The study of heritage languages has emerged as a separate subfield of linguistics over the past 15 years. Yet it is still riddled with uncertainties concerning fundamental concepts. In fact, the definition of ‘heritage language’ itself is controversial. The only uncontroversial point is that it is the weaker language of bilinguals whose stronger one is the societal language, recognized as a cultural heritage when in danger of being lost. Some restrict the term to immigrant languages, while others include indigenous languages; for some it is the first language of heritage language speakers, while for others it can also be a second language acquired in childhood or in adulthood.
Polinsky and Scontras (Polinsky & Scontras, Reference Polinsky and Scontras2019) deserve credit for having synthesized much of the relevant research in their draft of a model of heritage language competence: one I find plausible and in many aspects convincing. They define heritage language speakers as unbalanced bilinguals who acquired the heritage language as L1 (first language) and the societal language simultaneously or successively in childhood, and they identify features in which the heritage language conforms or deviates from the baseline. Divergent attainment, including ‘shrunk structures’, are claimed to be triggered by limitations of online resources and/or inadequate input. This, however, can only be part of the explanation, for such effects are not commonly detected in simultaneous bilinguals (2L1). Here I focus on the second point: the role of input on grammatical development. This is an issue of interest beyond heritage language, for it implies that children exposed to a language from birth can nevertheless fail to attain native competence.
Children acquire native competences by mere exposure to linguistic data, provided onset of acquisition happens early and access to adequate input is available. The first but not the second condition is met by heritage language learners, according to virtually all heritage language scholars – including Polinsky and Scontras, who state that bilingual input differs in quantity and quality from that of monolinguals and surmise that this explains divergent grammatical knowledge (1.1). Yet when is input inadequate? After all, there is ample evidence that simultaneous bilinguals are able to attain native competences, even in the weaker language of unbalanced bilinguals (Meisel, Reference Meisel2007). Development can be delayed, but it follows the developmental trajectory of monolingual learners (Polinsky & Scontras, 3.1.1) and does not result in divergent competences.
The quality of input that heritage language learners receive differs indeed from that of monolinguals in that the former interact with a smaller group of interlocutors. However, this is also true for family bilingualism where children are nevertheless able to develop two native competences. It is therefore more likely that quantitative differences are the cause of divergent development. Trivially, zero input results in acquisition failure. The challenge is to quantify the necessary minimum.
Heritage language research has not yet provided valuable information on this issue. I wonder why studies that view reduced input as the main cause of divergent attainment refer to the simplistic formula that bilinguals receive less input because there are only so many hours in a day. The math is clear, but it tells us nothing about input quantity – for hours spent with speakers of a language do not represent reliable information about quantity of exposure. Adults vary enormously in the amount of speech they produce interacting with children. Weisleder and Fernald (Reference Weisleder and Fernald2013) report that one (monolingual) toddler heard more than 12,000 words of child-directed speech, while another heard only 670 words during the entire day. Amount of exposure to each language of a bilingual can therefore be larger than that of a monolingual, if parents talk a lot to their children. De Houwer (Reference De Houwer, Grüter and Paradis2014) demonstrated that this is indeed the case in at least some families.
An even more serious reason to dispense with the 50% formula is that it blurs the distinction between exposure and input (Carroll, Reference Carroll2017). In order for linguistic data to qualify as input to the acquisition device, they need to be presented in child-directed speech; overheard speech will not do. Moreover, the relevant cues for a particular learning task must be available during developmental phases when the device is ready to process them. Consequently, reduced input affects acquisition differently (depending on when this happens in the course of development), and on the particular grammatical phenomenon. This explains why different grammatical phenomena are more or less sensitive to input quantity (Polinsky & Scontras, 3.1.1). We are dealing with a three-dimensional variation space, defined by the developmental schedule, particular grammatical features, and quantity and quality of input. Effects of reduced input on heritage language will therefore vary across heritage language learners and within individuals, with some vulnerable grammatical domains unaffected and others exhibiting divergent structures.
In sum, the question of how reduced input affects grammatical knowledge is still wide open. Exposure to a heritage language over a longer time may lead to more balanced bilingualism; but this concerns proficiency, not competence. Empirical research suggests that even when relative frequency of exposure amounts to no more than 30%, this need not lead to divergent attainment; see Meisel (Reference Meisel2019:175ff) for a summary.
No one denies that the nature of the input influences heritage language acquisition. It may even trigger structure shrinking. Currently, however, solid empirical evidence in support of this cause-effect relationship is lacking. Heritage language studies must provide quantified data, preferably based on analyses of ongoing acquisition processes, longitudinally or cross-sectionally. Self-reports on within-family language distribution in heritage language speakers’ long past childhood will not do. In other words, heritage language studies must meet the same methodological standards as L1 and L2 acquisition research. If they do, they will be in a unique position to determine what constitutes ‘adequate’ input. This would be a significant contribution to acquisition theory.
