Socio-economic factors, school type and the uptake of languages: Northern Ireland in the wider UK context

ABSTRACT The turning point for ‘languages for all’ is widely associated with the end of compulsory language learning up to age 16 in England (2004). Uptake has since dropped dramatically at both GCSE and A-Level where these examinations are used, i.e. in England, Wales and Northern Ireland (henceforth NI). There is also significant evidence that languages uptake is characterised by socio-economic inequalities but the evidence is strongest in England and particularly poor in NI. This paper opens by outlining the broad picture for languages around the UK, including the differing policy contexts and levels of uptake in England, Wales and NI relative to the size of the cohort in each jurisdiction. The core of the article uses new school-level entry data for NI in 2018 to explore the question of variations in uptake at GCSE and A-Level in relation to school-type and social deprivation indicators, notably ‘free school meal entitlement’ (FSME). Our findings show that stark inequities in access to language learning at Key Stages 4 and 5 are indeed attested in NI and take a very clear shape that is related both to school type and socio-economic profile. The paper also discusses the nature of these inequities and their implications for policy-makers in the devolved government.


Introduction
Modern languages education in the United Kingdom is regularly described as being in a state of crisis (Lanvers 2017). As the uptake of languages amongst undergraduate non-specialists rises, the numbers enrolled on specialist language degrees are falling across the board (British Academy 2018). Overall figures for the numbers of students taking modern languages GCSE and A-Level qualifications are also shown to be consistently in decline (Collen 2020;Tinsley 2019aTinsley , 2019b. Although uptake of Spanish in schools, unlike French and German, appears to be rising, this positive trend is not reflected in uptake at undergraduate level (British Academy 2018). A deficit of language skills in trade and industry is thought to be costing the UK economy billions (Foreman-Peck and Wang 2013) and this is likely to remain the case in the context of Brexit (Ayres-Bennett and Carruthers 2019, 2020; CBI and Pearson 2019; Thomas et al. 2018). In addition to economic growth, a recent joint statement from the four National Academies (British Academy, Academy of Medical Sciences, Royal Academy of Engineering, and Royal Society) highlighted important cultural and social benefits that could result were the UK to embrace fully its potential as a 'linguistic powerhouse' (British Academy 2019). One key aspect of achieving this objective is to improve the uptake of languages in education, both at school and university; new proposals for a National Languages Strategy have recently been published by a coalition of five bodies involving the British Academy, the Arts and Humanities Research Council, the Association of School and College Leaders, the British Council and Universities UK (British Academy 2020).
There is some evidence that the overall declining uptake of languages masks significant social inequalities in language learning opportunities. The Language Trends reports for England (and more recently for Wales and NI) consistently show that the most favourable conditions for language learning, such as a higher time allocation, are more likely to be achieved in schools with fewer disadvantaged students and at grammar or private schools (Collen 2020;Tinsley 2019aTinsley , 2019b. This is particularly concerning given that academically selective systems are more likely to have high levels of socio-economic stratification (OECD 2012) and this is an issue to which we will return in detail.
In this paper, a contextual section will set out the educational policy position in England, Wales and NI and will offer a brief comparison of uptake levels across the three jurisdictions at GCSE and A-Level relative to cohort size. We will focus on NI, England and Wales because these are the three jurisdictions where GCSEs and A-Levels are the main secondary-level qualifications: Scotland has a different assessment system, with National 1-5, Higher and Advanced Higher qualifications (Scottish Qualifications Authority 2020). Nonetheless, some aspects of the policy environment in Scotland will be drawn upon in our paper where relevant (for further discussion see Scott 2015;Doughty, Spöring, and Holmes 2019). A brief discussion of previous research on the impact of social factors on language learning will lead into an account of the questions underpinning this paper. The main body of the paper will use new data (DE(NI) 2019a) to explore the question of social inequalities in uptake in NI, looking in particular at school-type and the percentage of pupils in each school with Free School Meal Entitlement (FSME), a widely-used proxy for levels of socio-economic prosperity or deprivation. This will be the first such analysis to use a fully comprehensive dataset that includes all NI secondary schools, thereby offering a more complete picture of the impact of social factors on uptake. A final section will address the policy implications of our findings in relation to social inequalities and associated educational inequities in access to language learning.

