Acquisition of Derivational Morphology
Abstract and Keywords
Children acquire some derived word forms early, initially as unanalyzed wholes. But from about age 2 onwards, they start to make use of attested derived word-form patterns when they construct new words to convey specific meanings. These spontaneous coinages offer one source of evidence for children’s identification of affixes and their meanings. Two further sources are elicited interpretations of novel words never heard before, and elicited coinages to express novel meanings. All three have been studied extensively for a number of languages. Children’s acquisition of derived word-forms depends on their ability to identify core stems and affixes, on the semantic transparency of the affixes (known meanings), and on productivity (forms well attested in adult speech). Order of acquisition for specific derivational meanings depends largely on adult productivity, and this varies with language typology.
As children begin to master a vocabulary, they typically attempt both simple and derived word-forms from very early on. All early words in production are unanalyzed, so whether they are simple or derived makes little difference. Indeed, among words children pick up early in English are many denominal verbs, with no affixation (e.g. to comb, to bicycle, to fish, to picnic). They also pick up a variety of other complex word-forms in their first years of talking, from breakfast, somersault, and tallest, to mountaineer, teacher, and complicated. In order to study children’s (and adults’) knowledge about derivation, one needs to look at (a) spontaneous analyses of word parts and elicited glosses that indicate understanding of novel word meanings, and at (b) novel coinages in their spontaneous and elicited production of language. Simply tracking and counting all the conventional words children say that contain an affix of some kind does not tell one directly what children know about derivation per se. Children pick up words wholesale as they acquire a vocabulary, so they often make use, early on, of many an established word that contains an affix of some kind, some productive and some not. Compare the affixes in washing-machine, skier, and bicyclist, with those in strength, kingdom, and motherhood. Mere production of established words does not demonstrate knowledge of the parts, here the affixes -ing, -er, and -ist compared to -th, -dom, and -hood (where the latter three are less productive; but see Lieber 2009b). For children, all these words start out as unanalyzed chunks, and are only later unpacked or analyzed into parts. The parts—roots and suffixes, here—must also become linked to a specific meaning for each element.
Knowledge of a derivational affix is impossible to assess from observations of established words alone. Children may recognize fairly early on that some words they use form small families with such near neighbors as in washing-machine, wash, and washing, or skier, skis, skiing, and skied, or bicycle, bicycling, bicycled. Eventually, they should also (p. 425) recognize the -er in skier as being the same as the -er in swimmer, teacher, and writer. How soon do children start to gather such groups of related words together? How soon do they recognize relations of form and meaning among the words they are acquiring? Can this kind of knowledge be readily tapped? The problem is evident when we consider another small family—strength, length, breadth, and health. The suffix -th is no longer productive, and the adjectival forms it was attached to, to construct these nominals, are pretty distant from their adjectival sources—strong, long, broad, and hale. At what age can speakers of English identify -th as a possible suffix? How many instances of nominals in -th do they need to hear? And even then, do they necessarily identify the final segment as a nominal suffix? They may have an easier time, though, with affixes that are currently productive in the language—affixes like English -ie (diminutive), -er (agentive, instrumental), both widely used, or -ness (nominal), -ful (adjectival), or -ly (adverbial), also all well represented in the established lexicon.
In this chapter, I will review children’s knowledge of derivation in word formation, what different studies can tell us about order of acquisition within a language, factors that influence that order of acquisition, and consistent differences in what children understand compared to what they can say.
24.1 Acquiring Derived Forms
What do children know about the morphemes that contribute to word meanings as well as forms? One source of information is what children have noticed about the words they hear or produce, typically in the form of spontaneous analyses of words into parts, for example, A horsie is a little horse. Such analyses show that children can distinguish the root or core word stem from the affix (horse+ie), but their spontaneous analyses are not particularly frequent although they do indeed occur (see, e.g., Clark 1978, Slobin 1978). They also depend on the child’s knowing enough language to be able to offer a gloss.
Another approach to assessing what they know has been to present children with unfamiliar, novel words, and then ask them what these might mean. If the children can analyze the word-forms they hear into root and affix, say, and if they understand the meaning of the affix, they can often offer a gloss or interpretation. For example, asked what a boy who’s a jumper could be, one 3-year-old readily replied “He jumps things.” Notice that to arrive at this interpretation, the child must removed the affix -er, turn the root back into a verb with an added 3Psg inflection, and then supply it with a singular subject, he, and an indeterminate direct object, things (Clark and Hecht 1982). This suggests that this 3-year-old understands that -er marks the agent of an action of jumping. If children can identify affixes as adding specific meanings to root word-forms, this provides clear evidence that they are able to analyze the forms in question. Then, in addition, if the same children can make use of the affix -er to construct innovative verb-root+affix combinations themselves, to express new meanings, we have further evidence that they have acquired the derivational suffix -er as an agentive suffix (p. 426) (see Clark and Hecht 1982, Mulford 1983, Clark and Berman 1984, Clark 1993). Once children can do this, these affixes are productive for them, and their meanings can contribute to a growing family of words that share that affix as part of their meaning. So in English, the word-shape Verb+er offers a template for constructing nouns from verbs to mean “the agent of the action denoted by the verb.”
