It is widely accepted that adult second/foreign language (L2) learners rarely achieve nativelike proficiency in their L2 and that younger children are generally more successful in L2 learning than older children or adults (Flege et al., 1999; DeKeyser, 2012, 2000; Granena and Long, 2013). This is often attributed to a critical period or, more broadly, age constraints on language acquisition (see Lenneberg, 1967; Johnson and Newport, 1989; DeKeyser, 2000; Abrahamsson and Hyltenstam, 2009; DeKeyser et al., 2010; Granena and Long, 2013; Han, 2013). According to the Critical Period Hypothesis, adults and children are not equipotential when it comes to language learning: specifically, one or more of the learning mechanisms available to children is no longer available to mature learners. Following from this, it has been proposed that there is a “fundamental difference” in the way that children and adults acquire language, as well as the final outcome, with children exhibiting uniform success in contrast to adults who show great variability in ultimate attainment (Bley-Vroman, 1989; DeKeyser, 2000; Han, 2013).

This individual variation in adults has been the focus of individual differences research according to which there might be different factors responsible for (non)nativelike ultimate attainment which may or may not be confounded with age of acquisition (Muñoz and Singleton, 2011; Paradis, 2019; Li et al., 2022). This would help explain why there are cases in which early L2 onset does not seem to suffice for L2 nativelike proficiency (McDonald, 2000; Soto-Corominas et al., 2020). Among others, the factors behind this could either be external, like input (Flege et al., 1999; Flege, 2018) or learner-internal, for example language aptitude (Abrahamsson and Hyltenstam, 2008; Sparks et al., 2011; Li, 2015).

Building on the aforementioned research on individual differences, the current study aims to investigate the predictive power of input and aptitude over both the first (L1) and second language (L2) grammar of Greek-English adult bilinguals. By input we refer to the language experience that learners accumulated throughout their lives and up to the moment of testing; in this paper, the terms input and exposure will be used interchangeably. In L2 research, the term aptitude is usually defined as “an individual’s initial state of readiness and capacity for learning a foreign language, and probable facility in doing so given the presence of motivation and opportunity” (Carroll, 1981, p. 81). However, given some recent research suggesting that the abilities referred to as “aptitude” are also relevant for L1 acquisition and processing, we will use the term more broadly to refer to the capacity for learning language in general, i.e., both the L1 and the L2.

According to the original formulation of the Fundamental Difference Hypothesis (FDH; Bley-Vroman, 1989), learning a language in adulthood is different from child language acquisition because adults no longer have (full) access to Universal Grammar. More recently, the hypothesis has been reformulated in terms of explicit and implicit learning mechanisms: Children acquire the grammatical rules of their language effortlessly and without any conscious effort by relying on (possibly language specific) implicit learning mechanisms (Ullman, 2001; DeKeyser and Larson-Hall, 2005; Montrul, 2008; Kidd, 2012; Kidd and Arciuli, 2016), whereas adults need to rely, at least partly, on domain general explicit learning mechanisms. This is because implicit learning abilities are thought to decrease as a function of age (Ullman, 2001; Paradis, 2011; Granena and Long, 2013); because of this, adults have to resort to explicit learning strategies to compensate (Norris and Ortega, 2001; DeKeyser et al., 2010; Spada and Tomita, 2010). However, this distinction between adults and children has been questioned in two ways.

Firstly, there is not sufficient evidence showing that adults’ implicit skills are indeed worse than those of children. Adults certainly are capable of acquiring new knowledge implicitly, as demonstrated by a number of studies on (semi) artificial language learning (Marsden et al., 2013; Rebuschat, 2013; Grey et al., 2014). Whether these abilities are inferior to those of children is unclear. While a number of studies (e.g., Howard and Howard, 2013) report a decline in implicit learning as a function of age, most of these studies compared younger and older adults and hence do not speak to the question whether children are better implicit learners than adults. There are very few studies which compared the implicit learning abilities of young adults and children, and most of these found that implicit learning abilities continue to improve throughout childhood and adolescence and begin to decline only in later adulthood (see, e.g., Thomas et al., 2004; Lukács and Kemény, 2015).¹

