The controller-first constraint beyond the Basic Variety: how do instructed learner varieties solve contexts of competition?

1 Introduction

Interest in exploring L2 grammars and uncovering their common regularities and stages by looking at the learners’ production increased steadily in Europe from the 1970s on and crystalized in several collaborative projects.^[1] This body of research focused on migrants that found themselves working in an environment whose language(s) they were actually learning at the time of the data collection in naturalistic, non-guided contexts. The existence of the Basic Variety is one of the best-known discoveries among second-language acquisition researchers and students. It was introduced by Klein and Perdue (1997) to counterbalance the prevalent “target deviation perspective” on language acquisition and biased teaching methodologies, such as error analysis. The Basic Variety is defined as a relatively stable system that is characterized by a small number of organizational principles and largely – though not totally – independent of the specifics of source and target language organization (Klein and Perdue 1997: 303). One of the principles determining the utterance organization in the Basic Variety is the semantic controller-first constraint, which says that “the NP-referent with highest control comes first”. In the Basic Variety, there is no verb inflection yet, although forms may show some unsystematic variation. A verb form may come with one, two, or (rarely) three noun phrases, resulting in the patterns shown in Table 1. The semantic controller-first constraint comes into play when at least two entities participate in an event, which may thus prompt the production of verbs admitting two or more arguments (i.e., transitive and ditransitive verbs).

In Table 1, the sentence The girl stealed the bread illustrates the pattern shown in PH1b by realizing the preverbal NP the girl as well as the postverbal NP the bread. The sentence Charlie give present for young children illustrates the subsequent pattern PH1c by realizing the preverbal NP Charlie as well as the postverbal NPs present and young children. We will henceforth refer to NPs as arguments, intended as event participants of the semantic representation of a verb. A verb’s arguments are syntactically mapped into utterances by their instantiations of thematic roles (i.e., subject, direct object, indirect object). If we assume that the controller-first constraint guides the type of language regularly developed during second language acquisition (SLA), then we must assume that learners will rarely produce constructions topicalizing the undergoer. This means that the controller-first constraint precludes the production of the following constructions:

Now, Klein and Perdue (1997: 329) also report the existence of discourse contexts in which the protagonist in topic function loses control, or where a topical NP-referent may simply not be the controller, as is the case in English passive sentences. They refer to these communicative situations as “contexts of ‘competition’” (Klein and Perdue 1997: 332, following Bates and MacWhinney 1987). Because of the missing verbal morphology characterizing the Basic Variety,^[2] the production of sentences similar to (1a), (2a), and (2b) seems to be unrealistic at this stage. Our study aims to test whether and to what extent post-basic varieties override the controller-first constraint by describing the grammatical and lexical means used by learners to topicalize undergoers.^[3] In so doing we share a common concern with Klein and Perdue that goes beyond SLA: the “speculation that the BV simply and directly reflects the necessary […] properties of the human language capacity” (Klein and Perdue 1997: 304). Our hypotheses are as follows:

1. The semantic controller-first constraint is an elementary principle of the human language capacity that is more or less salient across natural languages depending on the syntactic templates most frequently used in each language.

2. Compared to naturalistic learning by immersion, explicit teaching in the classroom not only reflects natural principles of the human language capacity, but also the effect of a particular curricular program targeting very specific features of the target language.

In Section 2, we introduce some basic concepts of argument linking. This is meant to provide the basis for establishing the semantic controller-first constraint as an elementary principle of the human language capacity. In Section 3, participants and methods are presented. Section 4 unveils some preliminary results from instructed French L2-varieties in contexts of ‘competition’ (i.e., where the controller-first constraint is in conflict with the segment of information highlighted by the speaker). L2 data are subsequently compared with the target language (Section 5) and the learners’ L1 (Section 6) so that a baseline is provided for further interpretation of the results. In Section 7, we discuss the results, explain them and put them in relation with previous studies having contributed to a better understanding of the syntax-semantic mapping. In Section 8, we summarize the main contributions and limitations of our survey, and point out further perspectives for future research.

