4.2 Pure tone perception test
4.2.2 Set 2
Figure 7 shows the mean RV for each Set 2 stimulus in every group. The rows and columns represent F0 levels and DURATION levels as Figure 4 does. The three boxes in each boxplot also stand for the three levels of AMPLITUDE as Figure 4.
Figure 7: The mean RV for each stimulus in Set 2 in each group (4) English speakers
(5) Mandarin speakers with an intermediate English proficiency level (MI)
(6) Mandarin speakers with an advanced English proficiency level (MA)
F0
DURATION
F0
DURATION
F0
DURATION
A five-way ANOVA (F0 (P, O, M) x DURATION (P, O, M) x AMPLITUDE (P, O, M) x GROUP (MI, MA, EN) x BLOCK (1, 2)) was conducted on the Set 2 data.
F0 [F(2, 971) = 90.268, p < 0.001], DURATION [F(2, 971) = 240.881, p < 0.001], GROUP [F(2, 971) = 19.512, p < 0.001] were found to have a significant effect on the RVs. AMPLITUDE [F(2, 971) = 0.114, p = 0.893] and BLOCK [F(1, 971) = 1.832, p
= 0.17616] did not have a significant effect on the RVs. Four interactions were found:
F0 x DURATION [F(4, 971) = 58.422, p < 0.001], F0 x GROUP [F(4, 971) = 4.036, p
< 0.01], DURATION x GROUP [F(4, 971) = 7.651, p < 0.001] and F0 x DURATION x GROUP [F(8, 971) = 2.709, p < 0.01].
Pairwise comparisons with Tukey’s HSD method were done respectively on the three significant factors, i.e., F0, DURATION and GROUP. In the three levels of F0, the mean RV for P (2.846) was greater than O (2.730), which was greater than M
(2.317). The P-O [p < 0.05], O-M [p < 0.001] and P-M [p < 0.001] differences were all significant (P > O > M). This suggests that the increase of F0 in the second tone (either one increment or two increments) could facilitate tone differentiation. In the three levels of DURATION, the mean RV for P (2.960) was greater than O (2.820), which was greater than M (2.114). The P-O [p < 0.01], O-M [p < 0.001] and P-M [p <
0.001] differences were all significant (P > O > M). This suggests that the increase of duration in the second tone (either one increment or two increments) could facilitate
tone differentiation. In the three levels of GROUP, the mean RVs for EN (2.562) and MA (2.550) were lower than MI (2.780). The MI-MA difference [p < 0.001] and MI-EN difference [p < 0.001] were significant, while the MA-EN difference were insignificant [p = 0.943] (MI = MA > EN). This suggests that the Mandarin speakers in the MI group are more likely to differentiate to the two tones than the English speakers and Mandarin speakers in the MA group.
4.2.2.1 F0 x DURATION
For each interaction between the factors, the data was stratified to examine the simple main effects. Pairwise comparisons with Tukey’s HSD method were
conducted when there were significant simple main effects. For the F0 x DURATION interaction, a one-way ANOVA (F0) was performed at each level of DURATION (P, O and M) respectively. First, at the P level of DURATION, F0 was not found to have a significant effect on the RVs [F(2, 374) = 0.108, p = 0.897]. Second, at the O level of DURATION, F0 was found to have a significant effect on the RVs [F(2, 375) = 5.257, p < 0.01]. Pairwise comparisons show that the RVs of F0-P (2.897) and F0-O (2.873) were greater than F0-M (2.690). The O-M difference [p < 0.05] and P-M difference [p < 0.01] were significant. The P-O difference was insignificant [p = 0.938]. This suggests that when the second tone was longer in duration by one
increment than the first tone (i.e., the O level), the increase of F0 by one increment (i.e., the F0-O case) in the second tone could facilitate the differentiation between the two tones, while the increase of F0 by two increments (i.e., the F0-P case) would not further strengthen the tone differentiation. Third, at the M level of DURATION, F0 had a significant effect on the RVs [F(2, 375) = 106.8, p < 0.001]. Pairwise
comparisons show that the RV of F0-P (2.675) was greater than F0-O (2.365), which was greater than F0-M (1.302). The P-O [p < 0.01], P-M [p < 0.001] and O-M [p <
0.001] differences were all significant. This suggests that when the two tones had the same duration, the increase of F0 in the second tone would facilitate tone
differentiation.
