This chapter presents the results of data analyses to answer the research questions. The purpose of the present study is to investigate the underlying linguistic knowledge of heritage language learners (HLLs) as compared to non- heritage language learners of Korean (NHLLs), and to provide more specialized support for pedagogical methods, upon which aural interpretive communication skills of HLLs can be used to generate better language learning outcomes. Three research questions were established in order to fulfill the research purpose.
The questions and hypotheses were:
1. Are there differences between English-speaking heritage language learners (HLLs) and non-heritage language learners (NHLLs) of Korean in the frequency of errors in case and postposition markers and affixal connectives, demonstrated in their writing?
The hypothesis is that there are significant differences between HLL and NHLL in the correct application of Korean case and postposition markers and affixal connectives in the writing performance.
2. Are there differences between HLLs and NHLLs in sources that cause errors in using case and postposition markers and affixal connectives in their writing?
The hypothesis is that there are significantly different sources of errors in using case and postposition markers and affixal connectives between HLLs and NHLLs.
3. Are there differences in the ability to identify grammatically correct case and postposition markers and affixal connectives between HLLs and NHLLs?
The hypothesis is that there are significantly different sources of errors in using case and postposition markers and affixal connectives between HLLs and NHLLs.
Participants and Demographic Data
All of the participants were currently taking or have taken a Korean language program at the University of Kansas (KU) in Lawrence, Kansas. All participants were placed in an appropriate level of class corresponding to their overall language proficiency based on the departmental policy of assigning proficiency levels according to leaners’ previous learning experiences or a proficiency test. There was a difference of 9 in the total number of participants between HLLs and NHLLs. The total number of participants at beginning level was 18; the total number of participants at intermediate level was 16; and the total number of participants at advanced level was 14. The size of NHLLs (N = 12) was twice as big as HLLs (N = 6) at the beginning level. However, the sizes of HLLs and NHLLs at the advanced level are identical. As a whole group, students’ gender was balanced with 49 % of females (M = 24) and 51 % of males (M = 25). However, there was a gender imbalance observed within group. While HLL group has more female participants (N = 12) than male participants (N = 8), NHLL group has a larger number of male (N = 17) than females (N = 12).
The average length of study of the language ranged from 2.42 semesters among NHLL at the beginning level to 5.29 semesters among both NHLLs and HLLs at the advanced level. The average age of contacting the target language for the first time is much lower among HLLs (M = .42, SD = 1.26) than that among NHLLs (M = 18.07, SD = 3.18). None of the NHLLs had their first contact with the language before 10 years old. In contrast, all the HLLs had their first
contact with Korean language before school age. All of the HLLs at the intermediate level were exposed to Korean since birth. Among the three sub-HLL groups by proficiency level, all of the intermediate level of HLLs were exposed to Korean since birth. The details about the
Table 6
Demographic information
Language status Non-heritage learner (N = 29) Heritage learner (N = 20)
Beginner Intermediate Advanced Beginner Intermediate Advanced Language level N 12 10 7 6 7 7 % 25 21 14 12 14 14 Gender Female 6 4 2 4 3 5 Male 6 6 5 2 4 2 Length of Study of Korean M 2.58 4.3 5.4 2.67 4.42 5.29 SD .90 .67 .79 .82 1.13 .95 Age of contacting Korean M 18.33 17.10 19 .67 .00 .57 SD 3.28 3.78 1.83 1.63 .00 1.51 Relative use of Korean on a daily basis (%) M 8.58 4.00 13.42 7.5 12.57 12.43 SD 7.81 2.67 16.72 5.24 7.16 6.43
Research Questions and Results
Study Variables. The independent variable is heritage language status (HLL vs. NHLL).
