Linguistic distance and crosslinguistic influence: Commentary

I Introduction

The articles in this issue offer different perspectives on the effects of linguistic distance and crosslinguistic influence on second language (L2) acquisition. I will address three issues that arise in this context:

II Definitions of linguistic distance

Regarding the first issue, there have been several different approaches to linguistic distance in the literature, as well as in the articles in this issue. Perhaps the best-known early perspective is that of Greenberg (1963), who made detailed typological comparisons between languages as a way of establishing relatedness, and similarities and differences between languages, as well as shared, universal properties. Included in the typological universals that Greenberg proposed is a series of generalizations known as implicational universals which take the following form: if a language has some property X, it will also have property Y, but not vice versa. Within a language, then, one property would be crucially related to another; such relationships would also generally hold across languages. These comparisons are descriptive and external, in the sense that it is linguists who determine the relatedness between languages and the degree of distance between them.

Typological/implicational universals have long been of interest to L2 researchers. Researchers like Eckman (1984) and Hawkins (1987) argued that learner grammars conform to such universals. A number of studies in the 1980s and 1990s explored whether Keenan and Comrie’s (1977) Noun Phrase Accessibility Hierarchy (NPAH), an implicational universal relating to the availability of different relative clause types, applies in L2 acquisition (e.g. Doughty, 1991; Eckman et al., 1988; Gass, 1982). These researchers argued that the NPAH was able to predict and explain the order of L2 relative clause acquisition; studies showed that relative clause types low on the hierarchy were precisely those that learners found hardest to acquire, regardless of the situation in the first language (L1) with respect to available relative clause types.

As several of the articles in this issue note, linguistic differences can be considered from a global perspective (language typology as a whole) or a local perspective (consideration of specific structures). The typological perspective on establishing distance (or lack thereof) by making macro comparisons is adopted by two of the articles in this issue, namely the article by Öksüz, Alexopoulou, Derkach and Tsimpli (2025) and the one by Van der Slik and Van Hout (2025). In both cases, a number of linguistic properties (lexical, phonological, morphological and syntactic) are used to measure the distance between L1 and L2. In addition, Öksüz et al. (2025) look at crosslinguistic differences in article availability (presence or absence of articles in L1). While their results show that having articles in the L1 was the strongest predictor of accuracy in the L2, there were also effects of typological distance, but only within languages with articles.

In contrast to such typological definitions, several of the articles define linguistic distance in terms of structural differences or similarities, which may be overt or abstract, as suggested by Castle, Jensen, Mitrofanova and Westergaard (2025). As far as structural definitions are concerned, there is disagreement as to whether structural distance between languages should be considered globally (along the lines of the typological approach) or on a case-by-case, more localized basis. Within the third language (L3) field, precisely such a distinction is maintained by two different theories. On the one hand, the Typological Primacy Model (TPM) (Rothman, 2011, 2015), advocated by Puig-Mayenco, Kubota, Naganawa and Merlo (2025), assumes that similarities or differences between languages are to be considered on a holistic basis. On the other hand, according to the Linguistic Proximity Model (Westergaard, 2021), L2 and L3 acquisition take place property by property, with structural distance determined at a local level, as discussed in the article by Castle et al (2025). The TPM is nevertheless tested by means of individual linguistic properties, making one wonder what the difference might be between these two positions. Puig-Mayenco et al. (2025) address these differences: languages that are typologically similar/related may nevertheless treat a particular property very differently, while languages that are typologically different may nevertheless treat some property in the same way. The two theories make different predictions as to what to expect in such circumstances.

A further issue relating to structurally-defined linguistic distance arises when there is micro-variation between languages, in other words, languages that are very close typologically but that show subtle differences between them in terms of how particular expressions are realized. Such cases are discussed by Ionin (2025) with respect to plural marking, and by Gil (2025) regarding wh-indefinites. We consider these below when discussing crosslinguistic influence.

III The relationship between linguistic distance and linguistic knowledge

It is interesting that interlanguages largely conform to typological universals, as shown, for example, in the earlier L2 studies on NPAH. However, it is not clear how this comes about, in other words, what the connection is between language typology and human cognition, and how we get from descriptions of what learners know to explanations of how they come to know it. Language learners are not language typologists, after all.

