Emerging Constructions in the L2 Italian Spoken by Low Literate Migrants

Abstract

The emergence of autonomous interlanguage constructions is widely recognised in the literature on L2 Italian. These constructions involve the overgeneralisation of functional forms learners are in the process of acquiring, e.g., siamo in siamo mangiare ‘be:1PL eat:INF’ (target Italian: mangiamo ‘eat:1PL’); facciamo in facciamo cucinare ‘do:1PL cook:INF’ (target Italian: cuciniamo ‘cook:1PL’); per in piaciare per uscire fuori ‘like:INF for go:INF out’ (target Italian: mi piace uscire fuori ‘to.me like:3SG go:INF out’. ‘Be’/‘do’ forms are assigned a morphosyntactic function to convey temporal/aspectual/person information instead of inflecting the verb, while per ‘for’ is a generic subordinating marker. Based on new corpus data, I claim that such constructions may correlate with a learners’ degree of first language (L1) literacy. In detail, and consistent with, both literate and non-literate learners overgeneralise functional forms while working on the newly acquired morphosyntax; this shows that the non-literates are perfectly able to subconsciously identify functional forms in the input. Non-literates, however, show a stronger tendency than literates to select lexical-syntactic sub-patterns.

1. Introduction: Literacy in New Migration and the Status of the Research

Low literacy represents an important factor among adults in the recent migration towards Italy (and Europe in general). While official data are still lacking,1 some information on literacy rates in migrant populations derive from the surveys conducted in the last years at the University of Palermo, in Sicily, that is, the Italian region on the Mediterranean Sea which is the most involved in the migrant welcome system.

Between 2017 and 2018, the researchers at the School of Italian language for Foreigners (henceforth, ItaStra) of Palermo developed a literacy test, which has been administered to a sample of 774 migrants, both middle-term residents (one to five years) and the newly arrived, from North Africa, Sub-Saharan Africa, China, Bangladesh, and to a lesser extent, parts of the rest of Europe (Serbia). The literacy test is based on various L1s (Bambara, Mandinka, Pulaar, Wolof) or languages of schooling (Arabic, English, French, Italian) and it includes different writing systems (Arabic, Bengali, Chinese, Hindi, Pulaar, Tamil) along with the Roman alphabet. About 31% of the resulting sample was not fully literate in their L1 or in an early-learnt post-colonial language, with a lower percentage for women (26.72%) and a higher one for men (32.84%) and for adolescents/minors (35.88%) (Amoruso and Lo Maglio 2018; D’Agostino and Lo Maglio 2018, pp. 24–25). These results mainly refer to migrants from Sub-Saharan Africa and Bangladesh and confirm UNESCO’s estimates that the lowest literacy rates worldwide are observed precisely in these parts of the world due to poor access to school, early school leaving or poor-quality education (UNESCO 2017). As the data were elicited through a dedicated tool—rather than based on migrants’ self-report or inferred from reported schooling experience—the ItaStra data are the first reliable data on the degree of literacy of older migrants.

A point that deserves attention is the frequent correlation between migrants’ literacy and their L2 competence, which in Italy—as elsewhere—can be very low after a medium or even long length of residence. A language test administered in parallel to the literacy test revealed that 58.5% of the entire sample and 49.22% of medium-length residents had proficiency in L2 Italian at A1 of the CEFR (Common European Framework of Reference for Languages) or even below A1 (D’Agostino and Lo Maglio 2018, p. 24). This may be due to low exposure to the target language, as older migrants generally receive little exposure to the L2 in large part due to low levels of interaction with native speakers.2 Whether or not low literacy skills further slow down the acquisition or produces different results, with respect to those described in the existing literature on second language acquisition, still needs to be verified.

Research on the impact of low L1 literacy on L2 acquisition is, however, still peripheral to the wider field of second language acquisition (SLA) and has mainly been carried out by researchers involved in the international association LESLLA (Literacy Acquisition and Second Language Learning for Adults). LESLLA studies suggest that some aspects of second language development may, in fact, be related to learners’ degree of L1 literacy (see Tarone 2010; Tarone and Bigelow 2005; Tarone et al. 2009; Vainikka and Young-Scholten 2007; Vainikka et al. 2017; Young-Scholten and Strom 2006)3. These studies have, however, largely focused on English as well as on Dutch (e.g., Julien et al. 2016; Oldenkamp 2013; Sanders et al. 2014; van de Craats 2011).4

In Italy, as in other countries with large immigrant populations, there is research not just on L2 acquisition of, for example, morphosyntax, but also on migrants who were L2s in naturalistic contexts (such as the studies collected in Giacalone Ramat 2003a), but no explicit attention has been dedicated so far to the variable literacy.

To fill this gap, a research project on the acquisition of L2 Italian morphosyntax by low- or non-literate adult learners was launched in 2016 as a part of ItaStra research activity. The project involved the collection of oral production data in the form of recordings and transcriptions of young adult migrants’ speech at different points in their acquisition path. The entire group of participants was recruited during administration of the literacy test at ItaStra, and they were both literate and low- or non-literate and all were acquiring L2 Italian naturalistically.5 The analysis of this Italian LESLLA interlanguage corpus allows the researcher to verify the degree to which the existing descriptions of L2 Italian also apply to LESLLA learners, in terms of route, rate and outcome of the second language acquisition process.

In what follows, attention will be focused on a well-known phenomenon in L2 Italian, namely the overgeneralisation of functional items while learners are in the process of acquiring verbal morphosyntax. This may result in the emergence of independent interlanguage constructions, that is, forms and constructions which do not exist in the L2 input and whose occurrence seems to be systematic in learners’ interlanguages. In this paper, I will look at two different types of constructions: (1) constructions resulting from the overgeneralisation of auxiliaries (e.g., essere ‘to be’) and light verbs (e.g., fare ‘to do’); (2) constructions resulting from overgeneralisation of prepositions (or adverbs), such as per ‘for’. These cases are exemplified in (1)–(2) and (3), respectively:

(1)	Auxiliary construction
	siamo	mangiare
	be:1PL	eat:INF
	‘We are eat.’
	mangiamo	(target form)
	eat:1PL
	‘We eat.’

(2)	Light verb construction
	facciamo	cucinare
	do:1PL	cook:INF
	‘We do cook.’
	cuciniamo	(target form)
	cook:1PL
	‘We cook.’

(3)	Preposition construction
	piaciare	per	uscire	fuori
	like:INF	for	go:INF	out
	‘To like for go out.’
	mi	piace	uscire	fuori	(target form)
	to.me	like:3SG	go:INF	out
	‘I like to go out.’

