Prosodic constraints on children’s use of grammatical morphemes

Abstract

It has long been known that children may use a particular grammatical morpheme inconsistently at early stages of acquisition. Although this has often been thought to be evidence of incomplete syntactic representations, there is now a large body of crosslinguistic evidence showing that much of this early within-speaker variability is due to still developing phonological and/or prosodic representations. This article reviews recent research showing how a phonological approach to the emergence of grammatical morphology makes it possible to make crosslinguistic predictions about what will be acquired early and what acquired late. This is known as the Prosodic Licensing Hypothesis, providing a unified framework for understanding the course of morphosyntactic development across languages. The implications are both theoretical and methodological, suggesting that children may know more about the grammar of their language at an earlier age than is often assumed, and that this can only be revealed by taking prosodic phonology into account in designing early tests of syntactic development.

Keywords

Determiners grammatical morphemes inflectional morphemes prosodic hierarchy prosodic morphology

Introduction

Much of the field of language acquisition owes its beginning to observations of children’s early speech. Diary studies as early as Darwin (1877) began to provide early insight into the course of language development, and how a child’s grammar changes over time. Once speech could be captured on audio recorders, and larger sets of longitudinal data collected and transcribed, early researchers like Brown (1973) could begin to systematically examine changes in children’s language – including the acquisition of grammatical morphemes, and how these developed over time. Using Mean Length of Utterance, or MLU, the field began to have a measure of a child’s grammatical abilities, providing a tool for comparing language development across children. Brown’s (1973) early study of Adam, Eve, and Sarah thus provided a baseline for the acquisition of grammatical morphemes in English, showing that, though some children are faster at acquiring language than others, the overall patterns of development are largely the same. These findings suggested that some grammatical morphemes are learned earlier than others, and that this might be true crosslinguistically. Brown (1973) suggested that the semantic transparency of grammatical morphemes would help determine which would be acquired first, with plural morphemes acquired before tense/agreement morphemes – at least in English.

The later acquisition of tense/agreement marking led Radford (1990) to propose that this was due to syntactic deficits, where children initially lacked the syntactic structures needed to support tense/agreement marking. This view was then formalized by Wexler (1994) and colleagues in terms of the Optional Infinitive Hypothesis, where children’s lack of access to syntactic processes left them occasionally producing verbal inflection, and occasionally not. This suggested a ‘random’ approach to the use of grammatical morphemes that became stable only later, after syntactic abilities had matured.

However, several independent developments in the early 1990s began to lead to a different explanation for children’s variable use of early grammatical morphemes. Using corpus data from the spontaneous speech productions of three 2–3-year-olds learning the morphologically rich Bantu language Sesotho, Demuth (1992a, 1994) showed that children’s early use of singular/plural noun class prefixes was prosodically conditioned. Thus, noun class prefixes tended to be omitted when preceding a disyllabic nominal stem (e.g., mo-sadi ‘woman’), but were likely to be produced when preceding a monosyllabic nominal stem (e.g., mo-tho ‘person’), resulting in words that took the form of a disyllabic foot. This strongly suggested that the variable use of these nominal inflections was not random, but principled, and that these principled regularities were phonological, not syntactic or semantic.

Around the same time, converging evidence for this approach was emerging from a set of experiments with children learning English. Using an elicited imitation experiment, Gerken (1994, 1996) showed that the use of object articles in children aged 2;3 was prosodically conditioned, with articles more likely to be produced when they could be prosodified as part of a (strong-weak) trochaic foot. Thus, when preceded by a monosyllabic verb, where it was prosodified as part of a foot (Tom [likes the]_Ft [doggie]_Ft), the article tended to be produced, whereas it tended to be omitted when following a disyllabic verb, where it was left unfooted (Tom’s [watching]_Ft the [doggie]_Ft). Note that in the first example, the article is prosodified to the left, along with the preceding monosyllabic word [likes the]_Ft. However, in the second example [watching]_Ft the [doggie]_Ft, where the disyllabic word watching precedes the article, the article ends up being prosodified with the following word – at a higher level of prosodic structure. Gerken (1996) proposed that children’s pattern of development was due to their limited prosodic representations, where unfooted syllables could only be produced once children had access to higher level of the Intonational Phrase (IP). This raises the interesting issue of the nature of children’s early prosodic organization, how this develops over time, and the implications for understanding prosodic constraints on children’s realization of grammatical morphemes.

The Prosodic Hierarchy is shown in (1) (Nespor & Vogel, 1986; Selkirk, 1984), providing a framework for understanding the prosodic organization of children’s early utterances.

(1) The Prosodic Hierarchy

Utt (Phonological Utterance) I think Sue likes bananas

IP (Intonational Phrase) Sue likes bananas

PP (Phonological Phrase) likes bananas

PW (Phonological Word) bananas

Ft (Foot) nanas

σ (Syllable) nas

µ (Mora) na

Much is known about prosodic development at the lower levels of prosodic structure (e.g., at the level of the PW, which contains a foot (for example, see Demuth, 1995, 1996b, 1996c; Fikkert, 1994; Gerken, 1996). Much less is known about acquisition at the higher levels of prosodic structure. However, Snow (1994) showed that, even by the age of 2, English-speaking children have begun to use phrase-final lengthening and declining F0, suggesting at least an IP boundary. Branigan (1979) also suggested that, even at the one- and two-word stages of development, children’s early PWs are IPs. Frota (2012) and Frota, Cruz, Matos, and Vigário (2016) also propose an unfolding of higher levels of prosodic structure, which becomes gradually more articulated (see also Demuth, 1996c). All this research suggests that young children are operating at multiple levels of the Prosodic Hierarchy simultaneously, developing more complex, fuller prosodic structures over time.

