2.5 Discussion of grammatical acquisition
2.5.1 Manner of NPx acquisition
2.5.1.1 Insights into potential semantic motivations
Demuth (2003), Connelly (1984), and Tsonope (1993) find that Bantu language speaking children do not make semantic overgeneralisations and that the only NPx errors are errors of commission when a singular may take its plural in one of two classes (for example, in Nguni and Sotho languages, Cl. 9
37 plurals can occur in Cls 10 or 6), or as Connelly (1984) finds, children using Cl. 9 plural (Cl. 10) on nouns in which no plural prefix exists in adult speech. Sibanda (2014) also notes the overgeneralisation of the Cl. 5 plural (Cl. 6) to Cl. 10 nouns (the plural of Cl. 9) which usually have a zero prefix. This suggests that children do not have access to the productive semantics of the noun class system but rather treat it as a grammatical system, suggesting that overgeneralisations are “morphologically motivated” (Connelly, 1984: 147). Connelly (1984) finds that these types of overgeneralisations happen at a stage after which the main aspects of the morphological system have been acquired. This moreover suggests that the child has already reached a stage in which it is using grammatical rules, pointing toward the fact that overgeneralisations are morphological and not semantic.
On the other hand, Suzman (1996) points to the fact that NPx are learnt in a gradual and accurate way, derived from the initial basis of a much simpler gender system in isiZulu. She provides evidence of the overgeneralisation of noun classes when children start off with a binary understanding of the classes, placing ‘people’ in Cls 1 or 1a and all other ‘things’ in Cl. 9 – a semantic overgeneralisation of i- commencing Cls 5, 7 and 9 (Suzman, 1991; 1996; 1999). But, the extent to which this is due to grammatical processing still being assimilated, or due to an accommodation arising from ill-formed phonological processing (Connelly, 1984) is ambiguous. The latter would imply that apparent, potential semantic overgeneralisations of Cl. 9 are not overgeneralisations, per se, but the use of a NPx in a collapsed form – for example, i' tulo instead of. isitulo ‘chair’.
Kunene (1979, in Demuth, 2003) and Idiata (1998, in Demuth, 2003) find overgeneralisation of subject markers in story-telling tasks where human/animate referents that do not belong to Cls 1 and 2 are subsequently referred to with Cls 1 and 2 agreement. Despite this being the norm in adult speech as well, I believe it may suggest a basic understanding, at the very least, of the relatively consistent semantics of Cls 1 and 2.
2.5.1.2 Insights into potential morphological motivations
According to Suzman (1991) the acquisition of NPx in isiZulu is earlier than in Sesotho and Siswati, suggesting that the pre-prefix on isiZulu nouns may facilitate the earlier emergence of NPx as a whole (Demuth, 2003). For example, Cls 1a and 9 in Sesotho have a zero prefix (Ziesler and Demuth, 1995), whereas Nguni languages retain the pre-prefixes u- and i- for these classes respectively. Additionally, Cls 1, 2, 4, 7, and 8 all retain a pre-prefix in isiXhosa and isiZulu, whereas in Sesotho they consist of only the basic prefix. Another reason could be that Setswana and Sesotho speaking children hear NPx less
38 frequently since classes which are empirically important in young children’s speech (Cls 1a and 9) have zero prefixes (Suzman, 1996). Nguni nouns omit the pre-prefix in the vocative29 as well as after negatives, demonstratives, lists of things, and in a number of other environments (see Pahl, 1967: 27 and Suzman, 1991: 42). Certain nouns in Sesotho (from Cls 5, 7, 8, 10, and 14), on the other hand, lose the whole NPx in adult speech when a nomimal modifier is used30 (Demuth, 1988, 1992; Doke and Mofokeng, 1957 in Ziesler and Demuth, 1995) or if agreement is marked elsewhere in the sentence (Demuth and Ellis, 2010). This means Sesotho children may hear full NPx less frequently (Suzman, 1996).31 Demuth and Ellis (2010) suppose this may be due to different syntactic analysis of NPx across Bantu languages: in Sesotho being grammatical function items32 and in other languages being bound morphemes that are listed in the lexicon.
Idiata (2005) summarises the work on Bantu agreement acquisition, noting that it appears that correct agreement markers appear before NPx fully emerge in children’s systems. This suggests that children have access to gender features even before the appearance of NPx. Connelly (1984) elaborates on this by considering the acquisition of plural NPx and suggests that number agreement occurs before children can mark it in noun morphology. His data highlights that number is indeed expressed as a concept in utterances before obligatory plural NPx are in place33 (Connelly, 1984). Number can be expressed through the use of number adjectives (he finds two cases of this) or by means of plural agreement morphemes in the rest of the utterance34 (Connelly, 1984). For the latter, he finds plentiful evidence, with examples indicating the plural subject markers occur not only when the plural NPx is
29 The vocative denotes “a case of nouns, in some inflected languages, used when the referent of the noun is being addressed” (Collins, 2018).