In (1) and (2), the ‘be’ and ‘do’ forms are assigned the morphosyntactic function of conveying temporal, aspectual and person information instead of the uninflected verb (the dependent infinitive form). In other words, they express at the analytical level what the target language expresses in a synthetic way through inflection. They are not target-like and they do not derive from the input, because in Italian neither auxiliaries nor fare is used with an infinitive form. In (3), per is a generic subordinating marker employed in inter-clausal linking. Again, this is not target-like and is not from the input, because per is only used in adverbial (purpose) infinitive clauses and not for completive clauses. What all these constructions have in common is that they are learners’ strategies to reconstruct function/form relations which they seem to be aware of—even if they still lack the morphosyntactic means to encode them. Both (1)-(2) and (3) are lexical-syntactic strategies that express these relationships in a more transparent way than the target language morphosyntax does. These lexical-syntactic strategies are referred to as “interlanguage constructions” (henceforth, ICs).

The existence of ICs (at least, those in (1) and (3)) has been observed not only in L2 Italian research (Andorno et al. 2003; Banfi and Bernini 2003), but also in L2 Dutch (Julien et al. 2016) and L2 English (Vainikka et al. 2017) and it has been interpreted in fact as a temporary strategy to make up for the lack of the target morphosyntactic means.

The analysis of this Italian LESLLA corpus will take the steps from this tradition of studies, but the results will be interpreted taking into account learners’ degree of literacy. Against this background and in the light of current insights on the role of literacy in second language acquisition (in particular, Vainikka et al. 2017), I will argue that:

• ICs involve overgeneralisation of functional forms belonging to closed lexical classes; learners identify these forms in the input as elements conveying grammatical meanings and overextend them to non-target contexts.
• This means that, when ICs emerge, learners must have already entered a “grammatical” phase of second language acquisition.
• Both literate and low-/non-literate learners develop ICs while working on acquiring new structures; this shows that not only literate but also low- and non-literate learners are able to subconsciously identify functional words in the input.
• Learners may favour this due to the specific status of function words, which are more transparent than bound morphemes both at the phonological and at the semantic levels.
• In general, non-literates show a stronger tendency to select lexical-syntactic sub-patterns.

The rest of this article is organised as follows: in Section 2, I report the main results in LESLLA studies on the acquisition of L2 competence and, in particular, morphosyntax. In Section 3, I summarise the path of acquisition of L2 Italian as it has been described in the functionalist theoretical framework of the Basic Variety (Klein and Perdue 1997), which is the benchmark of Italian studies of immigrant adults. Section 4 is dedicated to previous literature on L2 ICs. In Section 5, I present the profile of LESLLA participants involved in this study and the method for collecting data. Data analysis is discussed in Section 6. While the background of the analysis is the functionalist perspective described in Section 3, I will also refer to hypotheses elaborated within a different theoretical approach, namely the Organic Grammar, which as reported in Section 2, provided convincing claims on the impact of literacy in second language acquisition (Vainikka and Young-Scholten 2007; Vainikka et al. 2017). This is reported in Section 7, dedicated to discussion and final remarks.

2. LESLLA Studies

L2 linguistic competence is far from being a major topic of research within Literacy Education and Second Language Acquisition by Adults (LESLLA) studies. Since its establishment in 2005, LESLLA research has mainly been devoted to the development of literacy skills and to related pedagogical challenges (Young-Scholten forthcoming). The few scholars interested in the role of L1 (alphabetic) literacy in second language acquisition fall into two main perspectives:

(1)

the cognitive perspective, inspired by the findings of psycholinguistic experimental studies on phonological awareness and working memory (cf. the research conducted by Tarone and colleagues since the early 2000s);

(2)

the linguistic paradigm, represented by the Organic Grammar (OG), a weak continuity generativist approach (Vainikka and Young-Scholten 1998, 2007; Vainikka et al. 2017; Young-Scholten and Strom 2006).

One of the main divides between the two perspectives is the role assigned to explicit knowledge in adults’ L2 acquisition. In the first case, acquisition is viewed as a largely conscious process involving explicit attention to linguistic forms (Schmidt 1990). In the second case, acquisition is conceived of as an entirely subconscious process guided by internal linguistic mechanisms and triggered by mere exposure to primary linguistic data (that is, the language spoken by natives). When it comes to the analysis of non-literate learners’ second language acquisition, there is consensus in both perspectives that they acquire L2 linguistic competence more slowly compared to educated, literate learners; however, the interpretation of this observation varies considerably and a greater emphasis is given to phonological awareness or to phonological competence, depending on the theoretical stance assumed.

It is commonplace in psycholinguistics research that the ability to segment the speech into non-semantic units—in particular, into phonemes—and to consciously manipulate these units depends on alphabetic literacy (cf. Goswami and Bryant 1990). On the other hand, literacy does not affect the ability to process oral speech in semantic units. Reis and Castro-Caldas (1997, p. 445) found that “literate individuals develop a strategy where visual-graphic meaning is given to units that are smaller than words and units with no semantic meaning. These segments are introduced sequentially in a working memory system with a new content of visual experience. Then, we can play with those written symbols, each coded to a sound, for example, to form pseudowords with no semantic meaning.” Thus, phonemic awareness also entails the ability to segment the oral input into words as phonological units, independent of lexical semantics, that is, to identify word boundaries in the speech continuum (Reis et al. 2007; see also, in LESLLA research, Kurvers et al. 2007; Onderlinden et al. 2009).6

The advocates of the cognitive perspective conclude that: (1) if literacy affects the (conscious) processing of the oral input and (2) if conscious processing of the oral input is crucial to acquiring the second language, (3) then literacy must have important consequences in second language acquisition. For instance, low- and non-literate learners will struggle more than literate ones to acquire functional units smaller than words such as morphemes (e.g., English third person singular -s), as well as purely functional words lacking semantic content (e.g., articles or auxiliaries) (Bigelow and Tarone 2004; Tarone 2010; Tarone and Bigelow 2005; Tarone et al. 2007, 2009).

Advocates of the generativist perspective argue instead that the relationship between phonological awareness and second language acquisition cannot be taken for granted. Along these lines, Young-Scholten and Strom (2006) found a correlation between low-/non-literacy and morphosyntax development but caution against interpreting this as causal. Other factors—including linguistic factors—can still be at work. One of these factors is phonological competence. Vainikka and Young-Scholten (2007, p. 143) propose that literacy affects the acquisition of phonological competence in a second language, which in turn may result in incomplete analysis of morphological constituents. This hypothesis was elaborated in Vainikka and Young-Scholten (1998) adopting their Organic Grammar approach. Resting on one line of generative grammar assumptions (weak continuity), they argue that at the beginning of the acquisition path learners do not project any functional syntax, despite the opportunity of transferring this from their L1s. Learners’ initial interlanguages are “minimal trees” whose headedness is based on their L1s. When it comes to acquire functional elements, because they differ across languages, they must be acquired in response to the input. Unlike in L1 acquisition, where bound morphemes (e.g., inflectional affixes) trigger L1 morphosyntax acquisition, it is free morphemes that do so in L2 acquisition. This relates to the phonological difference between bound and free morphemes: “Free morphemes such as auxiliaries typically constitute at least a phonological foot, while bound morphemes typically involve units smaller than a foot. Lack of phonological attainment may in turn result in incomplete analysis of sub-foot constituents in the learner’s L2” (Vainikka and Young-Scholten 1998, p. 106). Thus, free morphemes are more easily perceived and acquired, and it is possible that a bound morpheme can never act as a trigger, resulting in a fossilised non-target grammar if a parameter setting, for example, can only be triggered by a bound morpheme. This can be the result of little exposure to the target language, which frequently happens in adult migration contexts.

