Academic words constitute 8% - 10% of academic text vocabulary and thus are crucial for comprehending expository texts (Nation, 2001; Hyland & Tse, 2007) . The concept of academic vocabulary remains controversial, where multiple and inconsistent definitions have been proposed. Here, academic vocabulary refers to a general cluster of words that are used across different content areas (e.g. History, science, technology), are necessary for learning and are difficult to acquire (Townsend, 2009) .

Previous work indicates that students’ poor academic vocabulary knowledge affects their academic gains (Cohen, Glasman, Rosenbaum-Cohen, Ferrara, & Fine, 1988; Coxhead, 2000; Nation, 2001) . Consequently, several attempts have been made to establish academic word lists that serve to develop appropriate intervention programs for fostering academic vocabulary, mostly among English learners (e.g. Campion & Elley, 1971 ; Coxhead & Nation, 2001 ; Gardner & Davies, 2014 ; Ghadessy, 1979 ; Lynn, 1973 ; Praninskas, 1972 ; Xue & Nation, 1984 ). However, the examination of academic language development among Arabic native speakers remains scarce and unestablished. Additionally, taking into account the specific linguistic features of the Arabic language and its impact on literacy development, a finer toned examination is advisory.

Several properties of the Arabic language seem to obstruct Arabic academic vocabulary acquisition, especially its diglossic nature (i.e. the linguistic gap between its spoken and written form; Ferguson (1959) ). There is a large body of evidence that points to the great impact of Arabic diglossic nature on language development, including oral language and later reading skills (e.g., Abu-Rabia, 2000 ; Saiegh-Haddad, 2017 ), accounting in part to the low attainments in reading literacy tests of Arabic native speaking pupils both on national and international scales (Zuzovsky, 2005, 2010; Abu-Asba, 2005; Karmarski & Mevarech, 2004; Mevarech & Karmarsky, 2007) . Moreover, the complex linguistic characteristics of Arabic morphological, semantic, syntactic and phonological structures may constitute yet another challenge for literacy acquisition in general and academic vocabulary development in specific (for a full review about Arabic linguistic properties see Saiegh-Haddad & Henkin-Roitfarb, 2014 ).

The Arab population in Israel consists 25% of the Israeli population, with 92% of the former living in an exclusively Arab inhibited towns and villages whereas the others live in an intermixed Jewish-Arab cities (Yonay, Yaish, & Kraus, 2015) . The Muslim community, including the Bedouins, constitutes about 83% of the Arab population in Israel, the Christians 9% and the Druze 8% (Central Bureau of Statics, 2016) . The Christian population is characterized by higher academic achievements, higher marriage mean age, lower birth rate and larger income (Khattab, 2002; Kraus & Yonay, 2000) . The Druze and the Muslim community, excluding the Bedouin population, are found to share equivalent socio-economic conditions (Yonay, Yaish, & Kraus, 2015) . In contrast, the Bedouin population is nomadic agrarian that live in recognized and unrecognized villages. Among the latter, poverty rate was nearly 80% in 2004. In general, when comparing the Socio-economic status (SES) of the Arab to the Jewish population in Israel, immense gaps are encountered, favouring the latter. In comparison to 14% of Jewish families, 47% of the Arab population in Israel live in poverty (National Insurance Institute of Israel, 2016) with the Bedouin population ranking on the bottom of the Israeli’s SES scale (Abu-Bader & Gottlieb, 2009; Rudnitzky, 2012) .

The socio-cultural heterogeneity of the Israeli population echoes also in the structure of the education system. Along the years, separate Arab and Jewish educational systems have been operated. The former is further divided into three sub-systems: general Arab (Muslims and Christians), Druze and Bedouin. To meet the language and cultural differences both between the Arab and Jewish population as well as those between the three Arab subgroups, different curriculums are applied. This separation in turn makes education in Israel highly susceptible to political agendas, ethno-religious Hierarchy and discrimination (Adalah, 2011; Friedlander, Okun, & Goldscheider, 2016) .

Socio-culture factors are found to impact academic achievements in general and literacy status in specific among the Israeli population, as seen on both national and international tests. For example, as indicated by PISA results (OECD, 2014) , 50% of Arabic speaking pupils are below the expected literacy proficiency level, where poorer performance has been encountered among the Arab sector in Israel when compared to the Jewish sector (Israeli Ministry of Education, 2015) . Within the Arab sector, higher literacy achievements are encountered among the Druze pupils when compared to the general Arab subgroup, across all disciplines, whereas the poorest performance was noted among the Bedouin subgroup (Knesset Research and Information Center, 2013; MJB, 2015; Edi-Rokah et al., 2011) . These observed literacy gaps seem to widen in middle school (Edi-Rokah et al., 2011) . Furthermore, according to the Israeli Ministry of Education 2013-2015 reports (MJB, 2015) , only 33% of Arab high school pupils were eligible for a matriculation certificate in comparison to 50% in the Jewish sector. Moreover, higher dropout and violence rates are observed in Bedouin junior highs, having only 20% matriculation eligibility rate.