Along with this prosodic development comes the ability to produce grammatically more complex structures. This includes the production of grammatical morphemes of various types. The prosodic structure of the morpheme itself, and the prosodic context in which it occurs, both help determine how a given grammatical morpheme will be prosodified in a given language. Selkirk (1996) again provides a useful typology for how grammatical function words may be prosodified across languages. This is shown in (2).

(2) The prosodic structure of grammatical function items

Given the typology shown in (2), Lleó and Demuth (1999) and Demuth (2001b) suggested that much of the variability found in children’s early production of grammatical morphemes might be conditioned by phonological/prosodic factors rather than syntax alone. If so, this has important implications for our understanding of syntactic development. If experiments designed to tap children’s knowledge of syntax included stimulus sentences from both prosodically footed (easy, earlier acquired) and prosodically unfooted (harder, later acquired) contexts, this could easily lead researchers to think that children’s use of grammatical morphology was ‘random’, providing support for theories that suggest the later acquisition of syntactic competence. To our knowledge, very few researchers examining syntactic development today take phonological/prosodic factors into account when designing syntactic experiments and/or the analysis of spontaneous speech data. This is largely due to the persisting trend of both phonological and syntactic theory, and practitioners, taking a largely modular approach to their domain of study. This article argues that some cross-talk between these disciplines is long overdue.

Some understanding of both (prosodic) phonology – and acoustic phonetics – is also critical for addressing how and when grammatical morphemes begin to appear in the speech of children with typical language development, and why the same patterns may be protracted in older children with language disorders/delay. Many grammatical morphemes (e.g., articles, plurals, some verbal inflections, auxiliaries) begin to appear before the age of 2 for typically-developing children, when their segmental inventory (e.g., Smit, 1993) and prosodic word structure (Demuth, 2001a) is not yet well formed. This is also an age at which perception experiments can be conducted if production experiments are not yet possible.

The development of longitudinal acquisition corpora, and the addition of the tools needed to computerize, annotate, and store the data (e.g., the CHILDES database; MacWhinney, 2000), has also provided a wealth of information regarding children’s language development. This has been greatly augmented by the addition of resources for analyzing children’s speech, including access to the acoustic and video files, so that researchers can actually hear what the children said, and see on a spectrogram if there might be acoustic ‘traces’ of a morpheme (cf. Theodore, Demuth, & Shattuck-Hufnagel, 2011). This now allows us to examine more closely the course of development with respect of grammatical function words, both across different children learning the same language, as well as crosslinguistically. Of course, being able to predict which grammatical morphemes in a given language will be acquired earlier or later depends on knowing about how these are prosodically realized within a given language. This may mean that syntacticians need to team up with phonologists who know, for example, about the Prosodic Hierarchy, and how grammatical morphemes are prosodified in a particular language. This can then help inform future research regarding the various factors that influence the course of grammatical development in a particular language, and if these are truly syntactic, or are (also) influenced by phonological/prosodic, frequency, or other factors.

The remainder of this article is organized around three case studies examining children’s acquisition of grammatical function words. The first explores in more depth the acquisition of noun class prefixes in Sesotho, showing how children gradually relax their prosodic constraints, and also learn that some prefixes can be ‘optionally’ dropped at the intersection of certain phonological, syntactic and discourse/pragmatic conditions. The second reports on a series of studies examining the emergence of articles/determiners in English, French, and Spanish, showing both crosslinguistic differences in the course of acquisition, but also individual differences within each language. The third reviews another series of studies examining children’s acquisition of 3^rd person singular inflectional grammatical morphemes, showing interesting differences in acquisition as a function of allomorph type. All point to the important role of phonological/prosodic factors in understanding the course of morphosyntactic development.

Study 1: The acquisition of Bantu noun class prefixes

As noted above, Sesotho-speaking children around the age of 2 tend to omit noun class prefixes, producing forms like ba-sadi ‘women’ as bare stems, e.g., sadi. Once the Sesotho Corpus (Demuth, 1992b) was computerized and contributed to the CHILDES database, it was possible to exhaustively assess where and when (i.e., in which prosodic contexts) children used noun class prefixes, and where they did not. The data were drawn from spontaneous speech productions during interactions with family members (mothers, grandmothers, siblings) from three children between the ages of 2 and 3, constituting about 30 hours of speech per child. Demuth and Ellis (2009) found that, as anticipated, the children were much more likely to use noun class prefixes before monosyllabic nominal stems (e.g., ba-tho ‘people’) compared to disyllabic nominal stems (e.g., ba-sadi ‘women’). That is, their PWs were truncated to a disyllabic foot.

Note that Sesotho, and most Bantu languages, do not have lexical stress. Rather, they exhibit lengthening of the penultimate syllable at the end of a phrase, resulting in the second to last syllable of a phrase-final word being twice as long as other syllables (cf. Doke & Mofokeng, 1985, p. 41). Since Bantu languages are also null-subject languages, most nouns occur utterance-finally, with a long-short (prominent-less prominent) foot of prosodic structure on the last two syllables, regardless of morphological structure. This suggests that children’s early words are truncated to a strong-weak trochaic foot (sadi for ba-sadi), much as English-speaking children truncate words with initial unstressed syllables (e.g., banana > nana) (see Demuth, 2014 for a review). The Sesotho-speaking children then go through a ‘filler syllable’ stage of development (cf. Peters, 1983), producing part of the prefix (e.g., a-sadi ‘women’) around 2;6 years, and finally full target forms by around the age of 3 (ba-sadi).

