Chapter 4. Results on Specificity Effect
4.1 Production task
4.1.1 Performance in each trial testing novel word
Before discussing results from inferential statistic analyses, this section provides a
general description for the trials that were designed to test the performance on the target
words. Results were summarized in Table 4.1. The accuracy in Trial 1, Trial 3 and in
Trail 4 in the production task (see Table 3.3 on p. 47) was coded for further analysis. If
the target word was produced for the certain action shown in the video of the trial, the
response would be coded as “target word”. On the other hand, other responses or no
responses would be coded as “others”.16
Overall, the performance on learning a general word was more stable across the
pre-test session and the post-test session when compared to the performance on learning
specific words. Results from the General Condition showed that about 70% of responses
were target words in each trial. On the other hand, results from the Specific Condition
showed that the proportion of target responses dropped in the post-tests for both the test
with the baseline label, i.e., Label 1, and the other test with Label 2. It seemed that the
contrast marked by an additional label (Label 2) was not learned and this contrast also
impeded the performance of the baseline label. Nevertheless, it was also found that
participants under the Specific Condition still had better performance in the post-test
with Label 1 than that with Label 2 (Label 1: 52% vs. Label 2: 33% ).
In addition, the performance on mapping Action 2 onto Label 1 in the General
Condition could be compared to the performance on mapping Action 2 onto Label 2 in
16 It should be noted that these two trials were designed to test the production for the baseline verb, and thus the response of the other specific verb in the Specific Condition would be coded as “others”.
the Specific Condition. Results showed that children under the General Condition
perform much better than those in the Specific Condition (General: 70% vs. Specific:
33%). It seemed that mapping a similar action onto the same label is easier than
mapping it onto a different label.
Table 4.1: The number and proportion of target responses in the production task
General Specific
Target break carry Total Target break carry Total
Session Stimuli n=30 n=30 n=60 n=30 n=30 n=60
Pre-test Action 1 Label 1 23 17 40 (67%) Label 1 23 23 46 (77%)
Post-test Action 1 Label 1 22 20 42 (70%) Label 1 13 18 31 (52%)
Post-test Action 2 Label 1 24 18 42 (70%) Label 2 9 11 20 (33%)
Change from pre-test to post-test
To explore whether children’s responses in the pre-test session significantly
differed from those in the post-test, performance of each child was analyzed through
McNemar tests. In this section, for the Specific Condition, it was explored whether the
additional label would facilitate or impede the performance on the baseline label. For
the General Condition, it was examined whether the additional kind of exemplar for the
same label facilitated the performance of the label in the test trial with the baseline
action.
Table 4.2 summarized the frequency of responses for the pre-test and the post-test
on the baseline verb. Most children under the General Condition did not change their
response after being exposed to a novel action that was also labeled by the baseline verb.
The proportion of target responses in the pre-test and the post-test was similar. On the
other hand, the McNemar tests showed that in the Specific Condition the proportion of
children responding with target words in the pre-test significantly differed from the
proportion in the post-test (p < .0005): Forty-six out of 60 children responded with
target words in the pre-test while only 31 out of 60 in the post-test did so. This
difference is attributable to the fact that 16 children produced the target word in the
pre-test successfully but failed to respond with the target word in the post-test whereas
only one child changed their response from the target word to others.
Taken as a whole, showing another similar action and mapping it onto the same
word did not affect the performance on producing the baseline verb, as shown in results
from the General Condition; in contrast, exposing children with a different word that
labeled a similar action seemed to affect the production of the baseline verb, as shown
in results from the Specific Condition. One possibility might be that some of the
children in the Specific Condition were confused by the two newly-learned words. They
produced the other word, which was supposed to label the other action rather than the
action shown in the test trial. In addition, the lower frequency of the baseline verb in the
Specific Condition could be a competing account for the poorer performance in the
Specific Condition when compared to General condition. However, this possibility
seems not plausible since the frequency of another specific verb was actually lower than
baseline verb in the Specific Condition. If frequency alone could account for the
learning process in this task, children should have been more likely to produce the
baseline verb instead of changing their responses from the baseline verb to the other
verb, which was presented with lower frequency.
Table 4.2: Performance on the baseline verb in the production task (pre-test and post-test)
Pre-test
Target word Others Total
General Post-test Target word 37 5 42
Others 3 15 18
Total 40 20 60
McNemar test: p =.727
Specific Post-test Target word 30 1 31
Others 16 13 29
Total 46 14 60
McNemar test: p < .0005