You have
Access
The study of heritage languages has emerged as a separate subfield of linguistics over the past 15 years. Yet it is still riddled with uncertainties concerning fundamental concepts. In fact, the definition of ‘heritage language’ itself is controversial. The only uncontroversial point is that it is the weaker language of bilinguals whose stronger one is the societal language, recognized as a cultural heritage when in danger of being lost. Some restrict the term to immigrant languages, while others include indigenous languages; for some it is the first language of heritage language speakers, while for others it can also be a second language acquired in childhood or in adulthood.
Polinsky and Scontras (Polinsky & Scontras, Reference Polinsky and Scontras2019) deserve credit for having synthesized much of the relevant research in their draft of a model of heritage language competence: one I find plausible and in many aspects convincing. They define heritage language speakers as unbalanced bilinguals who acquired the heritage language as L1 (first language) and the societal language simultaneously or successively in childhood, and they identify features in which the heritage language conforms or deviates from the baseline. Divergent attainment, including ‘shrunk structures’, are claimed to be triggered by limitations of online resources and/or inadequate input. This, however, can only be part of the explanation, for such effects are not commonly detected in simultaneous bilinguals (2L1). Here I focus on the second point: the role of input on grammatical development. This is an issue of interest beyond heritage language, for it implies that children exposed to a language from birth can nevertheless fail to attain native competence.
Children acquire native competences by mere exposure to linguistic data, provided onset of acquisition happens early and access to adequate input is available. The first but not the second condition is met by heritage language learners, according to virtually all heritage language scholars – including Polinsky and Scontras, who state that bilingual input differs in quantity and quality from that of monolinguals and surmise that this explains divergent grammatical knowledge (1.1). Yet when is input inadequate? After all, there is ample evidence that simultaneous bilinguals are able to attain native competences, even in the weaker language of unbalanced bilinguals (Meisel, Reference Meisel2007). Development can be delayed, but it follows the developmental trajectory of monolingual learners (Polinsky & Scontras, 3.1.1) and does not result in divergent competences.
The quality of input that heritage language learners receive differs indeed from that of monolinguals in that the former interact with a smaller group of interlocutors. However, this is also true for family bilingualism where children are nevertheless able to develop two native competences. It is therefore more likely that quantitative differences are the cause of divergent development. Trivially, zero input results in acquisition failure. The challenge is to quantify the necessary minimum.
Heritage language research has not yet provided valuable information on this issue. I wonder why studies that view reduced input as the main cause of divergent attainment refer to the simplistic formula that bilinguals receive less input because there are only so many hours in a day. The math is clear, but it tells us nothing about input quantity – for hours spent with speakers of a language do not represent reliable information about quantity of exposure. Adults vary enormously in the amount of speech they produce interacting with children. Weisleder and Fernald (Reference Weisleder and Fernald2013) report that one (monolingual) toddler heard more than 12,000 words of child-directed speech, while another heard only 670 words during the entire day. Amount of exposure to each language of a bilingual can therefore be larger than that of a monolingual, if parents talk a lot to their children. De Houwer (Reference De Houwer, Grüter and Paradis2014) demonstrated that this is indeed the case in at least some families.
An even more serious reason to dispense with the 50% formula is that it blurs the distinction between exposure and input (Carroll, Reference Carroll2017). In order for linguistic data to qualify as input to the acquisition device, they need to be presented in child-directed speech; overheard speech will not do. Moreover, the relevant cues for a particular learning task must be available during developmental phases when the device is ready to process them. Consequently, reduced input affects acquisition differently (depending on when this happens in the course of development), and on the particular grammatical phenomenon. This explains why different grammatical phenomena are more or less sensitive to input quantity (Polinsky & Scontras, 3.1.1). We are dealing with a three-dimensional variation space, defined by the developmental schedule, particular grammatical features, and quantity and quality of input. Effects of reduced input on heritage language will therefore vary across heritage language learners and within individuals, with some vulnerable grammatical domains unaffected and others exhibiting divergent structures.
In sum, the question of how reduced input affects grammatical knowledge is still wide open. Exposure to a heritage language over a longer time may lead to more balanced bilingualism; but this concerns proficiency, not competence. Empirical research suggests that even when relative frequency of exposure amounts to no more than 30%, this need not lead to divergent attainment; see Meisel (Reference Meisel2019:175ff) for a summary.
No one denies that the nature of the input influences heritage language acquisition. It may even trigger structure shrinking. Currently, however, solid empirical evidence in support of this cause-effect relationship is lacking. Heritage language studies must provide quantified data, preferably based on analyses of ongoing acquisition processes, longitudinally or cross-sectionally. Self-reports on within-family language distribution in heritage language speakers’ long past childhood will not do. In other words, heritage language studies must meet the same methodological standards as L1 and L2 acquisition research. If they do, they will be in a unique position to determine what constitutes ‘adequate’ input. This would be a significant contribution to acquisition theory.