In the process of this investigation two questions were raised. First, before full acquisition, do children really know which noun class a truncated noun belongs to? Sesotho has 17 noun class (gender) prefixes, and other Bantu languages have more (cf. Demuth, 2003), marking both singular and plural forms of a noun; perhaps this takes a long time to learn, so the prefix is initially omitted. One way to test this was to look at children’s use of modifiers, which, like Spanish and many other languages, must inflect for the gender/number of the noun. Critically, most modifiers young children use contain a monosyllabic agreement prefix, and this attaches to a monosyllabic modifier stem (e.g., li-khoho tsa-ka ‘chickens my’ my chickens). Although 2–3-year-olds do not use many nominal modifiers, there were enough in the corpus to examine this issue. Demuth and Weschler (2012) therefore conducted an exhaustive search of children’s possessive, numeral and adjectival modifiers, and found no errors. Thus, children produced forms like sadi ba-ne for ba-sadi ba-ne ‘those women’, demonstrating excellent knowledge of the noun class to which the noun belonged, even while producing a bare nominal stem. This provides strong evidence that Sesotho-speaking children have excellent knowledge of the gender/number of nouns even when morphological marking on the noun itself is absent.

Thus, it appears that Sesotho-speaking children’s early PWs take the shape of a disyllabic foot. Function words are initially produced only as an ‘internal clitic’ as part of a foot, as in (2c) above. Around the age of 2;6–3;0, these children begin to produce more complex, trisyllabic words, with ‘affixal clitics’ like that in (2d). This suggests that children’s early prosodic grammars adhere to the Strick Layer Hypothesis (Nespor & Vogel, 1986; Selkirk, 1984, 1996), where the Exhaustivity constraint ensures that a prosodic constituent must immediately dominate only constituents of the next level down in the Prosodic Hierarchy. The function word representations in (2b) and (2d) above violate this constraint. Gerken (1996) and Demuth (1996b, 1996c) both proposed that function words are omitted in children’s early speech due to a ‘lack of access’ to higher level prosodic structures, consistent with the Exhaustivity constraint. Demuth (1996a) also proposed that children’s early words are ‘unmarked’, exhibiting simple PW structures. All are consistent with Frota et al.’s (2016) proposal for early ‘phrase match’, where higher and lower level prosodic boundaries are aligned.

In conducting the Sesotho study we also examine adults’ use of noun class prefixes. It had been noted that adults sometimes omitted noun class prefixes as well, and it had originally been thought that this was a type of ‘motherese’ used in child-directed speech. Certainly it is common for disyllabic truncated words to occur in the ‘baby talk’ of various languages in the world (cf. Turpin, Demuth, & Campbell, 2014 for a recent review). Demuth, Machobane, and Moloi (2009) therefore examined the Sesotho corpus again to determine the contexts under which adults omitted noun class prefixes. They found that Sesotho grammar does permit null noun class prefixes, but only at the intersection of the following three phonological, syntactic, and discourse/pragmatic conditions: (1) the noun class begins with a coronal consonant (e.g., le-, but not ba-), (2) the noun is in a c-command position (e.g., followed by a modifier/agreement form), and (3) the referent is salient in the discourse context (visually, or formerly mentioned). Thus, utterances such as …tsatsi le-na… (for le-tsatsi le-na) ‘this day’ are well formed, but utterances such as …*sadi ba-ne… (for ba-sadi ba-ne) ‘those women’ are not. Adults only produce null noun class prefixes of the first type, whereas children at 2 years produce both. By 3 years, however, children have relaxed their prosodic constraints, and only omit noun class prefixes when they occur at the union of these three conditions. This is a very intriguing case of early acquisition at the intersection of phonology, syntax, and discourse, and one that might be thought to be a significant challenge for learners. However, it is not: one child showed phonological overgeneralization around 2;6 years, another showed syntactic generalization around the same age, and the third child (the least verbal but most syntactically advanced) showed adult-like forms from the beginning of her recordings at 2;4 years. In line with Szagun and Schramm (2016), this suggests that aspects of morphosyntactic competence may correlate with children’s knowledge of syntax.

In sum, the acquisition of Sesotho noun class prefixes shows early sensitivity to both prosodic (and other) constraints, where these grammatical morphemes appear early when they can be prosodified as part of a disyllabic foot, but are omitted when they cannot. This pattern is repeated in the acquisition of articles across many languages, as shown below.

Study 2: Acquisition of articles/determiners in English, French, and Spanish

Recall that Gerken (1994, 1996) showed that English-speaking children tend to produce object articles more reliably when these can be prosodified as part of a foot with the previous monosyllabic verb (e.g., Tom [likes the]_Ft [doggie]_Ft). Since these results came from an elicited imitation experiment, we wanted to know if the same would be found in children’s spontaneous speech. The answer was yes. The Providence Corpus is composed of longitudinal, biweekly 1 hour-long audio/video recordings of spontaneous speech interactions between six mother–child dyads from the ages of 1 to 3, totaling around 360 hours of speech (Demuth, Culbertson, & Alter, 2006). Using transcribed (and audio) data from the five children who produced a sufficient number of articles, Demuth and McCullough (2009) found that they showed the same effects as reported by Gerken (1994, 1996), tending to produce articles first when these occurred in a context where they could be prosodified as part of a disyllabic trochaic foot, and only later starting to produce unfooted articles. In keeping with Gerken’s (1996) proposals, this behavior can be explained by early function words being prosodified as part of a disyllabic foot, as in (2c). Prosodifying function words at the higher level of the PW or IP, in the form of an unfooted syllable (2d), appears later in development. Once again, the acquisition of function words is closely linked to developments in prosodic phonology, whereby the strict domination of feet by PWs is eventually relaxed.

