Abstract
In this article, we present an analysis of the complexity of grammatical constraints and their impact on early language acquisition of inflectional morphemes in Malayalam. We use the natural speech production data of two monolingual children acquiring Malayalam between the ages 1;9–2;10 and 2;3–3;0 and three bilingual children acquiring Malayalam-English between the ages 1;9–2;8, 2;0–3;0 and 1;10–2;11 to recover the underlying grammatical constraints that govern the correct productions as well as errors across monolingual and bilingual contexts. We find rules that reference lexico-semantic properties to be particularly challenging to young children.
Keywords
Introduction
Children show rapid progress in modulating lexical categories for different inflections to produce different lexical forms soon after they produce their first words. They add markers such as tense, aspect, mood and agreement on verbs, and markers such as case, number and gender on nouns to represent the grammatical distinctions relevant to their target language(s). Often, children’s novel utterances correspond to those that their adult counterparts could produce. However, even when their productions adhere to the rules of the target grammar(s), children are also seen to produce certain kinds of errors that are typically not attested in adult utterances. Such errors in early language productions have been broadly divided into two types: (a) errors of omission, and (b) errors of commission. In the first type, a grammatical item or an affix is not employed in contexts where it is obligatory, that is, children produce less material than adults typically would. In the second type, there may be substitutions or overgeneralisations of affixal material, that is, children produce more and different material than the adults would in the same syntactic contexts.
Studies of early language acquisition have offered several explanations for the ease/difficulty in acquiring the inflectional morphology of particular languages: (a) semantic distinctions that are conceptually complex may take longer to learn, (b) uneven or irregular paradigms can impede acquisition, and (c) typological properties can play a role in facilitating or hindering the process of morphological acquisition (Clark, 2016). In time, children learn the complete morphosyntactic function of each inflectional type in their target language. While semantic complexity is one of the main factors determining the relative ease/difficulty in acquiring an inflectional type (Slobin, 1973), precisely what constitutes semantic complexity cross-linguistically is not well-established.
In this article, we look at how inflections in Malayalam are acquired by young children and what their production errors tell us about the growth of grammar. We use longitudinal data across monolingual and bilingual contexts to consider the acquisition patterns of nominal and verbal inflections. The bilingual data provides for a fuller picture of the acquisition patterns, since many errors are common to both contexts and a few are limited to either the monolingual or the bilingual context. In addition, the effect of typology is easier to observe in bilinguals who are simultaneously acquiring two typologically different languages, here Malayalam and English.
While children acquire the target inflections in Malayalam with great ease, their productions also contain errors that are worth analysing because (a) they are restricted to certain inflections across the children, and (b) they allow us to confirm the underlying grammatical constraints that govern the different morphological processes. We can then cross-verify whether specific grammatical processes themselves are inherently difficult for the children to acquire, or if there are other factors or constraints that impede full acquisition. To this end, we focus our attention on the different types of inflectional errors.
Malayalam (South Dravidian) has regular and productive agglutinative morphology as is typical of the language family. Nominal inflections include those for case, number and gender, while verbal inflections include those for tense, aspect and mood but not subject–verb agreement, which is atypical for Dravidian languages. Nouns and verbs carry the bulk of the grammatical information, often through inflectional markers, and we limit the discussion here to these two categories. The acquisition of Malayalam is, in principle, interesting, since children acquiring such agglutinative systems are observed to both acquire and productively use the target inflections much earlier compared to their counterparts learning languages with sparser inflectional systems (Aksu-Koç & Slobin, 1985; Dressler, 2012).
