Language Awareness in Italian Secondary Schools How Do Students Approach Language Awareness Tasks and What Kind of Language Awareness Emerges from their Reflections?

The present paper aims to explore how students of Italian secondary schools approach linguistic tasks and what sort of knowledge about language emerges from their analyses. These questions arise from the observation of the results of the Invalsi tests. In order to examine how students construct their answers to a sample of language awareness questions, they were asked to explain how to answer them by making a tutorial video. The videos were discussed during a semi-structured interview. The data were transcribed and analysed by means of Qualitative Content Analysis. The results indicate that the principal difference between students relates to the stability of their “engagement with language” and their ability to switch their attention between different levels of linguistic analysis.


Introduction
The present study focuses on language awareness in L1 Italian of upper-secondary school students, as part of the skills students are expected to acquire according to the National Guidelines developed by the Italian Ministry of Education. 1 Given its importance in students' curriculum, language awareness is assessed within the Italian test administered by INVALSI, 2 the research institute responsible for the external assessment of learning outcomes at various levels in the Italian school system. The present study focuses on the Italian tests administered in the second class of upper-secondary schools, corresponding to the 10th year of schooling, in the period between 2011 and 2017. At this time the tests were administered in paper-and-pencil format, and published on the institute's website, together with the related results. The tests administered from the year 2018 are not available for analysis because their administration is computerbased, and the tests are no longer published. The Italian tests that form the subject of the present study were composed of two parts: a reading comprehension section and a language awareness section. The results of the language awareness questions reveal significant variations in the achievements of students from different school types (general upper secondary education, technical and vocational education). When looking at the percentage of correct answers, for instance, there is always a gap between general secondary schools on one side, and vocational schools on the other, which suggests that students attending a vocational school are disadvantaged when it comes to reflection on language.
The dimension of the gap between the two school types is inconstant; for the questions examined in the present study it varies between 14% and 45%. However, this variability cannot be directly linked to the overall difficulty of the questions at the national level, nor to the linguistic phenomena on which they focus (Toth 2019). Among two questions targeting the same grammatical phenomenon, for instance, one may generate a large gap between school types, the other a moderate one. The same variability is observable when looking at the difficulty of these questions at the national level.
The absence of systematic patterns suggests that students' language awareness should be viewed from a complex system perspec-tive (Larsen-Freeman 2013), and the test outcomes should be considered as results emerging from the interaction of multiple factors (Svalberg 2016a, 4). An attempt to take a closer look at how this complex system works is motivated by the relevance attributed to language awareness in both the ministerial guidelines and the academic literature (e.g. Bialystok 2001;Lo Duca 2004, 2018aSvalberg 2015Svalberg , 2016aSvalberg , 2016b. The present study aims therefore to deepen our understanding of how students approach language awareness tasks and elaborate their answers, in order to gain a better understanding of which factors influence the outcome of their reflections and how these factors work together.

2
Language Awareness in the Italian Curriculum As pointed out by Balboni (1993), the Italian education system is centralized, in the sense that curricular guidelines are prepared by the Ministry of Education and are the basis of the teaching programmes implemented by schools. The present study focuses on upper-secondary education, which lasts five years and is organized into a twoyear period and a three-year period. Even though in upper-secondary education curricular guidelines are differentiated according to the school type, the ministry also defines knowledge and competencies all students are expected to possess at the end of compulsory education, which coincides with the end of the two-year period. The ability to reflect on language is part of these competences, and its conceptualization in the ministerial documents is congruent with the definition of language awareness (LA) as "explicit knowledge about language, and conscious perception and sensitivity in language learning, language teaching and language use" (Association for Language Awareness, https://www.languageawareness.org/?page_id=48). However, as pointed out by several scholars (Cots 2017;Jessner 2017), the definition and conceptualization of LA varies across disciplines, and researchers use the terms 'knowledge about language' (KAL) and 'metalinguistic knowledge' (MLK) when referring to concepts similar to LA. In the present study, the terms LA and KAL are used as synonyms, while MLK is seen as a "more specific conscious knowledge of the formal aspects of language" (Cenoz, Gorter, May 2017, ix). As pointed out by Lo Duca (2004, 2012a, 2012b), the majority of Italian teachers seem to follow a traditional approach to language, conceiving knowledge about language as the memorization of a set of rules and labels rather than the observation and analysis of linguistic data. As Van Rijt et al. (2019, 79) note, the same tendency has been observed in a significant number of studies conducted across the globe. However, a traditional approach to language often leads to the memorization of partial definitions and the acquisition of frag-EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 370 mentary knowledge, as exemplified by Lo Duca (2011) on the classification of words into parts of speech. Such classification is traditionally based on semantic definitions, by treating words in isolation, rather than examining their function in a sentence and their relationship with the context. Consequently, students tend to classify words according to semantic criteria, and do not pay attention to morphosyntactic features, encountering difficulties with words that do not correspond to a semantic definition of the part of speech they belong to (Lo Duca, Ferronato, Mengardo 2009;Lo Duca, Polato 2010;Lo Duca, Cristinelli, Martinelli 2011).
Thus, a traditional approach to language seems to discourage analytical reflection and may lead to a mechanical application of labels and rules of thumb to linguistic data, as indicated by studies on L1 Italian and L1 Dutch (Lo Duca 2018a; Van Rijt et al. 2019), as well as by studies on the results of the INVALSI test on language awareness (e.g., Lo Duca 2018b; Toth 2019).