Assuming, then, that learners do not (usually) have explicit knowledge about language typology, how do language distance or implicational relationships between linguistic properties translate into unconscious linguistic knowledge? The learner is typically acquiring one additional language, not making comparisons across multiple languages. At least some of the researchers who adopt a typological perspective on L2 (Eckman, 1984; Hawkins, 1987) denied the possibility of ‘inbuilt’ universals – such as Universal Grammar (UG) – leaving it unclear as to how learners ‘know’ these universals or how they can guide language acquisition. (For discussion, see White, 1989, 2004; for related comments on the typological perspective, see Gregg, 1989.)

This question also applies to the macro approaches to language distance adopted in some of the articles in this issue, for example, by Van der Slik and van Hout (2025). How do external measurements of linguistic distance relate to the internalized grammar of the language learner? An early proposal by Kellerman (1978, 1979) addressed this issue. Kellerman argued that it is not actual language distance or relatedness that matters but rather distance as perceived by the learner, which he called ‘psycho-typology’. Actual relatedness (as determined by linguists) and perceived relatedness (as understood by a learner) might not coincide.

This raises the question of how one can assess what the learner’s perceptions of language distance are. Kellerman, working on idioms in L2 English, asked people whether translations of L1 Dutch idioms were/were not possible in L2 English. Where the two languages coincided in having the same idioms, Kellerman found that these were largely rejected. Learners perceived these forms as marked and, hence, unlikely to be transferable.

While this kind of post-hoc questioning of learners may work for idioms, it is not so clear that it would work to determine the psycho-typology of other grammatical properties, nor is it clear whether any predictions can be made, given that one cannot tell in advance what degree of relatedness an individual might perceive between languages. For example, in a grammaticality judgment task, L2 sentences based on L1 morphosyntactic properties are often included. If learners accept certain L1-like forms, does it mean that they perceive the two languages to be typologically related or simply that properties of the L1 grammar are represented in the interlanguage grammar, regardless of perceived relatedness or distance, as assumed by Schwartz and Sprouse (1996), for example?

The issue of typology versus psycho-typology has resurfaced in the context of the TPM, which argues that L3 acquisition is influenced by the typologically closer of the two prior languages (Rothman, 2011, 2015). Recognizing that actual typology might not necessarily be crucial, the definition of TPM was presented in terms of actual or perceived typological relatedness: ‘Syntactic properties of the closest (psycho)typological language, either the L1 or L2, constitute the initial state hypotheses in multilingualism’ (Rothman, 2011: 112). Again, this presupposes either that learners are somehow sensitive to language relatedness or that they have perceptions of relatedness that influence language acquisition. The post-hoc problem remains: how can one determine in advance which language might be perceived as psychotypologically closer? The article by Puig-Mayenco et al. (2025) shows transfer in the initial L3 stages to be from the language that is typologically closer to the L3 (actual closeness, rather than perceived), so from L2 English to L3 Spanish, rather than from L1 Japanese. The issue of psycho-typology is not addressed, so the question of how relatedness is established in the mind of the learner remains open. Similarly, Öksüz et al. (2025) specifically exclude psycho-typology (note 1), raising the question of how external typology translates into internalized knowledge. Given that these authors argue for a DP/NP division between the languages under consideration, it could be that UG-based knowledge of this distinction plays a role, but this is not addressed.

For researchers defining linguistic distance as structural distance rather than typological (or psycho-typological) distance, the ability to assess structural distance is seen as something internal to the learner, so necessarily implicated in grammar construction. For example, in the article by Ionin (2025), grammatical features are seen as properties of the internalized linguistic representation, in line with the Feature Reassembly Hypothesis (Lardiere, 2009). In the case of English plurals, assuming that learners start out with features available in the L1 (Japanese, Korean or Mandarin), some features have to be removed in order to arrive at an appropriate L2 representation.

This leads us to the remaining question, namely how structural distance as an internalized property affects crosslinguistic influence or transfer. This issue is explored in several of the articles, discussed in the next section.