Policy context
Our focus in this paper is on educational issues. However, for contextualisation in the wider language policy field, it is important to point out that there is a high degree of policy divergence in the approaches adopted in relation to minoritised languages across the UK. Both Wales (Welsh Language (Wales) Measure 2011) and Scotland (Gaelic Language (Scotland) Act 2005) have legislated for the protection, promotion and official status of their indigenous languages. Softer, or non-legislative, policy provision also serves to enhance the position of indigenous languages, particularly in Wales, through multiple policy documents (for example, in the delivery of health and social care services (Welsh Government 2016)) and Welsh language strategies (Welsh Government 2017a). The position in NI differs significantly. The introduction of an Irish Language Act (St Andrew's Agreement 2006) remains politically sensitive (Dunlevy and Ó Mainnín 2017), although the recent New Decade New Approach (Northern Ireland Assembly 2020) contains a renewed commitment to legislate to support Irish. There has been a legal obligation since 1998 to encourage and facilitate Irishmedium Education (Carruthers and Ó Mainnín 2018).
Although policy formation in the field of education is devolved and thus in many respects also 'divergent' across the UK, the more complex reality is that there are elements of divergence and convergence, both for education policy in general and for languages education specifically. Divergences often reflect longstanding differences, whereby particular national values, ideologies and discourses continue to influence policy directions within a pattern of policy continuity (Gallacher and Raffe 2011). Nonetheless, since many ideological and cultural beliefs are shared across the UK (Donnelly and Osborne 2005), the extent to which policies are similar, or converge, remains relatively high.
A recent analysis of language education policy illustrates the extent of policy convergence and divergence across the jurisdictions (Ayres-Bennett and Carruthers 2019). Examples of convergence include: a shared approach in the secondary phase across the four jurisdictions, where language learning is compulsory from age 11-14 years and alignment of the move from previous 'compulsory' status in Key Stage 4 (upper secondary (14 to 16) education) to 'entitlement' or similar. Examples of divergence include: different approaches to statutory vs non-statutory provision at primary level, with England, Wales and Scotland having statutory provision whilst NI does not; and one striking example in the form of Scotland's positive recognition of community languages in the '1 + 2 policy' (Scottish Government 2012), where Chinese and Urdu enjoy particular support.
It has been argued that supra-national political influence leads to greater policy convergence as nation-states align policies with international norms (Gallacher and Raffe 2011). In this regard, UK language education policy provides an interesting case study. The European policy position is that every European citizen should have proficiency in three languages, known as the 'mother tongue plus two' model set out in the 2002 Barcelona Objective. European policy documents from the mid-1990s onwards called for young people to learn at least two 'foreign' languages at school level and more recent documentation has emphasised that regional and minoritised languages may be included in assessing compliance with the European benchmark (European Commission 2012). In the context of language education policy, a certain degree of alignment with the EU objective can therefore be seen in the Scottish '1 + 2' strategy (Scottish Government 2012) and in the Welsh 'bilingual + 1' strategy (Welsh Government 2015). However, no similar policies have been implemented in England or NI.

The decline in uptake of languages in UK schools
The declining uptake of languages across all educational levels in the UK has been well documented in a number of reports and policy briefings (Ayres-Bennett and Carruthers 2019Carruthers , 2020British Academy 2018Collen, 2020;Tinsley 2019aTinsley , 2019b and academic research has repeatedly called attention to the 'crisis' (Coffey 2018;Graham et al. 2016;Grenfell and Harris 2013;Lanvers 2017). At the turn of the twenty-first century, the focus was on declining numbers at A-Level and beyond (Marshall 2000). Following the end of post-14 compulsory status in 2004, uptake at GCSE has come to the fore as a key issue (Coleman 2009), given its potential implications for numbers at A-Level and beyond (Graham and Santos 2015). However, it is important to note that it has been argued, perhaps most notably by Macaro (2008), that the decline in uptake at GCSE actually preceded the 2004 decision.