2 The notion of controller: origins and related theoretical underpinnings

When talking about the notion of control, Klein and Perdue (1997: 314) mention Comrie’s (1989 [1981]) idea of a continuum. Comrie (1989 [1981]: 59) established a “continuum of control” as a universal scale on which various semantic roles – i.e., agent, force, instrument, and patient – can be situated. This concept replaces the assumption of a one-to-one relation between semantic case and morphological case marking. As Comrie (1989 [1981]: 60) shows, there is no such one-to-one mapping across languages, and further fine-grained distinctions capturing degrees of control are needed to explain the syntactic realization of an argument with a certain case or another. It thus becomes clear that the notion of control refers to a relation between the predicate – i.e., the verb – and one of its arguments. Other proposals that have proposed solutions to the question of how many semantic roles should be assumed are Dowty’s (1991) proto-roles and the macroroles of the Role and Reference Grammar (RRG; Foley and Van Valin 1984; Van Valin and LaPolla 1997; Van Valin 2005). Both of them merge all possible semantic roles under a binary opposition: proto-patient versus proto-agent (following Dowty 1991), and actor versus undergoer (following RRG). The notion of control becomes explicit again in this basic opposition, in which the actor is defined as the participant which “performs, effects, instigates, or controls the situation denoted by the predicate”, whereas the undergoer “is affected by it in some way” (Foley and Van Valin 1984: 29).

In their explanation of the controller-first constraint, Klein and Perdue (1997: 314) also refer to the control asymmetry between referents of noun phrases. This is an important observation that specifically concerns canonical transitive verbs, intended as two-place predicates with two clearly distinguished argument roles (Wunderlich 2006b: 63). The asymmetry captures the fact that the logical representation of prototypical transitive verbs – e.g., to chase, to eat, to hit, to kill, to kiss – make available two hierarchical argument slots: a higher argument and a lower argument. Table 2 summarizes the asymmetry of transitive verbs.

Satisfying the controller-first constraint requires that the argument with highest control be referred to before the argument with lowest control. In English – as well as in Romance languages – this can be done by using SVO active voice sentences in which the controller is syntactically realized as subject (e.g., The girl stealed the bread, La chica robó el pan). Sentences of this kind are typically presented to illustrate the default, non-marked word order. Still, when we talk about word order we talk about a certain argument being realized as subject/nominative versus object/accusative, so we are actually talking about argument realization as well. This said, the controller-first constraint does not imply the controller being realized as subject. Case marking is just one device among others to mark agency. In a sentence like (3), the fact of having an active voice sentence in the default word order does not imply a reading of the subject as a controller but rather as a patient.^[4] This is so because the lexical decomposition of the verb is another important parameter contributing to the meaning of a sentence. The English verb to die can be read as an accomplishment if the predicate denotes a durative change, or as an achievement if it denotes a punctual change. Accomplishment, achievement, and causative predicates denote a change of state, and therefore are somehow related to affected arguments, as can be observed in (3) for the NP the dog.

Word order, semantic roles, argument realization, and the modifiers of the verb are interacting factors at play to mark affectedness and agency, and thus necessary ingredients for the analysis of the controller-first constraint. Table 3 offers a schematic overview of the factors that influence argument realization. The particular device that maps argument roles onto morphosyntactic patterns is called argument linking (Wunderlich 2006b: 60). Not all languages share the same argument-linking type, yet each language represents an economic way of avoiding ambiguities with verbs having more than one nominal argument. A common assumption shared in the domain of functional linguistics is the Actor-Undergoer Hierarchy (cf. Figure 1). This concept is based on the idea that the ‘argument of DO’ (agent) is the unmarked – or default – choice for actor and ‘argument of pred´ (x)’ (patient) is the unmarked choice for undergoer.

Figure 1:

The Actor-Undergoer Hierarchy (Van Valin and LaPolla 1997: 146).

The Actor-Undergoer Hierarchy is intended as a universal mechanism of the language faculty beyond the syntactic templates available in a given language, and beyond its argument-linking rules, too. In this sense, the controller-first constraint is compatible with RRG and other theories that reject a purely syntactic perspective. Section 3 introduces language-specific syntactic templates that either neutralize or violate the controller-first constraint in French.

3 Empirical survey targeting contexts of competition

Discourse contexts in which the protagonist in topic loses control – or where the protagonist is simply not the controller – are relevant for the topicalization of undergoer referents. Section 3.1 offers a crosslinguistic overview of the constructions and grammatical means for topicalizing non-agentive referents in French and Spanish. Section 3.2 introduces the participants and methods that made the elicitation of spoken data in contexts of competition possible.

