Page 16. An /ŋ/-initial word was erroneously coded as /n/-initial in the word list file that was joined with the production data. In this correction we update the analyses and provide new values and figures for the merger. We report these corrections with some of the original context.
Correction of this error alters the exact values reported for the initial two-raters inter-rater reliability for the /n/- and /ŋ/-initial words, which are reported in an updated version of Table 5. The italicized values indicate a change from the original text.
|Historical Phoneme||Number of Items Coded||Kappa Statistic||p-value||Percentage Agreement|
|vowel onset||1068||0.59||< 0.001||81%|
Page 22. The section for the production results for the [ŋ]↔Ø merger in 3.1.3 are updated here. The amount of production data for the [ŋ]↔Ø merger is increased (n = 1170). Adjustments for the regression model output (Table 8), group-level visualization (Figure 6), and individual scatterplot (Figure 7) are presented.
|Historical Pattern: Age Group||0.1576||0.1363||1.1570||0.247|
|Historical Pattern: Gender||0.02157||0.1406||–0.1530||0.878|
|Age Group: Gender||0.4777||0.5540||0.8620||0.389|
|Historical Pattern: Age Group: Gender||–0.1472||0.1430||–1.0300||0.303|
The most relevant change is that the historical pattern is not statistically significant (p = 0.239), meaning there is no clear evidence of group-level contrast between [ŋ] and null-initial (Table 8). This is consistent with individual-level patterning where none show robust contrastiveness (Figure 6). Visual interpretation of Figure 7 does not change: Although the corrected data show two more older individuals merging towards null (top right quadrant), younger speakers are still the majority of individuals merged towards null as six of the nine participants are younger.
Comparing across the three mergers, [ŋ]↔Ø is the only one that does not maintain clear patterns of community-wide contrast.
Page 32. We provide an updated Figure 15 that provides unchanged panels for the [n]→[l] and [ŋ̩]→[m̩] mergers, but are presented with the corrected panel for [ŋ]↔Ø merger to facilitate comparisons across the mergers. As before, there was no relationship between merger in perception and production for the [ŋ]↔Ø merger [t(47) = –0.14, r = 0.02, p = 0.89]. Figure 16 presents the direction of misalignment within an individual for each merger, though the values in the third panel for the [ŋ]↔Ø merger are the only ones that have been updated.
Page 39. Table 14 updates the full set of findings for [ŋ-]↔Ø. The changes are noted in red font. What changes here is that, with the correction, we find no evidence in the lexical classes contrasting in production at the community level. This indicates that the community-level production-perception relationship is mismatched. At the individual level, there are some small adjustments to the number of individuals who adhere to the particular patterns, but the overall lack of a production-perception correlation at the individual level is maintained, though the actual statistical values need small adjustments.
|Production (PROD)||Perception (PERC)||Production-Perception|
|Community||Overall||[ŋ]-variant more likely than Ø||[ŋ]-variant more likely than Ø||• PROD and PERC are overall mismatched|
|Lexical classes||No contrast between Ø and /ŋ/||Contrast between Ø and /ŋ/|
|Age||None||Trend towards younger more contrastive than older|
|Individual||Overall||• Few (n = 7, 14%) use mainly Ø for both (innovative)
• Few (n = 5, 10%) use mainly Ø for both somewhat variably
• Most (n = 37, 76%) use mainly [ŋ] for both (hypercorrective)
|• Majority (n = 50, 98%) had little to no contrast (merged)
• Few (n = 1, 2%) had intermediate contrast
• None had close to discrete contrast (categorical)
|• PROD and PERC are overall not correlated (r = 0.02)
• Perception leads early: Those who contrast more in PROD tend to contrast relatively less in PERC
• Perception lags late: Those who contrast less in PROD tend to contrast relatively more in PERC
|Age||None||Those most contrastive are generally younger|
Page 41. We update the language surrounding the overall status of this merger in production: For the [ŋ]↔Ø merger, the data suggest that all groups produce less null-initial for both historical categories, at a mean rate of 24% null (i.e., 76% [ŋ]). No one robustly maintains the contrast in production.
The authors have no competing interests to declare.