Secondly, research on ultimate L2 attainment comparing children and adults is inconclusive. Although Critical Period (CP) effects are reported in many studies (Asher and García, 1969; Ramsey and Wright, 1974; Johnson and Newport, 1989; Hyltenstam, 1992; Flege et al., 1999; Morford and Mayberry, 2000; Granena and Long, 2013), in some studies adults have been found to outperform younger learners in L2 learning (Asher and Price, 1967; Cenoz, 2003). This opposite pattern has been attributed to the fact that ‘the younger the better’ position might be domain-specific, with research supporting that this advantage is mainly associated with phonetic/phonological performance (Kirch, 1956; Flege et al., 1995, 1999) whereas grammar is not necessarily influenced by the same maturational constraints (Snow and Hoefnagel-Höhle, 1978). Indeed, the results of studies which investigated grammar are mixed: while some studies report a (near-)categorical difference between younger and older learners (Johnson and Newport, 1989; DeKeyser, 2000; DeKeyser et al., 2010; Abrahamsson, 2012), others (e.g., Sasaki, 1996; Street, 2017; Dąbrowska, 2019; Dąbrowska et al., 2020) have found considerable overlap between groups. These differences appear to be partly attributable to which aspects of grammar are investigated (L2 speakers do relatively well compared to native speakers on aspects of grammar which make a relatively clear contribution to meaning, in contrast to performance on grammatical distinctions which are largely redundant from a semantic point of view) and partly to the composition of the native control group (there is much more overlap between groups in studies in which the native control group was demographically more diverse – see Dąbrowska et al., 2020).

It is well known that some learners acquire a second language much faster than others. Such differences in speed of learning are attributed to individual differences in language aptitude, which, it has been argued, can sometimes compensate for age-related disadvantages in ultimate attainment. Foreign language aptitude involves a number of distinct abilities, including phonetic coding ability (the ability to discriminate and identify new language sounds), associative memory (the ability to learn a large number of items in a short time), grammatical sensitivity (awareness of the grammatical function of the different elements constituting a sentence) and inductive language learning ability, or the ability to identify grammatical/meaning patterns in an unknown language sample (Carroll, 1964, 1990). The last two abilities are the most relevant for the acquisition of grammar (Li, 2015).

Language aptitude tests such as the MLAT (Modern Language Aptitude Test, Carroll and Sapon, 1959) or PLAB (Pimsleur Language Aptitude Battery, Pimsleur, 1966; Pimsleur, 2004) were originally developed to predict the learning outcomes of the early stages of L2 learning in instructional settings. However, later research has also applied them to later stages of acquisition and ultimate attainment and to naturalistic settings (DeKeyser, 2000; Abrahamsson and Hyltenstam, 2008; Granena and Long, 2013). In fact, it has been argued that aptitude may be even more relevant in such settings because naturalistic learners must deduce the regularities and patterns of the language on their own without explicit instruction (Skehan, 1991; Granena, 2014). It is, indeed, in immersive settings where aptitude has yielded more intriguing but also conflicting results. Researchers concerned with this learning environment have mainly focused on the relationship between age and aptitude as a key point of investigation for the validity of the FDH.

More specifically, DeKeyser (2000) conducted a study with L1 Hungarian speakers in the United States in order to test the FDH. The participants were administered a grammaticality judgement task in English and an aptitude test in Hungarian. The results revealed that there was a correlation between aptitude and grammatical proficiency for the late arrivals, i.e., adult arrivals (0.33), but no correlation for the younger group (0.07). One possible explanation for this is that, when asked about which language they felt more comfortable with, the majority of the participants from the younger group in DeKeyser’s study reported English instead of Hungarian. This could have had affected their performance in the aptitude task (administered in Hungarian), which could in turn explain the lack of an effect of aptitude in the younger group.

By contrast, Abrahamsson and Hyltenstam (2008) tested 42 L2 speakers of Swedish with L1 Spanish exploring the relationship between aptitude and L2 proficiency among near-native L2 speakers. They found a striking correlation as high as 0.70 between aptitude and grammatical proficiency and what was even more surprising is that they found a relationship for the younger (age of onset <12) instead of the older (age of onset >12) group. In this study, the lack of an effect for the older group was attributed to the small sample size (n = 11 participants) as well as the narrow distribution of scores in the aptitude task for the same group. Finally, Granena and Long (2013) compared three groups of participants dividing them based on their age of onset (i.e., 3–6, 7–15, and 16–29 years) and did not find any relationship between morphosyntactic proficiency and aptitude in any of the three groups.