Given the very consistent findings of prosodic constraints on English-speaking children’s early use of articles from both experimental and corpus data, the next question to ask was how articles are acquired in a prosodically different language. Lleó and Demuth (1999) noted that Spanish-speaking children acquired determiners earlier than German-speaking children, and this was thought to be due to the different prosodification of determiners in these languages. Note from the discussion above that English articles can prosodify to either the right or left, depending on the prosodic context. Monosyllabic articles in Spanish prosodify to the right, along with the following noun, forming one PW ([la [casa]_Ft]_PW ‘the house’), or possibly a more complex unit at a higher level of prosodic structure if the noun consists of more than two syllables/a foot. In contrast, articles in German constitute a foot/PW, and can stand on their own (e.g., the structure in (2a) ([das]_Ft/PW [haus]_Ft/PW ‘the house’) (see Hall, 1999 for proposals that these ultimately attach at the level of the PP). Lleó and Demuth (1999), in a series of longitudinal case studies, showed that Spanish-speaking children started to produce articles several months before their German-speaking peers. It was hypothesized that this was due to the fact that the function word plus noun constituted two PWs for German, but just one PW in Spanish, being realized as either an internal clitic (2c) or an affixal clitic (2d). See Vigário (2003) for proposals for the latter structure in European Portuguese.

The prosodic structure of the language, and the prosodic form of the particular lexical item and the grammatical morpheme itself, thus all interact to help determine when a particular function word will appear in children’s speech. For example, French, with phrase-final lengthening, or (iambic) ‘stress’ (Goad & Buckley, 2006; Scullen, 1997), contrasts prosodically with English, which tends to have lexical, trochaic stress (cf. Hayes, 1995). Like Spanish, French articles also prosodify with the following noun. Demuth and Tremblay (2008) examined the spontaneous speech development of articles (and other determiners) longitudinally using data from two of the children from the Lyon Corpus (Demuth & Tremblay, 2008) between the ages of 1;6 and 2;6. Similar to Sesotho, they found that articles tended to occur first with monosyllabic nouns (e.g., le nez ‘the nose’), and a few months later with disyllabic nouns (e.g., les poubelles ‘the garbage’).

In French, as also in English and Sesotho, some children tend to use phonologically underspecified ‘filler’ syllables as precursors to segmentally more well-formed determiners. Veneziano and Sinclair (2000) noted also that these filler syllables occurred more often before monosyllabic nouns from the age of 1;8–1;10 in the French-speaking child they studied, with well-formed articles emerging around 2;3 years. Bassano, Maillochon, and Mottet (2008) have also found prosodic (and semantic) conditioning on the use of early French articles. Thus, despite a very different prosodic system, French-speaking children’s first use of articles also appears to be prosodically licensed as part of a (disyllabic) foot.

Recall that, like French, Spanish articles prosodify with the following noun. Spanish also has default trochaic (strong-weak) lexical stress (Harris, 1983). We would therefore predict that Spanish would also show early cases of disyllabic PWs, where the first articles would be used with monosyllabic words – just as was found for both Sesotho and French. However, Spanish has very few monosyllabic words: unlike French, most Spanish words are disyllabic, with about a third of the words a child hears containing three syllables or more (cf. Roark & Demuth, 2000). Preliminary evidence from three children suggested that Spanish-speaking children’s first articles might appear with disyllabic words (Lleó & Demuth, 1999), and further evidence from another child found that when articles first begin to be produced with three-syllable target words, the lexical item is truncated to two syllables (e.g., la muñeca [aˈmweka] ‘the doll’; Gennari & Demuth, 1997). In order to examine these issues more fully, longitudinal data were needed covering the potential time period for the emergence of Spanish articles – about 1;5–2;5 years. Longitudinal, phonologically transcribed data from two children were found and analyzed, supporting these preliminary findings, but with very interesting individual differences emerging (Demuth, Patrolia, Song, & Masapollo, 2012). The girl, who was more precocious, with an MLU of 3 at age 1;8, truncated early words to include the article, initially producing forms such as a beza for la cabeza ‘the head’ before full (untruncated) forms began to be produced a few months later. In contrast, the boy, who was much slower to develop, with an MLU of only 1.7 at 1;8 years, initially produced only morphologically simple three-syllable PWs such as caballo ‘horse’, with no articles. His articles only began to appear around 2;1–2;3 years, with morphologically complex PWs beginning to appear with no truncations of the noun (un caballo ‘a horse’).

There are two intriguing observations to be made from these Spanish data. First, these children are producing much larger three-syllable PWs compared to their Sesotho-, English- and French-speaking peers before the age of 2. Although the Spanish-speaking children also pass through a brief month or so of truncating PWs to a disyllabic trochaic foot, they quickly start producing larger 3-syllable PWs with an unfooted syllable, either truncated to include a (proto-) article, or not. Demuth et al. (2012) suggest that this is because Spanish has many more trisyllabic and longer words than is typically found in languages such as English or French. They suggest that this may encourage learners to develop a larger three-syllable ‘prosodic window’ earlier – just to be understood. Thus, the frequency of different PW shapes in the ambient language helps determine the prosodic shape of children’s early morphologically simple and complex PWs (cf. see Demuth et al., 2012; Levelt, Schiller, & Levelt, 2000; Lleó, 2003; Prieto, 2006, for further discussion). We might then expect that the acquisition of PWs (and articles) in Italian – another Romance language with many long words – would pattern similarly to that found in Spanish, and this appears to be the case (cf. Giusti & Gozzi, 2006).

The second observation regarding the appearance of Spanish articles in the Demuth et al. (2012) study is that of individual variation. Both children used the same number of syllables in their early PWs, but only the girl’s were morphologically complex. This suggests that (1) both children had a similar sized three-syllable PW window, but that the boy had not yet acquired the syntactic knowledge needed to use articles. This is confirmed by his lack of article used even with two-syllable words before 2;1. Thus, one possibility is that the boy had the prosodic ability to produce articles, but not the syntax. Alternatively, under the ‘unfolding of prosodic structure’ hypothesis (Frota et al., 2016), he might not yet have the full, more complex prosodic structure needed to represent these function words at a higher level of the Prosodic Hierarchy. By understanding the various predictions made by this prosodic approach to the emergence of function words, we can better determine if a child’s lack of grammatical morphology is due to syntactic or prosodic limitations. This then brings us to case study 3, which examines the use of inflectional morphology.

