the singular and without a determiner remain at 42.4%, more than 36 percentage points. When the noun is in the plural, the only example also with a determiner (dando eses berros “giving those shouts”) is identified as DO in 70.3%, practically the same proportion as that reached by the two examples in the plural and without a determiner, with 69.6%. If we count all the examples with the plural noun, the DO identification rate is 69.8%, while all the singular nouns have 54.6%.

Among the informants there are 14 specialists in Galician philology: on average they identify the function of DO with the nominal unit in 59.3% of occasions; this figure falls by more than 20 points in the case of LLV, to 38.8%, and increases by 24 points to reach 83.3% with CLV. Facer acenos reaches 100% interpretation as DO, followed by dar berros with 92.9%, and ter vergonza and ter tose with 85.7% each. The lowest is ter lugar, with only 7.1% interpretation as DO, followed at a great distance by (non) ter idea and non dar golpe with 35.7% each.

There are 48 informants who, as students or professors, belong to philological or linguistically based degree programs (the 14 specialists indicated above are included). The average identification of the syntactic function as DO is 59.5% (this is the highest mean among the participants’ combinations we analyzed), which falls to 44.5% with LLV and rises to 77.1% with CLV. The three examples most frequently identified as DO are ter tose (89.6%), and facer acenos with 85.4%; the lowest rates of interpretation as DO correspond to ter lugar (4.3%) and (non) ter idea (43.5%).

If we select the informants who, having Galician as their mother tongue (solely or not), state that they use Galician or more Galician in three basic social contexts (family, friends and educational institutions/workplace; n = 23), the average interpretation of this function as DO by these respondents is 53.8%; for LLV it decreases to 39.8% and for CLV it rises to 70.3%. The two expressions most readily identified as DO are dar o aviso with 90.9% and facer acenos with 87%. The three with the lowest interpretation rates as DO are ter lugar (17.4%), non ter idea (28.6%) and (non) facer caso (30.4%), all three LLVs.

If we apply the same selection as in the previous paragraph but focusing on informants who speak Spanish or more Spanish (and who include Spanish among their mother tongues; n = 29), we obtain a mean identification as DO of 57.8%, 50.2% for LLV and 66.7% for CLV. In any case, the most interesting aspect of these informants is the Certainty of Response Index (CRI), which we discuss below.

Focusing on the participants who self-rated their command of Galician with the highest scores (9 and 10 out of 10; n = 35), we see that the mean assignment as DO is 56.7%, which drops 13 points for LLV to 43.7% and rises 15 points for CLV to 71.9%. The CLV example dar o aviso displays the highest assignment as DO (91.4%), followed by ter tose (82.9%) and dar berros (80%). The constructions with the lowest rates as DO are ter lugar (11.4%), followed by non dar golpe and by golpes que deron with 34.3% each.

Beyond the first example in this question, with a canonical DO without an LV construction, we also included another, the last item, which included a relative clause with que and an explicit antecedent immediately preceding it (Faloume dos golpes que deron na parede “He told me about the blows they gave on the wall”). The relative construction created a confound with the possible reading as DO since, as Figure 16 shows, the majority (more than three quarters of the responses) do not analyze it as DO (it is also possible that the orthotypographic choice caused some confusion, since although the highlighted form, like the other examples, was the relative que, in bold and underlined, the antecedent os golpes was underlined to facilitate understanding. In this respect, it remains possible that some survey respondents incorrectly assumed that they were being asked to assess whether os golpes was the DO of Faloume.). From the perspective of LV structures, this result is a clear outlier because, in the case of a similar construction (dar berros), 70.3% do interpret the nominal unit as DO16.

We also consider the analysis of the Certainty of Response Index (CRI) to be very revealing, since “Yes” responses were always more frequent than “I think so” responses (except in the case of ter lugar), whereas “No” responses were less frequent than their respective “I don’t think so,” except in the following two examples: ter lugar and the construction with a relative (golpes que deron). To calculate the CRI we obtain the ratio between the percentages of “Yes”/“No” responses and those of “I think so”/“I don’t think so,” respectively17; we express that ratio on a base-1 scale: the closer the result approaches that value, the greater the certainty in the response (one would imply absolute certainty) and the closer that value approaches zero, the greater the uncertainty in the responses. The CRI for the 13 LV examples of Question II (Appendix A) for each example is as follows (cf.