In short, both the ability to show understanding of an affix in unfamiliar words and the ability to produce new word-forms with that affix added to other verb-roots provide strong evidence that children have acquired the relevant knowledge about deriving new agentive nouns in English. And this provides stronger evidence for knowledge acquired than merely saying or repeating an established word that happens to contain a particular affix.
24.2 Order of Acquisition
English-speaking children appear to pick up several productive affixes very early. The first to emerge in their speech, well-modeled by parents and other adults, is the diminutive -ie, added to monosyllabic nouns, as in horse/horsie, bear/bearie, dog/doggie. This is supported perhaps by the /-i/ suffix on such established forms as kitty and puppy, as well as the address terms mommy/mummy and daddy (e.g. Ferguson 1964). Diminutive forms in other languages also tend to be acquired early by young children, probably because of extensive uses of diminutive forms in parental speech (see, e.g., for Dutch: Gillis 1997; Finnish: Laalo 1998; Greek: Stephany 1997; Italian: Bates and Rankin 1979; Quiché Mayan: Pye 1986; Polish: Haman 2003, Dabrowska 2006; Russian: Kempe et al. 2003, Kempe and Brooks 2001; Spanish: Melzi and King 2003, King and Melzi 2004; see also Dressler 1994). However, very young children may not use diminutives productively: rather, they acquire diminutive word-forms, and only later come to analyze them and use the diminutive affix productively (see Svaib 1993).
The English adjectival suffixes in -y and -ed also emerge early on, as 2-year-olds use innovative forms to describe both inherent properties (e.g. crowdy [= crowded], nighty [= dark]) and temporary states (e.g. crumbed [= covered in crumbs], floured [= covered in flour], fastened up [= speeded up]) (Clark 2001). One sees a similar use of past participial adjectives for resultant states in French children’s early uses of c’est+PP (Veneziano and Clark 2012).
Next in line to emerge in English is the agentive (and instrumental) suffix -er. This appears with regularity first only in agentive nouns like cooker (for conventional cook) or lifter (for “someone who lifts things”), and then in later-emerging instrumental nouns as well, such as the innovative climber (for ladder) or hanger (for hook) (Clark and Hecht 1982, Clark 1993). The next derivational morpheme to emerge is the negative prefix un- on verbs, used to convey the reversal or undoing of an action (lock/unlock) or the meaning of privation (saddle/unsaddle).1 However, children take some time before (p. 427) they make use of negative un- in their own innovations (Clark et al. 1995). By age 4 to 5, English-speaking children begin to master another agentive suffix, -ist, used in such child forms as trumpetist (for trumpeter, in parallel to such forms as pianist, flutist, and violinist (see Clark and Cohen 1984).
Most other derivational affixes in English are not acquired until after children enter elementary school, and some are only mastered once children have access to another source of information besides speech about the relations between such pairs as danger–dangerous, library–librarian, magic–magician, or volcano–volcanic (e.g. Deacon and Bryant 2005, 2006). Stress-shifts with their concomitant changes in the value of the affected vowels, as well as consonantal shifts from /-k/ to /-ʃ/ make the relations between base words and such derived forms as adjectives in -ic, -al, or -ous, or nouns in -ation or -ity, for example, hard for children to discern (see, e.g., Lewis and Windsor 1996, Jarmulowicz et al. 2008).
What children acquire when also depends on which language they are learning. In some languages, derivation is king for the construction of new words. In others, derivation shares that role with compounding, and in others still, compounding provides the major option for lexical innovation. What is favored in any particular language depends largely on typology. For example, in Germanic languages, we see both compounding and derivation used in the construction of new words, with compounding somewhat favored over derivation, but in Romance and Semitic languages, derivation is heavily favored for most lexical innovations. One result of this is that the productive options displayed to young children in the language they hear from adults will differ from one language to the next, and this affects the kinds of lexical innovations observable in children.
Finally, several factors appear to affect what children learn first, and hence their general course of acquisition for constructing new words. One that has already been mentioned is productivity, here the productivity of the affixes available. In general, more productive affixes are more frequent—they occur on more stems (types) and hence appear on more tokens in everyday speech. This should make them more available for children to identify and analyze in the adult speech addressed to them. But how to measure productivity remains an issue. One approach has been to look at the frequency of different affixes in the established lexicon (e.g. Baayen and Lieber 1991), but while this offers a good record of past productivity extrapolated to current usage, it may not measure the particular word-forms that adult speakers currently favor in their innovations. Another approach has been to look at normative recommendations for which word-formation pattern(s) to use for specific meanings, but speakers often depart from such recommendations (e.g. Berman 1987). And another has been to focus on those word-forms adult speakers actually use in their lexical innovations, spontaneous and elicited, viewed as natural productivity (see Clark and Berman 1984; see also Schultink 1961).
