css.php

10 Replies to “Cognate Status, Voicing, and Spanish Language Dominance: How Heritage Language Learners’ Perceive Intervocalic [z] in Cognates versus Non-Cognates (Rachel Sangster Garza)”

  1. Great job and fascinating topic. l am arious to know if you formed the same results in tern of cognates vs. non- cognates for the ” native speakers non- heritage group”. Also, I have a couple of suggestions for analysis since you seem to be still collecting data. I am not sure if the items were presented in a randomized order each time, but if not, might be necessary to study “the floor effect”, all the most since as you comment the “r” items carried more weight in deciding on “native_ likeness”. I recently read a study that used a match methodology and found a floor effect in every single case, sometimes as the primary effect. More importantly perhaps, ratings are ordinal, not ratio- scale. There are many analysis that can handle this kind of data, but multiple linear regression is not one of them. You may want to consider polinomial or ranked logistic regression or perhaps conditional trees. I could send some references or sample studies using these. For correlation ranked Spearman rather than Pearson and use the subject in a mixed effect design so you can use all measures rather than the average of them which treats the data as ratio scale and makes you lose information. Hope this is helpful, happy to provide some references and more in depth comments on why this scale is ordinal and ratio treatment would misrepresent it.

    1. Thank you so much for your comments and questions. As far as the native speakers/non-heritage group, unfortunately I wasn’t able to get enough responses to run a regression. So, statistically I am currently unsure what the results of this group would be as far as how the independent variables of the study might have explained the native speakers’ ratings of each pronunciation. It definitely warrants further data collection! However, each group followed the same trends as far as the average ratings that they gave per category of word. In other words, all speakers gave the lowest ratings of Spanish-like accent when there was an [-z-] in a non-cognate but the highest ratings when there was an [-s-] in a cognate (average ratings are displayed on slide 14).

      As far as randomization, each survey was randomized using Qualtrics. So, in theory, each participant did not hear the recordings in the same order. However, I do think that your comment about “floor effect” is valid and that is absolutely something I would like to take into account going further. And thank you for your suggestions on which statistical methods could handle this data. Since creating this presentation I have learned more varied statistical methods and I agree that either a logistic regression or conditional trees would be better suited.

      Thank you again for your time and your feedback. This is very helpful, and I have made notes for continuing this project.

  2. Great job and fascinating topic. I was wondering if you found the same difference between non-cognate and cognates [z] in the non-heritage native speakers group.

  3. Since it looks like you may still collect and analyze more data, I have a couple of suggestions for the analysis. First, if the presentation of the items was not randomized for each subject, have you considered adding the position of the stimulus (before or after a cognate or the r distractors and earlier on in the test or later on in the test) as a variable? There are studies that have included position and all the ones I have seen found it to be significant, showing “a floor effect”, that is, participants judge the item in comparison with the previous one they heard. Second, the data is ordinal rather than ratio scale. There are analysis methods that are designed for ordinal data. For instance, for correlation the rank Spearman correlation rather than Pearson, and instead of multiple linear regression, a polynomial logistic regression or conditional inference trees, or many other options actually. Happy to send some references and sample studies your way on any of this. Finally, you could include the participant as an independent variable (random) so you do not have to lose information by using the average rather than all the data points (and also again treating the variable as ratio-scale rather than ordinal). Hope this is helpful! Again, very interesting study!

    1. Thank you for your comments! I’m going to address here what I may have missed in my previous response.

      The presentation of the items was randomized per participant, however I am interested in the idea that you have presented here as a way to evade “floor effect”. I do think position as well as quantity of stimuli could have posed some effects here. Going forward I hope to address the issue of “floor effect” more than I did here.

      Thank you again for your comment on the data. I hadn’t considered making the participant group an independent variable. That would have been better here. I have taken notes about the data analysis that you recommend and I would love to see any references that you might have.

      Thank you again for your comments and for taking the time to view my presentation.

  4. I also have some suggestions for the data analysis, since the data is ordinal rather than ratio-scale. Let me know if you would like some references and pointers there. I tried to post this, but I am not sure long comments work in here.

  5. Interesting study, clear and compelling presentation. The findings relate nicely to work I’ve been doing on the use and acceptability of different types of calques in Miami. I have found that among second- and third-generation speakers, use and acceptability of calques is significantly conditioned by Spanish language ability (regression analysis), along with level of formal education (not in Spanish language, but in general).
    I look forward to seeing more of your work in the future! 🙂

  6. Apologies for the multiple postings. yours was the first presentation I saw and I was figuring out how this works. You can email me for references to irene.checa@uwyo.edu.

Comments are closed.