Collaborative Construction of Language Awareness
In addition to examining how students approach linguistic data and tasks, several studies (e.g., Galloway, Stude, Uccelli 2015;Lo Duca 2004, 2012a, 2012b, 2018a, 2018bMyhill, Jones, Wilson 2016;Svalberg 2015;Toth 2019;Watson, Newman 2017) point towards the importance of discussion based on the observation of linguistic data in developing the language awareness of students. Discussion enhances metalinguistic understanding by directing the student's attention to relations between form, meaning and function (Myhill, Jones, Wilson 2016, 37), and exploiting the student's implicit competences and natural curiosity towards language (Lo Duca 2004, 2018a. These results support a "language awareness approach", as described by Svalberg (2016b), focusing on the collaborative construction of knowledge instead of its transmission, emphasizing the learner's active participation.
In Svalberg's (2009Svalberg's ( , 2016a) studies, the collaborative construction of language awareness is called 'engagement with language' (EWL), and described as a complex and cyclical process, in which the learner draws on their existing language awareness (LA) to reflect on linguistic data and analyse how language works, resulting in new or enhanced LA (Svalberg 2009, 248). This process has not only a cognitive, but also an affective and a social dimension; in addition to focused attention directed to the language, it requires a positive and purposeful attitude towards the language, and a willingness to engage in collaborative knowledge construction (Svalberg 2009(Svalberg , 2016a. The three dimensions are not easily distinguishable, given that they interact and influence each other, forming a complex system (Larsen-Freeman, Cameron 2008).

371
A pedagogical implication of this conceptualization is that language teaching should be designed in a manner that fosters highquality EWL, in order to ensure the creation of new or enhanced LA. According to Svalberg's study, cognitive conflict, i.e. when the learner's knowledge is "in some way insufficient to analyse the language to the group's, or the individual's satisfaction" (Svalberg 2015, 541), is among the factors that increase the learner's engagement, by initiating focused attention, and creating occasions for collaborative reflection on language and knowledge construction. The concept of cognitive conflict, and its role in fostering EWL, seems to be related to the concept of noticing in second language acquisition (Schmidt 2001;Mackey 2006), referring to episodes that trigger the learner's attention on a "gap between what they produce/know and what is produced by the speakers of the L2" (Mackey 2006, 408). Both concepts focus on the importance of triggering selective attention, directed to specific aspects of language, as a prerequisite to knowledge construction through social interaction.
To sum up, the studies reviewed in the present section point towards a conception of language awareness instruction as knowledge construction rather than knowledge transmission and emphasize the role of the learner's agency and focused attention. This approach promotes the exploitation of the learner's existing LA and curiosity towards language, by guiding them during the creation of new or enhanced LA.