As each heritage group consists of students at three proficiency levels, language proficiency level (beginner, intermediate, and advanced) are the moderator variable. In testing research questions 1 and 3, the dependent variables are error rates in four case markers, six postposition particles and four affixal connectives. The 14 inflectional morphemes were selected based on the findings from a pilot error analysis. The 4 case marker morphemes are Subject, Topic, Object, and
Possessive markers, the 6 postposition markers are Indirect object, Locative, Instrumental, Companion, Temporal, and Directional postposition markers. The 4 affixal connective have equivalent meanings of “AND,” “BUT,” “SO,” and “WHILE,” respectively. In testing research
question 2, the dependent variables are four error sources - omission, addition, replacement, and malformation (more details about these morphemes can be found in Table 3).
Research question 1. Research question 1 is: “Are there differences between English- speaking Korean heritage language learners (HLLs) and non-heritage language learners (NHLLs) of Korean in the occurrence and frequency of errors in case and postposition markers and affixal connectives, demonstrated in their writing?”
The hypothesis 1 is that there are differences in in the correct application of Korean morphemes in writing performance between HLL and NHLL. To test hypothesis 1, a 2-way ANOVA was conducted with participants' heritage status as the focal independent variable, proficiency level as the moderator, and error rates of 14 types of Korean inflectional morphemes as the dependent variables.
Descriptive report
The percentages of errors by type range from 8 % in Object marker error (SD = 8.18) and 6 % in Subject marker (SD = 5.52) to 0% in Instrumental marker error. Ranges of error numbers are following: 3 -11 errors among Beginning level of HLLs, 2 - 20 among intermediate level of HLLs, and 2-31 among advanced level of HLLs, and 0 - 14 errors among Beginning level of NHLLs, 4 – 20 among intermediate level of NHLLs, and 2 - 22 among advanced level of NHLLs.
Errors in case markers are the most frequently committed by both HLLs and NHLLs. Among the errors in four case markers, the highest rate of errors is found in Object marker errors in both HLLs and NHLLs, with a slightly higher score by HLLs (M =7.99, SD = 7.12) than NHLLs (M = 7.51, SD = 8.96). The most contrasting results in the frequencies of errors between HLLs and NHLLs are found in Topic marker and Locative postposition marker. NHLLs (M = 5.04, SD = 3.85) obtained a higher score in Topic marker than HLLs (M = 2.84, SD = 2.79); and
HLLs (M = 3.71, SD = 4.49) earned a higher score in Locative particle than NHLLs (M = 2.25, SD = 3.62). No error was found in Instrumental marker either by HLLs and NHLLs. The descriptive statistics are given in Table 7 and Figure 1.
Table 7
Error rates by 14 types of morphemes identified on the writing samples
Language status Non-heritage learner (N = 29) Heritage learner (N = 20) Error rate Data Beginner
(N = 12) Intermediate (N = 9) Advanced (N = 8) Beginner (N = 6) Intermediate (N = 7) Advanced (N = 7) Subject case M 6.28 5.74 4.50 5.62 7.51 3.60 SD 6.93 4.57 3.58 5.34 4.92 4.09 Topic case M 6.17 5.12 3.25 5.04 4.11 1.79 SD 4.51 2.91 3.46 3.85 3.69 1.83 Object case M 9.41 9.00 3.00 7.51 7.72 11.80 SD 11.91 7.15 2.70 8.96 5.72 8.91 Possessive M .38 .87 1.19 .75 .00 .97 SD 1.31 1.94 1.86 1.66 .00 1.66 Indirect Object M .89 .20 1.56 .86 .76 .00 SD 2.09 .60 2.17 1.81 1.86 .00 Location M 2.80 1.61 2.15 2.25 2.03 2.68 SD 4.97 1.93 2.90 3.62 3.25 4.72 Direction M 1.23 .00 .00 .51 1.34 .00 SD 1.93 .00 .00 1.36 2.09 .00 Instrument M .00 .00 .00 .00 .00 .00 SD .00 .00 .00 .00 .00 .00 Companion M 1.93 .93 .46 1.21 1.40 2.65 SD 2.56 1.39 .86 1.93 2.18 4.95
