After the foregoing discussion about the construction effects, in this section, close attention is switched to the interplay of construction effects and property effects
on the L2 acquisition of the three constructions. In this study, two properties of each construction were examined. One property pertained to verb types, and the other property concerned predication by resultatives, particle movement, and word order of datives/double objects. The subjects’ acquisition of these properties is shown in Figure 4-4.
0.47 0.5
0.69 0.69
0.83 0.87
0 0.2 0.4 0.6 0.8 1
R-verb R-pn P-verb P-mt D-verb D-to
Property
Mean score
Figure 4-4. Mean scores on different properties of the constructions
When property effects were examined vis-à-vis constructions effects, it was found that the subjects performed significantly better on both properties of dative/double object constructions than those of the other two constructions (p=.000** throughout) and that their performance on the properties of verb-particle constructions was evidently better than those of resultative constructions (p=.000**
throughout). As Figure 4-4 shows, the subjects consistently performed the best on dative/double object constructions, the next on verb-particle constructions, and the worst on resultative constructions, irrespective of which properties were considered. It indicated that construction effects might override property effects such that the acquisition of any properties was essentially contingent on their corresponding constructions.
The results confirmed the difficulty hierarchy established in (2), in which resultatives were the hierarchically highest constructions and datives/double objects the lowest. As construction effects could overrule property effects and the hierarchically higher constructions exhibited greater degrees of difficulty than the lower counterparts, any properties of the hierarchically higher constructions would be acquired later than those of the lower ones, as Figure 4-4 shows. In addition to corroborating the hierarchy phenomenon, the results echo the syntactic concept of linear precedence. The structurally more dominant nodes constantly take precedence over the lower ones in the linear order. That is, in the difficulty hierarchy, any properties of the hierarchically higher constructions consistently precede those of the lower constructions on the degree of complexity.
The data presented so far displayed esthetically consistent patterns concerning the relationships of the three telicity-related constructions in L2 acquisition. It merits deeper exploration whether the consistent relationships would hold true when English proficiency as well as property effects interacted with construction effects. Figure 4-5 shows the acquisition of the two properties of each construction by the three experimental groups.
Figure 4-5. Each group’s scores on different properties of the constructions
Strikingly, all the three groups exhibited parallel patterns of the acquisition of the three telicity-related constructions. The findings indicated that our subjects performed significantly better on dative/double object constructions than the other two constructions and better on verb-particles than resultatives, regardless of the properties examined (p<.05* throughout). It is evidenced that construction effects could override property effects throughout different L2 proficient levels and that the consistent relationships of the three constructions could be substantiated as well.
In the within-construction analysis, the group of higher English proficiency performed similarly between the two properties of resultative constructions (p=.24) as well as between those of verb-particle constructions (p=.63). However, they performed significantly differently on the two properties of dative/double object constructions (p=.001**). The M and L groups demonstrated no significant differences between the two properties of the three constructions (resultatives: p=.51 and p=.20, respectively; verb-particles: p=.74 and p=.84, respectively; datives/double objects: p=.06 and p=.56, respectively), indicating that properties might play a marginal role in influencing the L2 acquisition of each construction.
With respect to the factor of English proficiency, group differences on the properties of the telicity-related constructions are shown in Table 4-4.
Table 4-4. Group differences on different properties of the constructions Groups R-verb R-pn P-verb P-mt D-verb D-to
H M .09 .000** .17 .06 .006** .02*
L .001** .001** .003** .004** .02* .000**
M L .23 .96 .28 .62 .92 .17
Note: The mean difference is significant at the .05 level (*p<.05, **p<.01).
It was found that the H and M groups performed significantly differently on three properties, i.e. predication by resultatives (p=.000**), verb types of datives/double
objects (p=.006**), and word order of datives/double objects (p=.02*). Also, marked differences were observed between the H and L groups on all the properties at issue (p<.05* throughout). However, no significant differences were obtained between the M and L groups on the six properties (p>.05 throughout). The findings showed that the M and L groups were at a similar developmental stage and that English proficiency was a determinant in affecting our subjects’ acquisition of the telicity-related constructions. The more proficient the subjects were, the better they performed on the three constructions. The greater difference their English proficient levels displayed, the larger divergence their performance demonstrated.
In the six properties discussed above, there was an identical property (i.e. verb types) examined throughout the three telicity-related constructions. With respect to this property, the acquisition of the three constructions also exhibited the same trend as observed so far. This property was concerned with two contrasting subproperties, i.e. dynamic versus stative verbs. To avoid a cursorily-drawn conclusion, this study stepped further to investigate the relationships of the three constructions with regard to the property in common. Given the fact that the telicity-related constructions are incompatible with any stative verbs, this study drew on the grammaticality judgment task to investigate the subjects’ competence on the property, i.e. grammatical dynamic verbs and ungrammatical stative verbs. Figure 4-6 below presents the subjects’
accuracy on the grammaticality judgment of the three constructions with respect to dynamic versus stative verbs.
