Acquisition of Comparison Constructions

Abstract

This article presents a study on the time course of the acquisition of comparison constructions. The order in which comparison constructions (comparatives, measure phrases, superlatives, degree questions, etc.) show up in English- and German-learning children’s spontaneous speech is quite fixed. It is shown to be insufficiently determined by factors like frequency. Instead, a model of parametric variation in the grammar of comparison is shown to predict the order we observe.

Notes

1. ¹The report of the acquisition study is split between the present article and Tiemann, Hohaus & Beck (2012). Tiemann, Hohaus & Beck focus on one particular construction (pronominal measure phrases like John is that tall) in acquisition as well as across languages. They also provide a more comprehensive overview of the theoretical motivation for the acquisition study. The present article focuses on the differences between English and German than-constituents that the study uncovered. It also provides a more detailed and critical discussion of the significance of the findings of the acquisition study. There is some overlap between the two articles regarding background discussion and description of basic results.

2. ²A good argument for a quantificational analysis of comparatives comes from a scope ambiguity between modal quantifiers and the comparative operator exemplified in (i) and (ii). (See also Heim 2001 for discussion.)

• (i) (Context: The draft is 20 pages long.)

• The paper is allowed to be exactly five pages longer than that.

• (ii)a. The following is allowed:

• The paper’s length is exactly five pages more than 20 pages.

• (allowed taking scope over comparative)

• b. The maximally permitted length for the paper is five pages more than 20 pages.

• (comparative taking scope over allowed)

3. ³We use a simple version of the operator POS that suffices for present purposes. See Stechow (2009) for interesting discussion.

4. ⁴This summary leaves out data types that turned out to be unusable in a corpus study of child language acquisition, i.e., subcomparatives, negative island effects in than-clauses, and differential comparatives. See Beck et al. (2009) and Beck, Hohaus & Tiemann (2012) for a more comprehensive picture, and thus more solid motivation for the parameters proposed.

5. ⁵Calculating the number of possible acquisitional orderings reduces to the following combinatorial problem: We need to calculate the number of possibilities A_N to put N objects in up to N bins, in this case for N = 3. In order to compute A_N, one counts for all fixed k from 1 to N the number B_k of possibilities to put N objects into k bins such that no bin is empty, and then obtains A_N as the sum of B₁ + … + B_N . The resulting formula is

where C_k is the set of all k-tuples n = (n₁, … , n_k) of positive integers with n₁ + … + n_k = N. The multinomial coefficient stands for the number of possibilities to put N objects into k bins such that the j-th bin contains n_j objects.

6. ⁶We should point out that the data collected reflect the acquisition of comparative morphology, which is not necessarily the same as the acquisition of comparative meaning. Ross, for instance, seems to go through a brief phase at age 3;05 in which he uses comparative forms with a nonadult, presumably Positive, interpretation. Here is an example: (i) *FAT: in fact # the fastest ones are the little ones.*CHI: yeah .*FAT: they’re the fastest ones # yeah .*CHI: and the big ones are faster too .*FAT: the big ones are fast too .*CHI: yeah .(Ross, age: 3;5.26, file: 41a1.cha)Otaki (2010) makes a similar observation; see also Ryalls (2000), Gitterman & Johnston (1983), and Ehri (1975). We conjecture, however, that the children get past this stage relatively quickly and then have both a Positive and a comparative interpretation at their disposal. Our assumption is motivated by the scarcity of mistakes in the children’s transcripts and the pattern in which they appear; see also Hohaus & Tiemann (2010) for discussion.

7. ⁷We follow the vision of Roger Brown in that we want to provide an explicit description of the child’s semantic knowledge at each stage. See Brown (1973:56–57) for a passionate discussion of such an approach.

8. ⁸This subsection has only discussed one, most simple, input-driven model. More sophisticated input-driven models of language acquisition might be more successful in accounting for the observed order of acquisition. For instance, in addition to input frequency, alternative models might want to take into account a broader range of input and consider competitors to the target constructions the children are exposed to in the adult input. (This is specifically relevant in the light of recent results discussed in Syrett 2013.) However, such a model would need to be spelled out in much more detail in order to test its potential for explaining the observed acquisition facts.

9. ⁹The presentation is a slight simplification. We should not really talk about predicative versus other adjectives/adverbs, but rather about whether the gradable property that the comparative operator works with is the lexical adjective or not. An attributively used adjective can, besides the data discussed in the text, give rise to a DP-internal reading in which indeed the comparative operates on the lexical adjective. An example is given in (i). This issue is ignored by the presentation in the main text.(i) Mary owns a faster computer than this one.‘Mary owns a computer which is faster than this one is fast.’

10. ¹⁰In terms of their appearance in the adult input in the MacWhinney corpus, predicative comparatives with a phrasal than-constituent are significantly more frequent than attributive comparatives with a than-constituent and than adverbial comparatives with a than-constituent (p < .01 by Pearson’s χ² test). Attributive comparatives with a than-constituent are not significantly more frequent than adverbial comparatives with a than-constituent (p > .1). We thus cannot exclude the possibility that the observed sequencing is an effect of the differential frequency of these constructions in the adult input.

11. ¹¹A possible prediction of adding both -er1 and -er2 to the inventory of comparative operators in English is that predicative comparatives with than-phrases might have a reduced processing complexity when compared to other comparatives because of the simpler analysis available. We have not yet followed up on this prediction.