Temporal M .95 .19 .11 .48 .64 .00 SD 2.28 .56 .30 1.52 1.57 .00 AND M .87 .37 .00 .48 1.92 .26 SD 1.64 .73 .00 1.16 4.71 .69 SO M .81 .98 1.30 1.00 1.90 1.31 SD 1.90 1.48 1.31 1.58 3.20 1.79 BUT M .64 .29 .22 .42 .76 .65 SD 1.58 .88 .62 1.15 1.86 1.72 WHILE M .79 .37 1.54 .87 .00 .65 SD 1.85 .73 1.54 1.52 .00 1.13
Figure 1 Comparison of error frequencies in 14 types of errors between HL and NHL at 3 proficiency levels 0 2 4 6 8 10 12 SUB TOP OBJ POS INO LOC DIR INS COM TEM AND SO BUT WHILE BNHLL INHLL ANHLL BHLL IHLL AHLL
Statistical Tests Set 1
A two-way ANOVA was performed to test statistically significant differences between HLLs (N = 20) and NHLLs (N = 29) in the frequencies of errors in the use of Case markers, Postposition particles, and Affixal connectives on their writing task. The results of the statistical analysis are seen in Table 8 - 10.
Error rate in Case markers
The main effect of heritage status was insignificant in the error rate in Case marker, F (1, 43) = .09, p = .77, with a trivial effect size,= .002; the main effect of proficiency level yielded
an F ratio of F (1, 43) = 2.04, p = .13, with a large effect size, = .09, indicating that
proficiency level to error rate in Case markers was non-significant; and there was also no
significant interaction between heritage status and proficiency level, F (2, 43) = .29, p = .24, with a small effect size,= .01.
Table 8
ANOVA: Error rates in Case markers in error analysis (EA)
Source df F η2 p Heritage status 1 .30 .64 .13 Proficiency level 2 6.92 .13 .87 Heritage* level 2 .29 .75 .01 Error 43
Error rate in Postposition markers
The two-way ANOVA test showed no significant main effect for heritage factor, F (1,43) = .77, p = .38, with a small effect size,= .02; there was no significant main effect of language proficiency level, F (2, 43) = 1.06, p = .35, with a small effect size,= .05; and no significant interaction effect was identified between proficiency level and heritage status, F (2,43) = .92, p
= .41, with a small effect size,= .04. The results indicated that language proficiency level, heritage status, and interaction between level and heritage status were not significantly related to the error frequency in Postposition markers.
Table 9
ANOVA: Error rates in Postposition markers in EA
Source df F η2 p Heritage status 1 .84 .46 .29 Proficiency level 2 1.15 .46 .54 Heritage* level 2 1.15 .46 .54 Error 43
Error rate in Affixal connectives
The two-way ANOVA test showed no significant main effect for heritage factor, F (1,43) = .02, p = .89, with a small effect size,= .02; there was no significant main effect of language proficiency level, F (2, 43) = 1.12, p = .34, with a small effect size,= .05; and no significant interaction effect was identified between proficiency level and heritage status, F (2,43) = 1.01, p = .37, with a small effect size,= .05. The results indicated that language proficiency level, heritage status, and interaction between level and heritage status were not significantly related to the error frequency in Affixal connectives.