Two further factors are semantic transparency and formal simplicity. Before children can make use of any derivational affixes in their own novel word formations, they need to have identified an affix as an affix, and analyzed its meaning in the forms (p. 428) where it appears. Once they have done this, children can add that affix to their repertoire of devices available for use in the construction of new words with the relevant meaning-type. In other words, the affix has become transparent for the child in the meaning it adds to a stem, and hence available for constructing a new word-form to convey a new meaning. Simplicity of form in new words is also a factor early on. If children can construct a new word with minimal, or even no, changes in the form of the source word, they do so. This seems to make some word-formation devices highly accessible very early, with the result that children learning languages with conversion make early use of that option in their lexical innovations.
24.3 Using Derivational Affixes in Lexical Innovations
As children learn words, they gradually build up a vocabulary. At age 2, they have typically mastered between 200 and 600 words for production (a good number more in comprehension). Their repertoire gets steadily bigger as they get older, and by age 6, their comprehension vocabulary is typically estimated to be around 14,000 words, with somewhat fewer in production (Clark 2009). But children often try to talk about things for which they have yet to acquire the conventional terms, so they coin new words on such occasions that later yield to established adult forms. These coinages, both spontaneous and elicited, offer invaluable information about what children know about the internal structure of words and about the meanings of the component roots and affixes.
In some languages, conversion or derivation without any affixation added (sometimes called zero derivation) provides a ready means for constructing new verbs from nouns, as in English. Young children make early use of this option in some languages, possibly because the change of word class without any affixation is so simple. Such verbs, for adults, fall into a small number of categories; within each category, they lie on a continuum from well-established verbs at one end to innovations at the other; and to use denominal verbs, speakers of English rely on a convention that places constraints on the actual process of innovation (Clark and Clark 1979). Children acquiring English readily form new verbs from nouns, as shown in Table 24.1, as do children acquiring other West Germanic languages like German, Dutch, and Swedish (Clark 1982, 1993). In all four languages, children as young as age 2 favor, as the source nouns for new verbs, words for instruments, objects (locata), and locations—common adult categories of denominal verbs in those languages. (p. 429)
Table 24.1 Some innovative denominal verbs
1 S (2;4, wanting to have something weighed): you have to scale it first. [= weigh]
2 S (2;7, having hit his baby sister, explaining why she was crying): I broomed her. [= hit with a broom]
3 D (2;9, of a sock): did you needle this? [= mend]
4 D (2;9.10, after mention of boats): and we might see a man oaring a boat with oars. [= row]
5 S (2;11, not wanting his room to be cleaned): don’t broom my mess. [= sweep]
6 D (3;2.9, picking up a Cuisinart blade from the sink)
7 DH (2;3, talking about getting dressed): Mummy trousers me. [= puts on my trousers]
8 J (2;6, asking a teacher to toss a pillow at him in a mock pillow-fight): pillow me! [= throw a pillow]
9 EB (3;4, deciding not to wear her new nightgown outside in the patio): mine will dust.
10 CB (3;11): I’m crackering my soup. [= putting crackers into]
11 CA (4;6, waving a funnel at her younger sister): I’m going to funnel you. Ff, ff, ff. You’re all in there.
12 SO (~5;0): I’m going to basket those apples. [= put into a basket]
13 CB (5;5, asking Mo to stop her sister from putting popcorn “beads” on her thread): Mom, will you keep Eva from threading on mine?
14 K (7;0, asking Mo to wrap a towel round her as she got out of the bath): towel me, mommy.
In other languages, children make much less use of the conversion option and, if they use it, do so only at a later age. For example, children acquiring French do construct some new denominal verbs, always placing them in class 1 (verbs in -er, the class that accounts for around 90% of all French verbs), but this is relatively rare before age 4. Some typical examples include innovative pincer ‘to paint,’ from une pince ‘paintbrush’ (used in place of adult peindre); mètrer ‘to measure,’ from un mètre ‘ruler, tape-measure’ (used in place of adult mesurer), or pantoufler ‘to put on slippers,’ from une pantoufle ‘slipper’ (see Clark 1982).
