![QUANTITATIVE RESEARCH IN LINGUISTICS
22
operationalize this measurement. And as such, it is paramount that we have
a good grasp as to what our variables could potentially look like, before
we develop a method to capture as much of the variation in this variable
as possible. Phonology is a good linguistic example to illustrate this.
Traditionally, sociolinguists and dialectologists have focused on variation
in pronunciation, and have tried to explain this variation along various
dimensions such as social categories or geographic areas. Sebba (1993:
23) in his study on ‘London Jamaican’ discusses some of the phonological
differences between Jamaican Creole and London English, and identifies
for one particular speaker in his sample two distinct realizations for
the pronunciation of the /u/ sound in ‘who’ and the initial consonant in
‘the’: [hu:] versus [hʊ:], and [di] versus [đə]. If, like Sebba, we consider
this phonological difference to be important, we must use a measure that
accounts for this. In other words, we need an operation definition that
allows us – in practice – to clearly classify a particular vowel as [u:]/[ʊ:] and
a particular consonant as [d]/[đ]. At the same time, if such differences are
not of interest for our study, we do not need to account for them. The trick
is to develop a methodology which allows us to measure as accurately as
required but does not make things unnecessarily complex.
A word of warning: some concepts, particularly abstract ones and those
that comprise several variables and sub-concepts, are notoriously difficult
to measure. Motivation, as outlined above, is one such concept, as are
attitudes. A quantitative analysis of these takes time, effort and a lot of trial
and error. Possibly more than for any other study, the thorough evaluation
of previous work and in particular the methodologies used (including all
the difficulties and pitfalls!) is vital. It is usually a good although resource-
and time-consuming idea to support quantitative data on these issues
with qualitative findings from interviews or focus groups. Often, the
combination of qualitative and quantitative results provide a far better and
reliable insight than one approach on its own.
2.4. Independent and dependent variables
Following our discussion of variables in general, we now turn to the
relationship variables can have to each other. More specifically, we focus
on independent and dependent variables. The crucial difference between
independent and dependent variables is that the latter can be influenced by
the former, but not vice versa.
A common independent variable in linguistic research is age. Research
on second-language acquisition has shown that age influences – to some
extent – acquisition success, with older learners being more likely to achieve
lower proficiency in the second language than younger ones. In this case,
language acquisition success depends on age, and is hence the dependent](https://image.slidesharecdn.com/25339233-250516144922-7564ef89/85/Quantitative-Research-In-Linguistics-An-Introduction-Sebastian-M-Rasinger-31-320.jpg)
![if to the left, they give a sharp jerk. Sometimes when the master is
angry he will pull the poor carabao’s nose, so that he tears the piece
of flesh out altogether; not at all an uncommon occurrence, and
nothing distressing to a Filipino.[3] In the days of the rebellion
against Spain, a few years ago, when the Filipinos caught the hated
Spanish friars, they ran a rope through the priests’ noses, tied their
hands, and led them about like the carabaos, so that people might
spit upon the hated tyrants, and insult them at their own pleasure.
The carabaos are as gentle and amenable as horses with the
natives; quite tiny children ride and bully the huge beasts, looking so
comically small on the big backs, with their tiny brown legs hardly
reaching to each side of the broad ribs, and driving whole herds with
the most perfect independence and self-possession. The carabaos
are not at all safe as regards white people, however, for they can
smell and detect them at an immense distance; and they will
occasionally charge them ferociously, so that they are very
dangerous in the open country. I have heard some horrible stories of
carabaos killing and trampling on white men in out-of-the-way
places. They don’t gore, I suppose because their horns are so flat,
but they trample to death, which does just as well.
These great grey, lumbering animals are very picturesque, and
redeem many a Philippine scene from utter dulness as they go
shambling along, drawing the native two-wheeled cart, with its big
hood of brown matting filled with bundles of emerald-green sacate
grass. They can shamble at an amazing pace, and that is their usual
gait; but they can gallop, too, as quickly as a horse.
Besides the herds of carabaos, we saw several natives down in the
bed of the river, going out to certain spots where the shelve of the
sand was more abrupt for their supply of water. These were women,
of course, for women do all the household tasks, even the most
burdensome, their lords being busy standing about the roads or
Plazas, or attending a cock-fight.