Language Awareness in the INVALSI Tests
The questions focusing on knowledge about language in the INVAL-SI tests are formulated with the aim of stimulating an explorative approach to language, reflection on linguistic data and analysis of form-meaning connections (INVALSI 2018). Consistent with the principles formulated by Lo Duca (2004Duca ( , 2018a and the language awareness approach (Svalberg 2016b), they require the student to exploit both their explicit knowledge about language and their implicit linguistic competence.
As pointed out by several scholars (Jessner 2017;Ellis 2017), there are different conceptualizations of explicit and implicit competences, frequently based on a distinction between procedural and declarative knowledge. While earlier research questioned the possibility of an interface between explicit and implicit competence (e.g., Paradis 2009), recent research (Ellis 2017;Rebuschat 2015) has come to the conclusion that the two knowledge systems influence each other. This idea is consistent with Bialystok's (2001) conceptualization of metalinguistic processing as a gradable concept, based on the idea that there is "no absolute threshold beyond which processing can be claimed to be 'metalinguistic'" (Bialystok 2001, 130 Consistent with the idea that explicit and implicit knowledge can mutually influence each other (Ellis 2017, 118), and that language awareness is "partly conscious and partly intuitive" (Svalberg 2016b, 399), the linguistic tasks in the INVALSI tests aim to motivate students to exploit both their implicit and explicit knowledge. Tasks implying less explicit analysis require students to understand logical-semantic relations between clauses forming a sentence, or morphosyntactic relations between sentence components, without having to make this understanding explicit. Their answer may be intuitive, based on their linguistic sensitivity. For instance, students may be asked to substitute a conjunction without changing the overall meaning of a sentence, or to replace a synthetic relative pronoun with an analytical form, respecting the concordance in gender and number with the antecedent. Other tasks require a more explicit analysis and a reasoning in abstract terms, by taking into account both morphosyntactic and semantic features. For instance, in order to distinguish between the passive and active voice, especially in non-prototypical sentences, students need to direct their attention to the syntactic structure of the sentences and the semantic role of the constituents, and be familiar with abstract concepts, such as transitivity. This type of task requires a higher level of abstraction than tasks focusing on manipulation of linguistic data.

Study Objectives
The present study aims to examine how secondary school students reflect on language when solving language awareness tasks administered within the INVALSI tests. Thus, the over-arching research question that guided the present study is "How do students from different school types approach language awareness tasks from the INVALSI tests?". Consistent with the view advocated by the language awareness movement (Svalberg 2016b), the present study assumes that knowledge about language is primarily observable in interactions. The study is therefore based on data collected from instances of group work and focus group interviews, analysed with the aims of examining what kind of language awareness emerges from these interactions. In addition, the study aims to examine to what extent are the kind of LA emerging from students' interactions and the quality of engagement with language shown by the students influenced by the type of secondary school they are attending. Thus, the main research question can be broken divided into three subordinate questions: 1. How do students engage with language during a collaborative task and a focus group interview? 2. What kind of language awareness emerges from students' reflection on language?

373
3. To what extent does students' engagement with language and language awareness differ in relation to school type?

Data Collection Methods
The data collection was directed to elicit reflection on linguistic tasks in order to examine students' language awareness and engagement with language. Work with peers was initiated in order to obtain data that were as least likely as possible to be influenced by the presence of the researcher, by asking selected groups of students to make a tutorial video in which they explain how to answer a series of eight INVALSI questions representing various levels of explicitness and difficulty. The data collection took place in two classes, each of approximately 20 students, one in a general upper secondary school and one in a vocational upper secondary school. Each class of students was divided into four groups, by asking their teacher to create groups that should be able to collaborate well together. While students were working on their tutorials, both the teacher and the researcher were in the class and made themselves available for technical support and clarifications but did not intervene in the students' work. The researcher was not part of the school staff and has never met the students before the day of data collection.
It was assumed that, contrary to the interviews, which were produced jointly by the interviewer and the interviewee (Talmy 2010, 141-2), group work data would mirror how participants engage with language when working with peers. For the same reason, the guidance given to the students was as generic as possible. They were asked to imagine making a tutorial video for their friends who have to take the INVALSI tests, indicate the correct answer for each question, explain why the given answer is correct and, in the case of multiple-choice tasks, explain why the other options should be excluded. In addition, the students were requested to make sure that each question was commented on by at least three students and that each member of the groups had the possibility to comment on at least two or three questions. The latter request aimed at eliciting peer interaction on the videos, without explicitly asking students to film the whole working process, which could have biased their conversations.
When the students completed the tutorials, the videos were discussed within focus group interviews, moderated by the researcher, which aimed to deepen the analyses proposed on the videos, by pushing students to the limits of their ability to reflect on language (Watson, Newman 2017, 392). Since interviews present a case when meaning is co-constructed by the researcher and the interviewee, they were analysed separately from the video data.

