Linguistic distance and crosslinguistic influence in the L2 and L3 acquisition of morphosyntax

Abstract

In this introduction to the special collection on linguistic distance and crosslinguistic influence in second and third language acquisition, we contextualize the articles of the collection in current debates. We propose a three-tiered framework for typological comparison: individual feature similarity, parameter-setting similarity, and grammar-level linguistic distance. We review key theoretical debates in generative second language acquisition (SLA) regarding whether second language (L2) acquisition entails parameter resetting or feature (re)assembly, highlighting ongoing difficulties with functional morphology despite successful acquisition of syntactic word order. We connect recent large-scale empirical studies showing robust effects of linguistic distance on L2 outcomes with smaller experimental studies examining specific morphosyntactic features. The articles in the collection address four key questions: whether broad typological effects can be distinguished from feature-level effects; whether L1-L2/3 typological similarity consistently facilitates acquisition; how linguistic similarity should be measured; and how prior knowledge effects vary across proficiency levels. After introducing each article, we synthesize findings showing that broad typological effects often override individual feature similarities in determining transfer patterns. In contrast, linguistic proximity can sometimes impede acquisition when micro-variation requires costly ‘restructuring’ of existing first language (L1) representations rather than the development of a new item in L2. Large-scale studies reveal that lexical and morphological distance measures are more predictive than phonological or syntactic ones, with distance effects persisting across proficiency levels while, at the same time, shifting in relative importance during acquisition. We propose that considering multiple levels of typological granularity is essential for understanding crosslinguistic influence in multilingual development.

I Introduction

Since early studies on contrastive analysis, similarities and differences between first language (L1) and second language (L2) have been crucial to our understanding of L1 influence on L2. ¹ Across theoretical frameworks the availability of L1–L2 congruent items is seen as the basis for transfer to L2. Within the generative approach, crosslinguistic similarity is captured in terms of parameters, clustering of abstract features on functional heads. There is now a large body of research across the cognitive/functionalist and generative perspectives demonstrating a facilitative/positive effect of transfer in L2 ² but at the same time identifying cases where similarity does not appear to help acquisition. ³

L1–L2 similarity is not always or necessarily the basis of transfer. Early on, it was established that learners appear to transfer properties of the L1 to their L2 even when the L2 properties differ from those of the L1. For example, French learners of English seem to incorrectly assume that adverbs can appear postverbally in English, as they do in French (White, 1990, 1991). Such negative transfer ⁴ leads to errors and is generally thought to impede rather than facilitate acquisition since, in order to acquire the relevant L2 properties, learners need to adapt some aspect of the transferred property. Importantly, positive and negative transfer might co-exist. For example, cognates might facilitate the acquisition of vocabulary for a Spanish learner of English, but, at the same time, alongside the cognate lexical item, e.g. information, the count feature specification of L1 Spanish information might also transfer, making it harder to acquire that, in English, information is a mass noun. Similarly, Romance learners of English might benefit from the availability of the definite article in their L1s but, alongside the similar and helpful features they might also transfer features that are not part of the L2 English article, e.g. features relating to the use of definites to refer to kinds (Ionin and Montrul, 2010).

Recently, the discussion of the role of crosslinguistic influence and transfer has been extended to third language (L3) acquisition. Under some influential proposals, the typological similarity between the L3 and the two prior languages, L1 and L2, can determine the source of transfer to L3, that is, whether the L3 learner will draw from the L1 or the L2 or both (Rothman et al., 2019; Slabakova, 2017; Westergaard, 2021).

Against this background, recent empirical work employing big data has provided compelling empirical evidence that the linguistic distance in the sense of similarity between the L1 and an additional language can influence broad acquisition outcomes. For example, Schepens et al. (2020) analysed speaking proficiency scores of 50,000 learners of Dutch from 62 L1s. They found that a learner’s L1 accounts for 28%–69% of variance in scores in a model controlling for education, age of acquisition and length of residency in the Netherlands. Importantly, 80% of the variance due to L1 was captured by a combination of phonological, morphological and lexical measures of linguistic similarity/distance. Van der Slik et al. (2019), based on the proficiency scores of 9,000 adult learners of Dutch from 33 L1s, show that learners with L1s of lower morphological complexity than Dutch make less progress over time in their learning, while Schepens et al. (2023) show that linguistic similarity mitigates against age-related decline in adult language learning. A key element of these studies is the use of big data from assessment, which enables empirical studies with a wide range of typologically diverse languages and can therefore allow us to identify effects that could not be identified in small crosslinguistic studies.