There are, of course, a number of studies seeking to account for errors (especially of inflections) in young children’s productions (Clahsen et al., 1992, 2002; Maslen et al., 2004). Two main kinds of models have been proposed. One type, for example, the small clause hypothesis (Radford, 1990), sees child grammars as different objects from adult grammars, with this difference driving the ‘missteps’. The second kind, for example, the full competence hypothesis (Poeppel & Wexler, 1993), sees child grammars as being congruent to adult grammars and aims to account for the errors in other ways. Within the generative framework, models such as the Prosodic Constraints model (Demuth, 1996, 2001), the Truncation Model (Rizzi, 1993), and the Agreement/Tense Omission Model (ATOM) (Schütze & Wexler, 1996) have been proposed to account for specific patterns in early child productions, such as the absence of tense morphology or the elision of subjects in non-pro-drop languages. These studies discuss, generally, data from various Western languages. Acquisition studies involving errors of commission have mainly centred on lexical access of irregular and regular inflections such as the dual-route model (Marcus, 1995; Pinker & Prince, 1994) and the single-route model (Daugherty & Seidenberg, 1994; Elman et al., 1996; Rumelhart & McClelland, 1986), and have concentrated primarily on English. We analyse both kinds of errors and the correct productions in Malayalam, an understudied language, to understand the development and organisation of the grammars. As Saaristo-Helin et al. (2011) point out, early grammars of less well-known and less well-studied languages need to be given more attention for a better understanding of the grammatical processes and the human language faculty as a whole.
There are very few acquisition studies on Dravidian languages, in general (Lakshmanan, 2006; Lakshmi Bai, 2004; Raghavendra & Leonard, 1989; Sarma, 2014; Usha Rani & Sailaja, 2004), and for Malayalam, in particular (Gayathri, 2019; Girija Devi, 1972; Raghunathan, 2021). The patterns in early acquisition data allow us to uncover the grammatical constraints that operate in Malayalam and to evaluate how they influence the developmental trajectory. Specifically, the production errors can be used to understand both the formal or semantic complexity of inflectional assignment, and the relative ease or difficulty in their acquisition. Since many of the affixes are phonologically salient, semantically transparent, productively applied and easily segmentable (Aksu-Koç & Slobin, 1985; Dressler, 2012), as is typical of agglutinative languages, we may expect children’s grammars to not deviate from adults’ grammars. The observed fact is that while early Malayalam productions are remarkably close to adult productions, they still show certain differences and a detailed investigation can clarify the causes of such deviations. We concern ourselves with three research questions, as follows:
RQ1. What underlying properties determine the ease or difficulty of acquisition of various nominal and verbal inflections in Malayalam?
RQ2. Is there a significant difference in the relative ease or difficulty of acquisition of nominal versus verbal inflections?
RQ3. Do monolingual and bilingual contexts show differences with respect to the acquisition of the inflectional markers in the same language?
The outline of the article is as follows. First, we provide an overview of the subjects and the methods of data collection and organisation. We then discuss the specific errors attested in early Malayalam for nouns and verbs. We present the results of our study, followed by a detailed discussion of the research questions, and offer our conclusions.
Subjects and method
The longitudinal data that we analyse here include the spontaneous speech productions of (a) two monolingual children, a girl and a boy, acquiring Malayalam, and (b) three bilingual children, two girls and a boy, acquiring Malayalam and English. 1 All the children belong to middle-class families with college-educated parents. The monolingual children were being raised with Malayalam as the only language of communication at home and in the larger social environment. The bilingual children heard both languages at home through the year of recording. The linguistic experiences of the bilingual children came to vary over the course of the year although they received similar kinds of input overall: one heard more Malayalam early on because of visits with the extended family, the second child received a little more exposure to English during the day at a crèche, and the third child heard some Kannada and Tamil as well from caregivers at home. All the recordings were transcribed and organised in ELAN (Wittenburg et al., 2006).
Child Am was recorded between the ages of 1;9.14 and 2;10.3, over 26 sessions, and child Hm was recorded between the ages of 2;3.28 and 3;0.17, over 18 sessions. The recording sessions range in duration from 30 to 50 minutes. The children were recorded in their own homes using an audio recorder while generally interacting with the researcher and, occasionally, with their caregivers. Additional materials like puzzles and picture books were used to encourage the children to speak and to facilitate interaction with the researcher.
The monolingual data (Table 1) consist of a total of 3868 and 1357 utterances, of which 3673 and 1072 were analysed for Am and Hm, respectively. The Mean Length of Utterances (MLU) calculated based on the number of morphemes in the utterances during the period of the study are 4.4 and 4.8 for Am and Hm, respectively. There are 9953 words and 3394 inflectional tokens in Am’s transcripts and Hm has 2809 words and 1095 inflectional tokens.