3.2 Participants and methods

Twenty Spanish-speaking learners of French took part in the survey, aged between 20 and 42, and coming from different Spanish-speaking countries, nineteen women and one man.^[7] All of them were university students benefiting from guided instruction in French at the time of the data collection: six Erasmus students, three regular MA students, and eleven students that were taught at the Institute of French as a foreign language. All participants completed a placement test ranging from proficiency levels A1 to B2 (cf. Figure 2). Additionally, they provided information about other known languages and broader bio-linguistic data by completing a questionnaire. This additional information revealed individual differences concerning the age at which participants were exposed to French for the first time. Eleven participants reported having started studying French in their twenties or later, seven reported having started at the age of puberty in an academic context, whereas two reported having benefited from input in French from childhood (cf. Tables 4 and 5).^[8] Levels of proficiency in French are displayed in Figure 2, and means of learners’ age and learners’ age of onset in French are provided in Table 6.

Figure 2:

Level of proficiency in French, age, and age of onset across learners.

Table 4:

Age of participants and age of onset in French reported by B1/B2 learners.

B1/B2 (n = 12)	#1	#2	#4	#5	#7	#8	#12	#15	#16	#19	#20	#21
Age	21	27	20	21	20	20	35	30	42	25	27	23
Age of onset	10	15	10	12	11	2	34	29	40	24	26	22
French input as a child	Yes	No	No	No	No	Yes	No	No	No	No	No	No
Studiesa	E	MA	E	E	E	MA	FFL	FFL	FFL	FFL	FFL	MA
Proficiency	B2	B1	B1	B1	B1	B1	B1	B1	B1	B1	B1	B1

1. ^aE, Erasmus; MA, Master; FFL, Institute of French as a foreign language (University of Nantes).

Table 5:

Age of participants and age of onset in French reported by A1/A2 learners.

A1/A2 (n = 8)	#3	#6	#9	#10	#11	#13	#17	#18
Age	22	22	28	34	20	34	29	27
Age of onset	12	12	14	33	19	33	28	26
French input as a child	No	No	No	No	No	No	No	No
Studies28	E	E	FFL	FFL	FFL	FFL	FFL	FFL
Proficiency	A2	A2	A2	A1	A1	A2	A2	A2

Table 6:

Means of learners’ age and learners’ age of onset in French.

	Age of onset		Age
	Mean	Standard deviation	Mean	Standard deviation
A1/A2 (n = 8)	22.1	8.99	27.0	5.37
B1/B2 (n = 12)	19.5	11.38	25.9	6.89

The French control group consisted of 21 native French informants, aged between 19 and 40: 15 women and six men. Most of them were students in the third year of a BA degree (10 participants) or in their first or second year of an MA degree (10 participants), and only one was already employed after having completed her MA studies. The data collection for this group took place at the University of Nantes. The Spanish control group consisted of 21 Spanish-speaking informants: 11 women and 10 men aged between 18 and 37, coming from different Spanish-speaking countries (13 Spanish, 4 Colombian, 2 Mexican, 1 Peruvian, and 1 Bolivian). The data collection of this group took place partly at the University of Nantes (11 informants), and partly in the Spanish region of Valladolid (10 informants).

The three groups completed an offline task in which they were presented with a blue-dressed cartoon figure (i.e., the NP-referent in topic) and were asked to retell what happened to that figure. In total, eight vignettes were displayed individually on a screen. Four vignettes were target items representing the blue-dressed cartoon figure as the undergoer and the other four were fillers representing him as an actor (cf. Figures 3 and 4).^[9] When the informant had finished describing the first vignette, we passed on to the second one, and so on. There was no time limit for the display of the visual stimuli. In order to avoid biasing effects linked to the order of presentation of the items, the vignettes were presented in two different orders, resulting in version I and version II of the data set. Out of the 62 participants, 31 completed version I and 31 completed version II. The audio files of the 63 participants taking part in the survey were time-aligned and transcribed using the free toolkit EXMARaLDA. The target items’ verbalizations were categorized in an Excel file and stored as tab-delimited text files in order to be processed as input files by R (using the RStudio environment). Responses varied in length across informants. A major methodological issue was to individuate the segment encoding the loss of control by the NP-referent in topic. In order to do so rigorously, we applied the following hierarchical criteria:

1. Utterances expressing unpleasant, non-volitional actions by means of the verb accordingly to Gaatone’s (1983) analysis of French;

2. utterances in which the two hierarchical argument slots depending on transitive verbs – i.e., a higher argument and a lower argument (Wunderlich 2006b: 63) – are fulfilled and thus, made explicit by means of proper nouns, subject pronouns, object pronouns, etc.;

3. whenever several utterances fulfill these criteria, only the utterance produced first was considered.