Our goal in this collection is to take a step in connecting findings from these large-scale empirical studies with insights from small-scale and experimental studies targeting the impact of fine-grained L1–Ln variation on Ln (additional language acquisition) learning. Both approaches provide crucial information on transfer processes and language learning. We suggest that considering the role of typological similarity and linguistic distance in depth can provide a way to understand how the large-scale effects might connect to more qualitative insights from smaller studies. Understanding this connection will illuminate the nature and consequences of crosslinguistic influence and may have concrete didactic implications for the increasingly multilingual classroom.

Below, we first contextualize the collection in current theoretical debates and then consider notions of typological similarity before turning to the questions raised in this collection and the articles.

II Theoretical background: Transfer and learnability in SLA theory

The role of prior linguistic knowledge in additional language acquisition (Ln) is central to generative SLA theory. The existence of a full-fledged grammar at the initial state of L2 is a crucial difference from monolingual first language acquisition. Similarly, the presence of an L1 and an L2 is a crucial property of the initial state of L3. The study of transfer is, thus, inseparable from the question of learnability in L2/Ln both in terms of the path followed by learners as well as ultimate attainment and the learnability of individual features.

Accounts of transfer rely on an understanding of crosslinguistic variation, which, within generative syntax, is captured as variation in parameter settings, that is, variation in the abstract features associated with functional categories. In effect, a significant part of acquiring an L2 grammar amounts to identifying (and acquiring) the features which vary from prior grammars and the relevant functional categories. Several aspects of this process are currently under debate. One question is whether this process involves parameter re-setting or feature (re)-assembly (Hawkins, 2019; Lardiere, 1998; Schwartz and Sprouse, 1996; Slabakova, 2016; White, 1985b). The question arises because of the apparent dissociation between syntax and morphology in L2 grammars. For example, acquiring a new setting of the head parameter does not involve the acquisition of new features: all an English learner of Japanese has to do is (re)-set the directionality of the head parameter for Japanese as head-final, thus ‘switching’ from an SVO language to SOV. Indeed, this kind of parameter setting appears to be successfully acquired from early stages of development (Vainikka and Young-Scholten, 1996; VanPatten and Smith, 2015). However, when L2 parameter settings involve features unavailable in the L1, e.g. agreement tense features on the verb or gender agreement features in the nominal domain, there is persistent optionality in the suppliance of agreement morphology until late stages of acquisition. This difficulty with verbal agreement morphology appears to contrast with learners’ relative ease in acquiring orders arising from V-to-T raising in Spanish (VanPatten, Keating and Leeser, 2012). ⁵ Similarly, L1 Turkish and L1 Moroccan Arabic learners of L2 Dutch appear to acquire the word order of possessor nominals relatively well and early but find aspects of the morphology and case marking more challenging (van de Craats et al., 2000). The dissociation between syntax and morphology has been captured by Slabakova’s Morphological Bottleneck Hypothesis, which identifies functional morphology as the main acquisitional challenge for L2 learners (Slabakova, 2009).

The source of this acquisitional challenge is currently an open question. The Feature Assembly hypothesis (Lardiere, 2009) takes the dissociation of syntax and morphology as evidence that, rather than parameter resetting, L2 acquisition involves mapping of individual features on functional projections and morphological functional items, modulated by the availability of the relevant features in the L1. The process might be one of feature assembly, when new features might be selected from a UG inventory of features, or re-assembly where features available in the L1 may need to be re-organized in the L2 grammar. The challenge for the L2 learner, therefore, involves deciphering the form–meaning mappings that provide cues to the re-assembly of abstract features. Note that under this approach, there is no distinction between semantic and formal morphosyntactic features (e.g. animacy or person features on pronouns vs. person features on verbal morphology). Alternatively, the difficulty learners show with functional morphology is attributed to the hypothesis that formal features are not available to L2 learners if not activated in the L1, either in their entirety (Failed Functional Features Hypothesis, Hawkins and Chan (1997)) or just those that are not interpretable at the interfaces (Interpretability Hypothesis, Mastropavlou, 2017; Tsimpli and Dimitrakopoulou, 2007; Tsimpli and Stavrakaki, 1999). The Missing Surface Inflection Hypothesis postulates that acquisition of abstract L2 representations is unproblematic; rather, the observed optionality is due to difficulty accessing the morphological realizations of the relevant abstract features (Prévost and White, 2000). van de Craats et al. (2000) suggest word order variations are more salient in the input, unlike morphology, so that they are easier for learners, who generally are inclined to conserve their L1 representations in their L2 and resist restructuring of morphological/functional items.