Exposure is even poorer for low-/non-literate learners who cannot access written texts and, hence, have fewer chances to experience the target language morphosyntax and only rely on aural input. According to Vainikka et al. (2017), this could explain why the non-literate learners they studied overgeneralised function words as what they claim are “placeholders”. They also observe that these adults, who were in a literacy class, overgeneralised multi-word sequences not directly related to the actual verbal head to mark morphosyntactic functions in L2 English (e.g., in the to mark progressive aspect, as in in the drink; in the no cooking). The occurrence of such sequences shows that learners are able to subconsciously identify functional vs. lexical forms in the input, presumably based on what was presented in the classroom. It also shows that they select a “heavier” form compared to inflection which Vainikka and Young-Scholten say is most likely due to “a greater reliance on auditory as compared to visual memory” (p. 248).

These hypotheses will be particularly relevant to the analysis of L2 Italian data.

3. The Path(s) of Development of L2 Italian

The most systematic part of Italian research on Second Language Acquisition has taken place in the functional-typological framework. According to this theoretical model, language consists of bidirectional function-to-form mappings that can favour synthetic (morphological) or analytical (lexical-syntactic) options, according to patterns which vary and are cross-linguistically recurrent (Comrie 1989; Givón 1990). Interlanguages fall in the category of typological descriptions: although influenced by the properties of the source and target languages, they are nevertheless autonomous linguistic systems that learners develop spontaneously as a consequence of exposure to the target language. Therefore, interlanguages should adhere to organisational principles consistent with those underlying any other natural language (Giacalone Ramat 2000, 2008, 2013; Tomlin 1990).

An influential description of the organisational principles governing the structure and the development of the interlanguages was proposed by Klein and Perdue (1997), based on the results of the European Science Foundation Project (Perdue 1993). The focus of this description is the notion of Basic Variety (henceforth, BV), a simple yet structured initial stage of interlanguage. It exhibits organisation in which utterances contain verbs and are structured according to their valency, but there is no trace of inflection. Nouns and verbs occur in an invariant form (e.g., the infinitive form or the stem), while information about temporality, aspect, modality, person, number and gender are conveyed by non-inflectional means; that is, lexical items such as the numeral tre ‘three’ in (4), which quantifies the uninflected noun lingua ‘language’.

(4)	Io	parla	tre	lingua
	I	speak:BASIC FORM7	three	lingua:SG
	io	parlo:1SG	tre	lingue:PL	(target form)
	‘I speak three languages.’		(ItaStra LESLLA Corpus)

Before BV emerges, second language learners use a pre-grammatical variety of interlanguage, the so-called pre-Basic variety, characterised by unstable syntax and word order governed by pragmatic principles (i.e., topic/focus organisation). The BV then represents the very first appearance of the L2 grammar which, irrespective of the organisational principles of the target language, is mainly governed by semantic and lexical(-syntactic) principles (lexical encoding of grammatical categories). Many learners, especially in on-going conditions of low exposure, tend to fossilise at the BV stage (cf. Klein and Perdue 1997, p. 332), which is in fact claimed to be effective on the communicative level. Under adequate conditions of exposure, however, the acquisition of the target language develops in a series of successive post-Basic varieties through which morphosyntax becomes more and more complex.

Based on these theoretical assumptions, Italian research has provided descriptions of many aspects of learners’ morphosyntax.

Table 1. The continuum of the L2 Italian varieties (based on Andorno et al. 2003; Banfi and Bernini 2003).

Variety	Main lexical Categories	Functional Elements	Inflection (Tense-Aspect-Mood)	Inflection (Person)	Utterance Organisation
Pre-basic	no distinction first, second, third personal pronouns	some marks of negation (no, non) (some) conjunctions (e ‘and’, anche ‘also’, poi ‘then’) unanalysed existential c’è ‘there is’	none	none	pragmatic principles (topic/focus) interclausal links (juxtaposition, coordination)
Basic	N and V are distinct basic form of V (theme, unanalysed form of the present, infinitive)8	(some) prepositions (new) conjunctions adverbs	none		argument structure ➔ semantic-pragmatic principles (Agent-V-Object) (➔ syntactic principles: SVO) interclausal links (juxtaposition, coordination)
Post-basic		copula (some forms)	-to (past participle)		syntactic principles (SVO) prepositions governing Ns
		perfective auxiliaries (essere, avere)		differences in the present tense	interclausal links (subordination: adverbial, i.e., causal ➔ temporal ➔ final adverbials)
		imperfect (some forms of ‘be’: ero, era)			interclausal links (subordination: completive ➔ relative)
		auxiliary stare ‘stay’ + gerund (progressive)	-v- (imperfect morpheme for thematic verbs)
			gerund
			future
			conditional mood subjunctive mood

summarises the main results on the acquisition of the verb and the utterance organisation in L2 Italian, as they have been reported respectively by Banfi and Bernini (2003) and Andorno et al. (2003), also on the basis of the research conducted in the previous decades.

As Table 1 shows, the acquisition of the Italian verbal system develops along the following sequence:

Basic form (pseudo-present form or theme, infinitive) > past participle (and auxiliary + past participle, i.e. the perfective past tense passato prossimo) > imperfective past tense (imperfetto) > future > conditional mood > subjunctive mood (Banfi and Bernini 2003, p. 90).

This sequence is implicational because the emergence of a form on the right of the development chain entails the presence of the forms on the left.9 The sequence also suggests that the path of development is aspect > tense > mood: the first functional opposition is between the perfective form—to and the non-perfective form represented by the uninflected basic form (e.g., cerca-re/cerca ‘to seek/seek’ vs. cerca-to ‘sought’); then an aspectual difference emerges within the past, namely perfective vs. imperfective past tenses; the progressive construction stare + gerund is a further aspectual specification that emerges after the imperfective and before the future tense (Giacalone Ramat 1997, 2003b); non-factual moods (conditional, subjunctive) emerge later on (Giacalone Ramat 1995, 1999a). Before a given morphological form emerges, the functions associated with it can be conveyed by other lexical items such as adverbs (e.g., ieri ‘yesterday’ to express past tense) or verbs (e.g., pensare ‘to think’ for epistemic modality) (Bernini 1995, pp. 302–303; Giacalone Ramat 1992, pp. 309–11).