Table 10
ANOVA: Error rates in Affixal connective in EA
Source df F η2 p Heritage status 1 .02 .90 .01 Proficiency level 2 1.11 .47 .53 Heritage* level 2 1.01 .37 .04 Error 43
Statistical Tests Set 2
No differences were found from the three broad categories of morpheme errors, so each of the 14 possible morpheme errors were tested individually in an attempt to examine in-depth whether there are significant differences between heritage language learners (HLLs) and non- heritage language learners (NHLLs) in the use of these morphemes. A series of two-way
ANOVA were performed to test statistically significant differences between HLLs (N = 20) and NHLLs (N = 29) in the rates of errors in the use of a total of 14 inflectional morphemes: four case markers (Subject, Topic, Object, and Possessive), six postposition markers (Indirect object, Locational, Directive, Instrumental, Companion, and Temporal), and four affixal connectives (“AND,” “BUT,” “SO,” and “WHILE”). However, as there was no error found in Instrument postposition marker on any of the writing samples, this error category was not tested. Thirteen out of 14 dependent variables were explored using two-way ANOVA to test whether there are significant differences between heritage status and proficiency level in the error rates in the 13 types of morphemes.
Error rates in Subject marker
The main effect of heritage status was insignificant, F (1, 43) = .00, p = .98, with no effect size, = .00; the main effect of proficiency level yielded an F ratio of F (2, 43) = .85, p = .44, with a small effect size, = .04, indicating that proficiency level to error rate in subject marker was non-significant. There was no significant interaction effect between heritage status and proficiency level, F (2, 43) = .47, p = .63, with a small effect size, = .02.
Error rates in Topic marker
The main effect of language status yielded an F ratio of F (1, 43) = 3.75, p = .06, with a medium effect size, = .08, indicating that heritage status to scores in topic marker error was
non-significant. The main effect of language proficiency level yielded an F ratio of F (2, 43) = 1.66, p = 20, with a medium effect size,= .07, indicating that language proficiency level is insignificant to rate in topic marker error. There was no significant interaction between heritage status and proficiency level in the error of topic marker, F (2, 43) = .68, p = .51, with a small effect size, = .03.
Error rates in Object marker
No significant main effect was presented for heritage effect, F (1,43) = .13, p = .72, with a trivial effect size,= .003; the main effect of language proficiency level is not significant with an F ratio of F (2, 43) = 2.80, p = .07, with a medium effect size,= .12, indicating that
language proficiency level is not significantly related to errors in object marker; and there was no significant interaction effect between proficiency level and heritage status, F (2,43) = .32, p = .73, with a small effect size,= .01.
Error rates in Possessive marker
The main effect of heritage status was insignificant, F (1, 43) = .14, p = .71, with a trivial effect size,= .003; the main effect of proficiency level yielded an F ratio of F (2, 43) = 2.97, p = .06, with a medium effect size, = .12, indicating that proficiency level to error rate in possessive marker was non-significant. There was no significant interaction effect between
heritage status and proficiency level, F (2, 43) = .53, p = .59, with a small effect size, = .02. Error rates in Indirect Object marker
The main effect of language status yielded an F ratio of F (1, 43) = 1.98, p = .17, with a small effect size, = .04, indicating that heritage status to scores in indirect object particle error was non-significant. The main effect of language proficiency level yielded an F ratio of F (2, 43) = 1.11, p = .34, with a small effect size,= .05, indicating that language proficiency level is
insignificant to rate in indirect object particle error. There was no significant interaction between heritage status and proficiency level in the error of topic marker, F (2, 43) = 1.06, p = .35, with a small effect size,= .05.
Error rates in Locative marker
No significant main effect was presented for heritage effect, F (1, 43) = 1.54, p = .22, with a small effect size,= .03; the main effect of language proficiency level is not significant with an F ratio of F (2, 43) = 1.16, p = .32, with a small effect size,= .05, indicating that language proficiency level is not significantly related to errors in locative particle; and there was no significant interaction effect between proficiency level and heritage status, F (2, 43) = 1.44, p = .25, with a large effect size,= .06.
Error rates in Directional marker
The two-way analysis of variance showed no significant main effect for heritage factor, F (1,43) = .54, p = .47, with a small effect size,= .01. However, the main effect of language proficiency level yielded an F ratio of F (2, 43) = 3.19, p = .05, with a medium effect size,= .13, indicating that language proficiency level is significantly related to errors in directional particle. There was no significant interaction effect between proficiency level and heritage status, F (2, 43) = .36, p = .70, with a small effect size,= .02.