While the articles in this collection do not focus on these theoretical debates, the level of typological comparison and assumptions regarding the features involved in crosslinguistic variation bears directly on the theoretical approach adopted and the conclusions that can be drawn regarding crosslinguistic influence, as we will see shortly. We consider typological similarity in the next section.

III Typological similarity between L1 and L2/L3

We now turn to the question of defining typological similarity between L1 and L2/L3. Assuming a universal inventory of features, typological similarity can be broadly defined in terms of shared features between two grammars. Importantly, the level of granularity at which typological comparisons are considered will impact on theoretical accounts and conclusions. For current purposes, we can distinguish three levels of typological comparison: (1) individual feature similarity, (2) parameter-setting similarity and (3) grammar level similarity or linguistic distance.

Individual feature similarity involves the availability of a (congruent) individual feature in the L1 or L2 and differences or similarities in the properties of the shared feature: e.g. whether the shared feature is lexicalized by a similar functional item (e.g. tense affix, question particle) and associated with a corresponding functional category, whether it is overt or covert, whether it is obligatory or optional and, if the latter, what might be constraining optionality. These possibilities provide a variety of L1–L2 combinations as illustrated by the studies included in this collection. ⁶ For example, as Ionin (2026) explains in her article, while plural marking is obligatory in English, it is optional in Korean, Japanese and Mandarin, where its presence interacts with animacy and definiteness. Gil (2026) contrasts constituent questions in English, which lexicalize a wh/Q feature through a series of dedicated wh-pronouns, with Korean and Japanese. In these languages, constituent questions involve in-situ wh-indefinites, which can also function as existential quantifiers. Case marking might be expressed morphologically in a consistent manner on all nominal elements (nouns, adjectives/determiners) as in Greek, or it might only appear on the noun, as in Polish, or not at all as in Norwegian. Castle et al. (2026) focus on the implications of such variation in L3. Puig-Mayenco et al. (2026) discuss English negative words like nobody and nothing which may not co-occur with negation (not). In comparison, Japanese negative words like dare-mo (‘nobody’) and nani-mo (‘nothing’) require verbal negation (see English anyone and anything). Spanish instantiates both options: nada (‘nothing’) is compatible with negation, patterning with Japanese, while preverbal nadie (‘nobody’) patterns with English in a paradigm sensitive to word order. Finally, languages may lexicalize definiteness through a definite and indefinite article, as we find in many European languages, contrasting with articleless languages like Russian, Korean or Japanese, a type of variation considered in Öksüz et al. (2026).

The case studies above illustrate the variety of acquisitional scenarios that emerge through different L1–L2(–L3) combinations varying in individual features. However, individual features are not randomly distributed across the world’s languages. Rather, typologists have drawn several patterns of correlations in the distribution of individual features, for example, distinguishing languages with articles from articleless languages (Dryer, 2013), patterns of occurrence of plural marking and their relation with noun categorizing devices (Aikhenvald, 2017; Haspelmath, 2013), patterns of optional and obligatory double negation (Dryer, 2013), etc. Crucially, since Greenberg (1966), systematic correlations in the distribution of individual features have been captured through implicational universals. Within generative syntax, parameters have sought to provide explanations for typological generalizations through connecting surface morphological expressions to variation in abstract features. For example, the wh-parameter captures a cluster of properties of constituent questions in English-type languages (wh-pronoun, movement, person features on T and D) which contrasts with Korean and Japanese-type languages (indefinite variable, no-movement, question particle, absence of person features on T/D; see, for example, Huang and Roberts, 2016; Roberts, 2019). Similarly, within the typology of nominal systems, Japanese, Korean and Mandarin instantiate a distinct type, Generalized Classifier languages (for discussion, see Ionin, 2026) contrasting with obligatory number languages like English; the contrast between the two types has been linked to independent properties of the nominal system and accounted for by parameters proposed by, amongst others, Chierchia (1998), Longobardi (1994) and Bošković (2008). Chierchia’s nominal parameter and Longobardi’s account further predict that articles and a D head are only expected in obligatory number languages, not in Generalized Classifier languages. The acquisitional question arising then is whether the generalizations obtained through the parametric approach to crosslinguistic variation bear on the L2 acquisition, that is, whether learners show evidence of acquiring clusters of properties rather than (or in addition to) individual feature assembly (notwithstanding the observed dissociation between syntactic and morphological manifestations of a single parameter). ⁷

Further variation might be observed within languages with the same parameter setting. Roberts (2019) aims to capture such variation in terms of hierarchies of parameters (macro, meso, micro and nano), thus, distinguishing macro-parametric variation, capturing broad typological variation from meso and micro variation, capturing finer-grained variation within the same broad language type (see Huang and Roberts, 2016). For example, the classic wh-parameter is a macro-parameter broadly distinguishing English from Korean and Japanese. At the same time, a nano-parameter might be relevant for the contrast between Korean and Japanese wh-indefinites, whose bare form can be used for constituent questions in Korean but not in Japanese (for discussion, see Gil, 2026). Similarly, a macro-parameter can distinguish Generalized Classifier languages from obligatory number languages like English, while micro-variation could be assumed to capture the way animacy and definiteness features interact with plural marking in Japanese, Korean and Mandarin, as in Ionin (2026).