Three further observations are needed in order to interpret Table 1 properly:

• Different columns represent different paths of developments that are related but not necessarily dependent on each other; in other words, it may be the case that one path develops more slowly than another, as in the case of the acquisition of person markers. These markers emerge immediately after the first aspectual opposition perfective vs. non-perfective, but they stabilise much more slowly than temporal-aspectual markers (Banfi and Bernini 2003, p. 95).
• It follows from 1 that a category should be considered acquired even if the form in which it is encoded does not perfectly reflect the target form; e.g., io era ‘be:IPFV.3SG, I was’ (target form: io ero) reflects the acquisition of the imperfective past tense of ‘to be’, but not that of the grammatical person. In other words, the acquisition of a grammatical category does not correspond automatically to the acquisition of the whole target morphology.
• Even within the same column, a category first appears in individual forms, rather than involving the entire paradigm; e.g., the imperfect tense generally emerges as individual forms of ‘to be’ (era, ero) and only later it spreads to other verbs (Andorno 2006).

At the very beginning of the acquisition path, L2 utterances have a pre-syntactic organisation; this is because words do not yet belong to a specific lexical category and, hence, the distinction between noun and verb has not yet emerged. The position of the words within the utterance reflects their information status (topic, focus) rather than grammatical relations (subject, object), as in (5):

(5)		[parlare	italiano]TOPIC	[poco,	no tutti]FOCUS
		speak:INF	Italian	a little,	not all
		parlo	poco	italiano,	non molto	(target form)
		‘I speak Italian a little, not much.’			(ItaStra LESLLA Corpus)

Nouns and verbs start to be used as distinct categories in the Basic Variety. According to Bernini (2005), an important role in triggering this passage is played by the existential construction which is very frequent, and, hence salient in the input—especially in the third singular person c’è ‘there is’ (locative clitic ci ‘there’ + third singular person of ‘to be’). This form appears early in the interlanguages, but in contrast to the target language it is an unanalysed chunk (and should be better indicated as ce), as testified by its frequent overgeneralisation as in c’è erano ‘there is (they) were > there were’. Besides the existence of a state of affairs, c’è may also indicate ‘location’ (‘NP c’è NP’, e.g., l’uomo adesso c’è America ‘the man now there is America > now the man is in America’) and ‘possession’ (NP [+ human] c’è NP, e.g., lui c’è un molie ‘him there is a wife > he has a wife’). In other words, “c’è expresses a generic relationship whose nature is specified by that of the NPs involved” (Bernini 2005, p. 168, translation the author’s). As it occurs as the first and unique predicative element already in the pre-Basic variety, the existential construction can be considered as the forerunner of the distinction between nouns and verbs. At a slightly later stage, c’è will act as a “bridge” for the emergence of essere ‘to be’ (both copula and auxiliary).

When the verb emerges as a distinct category, utterance starts to organise around an argument structure, even if the verb still lacks inflection. This phase of the utterance organisation is, in fact, governed by semantic principles, namely the argument with more control of the event (typically the Agent) tends to occur first, while the controlled arguments tend to be post-verbal (Agent–V–Patient). This semantic order overlaps with the target morphosyntactic order (SVO) of Italian because controllers tend to be encoded as subjects (Andorno et al. 2003, p. 132). This explains why SV(O) order stabilises early in Italian interlanguages irrespective of the learner’s L1 word order.

Utterance organisation becomes more clearly syntactic in the post-Basic varieties. This can be observed in the emergence of prepositions governing nouns to signal their semantic role, although they are frequently overgeneralised to non-target contexts (io scrivo con [target: in] lingua arabe ‘I write in Arabic’, Andorno et al. 2003, pp. 133–34; Bernini 1987, pp. 136–37). At this stage, prepositions start to also be involved in subordination. This development follows a phase of mere juxtaposition between clauses and is accompanied by frequent overgeneralisation of the prepositions to non-target contexts.

4. Previous Studies on Interlanguage Constructions

4.1. Non-Target Analytical Constructions

The presence of interlanguage constructions has been observed early in L2 Italian studies. Bernini (1989) described some examples of analytical constructions in which overgeneralised forms of essere ‘to be’ and, to a lesser extent, avere ‘to have’ co-occur with unanalysed lexical forms, so that the grammatical meaning and the lexical meaning are encoded separately. In other words, essere and avere are auxiliaries conveying grammatical information instead of the lexical verb. This means that when these constructions emerge, learners must have already individuated in the input some forms of auxiliaries (and copula, in the case of essere) and are in the process of acquiring them. For instance, as already seen in Section 3, the acquisition of the imperfective past tense starts with the acquisition of individual forms of ‘to be’ which can be used to create provisional analytic past forms of lexical verbs, as in (6), where ‘to be’ marks continuous aspect in the past instead of the uninflected lexical verb (Giacalone Ramat 1992, p. 307).

(6)		Lui	era	ha	una	macchina
		he	be:IPFV.3SG	have:BASIC FORM	a	car
		lui	aveva		una	macchina	(target form)
			have:IPFV.3SG
		‘He had a car.’

According to Banfi and Bernini (2003, p. 106), lexical forms may occur as:

• pseudo-present verbal forms, as ha in 6 (other examples are: ero sono ‘I’m was > I was’; era si chiama ‘he was he is called > it was called’; avevo credo ‘I had I think > I thought’);
• infinitives, e.g., sono studiare ‘I am to study > I study’;
• unanalysed constructions, e.g., sono sto facendo ‘I am I’m doing > I’m doing’, siamo non ha fatto ‘we are (he) didn’t do > we didn’t do’;
• nouns, e.g., siamo partenza ‘we are departure > we leave/are leaving’, sono paura ‘I am fear > I’m afraid’).

All the examples resulting from combination of essere/avere with the lexical or lexicalised forms in the list above are not target-like because Italian auxiliaries can only be linked to the past participle of the lexical verb to form analytic perfective past tenses, as in (7).

(7)		sono	andato	a casa
		be:1SG	go:PST.PTCT	to home
		‘I went home.’

Bernini (1989) and Banfi and Bernini (2003) report on data elicited from learners with various linguistic and socio-cultural backgrounds.10 The presence of ICs has been also observed by Andorno (2010), Banfi (2003) and Valentini (1992) in Chinese learners acquiring L2 Italian in both naturalistic and instructed learning contexts. None of these studies explicitly include learners’ degree of literacy to account for patterns found.

There have, however, been studies of LESLLA learners of L2 Dutch which analysed the overgeneralised use of is ‘be:3SG’ and ga ‘go:1SG’ in the interlanguages of low-educated (although not necessarily non-literate) Moroccan learners (van de Craats 2009; van de Craats and van Hout 2010). They find that is and ga are “dummy auxiliaries” (or placeholders); that is, desemanticised verbs, which convey morphosyntactic information instead of the verb they occur with. Consistent with what has been observed in L2 Italian, dummy auxiliaries emerge at a stage in which lexical verbs still lack morphology, but learners are already aware that verb phrases must contain grammatical information. These forms are in fact more transparent and easier to be perceived than inflectional morphology, frequently associated instead with phonological features such as consonant clusters which are difficult for Moroccan learners of Dutch. As soon as proficiency in the target language increases, dummy auxiliaries decrease. While the is-pattern has been widely described in both L2 Dutch and other L2s with equivalents of the verb ‘to be’ (cf. inter al. Fleta 2003; Huebner et al. 1992; Jordens and Dimroth 2006; Schimke 2013; Starren 2001; von Stutterheim 1986), the ‘go’-pattern has only been observed in Moroccan learners of Dutch, presumably as a result of the similarity with the Moroccan Arabic future auxiliary ġadi. A more recent study carried out with learners with different linguistic backgrounds (Julien et al. 2016, p. 69) scaled down the role of the L1 in learners’ selection of dummy auxiliaries, showing that this is directly influenced by target language input and the learners’ level of proficiency. This study is based on 40 adult learners of Dutch with a low educational level, as none of them had attended university and six of them had never attended school.