30 Demuth and Ellis (2010) find that in adults’ conversations with 2-3 year olds, adults dropped 20-35% of NPx when the noun was marked with agreement elsewhere.
31 See Demuth and Weschler (2012) for an analysis on acquisition of this phenomenon which is beyond the scope of this thesis given it is not a feature of Nguni languages.
32 Grammatical function refers to the syntactic roles played by morphemes in a particular clause or sentence (Nordquist, 2018).
33 The case of missing NPx with plural agreement could support Suzman’s (1991) claim that subject marker agreement is often incorrect, however, Connelly (1984) does not view this as an ‘error’ as the concords are correct by reference to the context.
34 Note that this does not dispute Demuth’s (2003) assertion that there is no delay in plural classes, due to the way in which plurality is marked in the rest of the utterance.
39 absent but when the singular NPx is absent too (although in one case it is present) (Connelly, 1984).
Connelly (1984) stipulates that the missing singular NPx makes the plural noun both phonologically and morphologically ill-formed, since there must be either a singular or plural prefix of the correct class on any noun stem. A reason for this is unlikely to be that children favour less complex singular forms since in Bantu languages singular and plural forms are equally as complex. It may thus be due to the higher proportion of singular nouns in children’s speech (Connelly, 1984).
2.5.1.3 Insights into potential phonetic, phonological, and prosodic motivations
Tsonope (1993) takes issue with the assumption that children have access to gender features even before the appearance of NPx. He posits that considering NPx as a ‘form’ and noun stems as ‘content’
as a potential explanation for why children appear to acquire noun stems before NPx is problematic.
He argues that 1) prefixes do bring about profound meaning changes, and 2) if the child were aware of a NPx and stem distinction, it would be very unlikely that they would delete the very NPx that they are aware of (Tsonope, 1993). A counter argument for (1) is provided by Sibanda (2014), who argues that if a word takes a variety of prefixes it is likely that a child will hear the stem more often and, confronted with the task of choosing the appropriate NPx, will rather omit it.
Notwithstanding, Tsonope (1993) furthermore believes that a form/content distinction does not explain why monosyllabic stems are acquired with their prefixes on.35 He attempts to explain the three- stage process by drawing on the deletion of weak-initial syllables in English, for example, banana (nana) and potato (tato) (Allen and Hawkins, 1978, 1980 in Tsonope, 1993: 113; Demuth, 2003). In children’s attempts to make trochaic patterns,36 an initial weak syllable must either be made heavy, or deleted (Tsonope, 1993). Demuth (1992) similarly proposes that agreement appears correctly marked before NPx are consistently marked, because children prefer disyllabic feet and thus potentially omit NPx.37
35 See recent work by Demuth and Ellis (2010) suggesting that children might be more likely to use NPx with monosyllabic nouns and that there is a greater likelihood of prefixes being truncated on polysyllabic nouns for children aged 2;3.
36 A trochee is a rhythmic unit consisting of a stressed/long syllable followed by an unstressed/short syllable (Collins, 2018).
37 In Sesotho, both possessives and demonstratives are disyllabic, making this an appropriate hypothesis, however, this is not the case in isiXhosa as most often the common demonstratives and possessives are monosyllabic.
40 Tsonope (1993: 114) accordingly explains the three-stage process of NPx acquisition as noted in Table 2-1 as:
1. Children produce noun stems in a manifestation of a ‘template-based phonological heuristic’;
2. Children increasingly hear noun forms that depart from this shape and in an attempt to correct the mismatch, elaborate on their template with a shadow vowel.
Suzman (1996) elaborates on this prosodic argument by arguing that elision rules in isiZulu (same as in isiXhosa), do not allow for CCV sequences, necessitating even in child speech, the use of a pre-prefix, for example, thatha isicathulo -> thath’ isicathulo ‘take the shoe’ (see full example in Suzman, 1996:
95), which may facilitate the earlier acquisition of prefixes in isiZulu and other languages that have retained the pre-prefix. Suzman argues that her findings of this trend in isiZulu support Peter’s (1994 in Suzman, 1996: 96) proposal that prosody assists children in segmentation (finding boundaries) but also provides a focusing tool for certain aspects of morphological structure. This is likely to assist children in moving from non-productive to productive word combinations and is also linked to the notion of prosodic bootstrapping as described in Section 2.2.4.
Suzman (1996) also suggests that the placement of the noun in the sentence38 will be correlated with use or non-use of the NPx. That is, the NPx is more likely to be used when a noun follows a verb, rather than when it precedes it (Suzman, 1996). This is corroborated by her findings that children used the NPx for Cl. 7 with the consonant in place when following siphi (e.g. siph’ is’tulo) but with only the pre- prefix i- when following the construction nayi ‘here is’ (Suzman, 1991: 68). This suggests children learn from a young age that yi- or i- triggers i- (NPx) and ya- (PM) and that “the subcategorisation frames associated with basic syntactic categories facilitate the acquisition (of new instances) of the category”
(Suzman, 1991: 68).