Table 7. CRI of responses to Question II of the survey (Appendix A).

Q II	2	3	4	5	6	7	8	9	10	11	12	13	14
CRI “Yes”	0.64	0.55	0.64	0.56	0.59	0.53	0.66	0.65	0.59	0.56	0.29	0.61	0.59
CRI “No”	0.43	0.48	0.44	0.32	0.32	0.43	0.42	0.35	0.28	0.30	0.60	0.31	0.55
Global CRI	0.55	0.53	0.53	0.39	0.52	0.48	0.62	0.51	0.54	0.48	0.55	0.51	0.57

):

The overall CRI value is 0.52 (on a scale of 1), which, although it means that confident responses are slightly more frequent, in fact there is almost a technical tie between them and the hesitant choices and, therefore, the value of uncertainty is very high. The mean CRI for “Yes” rises to 0.59, whereas when “No” is chosen, the CRI falls to 0.43, quite distant from the former.

It is very interesting to observe the specific examples in light of their CRI, according to whether the noun’s function is identified as DO or not:

−

For the “Yes” responses, the constructions with the highest CRI (between 0.64 and 0.66) are ter vergonza, (non) facer caso, facer memoria and dar o aviso; the rest oscillate between 0.53 and 0.61, and there is only one with an extremely low CRI, 0.29, ter lugar (that is, those who chose to assign the DO function to lugar did so with very little certainty).

−

For the “No” responses, there are only two examples with a high percentage of certainty when the noun is not interpreted as a DO: ter lugar (with 0.60), and the example with a relative clause, golpes que deron (with 0.55). There are five examples that fall within a CRI range of 0.42 to 0.48 and the majority, six, have a low CRI between 0.35 and 0.28; the latter corresponds to tiveches unha tose louca, the others are three LLVs (facer memoria, (non) ter idea and dar cabazas) and two CLVs (facer acenos and dar berros).

−

Regarding the overall CRI, the highest, at 0.62, corresponds to dar o aviso and the lowest, with less than half that figure, corresponds to (non) ter idea. The second highest is relatively close to the first and corresponds to golpes que deron (0.57), and the penultimate lowest is non dar golpe (0.48). Note that the two highest CRI values are all CLVs with the verb dar and the two lowest are LLVs. The rest of the examples move within a narrow range, between 0.51 and 0.55.

As we noted above, among the 29 informants who speak only Spanish or more Spanish, there is an interesting result regarding their CRI: the overall mean is low, 0.43, which we can interpret as substantial uncertainty; this figure is identical to that obtained if we analyze the CRI of the “Yes” responses alone (it is DO), and it decreases considerably to a CRI of 0.37 when we analyze the “No” responses. This low certainty in the responses is not only due to the fact that these speakers show a clear preference for Spanish over Galician, but also because they are all students, with an average self-assessment of their command of Galician of 6.7 (out of 10) and, although there are some philologists (six) and five translation students, the majority are studying non-philological degrees.

Finally, if we look at the CRI of the examples with more abstract nouns (cf. supra) we obtain for them a CRI of 0.51 (same result as the subset of the three concrete nouns in abstract use), which contrasts with the other constructions with non-abstract nouns, which reach a CRI of 0.54. Specifically, the difference does not lie in the CRI of the “Yes” responses −0.56 for abstract nouns (just like concrete nouns in constructions with abstract sense) and 0.60 for non-abstract ones—but rather in the fact that with concrete nouns there is slightly greater uncertainty when discarding that the syntactic function of the noun is DO: the CRI of “No” responses is 0.43 for abstract nouns and 0.54 for concrete ones (0.41 for concrete nouns in constructions with abstract sense).