Finally, turning to linguistic distance, the typological literature has been concerned with measuring linguistic distance as a method to establish phylogenetic relations between languages (see review in Longobardi and Guardiano, 2009). The dominant approach relies on the analysis of the number of cognates shared between a pair of languages and the degree of phonological and orthographic differences between cognates in a pair of languages e.g. on the basis of Levenshtein distances (Schepens et al., 2012; Shatz et al., 2023). Such lexico-statistical approaches have been successfully exploited for automated measurement of language distance and have been complemented by other approaches, e.g. word distributions or n-grams (Gamallo et al., 2017), as well as semantic features (Bella et al., 2021). In other words, linguistic distance does not provide us with an account of variation, but, rather, with surface or aggregate measures quantifying the degree of variation between two languages.

In the context of SLA, measures of linguistic distance have been used as measures of L1–Ln similarity to investigate broader typological effects on L2/L3 learning (Schepens et al., 2020; Shatz et al., 2023, 2024), in effect offering a way to quantify crosslinguistic influence on L2/L3. ⁸ Further, measures of linguistic distance allow us to consider the relative effect of subdomains of the linguistic system such as morphology (van der Slik et al., 2019) or phonology (Schepens et al., 2020) on crosslinguistic influence. For instance, Schepens et al. (2020) show that when considering phonological distance, it is subcategorial properties (i.e. phonological features) that have the highest impact on L3 attainment. In addition, they show that a combination of measures of lexical, morphological and phonological distance is necessary to adequately capture the contribution of L1 to L3 Dutch. In the generative context, since Longobardi and Guardiano’s influential article (Longobardi and Guardiano, 2009), generative parameters have been used to measure syntactic distance in the nominal domain (Ceolin et al., 2020), an approach recently extended by Baker and Roberts (2021) beyond the nominal domain.

With this background in place, we can return to the role of typological similarity between the grammars involved in L2/L3 acquisition. Considering typological similarity at three separate levels of granularity allows us to distinguish the role of broader typological effects captured by linguistic distance from variation related to parameters and individual features. For example, Russian is closer to English than Arabic in terms of linguistic distance; however, Russian lacks articles, unlike Arabic, which has a definite article similar to English. In other words, Arabic has the same parametric setting as English, being a +DP language, while it is typologically more distant from English as a Semitic language; conversely, Russian is a –DP language, typologically closer to English as an Indo-European language. Looking at variation at the individual level in this set of languages, we find that Arabic is similar to English in lexicalizing the definite article but differs from English in lacking an indefinite article (though both are +DP languages). By contrast, Turkish, a –DP language, presents the opposite paradigm; it has an indefinite article (based on the numeral) but no definite article. Öksüz et al. (2026) aim to clarify the potential effect of similarity at these three distinct levels: linguistic distance, parameter setting and individual features.

Turning to the use of wh-phrases in constituent questions, discussed by Gil (2026), L2 Korean is closer to L1 Japanese than L1 English, both in terms of linguistic distance and in terms of macro-parametric variation, since both Japanese and Korean are wh-in-situ languages. However, there is micro-variation between Korean and Japanese regarding the distribution of wh-indefinites, raising the question of whether the macro-parametric and typological similarity between Korean and Japanese facilitates or impedes the acquisition of micro-variation in connection to the Korean bare wh-indefinite variant for Japanese learners.

Negation provides a further case of interplay between linguistic distance and micro-variation considered by Puig-Mayenco et al. (2026). Specifically, while Spanish is closer to English in terms of linguistic distance, there is more overlap between Spanish and Japanese in the realization of negative concord when it comes to individual negative words, raising the question of whether L1 Japanese learners of L3 Spanish will rely more on their L2 English or L1 Japanese.

In Ionin (2026), Korean, Japanese and Mandarin are of roughly similar linguistic distance from English, at least in relation to the nominal domain, as they have many similar parameter settings. At the same time, the three Asian languages show micro-variation in the distribution of the plurality marker and its interaction with animacy and definiteness. ⁹ The question then is whether such micro-variation will influence the acquisition of the plurality feature in a typologically distant language.