Figure 3:

Retelling task’s target items.

Figure 4:

Retelling task’s fillers.

4 Contexts of competition and borderline cases of learner varieties

There are some contexts where the controller-first constraint comes into conflict with other constraints that allegedly govern the BV as well, such as the pragmatic constraint “Focus expression comes last” (Klein and Perdue 1997: 317). Example (4) illustrates this conflict with a violation of the controller-first constraint. Sentences like (4) are typically produced either in contexts where the NP-referent in topic may not be the controller, or in contexts where the protagonist in topic loses control. By topicalizing^[10] the object Brot, the speaker may emphasize brand-new information about the stolen object (informative focus), or highlight a specific object over a set of objects in competition (contrastive focus: e.g., ‘it was not milk but bread’). In any case, the pragmatic focus-last constraint clearly outweighs the semantic controller-first principle in this case. Example (5) illustrates the production of cleft constructions in learner varieties by native Spanish speakers. The French morpheme [se] – whose orthographic form is c’est – is a focus marker used to introduce a new referent different from the topic (i.e., la dame, ‘the woman’) about whom the speaker provides additional information by means of a subordinate clause introduced by the relative pronoun qui, ‘who’.

The examples above illustrate two different strategies to resolve the conflict: the former ranks the pragmatic constraint over the semantic one (4), whereas the latter uses a cleft construction, which allows the speaker to mark either what is in focus or who is the controller (5). Beyond cleft constructions, this can be done as well by using specific constructions that mark non-controller referents over controller referents, as previously shown in (1a) and (2a). The production of any construction which allows the topicalization of non-controller referents was targeted by a pilot study exploring learner varieties based on L2 guided instruction (Laboratoire de linguistique de Nantes [LLING] 2023). This survey was based on an offline task in which participants were asked to retell what happened to a blue-dressed cartoon (i.e., the NP-referent in topic), and whose target items presented the participants with contexts where the NP-referent in topic loses control. The analysis of learner varieties was complemented by analyses of L1 grammars corresponding with the target language (French controls) and the learners’ first language (Spanish-speaking controls) in order to get a baseline enabling the interpretation of the results.^[11] In what follows, we provide some examples that illustrate either neutralizations (6) or violations (7) of the semantic controller-first constraint.

Example (6) does not provide the hearer with the information of a control asymmetry, because the two arguments rank equally. A symmetrical reading of this type is typically expressed by reciprocal se–verbs^[12] in French and Spanish (Lewandowski 2021). By contrast, example (7) expresses a control asymmetry in which the argument with less control is syntactically realized as subject. This is possible in French beyond passive sentences by means of the verbal periphrasis se faire + infinitive. Examples (6) and (7) come from native French speakers and therefore, may be taken as instances of natural languages, which are borderline cases of learner varieties (Klein and Perdue 1997: 308; Klein 2018: 467). Frequency-based analyses run on data from learner varieties confirmed this point at least for the construction illustrated in (7). The overriding of the controller-first constraint by means of se faire + infinitive was significantly less frequent in the learner varieties than in native French (X (2, N = 164) = 35.258, p < 0.001).^[13] Qualitative analyses on se–reciprocal verbs revealed that they were used differently by learners and French controls. In learner varieties, se-reciprocal verbs were always linked to a singular NP-referent working as the syntactic subject and thus, marked a control asymmetry in which the NP-referent in topic had the highest level of control (8). In French, these verbs were either used to mark an equal ranking of arguments – as shown in (6) – or to mark a control asymmetry in which the referent with lowest control was introduced by a prepositional phrase, as exemplified in (8).

Observations like these raise the following two questions: are the learner varieties under analysis more controller-first oriented, compared to French? And if so, what role do typological features of Spanish play? This second question is motivated by the assumption that learners transfer the priorities of their source language to their learner varieties, as claimed by Klein and Perdue (1989) and other SLA scholars sharing a cognitive approach to grammar (Lambert et al. 2022; Slobin 1996). In the next section, we present the most frequent constructions found in the learner varieties in contexts of competition.