5. A Multiparametric Heterorganic Quantification of Syntactic Analyzability

Following the analysis of the examples in our corpus, we present a proposal for a multiparametric heterorganic quantification of analyzability, which allows us to obtain a Syntactic Analyzability Index (SAI). The higher this index, the closer a phrase is to unconstrained discourse, and the greater the likelihood of interpreting an LV + noun sequence in terms of the prototypical syntactic structure of PRED + DO. Conversely, the lower the index, the more likely it is that the sequence is a strongly fixed one, which makes it more difficult or prevents the interpretation of the noun as a direct object (DO). To obtain the SAI, we combine the following two complementary types of values: (i) we start from the morphosyntactic fixation scale (MFS) and (ii) combine it with the results obtained from the survey question on functional assignment.

5.1. Morphosyntactic Fixation Scale (MFS)

In Section 4.1, we already referred to the gradation and scalability of syntactic fixation (Alba-Salas, 2020, p. 26). A comprehensive bibliographic summary on the concept of structural fixation can be found in Martínez López and Reyes Tejedor (2024, pp. 682–683). These authors argue that “la fijación raras veces es absoluta” [“fixation is rarely absolute”] and that “como todo parece indicar, se dé con mayor intensidad en el ámbito de las locuciones” [emphasis added, “everything seems to suggest that it is more common with locutions”] and that it occurs with “menor intensidad en el de las colocaciones” [“it is less common with collocations”] (Martínez López & Reyes Tejedor, 2024, p. 683).

Scalar proposals are not new. Fraser (1970) proposed a “Frozenness Hierarchy”, a scale of syntactic fixation identifying seven levels, each associated with a different syntactic operation. The “LØ” level implies that “no operations whatsoever may apply to an idiom so characterized” (Fraser, 1970, p. 39), whereas levels 5 and 6 are “L5—Reconstitution” and “L6—Unrestricted.” Moreover, he establishes a dependency between them: “Any idiom marked as belonging to one level is automatically marked as belonging to any lower level.” This proposal was heavily criticized from a methodological perspective by, among others, Zuluaga (1975, p. 235) and Álvarez de la Granja (2003, pp. 218–221). The latter author emphasized the importance of working with real examples and with the different behaviors that, in this regard, various types of verbal locutions may exhibit compared to, for example, collocations with LV (Álvarez de la Granja, 2003, pp. 219–220).

Zuluaga raises several methodological issues in Fraser’s proposal and argues that it is necessary to determine the “grado de fijación” [“degree of fixity”] of an expression relative to others “en el marco de una investigación detallada” [“in the context of rigorous investigation”] (Zuluaga, 1975, pp. 234–235). The goal of our study is to understand the behavior of different units that share LV over time and, in this way, to distinguish, as objectively as possible, fixed combinations from unconstrained ones and “establecer los tipos de fijación que presentan” [“establish the degrees of fixation they exhibit”] (Zuluaga, 1975, p. 235), using a methodology that can be applied to other LV units. These parameters (“insertion of elements that do not belong to the phraseological unit” and “changing the order of the components”) were already proposed by Zuluaga (1975, pp. 235–236) to assess different degrees of fixation.

Berlage (2025, pp. 159–162) uses the ratios of determiners and modifiers in composite predicates to define a scale: their absence is associated with “syntactically fixed” composite predicates, and their presence implies “syntactically flexible” composite predicates. Our alternative advances this scalar conception, treats the functions of determination and modification differently, and increases the parameters used. In this way, we provide a methodological tool that increases its explanatory capacity to rigorously demonstrate this greater or lesser “prevalence” of syntactic fixation in LV constructions (cf. Wulff, 2013, pp. 285–286 regarding the relevance for speakers of variables used to judge idiomaticity). The criteria and weighting for evaluating morphosyntactic fixation on the MFS are as follows:

−

Numerical variation [0.5]: Variation in the noun number is valued proportionally18 based on the percentage of the least represented number (a 50% variation will imply the maximum score/variation, and a 0% will correspond to the minimum). As the main purpose of this section is to quantify as objectively as possible the (non)variation in the structures studied as a whole, we do not introduce any correction factor, for example, based on the nature of the noun (such as collective nouns, plurale tantum nouns or (un)countable nouns) which, obviously, may limit the possibilities of variation and which will have to be developed in the specific study of each construction).