Children acquiring Romance languages make little use of conversion to form new verbs, but they do rely on it rather earlier in forming new nouns (from verbs), as in le tomber ‘the fall/ event of falling,’ from tomber ‘to fall,’ or le jeter ‘the throw/ event of throwing,’ from jeter ‘to throw.’ Most nouns from verbs like these (p. 430) denote activities or events, and are not common among children’s early coinages (Clark 1993). Although Hebrew has historically relied on present-tense (benoni) verbs as a source of nouns, as in /ʃofet/ ‘a judge; judges-3Psg’ or /mocec/ ‘a pacifier; sucks-3Psg,’ young children make relatively little use of such conversion in their early coinages (Berman 1985). In Hungarian, though, they do construct some early nouns from verbs and adjectives, as in savanyú ‘a grape,’ from the adjective “sour” (for adult szōlo), or farag ‘knife,’ from the verb “to carve” (for adult kés) (MacWhinney 1978).
Although conversion would appear to offer children a very simple way to construct a new verb or a new noun from an existing word, its availability to young children depends on how productive this option is in the language. Moreover, as an option for new-word formation, it is not equally available across languages. Some languages make extensive use of it, as English does—for constructing new verbs—but others seldom use it, instead favoring derivational affixes.
24.3.2 Derivational Affixes
Before children can start to make use of derivational affixes in constructing new words with the relevant meaning added by the affix, they must learn to identify affixes apart from the stems they appear with in established words, and analyze their meanings. How soon children do this depends on the frequency of the established words containing each affix, and on how productive that affix is in adult speech. Some derivational suffixes emerge in children’s production early on, typically beginning between age 2 and 3, while others are still not mastered at age 10.
Overall, languages are biased towards suffixes (Cutler et al. 1985, Hupp et al. 2009). Most inflections appear as suffixes, as do most derivational affixes. This probably reflects a processing preference or bias, whereby the core word or stem is readily recognized from its onset, especially when the onset is not obscured by any prefixed material (see Kuczaj 1979, Cole et al. 1989, Lahiri and Marslen-Wilson 1991). Once a core stem has been identified, it should then be easier to process any modulations of meaning contributed by derivational suffixes added onto the end of the core stem. The presence of prefixes, however, may well delay children’s recognition of the pertinent core word or stem, and so make it harder for them to identify both the core forms and specific prefixes (see, e.g., Pye 1986, Mithun 1989).
126.96.36.199 Diminutive Suffixes
Diminutive suffixes appear early, but mainly in conventional diminutives such as English kitty, dollie, and birdie, as well as terms like doggie, horsie, and bearie. Children sometimes notice the diminutive ending and extend it, as when Leopold (1949) observed his daughter, aged 3;3, adding the English diminutive -ie to a variety of nouns not normally so modified, for example, wallie (wall), chairie (chair), lappie (lap), and (p. 431) even the plural books-ies (vol. 4: 45). In my diary data, I observed similar uses, as in the following exchange:
D (3;6.25, out of the blue one morning): These are my pantsies. D’you know what I call a cat?
Fa : What?
D : A cattie. D’you know what I call a bear? A bearie!
D’you know what I call a dog? A doggie!
But when asked what they could call a “very tiny wug,” Berko (1958) found that few children aged 4 to 7 could supply a diminutive ending. In a more extensive elicitation task, Svaib (1993) found that 4-year-olds produced significantly fewer diminutives in -ie than 5-year-olds or adults (24% vs. 48%, 49%). When all the diminutive affixes used were counted, 4-year-olds still produced fewer (32%) than 5-year-olds (62%) or adults (83%). An analysis of diminutive usage in four children in CHILDES showed that they used very few diminutives as they went from age 2 to age 4, and they used them on only a very small number of lexical items. But when Svaib asked 4-year-olds to pretend they were talking to a 2-year-old as they provided labels for pictures, they produced many more diminutive forms (49%, up from 24%). Diminutive use may therefore be largely register-dependent in English, with 4-year-olds able to form diminutives in the appropriate context.
Children acquiring Italian show mastery of several diminutive suffixes by age 2;0, but they use them only for small size, and take considerably longer to master other uses to connote value, affection, humor, and sarcasm in Italian (Bates and Rankin 1979, King and Melzi 2004). Diminutive affixes are also very frequent in languages like Russian, and some researchers have argued that their use in adult speech to children also aids in the acquisition of noun gender (Bogoyavlenskiy 1973, Kempe and Brooks 2001).
188.8.131.52 Agentive and Instrumental Suffixes
Suffixes for agentive meanings start to emerge in English in children’s innovations around age 3, and become moderately well established by age 5. However, their acquisition continues well into the school years as children learn to read and make use of morphological information to interpret unfamiliar words (e.g. Sterling 1982, Wysocki and Jenkins 1987, Tyler and Nagy 1989, Windsor 1994, Lewis and Windsor 1996, Deacon and Bryant 2005, 2006, Roy and Labelle 2007, Rabin and Deacon 2008). The use of affixes to express agentive meaning in words newly constructed lags behind children’s earlier, extensive reliance on compounding for this purpose (Berman 2009). Indeed, in English-speaking children, compound nouns typically account for 70–80% of their innovative nominals up to age 4 (Clark 1993, Becker 1994).