Data Analytical Methods
Consistent with the research questions, the analysis of the data aimed to extract information relating to the students' engagement with language and their language awareness. The data were analysed by means of qualitative content analysis, following the principles described by Mayring (2014). Since "content analysis requires a written text as a basis" (Mayring 2014, 44), both the interview data and the videos were transcribed and analysed as textual data. The transcription followed the conventions formulated for the corpus VOICE (VOICE Project 2007), with some adjustments related to the specificities of the Italian language.
The data coding followed a deductive-structuring approach (Mayring 2014, 95-102), classifying segments of interactions according to the set of codes reported in Appendix, developed in a pilot study, in deductive and inductive cycles (Toth 2019). Thus, the system of categorization presented here is based on previous research on language awareness and engagement with language, described in the theoretical introduction to the present study, integrated with categories generated from the data, collected and analysed during the pilot study.
In the present study, a total number of 508 data segments were coded: 141 for socio-affective engagement and 367 for cognitive engagement. The whole dataset was coded by the researcher, while 30% of the data were coded simultaneously by a research fellow, who was working with a more detailed version of the code system (Appendix), with a definition and an anchor example provided for each code. The material coded by the two researchers was compared, and differences were discussed until agreement on code-assignment was reached. Subsequently, the two researchers recoded the data independently and the two versions of the coding were compared for a second time.
The second comparison showed that the codes assigned by the two researchers were consistent, with 96% intercoder agreement. The rest of the data was subsequently coded by the researcher.

4
The Findings The description of the findings focuses on the students' engagement with language and the language awareness emerging from their reasonings, aiming to identify to what extent students from general upper secondary schools differ from those attending vocational secondary schools.
In the present study, the social and affective dimensions were condensed into a single category, to code sections containing explicit indications of particularly low or high socio-affective engagement. The latter was evidenced from an increased number of turn-takings, student's asking each other questions and reacting to the ideas of their peers, collaborative hypothesizing, initiation of ideas, etc. In contrast, low socio-affective engagement was deduced from discouraging comments, lack of reaction to peers' ideas, lack of supporting peers in their reasoning and explicit comments expressing boredom or frustration. Segments labelled as cognitive engagement showed students talking about language, noticing a linguistic feature, completing peers' utterances, justifying an argument, etc. (Baralt, Gurzynski-Weiss, Kim 2016; Philp, Duchesne 2016).

Socio-Affective Engagement
One of the most evident features emerging from the transcripts of the tutorial videos was the students' scant adherence to the guidelines suggested by the researcher. Instead of working together, most of the groups assigned one or two questions to each group member and students completed the task on their own. Only two groups, one in the vocational school and another in the general school, endeavoured to solve the tasks and comment on the questions together.
In spite of the scarcity of interaction data, some information about students' EWL evidenced from the video material points towards differences in their socio-affective engagement in relation to school types. Students in the general education school answered all the questions they were assigned and registered their answers on a video. In contrast, students in the vocational school skipped several questions. Thus, the total number of tasks commented on the videos was 32 in the general education school, 16 in the vocational school. Apart from the quantity, there were significant differences in the quality of the videos made in the two school types.
Videos made by the students of the general education school show that even though they preferred working autonomously, the other group members kept focusing their attention on the task. This is evidenced from interventions following small mistakes such as misspelling of a word or an imprecise answer, where students correct EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 376 or complete their peers' utterances. In addition, only a very limited background noise was observable in the videos made in this school, indicating that the whole class was working on the task; students used a quiet voice and did not disturb each other.
This contrasts with the videos made in the professional school, which are characterized by a constant background noise made by students who refused to collaborate. On the videos, these students are seen walking around the class, disturbing their peers and sometimes interrupting their work. In addition, low socio-affective engagement is evident from some really hasty reasonings recorded on the videos, where students justify their answer by saying that they felt like it was correct, as observable in Excerpt 1. The ironic comment of the second student suggests that they realize the incompleteness of their answer, but they do nothing to complete it, just rush to the next task. To sum up, the most evident difference between students from the general education school and the professional school emerging from the tutorials appears to be the stability of their socio-affective engagement. While students from general education school maintain a stable engagement throughout the work, students from the vocational school fluctuate between low and high engagement. The analysis of students' socio-affective engagement during the interviews confirms the same pattern that emerged from the tutorial videos. While the general education school students maintain a fair level of engagement throughout the interviews, the vocational school students sometimes interrupt the discussion with impertinent conversation. The focus on the task can be maintained only if the interviewer keeps asking questions and guiding their attention.