Study 3: The acquisition of 3^rd person singular tense marking in English

As noted in the introduction, the acquisition of tense marking has long been an active area of research, with Radford (1990), Wexler (1994), and many others suggesting that bare verbs are produced until a child has knowledge of the syntactic structures needed to license the use of these inflectional morphemes. It was therefore of great interest that Song, Sundara, and Demuth (2009), using longitudinal spontaneous speech data from three children in the Providence Corpus (Demuth et al., 2006) aged 1;6–3;0, found that much of the within-speaker variable use of 3^rd person singular –s was due to (1) the prosodic position in which the morpheme appeared in the utterance, and (2) whether it occurred in a word that contained a consonant cluster or not. The later was first suggested by Marshall and van der Lely (2007), who showed that children with specific language impairment (SLI) were worse at producing past tense morphemes on verbs that had more complex codas (e.g., danced) than on those with simple codas (e.g., played), suggesting that phonotactics (and thus phonology), rather than syntax, was responsible. Hsieh, Leonard, and Swanson (1999) also noted that verbs tended to occur in the middle of utterances (at least in SVO English), whereas nouns tended to occur utterance-finally, where phrase-final lengthening would make the final syllable (and morpheme) longer. Song et al. (2009) confirmed this using data from the three children in the Providence Corpus, finding that almost three-quarters of children’s verbs occur utterance-medially. In contrast, more than half of their nouns occur utterance-finally. Although English is not a null-subject language, most subjects in children’s everyday speech are 1^st and 2^nd person pronouns, with very few nouns appearing utterance-initially (Evans & Demuth, 2012).

Song et al. (2009) thus suggested that phrase-final lengthening (which in early child speech is typically also IP-final) might enhance children’s production of verbal inflectional (tense), providing more time to produce all the segments of the utterance-final syllable, including the function morpheme. This was confirmed in data from both spontaneous speech corpora and from an elicited imitation task with children aged 2;3 that controlled for utterance length and syntactic complexity. Similar results have been found in children the same age using the same type of elicited imitation task with English plural inflections and coda clusters (Theodore et al., 2011), as well as in the development of coda consonants in European Portuguese (Jordão & Frota, 2010).

This led Sundara, Demuth, and Kuhl (2011) to propose that, since utterance-final morphemes were about twice as long as the utterance-medial ones, children would also be better able to perceive (and learn) these inflectional morphemes when they occurred utterance-finally compared to utterance-medially. Using simple sentences such as ‘She cries now’ vs. ‘Now she cries’, and their ungrammatical (uninflected) counterparts, Sundara et al. showed that 2-year-olds were much more sensitive to the morpheme when it was missing utterance-finally compared to utterance-medially. This confirms that the prosodic position in which an inflectional morpheme occurs in the utterance has a large impact not only on perception and production, but probably learning too. Many follow-up elicited imitation studies have confirmed these results, for both 3^rd person singular –s, as well as plural –s, showing that children’s production of utterance-final grammatical morphemes is significantly better than producing the same inflected words utterance-medially (e.g., Mealings, Cox, & Demuth, 2013; Theodore et al., 2011).

Mealings and Demuth (2014) also showed that increasing the length of the utterance had a negative effect on the production of 3^rd person singular –s, suggesting that aspects of complexity play a role in determining how and when children will omit grammatical morphemes, even when other factors are controlled (cf. also Valian, 1991). It is not entirely clear if this is due to constraints on working memory, or if this is due to increased prosodic complexity. This is obviously an area for further research.

The frequency of different allomorphs also appears to play a role in when and how certain forms of a morpheme are used. For example, Brown (1973) noted that –es forms of both the plural (e.g., peach-es) and 3^rd person singular (e.g., wash-es) were later acquired, and Berko (1958), using her famous ‘wug’ task, showed that children only acquired productive knowledge of this form of the plural between the ages of 4 and 7. However, this syllabic allomorph is very low in frequency, constituting only 5% of plural and 3^rd person singular allomorphs children hear. This helps to explain why it is acquired later, both by typically-developing children (e.g., Davies, Xu Rattanasone, & Demuth, 2017), and by those with SLI (Tomas, Demuth, & Petocz, 2017). Thus, frequency factors may also play a role in determining the likelihood that children will produce a particular grammatical morpheme, even once the prosodic and syntactic abilities to do so are in place.

General discussion

The above three studies provide a brief overview of a series of investigations all illustrating how prosodic factors play a critical role in understanding children’s morphosyntactic development. These studies are part of what led to the Prosodic Licensing Hypothesis (Demuth, 2014), pointing to the critically important role of the Prosodic Hierarchy in understanding morphological development. A better understanding of prosodic development in general is still needed to more fully understand many of these issues.

Not discussed above, but equally important to understanding many aspects of children’s early productions, is Selkirk’s (1996) observation regarding ‘alignment’, where edges of prosodic units may be aligned. Demuth (1995, 1996a) addresses some of these issues in terms of early constraints children learning both English and Dutch appear to have, ‘aligning’ primary stress with the left edge of their PWs (thus, baNAna becomes NAna). Although Demuth is agnostic about the higher levels of structure, her proposals are consistent with Frota and colleagues’ ‘prosodic unfolding’ hypotheses, with alignment of the foot/PW with the left edge of the IP (Frota et al., 2016). This is further supported by the presence of different tunes in early speech, with nuclear accents and IP boundary tones close to adult-like in 2-year-olds in European Portuguese and Catalan (Frota et al., 2016; Prieto, Estrella, Thorson, & Vanrell, 2012). They argue that what is acquired later is the unfolding of the different levels of prosodic structure, where the edges of the different levels of prosodic structure no longer need to match/align, yielding more complex structures. With the exception of Snow (1994), the present article has also not discussed the role of pitch and F0 in children’s early prosodic organization; this is another area for further research.