While agent nouns in -er start appearing quite frequently in spontaneous innovations around age 3, instrumental nouns with this suffix emerged a few months later, as shown in Table 24.2. Earlier, English-speaking children strongly favor root compounds made up of two nouns, as in plánt-man (3;0, for ‘gardener’) or fíretruck-man (2;8, for ‘fireman’), (p. 432) (p. 433) in their spontaneous innovations and in elicitation tasks for both agent and instrument nouns. But around age 4, in elicited forms, they shift to reliance on the -er suffix, first for agent nouns, and somewhat later for instrument nouns (see Clark and Hecht 1982).
Table 24.2 Some innovative agent and instrument nouns
1 D (2;5.26, as he reached across the kitchen counter): I’m a big reacher.
2 D (2;8.1, reading In the night-kitchen): where’s the cookers? [= cooks]
3 D (3;2.17, 3;2.16, after knocking a Lego piece off a boat being built)
4 D (3;2.20, poised to roll up bedside sheepskin rug): I’m going to roll up this rug.
5 D (3;5.27): did you know Gabriel is a stealer? sometimes he steals Judy’s things.
6 D (3;6,23, playing with some tiny dolls): and these are the guarders, in case bad guys come.
7 D (3;7.1, talking about being on the blue chair at preschool the week before): guess what.
8 D (3;8.13, after story about Oliver Pig): Pigs are really good throwers.
9 D (3;10.26): someone has been messing up my room.
10 D (3;2.28, playing with Lego): hey! this is another kind of plane. it has two shooters. [= guns]
11 D (3;3.3, after putting staples into a piece of paper): take these staplers out of here! take these staplers out of here.
12 D (3;4,8, adding Lego blocks to a Lego wall around a field with a Lego donkey in it) these are lockers which lock the donkey up in the farmyard.
13 D (3;6,22, smoothing out a piece of paper): I have a stamper for making things smooth.
14 D (3;9.4, looking out at the balcony where he’d been using pliers the day before): oh, my pinchers are outside.
15 D(3;9.17, talking about someone’s glasses that broke): it has only one (gesture to ear)—holder.
16 D (3;9.18, looking at a deep valance on the coffee-shop window at the airport): Mummy, why do they have that big shader?
17 D (3;9.19, watching as a mechanic placed two chocks by the plane wheels): why did he put two loggers?
18 D (3;9.29, taking about the paper-punch Mo had been using): this isn’t real, the puncher. it isn’t a real toy.
How soon children begin to rely on affixes for new-word formation depends in part on the typology of the language. Children acquiring French start to produce a variety of affixes in their lexical innovations between age 2 and 3 (e.g. Aimard 1975). The same holds for Hebrew, Russian, and Polish, where children make early use of a variety of affixes in their coinages (e.g. Gvozdev 1961, Chmura-Klekotowa 1964, 1971, Berman and Sagi 1981). That is, derivational affixes are favored in some languages, while compounding is favored in others. Over and above that, adult speakers prefer some options over others within a language as they coin new words. These choices can also change over time (see Schultink 1961, Dubois 1962, Chmura-Klekotowa 1970, Adams 1973).
Children also rely on the suffix -er in synthetic compounds in English, but they have some difficulty in learning just how to construct such compound forms. The difficulty here appears to result from the mismatch in word order for the verb and for the noun denoting the object-affected in predicates, compared to in synthetic compounds. Compare He builds walls with a wall-builder. Children learn first to attach the suffix -er to the appropriate core element (the verb), and produce forms like the dríver-wagon, or the búilder-wall, before they master the conventional word-order for the wagon-driver or the wall-builder (Clark et al. 1986). Children acquiring other Germanic languages like Dutch and German also seem to learn where to place affixes first, and only later master the word order within such compounds, and they make similar errors in word order (Clark 1993). Interestingly, the same children who make word order errors in production consistently treat the second element in such compounds, in comprehension, as the head noun. So, when given a compound like ball-hitter, they readily gloss it as meaning “someone who hits balls,” but if given a form modeled on one of their own productions, for example, a thrów-button (based on a púll-wagon, a form elicited from the child for “someone who pulls wagons”), they still take the second noun as the head and therefore gloss it as referring to “some kind of button” (Clark 1984). With instrument nouns, 4- and 5-year-olds make similar errors in production, producing innovative forms like fíx-bike, followed next by fíxer-bike, for adult bíke-fixer.
In English, -er is one of the earlier affixes acquired. Other affixes for constructing nominals like -ness and -ity, adjectival affixes like -ous, -al, and -ic, and verbal suffixes like -ize and -ify, all appear only considerably later in children’s speech, and only occasionally appear in spontaneous coinages. For the most part, children appear to start learning these Latinate suffixes and identifying the meanings they add to the roots they attach to only after they begin to read (see, e.g., Tyler and Nagy 1989, Lewis and Windsor 1996, Deacon and Bryant 2005, 2006).
