Avoiding dative overgeneralisation errors: semantics, statistics or both?

Abstract

How do children eventually come to avoid the production of overgeneralisation errors, in particular, those involving the dative (e.g., *I said her “no”)? The present study addressed this question by obtaining from adults and children (5–6, 9–10 years) judgements of well-formed and over-general datives with 301 different verbs (44 for children). A significant effect of pre-emption—whereby the use of a verb in the prepositional-object (PO)-dative construction constitutes evidence that double-object (DO)-dative uses are not permitted—was observed for every age group. A significant effect of entrenchment—whereby the use of a verb in any construction constitutes evidence that unattested dative uses are not permitted—was also observed for every age group, with both predictors also accounting for developmental change between ages 5–6 and 9–10 years. Adults demonstrated knowledge of a morphophonological constraint that prohibits Latinate verbs from appearing in the DO-dative construction (e.g., *I suggested her the trip). Verbs’ semantic properties (supplied by independent adult raters) explained additional variance for all groups and developmentally, with the relative influence of narrow- vs broad-range semantic properties increasing with age. We conclude by outlining an account of the formation and restriction of argument-structure generalisations designed to accommodate these findings.

Keywords:

• child language acquisition
• retreat from overgeneralisation
• entrenchment
• pre-emption
• semantic verb class hypothesis

Acknowledgements

This research was supported by grants from the Economic and Social Research Council (RES-062-23-0931) and the Leverhulme Trust (RPG-158).

Notes

1. A reviewer asked when children typically acquire the PO-/DO-dative distinction. The age at which children acquire the subtly different semantic properties of the two constructions is one of the questions investigated in the present study. But at what age are children able to correctly use and understand the two constructions in a more general sense (i.e., understand that the PO-dative places the theme before the recipient, whilst the DO-dative displays the opposite pattern)? Studies of spontaneous speech (e.g., Campbell & Tomasello, 2001; Gropen, Pinker, Hollander, Goldberg, & Wilson, 1989; Snyder & Stromswold, 1997) suggest that both emerge in production at around 2;0. There is some suggestion that the DO-dative appears earlier, though this likely largely reflects the use of semi-productive chunks such as Gimme X. A recent forced-choice comprehension study (Rowland & Noble, 2010) found that 3-4 year-olds demonstrated more robust performance with PO- than DO-datives, but could still correctly interpret DO-dative sentences when given additional cues (e.g., when the recipient was a proper noun).

2. Baker's paradox is also sometimes known as the 'no negative evidence' problem (e.g., Bowerman, 1988; Marcus, 1993), as the assumption is that caregivers do not provide children with evidence regarding the (un)grammaticality of their utterances (e.g., McNeill, 1966). In fact, this is something of a misnomer, as evidence suggests not only that caregivers spontaneously correct such errors, but also that children are sensitive to this feedback (e.g., Chouinard & Clark, 2003; Clark & Bernicot, 2008; Saxton, Backley & Gallaway, 2005; Strapp, Bleakney, Helmick, & Tonkovich, 2008). However, this type of feedback is unlikely to be sufficient as an explanation of the acquisition of restricted generalizations. One problem is that both children and adults rate as ungrammatical certain uses of low frequency and even novel verbs (e.g., Ambridge, Pine, Rowland, & Young, 2008; Wonnacott, Newport & Tanenhaus, 2008), for which they cannot have received such feedback. A second problem is that such errors - and hence opportunities for feedback - are relatively rare; indeed some children may not produce them at all. Yet Baker's paradox applies equally to children who do not produce such errors, where the question becomes how they avoid them whilst retaining the ability to produce novel utterances.

3. In the corpus analysis conducted as part of the present study, we make the simplifying assumption that any occurrence of a PO-dative constitutes an instance of case where a DO-dative would have been equally felicitous (and vice versa). This is an oversimplification as factors such as the relative animacy, accessibility, definiteness, and length of the recipient and theme affect the relative felicitousness of the two constructions (see Bresnan et al., 2007). However, hand-coding each dative sentence on each of these dimensions would have been not only prohibitively time-consuming, but also—for some dimensions—difficult to do objectively (though see Goldberg, 2011, for a study along these lines). Consequently, the present study constitutes a particularly strong test of the pre-emption hypothesis, because the noisiness of the pre-emption measure that results from this simplification counts against the likelihood of any such effect being observed.

4. For the purposes of these descriptive counts, a verb was classified as PO-only or DO-only if either or both of the two authors listed it as such, otherwise it was classified as alternating. Agreement between the two sources was high with only seven verbs classified as alternating by Levin and PO-only by Pinker (carry, haul, pull, push, schlep, repeat, recount) and two as DO-only by Levin and alternating by Pinker (guarantee, declare). No verb was classified as PO-only by one author and DO-only by the other.

5. The log transformation of frequency counts is standard practice in psycholinguistics as these data follow a highly-skewed non-normal, distribution. It is necessary to transform N + 1, as opposed to N, as some verbs have a frequency of zero, which has no log.

6. Note that, to our knowledge, no researcher has proposed a statistics, with semantics/morphophonology for exceptions model. Such an account would claim that there exist verbs that appear in (for example) the PO-dative construction only (statistics), but whose semantic/morphophonological properties are such that speakers would also consider them to be grammatical in the DO-dative construction.

7. One study that predicted, but did not find, an effect of construction-slot semantics is that of Matthews and Bannard (2010). These authors found that two- and three-year old children showed no increased ability to repeat a familiar string when the relevant slot-filler was semantically similar to the corresponding filler in a high frequency (and hence presumably well-learned) string. For example, considering the familiar string a piece of toast, children made just as many errors when repeating a piece of meat as a piece of brick. However, one important difference from the present study is that the semantic properties of the fillers (e.g., brick), though atypical of the semantic properties of the slot, were not incompatible with them. Another is that the present participants were considerably older; and it would seem likely that 2–3-year-olds have learned less about construction semantics.

8. See Lidz, Gleitman, and Gleitman (2004), Naigles, Fowler, and Helm (1992), and Naigles, Gleitman, and Gleitman (1993) for studies investigating children's interpretations of utterances with mismatching verb and construction semantics.