For the F0 x DURATION interaction, the other one-way ANOVA (DURATION) was performed at each level of F0 (P, O and M). First, at the P level of F0,
DURATION was found to have a significant effect on the RVs [F(2, 374) = 11.33, p <
0.001]. Pairwise comparisons show that the RVs for DURATION-P (2.968) and
DURATION-O (2.897) were greater than DURATION-M (2.675). The O-M [p < 0.05]
and P-M [p < 0.001] differences were significant, while the P-O difference [0.511]
was not significant. This suggests that when the second tone was higher in F0 by two increments (i.e., the P level), the increase of duration in the second tone by one
increment (i.e., the DURATION-O case) could facilitate tone differentiation, while the
duration increase by two increments (i.e., the DURATION-P case) would not further strengthen tone differentiation. Second, at the O level of F0, DURATION had a significant effect on the RVs [F(2, 375) = 34.05, p < 0.001]. Pairwise comparisons show that the mean RVs for DURATION-P (2.952) and DURATION-O (2.873) were greater than DURATION-M (2.365). The P-M [p < 0.001] and O-M [p < 0.001]
differences were significant, while the P-O difference was insignificant [p = 0.560].
This suggests that when the second tone was higher in F0 by one increment than the first tone, the increase of duration in the second tone by one increment (i.e., the DURATION-O case) could facilitate tone differentiation, while the duration increase by two increments (i.e., the DURATION-P case) would not further facilitate tone differentiation. Third, at the M level of F0, DURATION had a significant effect on the RVs [F(2, 375) = 283, p < 0.001]. Pairwise comparisons show that the mean RV for DURATION-P (2.960) was greater than DURATION-O (2.690), which was greater than DURATION-M (1.302). The P-O [p < 0.01], O-M [p < 0.001] and P-M [p <
0.001] differences were all significant. This suggests that when the two tones had the same F0, the increase of duration in the second tone could facilitate tone
differentiation. Table 10 summarizes the analyses on the interaction between F0 x DURATION.
Table 10: Examining the simple main effect in F0 x DURATION for Set 2
Effect ANOVA result TukeyHSD result
F0 at DURATION-P F(2, 374) = 0.108, p = 0.897 (P = O = M)
F0 at DURATION-O F(2, 375) = 5.257, p < 0.01** O > M*, P > M** (P = O > M)
F0 at DURATION-M F(2, 375) = 106.8, p < 0.001*** P > O**, O > M***, P > M*** (P > O > M) DURATION at F0-P F(2, 374) = 11.33, p < 0.001*** O > M*, P > M*** (P = O > M)
DURATION at F0-O F(2, 375) = 34.05, p < 0.001*** O > M***, P > M*** (P = O > M)
DURATION at F0-M F(2, 375) = 283, p < 0.001*** O > M***, P > M***, P > O** (P > O > M)
Significance code: *** for p < 0.001, ** for p < 0.01, * for p < 0.5. The comparisons between levels shown in the parentheses are based on whether the values for the levels are significantly different.