At first glance, the results on stative verbs shown in Figure 4-6 seemed to be at odds with the formerly observed tendency, although the results on dynamic verbs still demonstrated a consistent trend. On the one hand, the subject were significantly better in judging grammatical dynamic verbs of datives/double objects than those of verb-particles and resultatives (p=.000** both), and they performed significantly
0.4
Figure 4-6. Accuracy on the grammaticality judgment of different verb types
better on grammatical dynamic verbs of verb-particles than those of resultatives (p=.000**). Nonetheless, the findings concerning ungrammatical stative verbs exhibited a reverse pattern. The subjects’ judgment was significantly worse on ungrammatical stative verbs of datives/double objects than those of verb-particles and resultatives (p=.000** both) although their performance on those of the latter two constructions was not significantly different (p=.13).
Since the data on the grammatical dynamic verbs still exhibited the corresponding pattern to that observed above, it was speculated whether the inverse tendency on ungrammatical stative verbs might stem from the artificial selection of the test verb items in the methodological design. Fortunately, when this study probed into the raw data of dative/double object constructions, it was found that the subjects performed unanimously better on one verb item than the other, as shown in Table 4-5.
Table 4-5. The two ungrammatical stative verb items in datives/double objects
Stative verb Mean score SD p
possess .28 .45
.000**
know .90 .29
Note: The mean difference is significant at the .05 level (*p<.05, **p<.01).
The far lower score on the stative verb possess than that of the verb know might result from the fact that the subjects were not familiar with the former. Given the distinct discrepancy between the mean scores of the two verb items and the subjects’ overall consistent performance on datives/double objects, it seemed that the skewed results on ungrammatical stative verbs could be attributed to the poor selection of the test items.
Figure 4-7 shows the results with the data on the item possess removed.
0.4
Figure 4-7. Accuracy on the grammaticality judgment of different verb types (Reanalysis)
It turned out that the subjects performed better on ungrammatical stative verbs of datives/double objects than on those of verb-particles (p=.003**). Nonetheless, they were not better at judging ungrammatical stative verbs of verb-particles than those of resultatives. Besides, the differences between verb-particles and resultatives did not arrive at the statistically significant level (p=.13), and this was the case with those between datives/double objects and resultatives (p=.06). Thus, the findings on resultative constructions still posed a challenge for the above-observed trend of the acquisition of the three constructions.
However, when the results on grammatical dynamic verbs of resultatives were compared with those on ungrammatical static counterparts, the comparison pointed to
an inclination of the subjects. The score on the judgment of grammatical resultative sentences only reached 0.4 point, indicating that the subjects tended to indicate
‘incorrect’ for grammatical resultative sentences. On the other hand, the score on the judgment of ungrammatical resultative ones was twice as high as the former (Mean=.84), which also implied that the subjects tended to consider ungrammatical resultatives incorrect sentences. Then, the results implied that the subjects had a propensity to reject the sentences involving resultative constructions. This tendency was also observed on the other test property of resultatives, namely predication of the object versus the subject. In predication, resultatives can be predicated only of the object; thus, sentences in which a resultative complement modifies the subject are ungrammatical. It was found that the subjects judged grammatical resultative sentences (with predication of the object) around chance (Mean=.57, SD=.34) and were inclined to reject ill-formed counterparts (Mean=.72, SD=.33). It was also suggested that the subjects performed poorly on resultatives. In the case of datives/double objects, in contrast, the subjects performed much better on the grammaticality judgment task. The scores on the judgment of grammatical and ungrammatical verb types of the constructions surpassed 0.9 point, showing that the subjects were accurate at accepting grammatical dative/double object sentences and rejecting ungrammatical ones. As regards verb-particles, the subjects’ mean scores lay in between those on the other two constructions. By and large, the subjects’
inclination to reject resultatives suggested that they performed the worst on the constructions, which might result from the fact that English resultatives were novel and difficult constructions to them. Hence, the findings reported here are still generally in concert with the above-observed trend. The overall parallel results favor and lead to the conclusion that the three telicity-related constructions are consistently correlated and that construction effects essentially override property effects on the L2
acquisition of the three constructions.
Last, our attention is turned to other within-construction analyses. In the case of particle movement, the subjects performed significantly better on V-NP-Particle constructions (Mean=.83, SD=.29) than V-Particle-NP constructions (Mean=.69, SD=.31, p=.001**). This L2 result is in accordance with Snyder and Stromswold’s (997) L1 finding that V-NP-Particle constructions substantially emerge before V-Particle-NP in child language acquisition. With respect to word order of datives/double objects, when the direct object is followed by the indirect one, the case marker to should obligatorily appear before the latter. When the two postverbal NPs are in reverse order, the retention of the case marker before the indirect object will give rise to ungrammatical sentences. It was found that the subjects performed better in judging grammatical V-D.O.-to-I.O. constructions (Mean=.97, SD=.14) than ungrammatical V-to-I.O.-D.O. counterparts (Mean=.84, SD=.26). In spite of the high scores on ungrammatical constructions, the results demonstrated significant differences between the subjects’ judgment of grammatical versus ungrammatical sentences (p=.000**), indicating that negative L1 transfer takes effect to some extent, since Chinese licenses V-to-I.O.-D.O for a large number of ditransitive verbs.