Polinsky and Scontras (Polinsky & Scontras, Reference Polinsky and Scontras2019) highlight the intrinsic interest of heritage languages (HLs), and provide compelling arguments that the input to heritage language learners and online processing limitations operate together to explain heritage language properties. They conclude by noting that studying HLs can be connected more closely to theoretical discussions of vulnerable and robust aspects of language in mutually informative ways. Here we focus on another connection, with the study of inter- and intra-speaker variability. We suggest that HL-like scenarios may arise when learners are acquiring variable phenomena in non-heritage languages, a possibility that provides a foundation for crossover between these research areas.
Polinsky and Scontras distinguish three factors influencing heritage language scenarios: (1) how the baseline linguistic properties of the input compare to the language as spoken in non-heritage communities; (2) the number of model speakers providing input; and (3) the sheer quantity of input. While these are in principle independent, they co-occur in the typical heritage language scenario, making it unclear whether particular heritage language properties arise due to (1), (2), (3), or some interaction thereof. Thus, an initial step towards understanding the causal relations in heritage scenarios would be to isolate (1–3) in monolingual situations in which (some of) the complexities and confounds produced by unbalanced bilingualism can be avoided. With a different profile of effects of (1–3), it may become possible to disentangle their contributions. We focus our discussion on factor (2), a smaller number of input speakers.
Polinsky and Scontras observe that, relative to the typical language acquisition scenario, heritage speakers are exposed to increased instability in the input. We take the key idea to be that, with a limited number of speakers providing input, talker-specific idiosyncrasies might exert a disproportionate influence on heritage language learning. This point resonates with experimental work showing that variability can be a useful learning cue: input from multiple talkers has been shown to promote infant word learning, adult learning of L2 contrasts and words, and comprehension of foreign accented speech (e.g., Bradlow and Bent, Reference Bradlow and Bent2008). Similarly, individuals’ social network size has been shown to influence their performance on various linguistic tasks (e.g., Lev-Ari, Reference Lev-Ari2018), and interactions with strangers outside of a speaker's social network can have an additive effect on L2 competence (e.g., Dewey, Belnap & Hillstrom, Reference Dewey, Belnap and Hillstrom2013). Explanations for these effects typically invoke the learning of gradient distributions, and are therefore quite limited in scope, as they do not address the structural properties of language. But alternative explanations appeal to the depth of processing induced by interaction with many different people (e.g., Barcroft and Sommers, Reference Barcroft and Sommers2005), allowing for the possibility that factor (2) could play an important role not only in the pronunciation of heritage languages, but in their structural properties as well.
This line of work suggests that (2) can be isolated from (1) and (3) experimentally; what about in naturalistic settings? Variation in all levels of linguistic analysis is the norm, such that learners are exposed to different variants in the typical (i.e., non-heritage) acquisition scenario. We believe that under certain circumstances, monolingual speakers are exposed to variable aspects of their language in ways that share essential properties with heritage language acquisition due to the context involving factors (1–3). To the extent that these factors are weighted differently in the variable scenario, a comparison of their effects with learning outcomes in heritage language would shed light on the independent contribution of each factor.
Consider situations in which adults move across dialect boundaries or multiple dialects are in contact within a community. The children of these adults, acquiring features that are variable within or across those dialects, face a challenge that parallels factor (2) in heritage acquisition: they get input on those features from only a subpart of the larger community. Crucially, in a monolingual scenario there is not a restricted baseline (1) overall, nor are there the challenges introduced by unbalanced bilingual acquisition. Separating (2) and (3) may prove harder. However, there are some reasons to think that comparisons with monolingual variation might be fruitful. While there are aspects of the sociological context of HLs that produce tight connections between these factors, in the variable scenarios under consideration all of the input is in a single language. The monolingual nature of the input may weaken the effects of (3) relative to what happens in bilingual scenarios. For some directions in the literature that connect with these points, we would look to work on (i) native acquisition of variation in ethnolectally diverse contexts (e.g., Sharma and Sankaran, Reference Sharma and Sankaran2011); (ii) second dialect acquisition after mid-childhood immigration (e.g., Chambers, Reference Chambers1992); and (iii) acquisition of complex local dialect features by children whose parents speak a different dialect (e.g., Payne, Reference Payne and Labov1980).
If our suggestions are on the right track, heritage language and variationist theories can be mutually informative. In one direction, heritage language theories make predictions about which aspects of variation are likely to be vulnerable in scenarios that manifest a small speaker community (2). In the other direction, studying (2) in variation allows us to see the effects of this factor independent of (1) and (3), in a relatively restricted linguistic context, and without the complexities of unbalanced bilingual acquisition.