Finally, Castle et al. (2026) investigate different levels of linguistic similarity (structural vs. lexical) for crosslinguistic influence in L3. In their study, Polish and Norwegian represent different types of similarity to two L3 artificial languages, Aliensk A and Aliensk N, both using Norwegian as a lexifier. Polish has overt case marking, while Norwegian lacks case marking. The question is whether structural similarity in case-marking systems (overt case-marking in Polish) can facilitate L3 acquisition even when lexical similarity (Norwegian vocabulary) is present, potentially triggering transfer from caseless Norwegian. The study further tests micro-variation in case-marking distribution using the two artificial languages: ‘Aliensk N’ has Case on nouns (overt + abstract similarity to Polish) and ‘Aliensk A’ marks case on the articles (abstract similarity only to Polish which has case on nouns but has no articles).

The typological variation explored in the empirical studies of this special collection is our springboard for considering the different ways typological distance between the L1 and L2/L3 can affect the L2/L3 acquisition.

IV The collection: Questions and contributions

The aim of this collection is to consider the role of typological similarity at different levels of granularity aiming to shed further light on the way prior linguistic knowledge influences second and third language acquisition. In particular, we aim to address the following broad questions:

• Question 1: Can we distinguish effects of broad typological classifications (based on measures of linguistic distance) from similarities/differences at the level of individual features or parameters?

• Question 2: Does similarity always facilitate acquisition? Is dissimilarity always a challenge? Are features unavailable in L1 always challenging in the L2/L3?

• Question 3: How do we measure linguistic similarity? Can we build on syntactic typology (e.g. Longobardi and Guardiano (2009)) to obtain measures of morphosyntactic distance that can provide testable predictions for Ln acquisition? How do syntactic distance measures compare with measures of lexical, phonological and morphological distance?

• Question 4: What is the effect of prior linguistic knowledge across the L2 trajectory, and how might it interact with proficiency and amount of exposure?

Below, we first provide short introductions to the articles of this collection and return at the end of the section with an attempt to synthesize the answers to the above questions as emerging from this collection.

Ionin (2026) reviews a number of recent studies on the L2 acquisition of English plural marking by Mandarin, Japanese, and Korean speakers, finding that micro-variation within typologically similar L1 languages, namely General Classifier Languages, produces distinct transfer effects. For example, Korean learners of English outperformed Mandarin learners of English in plural contexts with demonstratives due to Korean requiring plural marking in definite contexts. At the same time, both Korean and Mandarin learners were equally successful in indefinite contexts. Similarity does not necessarily lead to facilitation – even favourable L1 features posed challenges when optional in L1 but obligatory in L2. Individual feature specifications ([+human], [+animate], [+definite]) would appear to matter more than broad typological classifications. Proficiency critically modulates transfer – in some studies, only advanced learners showed sensitivity regardless of L1. The discussion compares the predictions of the Feature Reassembly Hypothesis against the Morphological Congruency Hypothesis and finds more supporting evidence for a feature-based account.

Gil (2026) challenges the assumption that typological similarity facilitates L2 acquisition, showing that typological proximity in L1–L2 pairs can hinder acquisition compared to typologically distant language pairs. Japanese learners of Korean wh-indefinites and Chinese learners of English experientials underperformed compared to more typologically distant groups (L1-English–L2-Korean, L1-Korean–L2-English, respectively). Gil argues that micro-variation within typologically close languages involves ‘restructuring’, which requires feature reassembly of existing L1 structures. By contrast, learners in typologically distant language pairs face ‘developmental’ tasks which require feature assembly and are less challenging. Typological closeness becomes problematic when subtle differences exist within languages that share broad typological properties.

Staying with L2, Öksüz et al. (2026) present a large-scale corpus study of the accuracy in the use of the English article by 104,541 learners from 11 typologically diverse L1s. The study demonstrates that both item-level typological similarity (presence/absence of definite articles or positively set DP parameter) as well as broader typological similarity captured by linguistic distance measures independently predict L2 accuracy. The availability of a definite article in L1 was the strongest predictor, supporting the DP/NP language distinction, while Levenshtein lexical distance scores also significantly predicted accuracy beyond individual feature effects. Crucially, L1 influence persisted across proficiency levels, with learners from +DP languages showing greater improvement over time. However, similarity is not consistently shown to always facilitate acquisition – Chinese learners, though lacking a definite article, performed well initially but showed minimal improvement with proficiency, suggesting complex interactions between abstract syntactic features and overt morphological realizations in crosslinguistic influence.