Monolingual data overview.
The bilingual data (Table 2) include the spontaneous speech productions recorded at least once a month by the parents for the children whom we label Ab, and Db, and twice or more per fortnight by the researcher for the child we label Eb. The recordings are on average 40 minutes long. Child Db was recorded between 1;9.10 and 2;8.21, over 20 sessions; child Ab was recorded between 2;0.18 and 3;0.20, over 18 sessions; and child Eb was recorded between 1;10.7 and 2;11.15, over 25 sessions. Most of the interactions were within the children’s homes, with a few sessions being recorded at a park. Although the three bilingual children were situated in different geographical locations, they received similar inputs from their parents with adequate and equal exposure to both languages. The bilingual data consist of a total of 2004, 4543 and 1542 utterances for Ab, Eb and Db, respectively. The MLUs in Malayalam for Db, Eb and Ab during the period of the study are 3.7, 2.2 and 2.4, respectively. The MLUs in English for Db, Eb and Ab during the same period are 2.2, 2.5 and 3.6, respectively. There are 1328 Malayalam words and 583 inflectional tokens in Db’s transcripts, 1488 Malayalam words and 353 inflectional tokens in Eb’s transcripts, and 652 Malayalam words and 184 inflectional tokens in Ab’s transcripts.
Bilingual data overview.
We have analysed every instance of inflectional marking in both sets of data, excluding only non-words, false starts, fillers, songs, imitations and repetitions (except when used for emphasis). The criteria to determine productivity in inflectional marking are adapted from Cazden (1968). We take a marker to be productively used when a child uses it (a) correctly, (b) in the obligatory contexts at least 90% of the time and (c) in three out of four consecutive samples of data.
Errors in inflections were coded separately based on their deviations from target adult structures. When the omission of a marker resulted in ungrammaticality, it was identified as an omission error, and when a marker was incorrectly used in contexts that do not require them in the adult language, it was identified as a commission error. Productions were deemed to be errors based on native speaker judgements. Two of the authors are native speakers of Malayalam and they coded the data. Inter-coder reliability was established using Cohen’s kappa. Based on the proportion of errors, the estimated sample size required for 95% confidence in the kappa estimand is 1284 (Shoukri et al., 2004). We ran the test on a randomised selection of 1300 utterances and obtained a kappa of 0.94.
Data
In this section, we focus on the verbal and nominal inflection errors produced by the children. While there are no omission errors with verb inflections in the monolingual data, we find omissions of tense inflections (also called Root Infinitives/RIs) in the bilingual data. Commission errors are restricted to the past tense affix and are common to both sets of data. Compared to children’s productions of verbal inflections, their productions of nominal inflections show many omission and commission errors. Overall, however, the ratio of errors to correct instances is low as we will see in the following.
Verb inflections
Tense, aspect and mood are the three major verbal inflectional categories in Malayalam, and children are seen to use these categories productively and with ease. Am’s transcripts have 1773 tokens and Hm’s transcripts have 585 tokens of verb inflections. Inflectional marking is obligatory in Malayalam and bare stems make pronounceable phonological words only for bare imperative forms. The verb inflections attested in the monolingual data include the present progressive -un̪n̪u, past tense suffixes -i and -Tu, future marker -um, infinitive -aan, participial -uka, optative -aʈʈe, conditional -aal and the imperative markers -aan, -ee and -oo. None of these inflectional markers are omitted in the monolingual transcripts. Table 3 summarises the Malayalam verbal production data of monolingual and bilingual children.
Monolingual and bilingual Malayalam verb inflection production data.