You have
Access
Polinsky and Scontras (Polinsky & Scontras, Reference Polinsky and Scontras2019) highlight the intrinsic interest of heritage languages (HLs), and provide compelling arguments that the input to heritage language learners and online processing limitations operate together to explain heritage language properties. They conclude by noting that studying HLs can be connected more closely to theoretical discussions of vulnerable and robust aspects of language in mutually informative ways. Here we focus on another connection, with the study of inter- and intra-speaker variability. We suggest that HL-like scenarios may arise when learners are acquiring variable phenomena in non-heritage languages, a possibility that provides a foundation for crossover between these research areas.
Polinsky and Scontras distinguish three factors influencing heritage language scenarios: (1) how the baseline linguistic properties of the input compare to the language as spoken in non-heritage communities; (2) the number of model speakers providing input; and (3) the sheer quantity of input. While these are in principle independent, they co-occur in the typical heritage language scenario, making it unclear whether particular heritage language properties arise due to (1), (2), (3), or some interaction thereof. Thus, an initial step towards understanding the causal relations in heritage scenarios would be to isolate (1–3) in monolingual situations in which (some of) the complexities and confounds produced by unbalanced bilingualism can be avoided. With a different profile of effects of (1–3), it may become possible to disentangle their contributions. We focus our discussion on factor (2), a smaller number of input speakers.
Polinsky and Scontras observe that, relative to the typical language acquisition scenario, heritage speakers are exposed to increased instability in the input. We take the key idea to be that, with a limited number of speakers providing input, talker-specific idiosyncrasies might exert a disproportionate influence on heritage language learning. This point resonates with experimental work showing that variability can be a useful learning cue: input from multiple talkers has been shown to promote infant word learning, adult learning of L2 contrasts and words, and comprehension of foreign accented speech (e.g., Bradlow and Bent, Reference Bradlow and Bent2008). Similarly, individuals’ social network size has been shown to influence their performance on various linguistic tasks (e.g., Lev-Ari, Reference Lev-Ari2018), and interactions with strangers outside of a speaker's social network can have an additive effect on L2 competence (e.g., Dewey, Belnap & Hillstrom, Reference Dewey, Belnap and Hillstrom2013). Explanations for these effects typically invoke the learning of gradient distributions, and are therefore quite limited in scope, as they do not address the structural properties of language. But alternative explanations appeal to the depth of processing induced by interaction with many different people (e.g., Barcroft and Sommers, Reference Barcroft and Sommers2005), allowing for the possibility that factor (2) could play an important role not only in the pronunciation of heritage languages, but in their structural properties as well.
This line of work suggests that (2) can be isolated from (1) and (3) experimentally; what about in naturalistic settings? Variation in all levels of linguistic analysis is the norm, such that learners are exposed to different variants in the typical (i.e., non-heritage) acquisition scenario. We believe that under certain circumstances, monolingual speakers are exposed to variable aspects of their language in ways that share essential properties with heritage language acquisition due to the context involving factors (1–3). To the extent that these factors are weighted differently in the variable scenario, a comparison of their effects with learning outcomes in heritage language would shed light on the independent contribution of each factor.
Consider situations in which adults move across dialect boundaries or multiple dialects are in contact within a community. The children of these adults, acquiring features that are variable within or across those dialects, face a challenge that parallels factor (2) in heritage acquisition: they get input on those features from only a subpart of the larger community. Crucially, in a monolingual scenario there is not a restricted baseline (1) overall, nor are there the challenges introduced by unbalanced bilingual acquisition. Separating (2) and (3) may prove harder. However, there are some reasons to think that comparisons with monolingual variation might be fruitful. While there are aspects of the sociological context of HLs that produce tight connections between these factors, in the variable scenarios under consideration all of the input is in a single language. The monolingual nature of the input may weaken the effects of (3) relative to what happens in bilingual scenarios. For some directions in the literature that connect with these points, we would look to work on (i) native acquisition of variation in ethnolectally diverse contexts (e.g., Sharma and Sankaran, Reference Sharma and Sankaran2011); (ii) second dialect acquisition after mid-childhood immigration (e.g., Chambers, Reference Chambers1992); and (iii) acquisition of complex local dialect features by children whose parents speak a different dialect (e.g., Payne, Reference Payne and Labov1980).
If our suggestions are on the right track, heritage language and variationist theories can be mutually informative. In one direction, heritage language theories make predictions about which aspects of variation are likely to be vulnerable in scenarios that manifest a small speaker community (2). In the other direction, studying (2) in variation allows us to see the effects of this factor independent of (1) and (3), in a relatively restricted linguistic context, and without the complexities of unbalanced bilingual acquisition.
Author ORCIDs
David Embick, 0000-0002-6697-4428.