−

Variation in the determiner [0.5]: The same criterion as the previous one applies, but regarding the (non)use of the definite/indefinite article with the noun. According to Álvarez de la Granja’s data from her corpus of Galician phraseological units, the pattern verb + def. art. + noun accounts for 24.71%, and verb + noun for 12.71% (Álvarez de la Granja, 2003, p. 57). For Galician collocations with the verb dar, Martínez Vicente (2023, pp. 160–163) initially observes great flexibility, and the predominant pattern is one in which the collocation admits both the definite article and the indefinite article as well as its absence, in combination with the noun in the singular or plural (38.91%), followed by a pattern similar to the previous one but in which, when there is no article, the noun appears only in the plural (29.33%). For this reason, it seems fundamental to us to quantify (as with the numerical variation) the greater or lesser fixation in the presence or absence of the article, which manifests itself in greater or lesser variation in its (non-)use (since both options may be grammaticalized).

−

Existence of nominal modifiers [0.5]: based on the data obtained for Galician, 0.1 is assigned for every 15%, up to the maximum score with 75% or more of examples having some modifier.

−

Preposing and contiguity between noun and verb [0.5]: we assess cases of noun preposing up to a maximum of 0.2 (which is achieved with 20% of preposings), while non-contiguity (always of elements that do not belong to the nominal unit, heterostructural, cf. Section 3.2.5) is quantified up to 0.3 (a score achieved with 30% non-contiguity).

−

Complement governed by the noun (CGN) [0.5]: the existence of CGN is valued proportionally, reaching the maximum score with 100% CGN occurrences.

−

Appreciative derivation of the noun [0.5]: the maximum score for this concept is achieved with 10% derived nouns, which decreases proportionally (with a minimum of 0.1, as long as at least one example is documented).

−

Coordination schemes [0.5]: when the noun is part of a coordination scheme, the same scale is applied as in the previous criterion.

−

Finally, we also consider three other parameters not focused on the noun [0.5]: the existence of a simple verb19, indirect object (IO), and relativization examples (with a total score of 0.5 for all three). The presence of a simple verbal correlate equivalent to the LV construction is valued at 0.1; the presence of an IO (lexical, clitic, or doubled) is quantified proportionally up to a maximum of 0.1 with 100% of contexts having IO (and a minimum of 0.01 whenever at least one example is documented); finally, a 15% or more relativization examples20 are valued at 0.3, and the score decreases progressively to a minimum of 0.01.

The maximum score on the MFS is 4, which would correspond to the greatest discourse variation, typical of unconstrained noun phrases. A very low MFS score, on the other hand, is associated with a high degree of structural fixation. Obviously, it is expected that constructions interpreted as LLV will occupy lower positions on the MFS, while collocational LV constructions will have higher scores, but never reaching the highest possible scores (comparable, as we suggested, to unconstrained noun phrases). We also understand that as the MFS score increases or decreases, the possibilities of syntactically analyzing the noun as a direct object also increase or decrease. Figure 17 shows the MFS for the 26 examples from our LV corpus.

From Figure 17, some conclusions can be drawn:

−

First, the maximum score reached by these LV examples is 2 (exactly 2.05 for dar golpes), far from the theoretical maximum score of 4 (percentagewise, this represents 51.3% of the maximum value).

−

The gradual progression is evident, occurring without significant jumps or gaps, although there is some specificity at certain points on the scale.

−

The distinct scores of LLV vs. CLV are confirmed: seven of the eight LLV examples occupy the seven lowest positions on the scale (with the sole exception of dar cabazas, which, with only 20 occurrences, presents considerable variation, including IO and relativizations), while the 18 CLV examples consistently receive higher scores (except for haber/ter medo).

−

However, not all units are concentrated around the same point; rather, there are specific behaviors. Of the LLVs, the following three show maximum fixation and, consequently, a very low MFS value: ter lugar (0), dar gloria (0.09) and non dar golpe (0.11). Also with a low score of 0.48 are facer caso and facer memoria, followed by three others in the 0.7 range: two LLVs, (non) ter idea (0.7) and dar a lata (0.73), and one CLV, haber/ter medo (0.74).

−

Relatively close to the latter are dar cabazas and dar opinión, with 0.92 each, along with ter alerxia (0.95) and facer mención (0.99), all below the value of 1. Most are CLV examples with lower variation.