4.2.2.2 DURATION x GROUP
For the DURATION x GROUP interaction, a one-way ANOVA (DURATION) was performed at each level of GROUP (EN, MI and MA) separately. First, at the EN level of GROUP, DURATION had a significant effect on the RVs [F(2, 374) = 49.17,
p < 0.001]. Pairwise comparisons show that the RVs for DURATION-P (2.912) and
DURATION-O (2.730) were greater than DURATION-M (2.048). The O-M [p <
0.001] and P-M [p < 0.001] differences were significant, while the P-O difference was insignificant [p = 0.119]. This suggests that the increase in duration by one increment (i.e., the DURATION-O case) could facilitate the English speakers’ tone
differentiation, while the duration increase by two increments (i.e., the DURATION-P case) could not further facilitate English speakers’ tone differentiation. Second, at the MI level of GROUP, DURATION had a significant effect on the RVs [F(2, 375) = 38.88, p < 0.001]. Pairwise comparisons show that the RVs of DURATION-P (3) and DURATION-O (2.921) were greater than DURATION-M (2.421). The P-O [p = 0.506]
difference was insignificant, while the P-M [p < 0.001] and O-M [p < 0.001]
differences were significant. This suggests that with the increase of duration by one increment in the second tone (i.e., the DURATION-O case), MI’s differentiation between the two tones became more facilitated. However, the duration increase in the second tone by two increments (i.e., the DURATION-P case) could not further
strengthen MI’s tone differentiation. Third, at the MA level of GROUP, DURATION had a significant effect on the RVs [F(2, 375) = 101.4, p < 0.001]. Pairwise
comparisons show that the RVs of DURATION-P (2.968) and DURATION-O (2.810) were greater than DURATION-M (1.873). The O-M [p < 0.001] and P-M [p < 0.001]
differences were significant, while the P-O difference was not [0.137]. This suggests that with the increase of duration by one increment in the second tone (i.e., the DURATION-O case), MA’s differentiation between the two tones became more facilitated; however, the duration increase in the second tone by two increments (i.e., the DURATION-P case) could not further strengthen MA’s tone differentiation.
For the DURATION x GROUP interaction, the other one-way ANOVA (GROUP) was performed at each level of DURATION. First, at the P level of DURATION, GROUP had a significant effect on the RVs [F(2, 374) = 3.596, p < 0.05]. Pairwise comparisons show that the mean RVs of GROUP-MI (3) and GROUP-MA (2.968) were greater than GROUP-EN (2.912). The MI-MA [p = 0.604] and MA-EN [p = 0.209] differences were insignificant, while the MI-EN difference [p < 0.05] was significant. This suggests that when the second tone was longer in duration by two increments than the first (i.e., the P level), the Mandarin speakers from the MI group are more likely to differentiate the two tones than the English speakers. Second, at the O level of DURATION, GROUP had a significant effect on the RVs [F(2, 375) = 3.746, p < 0.05]. Pairwise comparisons show that the mean RVs of GROUP-MI (2.921) and GROUP-MA (2.810) were greater than GROUP-EN (2.730). The MI-MA [p = 0.251] and MA-EN [p = 0.493] differences were insignificant, while the MI-EN difference [p < 0.05] was significant. This suggests that when the second tone was longer in duration by one increment than the first (i.e., the O level), the Mandarin speakers from the MI group are more likely to differentiate the two tones than the English speakers. Finally, at the M level of DURATION, GROUP had a significant effect on the RVs [F(2, 375) = 10.86, p < 0.001]. Pairwise comparisons show that the mean RVs of GROUP-MI (2.421) and GROUP-EN (2.048) were greater than
GROUP-MA (1.873). The MI-EN [p < 0.01] and MI-MA [p < 0.001] differences were significant, while the EN-MA difference was insignificant [p = 0.314]. This suggests that when the two tones had the same duration (i.e., the M level), Mandarin speakers from the MI group and the English speakers were more likely to differentiate the two tones than the Mandarin speakers from the MA group. Table 11 summarizes the analyses on the interaction DURATION x GROUP.
Table 11: Examining the simple main effect in DURATION x GROUP for Set 2
Effect ANOVA result TukeyHSD result
DURATION at GROUP-EN F(2, 374) = 49.17, p < 0.001*** O > M***, P > M*** (P = O > M) DURATION at GROUP-MI F(2, 375) = 38.88, p < 0.001*** O > M***, P > M*** (P = O > M) DURATION at GROUP-MA F(2, 375) = 101.4, p < 0.001*** O > M***, P > M*** (P = O > M) GROUP at DURATION-P F(2, 374) = 3.596, p < 0.05* MI > EN* (MI > EN)
GROUP at DURATION-O F(2, 375) = 3.746, p < 0.05* MI > EN* (MI > EN)
GROUP at DURATION-M F(2, 375) = 10.86, p < 0.001*** MI > EN**, MI > MA*** (MI > MA = EN)
Significance code: *** for p < 0.001, ** for p < 0.01, * for p < 0.5. The comparisons between levels shown in the parentheses are based on whether the values for the levels are significantly different.