Note, however, that issues of ‘stress’ (or syllable prominence in the case of penultimate lengthening in Sesotho and phrase-final lengthening in French) have been briefly discussed above. Overall, grammatical morphemes that coincide with a ‘stressed’ syllable, as in the case of Sesotho agreement morphemes, and noun class prefixes that precede monosyllabic nominal stems, tend to be produced early. The same is true of ‘weak’ (unstressed) syllables that can be prosodified as part of a foot (e.g., English articles following a monosyllabic lexical item). Thus, although Leonard’s Surface Account suggests that grammatical morphemes that constitute ‘weak’ syllables will be likely to be omitted – at least by children with SLI – it appears that the prosodic context in which the unstressed grammatical morpheme appears plays a critical role in determining if it will be produced or not (see Leonard, 2014 for recent discussion).

In sum, there is much more research to be done on the prosodic organization of children’s early grammars, and the implications this has for understanding the development of morphosyntax. But this is true also of the field of linguistics in general, where prosodic factors of one kind or another may play a greater role in understanding syntax than often assumed (see, for example, Harford & Demuth, 1999). Much is certainly still to be learned about children’s acquisition of the higher levels of prosodic structure, and how this interacts with the lower levels of the foot and PW. Further acoustic and experimental evidence will be needed to provide a fuller account of these issues. This in turn will help provide the foundation for a developmental model of speech planning and production.

Conclusion

This article has reviewed some of the literature examining the role of phonological/prosodic factors in understanding children’s within-speaker variable (perception and) production of grammatical morphemes. It has shown that, across several typologically different Bantu, Romance, and Germanic languages, the prosodic shape a grammatical morpheme takes (e.g., a prosodic clitic or full PW), the prosodic context in which it occurs (i.e., shape of the adjacent PWs, position within the utterance, phonotactics), and the relative frequency of the morphological form, can all play an important role in determining when a child will tend to produce (or perceive) it. This is captured by the Prosodic Licensing Hypothesis (Demuth, 2014), where the likelihood of producing a grammatical morpheme can be predicted by these factors (see also Lleó, 2003). As a child’s prosodic (and more general phonological) competence develops, so does the ability to more reliably produce grammatical morphemes. This suggests that young children may ‘know’ much more about grammatical morphology than is often assumed. Computational modeling even suggests that the process of early word segmentation already involves a morphological parse of the incoming speech stream (Johnson, Christophe, Demuth, & Dupoux, 2014), and this would be consistent with the early perception results by Sundara et al. (2011). Though many of the determiner findings discussed here may be resolved by the age of 3 or so for typically-developing children, at least for very simple sentences, this may be delayed in children with SLI, and become variable again as sentence length increases and becomes more prosodically and syntactically complex (Valian, 1991). The fact that certain low-frequency allomorphs of a particular morpheme are much later acquired than others suggests that the study of children’s syntactic knowledge must proceed in a more integrated fashion, where syntax is only part of the larger whole of learning language.

Footnotes

Acknowledgements

I thank the editors and two reviewers for helpful comments.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work has been funded in part by the following grants: NIH grants R01MH60922 and R01HD057606, the ARC Centre for Cognition and its Disorders (CE110001021), and an ARC Laureate Fellowship (FL130100014).

References

Bassano

Maillochon

Mottet

(2008). Noun grammaticalization and determiner use in French children’s speech: A gradual development with prosodic and lexical influences. Journal of Child Language, 35, 403–438.

Berko

(1958). The child’s learning of English morphology. Word, 14, 150–177.

Branigan

(1979). Some reasons why successive single word utterances are not. Journal of Child Language, 6, 411–421.

Brown

(1973). A first language. Cambridge, MA: Harvard University Press.

Darwin

(1877). A biographical sketch of an infant. Mind, 2, 285–294.

Davies

Xu Rattanasone

Demuth

(2017). Two-year-olds’ sensitivity to inflectional plural morphology: Allomorphic effects. Language, Learning and Development, 13, 38–53.

Demuth

(1992a). Accessing functional categories in Sesotho: Interactions at the morpho-syntax interface. In Meisel

(Ed.), The acquisition of verb placement: Functional categories and V2 phenomena in language development (pp. 83–107). Dordrecht, The Netherlands: Kluwer Academic Publishers.

Demuth

(1992b). Acquisition of Sesotho. In Slobin

(Ed.), The cross-linguistic study of language acquisition (Vol. 3, pp. 557–638). Hillsdale, NJ: Lawrence Erlbaum.

Demuth

(1994). On the ‘underspecification’ of functional categories in early grammars. In Lust

Suñer

Whitman

(Eds.), Syntactic theory and first language acquisition: Cross-linguistic perspectives (pp. 119–134). Hillsdale, NJ: Lawrence Erlbaum.

10.

Demuth

(1995). Markedness and the development of prosodic structure. In Beckman

(Ed.), Proceedings of the North East Linguistic Society (Vol. 25, pp. 13–25). Amherst: GLSA, University of Massachusetts.

11.

Demuth

(1996a). Alignment, stress and parsing in early phonological words. In Bernhardt

Gilbert

Ingram

(Eds.), Proceedings of the International Conference on Phonological Acquisition (pp. 113–124). Somerville, MA: Cascadilla Press.

12.

Demuth

(1996b). The prosodic structure of early words. In Morgan

Demuth

(Eds.), Signal to syntax: Bootstrapping from speech to grammar in early acquisition (pp. 171–184). Mahwah, NJ: Lawrence Erlbaum.

13.

Demuth

(1996c). Stages in the acquisition of prosodic structure. In Clark

(Ed.), Proceedings of the 27th Child Language Research Forum (pp. 39–48). Stanford, CA: Stanford University, CSLI.

14.