5 Common traits of learner varieties in contexts of competition

Controller-first constructions were frequently produced in the learner varieties in combination with accomplishment, achievement, and causative predicates (18.7 %, example 9), and less frequently with the causative periphrases faire + infinitive or faire que + subjunctive (6.2 %, example 10). In both cases, the NP-referent in topic was syntactically realized as object, and semantically intended as an undergoer.

In addition, controller-first constructions were also used to mark the NP-referent in topic as the argument with highest control by subject realizations in predicates whose verbs carried the modifiers [BECOME], [INGR] or [CAUSE] (11.2 %, example 11), or activity verbs.^[14]

Some responses were not clear-cut controller-first nor undergoer-first. They contained state predicates (e.g., avoir peur, avoir un accident) in which the topical NP-referent was realized as subject and intended as an internal causation’s experiencer (example 12). 17.5 % of the learners’ responses were of this type.

The only undergoer-first construction relatively spread out across the learner varieties was the passive with être (13), representing 16.25 % of the learners’ responses. This type of construction was also present in French controls, although at a lower frequency (i.e., 8.3 %).

Figure 5 offers an overview over the distribution of the controller-first, middle-voice, and undergoer-first constructions across groups (in percentages). The controller-first category covers SVO active voice sentences in which the verb contains a modifier component that expresses a change of state, or a change in one of the properties of the affected argument (i.e., what in Section 2 has been broadly called “accomplishment, achievement or causative predicates”). Controller-first constructions were also marked either by SVO active voice sentences carrying an activity verb (e.g., Il est en train de courir, FR2_10#13) or the causative periphrases faire + infinitive and faire que + subjunctive. Se-reciprocal verbs were counted within controller-first readings when they provided the hearer with a control asymmetry, as shown in example (8). Conversely, they were counted separately when their use expressed a neutralization of the control asymmetry, as in example (6). The former reading was found both in learner varieties and in French, whereas the latter was only produced by French controls. Middle-voice constructions consisted in SVO active voice sentences with state verbs (e.g., Alex a peur de le chien, FR2_12#25). Undergoer-first constructions cover both passive sentences with être and se faire + infinitive. An additional construction counted as undergoer-first was the lexical collocation Verb + NP (e.g., Il s’est pris une voiture, FR_04#07; Alex est encore victime de violence, FR2_12#28).^[15] Beyond these readings, Figure 5 shows other categorizations taking into account intransitive predicates and productions in which the NP-referent in topic was not explicitly mentioned.

Figure 5:

Distribution of patterns across groups (in percentages).

To sum up, 47.5 % of the learners’ responses were clear-cut controller-first constructions, as against 21.25 % corresponding to undergoer-first constructions. In addition, the proliferation of state predicates – corresponding to 17.5 % of the learners’ responses – enabled a sort of middle-voice reading in which the NP-referent in topic appeared as experiencer (Maldonado 2009). These distributions are different from those found in French controls, for two reasons: first, because in the responses provided by native French speakers the amount of undergoer-first constructions amounted to 55.9 % (including both se faire + infinitive and passive sentences with être, plus lexical collocations verb + NP); and second, because state predicates presenting the NP-referent in topic as experiencer amounted to 8.3 % of the French controls’ responses.^[16] In the next section, we look into the Spanish controls’ production in order to explore whether the learners’ L1 may be taken as an explanatory factor concerning the salience of controller-first readings in the learner varieties under analysis.

6 Similarities between learner varieties and L1 Spanish

Figure 6 compares the frequencies of constructions used by learner varieties and Spanish-speaking controls. The most evident observation when looking at Figure 6 is that learner varieties and Spanish controls seem to converge in the high frequencies of controller-first readings, on the one hand, and in the relatively lower frequencies of undergoer-first readings, on the other hand. This similar pattern may be explained as an effect of the L1 during the process of “thinking for speaking” (Slobin 1996), or as a sort of conceptual transfer, intended as the influence of the language-mediated conceptual categories of one language on verbal performance in another language (Jarvis and Pavlenko 2010: 115). However, middle-voice constructions are dissimilar between Spanish-speaking controls and learner varieties, being more frequently used by the latter, as was also observed when comparing learners’ productions to French controls’ productions.

Figure 6:

Distribution of patterns across groups (in percentages).