4.2.2.3 F0 x GROUP
For the F0 x GROUP interaction, a one-way ANOVA (F0) was performed at each level of GROUP (EN, MI and MA) separately. First, at the EN level of GROUP, F0
had a significant effect on the RVs [F(2, 374) = 9.236, p < 0.001]. Pairwise
comparisons show that the RVs for F0-P (2.728) and F0-O (2.643) were greater than F0-M (2.317). The P-O [p = 0.676] difference was insignificant, while the P-M [p <
0.001] and O-M [p < 0.01] differences were significant. This suggests that the increase of F0 in the second tone by one increment could facilitate English speakers’
tone differentiation, while the F0 increase by two increments would not further facilitate English speakers’ tone differentiation. Second, at the MI level of GROUP, F0 had a significant effect on the RVs [F(2, 375) = 57.28, p < 0.001]. Pairwise comparisons show that the RVs of F0-P (3.000) and F0-O (2.984) were greater than F0-M (2.357). The P-M [p < 0.001] and O-M [p < 0.001] differences were significant.
The P-O [p = 0.971] difference was insignificant. This suggests that with the increase of F0 by one increment in the second tone (i.e., the F0-O case), MI’s differentiation between the two tones became more facilitated. However, the F0 increase in the second tone by two increments (i.e., the F0-P case) could not further strengthen MI’s tone differentiation. Third, at the MA level of GROUP, F0 had a significant effect on the RVs [F(2, 375) = 14.32, p < 0.001]. Pairwise comparisons show that the RV for F0-P (2.810) was greater than F0-O (2.563), which was greater than F0-M (2.278).
The P-O [p < 0.05], P-M [p < 0.001] and O-M [p < 0.05] differences were all
significant. This suggests that the increase of F0 in the second tone could facilitate MA’s tone differentiation.
For the F0 x GROUP interaction, the other one-way ANOVA (GROUP) was performed at each level of F0 (P, M and O). First, at the P level of F0, GROUP had a significant effect on the RVs [F(2, 374) = 9.326, p < 0.001]. Pairwise comparisons show that the RVs for GROUP-MI (3) and GROUP-MA (2.810) were greater than GROUP-EN (2.728). The MI-MA [p < 0.01] and MI-EN [p < 0.001] differences were significant, while the MA-EN was insignificant [p = 0.419]. This suggests that when the F0 of the second tone was higher than the first by two increments, the Mandarin speakers from the MI group were more likely to differentiate the two tones than the English speakers and the Mandarin speakers from the MA group. Second, at the O level of F0, GROUP had a significant effect on the RVs [F(2, 375) = 15.34, p < 0.001].
Pairwise comparisons show that the RV for GROUP-MI (2.984) was greater than GROUP-EN (2.643) and GROUP-MA (2.563). The MI-EN [p < 0.001] and MI-MA [p < 0.001] differences were significant, while the EN-MA difference was
insignificant [p = 0.588]. This suggests that when the F0 of the second tone was higher than the first by one increment, the Mandarin speakers from the MI group were more likely to differentiate the two tones than the English speakers and the Mandarin speakers from the MA group. Finally, at the M level of F0, GROUP was not found to
have a significant effect on the RVs [F(2, 375) = 0.224, p = 0.799]. Table 12 summarizes the analyses on the interaction DURATION x GROUP.
Table 12: Examining the simple main effect in F0 x GROUP for Set 2
Effect ANOVA result TukeyHSD result
F0 at GROUP-EN F(2, 374) = 9.236, p < 0.001*** O > M**, P > M*** (P = O > M) F0 at GROUP-MI F(2, 375) = 57.28, p < 0.001*** O > M***, P > M*** (P = O > M) F0 at GROUP-MA F(2, 375) = 14.32, p < 0.001*** P > O*, O > M*, P > M*** (P > O > M) GROUP at F0-P F(2, 374) = 9.326, p < 0.001*** MI-MA**, MI-EN*** (MI > MA = EN) GROUP at F0-O F(2, 375) = 15.34, p < 0.001*** MI > EN***, MA > MA*** (MI > EN = MA)
GROUP at F0-M F(2, 375) = 0.224, p = 0.799 (MI = MA = EN)
Significance code: *** for p < 0.001, ** for p < 0.01, * for p < 0.5. The comparisons between levels shown in the parentheses are based on whether the values for the levels are significantly different.