Error rates in Companion marker
The main effect of language heritage status yielded an F ratio of F (1, 43) = .51, p = .48, with a small effect size, = .01, indicating that heritage status to scores in Companion particle error was insignificant. The main effect of language proficiency level yielded an F ratio of F (2, 43) = .93, p = .40, with a small effect size,= .04, indicating that language proficiency level is not significant to rate in Companion particle error. There was no significant interaction between
heritage status and proficiency level in the error of Companion particle, F (2, 43) = .81, p = .45, with a small effect size, = .04.
Error rates in Temporal marker
No significant main effect was presented for heritage effect, F (1,43) = .28, p = .60, with a small effect size,= .01; the main effect of language proficiency level is not significant with an F ratio of F (2, 43) = 1.63, p = .21, with a medium effect size,= .07, which indicates that language proficiency level is not significantly related to errors in Temporal particle; and there was no significant interaction effect between proficiency level and heritage status, F (2,43) = .02, p =.98, with a trivial effect size, < .01.
Error rates in affixal connective “AND”
The main effect of heritage status was insignificant, F (1, 43) = .33, p = .57, with a small effect size, = .01; the main effect of proficiency level yielded an F ratio of F (2, 43) = 2.46, p = .10, with a medium effect size, = .10, which demonstrates that proficiency level to error rate in affixal connective ‘AND’ was non-significant. There was no significant interaction effect between heritage status and proficiency level, F (2, 43) = .47, p = .63, with a small effect size,
= .02.
Error rates in affixal connective “SO”
The main effect of language status yielded an F ratio of F (1, 43) = .29, p = .59, with a small effect size, = .01, indicating that heritage status to scores in affixal connective ‘CAUSE’ error was non-significant. The main effect of language proficiency level yielded an F ratio of F (2, 43) = .09, p = .91, with a trivial effect size, < .01, indicating that language proficiency level is insignificant to rate in topic marker error. There was no significant interaction between
heritage status and proficiency level in the error of affixal connective ‘CAUSE’, F (2, 43) = .61, p = .55, with a small effect size, = .03.
Error rates in in affixal connective “BUT”
No significant main effect was presented for heritage effect, F (1, 43) = .05, p = .82, with a trivial effect size, < .01; the main effect of language proficiency level is not significant with an F ratio of F (2, 43) = .82, p = .45, with a small effect size,= .04, indicating that language proficiency level is not significantly related to errors in affixal connective ‘BUT’; and there was no significant interaction effect between proficiency level and heritage status, F (2,43) = .19, p = .82, with a small effect size,= .01
Error rates in in affixal connective “WHILE”
The main effect of heritage status was insignificant, F (1, 43) = .33, p = .57, with a small effect size, = .01; the main effect of proficiency level yielded an F ratio of F (2, 43) = 2.46, p = .10, with a medium effect size, = .10, which demonstrates that proficiency level to error rate in affixal connective ‘WHILE’ was non-significant. There was no significant interaction effect between heritage status and proficiency level, F (2, 43) = .47, p = .63, with a small effect size,
< .01.
In summary, the two-way ANOVA performed to answer research question 1 indicated that there was no main effect of heritage status (HLL versus NHLL) to occurrences of 13 types of errors, and no significant interaction effect between heritage status and proficiency level as well. The main effect of proficiency level was determined significant only in the error frequency by Directional particle. In conclusion, there are no differences between HLLs and NHLLs in error rates in the use of 14 types of morphemes on their writing products.
Research question 2. Research question 2 is: “Do Korean heritage language learners (HLLs) and non-heritage language learners (NHLLs) of Korean exhibit differences in the sources of errors in using case and postposition markers and affixal connectives? The hypothesis 2 is that there are differences in the sources of errors on written product between HLLs and NHLLs.