The verb inflection errors that monolingual children make are all overgeneralisations of the default past tense exponent -i in place of the required exponent -Tu. The exponent selection is determined by contextual allomorphy. When there are affixes at v and Voice between the root and T(ense), the past tense suffix selected is always -Tu, otherwise the default -i applies. Children’s errors always show the use of the default -i in structures that require -Tu and never the converse. Am produces this error once each at 2;4.18 and 2;9.3 with root koll- ‘kill’ forming *[kolli] /kon̪n̪u/ ‘killed’ as shown in (1) and once at 2;3.16 with the root koɭɭ- ‘hit’ forming *[koɭɭi] /koɳɖu/ ‘hit’. Hm produces the overgeneralised form *[kolli] once each at 2;8.2 and 2;9.16, and twice at 2;8.16:
While the bilingual children productively use a range of verbal inflections including tense, like the monolinguals, we find that this is not the case with their tense inflections in English (Table 4). We find tense omissions and the presence of RIs in English throughout the period of recording (2), as has also been observed in monolingual English acquiring children (Guasti, 2017):
Bilingual English verb inflection production data.
In contrast to the lack of omission errors involving verbal suffixes in the monolingual productions, the bilingual children’s data include two instances of tense omissions in Malayalam by Eb, one each at 1;10.7 (3) and 2;4.5. We find the use of the default past tense exponent -i instead of the exponent -Tu in the bilingual data as well. Db makes this commission error with the same verb υeeυ- ‘boil’ twice at 2;6.5 (4) but does not repeat this error of commission later. At the final recording session at 3;0.20, Ab makes a similar error once with the past tense form of the verb t̪aɳukk- ‘cool’ (5):
Nominal inflections
Malayalam nouns are inflected for case, number and gender (e.g. maɖiy-an-maar-e; lazy-
Monolingual and bilingual Malayalam nominal inflection production data.
Malayalam has two subject cases, the nominative and the dative. Experiencer predicates assign dative case and select only [+animate] nouns as ‘subjects’. The dative case also ranges over different theta roles (experiencer, possessor, location, recipient etc.) and argument types (subject and indirect object) and is, thus, semantically complex. The genitive case overlaps with the dative case in signalling possession. Malayalam also shows differential object marking based on the noun’s animacy features. Direct object nominals must bear the accusative suffix overtly if they are [+ animate], but not if they are [−animate]. Malayalam nouns bear an inherent gender feature, and the three genders are congruent to natural sex distinctions, as in English. Absent subject–verb agreement, a common locus for gender features, the overt expression of gender is limited to the third person pronouns and to nouns that distinguish the male from the female (e.g. mak-
The errors in nominal inflectional assignment primarily involve case and number marking, while gender marking is error free. Table 6 provides the number of correct instances versus the error tokens of various inflectional types for which errors are attested in the monolingual productions. The accusative is the most frequently omitted case marker in both the monolingual children’s speech. In addition to the omission errors, there are also commission errors involving accusative marking. With the dative case too, we find both omission and commission errors, while genitive and locative markings only have omission errors and show no commission errors. For Am, the omission errors of the accusative, dative, and genitive cases persist till the end of the period of recording. Am is seen to omit the accusative case even at 2;10.3, (the last recording session), the dative at 2;9.3, and the genitive at 2;9.15 (although there are fewer genitive omissions overall). The last locative case omission, in contrast, is attested at 2;2.15 but not afterwards. Hm’s omission errors involving case markers are restricted to the accusative and the genitive. He does not omit any dative or locative affixes during the period of study. Unlike Am, his only omission error involving the genitive marking is attested at 2;4.14 and is not seen afterwards, but like Am, the omissions involving the accusative persist till the end of the study with the last accusative omission being attested at 3;0.2. Sample sentences with omission errors across case markers are given in (6) to (9):
Malayalam nominal inflections: token frequencies of correct instances versus errors in monolingual children.
OE: omission errors; CE: commission errors.