On this research background, my analysis will explicitly take into account the variable literacy, observing the linguistic behaviour of learners who have different literacy levels, but are similarly exposed to input in L2.

4.2. Non-Target Subordinating Markers

In his 1987 seminal study on prepositions in L2 Italian, Bernini observed the overgeneralisation of per ‘for’ and di ‘of’ as marks of generic subordination.11 This phenomenon accompanies the very emergence of prepositions in the domain of subordination, after a phase of mere interclausal juxtaposition. Bernini reconstructed the overall path of syntactic development of prepositions as represented in

Table 2. The acquisition of prepositions in L2 Italian (Bernini 1987).

Phase 1	Ø + N (e.g., andato Napoli ‘gone Napoli > I went to Napoli’)
Phase 2	P + N (frequent overgeneralisation, e.g., l’amico per mia sorella ‘the friend for [target: di ‘of’] my sister’)
	V + Ø + V (e.g., io vado lavorare ‘I go work > I go to [target a ‘to’] work’)
Phase 3	V + P + V (frequent overgeneralisation, e.g., questo è pesante per spiegare ‘this is hard for [target: da ‘to’] explain’); only in a very late stage this pattern becomes target-like.

(cf. also Table 1):

Valentini (1998) noticed a similar phenomenon in the selection of the adverb come ‘as, how’ as an unspecified subordinator in the L2 Italian spoken by Chinese speakers. Consistent with Bernini (1987), she places this phenomenon after the stage at which learners use juxtaposition (V + Ø + V) and before their target-like use of subordinators. The path of acquisition of target-like subordination in Italian is then reconstructed as follows: adverbial (causal > temporal > final) clauses > completive clauses > relative clauses (cf. Andorno et al. 2003, pp. 161–65; Berruto 2001, p. 27; Giacalone Ramat 1999b, p. 539; Valentini 2001, p. 84).

My data do not allow us to verify whether the development path proposed by Valentini (1998) also applies to LESLLA learners, because these learners did not reach stages of interlanguage advanced enough to systematically observe subordinates beyond the adverbial ones. On the other hand, LESLLA data allow us to examine in depth the initial stages of development, those where general subordinator constructions emerge. LESLLA data also show that the impact of these ICs may relate to the degree of literacy in L1.

5. The Italian LESLLA Corpus: Participants and Data Collection12

The participants in this study were 20 newly arrived learners recruited during literacy testing at the School of Italian for Foreigners (ItaStra) of the University of Palermo in 2017–2018. Their data are summarised in

Table 3. LESLLA learners’ data.

Learner	Age	Country of Origin	L1 /Other Languages	Schooling	Early Literacy (L1/School Ls)	Residence	Courses in Italy	L2 Italian	Late Literacy in Roman Alphabet
AC	20	Nigeria	Ika; English, Pidgin Eng.	12 years	Group 3 English	18 months	6 months	pre-basic	-
AL	27	Nigeria	Urhobo; Bini, English, Pidgin	10 years	Group 3 English	18 months	5 months	basic	-
AO	24	Nigeria	Esan; English, Pidgin	12 years	Group 3 English	12 months	2 months	basic	-
BD	18	Guinea	Pulaar; Wolof, French	2 years (Q)	Group 1	11 months	5 months	basic	Group 2
CO	26	Nigeria	Ika; Igbo, English, Pidgin	12 years	Group 3 English	12 months	10 months	post-basic	-
GO	27	Nigeria	Esan; Yoruba; English, Pidgin	15 years	Group 3 English	16 months	9 months	post-basic	-
HL	25	Nigeria	Esan; English, Pidgin	-	Group 1	11 months	3 months	none	Group 1
ID	25	Ivory Coast	French, Kojaka; Bambara Malinki	12 years	Group 3 French	11 months	7-8 months	post-basic	-
LO	25	Nigeria	Igbo; English, Pidgin	-	Group 1	11 months	none	pre-basic	Group 1
MC	18	Gambia	Mandinka; Creole	3 years	Group 1	21 months	10 months	basic	Group 2
MD	30	Senegal	Mandinka; French, English	10 years (Q)	Group 2 Arabic	11 months	10 months	post-basic	Group 2 French
MF	28	Mali	Bambara; French	-	Group 1	12 months	7 months	basic	Group 2
MJ	24	Nigeria	Igbo; English, Pidgin Eng.	11 years	Group 3 English	11 months	11 months	post-basic	-
MLG	25	Burkina Faso	Bissa; Mòoré, French	5 years (Q)	Group 1	11 months	6 months	post-basic	Group 2
MT	23	Mali	Bambara; French	-	Group 1	11 months	6 months	post-basic	Group 2
MTR	25	Ivory Coast	Bambara; Senufo, Wolof, French	2 years (Q)	Group 1	11 months	9 months	post-basic	Group 2
OT	23	Gambia	Mandinka; Wolof, English	12 years	Group 3	16 months	4 months	basic	-
RC	18	Bangladesh	Bengali	8 years	Group 3 Bengali/Eng.	13 months	9 months	basic	-
SM	27	Bangladesh	Bengali; English	12 years	Group 3 Bengali/Eng.	12 months	none	basic	-
YS	30	Senegal	Pulaar; Wolof, French	2 years (Q)	Group 1	10 months	5 months	post-basic	Group 1

, where learners are indicated by the initial letters of their name and are sorted in alphabetical order.

As Table 3 shows, learners were young adults, male, aged between 18 and 30 years, arrived from Western Africa (Gambia, Senegal, Guinea, Ivory Coast, Mali, Burkina Faso, Nigeria) and Bangladesh. They were native speakers of various African West Atlantic (Pulaar), Mande (Bambara, Bissa, Kojaka, Mandinka), Benue-Congo (Esan, Igbo, Ika, Urhobo), Gur (Senufo) languages, and Bengali.

They had not been exposed to Italian before their arrival in Italy, between 10 and 21 months before the first interview. Most of them had attended Italian language courses after arrival, in volunteer-led contexts. This experience (if not too short or discontinuous) provided an opportunity for interaction with native speakers, which is in general rare for migrants, and allowed many of them to have already entered the initial stages of L2 Italian acquisition when the study began. However, apart from possible language courses, their exposure to Italian was inconsequential and, in fact, none of them had gone far beyond the non-morphological stage at the time of the first interview, despite a length of residence of many months.