Devising a valid measure for vocabulary assessment can be challenging as the construct of word knowledge remains controversy (Pearson, Hiebert, & Kamil, 2007) . One of the earliest comprehensive definitions of word knowledge can be traced to Richards (1976) , who proposed that knowing a word involves broad knowledge of the word semantics, manner of utilization, its morphological properties, the syntactical forms associated with it and knowledge of its semantic association with other words, as well as the limitation of its usage in different contexts. However, such definition remains descriptive and incomplete, as it fails neither to account for the developmental aspects of the different word knowledge levels nor for its underlying shared mechanisms (Schmitt & Meara, 1997) . In an attempt to shift away from descriptive frameworks, Meara (1996) suggested that word knowledge encompasses three main distinctive competencies, learners’ vocabulary knowledge: the size of the lexicon, the automaticity by which lexical items are accessed and the richness of the learners’ semantic network. Alternatively, Cockrum & Shanker (2012) suggested that the purpose of assessing pupils’ vocabulary is to assess the adequacy of their vocabulary knowledge to their grade level and knowledge of words that are required for immediate instruction. They proposed a five word knowledge levels: 1) Absolute lack of familiarity with a word (no recognition of a word); 2) Recognizing the word but not knowing its meaning; 3) Inferring the word meaning from context while having only limited understanding of its meaning; 4) Adequate understanding of the word that appears in context; 5) Familiarity with the multiple meanings of the word (if it exists) and ability to utilize it for communication purposes (speaking and writing) and in thinking.

Other distinction of vocabulary knowledge relates to the breadth and depth of vocabulary. The former refers to number of words that one has some familiarity with (Nation, 2001) and the latter refers to the level of one’s understanding to the different aspects of the word. Breadth of vocabulary knowledge is by itself a multi-componential construct that encompasses pronunciation, spelling, morphological and syntactic properties, registers, meaning and semantic associations and collocational properties (Richards, 1976; Qian, 1999) .

For the first time, the current study attempted to investigate academic vocabulary knowledge among middle school Arab native speakers, shedding light on yet another possible underlying factor in the experienced literacy challenges. To this end, we compared academic vocabulary knowledge of 7^th and 9^th grade (the start and end of middle school). In addition, due to the gaps in literacy achievements between the Arab subgroups, we also endeavored on examining differences in academic vocabulary knowledge between the three Arab subgroups. Accordingly, three academic vocabulary measures were developed, utilizing a variation of test formats such as self-check tests, requiring the test taker to indicate whether he is familiar with a presented words or not, cloze task and sentence production task. For building the assessment tests, a corpus-based Arabic academic list for middle school was developed. The list contains the 55 most frequent word-roots appearing in middle school textbooks across four subject areas: history, civics, technology, science, and geography. It is worthy to note that the rationale for clustering the words based on their roots is due to Arabic morphological structure. As other Semitic languages, Arabic words are derived from “roots” (\جذرjaðer\), mostly trilateral consonant roots that are inserted within a word pattern, either derivational or inflectional. As it conveys the core meaning of the word, Arabic dictionaries are typically organized alphabetically according to word roots (Saiegh-Haddad & Henkin-Roitfarb, 2014) .

The aims of the study are as follows:

1) Examining varied levels of academic vocabulary knowledge among Arabic native speaking pupils in 7^th and 9^th grades.

2) Investigating the differences in academic vocabulary knowledge between the Arab subgroups in Israel: General Arab, Druze and Bedouin subgroup.

The current study included 1197 middle school Arabic native speakers (7^th and 9^th graders) from ten different schools in Israel, representing the different Arab subgroups in Israel (i.e., General Arab subgroup, Druze subgroup and Bedouin subgroup, see Table 1 below). In 7^th grade, age mean score is 12.98 (SD = .36) and in 9^th grade 14.96 (SD = .35).

In addition, four Arabic junior-high teachers assisted in mapping the academic vocabulary that is necessary to cope with the materials in the various disciplines (history, geography, civics, science and technology). All the participating teachers had obtained their M.A. degree in Arabic language and had more than 10 years of teaching experience. Another 46 teachers, who teach various disciplines, participated in rating the necessity of the mapped words for successful academic performance in middle school (see Section 2.3.1 developing corpus-based academic vocabulary list procedure). The teachers who participated in the mapping processes were selected from schools that were not among those selected for the test administration.