Rather than analysing raw data (which makes cross-country comparison impossible because of varying cohort size) or percentage of exam entries (where broad changes such as the move from four to three A-Levels can impact disproportionately on the statistics), our cross-country comparisons in Figures 1 and 2 below show calculations of the total number of entries for GCSE and A-Level languages as a percentage of the total cohort of students in each of the three jurisdictions. Using a comprehensive dataset, compiled from a range of sources (JCQ 2005(JCQ -2018a(JCQ , 2005(JCQ -2018bDfE 2010DfE -2018aDfE , 2010DfE -2018bDfE and DCSF 2005-2009a, 2005-2009bDE(NI) 2005Stats-Wales 2005-2018a, 2005-2018b, our data compare entry patterns over time across the three jurisdictions which use GCSE and A-Level examinations for the main shared curriculum languages (French, German and Spanish) and for 'Other languages', in addition to the indigenous languages of Welsh in Wales, and Irish in NI. Our approach is similar to that adopted by Ofqual to compare uptake (Churchward 2019) but provides additional detail beyond the figures for French, German and Spanish. It is fully acknowledged that the analysis makes no accommodation for dual or triple linguists; given that the only uptake data relate to individual languages rather than candidates, some pupils will be entered twice or more rarely three times if they take two or three additional languages and therefore the percentages are likely to be higher than the percentage of pupils 'taking a language'. Nonetheless, the data are directly comparable across the jurisdictions. Figures 1 and 2 give separate figures for French, German and Spanish, as well as a composite figure for all three which we call the 'big 3' total. For Wales and NI, figures are given for Welsh and Irish respectively. Welsh entries are given separately for Welsh as a first language (L1) and second language (L2) at GCSE level whilst a composite figure for Welsh is given at A-Level (because the JCQ data for L1 and L2 entries are only available for 2017 and 2018). Irish entries at GCSE are given as a composite figure for Irish (L2) and 'Gaeilge' (the 'Irish as first language' descriptor: L1) because these are not reported separately by JCQ. We know anecdotally that GCSE Gaeilge entries have grown as a proportion of the overall entries (and numbers in Irish have declined), although no full analysis is currently available. No A-Level 'Gaeilge' option exists and therefore  figures are for A-Level Irish only. Totals for 'Other' languages (i.e. all other languages entries which do not include French, German, Spanish, and Irish or Welsh where relevant) are reported. The final figure given in each case is the 'all languages total' which comprises the total entries for each language category reported by JCQ (i.e. French, German, Spanish, Irish, Welsh and 'Other').
As we might have predicted, the overall pattern is one of decline in mostbut not allinstances. Our analysis reveals a number of differences across the three jurisdictions. Broadly speaking, in terms of uptake relative to the size of the cohort, NI has better overall uptake levels of the 'big 3' foreign languages (French, Spanish and German) at GCSE and A-Level than England and Wales. Wales is in the weakest position in terms of 'big 3' uptake at both GCSE and A-Level but its position for 'all languages' is dramatically boosted by extremely strong uptake of Welsh (L2) at GCSE and, to a lesser extent, Welsh at A-Level. Our GCSE data confirm that the EBacc implementation resulted in a rise in languages entries in England in 2011 and 2012 but that this halted with the introduction of Progress 8 which has no language requirement (Gill 2017). French has undergone by far the steepest decline in all jurisdictions and while Spanish uptake has increased considerably at GCSE in England and NI, this is not well maintained at A-Level when we consider its position relative to the size of the cohort. German has declined across the board and even where the decline is less steep (as is the case in NI), the rate of uptake and critical mass of pupils is extremely low. Irish is reasonably steady although there is a small decline at GCSE. However, since our data cannot show the breakdown for Irish offered as an L1 (Irish-medium 'Gaeilge') and as an L2 ('Irish'), any shift in balance between the two will not be visible in our analysis. Uptake of 'other languages' is also steady, and in some cases rising, although there are large variations in the proportions of this figure accounted for by the individual languages (Tinsley 2019c).
These trends based on cohort percentages would not necessarily have emerged from an analysis based on uptake numbers alone. For example, raw numbers would suggest that A-Level Spanish is growing, whereas relative to the size of the cohort in all three jurisdictions, this is not the case. In Wales, the extent of the decline in the 'big 3' at A-Level is much clearer when viewed as a cohort percentage, as is the extent of the growth of Welsh as a second language and its impact on the 'all languages' figure. In NI, the cohort percentage uptake of Irish and German at A-Level is much steadier than raw pupil entry numbers would suggest, given that there has been a rise in the cohort size in recent years until around 2016.
In many respects, NI is in a better position, relatively speaking, than England and Wales: this is certainly the case for uptake of 'foreign' languages. On the 'big 3' indicator, NI is in the strongest position at both GCSE and A-Level, particularly relative to Wales. Moreover, the gradient of the decline at A-Level is distinctly less steep in NI than elsewhere. At this juncture, therefore, we note that there may be some reasons to be positive in relation to NI although the wider context is unquestionably one of decline.