Learners’ degree of schooling and literacy in L1 or in an early learnt language varies considerably. Based on the literacy test, learners were put into three levels:

• Group 1: no literacy (not able to read or write isolated words in any writing system).
• Group 2: low literacy (recognition of letters, in the Roman or other alphabets; slow deciphering of a few isolated words; writing his own name).13
• Group 3: literacy.

Home language literacy may result from school experience, but literacy in that or other languages may also be gained outside the school context, e.g., during migration or in the initial language courses in Italy (in this latter case, it is associated with learning Italian).14

Learners’ degree of literacy and L2 Italian level do not correlate, at least at the initial stages of morphosyntactic development, and the only variable at work in language acquisition is contact with the target language. We will see in Section 6 that the degree of literacy may, however, act in a subtler way in the stages immediately beyond the BV.

The development of learners’ interlanguages was traced through a longitudinal data collection of 13 months. During this period, learners were tested five times in individual sessions carried out in non-school settings, i.e., in the hosting centres. Sessions included a language and literacy test, interviews and narrative tasks. In detail:

• Session 1 included a preliminary interview, aimed at collecting biographic and sociolinguistic information, and the ItaStra literacy test.
• Session 2, took place a few days after and included: a) a wordless video (adapted from the web) which learners had to retell; the video showed in parallel the actions carried out by two persons, a man and a woman, during a day; and b) a set of images on paper representing daily events that learners had to select to talk about a day in the past. The tasks were aimed at eliciting data on the presence of the verb and the encoding of temporal-aspectual and person information.
• Session 3 was carried out after about six months. Learners were asked to talk about their life in Palermo, focusing on an event or a person. This task aimed at eliciting the same linguistic phenomena as Session 2 after a certain timespan.
• Session 4 took place after another four months and included a conversation about the learners’ life between the third and the fourth sessions and the retelling of Chafe’s film The Pear (Chafe 1980).
• Session 5 took place after another three months and included tasks performed in Sessions 1 to 4, namely a semi-guided conversation, the retelling of the video used in Session 2, the retelling of The Pear film.

All sessions were audio recorded and then transcribed using the software ELAN. Recordings and transcriptions—the analysis of which is still in progress—will be deposited in the archive of ItaStra. In the near future, they will be an open access resource.

6. Data analysis and Results

6.1. Sessions 1 and 2

Data analysis shows that the emergence of ICs first correlates with learners’ interlanguage stage. The first ICs emerge precisely in the transition from the Basic Variety to the post-Basic Variety. This transition stage is marked by the first appearance of past participle forms15 and, then, incipient forms of the copula (and possibly auxiliaries).

Table 4. ICs in learners’ interlanguage (Sessions 1, 2).

	Group 1			Group 2			Group 3
SESSIONS 1–2
	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic
Analytical ICs	-	Do-C	Do-C Be-C	-	-	Do-C Be-C	-	-	-
Subordinating ICs	-	-	-	-	-	-	-	-	-

shows the presence of ICs through the different L1 literacy levels (Groups 1 to 3).

Table 4 suggests that the interlanguage stage is a necessary, but not a sufficient, condition for ICs to emerge, as the level of L1 literacy appears to interact with the transition to post-Basic Variety. In particular, at this stage we find fare-constructions and essere-constructions (indicated as Do-C and Be-C, respectively).

In the first type, a form of fare ‘to do’ governs an uninflected verb (an infinitive or a basic form), as in (8) to (11):

(8)	BD2	io	poi	fatto	lavare
		I	then	do:PST.PTCP	wash:INF
		poi	mi	sono	lavato	(target form)
		then	myself	be:1SG	wash: PST.PTCP.M.SG16
		‘I then washed myself.’

(9)	MF2	fare	tirare	tenda
		do:INF	stretch:INF	curtain
		ha	aperto	la	tenda	(target form)
		have:3SG	open:PST.PTCP	the	curtain
		‘He raised the curtain.’

(10)	MT2	ha	fatto	pulire	denti
		have:3SG	do:PST.PTCP	clean:INF	teeth
		si	è	lavato	i denti	(target form)
		himself	be:3SG	brush:PST.PTCP.M.SG	the teeth
		‘He brushed his teeth.’

(11)	INT	la ragazza,	cos’	ha	fatto?
		the girl	what	has	done
		‘What did the girl do?’
	MT2	ha	fatto	mangiare,	anche
		have:3SG	do:PST.PTCP	eat:INF	also
		ha	fatto	prende	libro
		have:3SG	do:PST.PTCP	take:3SG	book
		ha		mangiato
		have:3SG		eat:PST.PTCP
		ha	preso	anche	un libro	(target form)
		have:3SG	take:PST.PTCP	also	a book
		‘She ate; she also took a book.’

Examples (8) to (11) show different interlanguage stages, namely that BD and MF still use their Basic Variety (or very initial post-basic: in BD2, fatto is a rare PP form), while MT already shows post-Basic features and, in fact, fare occurs in the perfective past tense (ha fatto). It should be observed that the construction may echo a form contained in the interviewer’s question (INT), as in 11 (‘what did the girl do’ ‘she did…’), but it also occurs autonomously, as in (8)–(10).

To my knowledge, no study has so far been devoted to the overgeneralisation of fare ‘to do’. Unlike in other languages, such as English for ‘do’, the Italian verb fare is not an auxiliary verb. It is rather a light verb, that is, a semantically weak verb that occurs in constructions with a second predicational element (e.g., nouns, verbs and adjectives) to form a complex predicate. Complex predicates are single units, i.e., their arguments map onto a monoclausal syntactic structure (Butt 2003). Fare is typically constructed with nouns (e.g., fare ricerca ‘to do research’, Quochi 2007) to which it attributes the abstract semantic component “activity”. I claim that, in learners’ interlanguages, fare is selected to play the same function with verbs that are still underspecified at the formal level. In other words, it makes explicit the predicative value of the uninflected, purely lexical verb it governs and, possibly, temporal, aspectual and modal information. Fare is a high frequency verb17 in the input, and its semantic lightness favours the generalisation to contexts where “verbal” information cannot yet be expressed morphologically. As in the case of the overgeneralised essere, this results in a construction heavier and more transparent than in the target language, that is, fare V.18

A few cases of essere-constructions emerge in a slightly more advanced variety, when learners are already in the process of acquiring the grammatical uses of ‘be’, i.e., copula and auxiliaries. Forms of ‘be’ can be observed in (12)–(14), where the negated è ‘is’, however, conveys neither temporal-aspectual nor person information, as both learners refer to the first person referent and, respectively, to past and future events. In these cases, è seems to express the (non-)existence of the state of affairs denoted by the dependent predicates and, thus, it might be linked to the existential construction.

(12)	MD2	non	è	continua	a	lavorare
		not	be:3SG	continue:INF	to	work:INF
		non	ho	continuato	a	lavorare	(target form)
		not	have:1SG	continue:PST.PTCP	to	work:INF
		‘I didn’t keep working.’

