Great White Shark Cage Diving in Gansbaai

I believe this photo is my most appropriate one yet, since it pretty much sums up what I saw when I first looked at my data – a large, Great White Shark, coming slowly towards me, although in the case of my data, I had no cage protecting me (I did with the sharks).

For those of you who have never done data analysis, let me give you some advice: If you are a Mathematics Major, but only study theoretical math, do not let people tell you that your “math background” will help you with data analysis. Yes, it theoretically should, and it does make things a little easier, such as when you’re able to understand half of the theory behind principal components analysis rather than none of it, and you can prove to yourself that the numbers are doing what they should. However, you still need to understand how to organize your data, how to move and extract large chunks of it without copying and pasting, and how to put it in beautiful charts and graphs so that someone can actually make something of it. These are all things I’ve learned a little about in the past few weeks.

Execel and Matlab = my new best friends. W&M has a license such that all students can download and use Matlab on their personal computers, so that’s what I did. I am lucky enough to be engaged to a beautiful woman who also happens to use Excel and Matlab a lot, so I got a tutorial from her on Excel macros and how to read in my data and extract what I want for analysis. If you’re curious, the “find,” “strcmp,” and “strfind” functions are also new acquaintances of mine.

While in South Africa and for an entire week after, I organized my data into Excel spreadsheets, and with some help got it all into a format that was Matlab-friendly. Then came the analysis – aka “type things into Matlab and hope it does what you want – which was long a grueling. You see, even with computer aid, there was still a lot of copying and pasting for me to do due to my crippling lack of computer know-how that would probably prevent it (I hope, or else a lot of scholars have really worn-out ‘C’ and ‘V’ keys).

The biggest issue I ran into with my data was how to deal with incompleteness. Some of the trials I ran had to be cut short, and thus a lot of my participants did not finish the matched-guise study What I ended-up having to do was eliminate their matched-guise studies, but luckily I made sure they all did the supplemental questionnaire, so for that analysis I still had 81 pieces of data!

For my final Monroe Project, I just did a lot with averages. I chose not to do any Factor Analysis just yet, since I need to study that a lot more before attempting it. For the sake of doing this initial analysis of the data, I  narrowed the 18 scales per recording (270 total) rated by each participant to six per recording (90 total). The six I chose were: poor-rich, unintelligent-intelligent, friendly-unfriendly, unlikeable-likeable, passive-active, and weak-strong. I chose these specifically by first being sure to choose two each from Zahn and Hopper’s (1985) three major factors (superiority, social attractiveness, and dynamism), while still maintaining what I felt was semantic distance within those factors. I also was sure to not include ones such as unclear-clear and untalented-talented, which seemed to confuse some participants (they asked about them during the study).

Narrowing the data as such truly made the analysis using Matlab loads easier. For the project I did some simple analyses using things such as gender, languages spoken, family languages, languages on the recordings, etc. Some results I found were expected, such as English speakers being rated higher on the superiority scales and Xhosa on the social attractiveness scales. Also, some confirmed results I know I’ll find in the qualitative data, such as Afrikaans experiencing a negative extrinsic attitude overall.

I only analyzed the quantitative data for my Monroe Project, since I plan to use the other, qualitative data for my Honors Thesis in Linguistics! So exciting and daunting at the same time, but no matter what this was an awesome experience that will definitely stay with me forever! =)

Thanks for reading,

Mike

Penguins at Boulders Beach

So, in my last post, I focused on building the matched-guise study.  Even though I probably left that I little confusing, I would like your permission to move onto what proved to be a more frustrating part of my study – the interviews. Given that this blog is primarily meant to shed some light on the W&M undergraduate research experience, I think such a post would be more relevant. Thanks for understanding!

To give a brief initial summary, once I reached South Africa, I spent the next day meeting with Linguistics professors at the University of Cape Town and visiting some of their classes to ask for participants. That part went very well, and lots of students signed-up for the matched-guise study, the primary, quantitative part of my study. Fewer than who signed-up came, but that was okay, since a bunch still showed and I was still gathering data. Furthermore, my advisor’s voice was always in the back of my head saying “Any data is good data!” – thus bringing me joy at the appearance of any participant at all.

After they had participated in the matched guise study, I gave the participants the option to sign-up for a sociolinguistic interview with me, based on their answers to the supplemental questionnaire that more directly elicited their language attitudes. A lot signed-up! I was happy.

I also went into the large Level 1 Linguistics class and pleaded for interviews. Even more signed-up! I was happier.

Then, learning the hard way, I found out the reality of asking for and scheduling  voluntary follow-up interviews for a two-week linguistics study: there is no accountability whatsoever, and you will not be in their lives for long (zero-guilt). So, out of the about thirty students who signed-up for interviews, I got about ten. Still a decent number (Any data is good data!), but not one that kept me a happy clam at the time. What made it worse was that I could not exactly do things other than sit and read articles, all alone in the room I was provided, as I waited for invisible participants. The thought that someone might actually show kept me sitting still for days. I was kind of unhappy. :/

However, in retrospect, getting the interviews I did was great! I also filled some of the time with interviews with faculty and past-grad students, which were fun to conduct.

My core interview schedule was based on my supplemental questionnaire, completed at the conclusion of the matched-guise study. This questionnaire directly elicited both intrinsic and extrinsic attitudes (based on Ramsay-Brijball’s 2004 study of students at the University of Durban-Westville in KzaZulu-Natal) towards Xhosa, English, Afrikaans, and codeswitching. To look at intrinsic attitudes, I asked them to rate on a scale their agreement with:

• Speaking _____ is important to my identity.
• Being able to switch between two or more languages is important to my identity.

To elicit extrinsic attitudes, simply replace “important to my identity” with “necessary to my success in life.” If you are confused about intrinsic vs. extrinsic, I like to think of the difference by how they are influenced – intrinsic attitudes are influenced from your own personal values and beliefs, while extrinsic attitudes are influenced by the outside world. This is where we get the difference between “important” and “necessary.”

The supplement also asked for personal information, such as majors and minors, age, gender, and then a linguistic profile. The profile had four questions:

• What languages or language varieties do you speak?
• What languages or language varieties do you speak with your family?
• What languages or language varieties do you speak with your friends?
• What languages or language varieties do you speak on campus?

You may wonder why these distinctions are important – it’s because the focus is attitudes. Someone who speaks both English and Afrikaans, but only speaks Afrikaans at home, could very likely have a different attitude towards the language than someone who regularly codeswitches between the two with his friends. This data will also be used to find different correlations with the matched-guise data, since the participants each had a code to keep track of these things.

So, my interviews basically consisted of tailoring follow-up questions to the answers provided on these supplements. I found out rather quickly that they did not need to last long, since I had very direct questions that did not allow for excellent follow-up material (answers I could use for further questions). However, every once and awhile I would get a talkative person (a.k.a. a breath of fresh air) who would give me great answers to form follow-ups.

In summary, follow-up interviews are unpredictable in every way. You literally never know who/what/when/why you’re going to get what you do, so just be happy when you get something! =)