Commission errors with the accusative marker in Am’s data involve the use of the case marker instead of the bare nominal (10) once at 2;0.16 and thrice at 2;5.16, and the overt marking of accusative on a [−animate] object once at 2;5.16 (11). Hm’s only commission error with the accusative marker is of the latter type (12):
Dative commission errors typically involve the use of dative subjects instead of nominative subjects and are found in Hm’s speech productions once at 2;8.16 (13), while Am has five such instances once each at 2;5.2, 2;8.1 and 2;10.3, and twice at 2;5.16. In addition, Am once incorrectly substitutes the dative case for an expected genitive on a possessive nominal (14):
Overall, plural productions in the monolingual data are far fewer compared to case and gender markers. The only plural marker that is incorrectly omitted in the monolingual data is -maar. While Am has only one recorded instance of omission of this marker at 2;1.1, Hm omits it thrice at 2;4.14 (15), 2;9.16 and 2;10.2. In fact, Hm has more error tokens than correct uses when it comes to this marker, unlike Am. Hm’s productions include one commission error involving -maar (16), but Am has none. Conversely, Am has a commission error involving the marker -kaɭ (17), while Hm does not have any:
Table 7 provides the correct versus error tokens of nominal inflectional types for which errors are attested in the bilingual Malayalam productions. The accusative is the most frequently omitted case marker in Malayalam in the bilingual data as well. Omission errors are also observed with the dative and genitive suffixes but not the locative. Db is seen to last omit the accusative at 2;5.20, the dative at 1;10.6 and the genitive at 2;1.18, and does not omit the locative marker. Eb’s last observed omission errors for accusative and genitive is at 2;11.1, a fortnight before the final recording. She omits the dative only once during the recording at 2;6.19 and does not omit the locative marker. Ab’s last accusative omission is at 3;0.9, a fortnight before the final recording session. Her final dative omission is observed at 2;8.10 and the genitive at 2;7.20. Examples are given in (18) and (19):
Malayalam nominal inflections: token frequencies of correct instances versus errors in bilingual children.
OE: omission errors; CE: commission errors.
In addition, there are two commission errors in Ab’s data at 2;1.25 (20) where the accusative marker is applied to a [−animate] noun:
While there are no plural omission errors in the bilingual data, Db produces a single commission error of the plural marker -kaɭ at 2;5.2 (21):
In Table 8, we provide a comparative summary of the English noun productions, and in Table 9, we include the token frequency details for errors in case and number marking in English in the bilingual data. When we compare Tables 7 and 9, we observe that the bilingual children do not demonstrate the same difficulty in using the accusative in their English productions that they do in Malayalam. As the data show, the two case markers are used correctly 90%–100% of the time by the three children. Although Db only has one instance of a visible accusative in English, he does use the unmarked accusative (whole nouns) in English in the correct word order (object after verb), which is also suggestive of an understanding of the object properties.
Bilingual English nominal inflection production data.
English nominal inflections: token frequencies of correct instances versus errors in bilingual children.
OE: omission errors; CE: commission errors.
In this section, we saw the different kinds of errors attested in inflectional marking in early Malayalam productions. Verb inflectional productions in Malayalam have fewer errors across both sets of children and involve, primarily, tense marking. However, nominal inflections include many more errors. We find only omission errors with the locative and genitive, whereas the dative, accusative and plural show both errors of omission and commission. In the next section, we focus on these three categories to show how animacy as a grammatical constraint makes inflectional acquisition difficult for early learners.
Results
In the previous section, we discussed the five children’s use of different nominal and verbal inflections in their productions, and we noted that there are many more production errors with nominal inflections than with verbal inflections. As has been observed for typologically similar languages in the literature (Aksu-Koç & Slobin, 1985; Dressler, 2010; Stephany & Voeikova, 2009), we expect greater ease of acquisition of target inflections in Malayalam, as well. It stands to reason, then, that the difficulty in the complete acquisition of certain inflections may be the consequence of the complexity of the rules that govern their assignment (RQ1). Fisher’s exact tests were run on the nominal and verbal inflectional productions of monolingual and bilingual children to verify whether the difficulty of producing inflectional morphology on a particular lexical category (which in our case is correlated with more errors) is dependent on the language-specific grammatical rules that underlie the inflectional assignment.
