Bare Conditionals: Parallelism

It has been found that quantifier types are related to donkey sentences, and this section is to see whether parallelism makes a difference in readings of bare conditionals. According to Cheng & Huang (1996), the unselective binding approach attributes bare conditionals to universal readings; therefore, bare conditionals, no matter in parallel or not, were coded with

universal readings.

To see whether parallelism was an influential factor, an overall between-type comparison was conducted. As presented in Figure 4-4, sentences with parallel wh-words were interpreted with higher mean scores than those with nonparallel wh-words (M=0.61 > 0.52), showing that parallel bare conditionals were easier to interpret than nonparallel ones for the subjects.

Figure 4-4. Subjects’ Overall Interpretations of Parallelism Types in Bare Conditionals

The Kruskal-Wallis rank sum test was conducted to examine the overall comparison, and

found that the differences of mean scores in parallelism for the four groups was significant (χ²(1)=12.13, p<.001). Hence, the degrees of the elicited universal readings of parallel and nonparallel bare conditionals were not the same, where parallel bare conditionals were significantly easier for the subjects to interpret with universal readings than nonparallel ones.

To look into the interpretations of each group more specifically, we conducted within-group comparisons between parallel and nonparallel sentences, as illustrated in Table 4-7.

Overall, Grade 2, Grade 4 and the adults processed parallel sentences with universal readings

Parallel Nonparallel

more than processing nonparallel ones (Grade 2: M=0.54 > 0.47; Grade 4: M=0.65 > 0.43;

Table 4-7. Each Group’s Interpretations of the Two Parallelism Types Parallelism

On account of comparisons between groups for parallel sentences, the Kruskal-Wallis rank sum test showed that the difference was not significant (χ²(3)=6.21, p=.10). In addition,

for the between-group comparison in nonparallel sentences, the Kruskal-Wallis rank sum test indicated that the four groups exhibited mean scores significantly differently (χ²(3)=24.38,

p<.001). Thus, pairwise between-group comparisons were needed with the utilization of the

two-sample Wilcoxon tests.

Table 4-8. The between-group Differences in the Two Parallelism Types

Note: The adjustment of p-values for six paired comparisons was .01.

As seen in Table 4-8, significant differences were found in the comparisons between KS and Grade 2 (p<.01), between KS and Grade 4 (p<.01) and between KS and the adults (p<.01), while no difference was found between Grade 2, Grade 4 and the adults (Grade 2-Grade 4:

p=.22; Grade 2-Adults: p=.69; Grade 4-Adults: p=.04). This finding is due to the fact that most

KS subjects interpreted nonparallel bare conditionals with universal readings significantly different from the other groups. Since they were unaware of the discrepancies of parallel and nonparallel sentence types, they tended to interpret both types of bare conditionals with universal readings, yet the older groups varied their interpretations based on parallelism types.

4.2.3 Discussion

The major finding about donkey sentences was that there was a type effect on the interpretation of quantifiers in donkey sentences. The distinction between universal and existential quantifiers was found in the interpretations of each type, supporting Kanazawa’s Generalization (1994). Mei ‘every’ was the easiest to interpret, followed by bushi meige ‘not

every’ and youxie ‘some.’

These results can be explained with a syntactic-semantic interface account. An operator of two anaphoric elements (i.e., a noun in the antecedent clause and the other in the consequent clause) binds them together and thus assigns a reading to the sentence. The operators of donkey

sentences are of the three quantifier types, mei ‘every,’ youxie ‘some’ and bushi meige ‘not every,’ which have different degrees of saliency or explicitness that determines how strong a

reading of a sentence can be. Mean scores of each type tell the saliency or explicitness of each quantifier. As a result, the subjects found mei ‘every’ to be the most salient or explicit one since all the four groups scored highly on this type; bushi meige ‘not every’ is the second explicit one, and finally youxie ‘some’ the least, which shapes a contrast with the other two quantifiers.

Although both youxie ‘some’ and bushi meige ‘not every’ are both existential quantifiers, bushi

meige ‘not every’ is more straightforward due to the existence of bushi ‘not,’ which has a

negative feature, while youxie ‘some’ is more murky and relies on a person’s subjective conscious to decide when to assign an existential reading to a sentence. Accordingly, the more explicit a quantifier is, the easier the interpretation of it is; thus, children consider sentences with mei ‘every’ and bushi meige ‘not every’ easier to interpret, and the ambiguous property of youxie ‘some’ takes time and experience to learn, leading it to be the most challenging one that hinders their acquisition.

For parallelism of bare conditionals, an overall between-type comparison found a type

effect where parallel sentences elicited universal readings more easily than nonparallel ones.

This result is in support of Cheng & Huang (1996), who argue for parallelism in that when structures are in symmetry, readings are more easily to process.

Furthermore, a developmental pattern of acquisition was found in the two construction-specific factors. First, from a structural facet, symmetry plays a role when it comes to interpreting bare conditional sentences. Parallel bare conditionals are less challenging than nonparallel ones, where in the present study, only KS did not process nonparallel ones in an adult-like way. It is referred that children can easily master symmetric sentences; however they have difficulty interpreting asymmetric ones as adults do (Philip 1995, Xu 2012). Second, from a semantic aspect, the explicitness of an operator influences interpretations of quantification deeply. The idea of mei ‘every’ in donkey sentences is quite explicit, offering the subjects clear

ideas of the amount of items it modifies, that is, totality of a set of items. Ta ‘he/she’ and

na-ge ren ‘that person,’ which are put in the consequent clause of nonparallel bare conditionals,

likewise, are more explicit in numbers, are more likely to be interpreted with existential readings. In contrast, youxie ‘some,’ bushi meige ‘not every’ and wh-words are less explicit

and vaguer with no absolute number of entities involved; hence, they are more challenging to interpret (Katsos et al. 2016³). When it comes to acquisition, explicit ideas are easier to interpret

3 Katsos et al. (2016) investigated the effects of four different constraints on the acquisition of quantifiers in thirty-one languages. The four constraints include monotonicity, totality, complexity and informativeness of quantifiers that influence the sequence of acquisition. They found that the totality of all or none was firstly acquired than quantifiers like some and most, and this supports our results in that mei ‘every’, which also presents totality in quantification, was an easier quantifier to interpret and it was also the quantifier firstly acquired.

than vague ideas (Gentner & Toupin 1986). Hence, the explicit referentiality of mei ‘every’ is the most salient and the easiest to assign universal readings to donkey sentences, and nonparallel bare conditionals with concrete ta ‘he/she’ and na-ge ren ‘that person’ are easily to be interpreted with existential readings.

Since concepts of quantification are quite abstract, it can be acquired when children are as old as seven to eight years old (Simons & Keil 1995, Caramelli, Setti & Maurizzi 2004); as

a result, KS, the preschoolers, only processed mei ‘every’ quite easily, while bushi meige ‘not every’ were not interpreted in an adult-like manner until the subjects were at seven years old,

and only when they were at Grade 4 were they aware of the distinction between parallel and

nonparallel bare conditionals. This explicitness could also explain the finding where KS interpreted nonparallel sentences with more universal readings than the older groups did. It can be reasoned that KS was still in the process of acquiring the indefiniteness use of wh-words;

hence, they were unaware of the differences between parallel and nonparallel sentence types, yet the older groups, on the other hand, might have interpreted nonparallel sentences with existential readings due to the influence of the referentiality of the explicit ta ‘he/she’ and

na-ge ren ‘that person.’ Therefore, the differences in explicitness of parallelism may cause the

differences in interpretations for children at different ages.