How soon do children master agent and instrument forms in other languages? Children acquiring Icelandic appear to go through very similar stages to those acquiring English: after an initial stage where root compounds dominate, children begin to add the agentive suffix -ari to verb roots instead, and by age 4, make this their dominant (p. 434) choice in elicitation tasks (Mulford 1983). Instrument nouns in -ari come in a bit later, in part because of competition with root compounds and conversion forms that are also frequently used for instrument nouns in Icelandic.
In French children produce innovative agent nouns with the suffix -eur at age 3 or even earlier. Indeed, they add -eur to both noun and verb roots, and among verbs often make errors early on in the form the root should take. Among the examples cited in various diary records, for instance, are le cerceaunier (3;0) for ‘the one who mends hoops’ (from the noun cerceau ‘hoop), le dormeur (3;8.14) for ‘the sleeper/the one who sleeps’ (from the verb dormir ‘to sleep’), le crêmeur (3;8.11) for ‘the eater of cream’ (from the noun crème), un salisseur (3;11) for ‘someone who dirties things’ (from the verb salir ‘to dirty’), le fermeur (5;6.10) for ‘the one who shuts things’ (from the verb fermer ‘to shut’), or un montagneur (4;6) for ‘someone who likes climbing’ (from the noun montagne ‘mountain’; cf. adult alpiniste). In fact, among the suffixes used in spontaneous innovations in French, agentive -eur accounts for some 40% of children’s innovative nouns (Clark 1993). In an elicitation study (based on Clark and Hecht 1982) with slightly older children, Seidler (1988) found that children between 4;6 and 6;8 consistently relied on -eur in constructing both novel agent and novel instrument nouns over 90% of the time. While their choices of verb stem for these coinages identified the appropriate stem over 80% of the time with class-1 verbs, they identified the correct stem only about 55% of the time for class-2 verbs and between 30% (instruments) and 50% (agents) of the time for class-3 verbs. So one difficulty children face in forming new agent and instrument nouns in a Romance language is identifying the appropriate stem for the innovative word (see also Lo Duca 1990, for Italian).
Finally, in Hebrew, for both agent and instrument nouns, some patterns in word formation are more productive than others. Children attend to relative productivity in adult speech and appear to identify first those affixes and word-formation patterns that are more productive. They readily interpret new agent and instrument nouns, but favor one suffix overall, -an, when they produce both new agent and instrument forms. And, just as with other languages, their comprehension of new forms runs ahead of their own ability to produce them (Clark and Berman 1984).
184.108.40.206 Prefixes Come Later
In word formation as in inflectional development, children attend to and make use of suffixes before they begin to use prefixes (Kuczaj 1979, Hupp et al. 2009). In languages with a variety of derivational affixes, children usually master several suffixes before they make much use of any prefixes. And their first coinages with prefixes tend to be negative verbs, verbs for talking about the reversal of an action. For example, English-speaking children start to use un- spontaneously for reversal around age 3. A selection of such uses appears in Table 24.3. But it typically takes them several years to identify the conditions on the use of un- in terms of the categories of actions that can be undone or reversed, compared to those that cannot. Among the latter are actions like break, run, scratch, or love. They also have to keep track of which reversal verb forms rely on negative particles like out or off, rather than the prefix un-, and which have suppletive reversal forms, as in bury ~ dig up, or hide ~ find (Clark et al. 1995). (p. 435)
Table 24.3 Using un- as a verbal prefix to talk about reversal
1 D (3;0.4, as climbed over Mo’s lap): oh, your hair is unpinning.
2 D (3;1.5, after Mo proposed that they go upstairs and D started to take his blocks off the bedside table and put them into a bag for carrying up): first I unbuild it, okay?
3 D (3;1.12, looking at pencil marks on a page that a family friend was proofreading): are you going to ungrow these? [= erase]
4 D (3;4.3, at the table, D was filling a sock with nuts from his tin in a brief spell of pretending to be Santa Claus, then he switched roles to become the recipient): I don’t know what’s in my stocking. (feels it) I’ll have to unhang it.
5 D (3;5.9, to Mo): show me how you uncatch your necklace.
6 D (3;8.1, of his pyjama top, to Fa): can you untake this off?
Children acquiring English and German typically rely on a general-purpose verb like open or aufmachen in their earliest attempts to talk about reversals. These are then replaced by reliance on negative particles and, in English, also by the prefix un-. Children acquiring French pick up dé- as a reversal prefix fairly early, before age 3, and use it in such forms as débâtir ‘unbuild’ and dégrandir ‘[get] unbig = shrink.’ But German (p. 436) lacks a reversal prefix and relies instead on particles like aus and ab to express a reversal meaning, with occasional uses of the prefix ver- to convey that an action was mistaken in some way. Table 24.4 gives some typical examples of children’s reversal forms in French compared to German (see Clark 1981).