Demuth

(2001a). A prosodic approach to filler syllables. Journal of Child Language, 28, 246–249.

15.

Demuth

(2001b). Prosodic constraints on morphological development. In Weissenborn

Höhle

(Eds.), Approaches to bootstrapping: Phonological, syntactic and neurophysiological aspects of early language acquisition (Language Acquisition and Language Disorders Series, Vol. 24, pp. 3–21). Amsterdam, The Netherlands: John Benjamins.

16.

Demuth

(2003). The acquisition of Bantu languages. In Nurse

Phillipson

(Eds.), The Bantu languages (pp. 209–222). London, England: Routledge.

17.

Demuth

(2014). Prosodic licensing and the development of phonological and morphological representations. In Farris-Trimble

Barlow

(Eds.), Perspectives on phonological theory and acquisition: Papers in Honor of Daniel A. Dinnsen (Language Acquisition and Language Disorders Series, Vol. 56, pp. 11–24). Amsterdam, The Netherlands: John Benjamins.

18.

Demuth

Culbertson

Alter

(2006). Word-minimality, epenthesis, and coda licensing in the acquisition of English. Language & Speech, 49, 137–174.

19.

Demuth

Ellis

(2009). Revisiting the acquisition of Sesotho noun class prefixes. In Guo

Lieven

Budwig

Ervin-Tripp

Nakamura

Ozçalikan

(Eds.), Crosslinguistic approaches to the psychology of language: Festschrift for Dan Slobin (pp. 93–104). New York, NY: Psychology Press.

20.

Demuth

Machobane

Moloi

(2009). Learning to license null noun class prefixes in Sesotho. Language, 85, 863–882.

21.

Demuth

McCullough

(2009). The prosodic (re)organization of children’s early English articles. Journal of Child Language, 36, 173–200.

22.

Demuth

Patrolia

Song

J. Y.

Masapollo

(2012). The development of articles in children’s early Spanish: Prosodic interactions between lexical and grammatical form. First Language, 32, 17–37.

23.

Demuth

Tremblay

(2008). Prosodically-conditioned variability in children’s production of French determiners. Journal of Child Language, 35, 99–127.

24.

Demuth

Weschler

(2012). The acquisition of Sesotho nominal agreement. Morphology, 21, 67–88.

25.

Doke

C. M.

Mofokeng

S. M.

(1985). Textbook of Southern Sotho Grammar. Cape Town, South Africa: Maskew Miller Longman.

26.

Evans

Demuth

(2012). Individual differences in pronoun reversal: Evidence from two longitudinal case studies. Journal of Child Language, 39, 162–191.

27.

Fikkert

(1994). On the acquisition of prosodic structure. Dordrecht, The Netherlands: Holland Institute of Generative Linguistics.

28.

Frota

(2012). Prosodic structure, constituents and their representations. In Cohn

Fougeron

Huffman

(Eds.), The Oxford handbook of laboratory phonology (pp. 255–265). Oxford, UK: Oxford University Press.

29.

Frota

Cruz

Matos

Vigário

(2016). Early prosodic development: Emerging intonation and phrasing in European Portuguese. In Armstrong

Henriksen

N. C.

Vanrell

M. M.

(Eds.), Intonational grammar in Ibero-Romance: Approaches across linguistic subfields (pp. 295–324). Philadelphia, PA: John Benjamins.

30.

Gennari

Demuth

(1997). Syllable omission in Spanish. In Hughes

E. M.

Green

(Eds.), Proceedings of the 21st Annual Boston University Conference on Language Development (Vol. 1, pp. 182–193). Somerville, MA: Cascadilla Press.

31.

Gerken

L. A.

(1994). Young children’s representation of prosodic structure: Evidence from English-speakers’ weak syllable omissions. Journal of Memory and Language, 33, 19–38.

32.

Gerken

L. A.

(1996). Prosodic structure in young children’s language production. Language, 72, 683–712.

33.

Giusti

Gozzi

(2006). The acquisition of determiners: Evidence for the full competence hypothesis. In Belletti

Bennati

Chesi

Di Domenico

Ferrari

(Eds.), Language acquisition and development: Proceedings of GALA 2005 (pp. 232–238). Newcastle upon Tyne, UK: Cambridge Scholars Press.

34.

Goad

Buckley

(2006). Prosodic structure in child French: Evidence for the Foot. Catalan Journal of Linguistics, 5, 109–142. (Special issue on the acquisition of Romance languages as first languages).

35.

Hall

T. A.

(1999). Phonotactics and the prosodic structure of German function words. In Hall

T. A.

Kleihenz

(Eds.), Studies on the phonological word (pp. 99–131). Amsterdam, The Netherlands: John Benjamins.

36.

Harford

Demuth

(1999). Prosody outranks syntax: An Optimality approach to subject inversion in Bantu relatives. Linguistic Analysis, 29, 47–68.

37.

Harris

(1983). Syllable structure and stress in Spanish: A nonlinear analysis. Cambridge, MA: MIT Press.

38.

Hayes

(1995). Metrical stress theory. Chicago, IL: University of Chicago Press.

39.

Hsieh

Leonard

L. B.

Swanson

(1999). Some differences between English plural noun inflections and third singular verb inflections in the input: The contributions of frequency, sentence position, and duration. Journal of Child Language, 26, 531–543.

40.

Johnson

Christophe

Demuth

Dupoux

(2014). Modeling function words improves unsupervised word segmentation. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Vol. 1: Long Papers, pp. 282–292). Stroudsburg, PA: Association for Computational Linguistics.

41.

Jordão

Frota

(2010). The intonational phrase constrains coda development. In Castro

Costa

Lobo

Pratas

(Eds.), Language acquisition and development: Proceedings of GALA 2009 (pp. 240–251). Cambridge, UK: Cambridge Scholars.

42.