Cognitive Engagement and Knowledge about Language
The present section examines the episodes when students display cognitive engagement, in the sense that they talk about the language as object, notice linguistic features, reflect on linguistic data and develop hypotheses (Baralt, Gurzynski-Weiss, Kim 2016, 222). The code-system used for the classification of these episodes [tab. 1] is

Zuzana Toth Language Awareness in Italian Secondary Schools
EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 377 composed of ten codes, referring to the way students' reasonings are structured. These codes can be grouped into two broad categories: codes 1-5 refer to cases when students' reasoning is incomplete, while codes 5-10 describe more elaborate reasonings. 367 data segments were coded; Table 1 shows how each code was distributed across school types and data types. The code lack of analysis refers to cases when students do not give any explanation for their answer and suggest, or explicitly state, that their choice was random; on the other hand, if students limited themselves to indicating their answer without giving an explanation for it, but neither stating nor suggesting that their answer was random, the segment was coded as analysis not made explicit. Episodes coded as attempt to analyse show that students started an explanation, but they could not complete it, possibly because they lost their train of thoughts or were not able to verbalize their intuition. Unclear focus refers to reasonings where students' attention seems to fluctuate between different linguistic features and disconnected information, without developing a coherent line of argument, while lack of explicit knowledge was attributed to episodes where students revealed an erratic assumption or misconception about the language. The codes focus on morphosyntactic features and focus on meaning were used to classify more elaborate answers, by indicating which level of linguistic analysis they focused on: meaning and semantics or morphosyntax. In other cases, students' reasoning was based on a reference to a metalinguistic knowledge or definition (code 8), or on the manipulation and observation of linguistic data (code 9). Finally, EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 378 the code noticing was applied to episodes where students acknowledged the incompleteness of their reasoning. As a consequence, they often directed their attention to linguistic features that were relevant for the analysis but had been neglected until then.

Incomplete Reflection on Language
The frequency of codes describing incomplete reasonings shows major differences between the students from the two school types. On the video tutorials, students from the general education school seem more likely to indicate the answers to the tasks without explaining the reason behind their choices (N=17). This tendency is exemplified in Excerpt 2, referring to a question where the students have to observe various occurrences of a polyfunctional word in Italian (lo) and indicate whether the word has the function of a determinate article or a pronoun.

Excerpt 2
Student 2: qualcuno lo potrebbe definire è pronome somebody may define it [it is] a pronoun Student 4: sì yes Student 3: lo smartphone articolo the smartphone article Student 4: ok (1) pronome articolo {indica le parole nel testo con la mano} ok (1) pronoun article {indicates the words on the paper} Student 3: lo superano <1> pronome </1> [they] overtake it <1> pronoun </1> Student 1: <1> pronome </1> <1> pronoun </1> As Excerpt 2 shows, students' answers are limited to indicating the morphosyntactic function of the word lo in the various contexts, without making any reference to the criteria they used to analyse the sentences. This tendency, however, is almost absent in the interviews made with the same students, during which students demonstrate their ability to identify the linguistic features they need to focus on in order to justify their choices. The interviewer's questions are answered extensively, and no prompting is necessary to develop a complete reasoning. Table 1 might suggest that vocational school students are less likely to omit explicit analyses on their tutorial videos than general education school students. However, the lower frequency of this code should be interpreted by taking into account the discrepancy EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 379 in the number of tutorial videos made in the two school types; 32 in the general education school and 16 in the professional school. Thus, the lower frequency of the code is related to the lower quantity of videos. In fact, the frequency of the code increases significantly in the interviews made in the vocational school, showing that, when confronted with questions targeting grammatical phenomena, students often avoid explicit analyses (N=45). In some cases, they make unsuccessful attempts to analyse the data (N=22), their reasoning does not have a clear focus (N=13) or they formulate an answer that demonstrates lack of explicit knowledge (N=10). In Excerpt 3, for instance, Student 1 cannot produce an answer to a question requiring the identification of a nominal predicate. When the interviewer explicitly addresses the difference between nominal and verbal predicates, the student makes reference to semantic criteria used to distinguish between active and passive voices, rather than referring to the characteristics of predicative and copulative verbs. and what is a verbal predicate? Student 2: è un verbo it is a verb Student 1: un verbo che compie l'azione a verb that undertakes an action To sum up, the distribution of incomplete answers suggests that the reason behind them may be different in the two school types. General education school students give some hasty answers on their video tutorials; however, when asked to provide further explanation, they show good analytical skills. On the other hand, vocational school students seem to be struggling when trying to analyse linguistic data, and the interviewer's scaffolding is necessary to direct their attention to relevant linguistic features and elicit more in-depth reflections on language.