IV The relationship between linguistic distance and crosslinguistic influence

Turning now to our final question, namely, the nature of the relationship between language distance and crosslinguistic influence, one can see opposing positions in the articles in this issue. On the one hand, there is what, intuitively, makes a lot of sense: the greater the distance between L1 and L2 (however defined), the harder the L2 is to acquire and/or the less accurate or proficient learners are, a position adopted by two of the articles here, namely Öksüz et al. (2025) and Van der Slik and Van Hout (2025).

Kellerman’s original proposal was, in fact, the opposite, although this is not always understood: the greater the perceived distance between L1 and L2, the lower the incidence of transfer. This is consistent with a position supported by several of the articles here, which argue for transfer in cases where L1 and L2 are typologically/structurally close rather than far apart. Gil’s article (Gil, 2025) provides such a case. She compares the acquisition of wh-indefinites in L2 Korean by speakers of Japanese and English. She shows that there is transfer from the closer L1, Japanese, which also has wh-indefinites with slightly different properties from those in Korean. In other words, microvariation between L1 and L2 ‘misleads’ learners into thinking the two languages are alike with respect to the property in question. In contrast, English-speakers, whose L1 is typologically distant from Korean, perform more accurately on Korean indefinites than Japanese speakers do. Assuming crosslinguistic influence, it is not surprising that similarity between L1 and L2 can be misleading rather than helpful. Sometimes, when the distance between L1 and L2 is great, it means that the learner comes with no preconceived ideas about how a particular linguistic property will work in the L2; this conveys an advantage.

Ionin’s article (Ionin, 2025) looks at a slightly different case of microvariation, considering situations where several different L1s differ minimally from each other and all are distant from the L2. In this case, all L1s under consideration (Korean, Japanese, Chinese) are classifier languages, unlike the L2 English. Classifier languages show different constraints on plural marking, with differing effects on plural marking in L2 acquisition, suggesting that these micro differences can be significant. At a more macro level, Puig-Mayenco et al. (2025) consider crosslinguistic influence to be more likely when languages are typologically closer, consistent with the findings reported here relating to micro-variation.

V Concluding remarks

One final point to be considered is how acquisition is determined. There is considerable variation in the articles in this issue as to what aspects of L2 are being assessed: unconscious knowledge (however manifested), accuracy, or proficiency. For some, proficiency is what is under consideration (for example, Van der Slik and Van Hout’s, 2025, assessment of performance using TOEFL scores). For others, it is accuracy that provides a measure of L2 acquisition (Öksüz et al., 2025). For yet others, it is unconscious properties of the grammar that are at issue (for L2, see Gil, 2025; Ionin, 2025; for L3, see Castle et al., 2025; Puig-Mayenco et al., 2025).

It is important to bear in mind that proficiency/accuracy are not the same as acquisition. Recall the old debate about what arbitrary percentage of correct production should count as showing that a particular morpheme has been acquired, 90% for L1 acquisition (Brown, 1973) or 60% for L2 (Vainikka and Young-Scholten, 1996). In fact, the important issue is not learner accuracy as such but rather the patterns that learners produce (see Grimshaw and Rosen, 1990). Do learners show a distinction between grammatical and ungrammatical forms, for example? Along these lines, both Ionin (2025) and Gil (2025) consider acquisition in terms of patterns of performance within the structures in question, rather than in terms of proficiency or general accuracy.

In conclusion, it is still an open question as to whether it is better or worse (from an acquisition point of view) for the L1 and L2 to be distant from each other. While actual distance (as determined by external measures) does seem to contribute statistically (some of the time) to L2 acquisition difficulties, when properties are considered at a more local level, it is much less clear that linguistic distance necessarily creates a disadvantage. Furthermore, it is still unclear (to me, at least) why typological distance should confer a disadvantage, or even an advantage, given that assessment of actual distance is beyond the control of the learner. It remains to be determined how external relatedness can be reflected in properties of the internalized grammar of the L2 speaker. This can only come about by a consideration of how the learner, rather than the linguist, analyses language.