4.2.2.4 F0 x DURATION x GROUP
For the F0 x DURATION x GROUP interaction, three two-way ANOVAs (F0 x DURATION, DURATION x GROUP and F0 x GROUP) were respectively conducted at each level of GROUP, at each level of F0 and at each level of DURATION. Since the significant two-way interactions F0 x DURATION, DURATION x GROUP and F0 x GROUP have been discussed in Sections 4.2.2.1, 4.2.2.2 and 4.2.2.3, they will not be displayed in text here again. Similar to Section 4.2.1.4, only the gist of the
analyses on F0 x DURATION x GROUP will be included in the text. Table 13 presents details of the analyses on the F0 x DURATION x GROUP interaction.
First, in the F0 (P, O, M) x DURATION (P, O, M) ANOVA, F0 and DURATION were found to have a significant effect on the RVs at all levels of GROUP. An
interaction was found between F0 and DURATION at all levels of GROUP. Pairwise comparisons suggest that: (1) The increase of F0 in the second note by one increment could facilitate tone differentiation. The F0 increase by two increments would further facilitate tone differentiation in the MA group. (2) The increase of duration in the second note by one increment could facilitate tone differentiation, while the duration increase by two increments would not further facilitate tone differentiation.
Second, in the DURATION (P, O, M) x GROUP (MI, MA, EN) ANOVA, DURATION was found to have a significant effect on the RVs at all levels of F0, while GROUP was found to have a significant effect on the RVs at the P and O levels of F0. An interaction between DURATION and GROUP at the P and O levels of F0.
Pairwise comparisons suggest that: (1) The increase of duration by one increment in the second tone would facilitate tone differentiation. The duration increase by two increments would further strengthen tone differentiation only in the two tones have the same F0. (2) At the P and O levels of F0, Mandarin speakers from the MI group
are more likely to differentiate the two tones than the English speakers and the Mandarin speakers from the MA group.
Finally, in the F0 (P, O, M) x GROUP (MI, MA, EN) ANOVA, F0 was found to have a significant effect on the RVs at the O and M levels of DURATION. GROUP was found to have a significant effect on the RVs at all levels of DURATION.
Pairwise comparisons suggest that: (1) The increase of F0 in the second tone by one increment could facilitate tone differentiation. The F0 increase by two increments would facilitate tone differentiation only when the two tones have the same duration.
(2) Mandarin speakers in the MI group were more likely to differentiate the two tones than the English speakers. When the two tones have the same duration, Mandarin speakers in the MI group were more likely to differentiate the two tones than the English speakers and the Mandarin speakers in the MA group.
Table 13: Examining the simple main effect in F0 x DURATION x GROUP for Set 2
Effect ANOVA result TukeyHSD result
F0 x DUR at GRO-EN F0: F(2, 368) = 13.07, p < 0.001***
DUR: F(2, 368) = 57.93, p < 0.001***
F0 x DUR: F(4, 368) = 11.32, p < 0.001***
F0: O > M***, P > M*** (P = O > M) DUR: O > M***, P > M*** (P = O > M) F0 at DUR-P (at GRO-EN):
F(2, 122) = 0.053, p = 0.948;
(P = O = M)
F(2, 123) = 0.34, p = 0.712;
DUR at F0-P (at GRO-MI):
O > M***, P > M*** (P = O > M)
GRO: F(2, 369) = 20.09, p < 0.001***
F0 x GRO: F(4, 368) = 0.237, p = 0.918
between levels shown in the parentheses are based on whether the values for the levels are significantly different.
Chapter Five: Discussion
This chapter discusses the results of the experiment. Section 5.1 discusses the results of the stress perception test. Section 5.2 discusses the results of the pure tone perception test.
5.1 Stress perception test
According to Wang (2008) and Zhang & Francis (2010), compared to English speakers, Mandarin speakers tend to rely on F0 as a cure for stress, because Mandarin speakers are under the influence of Mandarin tones, which use F0 as an important cue.