To test hypothesis 2, a 2-way ANOVA was conducted with participants’ heritage status (HLL vs. NHLL) as the focal independent variable, language proficiency level (beginning, intermediate, and advanced) as the moderator variable, and four error sources (Omission, Replacement, Addition, and Malformation) as the dependent variables. Data from the EA was further utilized to answer the research question 2. The data were analyzed using two-way
ANOVA. When 2-way ANOVA revealed a significant interaction effect between heritage status and language proficiency level, the paired independent variable was further investigated by proficiency level – beginning, intermediate, and advanced.
Statistical Tests
Two-way analysis of variance ANOVA was performed to test statistically significant differences between HLLs and NHLLs in the source of errors in the use of the four types of Case markers, six types of Postposition markers, and four types of Affixal connectives on their writing task. The means and standard deviations are presented in Table 11. The results of the statistical tests are given in Table 12 - 15.
Table 11
Rates of errors by its source
Heritage status Non-heritage learner (N = 29) Heritage learner (N = 20) Error source Beginner
(N = 12) Intermediate (N = 9) Advanced (N = 8) Beginner (N = 6) Intermediate (N = 7) Advanced (N = 7) Omission M 38.46 31.70 27.87 46.82 52.29 21.55 SD 28.13 17.30 15.70 28.90 27.16 13.59 Replacement M 34.66 52.15 54.20 31.30 28.98 47.61 SD 23.26 11.35 30.52 23.71 25.85 9.09 Addition M 4.34 7.15 3.79 11.52 .00 9.98 SD 7.69 7.69 9.87 24.03 .00 16.45 Malformation M 14.20 9.01 14.14 10.27 18.73 20.86 SD 18.43 7.81 23.13 15.53 17.41 12.29
Errors caused by Omission
The two-way ANOVA test showed no significant main effect for heritage factor, F (1,43) = 1.27, p = .27, with a small effect size,= .03 ; there was no significant main effect of
language proficiency level, F (2, 43) = 3.01, p = .06, with a medium effect size,= .12; and no significant interaction effect was identified between proficiency level and heritage status, F (2,43) = 1.33, p = .28, with a medium effect size,= .06. The results indicated that language
proficiency level, heritage status, and interaction between level and heritage status were not significantly related to the errors caused by Omission of necessary morphemes (Table 12). Table 12
ANOVA: Rates of Errors caused by Omission
Source df F η2 p Heritage status 1 1.27 .03 .27 Proficiency level 2 3.01 .12 .06 Heritage* level 2 1.33 .06 .28 Error 43
The two-way ANOVA test showed no significant main effect for heritage factor, F (1,43) = .02, p = .89, with a small effect size,= .02 ; there was no significant main effect of language proficiency level, F (2, 43) = 1.12, p = .34, with a small effect size,= .05; and no significant interaction effect was identified between proficiency level and heritage status, F (2,43) = 1.01, p = .37, with a small effect size,= .05. The results indicated that language proficiency level, heritage status, and interaction between level and heritage status were not significantly related to the errors caused by the Addition of a grammatically incorrect morpheme (Table 13).
Table 13
ANOVA: Rates of Errors caused by Replacement
Source df F η2 p Heritage status 1 2.92 .06 .10 Proficiency level 2 2.56 1.11 .09 Heritage* level 2 .92 .04 .40 Error 43
Errors caused by Replacement
The two-way ANOVA test showed no significant main effect for heritage factor, F (1,43) = .02, p = .89, with a small effect size, η2 = .02 ; there was no significant main effect of language proficiency level, F (2, 43) = 1.12, p = .34, with a small effect size, η2= .05; and no significant interaction effect was identified between proficiency level and heritage status, F (2,43) = 1.01, p = .37, with a small effect size, η2 = .05. The results indicated that language proficiency level, heritage status, and interaction between level and heritage status were not significantly related to