Social factors and the uptake of languages
Historically, language study in the UK has been considered the preserve of the elite and until the 1960s, access to qualifications in modern languages across the UK was limited to those attending private or grammar schools (McLelland 2018). Despite major changes towards 'democratisation' of language learning through developments such as comprehensivisation in the 1970s and the introduction of a common curriculum (Education Reform Act 1988; The Education Reform (Northern Ireland) Order 1989) where a modern language was compulsory until age 16, concerns persisted about equitable provision for all pupils (Dobson 2018;McKendry 2016;McLelland 2018). The decision to shorten the period of compulsory language learning in 2004 appears not only to have accelerated a decline in uptake but also to have (re-)entrenched social inequalities (Dearing and King 2007). As noted in our introduction, in England and Wales, there is evidence to demonstrate that factors such as Free School Meal Entitlement (FSME) and school type influence uptake profiles: uptake is statistically most likely to be higher in schools with the least disadvantaged students and/or in selective schools (Collen 2020;Tinsley 2019b;Vidal Rodeiro 2009). The recent (and first) Languages Trends NI (Tinsley 2019a) report found similar disparities by school selectivity and socio-economic profile, with lower uptake at GCSE in non-selective schools and those with higher proportions of FSME students (p. 25). In England, Vidal Rodeiro (2017) reports higher A-Level uptake amongst students who attend private schools and are from more affluent areas. Evidence of significant geographical disparities in uptake (Collen 2020) is depicted by the media as a 'North-South' divide (Coughlan 2017). However, it is likely that apparent geographical inequalities result from other demographic disparities, namely differential levels of socio-economic deprivation or cultural and linguistic diversity which have been shown to vary geographically (Graham 2017). For example, uptake in London is higher relative to other areas but this is unsurprising because it is also the most ethnically diverse area in England (Office for National Statistics 2018).
The recent Language Trends reports for each jurisdiction (Collen 2020;Tinsley 2019aTinsley , 2019b) also suggest that a number of other factors are relevant, notably perceptions around difficulty, severe grading, societal changes such as exiting the EU, messages that 'English is enough', as well as educational issues such as time devoted to language learning in the curriculum. Several of these factors have long been known to be mediated by existing educational inequalities. For example, young people's perceptions of 'value' (and their onward trajectories) can be shaped by the proportion of curriculum time dedicated to specific subjects and there is clear evidence that variations in time allocation for languages are, in turn, strongly related to school type (Evans and Fisher 2009;Harland et al. 2002).
Despite the mounting body of evidence emerging from numerous reports, the influence of social factors has not received as much attention in recent research as we might expect. One of the reasons for this is undoubtedly the variety of social factors that are pertinent to the discussion and in particular, the complexity of the notion of 'social class'. The latter has been a challenging construct in the discourse of sociolinguistics since Labov's pioneering work in the1960s; it has been recognised for many years that social class is not just economic but has important social, cultural and educational dimensions (Tagliamonte 2012, 25-70). More recently, particularly amongst applied linguists, scholars have drawn on Bourdieu to discuss the ways in which language learning and attitudes to multilingualism may be related to the complex interplay of economic and cultural capital, the latter including educational and linguistic capital (Block 2018, 89; see also Coffey 2018; Norton 2010). For example, Coffey (2018) uses Bourdieusian concepts to explore social variations in how young people construct their real or imagined identities as language learners, showing how different attitudes and motivations influence their potential language learning trajectories, and pointing up significant differences between the discourses of learners in state and independent schools in London. Lanvers (2017) also adopts a Bourdieusian lens in her systematic review of the literature on the relationship between social factors and access to language learning. She identifies multiple system-level and school-level factors, such as selectivity and messages from school leadership, that can magnify social inequalities and their effect on uptake. Elsewhere the approach is more quantitative: for example, in their evaluation of an educational intervention, Grenfell and Harris (2013) use multiple regression analysis to explore the influence of psychological and sociocultural (including socio-economic) factors on motivation and performance in MFL in two London schools. Nonetheless, the fact remains that in research on language learning, social issues are a 'neglected factor' (Coffey 2018, 463), representing 'striking lacunae' in the literature (Lanvers 2017, 517). As Block (2018, 147) puts it, 'social class in second language acquisition and social class issues in general seldom get a look-in'.
In the light of the UK policy context, cross-country uptake trends and previous research on the impact of social inequalities, this paper will focus on two specific social factors in the NI educational context, i.e. school type and levels of FSME. As noted above, both of these factors are shown to be relevant in Language Trends (Tinsley 2019a), but they are not explored in detail and in any case, the reliance on self-selection in sampling means that the selective sector is over-represented. Our interrogation of a newly-compiled NI dataset will thus complement and expand on previous analyses (such as Vidal Rodeiro 2017), looking at uptake in relation to school-level social factors at both GCSE and A-Level. The questions underpinning our analysis are as follows: . Do the headline figures for NI (which are slightly more positive than elsewhere in the UK) mask variations in uptake in different types of school? . What do the data in relation to FSME tell us about questions of social inequality in uptake of languages in NI? . What are the implications of the patterns emerging from the data for the future of languages in schools in NI and for education policy?