However, the current study found that the Mandarin speakers and English speakers did not differ in their use of acoustic cues, though Mandarin speakers tended to give a higher rating (i.e., ‘absolutely’ noun-like or verb-like) than English speakers did. It was found that all listeners in general used F0 for the noun form and duration for the verb form. Therefore, L1 influence was not able to explain the Mandarin speakers’ use of cues in the current study. The current study is not the only case where Mandarin speakers’ perception of stress does not seem to be affected by L1 influence. In
Chrabaszcz et al. (2014), Mandarin speakers and English speakers were found to have the same ranking of cues for stress. In response to their results, Chrabaszcz et al.
(2014) pointed out that the difference between L1 and L2 would not necessarily
hinder the L2 learners from achieving native-like stress perception. However, this explanation did not provide the reason for using or not using a cue.
In Fry (1958), it was found that the higher the duration ratio (of the first vowel to the second) or amplitude ratio (of the first vowel to the second) was, the more likely the English speakers would recognize a noun form. Motivated by Fry (1958), the current study would like to examine the stimuli’s acoustic parameters, and use the saliency of the increment in an acoustic parameter to explain the listeners’ use of acoustic cues. The saliency in this study is defined as the percentage of the
increment’s value over the unstressed syllable’s value in the non-synthesized original token produced by the English speaker. Take PROdawn’s F0 as an example. The increment is 70 Hz, and the unstressed syllable’s F0 is 172 Hz. In this case, the saliency of the F0 increment is (70/172)*100, which is 41%. Table 14 presents the saliency of the increment in each acoustic parameter.
Table 14: The saliency of each parameter Form
Parameter
Noun form
PROdawn
Verb form
proDAWN
F0 41% 5%
Duration 38% 229%
Amplitude 1% 3%
As can be seen in Table 14, among the acoustic parameters, the most salient parameter for the noun form is F0, while duration has the greatest saliency for the verb form.
The findings in the current study show that the listeners relied on F0 to perceive stress in noun, while they replied on duration for the verb perception. It turned out that the cue being relied on was the acoustic parameter with the greatest saliency, and both Mandarin and English speakers follow this rule. Therefore, the saliencies of the acoustic parameters might be a factor to account for stress perception.
In Beckman (1986), both amplitude and duration were found to work together as cues for stress. However, in the current study, amplitude was not a cue for stress. As can be seen in Table 14, amplitude was the parameter with the lowest saliency in the noun form and in the verb form. It is possible that the saliency of amplitude was too low to be noticed by the listeners.
Although Mandarin speakers and English speakers were not found to use the cues differently in the current study, they differed in their response values in general:
It was found that Mandarin speakers’ response values to the noun forms as well as the verb forms were greater than English speakers. Since the listeners responded based on how certain they were about the location of the stress (e.g., absolutely/probably on the first/second syllable), it appeared that English speakers tended to be less sure about their judgments. This might have to do with how Mandarin speakers and English
speakers acquire English stress. Mandarin speakers acquire English stress by listening to it consciously, while English speakers grow up acquiring stress naturally. The stress perception test was a scenario where listeners had to identify the stress consciously.
The test was more difficult to the English speakers than Mandarin speakers because the English speakers had to identify a feature that they may not be aware of. The finding that Mandarin speakers were more certain about their judgment implies that a perception test (where listeners have to make judgment consciously) might
underestimate English speakers’ ability in stress perception. (It is not that they cannot perceive stress. Rather, it is that they were not used to perceive it consciously.) The finding also sheds light on the importance of using a Likert scale. If a forced-choice test (as in Lai, 2008; Wang 2008; Zhang & Francis, 2010; Chrabaszcz et al., 2014) instead of a Likert-scale test were adopted, the difference in the certainty about the judgment would not be found. Therefore, a Likert scale has its worth in stress perception study.
The current study not only works on the difference between English and Mandarin speakers but also works on the difference between Mandarin speakers at two different proficiency levels. In Lai (2008), two groups of Mandarin speakers (beginning and advanced learner of English) were found to have different use of cues.
The current study not only works on the difference between English and Mandarin speakers but also works on the difference between Mandarin speakers at two different proficiency levels. In Lai (2008), two groups of Mandarin speakers (beginning and advanced learner of English) were found to have different use of cues.