Table 24.4 Innovative reversal verbs in French and German
make less hot
make less cold
make less tired
se restaurer, se reposer
free, let go
remove with tongs
(no conventional adult form)
kneel too long
give x to the wrong person
irren im Geben
run too much
sich verklingeln (6;3)
ring the wrong doorbell
(no conventional adult form)
Source: Based on Clark (1981).
24.3.3 Derivational Affixes in other Languages
Children acquire derivational affixes starting at different ages in different languages. Their first uses of affixes tend to be unanalyzed uses, where the affix is simply part of the word. So acquisition is better assessed by looking at productive uses where children coin new words with an affix, to fill some current gap in their lexicon. Such coinages may be illegitimate from the point of view of the adult’s conventional lexicon because there is a word for the particular meaning intended. Such coinages are fairly soon pre-empted by the conventional word. But in assessing what children know about word structure and affix meaning, such coinages are invaluable because they demonstrate what children have analyzed so far.
For some languages, there are extensive observations on what children do and do not know at particular ages. For Slavic, there are observational and structured elicitation studies in Polish (e.g. Chmura-Klekotowa 1970, 1971, Haman et al. 2009), and in Russian (e.g. Gvozdev 1961, Ushakova 1969, 1976, El’konin 1973). For Romance, there are mainly observational studies of French (e.g. Descoeudres 1922, Grégoire 1947, Méresse-Polaert 1969, Aimard 1975, Seidler 1988); Spanish (e.g. Kernan and Blount 1966, Montes Giraldo 1976); Portuguese (e.g. Costa 1976, Mediano 1976, Figueira 1984); and Italian (e.g. Lo Duca 1990). For Germanic, there are both observational and elicitation studies of several languages, including Dutch (e.g. de Vooys 1916, Tinbergen 1919, Kaper 1959, Schaerlakens 1980, Elbers 1988); German (e.g. Stern and Stern 1928, Augst et al. 1977, Panagl 1977, Brekle 1978); Icelandic (e.g. Mulford 1983), and Swedish (e.g. Gustafsson 1979, Söderbergh 1979). Finally, there are fairly extensive observational studies and some elicitation in Finnish (e.g. Bertram et al. 2000c) and Hungarian (e.g. MacWhinney 1978), as well as in Hebrew (e.g. Berman and Sagi 1981, Berman 1982, Clark and Berman 1984).
Even for closely related languages, there are differences in how readily children can discern and analyze root vs. affix. This depends in part on phonology, which can obscure or make salient the morpheme boundaries within words. Languages also differ in how much they favor derivation over compounding, or whether they rely exclusively on one or the other. And they differ in just what is productive for contemporary (adult) speakers.
24.4 Comprehension versus Production
In word-formation, one persistent finding is that children understand and offer appropriate glosses for word-forms before they are able to construct those target forms for themselves in coining new words. This is evident in the ability of 3-year-olds to provide good glosses for novel agent and instrument nouns in -er a good year before they start (p. 437) to produce novel forms in V-er when asked (Clark and Hecht 1982; see also Lewis and Windsor 1996, for data from rather older children).
Indeed, when 3- to 6-year-olds were given a series of novel synthetic compounds to gloss, they were highly successful in identifying the head (final) noun as the element that identified the kind of category being talked about. When the same children were next given definitions and asked to come up with new words for the relevant agents and instruments, they produced many word-order errors, typically placing what should have been the head in initial, instead of final, position. When typical errors from each child were used as templates for a further set of novel word-forms for that child’s interpretation, children consistently chose the second or final element in these compounds as the head (Clark 1984). In short, their comprehension was well ahead of their production, and the discrepancies between comprehension and production revealed both quantitative and qualitative differences.
24.5 Factors Affecting the Construction of New Word-forms
Two factors appear to be at work in children’s earliest coinages: simplicity of form and transparency of meaning. Simplicity of form captures the fact that children’s earliest innovative word forms usually involve minimal changes to the root or stem form. For example, young children acquiring English or German rely on conversion before they make use of any derivational affixes. (This applies, of course, only to languages that license conversion.) It suggests that children initially prefer to make the fewest possible changes in form. At the same time, they rely heavily on known meanings—stems and affixes that are semantically transparent—in their construction of new word forms (e.g. Clark 1982, Clark and Hecht 1982).