Socio-economic factors in Northern Ireland: uptake of languages by school type and FSME The Northern Ireland school system The comprehensivisation of secondary education in the UK did not extend to NI (Gallagher and Smith 2000), which has retained an academically selective system that separates pupils around the age of 11 into 'grammar' and 'non-grammar' schools (Gardner 2016)

School type and FSME
The reasons for exploring school type and FSME are three-fold. First, school type in the NI context is strongly linked to questions of cultural and educational capital. There is a clear 'symbolic' difference between school types in NI's academically selective system and this is reflected in the fact that many grammar schools are oversubscribed despite the requirement to opt into assessments for selection (Elwood 2013;Henderson 2020). Second, although the measurement of socio-economic status and social class are methodologically complex, FSME is a widely-used proxy for social deprivation and at the school level, 'FSME profile' can be used as an indication of the concentration of socio-economic deprivation in a school's population (Allen and Vignoles 2007). Third, secondary school data for NI demonstrate a strong link between school type and FSME representation (Henderson 2020) and therefore when analysed alongside each other, they provide a route into issues around both social and economic capital. The data in Tables 1 and 2 refer to the total population of schools in NI and certain traits are worth noting as relevant for the analysis that follows. Particular attention should be paid to the representation of only non-grammar schools in the highest FSME quartile and only grammar schools in the lowest FSME quartile, which reflects a broader pattern of association between academic selectivity and levels of FSME (Henderson 2020). We will return to this in the presentation of data below.

Accessing data
There are no publicly-available data regarding the uptake of languages across different school types in NI and we therefore requested these new data from the NI Department of Education for the purposes of this analysis. The data detail the number of entries for individual languages qualifications for the 2017/18 school year (DE(NI) 2019a), and in the following sections, we analyse the patterns of languages uptake and cohort size indicators for different types of school, according to their selective status, and in quartiles by their overall FSME population (DE(NI) 2018), with lower FSME percentages associated with higher levels of social deprivation and vice versa. This secondary data analysis thus provides a snapshot of student uptake in languages during the academic year under investigation. The purpose is to understand better the potential impact of school type and socio-economic factors on the current uptake of languages in NI's schools and it is fully acknowledged that this approach does not enable us to make year on year comparisons. Nonetheless, the 2018 snapshot can be read against the overall longitudinal patterns discussed above, where the direction of travel is clear. Moreover, a significant strength is that the findings around social inequities include every post-primary school in NI and can therefore be read with a high level of confidence. The detail provided by this new analysis also discusses the health and viability of languages beyond the end of the  compulsory period (up to age 14) in the selective and non-selective sectors. This is particularly important given that existing data from the Language Trends NI report (Tinsley 2019a), as noted above, rely on a dataset where grammar schools were overrepresented in terms of voluntary responses to the survey.

GSCE entries
Our data show that 14.6% of NI schools have no KS4 languages provision, a higher proportion than that reported by Language Trends NI which calculated the figure at 9% based on a large sample of schools. Figure 3 shows a clear difference in the proportions of entries across the two school types with an absolute majority (94.4%) of non-grammar schools having entries equating to less than half the total cohort of students. The data show that more than 20% of non-grammar schools have no pupils from the Year 12 cohort studying a language at GCSE and in 64.3% of this school type, 20% or fewer pupils pursue a KS4 language option. Conversely, more than two-thirds (68.2%) of grammar schools have language entries equivalent to at least 80% of the cohort. A significant number of grammar schools also have languages cohorts of more than 100% of the total number of students in their cohort indicating that at least some students are dual linguists at GCSE.
Greater insight into the inequalities within the system is offered when the data are grouped by FSME quartile. Figure 4 shows that more than a quarter of schools with the highest FSME populations have no uptake for any languages at GCSE level and no school in this category has more than half of its students studying a language at GSCE. Almost half of schools with the lowest FSME population have languages entries equating to more than 100% of the cohort, indicating that most or all students take a language for GCSE and at least some students are dual-linguists. These data show clearly that uptake of languages at GCSE is lowest in schools which serve the least advantaged student populations and highest in those that serve the most advantaged.