We see that it is the accusative that has the greatest number of errors in both the monolingual and bilingual data. Accusative case marking in Malayalam is dependent on the animacy constraint where only [+animate] objects are overtly marked and [−animate] objects are obligatorily left unmarked. We ran Fisher’s exact tests on the accusative productions in both the monolingual and the bilingual data to determine whether children have greater success with inanimate nouns than with animate nouns. Table 10 shows that children do achieve greater accuracy when the accusative marking is obligatorily null than with nouns that require such marking. The difference in accuracy across the data as a whole is statistically significant at p < 0.01.
RQ1: Accusative case productions of monolingual and bilingual children.
We have seen that verbal inflection errors involve tense marking and that nominal inflection errors range over case and number morphology. Both the monolingual and the bilingual data show that there are more errors of nominal inflections than verbal inflections (RQ2). Again, we ran Fisher’s exact tests on monolingual and bilingual nominal and verbal inflectional productions to determine whether there is a difference in the relative ease, correlated with the properties of agglutinative typology, or difficulty, assumed to be dependent on the grammatical rules, in the acquisition of verbal inflections versus nominal inflections. Table 11 shows that children have a statistically significant advantage (p < 0.01) on verbal inflectional marking over nominal inflectional marking, which in turn tells us that acquisition of nominal morphology is relatively difficult compared to verbal morphology in Malayalam.
RQ2: Nominal versus verbal Malayalam inflectional production in monolingual and bilingual children.
We also wished to determine whether early monolinguals have an advantage over early English–Malayalam bilinguals with respect to inflectional production in Malayalam (RQ3). We ran Fisher’s exact tests separately on nominal inflectional tokens and verbal inflectional tokens of monolingual and bilingual subjects to determine whether monolingual and bilingual children show differences with respect to the acquisition of verbal and nominal inflectional morphology in the same language. As shown in Table 12, monolingual children appear to have a statistically significant advantage (p < .01) over bilingual children when it comes to nominal inflectional marking, while such an advantage is not seen when it comes to verbal inflectional assignment (p = .07).
RQ3: Inflectional production in Malayalam in monolingual and bilingual children.
Despite the overall asymmetry between verbal and nominal inflections, bilingual children produce more nominal inflectional errors in Malayalam when compared to the monolingual children in the study. This difference between the two sets of children led us to check whether the bilingual children show any differences between their Malayalam and English nominal and verbal inflectional productions. To this end, we ran separate Fisher’s exact tests on Malayalam versus English nominal inflectional tokens and Malayalam versus English verbal inflectional tokens in the bilingual data to measure the dependency on the semantic complexity of the underlying grammatical rules and the effect of language typology. The results (Table 13) show that the ease of production of English nominal inflections is significantly better than Malayalam nominal inflections in bilingual children (p < .01).
Malayalam versus English nominal inflection production in bilingual children.
In contrast, the bilingual children have a considerable number of errors in their verbal inflectional productions in English, and very few errors in their Malayalam productions. We ran Fisher’s exact tests on the English versus Malayalam verbal inflectional production data of the bilinguals to see whether the greater success rate in verbal inflectional production in Malayalam, when compared to English, is dependent on the typological differences between the languages (Table 14). The agglutinative typology of Malayalam is expected to facilitate the acquisition of inflections compared to an inflection impoverished English. The results show that the children have a statistically significant advantage with Malayalam verbal inflections than with English verbal inflections (p < .01).
Malayalam versus English verbal inflection production in bilingual children.
In this section, we analysed whether early learners have difficulty with overt accusative marking, in particular, and nominal inflectional production, in general. We also checked whether monolingual learners have an advantage over bilingual learners with respect to nominal and verbal inflectional production and whether Malayalam nominal inflections pose a greater challenge to bilingual subjects than do Malayalam verbal inflections, which are produced with great accuracy by all the children. We also compared the production data of the bilinguals across both languages and found inverse correlations between languages and the inflectional type – verbal inflections being harder in English than in Malayalam and the converse for nominal inflections. We discuss the results and possible reasons for each error type in the next section.