To investigate academic vocabulary knowledge, three tests were devised, each addressing different knowledge level. The target words were selected from the Arab Academic Word List (AAWL) that was developed for the purposes of the current study (see Section 2.3.1 for the developing process):

¹Eight 7^th graders within the General Arab subgroup didn’t report their gender; ²Three 9^th graders within the General Arab subgroup didn’t report their gender.

1) Yes-No Academic vocabulary Test: the test included 93 words presented inself-check list, containing the 35 most frequent word-roots list (out of 55 roots), each accompanied by either two or three derived academic words as following: the most frequent word and additional one or two words that denote distinct meaning (see procedure section for the academic vocabulary list developing). In total, the students were asked to check-mark whether they know the meaning of each presented word. The total score was calculated as the total number of checked words. Aftermath, percentage scores (%) were calculated for each pupil. Alpha Cronbach value is .95.

Table 2 presents an example of the test. For the sake of convenience, the phonetic transcriptions as well as the word meaning in English were added after each word (third column from left). The students were asked to determine whether they are familiar with the meaning of each word or not by drawing a check-mark in the suitable column (fourth or fifth column).

The usage of the described test is due to the communality of using yes-no tests for assessing vocabulary knowledge in both L1 and L2, where it was found as a reliable predictor of reading ability (Anderson & Freebody, 1983) . For the purpose of the current study, an adaption of Meara (2010) yes-no vocabulary test was used. Unlike the original test that included several levels of testing and non-words to penalize over estimation, in the current test only academic words were included thus dispensing of non-words inclusion. In addition, the words are organized in relation to their roots due to the morphological structure of

Arabic (see Table 2).

2) Academic clozetest: The test included two sections. In the first section, twenty academic words were omitted from a 358 words informational text. The pupils were required to fill the blanks with the suitable target word, selecting it from a wordbank that appeared under the text. In addition to the target words, five distractors were included (overall 25 words: 5 distractors and 20 target words). One point was granted for each correct answer with a total of 20 points for this section. Alpha-Cronbach is .96.

In the second section, two comprehension questions were included, assessing the deeper understanding of the texts’ main ideas. First, the pupils were asked to identify the main idea of the text by selecting the adequate title for the completed cloze (multiple-choice question). One point was given for a correct answer. Moreover, the pupils were asked to justify their answer (an open-end question). A correct answer was given one point.

An overall percentage score was calculated for each pupil as follow: the cloze task constituted 50% of the overall score and the multiple-choice question as well as the open-end question constituted 25% each.

3) Academic Vocabulary Production in Context: the pupils were required to select seven different academic words, appearing in the first evaluation test (yes-no academic vocabulary test), and to compose a sentence for each selected word. Score range was 0 - 14. Each inappropriate response (i.e. either a missing response or incorrect utilization of the word in context) received: a) points; a partly correct response (i.e. correct allocation of the word in context but incorrect syntactical and\or morphological structures received); b) point and fully appropriate response; c) points (semantically and morpho-syntactically correct). Ultimately, a percentage score was calculated for each pupil. Alpha-Cronbach is .89.

Developing a corpus-based academic vocabulary list. Accredited Arabic middle-school teachers that underwent professional training by the head of research team conducted the mapping process. First, 8118 academic words were mapped from seven middle-school textbooks from different knowledge domains (history, civics, geography and science and technology). Excluding repetitions, 2532 academic words were identified, derived from 684 word-roots. Words and word-root frequencies were calculated across the different disciplines, leaving only words with frequency value above seven. We crossed-checked the high frequency words between the different knowledge disciplines, creating a unified lexicon that included 184 joint word-roots. From the remained 184 word-roots, the teachers rated the 40 necessary word-roots for adequate coping with the appearing expository texts in junior high schools. The selected most frequent words were cross-checked with the 50 most frequent words out of the 184 word-roots, yielding 45 word-roots. To complete a 50 word-root list, five novel high frequency word-roots (with accordance to the teachers’ rating) were added. Additional five non-crossed checked words were added as their frequency rates were very similar to the last five ranked word-roots in the 45 identified high frequency word-roots list. The fifty-five word-root list was then utilized for developing the evaluation tests.

Test administration. At the end of scholastic year, a 90-minuteacademic vocabulary battery was administrated by middle-school Arabic language teachers in the different selected schools during Arabic language lesson. Each pupil received a test booklet containing the three evaluation tasks with the relevant instructions.

All test scores were converted to percentage scores and Statistical Package for the Social Sciences (SPSS) was used for data analysis. To examine differences in pupils’ knowledge, application and production of academic vocabulary, 3 × 2 × 2 mixed factors analysis of variance (MANOVA) was performed. Additionally, multivariate analysis of variance was conducted with Bonefroni’s as a post hoc test.