A-Level entries
Similarly to Figure 3, Figure 5 shows a clear difference at A-Level in the proportions of entries across the two school types. For more than half (56.3%) of non-grammar schools that operate a sixth form, there were no candidates for French, German, Irish or Spanish in the 2017/18 academic year. An absolute majority (97%, all but 2 schools) of grammar schools retain some provision of modern languages at A-Level, although there are significant between-school variations in the proportions of students taking these qualifications, with only 12.1% of grammar schools having more than the equivalent of 20% of the cohort taking a language. Figure 6 shows A-Level uptake patterns between schools with the highest and lowest FSME populations to be in stark contrast. In schools within both the highest and above median FSME quartiles, the position of languages is extremely concerning, with (respectively) 84.6% and 75.6% of schools having no A-Level uptake whatsoever. Comparatively, the uptake is higher in schools with below median and lowest representation of FSME students but is nonetheless below 20% in 95% and 85.3% (respectively) of these schools. Similar to the analysis of GCSE uptake, the A-Level data provide significant cause for concern on questions of equality of access to language learning. However, they also clearly show that even in less disadvantaged schools, uptake is low when measured as a percentage of the cohort, reflecting the overall position of precarity discussed above with reference to GCSE entries.   Figure 7 shows the percentage of schools within each category (grammar or non-grammar) with cohorts for the four languages in bands of 10 students (1-10 students, 11-20 students and so on). The purpose is to identify the proportion of schools where languages are 'at risk' of being considered financially unviable. This analysis is unique in showing the differential implications of declining uptake for grammar and non-grammar schools. Whilst there is no universally accepted view of where the cut-off point for a viable class size would fall, we can suggest that a timetabled GCSE class for fewer than 20 students could be considered to be running with numbers that, in the context of the model for school funding, are at risk of being financially unviable. In this scenario, in addition to the 1.5% of grammar and 42.9% of non-grammar schools with no GCSE French provision, a further 16.7% of grammar and 46% of non-grammar schools are currently running GCSE French classes in a way that is not financially sustainable within the funding model. Put the other way around, French is running with cohorts of over 20 in 81.8% of grammar schools but only in 11.1% of non-grammar schools. The data suggest that despite declining numbers, significant proportions of schools are making every effort to continue to run classes to maintain languages provision within the school. However, it is evident that the potential negative impact of small cohorts on non-grammar schools is proportionately much greater. This is the case for French, Irish and Spanish: not only are fewer non-grammar schools running these subjects, but also, higher proportions of nongrammar cohorts have fewer than 20 students than in the grammar sector. The situation for German is somewhat different, with only a very small number of non-grammar schools (4.8%) maintaining this subject at GCSE level. In addition, across all languages, the range of cohort sizes is much larger for grammar than non-grammar schools although both school types are shown to have extremely small cohorts at the lower end of the range. Nonetheless, the average cohort size for grammar schools is at least double that for non-grammar schools for each of the four languages shown. Given the patterns of falling uptake, the number of grammar schools running GCSE French with fewer than 20 pupils could climb. Figure 8 offers significant insight into how declining GCSE uptake has differential consequences for schools according to levels of FSME. Across each of the languages, within the highest and above median FSME quartiles, in addition to larger proportions of schools with no provision, the absolute majority of languages cohorts are running with fewer than 20 students and cohorts with fewer than 10 students are not uncommon. For example, in the above median FSME quartile, in addition to the 40.4% of schools with no French cohort, half of those schools with a French cohort have fewer than 10 students. For each of the languages, the overall proportions of schools running a GCSE cohort is greater in the below median and lowest FSME quartiles. Furthermore, the highest and average cohort sizes are much larger for these schools when compared with schools where the FSME population is higher. German is in an extremely vulnerable position in all but the lowest FSME quartile; reflecting the patterns evidenced by the cross-country data presented in the context section above, these data offer a clear indication of a lack of long-term sustainability. Of equally grave concern is that, despite greater levels of provision, the pattern of unsustainability extends to cohorts in many schools in the below median FSME quartile, across all of the languages, with almost half of all 2017/18 cohorts in French, German and Irish running with fewer than 20 candidates. Average cohort sizes are only sustainable (in the short term) in schools with the lowest FSME representation. This raises a clear issue of equity in relation to provision and offers further evidence that, despite GCSE classes continuing to run for very small numbers of students across all FSME quartiles, a disproportionate share of small vulnerable cohorts are found in schools catering to the most socio-economically disadvantaged students.