Discussion and conclusion
Inflections in Malayalam exhibit several well-recognised and key morphological properties that facilitate language acquisition. These include phonological salience of the morpheme for identification, segmentability (non-fusional affixes), semantic transparency, bi-uniqueness (less allomorphy) and a high level of regularity and productivity (Dressler, 2012). The inflectional morphological system (as in many typologically similar languages) possesses several attributes that make it transparent for acquisition by the young child and is, perhaps, why young Malayalam learners exhibit adult-like mastery of their target morphosyntax from very early on with very few errors in their productions. Malayalam acquisition data, as seen here, offer strong support for a full-competence hypothesis of early grammars (Poeppel & Wexler, 1993).
Despite the overall success, we see that children make errors when they employ nominal and verbal suffixes. Though the total number of error tokens is small, they are nonetheless revelatory of the grammatical processes. While the monolingual data do not attest any omission errors with verbal inflectional marking, the two tense omission errors (RIs) in Malayalam from the bilingual data may involve a language switch, where the child may have begun a sentence in English and ended it in Malayalam verbs using the bare or RI equivalent of English as shown in (3). This is most likely due to the conflicting demands from the two languages and suggests a local compromise in marking finiteness in English and Malayalam. These errors are visible only with child Eb, whose use of RIs and bare imperatives in English is extensive and numerous. It must be noted that there are no other instances of tense omissions in any of the Malayalam data with the other children, both bilingual and monolingual.
The ease with which developing grammars handle agglutinative, inflectional morphology is underscored by the fact that early Malayalam does not show a Root Infinitive (RI)/Root Infinitive Analogue (RIA) stage in either the monolingual or the bilingual contexts. Malayalam acquisition data, then, pattern with other pro-drop languages like Italian and Tamil with no such discernible RI stage, unlike English and French (Guasti, 2017). In fact, the bilingual children’s differential capacities in Malayalam and English emphasise and reveal how this typological difference between the languages impacts acquisition. RIs are seen in the children’s English productions, even while they produce finite and fully-inflected verb forms in Malayalam. The lack of omission errors in verbs and the absence of an RI/RIA stage must, therefore, follow from the grammatical typology of Malayalam. The typological advantage of Malayalam over English is captured in Table 14 and shows us that bilingual children find it relatively easy to successfully produce the regular agglutinative verbal inflectional morphology of Malayalam compared to the sparse verbal inflectional morphology of English.
The other error involving verb suffixation in Malayalam that is seen across the monolingual and bilingual data is the overgeneralisation of the default past tense marker -i to contexts where the adult grammars select -Tu. While these commission errors can be seen as a choice of the default over the specific, the underlying reasons are of interest. The past tense suffix -Tu is determined via contextual allomorphy when affixes are present at the intervening v and Voice heads below T(ense). Absent such affixes, the default -i is selected. Synchronically, the morphosyntactic specifications for many v/Voice affixes have become unpredictable and increasingly opaque in the grammar because of reanalysis by analogy, paradigm levelling and semantic shifts. This opacity and the need to learn ‘exceptional’ tense marking may push the child towards the default in the same way that monolingual English children overuse the regular past tense affix -ed.
The asymmetry between the verbal production data with few errors, and the nominal production data, with many more errors, points to difficulties that may be specific to the nominal morphosyntactic rules. We found that most of the errors of nominal inflections centre on the accusative, dative and plural marking. Compared to these three nominal inflectional types, which involve both errors of commission and omission, genitive marking only involves omission errors that could be attributed to the overlap in contexts with the dative. However, what is common to the accusative, dative and plural affixation is their use of the animacy feature as a grammatical constraint within the morphological rules that govern inflectional categories, which seems to be challenging to young children. The accusative and the dative cases employ a two-way contrast of the animacy criterion [±animate], but plurals employ a three-way combination of the two features [±human] and [±animate]. 3 Since all the other major inflectional markers are used frequently, accurately and productively, the difficulty in the acquisition of the plural markers and the errors in accusative and dative case assignment point to an incomplete acquisition of the animacy constraint as it applies within these rules and as shown in Table 10. In addition, dative marking in Malayalam has greater semantic complexity, as it straddles several semantic roles including the experiencer, possessor and location.