In-Depth Reflection on Language
The most substantial differences between the two groups of students emerge when analysing episodes showing in-depth reasonings. As can be observed in Table 1, students from the vocational school nearly never refer to morphosyntactic features in their tutorials. They reflect on morphosyntax during the interviews, but only when their attention is guided by the interviewer's questions (N=16). In contrast, students from the general education school carry out morphosyntactic analyses when working with peers as well as during the interviews. Thus, the code focus on morphosyntax occurs 51 (22+29) times in the data from the general education school, while only 17 (1+16) times in the data from the vocational school. When vocational school students are confronted with a task that requires a focus on morphosyntax, they often limit themselves to giving an answer without explaining the reason for their choice or, in some cases, they explicitly state that they were guessing. When trying to make an explicit analysis, they show a strong tendency to focus on meaning and semantics (N=36 [9+27]) rather than morphosyntax, and sometimes demonstrate lack of explicit knowledge.
Excerpt 4, for instance, refers to a question focusing on the passive voice, and requires students to identify the sentence where the particle si, a polyfunctional word in Italian, is part of a passive construction. In the distractors there are sentences in which si occurs as part of an impersonal construction with an intransitive verb and as a reflexive pronoun. In their tutorial, lasting a few seconds, vocational school students indicated the option C [In questa trattoria si mangia benissimo / In this restaurant one can eat very well], where "si" is part of an impersonal construction, without giving any further explanation. Therefore, during the interview, students were requested to transform the sentence which they claim to be passive into the active voice. As exemplified in Excerpt 4, their attempts to transform the sentence focus on the level of semantics, by changing the word order and transforming the impersonal construction into a personal clause with an implied subject. Their reflections lack any reference to morphosyntactic features, such as the concept of transitivity and the syntactic roles of subject and direct object, essential to considering active and passive voices. Excerpt 4 suggests that if the students had considered morphosyntactic criteria, they would have noticed that the sentence cannot be transformed into the passive voice, because in this case the verb is used with an intransitive meaning. However, Excerpt 5 shows that the interviewer's scaffolding is necessary to guide the students' attention to the concept of transitivity, by asking them to analyse a more prototypical intransitive sentence (Anna cucina bene / Anna cooks well). Thanks to this example, the students notice that both verbs (mangiare [to eat] and cucinare [to cook]) can be used with both transitive and intransitive meaning, and finally conclude that sentence C, where mangiare has an intransitive value, cannot be passive. As exemplified in Excerpt 5, students refer to the concepts related to transitivity by means of periphrastic descriptions and examples, instead of using a grammatical terminology. For instance, they explain that the verb cucinare does not have a direct object by claiming that the sentence does not say what is being cooked. This episode suggests that students are able to notice morphosyntactic features, even though they do not possess the metalanguage necessary to label them. To sum up, Excerpt 5 demonstrates how guided discussion can foster students' metalinguistic understanding (Watson, Newman 2017, 392), by directing their attention on relevant linguistic features and enabling them to identify differences and similarities among linguistic data. As suggested by the concepts of noticing (Mackey 2006) and cognitive conflict (Svalberg 2015), perceiving a gap in their analysis and noticing a new feature seems to trigger students' focused attention and enhance their engagement with language. In fact, the tutorial video made by the students on this task lasted 12 seconds, while the reasoning guided by the interviewer lasted 6 minutes and 26 seconds, and consisted of 83 turn takings, where students made purposeful comments and drew on each other's ideas. Students from the general education school, on the other hand, show a good ability to switch their attention between the level of morphosyntax and the level of semantics, and to direct their attention to the linguistic features relevant for the analysis. Their answers tend to be based on multiple criteria, as exemplified by Excerpt 6, where students reflect on the function of the conjunction se, which EL.LE e-ISSN 2280-6792 9(3), 2020, 367-388 383 in Italian introduces both indirect interrogative clauses and hypothetical clauses.