Since languages differ in the word-formation options that are available and productive, children typically follow rather different routes from one language to another as they master the derivational options available for coining new words. Simplicity, transparency, and typology all play a role here, but a further factor, maybe the most important overall, is the (relative) productivity of each option for expressing a specific meaning-type. Adult speakers typically favor some options over others, e.g. English -er and Hebrew -an for expressions of agency, and these will therefore appear more frequently in coinages. But such preferences also change over time (e.g. Dubois 1962, Chmura-Klekotowa 1970, Guilbert 1975).
24.5.1 Productivity in Acquisition and Language Use
Productivity can be characterized in terms of formal constraints on affixes—which ones can be added where (e.g. Aronoff and Schvaneveldt 1978), and the ordering of (p. 438) multiple affixes on the same root. The focus here is primarily on word-form rather than on the meaning being conveyed (Cutler 1980, Lieber 1980). The fewer the constraints on the form-type, the more frequently speakers are likely to use both types and tokens in the lexicon. Frequency is important for acquisition because it is correlated with children’s exposure to particular word-form types (and tokens) and hence to the kinds of meanings each type expresses. But this approach does not readily distinguish between what is already “in” the lexicon, and what constitutes an “innovative” use constructed on a particular occasion to express a particular meaning.
Another view of productivity comes from normative or prescriptive accounts. These measures typically stem from the recommendations of an Academy about lexical usage and, in particular, about new-word uses in a language. And recommended word-types for lexical innovations may be disseminated through media and school systems, as in Israel, France, Sweden, among others (see Clark 1993).
Natural productivity represents a third perspective (Clark and Berman 1984). This measure is based directly on the lexical innovations adults produce (see also Plag 1999, Bauer 2001). It tracks the patterns and word-formation options currently favored by the speakers of a language. This view of productivity does not necessarily coincide with normative productivity since speakers may be unaware of, or disregard, prescriptive recommendations. Rather, it reflects actual current usage by the speakers of a language. This conforms to Schultink’s (1961) definition:
Onder produktiviteit als morfologisch fenomeen verstaan we dan de voor taalgebruikers bestaande mogelijkheid… onopzettelijk een in principe niet telbaar aantal nieuwe formaties te vormen. (By productivity as a morphological phenomenon, we therefore understand language users’ option …of coining an in principle uncountable number of new [word-]formations…)
(Schultink 1961: 113)
While adult coinages generally observe structural constraints on word-forms, the meanings they express with these lexical innovations may fill gaps or add finer specifications to the form–meaning combinations already available in the conventional lexicon. But there is no a priori limit on possible lexical innovations (see also Clark and Clark 1979). Note also that speakers do not necessarily rely on the least structurally constrained form-types more often than more constrained ones. Finally, as already noted, the preferences captured by natural productivity can change over time, but they leave traces from the past in the conventional lexicon. The frequencies of derivational patterns in everyday speech may therefore diverge from the frequencies calculated from corpora based on dictionaries or word lists. The latter are best considered as a cumulative record of productivity over time, rather than a reflection of speakers’ current preferences. For current (natural) productivity, one needs data on preferences by coinage-type for specific kinds of meanings.
Frequency matters here because children are attentive to it and track the frequency of forms in the language around them. This applies just as much to derivational affixes as to word stems, as shown, for instance, by children’s acquisition of agentive and instrumental affixes in Hebrew (see Clark and Berman 1984). Children are also (p. 439) sensitive to the relative productivity of affixes with closely related meanings, and they acquire the most productive affixes—the most frequent and typically the least constrained structurally—before less productive ones in the same domain. For example, children master English agentive -er before -ist, and they master both of these before -ian (Clark and Cohen 1984). More productive affixes appear to be easier to acquire: they are more frequent, more easily analyzed for meaning and form, and more readily retrieved from memory. But this can vary with just what meaning the child (or adult) is trying to convey.
Overall, there are some clear cross-linguistic similarities in children’s acquisition of derivational affixes for use in new words. They tend to acquire suffixes before prefixes. In doing this, they first identify the meanings of a number of suffixes and start to make use of them before they acquire any prefixes. This may well be because the status of a suffix is easier to establish in relation to word stems or roots than is the status of a prefix. Because words are processed from beginning to end, it is easier to identify words with suffixal material as words (Kuczaj 1979, Cutler et al. 1985, Cole et al. 1989, Lahiri and Marslen-Wilson 1991, Hupp et al. 2009). This could delay children’s identification of prefixes as forms added to a stem, and so also delay children’s identification of their meanings.
Children rely on simple options early on, with conversion when it is available. But this is used only for certain types of coinage. For example, in English, making new verbs from nouns, or in Hebrew, making new nouns from verbs (Clark 1982, 1993, Bolozky 1999). At the same time, children exhibit a strong tendency to rely on known elements as they construct new word-forms for new meanings. As a result, they go with transparency in their coinages. In doing this, they also tend to favor the most productive option available with the requisite meaning.
(1) This prefix also appears on adjectives to express the opposite semantic value, as in happy vs. unhappy.