Van der Slik and van Hout (2026) present a longitudinal study of 15+ million TOEFL scores (2005–20) from 127 countries, demonstrating that broad typological distance measures significantly predict L2 English proficiency. Three distance measures – lexical, morphological, and phonological – each contributed independently to explaining country-level scores. Crucially, linguistic distance effects proved dynamic across proficiency development: the effect of lexical distance decreased over time while morphological distance became increasingly influential at higher proficiency levels. This suggests that structural properties become more critical as learners advance, while vocabulary similarities matter more initially. The study argues for multiple distance measures as complementary predictors, showing that broad typological classifications and specific linguistic features operate through different mechanisms across the L2 acquisition trajectory, with dominance relationships shifting as exposure increases.

Turning to L3, Castle et al. (2026) use artificial languages as L3s to investigate whether structural similarity from one of the previously acquired languages can facilitate L3 acquisition even when the L3 is lexically similar to the other previously acquired language. Polish–Norwegian speakers (who also knew English) learned one of the two Norwegian-lexified artificial languages with case marking either on nouns (overtly similar to Polish) or articles (abstractly similar only). In comparison, Norwegian does have not case. Results were compared to a group of Norwegian speakers who did not know Polish and showed that Polish speakers outperformed Norwegian speakers only when case appeared on nouns, demonstrating that overt morphological similarity matters more than abstract structural similarity at early acquisition stages. Moreover, the results showed that factors such as proficiency, SVO bias, and lexical activation strength also modulate structural transfer effects. The findings support property-by-property co-activation models over a model assuming wholesale transfer, showing that individual morphosyntactic features can transfer independently. Importantly, the study shows that for linguistic properties such as case, overt similarity is necessary at early stages of acquisition, as abstract structural parallels alone provide insufficient facilitation.

Puig-Mayenco et al. (2026) examine Japanese–English bilinguals acquiring Spanish as L3, focusing on negative concord items to distinguish broad typological effects from feature-level similarities. Despite Japanese and Spanish both being negative concord languages (feature-level similarity), learners initially transferred English patterns (broad typological similarity), suggesting that holistic structural factors override specific linguistic features. Similarity does not always facilitate acquisition; learners showed both target-like and non-target-like development, with some restructuring toward Japanese-like patterns over time. Linguistic experience (the amount and depth of exposure) is also shown to modulate development rates, supporting the Cumulative Input Threshold Hypothesis. Lower L2 English use facilitated faster restructuring, indicating that entrenchment affects acquisition trajectories. The study demonstrates complex interactions between typological distance, feature similarity, and experiential factors in multilingual development.

In her commentary, White (2026) reviews investigations of the role of typology in SLA, reconsidering the seminal works of Kellerman (1983), Gass (1982) and Eckman (1984). She raises three key issues: how linguistic distance is defined, how external measures of distance translate into learners’ internalized knowledge, and the relationship between distance and transfer. White critiques the disconnect between external typological measures and learners’ internalized knowledge, questioning how linguists’ distance assessments translate into acquisition patterns. She highlights the unresolved issues with Kellerman’s psychotypology, particularly the inability to predict learners’ perceived language relatedness. She emphasizes the importance of examining acquisition patterns rather than mere proficiency measures, concluding that the fundamental issue is not whether similarity helps or hinders, but, rather, how external typological measures can meaningfully relate to learners’ internalized grammars.

Let us return to our questions, starting with question 1: can we distinguish effects of broad typological classifications based on linguistic distance measures from similarities/differences at the level of parameters and individual features?

The first observation is that, when we consider linguistic distance in broad samples, we can identify a clear and robust facilitative effect of linguistic distance. This is shown in broad outcomes like exam scores, where linguistic distance predicts lower proficiency scores as shown in van der Slik and van Hout (2026). However, to tease apart the effect of (global) linguistic distance from the effect of specific parameters and individual features, we need to consider effects on the acquisition of individual features as opposed to broad outcomes. Two studies in this collection attempt this, Puig-Mayenco et al. and Öksüz et al. As shown in Puig-Mayenco et al.’s article, L1 Japanese learners of L3 Spanish seem to draw from L2 English rather than L1 Japanese, at least at the initial stages of their L3 acquisition of Spanish negative polarity items. This is despite the fact that there is more overlap between Japanese and Spanish in the co-occurrence of individual negative words with negation. ¹⁰ It, therefore, seems that the broader typological similarity between English and Spanish guides transfer, although it is unclear whether this relates to parameters around licensing of quantifiers and negative polarity or broader macro-parametric differences. ¹¹

Turning to the acquisition of the L2 articles, measures of lexical and morphosyntactic distance/proximity capturing language family effects are linked with accuracy in the acquisition of the L2 English article, as demonstrated in Öksüz et al.’s study. However, this effect of language family is restricted within the group of +DP L1s. Importantly, the parametric setting (+/–DP) overrides linguistic distance: for example, L1 Russian learners (–DP) are less accurate than L1 Arabic learners (+DP). ¹² The parameter setting also overrides individual features: the accuracy of L1 Arabic learners, a +DP language lacking an indefinite article, matches the accuracy of learners from other +DP languages, while the accuracy of L1 Turkish learners, a –DP language which nevertheless has an indefinite article, matches the accuracy of –DP languages.