(13)	MD2	non	è	lavorare	più
		not	be:3SG	work:INF	anymore
		non	ho	lavorato	più	(target form)
		not	have:1SG	work:PST.PTCP	anymore
		‘I have not worked anymore.’

(14)	YS2	non	è	rimarci	per sempre
		not	be:3SG	remain:INF	forever
		non	rimarrò	qui	per sempre	(target form)
		not	remain:FUT.1SG	here	forever
		‘I will not remain here forever.’

Non-target subordinating markers do not occur in the recordings of Sessions 1 and 2.

ICs only occur in low-/non-literate learners’ speech.

6.2. Session 3

Table 5. ICs in learners’ interlanguage (Session 3).

	Group 1			Group 2			Group 3
	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic
Analytical ICs	-	-	Be-C	-	-	Do-C	-	-	-
Subordinating ICs	-	-	Per-C Come-C	-	-	Per-C	-	-	Come-C

shows the distribution of ICs in learners’ interlanguages as it emerged during Session 3.

On the whole, learners’ interlanguages do not show substantial development, although many of them have been included in linguistic and educational paths (to obtain the linguistic certification necessary for the stay permit) and/or found a job (mostly temporary, however).

The fare-construction still occurs in MT (as in 15), whose interlanguage shows the same stage as in Session 2, while it has disappeared in MF, whose interlanguage decisively shifted to the post-Basic Variety (most likely thanks to a systematic exposure to the target language in a new work context).

(15)	MT3	ho	fatto	con	mio amico	passeggiare
		have:1SG	do:PST.PTCP	with	my friend	walk:INF
		ho	passeggiato	con	un mio amico	(target form)
		have:1SG	walk:PST.PTCP	with	a my friend
		‘I went for a walk with my friend.’

On the other hand, BD, who also shifted from the Basic to the post-Basic Variety, developed the essere-construction, as in (16), together with copula and emerging auxiliaries, and this construction can be observed as well in YS, as in (17), while MD maintains the existential non è, as in (18).

(16)	BD3	domeniga	anche	poi	siete	un po’	legliere
		Sunday	also	then	be:2PL	a bit	read:INF
		domenica		poi	ho	letto	un po’	(target form)
		Sunday		then	have:1SG	read:PST.PTCP	a bit
		‘Sunday, then, I also read a bit.’

(17)	YS3	adesso	sono	estudiato	CPIA
		now	be:1SG	study:PST.PTCP	CPIA
		adesso		studio	al CPIA	(target form)
		now		study:1SG	at CPIA
		‘Now I study at the CPIA.19’

(18)	MD3	poi	un altro scuola	ma
		then	another school	but
		non	è	ricorda	nome
		not	be:3SG	remember:3SG	name
		poi	un’altra scuola	ma			(target form)
		then	another school	but
		non		ricordo	il nome
		not		remember:1SG
		‘Then another school, but I can’t remember the name.’

Data deriving from Session 3 also show the emergence of a few non-target subordinating markers, namely per ‘for’ and come ‘how, as’, in learners who have entered the post-Basic varieties. Per can be observed in BD, in MLG and in MF. It generally introduces a subject infinitive clause, as in (19):

(19)	BD3	A Palermo	per	vivere	non	è	facile
		at Palermo	for	live:INF	not	be:3SG	easy
		a Palermo	non	è	facile	vivere		(target form)
		at Palermo	not	be:3SG	easy	live:INF
		‘It’s not easy to live in Palermo.’

Come can be observed in CO, MC, YS. It can play the same function as per, as in (20), but it also introduces main clauses whose verb is uninflected, as in (21); in this second function, come is found in narrative sequences in which learners are listing habitual activities.

(20)	CO3	Io	no	ci	piace	come	guardare
		I	not	to.us	like:3SG	as	look:INF
			non	mi	piace		guardar=lo	(target form)
			not	to.me	like: 3SG		watch:INF=it
		‘I do not like to watch it.’

(21)	INT	Che cosa	fai	a	Palermo?
		what	do:2SG	at	Palermo
	CO3	Come	andare	scuola,	vedere	una persona
		as	go:INF	school	see:INF	a person
			vado	a scuola	vedo	persone	(target form)
			go:1SG	to school	see:1SG	person:PL
		‘What do you do in Palermo?’ ‘I go to school, I see people.’

Only one learner out of the eight who produce ICs in this session is literate, CO, who just overgeneralised subordinating markers.

6.3. Session 4

Table 6. ICs in learners’ interlanguage (Session 4).

	Group 1			Group 2			Group 3
	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic
Analytical ICs	-	-	Do-C Be-C Stay-C	-	-	Do-C Be-C	-	-	Do-C Go-C
Subordinating ICs	-	-	Per-C Come-C	-	-	Per-C Come-C	-	-	-

illustrates the development of ICs in learners’ interlanguages as observed in the fourth session.

The fare-construction is still present in BD and in MT, emerges in MC, MLG and CO (the only literate learner, but in this latter case we are only dealing with an occasional echo answer). The essere-construction persists in MD and YS and emerges in MC and MLG.

New auxiliaries emerge, namely andare ‘go’ and stare ‘stay’. The andare-construction in (22), however, is episodic and occurs only in the speech of the literate learner SM; the occasional auxiliary ‘go’ conveys temporal-aspectual information instead of the infinitive of the same verb. On the other hand, the overextended stare in (23) and (24) might be more systematic. In the target language, stare is constructed with the gerund (GER) to form the progressive construction; MC is in the process of acquiring this construction, which occurs in the (quasi)target form in (23) (i.e. stare + gerund, although the person agreement is still lacking), while in other two examples stare governs an infinitive form.

(22)	SM4	Tre	persone	vai	andare	giocare
		Three	persons	go:2SG	go:INF	play:INF
		tre	persone	vanno	a	giocare	(target form)
		three	person:PL	go:3PL	to	play:INF
		‘Three persons go to play.’

(23)	MC4	Loro	sto	camminando	avanti
		they	stay:1SG	walk:GER	on
		loro	stanno	camminando	avanti	(target form)
		they	stay:3PL	walk:GER	on
		‘They go forward.’

(24)	MC4	Lui	sto	guardare	questa	donna
		he	stay:1SG	look:INF	this	woman
		lui		guarda	questa	donna	(target form)
		he		look:3SG	this	woman
		‘He looks at this woman.’

(25)	MC4	Queste	persone […]	sto	guardare	questo	bambino
		these	persons	stay:1SG	look:INF	this	child
		queste	persone		guardano	il	bambino	(target form)
		these	persons		look:3PL	the	child
		‘These persons look at this child.’

The IC with stare does not occur elsewhere at this stage and this is not surprising, because, as already seen in Section 3 (Table 1), the target progressive construction emerges at a more advanced stage of interlanguage and is hardly attested in my data, only appearing in Session 5 (see 6.4).

As for the subordinating markers, non-target per + infinitive persists in BD and emerges in YS. Come is found in MT. No examples can be found in Group 3.

6.4. Session 5

Table 7. ICs in learners’ interlanguage (Session 5).