To add to this complex relationship, several studies have reported a relationship between aptitude and performance in the native language. Dąbrowska (2018) found a moderately strong correlation (0.46) between grammatical comprehension and (foreign) language aptitude as measured by the Language Analysis subtest from the PLAB in adult monolinguals. A more recent study by Llompart and Dąbrowska (in press) reports similar findings. In this last study, the researchers employed two different tasks (Language Analysis and Sentence Pairs, a test modelled on the Words in Sentences task of the MLAT) to measure aptitude, and an auditory grammaticality judgement task and a picture selection task to assess grammar. In the Language Analysis task, participants are required to deduce the meaning of a number of sentences in a foreign language by using the English translations of the words that form up those sentences. In the Sentence Pairs task, participants are presented with pairs of sentences, and they have to find the word of the second sentence that has the same function as a word in the first sentence which appears in capital letters. In line with Dąbrowska (2018), Llompart and Dąbrowska also report correlations over 0.40 between (foreign) language aptitude and native grammatical proficiency. Two other studies (Skehan and Ducroquet, 1988; Sparks et al., 2009) also found robust correlations between language aptitude measures (assessed when participants were in their teens) and earlier measures of L1 acquisition.

These findings are of interest because they challenge the view that child and adult language acquisition depend on distinct and fundamentally different systems, namely implicit and explicit learning mechanisms. If the existence of correlations between (explicit) language aptitude and attainment indicates that the latter relies on explicit learning, these findings suggest that explicit learning mechanisms may also be involved in L1 development (Dąbrowska, 2010; Llompart and Dąbrowska, 2020).

While the failure of most adult L2 learners to achieve nativelike proficiency is often attributed to maturational factors, other factors may also play a role. Among these, another important candidate is input (Huttenlocher, 1998; Huttenlocher et al., 2002; Clegg and Ginsborg, 2006; Hoff, 2006), which has been shown to be relevant for both L1 and L2 acquisition (Flege and Liu, 2001; Singleton and Ryan, 2004; Moyer, 2011). Furthermore, in connection with the current study, there is considerable evidence that variation in grammatical development can be explained to a great extent by differences in input (Gathercole, 2002; Montrul and Potowski, 2007; Chondrogianni and Marinis, 2011; Unsworth et al., 2014; Kaltsa et al., 2020).

It is clear that L1 learners as well as early L2 learners are more advantaged compared to late L2 learners in the amounts of input that they receive. Dąbrowska (2015) estimated that children receive at least 11,680 h of exposure to their L1 during the first years of their life (1–5 years old). Additionally, it has been shown that the benefits of exposure do not stop at a very young age but continue until adolescent years (Nippold, 1998; Berman, 2004, 2007; Nippold et al., 2005; Kaplan and Berman, 2015), and a recent large-scale study by Hartshorne et al. (2018) found that grammatical proficiency in native speakers continues to improve until about the age of 30. L2 learners, in contrast, not only receive less input overall (because they start learning later and typically continue to use the L1 alongside the L2), but also tend to get input of a poorer quality. Whether in instructional or immersive settings, L2 learners (in contrast to L1 learners) are often surrounded by other foreign language speakers and even their teachers may often be non-native speakers of the target language.

Traditionally, studies investigating the role of input have used length of residence (LoR) as a measure for the amount of input that speakers have received (Flege and Liu, 2001; Babcock et al., 2012; Saito and Brajot, 2013; Higby and Obler, 2016). However, as argued by Flege (2008), the use of LoR as an estimate of exposure assumes that exposure to the target language begins with arrival in the host country, and that immigrants with the same length of residence have received similar amounts of L2 input. Both of these assumptions are problematic. Many immigrants start learning or are exposed to the L2 before their arrival to the host country. Furthermore, individuals with the same LoR often differ tremendously in use of the target language: while some immigrants fully immerse themselves in the host environment from the beginning, others form closer bonds with their own communities, and may thus receive very little L2 input (Birdsong, 2009).

An additional methodological issue is the ‘age-length-onset’ problem (Stevens, 2006). In particular, Higby and Obler (2016) have argued that confounding issues between the three variables (age at testing, LoR and age of onset) give us a distorted picture of the exact contribution of each of these variables in L2 performance, making thus the relationship between aptitude and age more opaque. This was demonstrated in a study by Flege (2008) who conducted a Principal Component Analysis after encountering serious multicollinearity issues between the three variables. The analysis revealed that age of onset loaded on a different factor than LoR and age, which loaded on the same factor. This means that the unique contributions of LoR and age at testing remain unclear. These issues have long highlighted the need for more sensitive and reliable measures of exposure than the traditionally established ones. Flege (1991), for example, proposed the Full Time Equivalent measure to estimate quantity of input instead of relying on LoR alone, and more recently Unsworth (2013) proposed an alternative measure of current and cumulative exposure that is more detailed and informative. In response to this issue, we have also derived an alternative measure of input, which is described below.