Our RQ1 was to recover the underlying features that determine the ease/difficulty in the acquisition of Malayalam inflections. We have already seen that there are several morphological and typological factors that contribute to the ease of acquisition of the Malayalam inflections as evidenced by the preponderance of correct and productive use of inflections in general. All the case markers, except the sociative, are high-frequency morphemes and are used across the board with various predicates and nominals. The observed difficulty in the acquisition of certain case markers cannot be attributed to mere frequency effects in the target language. The difficulties seem to stem from an incomplete acquisition of certain grammatical properties such as the animacy constraint (for accusative, dative and plural), the overlapping of semantics of different forms (dative and genitive for possession) and the synchronic loss of aspects of morphosyntactic information (past tense marking with -Tu).
We saw that children have significantly greater difficulty in correctly marking inflections on nouns than verbs (RQ2) as shown in Table 11. We have also seen that monolingual subjects have an advantage over bilingual subjects in correctly assigning inflections on nouns, while no such advantage is discernible when it comes to producing inflections on verbs (RQ3) as shown in Table 12. This difference between verbs and nouns can also be attributed to the morphosyntactic rules that govern each category, since the other factors that contribute to ease of acquisition are common to both lexical classes. While nominal inflectional production in Malayalam is difficult for both sets of subjects, bilingual subjects are seen to have greater difficulty with nominal inflectional production than their monolingual counterparts, whereas no such significant difference in difficulty is discernible with verbs. This observation is strengthened by the fact that the bilingual subjects show a clear, comparative advantage in producing English nominal inflections which do not require reference to animacy compared to their Malayalam productions (Table 13). This cross-linguistic difference in the same children further underscores the relative difficulty of acquiring the animacy constraint that governs many nominal inflections in Malayalam.
As we mentioned earlier, several models have been proposed to account for the errors in inflection acquisition. The prosodic constraints model (Demuth, 1996, 2001) cannot successfully predict the omission errors in early Malayalam, since they are restricted to certain categories and do not occur across all nominal and verbal inflectional categories. While full-competence models (the Truncation model (Rizzi, 1993) and ATOM (Schütze & Wexler, 1996)) can successfully account for the tense omission patterns in the data, neither of these syntactic models can account for the omission errors of the accusative marker, in particular. Both the models predict the omission of inflections on higher nodes such as T, whereas the accusative is assigned by v. Therefore, these morphological errors which are frequent in the nominal productions in Malayalam must arise for other reasons.
As also mentioned earlier in the Introduction, the ease/difficulty in acquiring the inflectional morphology of a particular language is determined mainly by (a) the semantic complexity of the morphosyntactic rules that govern inflectional assignment, (b) the regularity of paradigms, and (c) language typology (Clark, 2016). We see that the different errors that are attested in the early inflectional production in Malayalam can be attributed to one of these three sources. The difficulties in the acquisition of case and number markers can be attributed to semantic complexity with the overlap of case-marking contexts and the grammatical expression of animacy. While the overlap of morphosyntactic contexts is generally construed as complexity-creating across languages, we have seen that the use of animacy as a grammatical constraint plays a significant role in Malayalam and appears to be hard to acquire. Number-marking, for example, does not come in reliably until much later (beyond the ages at which these children were recorded) and accusative case errors persist as well. It remains to be seen whether this difficulty with construing animacy in morphological rules holds cross-linguistically. Errors in past tense marking point to delays in learning the specific morphosyntactic contexts that require a particular affix rather than the default. We have also seen that typology does facilitate the acquisition of inflectional markers in Malayalam, which is especially visible in verbal inflections, and the children’s production accuracy is remarkable. Finally, the patterns in the bilingual data have enabled us to measure typological effects to some degree to show that the rich agglutinative typology of Malayalam does have a significant effect on the successful early productions of verbal morphology. Bilingual children with access to an alternate system appear to learn the grammar just as well or be tripped up by its complexities just like monolingual children.
Footnotes
Acknowledgements
We thank the two anonymous reviewers, and the editors of the journal, Prof. Judith Rispens and Prof. Chloë Marshall, for their valuable feedback. Any remaining errors are, of course, ours.
Author contributions
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) received no financial support for the research, authorship and/or publication of this article.