Taken together, the studies by Puig-Mayenco et al. and Öksüz et al. suggest that broader typological variation guides L2/L3 acquisition, with language family/linguistic distance effects also playing a role. Broad typological effects may override individual features, e.g. the overlap between Spanish and Japanese negative concord or the availability of indefinite articles in Arabic and Turkish. In turn, broader typological effects can be overridden by parameter settings (e.g. +/–DP). There are some further open questions. In Öksüz et al., language family effects only appear within +DP languages, raising the possibility that such effects are due to micro-variation within the +DP languages (e.g. in relation to generic and kind nominals), rather than global linguistic distance effects.

Further on question 1, the role of micro-variation is explored in more detail in the articles by Ionin, Gil and Castle et.al. Gil shows that, at least when it comes to specific features of micro-variation linked to wh-indefinites and experiential constructions, the broader typological similarity between L1 and L2 can be misleading and impede the acquisition of individual features involved in micro-variation. Thus, Japanese learners of Korean have difficulty dis-assembling features from specific functional elements. As a result, they are outperformed by English learners of Korean. In relation to question 1, such results suggest that the broader typological similarity between Japanese and Korean takes precedence over micro-variation between the two grammars. ¹³ Gil, though, proposes an alternative explanation. She suggests that the development of a new feature/structure is easier than restructuring a feature/structure already available in the L1. While the conclusion is undeniable in view of her results, it is worth considering the scope of this generalization. Gil’s findings are reminiscent of Ionin and Montrul (2010) who found that Korean learners of English were more successful acquiring that kinds in English involve bare nouns (tigers are wild) in comparison to Spanish learners, because of the association of kind and generic readings with the definite article in Spanish. ¹⁴

However, when the broader picture is considered, acquisition of new features (development) is more challenging than micro-variation, e.g. when overall acquisition of articles is considered, Spanish learners are more accurate than Koreans as shown in Öksüz et al.’s study. ¹⁵ It is indeed possible that English learners outperform Japanese learners of Korean in their overall accuracy of wh-indefinites, so that while development is overall challenging for L1 English learners, reassembly is challenging for specific features of micro-variation. If this analysis is correct, it would suggest that L1 Japanese learners have potentially transferred more detailed representations from their L1 and do not pay enough attention to smaller differences. Clearly, more research is needed in this area to better understand the interplay between development and re-assembly at different levels of typological variation.

A striking conclusion in Ionin’s review is that, while sensitivity to L2 plural marking appears inconsistent and fragile across the various L1 groups and studies, we do not see the strong transfer effects observed in the other studies. As Ionin points out, this could be because the reviewed studies have not targeted transfer per se. Another possibility is that there is a qualitative difference between plural marking in English and the General Classifier Languages, namely that plural marking in GCL is semantic, interacting with other semantic features, while English it is morphosyntactic, a difference that might be linked to macro-variation between the two types of languages. GCL learners, therefore, need to acquire another typological setting which somehow reduces the impact of the micro-variation found in the different L1s considered. This latter possibility would tie in with the findings of Puig-Mayenco et al. and, to a certain extent, Gil’s review, which show that similarity in macro-variation seems to take precedence over similarity in micro-variation.

In their study, Castle et al. consider aspects of variation that pertain not to the inventory of features per se, but rather their externalization. Specifically, whether prior knowledge of case marking on nouns can facilitate the acquisition of case marking on the article (when case is not marked on the noun). The study confirms an earlier conclusion that abstract case marking in the L1 is not sufficient to facilitate the acquisition of case marking in L3 and that some level of similarity on the exponent of case marking, i.e. superficial similarity in the authors’ terminology, is required for a facilitative effect in L3. At the same time, the acquisition of case marking appears to interact with other learning biases in the encoding grammatical roles (e.g. SVO word order, a potential preference for case marking on the article). Though the study does not speak directly to the question of linguistic distance and macro- or micro-variation, it highlights the interplay between crosslinguistic influence and more general learning biases.