−

The next group of CLVs is concentrated in the range from 1.16 for dar aviso to 1.24 for facer burla, with haber/ter vergonza (1.18) and dar testemuño (1.19) in between. These are followed by another subset of four CLV examples ranging from 1.41 for facer visita, to 1.47 for dar paseos, to 1.57 for facer acenos, to 1.61 for dar berros.

−

After this group, there is a small jump, and we find the last six cases of CLV, which display the highest MFS values and, consequently, a lower degree of fixation: they range from ter tose (1.91) and facer oración (1.92) to four examples that reach a score of two or higher: dar brados (2.01), dar couces and dar patadas with 2.02 each, and finally dar golpes (2.05).

In the construction of Figure 17 we used all available information from the medieval documentation to the present. To investigate what happens from a diachronic perspective with the MFS, we selected those LV examples with records from medieval Galician to the contemporary period, with a minimum of 75 occurrences in total. There are eight examples, one LLV (facer memoria [99]) and seven CLV (dar couces [165], dar testemuño [79], dar golpes [193], facer mención [157], facer oración [150], haber/ter medo [3020] and haber/ter vergonza [360]), which we also complement with dar gloria [97], (non) facer caso [1608], (non) ter idea [156] e ter lugar [996], with first attestations mostly from the 19th century. Figure 18 presents the results.

It is especially significant that the medieval period almost systematically presents a higher SMF (in six of the eight LV examples)21, with an average of 1.39 across all of them. In these eight examples with documentation in the four periods considered (all are CLV except facer memoria), the lowest average SMF, that is, with the greatest fixation, corresponds to the 21st century, with 0.89, followed by the period from the 16th to the 19th centuries with 1.13 and the 20th century with 1.19.

If we focus on the five LLV examples, in four of them it is again the earliest periods that concentrate greater variation, which changes markedly into significant fixation (facer memoria). This variation decreases but only slightly in (non) facer caso, or reaches total fixation, but starting from an already very high level of fixation (dar gloria and ter lugar). Only in one example, (non) ter idea, is the path followed the inverse, and in it, over the centuries, the SMF has almost doubled its value (see Section 3.2.8).

5.2. Usefulness of a Syntactic Analyzability Index (SAI)

As we have already suggested, we endorse a heterorganic conception of syntactic analyzability, in which the speaker plays a central role, reflected in the very definition of the concept (see Langacker, 1987, pp. 448 or 457, in the field of phraseological units, Baránov and Dobrovol’skij base the introspective approach on “intuición dos falantes” [“speakers’intuition”] Baránov & Dobrovol’skij, 2009, p. 264). Thus, to calculate the Syntactic Analyzability Index (SAI), we start with the data presented in the previous section (MFS) and combine it with the results from the survey on the syntactic interpretation of the nouns in LV structures. In Figure 19, both results are superimposed for the available examples.

In Figure 19, the correlation between the MFS and the assignment of the syntactic function of DO to the noun in the studied LV constructions is evident: the adequacy at the extremes is nearly perfect. The example with the lowest MFS, ter lugar (0), shows, by far, the most frequent interpretation in terms of non-DO (85.5%). Conversely, the four examples with the highest positions in the MFS display high levels of interpretation as DO, following a similar progression (ter vergonza, facer acenos and dar berros), culminating in ter tose (MFS of 1.91), which is identified as DO in 82.8% of cases. A similar adequacy is not observed with the intermediate examples, notably some LLV cases (such as dar a lata) or CLV cases (such as dar aviso). In this last example, it should be noted that the survey example that was asked about added a definite article and a prepositional modifier (o aviso do furacán tropical “the tropical hurricane warning”). The first example that was asked about in the survey is not usual, with the use of the p2 of the present subjunctive, directly addressing the reader, in an exhortative modality and with negative polarity (non me deas a lata “don’t annoy me”).

