We presented some of our work at the Second Language Research Forum held at the Michigan State University in September 2019. You can find our poster below. In addition, we answer some of the questions that we received during our presentation, in an effort to make the poster more accessible. If you have any further questions, please don’t hesitate to contact us!
Cite as: Campos-Astorkiza, R., O. Muxika-Loitzate, M. Towne, T. Zimmerman and K. DeLeon. 2019. Mixed methods to examine the acquisition of gradient allophonic variation. Presented at Second Language Research Forum, Michigan State University, Sept 20-22, 2019
- Could you give examples of words containing /b, d, g/ in Spanish?
These are examples of some of the words we analyzed:
lago /lago/ “lake” doce /dose/ “twelve”
bebé /bebe/ “baby” comida /komida/ “food”
- Which textbook do you use to teach pronunciation?
All of our data come from sections of the course Spanish Pronunciation that use the textbook “Sonidos en Contexto” by Terrell Morgan.
- What type of instruction do students receive?
These are the instructions that students receive as part of the module:
“Welcome to the recording part of your Spanish Pronunciation assignment. Before you begin, please make sure you are comfortably seated in a quiet room with no background noise. If you have a plug-in microphone for your computer, please attach it and use it.
The recording task is divided into 6 short lists of English words and 7 short lists of Spanish words. For the duration of each short list (2-3 minutes), please do not do anything else, or change windows on your computer, as this will disrupt the recording process. You will be able to pause between each list.
When you are ready to begin, click the BEGIN button below, then read the words as they appear. This list will all be English words”
After they complete the first list, the move to the next page and get the following instructions:
“This is the second of six English word lists. As before, please read the words that appear below.
Please don’t do anything else or change windows on your computer during the recording. When you are ready to continue, click the BEGIN button below.”
They get similar instructions in all subsequent pages.
- Have you considered analyzing the CV intensity ratio for each of the categories (voiced stops, approximants, and taps) separately?
For this presentation, we decided to group together approximants, voiced stops and taps for the analysis of the CV-intensity ratio in order to follow a similar methodology to previous studies on voiced stop weakening in Spanish (Carrasco et. al 2012, Rogers and Alvord 2014) and to make comparisons with their results. As we move forward, looking only at approximants from a continuous perspective will shed more light into our question related to the gradient nature of the weakening process under study. The categorical analysis captures the pattern for other categories and thus, would allow us to focus just on approximants for the continuous one.
- Do you have a control group in your study?
We don’t have a separate control group. In our study, we compare participants to themselves at the beginning and at the end of the semester. This methodology allows us to avoid having to choose a control group, which would be hard to decide on. For instance, one might think of native speakers as a possible control group. However, this kind of comparison (learners vs. native speakers) can be problematic since they rely on the assumption that learners’ aim is to sound exactly like native speakers and studies have shown that this might not always be the case (Solon 2018). Furthermore, there is ample literature on monolingual Spanish speakers’ production of voiced stops, which allows us to compare our data to those findings, especially in terms of what linguistic factors condition the gradient phenomenon of stop weakening which is the focus of our study.
- Why do you give students instant feedback about their vowels and then analyze /b, d, g/?
This poster is part of a bigger project where we actually plan to analyze more sounds, not just voiced stops. Some of the sounds for which we have data include vowels, voiceless stops and laterals. The instant feedback for the moment is for vowels but further feedback on students’ voiceless stop production is being developed. What kinds of feedback is given to students depends on what is feasible. For instance, it is possible to conduct instant acoustic analysis of vowels and generate vowel space plots. Similarly, automatic measuring of VOT for voiceless stops can be developed to provide instant feedback. On the other hand, automatic analysis of voiced stop production, in terms of stop vs. approximant productions, is a much more complicated task.
- Do you find that both the categorical and the continuous analysis are necessary?
Our findings suggest that the categorical and continuous analyses complement each other and for that reason, our conclusion is that both are necessary. The results from the categorical analysis shows that there is development in the types of productions that participants use, i.e. changes in stop vs. approximant production, including use of voiceless stops. More precisely, we find that participants show a change in the categories they use in different contexts. These changes reflect a pattern of acquisition that would be missed if we only considered a continuous analysis. On the other hand, some of the categories might be present gradient differences in the way they are produced. Most notably, approximants might be less or more weakened, mirroring the actual pattern of production that native speakers display. In order to better understand how participants acquire the weakening process, it is important to see if they also show this gradiency and what factors condition it. Our continuous analysis allows us to answer this question.
- Are students able to perceive differences between voiced stops and approximants?
This is a great question but unfortunately, we don’t have perception data to answer it. Our impressions from teaching students the alternation between voiced stops and approximants is that they can perceive the difference.
- Did you have inter rater reliability for the categorical analysis?
We didn’t have systematic inter-rater reliability. However, all team members that categorize the data received intensive training that included in-person sessions where each member would categorize the same tokens and then, we’d compare our categorization to other members’. In addition, any problematic tokens were reviewed by two team members (one of them was always prof. Campos-Astorkiza) and a consensus was reached on the exact label. While inter-rater reliability would be ideal, we think that our system of categorization is reliable and based on effective training.
Carrasco P., Hualde, J.I. and Simonet, M. (2012). Dialectal differences in Spanish voiced obstruent allophony: Costa Rican versus Iberian Spanish. Phonetica 69, 149-179.
Rogers, B., and Alvord, S. (2014). The gradience of spirantization: Factors affecting L2 production of intervocalic Spanish [b,d,g]. Spanish in Context, 11:3.
Solon, M. (2018) The acquisition of second language Spanish sounds. In K. Geeslin (ed.), The Cambridge Handbook of Spanish Linguistics. Cambridge University Press. 668-688.