Cohort size at A-Level
Extreme differences in the position of languages at A-Level between grammar and non-grammar schools are evidenced by Figure 9 which shows the percentage of schools within each category with cohorts for the four languages in bands of 5 students (1-5 students, 6-10 students and so on). A large majority of grammar schools continue to offer French (92.4%) and Spanish (80.3%), and a significant minority offer German (36.4%) and Irish (34.8%). The proportions of nongrammar schools with cohorts in each of the languages is very small for French (10.5%) and Spanish (9.5%), though the position is marginally better for Irish (13.7%). German is completely absent from the non-grammar sector. Overall, the size of A-Level cohorts is small with a larger average cohort size for French and Spanish in grammar schools when compared with nongrammar schools. Evidence of relatively large cohort sizes for Irish at non-grammar schools is accounted for by the inclusion of Irish-medium streams and schools in the analysis (where the majority of provision is in the non-grammar sector). Moreover, across sectors for each language, a large proportion of schools are running cohorts of fewer than 10 pupils. In the grammar sector, only 13.6% of schools have an A-Level class of more than 10 pupils in French, while the figure is 7.5% for Irish, 13.6% for Spanish and 1.5% for German. In each case the smallest cohort of students is a single candidate. This is particularly concerning and once again illustrates an obvious issue in terms of the provision of teaching staff. Furthermore, these figures are not outliers: six individual grammar schools had a cohort of one student. Even in cases where provision of lessons is shared between several schools, the current situation is untenable. Since having a minimum of 10 pupils is often considered the threshold for viability at A-Level, these figures suggest that A-Level in all four languages is in an extremely precarious position. The situation in both selective and non-selective schools is fragile: the non-selective sector is disappearing from A-Level provision and the bulk of the selective sector is maintaining languages at what most school leadership teams would consider to be unviable levels. As with the analyses presented above, Figure 10 shows that schools with the higher FSME representation (highest and above median quartiles) are in the weakest position, in this case with an extremely small number of schools having any A-Level provision at all. Where provision does exist in these quartiles it is for very small numbers of students: in French and Spanish no cohort has more than 10 students and while Irish is in a relatively strong position in some individual schools, these are likely to be schools with larger numbers of students who have current or prior experience of Irish-medium Education. Schools in the below median and lowest FSME quartiles are in a stronger relative position, with larger proportions of schools in these groupings maintaining A-Level provision, in many cases for multiple languages. However, even in these quartiles, only a very small minority of cohorts (i.e. between 2.4% and 19.5%) contain more than 10 students.
The overall picture at A-Level is therefore extremely concerning in several ways. First, provision in the non-grammar sector and for the most socio-economically disadvantaged pupils is in danger of disappearing completely. The fact that it is much more stable in grammar schools and lower FSME contexts may be accounted for by relatively high levels of school-level compulsory GCSE provision in these sectors. However, the viability at A-Level in many grammar schools is highly vulnerable and given the falling numbers at GCSE, the offer in the grammar sector could change rapidly if Principals are forced to take tough economic decisions that make classes of fewer than ten unacceptable. There is anecdotal information that this is already happening. The financial sustainability of A-Level languages is thus questionable in an absolute majority of schools.

Conclusions and implications for policy-makers
Like England and Wales, NI is undergoing a period where decline in uptake of languages at GCSE and A-Level is the dominant narrative. When the position of the 'big 3' (i.e. French, German and Spanish) in NI is set in the wider UK context, the figures suggest that NI is faring better than England and much better than Wales. Nonetheless, the broader NI patterns are moving in a negative direction and mask extremely sharp social inequities in terms of access to language learning which both reflect and reinforce differences in social and economic capital (Block 2014). This is evidenced by stark differential provision for schools of different types and socio-economic profiles. Our analysis, which is based on the full population of schools, depicts larger variations between the grammar and non-grammar sectors and between schools with low and high levels of FSME, than Language Trends, which adopted a self-selecting approach to sampling. In short, children in grammar schools and from schools with lower levels of FSME are much more likely to have a language in their profile than those from non-grammar schools and those in schools with higher levels of FSME. Similar to the pattern in England, the divide between those children who are and are not studying a language is to a large extent a socio-economic one (Graham 2017) but in the NI system, school selectivity magnifies socio-economic stratification (Henderson 2020).
In terms of viability, our statistics suggest that GCSE provision in grammar schools is the only apparently 'safe' zone for languages provision in NI schools. The data discussed in this paper suggest that every other area of provision within the languages landscape of NI is vulnerable: GCSE and A-Level are at risk in the non-grammar sector and many schools already have no cohorts at either KS4 or KS5; and A-Level provision in the grammar sector, where this is still available, is operating at non-viable levels in large numbers of schools. Moreover, even at GCSE in the grammar sector, numbers are not secure: where many NI grammar schools retained a 'compulsory' language at GCSE long after the 2006 government decision (2004 in England) to remove its statutory status (thus boosting uptake relative to England and Wales), this is currently an area of rapid change which is likely to impact negatively on GCSE numbers as increasing numbers of schools make a language optional for pupils.