	Group 1			Group 2			Group 3
	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic	Prebasic	Basic	Postbasic
Analytical ICs	-	Be-C Go-C	Do-C Be-C Stay-C	-	-	Do-C Be-C Stay-C	-	-	Be-C
Subordinating ICs	-	-	Per-C	-	-	Per-C Come-C	-	-	-

shows the distribution of ICs in the last session. ICs appear to be strongly rooted in the post-Basic Varieties of Groups 1 and 2.

The fare-construction persists in the learners of the Groups 1 and 2, while only one occurrence of the essere-construction can be found in a learner of the Group 3, MJ, as in (26):

(26)	MJ5	Io	sono	paura
		I	be:1SG	fear
		io	ho	paura	(target form)
		I	have:1SG	fear
		‘I’m scared.’

Groups 1 and 2 also exhibit frequent instances of the andare- and stare-constructions, the latter testifying a development of the overall interlanguage stage as observed in 6.3.

At this stage, no instance of non-target subordinating markers can be found in the interlanguages of the literate learners in Group 3.

7. Discussion and Conclusions

Data in Table 3, Table 4, Table 5 and Table 6 are summed up in

Table 8. Diachrony of ICs in learners’ productions (Sessions 1 to 5).

	Group 1			Group 2			Group 3
SESSIONS 1–2
	prebasic	basic	postbasic	prebasic	basic	postbasic	prebasic	basic	postbasic
Analytical ICs	-	Do-C	Do-C Be-C	-	-	Do-C Be-C	-	-	-
Subordinating ICs	-	-	-	-	-	-	-	-	-
SESSION 3
	prebasic	basic	postbasic	prebasic	basic	postbasic	prebasic	basic	postbasic
Analytical ICs	-	-	Be-C	-	-	Do-C	-	-	-
Subordinating ICs	-	-	Per-C Come-C	-	-	Per-C	-	-	Come-C
SESSION 4
	prebasic	basic	postbasic	prebasic	basic	postbasic	prebasic	basic	postbasic
Analytical ICs	-	-	Do-C Be-C Stay-C	-	-	Do-C Be-C	-	-	Do-C Go-C
Subordinating ICs	-	-	Per-C Come-C	-	-	Per-C Come-C	-	-	-
SESSION 5
	prebasic	basic	postbasic	prebasic	basic	postbasic	prebasic	basic	postbasic
Analytical ICs	-	Be-C Go-C	Do-C Be-C Stay-C	-	-	Do-C Be-C Stay-C	-	-	Be-C
Subordinating ICs	-	-	Per-C	-	-	Per-C Come-C	-	-	-

, which accounts for the diachrony of ICs in the interlanguages of learners who use them.

As Table 8 shows, the presence of ICs is strongly related to the transition to the post-Basic Variety, as ICs imply that learners are already able to:

• identify in the input functional elements that convey (or are supposed to convey) grammatical meanings;
• understand which segments of the utterances they produce do need morphosyntactic information, that is, that the verb must be associated with grammatical meaning and that inter-clausal linking must be syntactically marked.

Table 8 also shows that especially non-literate (Group 1) and low-literate (Group 2) learners select ICs to cover these functions, while literate learners (Group 3) use these constructions in a more sporadic and transient way. In LESLLA learners, ICs are not only more frequent (both as construction types and as individual tokens), but also more stable at the diachronic level (in the micro-diachrony of interlanguage development).

This persistence is surely related to learners’ slow morphosyntactic development. Although my data do not allow us to observe more advanced stages of interlanguage, previous studies have shown that as soon as morphosyntax develops, the use of ICs declines (see Section 3). As a corollary, if morphosyntax does not develop enough (for whatever reason), ICs are expected to persist longer. This is surely the case of the interlanguages employing ICs in my data.

However, literate learners do not exhibit more advanced stages of interlanguages than low-/non-literate learners and the development of morphosyntax and inflection is incomplete in both groups. A slow development of morphosyntax is more general and related to the low quantity and quality of exposure to the target language, which affects all the groups of learners. There should be different reasons if low-/non-literate learners show a stronger preference for lexical-syntactic patterns as ICs.

Interesting hypotheses derive from studies conducted under the Organic Grammar approach elaborated by Vainikka and Young-Scholten in several works since the 1990s, which are briefly described in Section 2. This approach is based on different theoretical premises from those adopted here (the one being a formalist perspective, the other one a functionalist one); however, the two perspective overwhelmingly converge in the results of the analysis of LESLLA interlanguages. This convergence is expressed in the preference for a mainly linguistic interpretation of the phenomena under examination.

As already mentioned in Session 2, the Organic Grammar approach (Vainikka and Young-Scholten 1998) convincingly suggested that the lack of phonological attainment in the second language can be a major cause of learners’ incomplete acquisition of morphosyntax; lack of phonological attainment, in turn, could be due to external causes, such as exposure. Vainikka et al. (2017) have analysed phenomena of overgeneralisation of function words and sequences (which they analyse as placeholders) to mark morphosyntactic functions in L2 English. They demonstrated that both literate and non-literate are able to individuate functional forms in the input and to work on their own projection of verbal syntax. Non-literate learners, however, tend to exhibit specific sub-patterns compared to literates; in particular, when they are in the process of acquiring L2 English, they tend to select forms not directly related to the actual verbal heads they have to fill (e.g., in the instead of a verbal element for progressive aspect). According to Vainikka et al. (2017, p. 248), this is due to “a greater reliance on auditory as compared to visual memory”. As they cannot access (or only have a very limited access to) written texts, low-/non-literate learners’ exposure to the target language is even poorer than that of the educated migrants.

This reasoning can help to explain my data. Low-/non-literate learners, who only rely on aural stimuli, exhibit a stronger preference for lexical-syntactic strategies to build the L2 grammar. These strategies involve the selection of forms, which can be more easily recognised in the input than bound morphemes, first because they are more salient on the phonological level (they are at least a phonological foot, in terms of Vanikka and Young-Scholten 1998). In addition, these forms are more transparent on the functional-semantic level. In fact, functional elements involved in ICs in L2 Italian are auxiliaries and light verbs (which also have a lexical counterpart, e.g., essere ‘existence’, fare ‘to do, to make’) and prepositions (which also express concrete, spatial meanings, e.g., per i campi ‘through the fields’). In other words, they are not grammatical forms tout court, but grammaticalising forms. Learners grasp their grammatical potential and build a part of their morphosyntax on these lexical-syntactic means. This process entails reinterpretation of the grammaticalising forms in their context of occurrence and generalisation to new contexts, as in other canonical process of grammaticalisation, as described by Hopper and Traugott (2003) and Heine et al. (1991), among other theorists writing about language change. As first suggested by Dittmar (1992) and Giacalone Ramat (1992), the micro-diachrony of acquisition is a path towards grammar, characterised by strategies and mechanisms to construct the languages that are universally documented. Literate and non-literate learners may choose contextually different strategies, as ICs in low-/non-literate learners suggest, although this does not alter the overall route they follow in developing grammar.