Demuth (1988) notes that NPx with a nasal tend to be acquired earlier and with more consistency than NPx in which there is no nasal. This is in line with what Jakobson (1968) proposed about the early acquisition of stops and nasals. However, despite Cl. 3 having a NPx with a nasal, Cl. 3 nouns are
38 Variation in isiXhosa sentence structure is common, as is pro-drop, a feature in which the subject noun does not have to be present in the sentence (Gxilishe, de Villiers and de Villiers, 2007a, 2007b; see also Suzman, 1991), but subject markers are always obligatory (Suzman, 1991).
41 reported to be lacking in children’s speech (see Suzman, 1991; Zawada and Ngcobo, 2008). Suzman (1991) supposes that this pattern of class acquisition may be due to the fact that nouns referring to people, common objects, and playthings (words children would be likely to encounter most often) occur in certain classes (like Cls 1, 5, and 9). Zawada and Ngcobo (2008) additionally suggest that due to the morphological similarity between Cls 1 and 3, yet the semantic similarities of Cl. 3 with Cls 5 and 9, children may be confused to the extent that these nouns are simply omitted from their speech until a later stage.
Sibanda (2014) highlights the fact that the phonological complexity of consonants vis-á-vis vowels seemed to be one of the reasons children dropped consonants in her data. Connelly (1984: 79) argues that later in the child’s development some nouns with greater phonological complexity are already
‘learned’ and in use, which tends to illustrate that learning is morphological rather than phonological (although he posits phonology is involved at a “less-important” level, and thus terms the process
‘morphophonological’).
2.5.1.4 Insights into potential morphophonological motivations
Kunene (1979 in Idiata 2005: 88) finds the overgeneralisation of Cl. 11 nouns to Cl. 5, using lunwabu >
linwabu ‘chameleon’ as an example. Gowlett and Dowling’s (2015) research reveals an incipient merger of Cls 11 to 5 in adult speech suggesting this no longer a phenomenon unique to child acquisition. It could be argued, however, that its presence in early childhood development initially may have induced this morphophonological historical change (Demuth, 2003). Dowling and Gowlett (2016: 296) present similar findings that explain the shift of Cl. 11 nouns to Cl. 5 was initiated by “the tendency towards the apparent ‘normalisation’ of the 11/10 gender in this instance to an 11/6 one” and then, by analogy, the shift of Cl. 11 nouns to Cl. 5 (the typical singular of Cl. 6). Possible reasons given are 1) that children are unaware of the distinction in the Cls 5 and 11 NPx lu- and li- or 2) children have an “internalised grammar” in which Cl. 11 concords have their plural in Cl. 6 (Dowling and Gowlett, 2016), reinforcing Connelly’s (1984) view that overgeneralisations are morphologically motivated. Dowling and Gowlett’s (2016: 300) study moreover has bearing on my rural study since they find that there appears to be “no difference between rural and urban children as regards the acquisition of Cl. 11”.
Suzman (1996: 100-101) suggests that shadow prefixes and overgeneralised prefixes are in
“complementary distribution in languages whose noun class systems have different phonological properties” (as Nguni languages and Sotho languages do), simply being alternative morphological
42 strategies of acquisition. She posits that, because in isiZulu all singular prefixes are either i- or u- commencing, children learn a system that is “superficially more transparent” than that encountered by Sesotho-speaking children (Suzman, 1996: 100). She says that isiZulu-speaking children are confronted with a range of i- commencing nouns, which gives them the tendency to overgeneralise the Cl. 9 prefix (or pre-prefix, in line with Herbert’s (1978) analysis that reanalyses the nasal of Cl. 9 as part of the stem). A Sesotho-speaking child, on the other hand, may tackle being confronted with prefixes commencing with different consonants, vowels, or zero morphemes by using a filler syllable identified as a shadow prefix, a partial prefix, or a place holder morpheme (Suzman, 1996). She says that consonant-commencing prefixes make the NPx less accessible to the child, and they resort to placeholder morphemes (Suzman, 1996). This is in line with findings of Tsonope (1993) who observes shadow vowels but no overgeneralisations of the NPx. Results of Suzman’s (1996) study show that the primary determinants of variation in patterns of acquisition across Southern African Bantu languages are the morphophonological properties of surface input.
Manner of NPx acquisition, summary: Overgeneralisations are unlikely to be semantically driven. The acquisition of NPx seems to occur through an interaction of morphological, phonological, and prosodic features, although the extent to which each of these factors dominates, remains unclear. This is in line with Yang’s (2000: 234 in Dowling and Gowlett, 2016: 301) alternative view that child language is not just an imperfect version of adult language but what he terms ‘a variational process’ in which children must adopt assumptions about how language works – assumptions which then change as new evidence is presented from their speech environments.
2.5.2 Morphological and phonological agreement overgeneralisations: insights into