Overall, the frequencies displayed in Figure 6 show the pervasiveness of the controller-first constraint both in learner varieties and in the learners’ L1. From a comparative viewpoint, this can be interpreted as Spanish being more controller-first oriented than French, at least in contexts where passive readings are expected. This explanation seems to be consistent with the claim made by Gauchola (2012: 163) about the higher frequency of use of passive periphrastic constructions in spoken French (i.e., either passives with être or se faire + infinitive) compared to Spanish. The higher frequencies of middle-voice readings compared to either French or Spanish may well be a developmental trait linked to the learners’ L2 proficiency. Qualitative analyses of the learner varieties at a proficiency level of A1 or A2 showed an overt preference for state predicates (e.g., avoir peur, *être peur, avoir un problème, avoir un inconvénient, avoir un accident, *être un accident), whereas at levels B1 and B2 a larger repertoire of causative and achievement verbs was used (e.g., frapper, taper, mordre, faire un croche-patte, fuir, *choquer).^[17]

If we reconsider the idea that natural languages are borderline cases of learner varieties (Klein 2018: 467; Klein and Perdue 1997: 308) in light of the frequencies presented in Sections 5 and 6, we must conclude that several observations that we made are consistent with this claim. Learner varieties differed from target language native speakers in the use of undergoer-first constructions, which were underused by the former. The only occurrence of the passive auxiliary se faire found in L2 grammars^[18] was produced by a B1 proficiency learner that reported to have lived in France during her early childhood, before moving back to Spain with her parents and older sibling. Among the 20 learners that took part in the survey, two reported in the pretest bio-linguistic questionnaire to have grown up in a multilingual environment where they were exposed to French input. The general profile of the learners’ sample corresponded to a learner with previous knowledge of English, and whose age of onset in French was at puberty or later. The two learners that reported early input in French were thus exceptional. Moreover, among these two cases, just one used the preferred French controls’ undergoer-first construction. This seems to support Klein’s (2018: 486) claim that ‘perfect mastery’ is just a special case of a learner variety. Undergoer-first readings differed not only in terms of frequency when comparing learners’ and French controls’ production, but also in the grammatical means used by each group. Learner varieties overused passive sentences with être, which is a construction not only available in Spanish, but formally similar from a morphological viewpoint^[19] and with no semantic restrictions.^[20] If we assume that developing learner varieties is managing to copy, with varying degrees of success, the ways in which other people speak (Klein 2018: 468), one may interpret the overuse of passive sentences by learners as a way to cope with the saliency of undergoer-first constructions in the target language, at least for causative, non-volitional predicates. But even in the case where learners were sensitive to the undergoer-first oriented saliency, they do not seem yet to have copied into their learner varieties the construction preferred by native French speakers in such contexts. This gap between target language uses and target language constructions, plus the structural similarities between constructions available in both the source language and the target language have been pointed out as a factor that may impede the learning of more target-like structures (Benazzo and Andorno 2017: 140).

7 Discussion

Two hypotheses outlined in the Introduction motivated the present study. In the first part of our discussion (Section 7.1), we will attempt to answer the question of whether the controller-first constraint can be regarded as an elementary principle of the human language capacity. In the second part (Section 7.2), we discuss whether and to what extent the oral production of the instructed learners under analysis can be explained in terms of (a) the specifics of the source language, and (b) the foreign language classroom setting.

8 Limits and outlook

More than 30 years after the first description of the Basic Variety (Klein and Perdue 1992: 30), its discussion continues to generate challenging questions. The motivation of our study was to shed light on how instructed learner varieties dealt with contexts of competition and, ultimately, to test whether the semantic controller-first constraint is copied in such contexts or, conversely, overridden. The results presented in Section 7 provided an empirically tested answer to this question. However, we are aware of the limits of the survey presented, as well as the necessity of complementing our data with online tasks simultaneously measuring L2 production and the learners’ preverbal attentional resources. In the aim of testing the pervasiveness of controller-first in contexts of competition, expanding the learners’ language configurations beyond Romance languages is essential, too. The results presented here are preliminary because they are based on a relatively small sample. Furthermore, the categorization of utterances according to the grammatical and/or lexical means of the verbal phrase did not allow us to analyze discourse functions of topicalization that were used transversally across some of the learners’ productions (cf. examples in Note 16). Despite these remarks, we consider the theoretical and empirical claims made in the present article robust and hope that it can serve as an attempt to rekindle the body of research devoted to the Basic Variety.