In the current study, we set out to explore the potential role of two predictors of L1 and L2 grammatical proficiency in L1 Greek-L2 English adult bilinguals: aptitude – more precisely, grammatical sensitivity – and input.

Grammatical proficiency was assessed by means of a grammaticality judgement task (GJT). Despite some criticism on the validity and reliability of GJTs (Platt, 1991; Devitt and Devitt, 2006; Tabatabaei and Dehghani, 2012; Orfitelli and Polinsky, 2017) many researchers still prefer them over other comprehension tasks (Linebarger et al., 1983; Van der Lely et al., 2011). More specifically, GJTs have been extensively used in both L2 (Plonsky et al., 2020) and L1 research (Ambridge, 2012, 2014), and have been employed in a large amount of research concerned with the Critical Period Hypothesis (e.g., Johnson and Newport, 1989; Flege et al., 1999; DeKeyser, 2000; Birdsong and Molis, 2001). This is because, in contrast to other grammar tasks, GJTs can test a wider variety of morphosyntactic structures with different degrees of difficulty and also different levels of association with prescriptive grammar instruction. Another advantage of GJTs is that they can target elements of “decorative” grammar, i.e., those aspects of a grammatical system that have low functional load (e.g.: grammatical morphemes such as agreement markers); these have been shown to be particularly challenging for L2 learners (White, 2003; Hopp, 2010; Dąbrowska et al., 2020).

Language aptitude was tested through the Sentence Pairs task (Llompart and Dąbrowska, in press), which is based on the Words in Sentences subtest of the MLAT and assesses grammatical sensitivity. Finally, as far as input is concerned, we computed a composite measure of exposure which was calculated based on the participants’ years of language use across different periods of their lifetime (for details see Method section). This measure of exposure was used in combination with LoR to test the sensitivity of the one against the other.

In addition, participants completed a Pseudoword Learning task which assessed explicit associative memory, i.e., an aspect of language aptitude which is most relevant for the acquisition of lexical knowledge. Including this task thus allows us to determine whether any effects of aptitude on grammar are component-specific (i.e., only grammatical sensitivity is relevant for grammar) or more general (other aspects of aptitude contribute to grammar).

Furthermore, age and education were included as covariates because it has been shown that they can influence grammatical performance in both L1 and L2 speakers. Dąbrowska (2019) found that age at testing showed a different pattern of correlations in L1 and L2 speakers: in L1 speakers, it was positively correlated with vocabulary and collocations; in L2 speakers, it was negatively correlated with grammar. With regard to education, it is a contributing factor in individual differences among both L1 (Chipere, 2003; Street and Dąbrowska, 2010; Llompart and Dąbrowska, 2020) and L2 speakers (Hakuta et al., 2003; Tarone and Hansen, 2007; Janko et al., 2019). Specifically, speakers with higher academic achievement typically perform better than those with low academic achievement, and the latter group also tends to present a much wider distribution of scores. Finally, education can be quite revealing of the quality of input that the participants have received, especially in the case of participants who started learning the L2 in instructional settings.

The aim of this study was to explore learner-internal and environmental factors that contribute to ultimate grammatical attainment in adult Greek-English bilinguals. Our main predictors were aptitude (and specifically, grammatical sensitivity), and exposure. Age, education, and explicit associative memory (as assessed by the Pseudoword Learning task) were treated as control variables. The present results are relevant in at least two major respects. First, there is a positive relationship between aptitude and proficiency in both the participants’ first and second language. Moreover, there is a strong relationship between exposure and grammatical proficiency in the second language. Below we will try to unravel the significance of these findings as well as their implications.

To sum up, the novelty of this study lies in the fact that it tested the same group of bilingual participants in their first and second language. This provided us with an ideal testing ground for the effects of linguistic and extralinguistic factors on grammatical knowledge. The results challenge the claim that language aptitude is only relevant for L2 learning and suggest that, in fact, it may be even more relevant to L1 acquisition. Last but not least, the current findings provide evidence that our measure of cumulative exposure can better account for individual differences in L2 grammatical proficiency than the measure routinely used in ultimate attainment studies, namely length or residence (LoR).

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Research Ethics Committees of the Friedrich Alexander University of Erlangen-Nuremberg and the University of Birmingham [ERN_16-0608AP23]. The patients/participants provided their written informed consent to participate in this study.

All authors designed the study and prepared the experimental tasks. LP collected and processed the data. ED and ML supervised and verified the analytical methods. All authors contributed to drafting and revising the paper and have approved the final version of the manuscript.

This project was funded by an Alexander von Humboldt Professorship (ID-1195918) awarded to ED.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.