Let us now turn to our second question, whether similarity always facilitates acquisition and conversely, whether dissimilarity is always a challenge, potentially meaning that features unavailable in L1 are always challenging in the L2/L3. Puig-Mayenco et. al. and Gil’s findings suggest that linguistic proximity, indicating grammar level similarity (e.g. English and Spanish vs. Japanese and Korean) as well as similarity regarding broad classifications around a particular area of grammar (e.g. wh-movement vs. wh-in-situ languages), typically linked to parameters, does not facilitate acquisition of more specific structures (e.g. negative quantifiers in L3 Spanish) or aspects of micro-variation within a given parameter setting (wh-indefinites within wh-in-situ languages). In fact, it seems that linguistic proximity and similarity in parameter settings is misleading for the acquisition of the negative quantifiers considered by Puig-Mayenco et al. and the micro-variation in the distribution of wh-indefinites considered by Gil. Similarly, in the case of articles, the potentially facilitative effect of linguistic proximity is overridden by the settings of the DP parameter (see L1 Russian vs. L1 Arab learners), that is, a positive setting of the DP parameter matters more than grammar level similarity as captured by linguistic distance. However, within languages with the facilitative +DP setting, linguistic proximity has an additional positive effect on accuracy in the use of the article. It, therefore, seems that there is an interplay between broader and finer grained L1–Ln similarity with the facilitative effect of broader similarity weakening when more specific structures are considered. Crucially, though, Puig-Mayenco et.al. and Gil’s results suggest that broader typological similarity does guide transfer. ¹⁶

Just as similarity is not always facilitative, dissimilarity, is not always challenging; indeed L1 English learners are more successful than L1 Japanese learners with L2 Korean wh-indefinites. Importantly, Ionin shows that features of macro-variation can be acquired, potentially reducing the influence of L1 micro-variation.

As suggested in White’s commentary (White, 2026), understanding when typological similarity facilitates acquisition and when it doesn’t and when differences prove challenging and when not is critical to our understanding of how prior linguistic knowledge impacts on the acquisition of subsequent grammars. To obtain a more definitive answer to this question we seem to require more empirical research investigating groups of languages showing variation at various levels of granularity within a single design, and potentially also considering a variety of morphosyntactic and semantic features, as it is conceivable that variation anchored to the properties of a specific functional category that might not be universally available (e.g. articles-DP, tense-TP, case, etc.) might show differences in acquisition when compared with variation linked to universally available categories (e.g. negative quantifiers, constituent questions, etc.).

Question 3 considers measures of linguistic similarity. Van der Slik and Van Hout (2026) and Öksüz et al. (2026) explore several measures of linguistic distance and evaluate how well they correlate with exam outcomes and accuracy in the use of the articles. Lexical and morphological measures appear to be the strongest predictors in the relevant models, while phonological and syntactic measures do less well. Rather surprisingly, when it comes to syntactic measures, Baker and Roberts’ measure, which is based on a wide range of parameters, is more predictive of accuracy with L2 English articles than Ceolin et al.’s measure that considers only parameters within the nominal domain.

Our final question targets the effect of prior linguistic knowledge across the L2 trajectory and potential interactions with proficiency and the amount of language exposure. Across all studies, the dominant pattern is that learners progress with proficiency, so that proficiency is the strongest predictor of accuracy and of the various performance measures adopted in individual studies. At the same time, it is important to note that, with proficiency, learners might move from strong L1 transfer to perhaps more indeterminate structures and optionality. At least this could be one interpretation of the results of Puig-Mayenco et al. Importantly, while everyone learns and progresses, there are residual effects of L1 even at advanced levels. This is most evident in Öksüz et al. where 11 L1s are compared. This study also reveals a different trajectory for learners from +DP and –DP L1s: learners from +DP L1s show a linear progression in their accuracy, while learners from –DP L1s show a progression that approximates a U-shape curve, suggesting a different acquisition trajectory.

We believe that the articles of this collection demonstrate the need to consider typological similarity at various levels of granularity: grammar level, as captured by linguistic distance measures, similarity in relation to typological classifications regarding specific areas of grammar, typically captured through parameters within the generative tradition, and variation around individual features. To do this, current theoretical accounts of the distinction between macro- and micro-variation would need to be incorporated to current L2 learnability hypotheses such as Feature Re-assembly, Failed Functional Features Hypothesis, Interpretability Hypothesis, Missing Surface Inflection Hypothesis, etc. Similarly, current debates around L1/L2 transfer, whether it involves wholesale transfer of prior grammar(s) or property-by-property transfer, will benefit from a clearer empirical understanding of the interplay between macro- and micro-variation in relation to crosslinguistic influence and transfer.