What are the implications for policy-making in NI? It is very clear from our analysis that policy matters and is of both symbolic and practical importance. It is of course most powerful at government level and this is clear from our comparison of uptake across the UK. For example, as is widely acknowledged, the 2004 decision to remove the statutory requirement to learn a language until the age of 16 appears in the long run to have had a negative impact on languages uptake. The EBacc requirement for a language brought a temporary boost to numbers in England but this was undercut by the criteria attached to policy around Progress 8 (Gill 2017). Policy in Wales in relation to Welsh has had a powerful positive effect on uptake (Welsh Government 2017b, 2019), although it is widely perceived to be having a negative effect on uptake of foreign languages (Lewis 2019) and the evidence shows that secondary-level pupils in the Welsh-Medium sector are less likely to study a modern 'foreign' language (Tinsley 2019b). In NI, school-level policy (rather than government policy) supported GCSE uptake for many years after the end of compulsory status but our data confirm that this is declining. In this context of lack of statutory support, it is clear that other measures will be vital in encouraging pupils to make positive decisions around language continuation.
Some interventions need to be enacted at UK level in order to address barriers to uptake that have been identified across all jurisdictions; many of these are set out in Towards a National Languages Strategy for the UK: Education and Skills (British Academy 2020) which includes recommendations for government and the education sector. In terms of specific interventions for NI, it is vital that both grammar and non-grammar sectors are supported although the two sectors are impacted in different ways on particular issues. For example, we know from previous research that questions of severe grading and difficulty are major barriers (Collen 2020;Tinsley 2019aTinsley , 2019b) and research consistently points to a modern languages 'grade deficit' (Thomson 2019). The 'severe grading' issue impacts extremely heavily at GCSE on the non-grammar sector and at both GCSE and A-Level in the selective sector. Not only does it affect pupil choices but also, accountability measures (such as the percentage of pupils obtaining GCSE C and above; or measures of excellence such as numbers of As and A*s) mean that Principals and careers teachers, including those who are highly supportive of languages, are wary of encouraging some students to take a language because of uncertainty around grading. In the non-grammar sector, widespread opting out of languages at GCSE can push languages into the 'unviable' bracket in many schools, ultimately cutting off access to all pupils in the school. In relation to grading, a review at GCSE and A-Level is ongoing in NI, along similar lines to the recent reviews of GCSE (Ofqual 2019) and A-Level (Ofqual 2018) in England. An adjustment to grading has now been agreed for GCSE French and German (in England) by Ofqual, so it is all the more urgent that the NI review is completed in order to restore confidence and remove a well-known barrier to uptake (Ayres-Bennett and Carruthers 2019).
Important measures to encourage pupils to opt into languages for positive reasons have been put in place in all three other jurisdictions. Scotland has long had its '1 + 2' policy in place (Scottish Government 2012), Wales has invested in a mentoring scheme (MFL Mentoring 2015; Tinsley 2018) involving undergraduates and pupils and targeted at raising GCSE numbers, England has launched a Centre of Excellence with hub schools (NCELP 2019). While the AHRC, through its Open World Research Initiative, ran a pilot Language Ambassadors scheme in 2019-2020 which aimed to raise A-Level numbers in several regions of the UK, including NI. A bespoke sustainable model is required urgently for NI, taking into account the evaluation of the schemes mentioned above. Any NI model needs to take account of the variation in uptake exposed in this paper. In our view, schools with viable GCSE cohorts and existing A-Level cohorts need to be supported particularly at KS4 and in successful conversion from GCSE to A-Level; those with low GCSE numbers and little or no A-Level provision need to be supported to build up the GCSE cohort in the first instance in KS3. This is not entirely about a binary divide between grammar and non-grammar schools: it must be acknowledged that languages are in a relatively strong position in some non-grammar schools and in a weak position in some grammar schools. Nonetheless, the patterns discussed in this paper suggest that the high levels of socio-economic stratification associated with selective systems have a differential impact on the uptake of languages in NI. Any government policy intervention needs to ensure that schools within both sectors are supported in the most effective ways: one size does not fit all.
Finally, potential positive policy interventions need to take account of less tangible but vital qualitative factors that impact on the choices young people make; these too are strongly conditioned by socio-cultural factors and the relevance or not of languages for their real or imagined futures (Coffey 2018). It is essential that the 'entitlement framework' (DE(NI) 2017) is meaningful in NI for all pupils, including those from socially disadvantaged backgrounds, such that all young people see themselves as having the capacity to benefit from the cultural and economic advantages that language learning brings.

Disclosure statement
No potential conflict of interest was reported by the author(s).