Based on the data shown in Figure 19, to refine the results and align them with an analyzability pattern, we calculate the global SAI for each example—on a scale of 100—by combining both values using the following formula:

$$\mathrm{S}\mathrm{A}\mathrm{I}={\displaystyle \frac{\left(\mathrm{M}\mathrm{F}\mathrm{S}\left[\mathrm{\%}\right]+\left(\frac{\mathrm{D}\mathrm{O}}{2}+\left(\left(\frac{\mathrm{D}\mathrm{O}}{2}\right)\times \mathrm{C}\mathrm{R}\mathrm{I}“\mathrm{y}\mathrm{e}\mathrm{s}”\right)\right)\right)}{2}}$$

where

−

MFS [%]: value in the MFS expressed as a percentage (cf. Figure 17 in Section 5.1).

−

DO: percentage at which the function of DO is identified in our survey (cf. Figure 16 in Section 4.2).

−

CRI “yes”: this is the index for the certainty of affirmative responses, applied to half of the percentage identifying the syntactic function as DO (cf. Table 7 in Section 4.2).

For example, for non dar golpe, the formula—with rounding to one decimal place—is:

$$\mathrm{S}\mathrm{A}\mathrm{I}={\displaystyle \frac{\left(2.8\left[\mathrm{\%}\right]+\left(\frac{42.8}{2}+\left(\left(\frac{42.8}{2}\right)\times 0.53\right)\right)\right)}{2}}=17.8$$

Thus, the SAI for non dar golpe is 17.8. Figure 20 presents all the SAI results, calculated in this way, for the available examples on a scale of 100.

The global average SAI for all the examples shown in Figure 20 is 34.6. This figure is equidistant between the average for LLVs (24.9) and CLVs (48.1), the latter figure nearly doubling the former, although it does not reach 50. At the extremes of this index, the examples with the lowest and highest analyzability indices stand out, both with the verb ter: the lowest (ter lugar) and the highest (ter tose). However, the latter is still far from the theoretical maximum achievable by unconstrained noun phrases (none of these LV examples reach a score of 60). The rest of the LLV constructions occupy positions with (very) low syntactic analyzability, although they are divided into the following two blocks: non dar golpe, (non) ter idea, (non) facer caso, and facer memoria, in contrast to dar a lata and dar cabazas, which, within the LLVs, show a specific behavior that places them relatively close to the CLVs. As for these, one case, haber/ter vergonza, is very close to the last two LLV examples, while the next three are positioned a clear step above (dar berros, facer acenos and dar aviso), although still several points below ter tose, discussed above. It is also worth noting that the general LV verb (dar, facer, or ter) does not determine a specific position on the index. If we compare Figure 17 and Figure 20, we see that, in general, there is correspondence with the main shifts concentrated in the following two cases: (non) ter idea, which decreases in the SAI, and dar aviso, which increases.

6. Coda

In conclusion, through our analysis, we observe the gradation of syntactic analyzability in the studied Galician LV constructions, a gradation that places the LLVs and CLVs in the expected positions on the scale, but this gradation is also reflected in the examples belonging to each group, showing a gradual evolution between them, without any significant radical breaks. Our proposal makes the greater or lesser syntactic fixation of LV constructions measurable and perceptible, both synchronically and, above all, diachronically, since it “renders visible” and to some extent objectifies (de)grammaticalization processes over time. This will not only also facilitate the projection of discrete cuts (if so desired) in the different constructional typologies, and, most importantly, will allow for easy application to other examples, other variants, and other languages. The more interlinguistic results can be objectively compared, the better we will understand the converging behaviors and the specificities of each language code. Furthermore, we believe that this methodological proposal is sufficiently flexible to admit new parameters or modify the ones presented here or their quantification. In this regard, it is true that the results of a survey can vary significantly from one instance to another; however, what is truly important is that the perception of speakers is also researched and integrated with the information obtained from the linguistic analysis of specific examples. Variations are acceptable because they will never be absolute (indeed, in Section 4.2, we observed that the overall mean variations between different informant typologies were not so pronounced), and they can always be minimized, as we did, for example, with the inclusion of the CRI. Incorporating the multifocal opinion of speakers into a solid corpus analysis of analyzability (and more) is essential and transforms the canvas of study of our language into a dynamic, polychromatic mural. Although it provides a snapshot of a specific moment, along with other images in diachrony, it will eventually constitute an authentic and rich linguistic mosaic, which will be more easily understandable.