Leonard

(2014). Children with specific language impairment (2nd ed.). Cambridge, MA: MIT Press.

43.

Levelt

C. C.

Schiller

N. O.

Levelt

W. J.

(2000). The acquisition of syllable types. Language Acquisition, 8, 237–264.

44.

Lleó

(2003). Prosodic licensing of codas. Probus, 15, 257–281.

45.

Lleó

Demuth

(1999). Prosodic constraints on the emergence of grammatical morphemes: Crosslinguistic evidence from Germanic and Romance languages. In Greenhill

Littlefield

Tano

(Eds.), Proceedings of the 23rd Annual Boston University Conference on Language Development (pp. 407–418). Somerville, MA: Cascadilla Press.

46.

MacWhinney

(2000). The CHILDES Project: Tools for analyzing talk, Vol. 2: The database (3rd ed.). Mahwah, NJ: Lawrence Erlbaum.

47.

Marshall

C. R.

van der Lely

H. K. J.

(2007). The impact of phonological complexity on Past tense inflection in children with Grammatical-SLI. Advances in Speech-Language Pathology, 9, 191–203.

48.

Mealings

Cox

Demuth

(2013). Acoustic investigations into the delayed acquisition of the syllabic -es plural in 2-year-olds’ speech. Journal of Speech Language and Hearing Research, 56, 1260–1271.

49.

Mealings

Demuth

(2014). The role of utterance length and position in three-year-olds’ production of third person singular -s. Journal of Speech Language and Hearing Research, 57, 484–494.

50.

Nespor

Vogel

(1986). Prosodic phonology. Dordrecht, The Netherlands: Holland.

51.

Peters

(1983). The units of language acquisition. Cambridge, UK: Cambridge University Press.

52.

Prieto

(2006). The relevance of Metrical information in early Prosodic Word acquisition: A comparison of Catalan and Spanish. Language and Speech, 49, 231–259.

53.

Prieto

Estrella

Thorson

Vanrell

M. M.

(2012). Is prosodic development correlated with grammatical development? Evidence from emerging intonation in Catalan and Spanish. Journal of Child Language, 39, 221–257.

54.

Radford

(1990). Syntactic theory and the acquisition of English syntax. Oxford, UK: Basil Blackwell.

55.

Roark

Demuth

(2000). Prosodic constraints and the learner’s environment: A corpus study. In Catherine Howell

Fish

S. A.

Keith-Lucas

(Eds.), Proceedings of the 24th Annual Boston University Conference on Language Development (pp. 597–608). Somerville, MA: Cascadilla Press.

56.

Scullen

M. E.

(1997). French prosodic morphology: A unified account. Bloomington: Indiana University Linguistics Club.

57.

Selkirk

E. O.

(1984). Phonology and syntax: The relation between sound and structure. Cambridge, MA: MIT Press.

58.

Selkirk

E. O.

(1996). The prosodic structure of function words. In Morgan

J. L.

Demuth

(Eds.), Signal to syntax: Bootstrapping from speech to grammar in early acquisition (pp. 187–213). Mahwah, NJ: Lawrence Erlbaum.

59.

Smit

(1993). Phonologic error distributions in Iowa-Nebraska Articulation Norms Project: Word-initial consonant clusters. Journal of Speech and Hearing Research, 36, 931–947.

60.

Snow

(1994). Phrase-final syllable lengthening and intonation in early child speech. Journal of Speech and Hearing Research, 37, 831–840.

61.

Song

J. Y.

Sundara

Demuth

(2009). Phonological constraints on children’s production of English third person singular -s. Journal of Speech, Language, and Hearing Research, 52, 623–642.

62.

Sundara

Demuth

Kuhl

(2011). Sentence-position effects on children’s perception and production of English 3rd person singular –s. Journal of Speech, Language, and Hearing Research, 54, 55–71.

63.

Szagun

Schramm

S. A.

(2016). Sources of variability in language development of children with cochlear implants: Age at implantation, parental language, and early features of children’s language construction. Journal of Child Language, 43, 505–536.

64.

Theodore

Demuth

Shattuck-Hufnagel

(2011). Acoustic evidence for position and complexity effects on children’s production of plural –s. Journal of Speech, Language, and Hearing Research, 54, 539–548.

65.

Tomas

Demuth

Petocz

(2017). The role of frequency in learning morphophonological alternations: Implications for children with Specific Language Impairment. Journal of Speech, Language, and Hearing Research, 60, 1316–1329.

66.

Turpin

Demuth

Campbell

A. N.

(2014). Phonological aspects of Arandic Baby Talk. In Pensalfini

Turpin

Guillemin

(Eds.), Language description informed by theory (Studies in Language Companion Series, Vol. 147, pp. 49–79). Amsterdam, The Netherlands: John Benjamins.

67.

Valian

(1991). Syntactic subjects in the early speech of American and Italian children. Cognition, 40, 21–81.

68.

Veneziano

Sinclair

(2000). The changing status of ‘filler syllables’ on the way to grammatical morphemes. Journal of Child Language, 27, 461–500.

69.

Vigário

(2003). The prosodic word in European Portuguese. Berlin, Germany: Mouton de Gruyter.

70.

Wexler

(1994). Optional infinitives, head movement and the economy of derivations. In Lightfoot

Hornstein

(Eds.), Verb movement (pp. 305–350). Cambridge, UK: Cambridge University Press.

Prosodic constraints on children’s use of grammatical morphemes

Abstract

Keywords

Introduction

Study 1: The acquisition of Bantu noun class prefixes

Study 2: Acquisition of articles/determiners in English, French, and Spanish

Study 3: The acquisition of 3rd person singular tense marking in English

General discussion

Conclusion

Footnotes

Acknowledgements

Declaration of conflicting interests

Funding

References

Study 3: The acquisition of 3^rd person singular tense marking in English