Excerpt 6
Student 1: invece la seconda frase <legge ad alta voce> se mi chiedessero la strada per il Duomo non saprei rispondere </legge ad alta voce> abbiamo detto che è un periodo ipotetico perché il SE si può in ehm sostituire con nel caso in cui e: anche: il fatto che ci sia il congiuntivo imperfetto e il condizionale ci: fa capire che è un periodo ipotetico by contrast the second sentence <reading aloud> if they asked me the way to the Cathedral, I would not be able to answer </reading aloud> we said that it is a hypothetical period because se may be replaced by in the case and: also: the fact that there is an imperfect subjunctive and a conditional it makes us clear that it is a hypothetical period Student 2: anche: la terza frase [Vogliono partire oggi, ma se non si sbrigano…] abbiamo deciso di mettere un periodo ipotetico siccome (.) ehm (.) la: l'apodosi è sottintesa (.) vogliono partire oggi? ma (.) se non si sbrigano puntini puntini quei puntini puntini indicano l'apodosi (1) also the third sentence [They would like to depart, but if they do not hurry…] we decided to say hypothetical period because (.) ehm (.) the apodosis is implied (.) they want to depart today? but (.) if they do not hurry three dots those three dots indicate the apodosis As can be observed in Excerpt 6, students take into consideration both the level of semantics, i.e. the meaning of the conjunction in the given context, and the level of morphosyntax, i.e. the verbal morphology and the syntactic structure of the sentence. In addition, they also try to manipulate one of the sentences, by replacing the word se with a different hypothetical conjunction. Their explanations are concise but extensive, suggesting a multi-layered analysis of linguistic data, and the ability to accurately select the relevant criteria to focus on. To sum up, the most marked differences between the students from the two school types, emerging from episodes of in-depth reflections on language, relate to their ability to select the linguistic features relevant for the analysis, and the ability to focus on morphosyntactic features. Professional school students show a strong tendency to focus on meaning; they avoid manipulation of linguistic data and never refer to morphosyntactic features during the group work. On the other hand, general education school students are able to select the relevant linguistic features on the basis of the task and alternate their attention between various levels of linguistic analysis.

Discussion
The data presented in this study provide insights into students' engagement with language during the work on language awareness tasks, and the kind of knowledge about language that emerges from their reasonings. The results corroborate the interrelatedness of the three dimensions of engagement with language, as emphasized in Svalberg's studies (e.g., Svalberg 2009Svalberg , 2015Svalberg , 2016a, and point towards the existence of school type-related differences in the quality of students' EWL and their LA. Consistent with the observations of Baralt, Gurzynski-Weiss and Kim (2016, 234), the data show that low socio-affective engagement may reduce cognitive engagement and hamper full exploitation of students' existing LA. This phenomenon is particularly observable in episodes displaying lack of analysis, when students give random answers to the questions, without careful examination of the linguistic data presented in the task (Excerpt 1). However, an increase in one of the three dimensions of engagement may positively affect the others, as can be observed during the interviews, especially when the students' attention is guided to incongruences in their analyses and/ or linguistic features which escaped their attention. These episodes are often accompanied by increased cognitive and socio-affective engagement, deducible from the purposefulness of students' comments and collaborative knowledge construction, as observed in Excerpt 5. In cases when students do not possess the metalinguistic terminology necessary to label the linguistic forms and structures noticed during the interview, they use colloquial terms or give examples to illustrate the linguistic phenomenon they noticed.
The interviews demonstrate the students' ability to notice new linguistic features in linguistic data, especially within the domain of morphosyntax. A didactic implication of this finding is to exploit this ability when introducing a new metalinguistic concept, by adopting a guidedinductive methodology. Metalinguistic terminology may be introduced when students perceive the need to label the patterns they found in the data. The role of guided discussion in the development and enhancement of LA emerging from the present data has also been observed in studies conducted on L1 Italian (Lo Duca 2004; 2018a) and L1 English (e.g., Galloway, Stude, Uccelli 2015;Watson, Newman 2017).
One of the most evident differences between students from the two school types appears to be the stability of their engagement. Students from the general education school maintain a fair level of engagement during both group work and the interview, while students from the vocational school tend to demonstrate low engagement during the work with their peers and are more likely to fluctuate between low and high engagement during the interview.
With respect to the kind of language awareness, a significant dif-