These women had long bamboo poles, with the divisions knocked
out and the end closed up, which they laid in the running stream to](https://image.slidesharecdn.com/25339233-250516144922-7564ef89/85/Quantitative-Research-In-Linguistics-An-Introduction-Sebastian-M-Rasinger-80-320.jpg)
Quantitative Research In Linguistics An Introduction Sebastian M Rasinger
- 1. Quantitative Research In Linguistics An Introduction Sebastian M Rasinger download https://ebookbell.com/product/quantitative-research-in- linguistics-an-introduction-sebastian-m-rasinger-50678466 Explore and download more ebooks at ebookbell.com
- 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Quantitative Research In Linguistics An Introduction Second Edition Sebastian Marc Rasinger https://ebookbell.com/product/quantitative-research-in-linguistics-an- introduction-second-edition-sebastian-marc-rasinger-5225968 Quantitative Methods In Cognitive Semantics Corpusdriven Approaches Cognitive Linguistic Research 1st Edition Dylan Glynn https://ebookbell.com/product/quantitative-methods-in-cognitive- semantics-corpusdriven-approaches-cognitive-linguistic-research-1st- edition-dylan-glynn-2524144 Qualitative Research In Applied Linguistics A Practical Introduction Juanita Heigham https://ebookbell.com/product/qualitative-research-in-applied- linguistics-a-practical-introduction-juanita-heigham-2217866 Quantitative Research In Communication 1st Edition Mike Allen B Scott Titsworth Stephen Hunt https://ebookbell.com/product/quantitative-research-in- communication-1st-edition-mike-allen-b-scott-titsworth-stephen- hunt-51259668
- 3. Quantitative Research In Education A Primer Paperback Wayne K Kolter Hoy Curt M Adams https://ebookbell.com/product/quantitative-research-in-education-a- primer-paperback-wayne-k-kolter-hoy-curt-m-adams-10016684 Doing Quantitative Research In Education With Ibm Spss Statistics 3rd Edition Daniel Muijs https://ebookbell.com/product/doing-quantitative-research-in- education-with-ibm-spss-statistics-3rd-edition-daniel-muijs-46665756 Reporting Quantitative Research In Psychology How To Meet Apa Style Journal Article Reporting Standards Second Edition Revised 2020 Second Edition Revised Cooper Phd https://ebookbell.com/product/reporting-quantitative-research-in- psychology-how-to-meet-apa-style-journal-article-reporting-standards- second-edition-revised-2020-second-edition-revised-cooper-phd-56445276 Conducting Quantitative Research In Education Saiyidi Mat Roni https://ebookbell.com/product/conducting-quantitative-research-in- education-saiyidi-mat-roni-10463488 Reporting Quantitative Research In Psychology How To Meet Apa Style Journal Article Reporting Standards 2nd Edition Harris Cooper https://ebookbell.com/product/reporting-quantitative-research-in- psychology-how-to-meet-apa-style-journal-article-reporting- standards-2nd-edition-harris-cooper-10812350
- 7. Figures and tables Figures 2.1 Quantitative-deductive approach 11 2.2 Qualitative-inductive approach 12 3.1 Longitudinal versus cross-sectional designs 40 3.2 Setup of Li’s study 42 3.3 Sampling 48 3.4 Sampling error 49 5.1 Bar chart 106–7 5.2 Pie chart 108 5.3 Bar chart showing feature distribution across LMC, UWC and LWC 109 5.4 Bar chart showing feature distribution between Detroit and Norwich 110 5.5 Line graph showing feature distribution between Detroit and Norwich 111 5.6 Line graph showing fixation times for boys and girls 112 5.7 Fixation times for boys and girls 113 5.8 Score A and Score B scatter plot 115 6.1 Arithmetic mean 122 6.2 Median in relation to the arithmetic mean 127 6.3 Example: Linguists and bid success 135 6.4 Normal distribution – mean, median and mode are identical 139 6.5 Right-skewed distribution 141 6.6 Left-skewed distribution 141 6.7 Z-score in skewed distributions 141 6.8 Normal distribution. Skew=0 142 6.9 Left-hand/negative skew. Skew= −0.64 143 6.10 Right-hand/positive skew. Skew= 1.31 143 7.1 Scatter plot for fictive data set 162
- 8. figures and tables ix 7.2 Scatter plot with line 163 7.3 Scatter plot showing positive and negative correlation between variables x and y 164 7.4 Correlation for two variables and 12 speakers 170 7.5 Scatter plot with regression/trend line 175 7.6 Scatter plot for multiple regression example 182 8.1 EG CG bar chart 200 9.1 Curvilinear graph 221 10.1 T-test – 1 IV, 1 DV 237 10.2 ANOVA – 2 IVs, 1 DV 238 10.3 ANOVA – 3 IVs, 1 DV 239 10.4 MANOVA – 2 IVs, 2 DVs 239 10.5 MANOVA – 3 IVs, 8 DVs 239 11.1 Task 8.3 – bar chart with error bars 271 Tables 2.1 Basic set of quantitative data 10 2.2 Quantitative analysis of police interview 15 3.1 Summary of different research designs 45 3.2 Project plan for (fictitious) project on language maintenance 56 3.3 (Real) Project plan 58 4.1 Data matrix for 5 respondents 77 4.2 Age-group coding 77 4.3 Age-group and gender coding 78 4.4 Attitudes towards language teaching 78 4.5 Fictional data set for questionnaire 83 4.6 Filtered data set for questionnaire 84 5.1 Consonant cluster use 94 5.2 Frequency table 101 5.3 Inverted frequency table 103 5.4 Frequency table with four classes 104 5.5 Cluster simplification: absolute and relative frequencies 105 5.6 Relative frequencies across three socioeconomic groups in two cities 109 5.7 Fixation time in seconds 112 5.8 Fictive set of pre- and post-test scores 114
- 9. figures and tables x 6.1 (ing) distribution across five social classes 122 6.2 Fictive example of /g/ drop 123 6.3 Social distribution of (ing). Romaine (2000) and fictive values 132 6.4 Exam scores for 20 groups 137 6.5 Mean and standard deviation for five fictive samples 144 7.1 Absolute frequencies of men/women and native/non-native speakers (N=100) 152 7.2 Joint probabilities 153 7.3 Absolute frequencies for third-person singular agreement 155 7.4 Production and comprehension of IN and OUT words 156 7.5 Calculation of expected values 157 7.6 Expected values 157 7.7 Calculating df 159 7.8 A fictive example 162 7.9 Fictive pairs of scores for two variables: Lengths of practice (in years) and score (on a scale from 0 to 100) 174 7.10 Simple regression, full information 181 7.11 Multiple regression result 183 7.12 Test-retest, same pattern 185 7.13 Test-retest, no pattern 185 7.14 Language vitality index per language group, based on the mean value of four language dimensions (in %) 187 8.1 Fictive example 193 8.2 EG/CG post-test comparison 195 8.3 Result of F-statistic 196 8.4 T-test between EG and CG. Excel output 198 8.5 Pre- and post-test results 201 8.6 Dependent t-test for experimental group 202 8.7 Performance of 12 speakers 204 8.8 Absolute frequencies of simplified consonant clusters 206 8.9 Observed and expected values 207 8.10 Absolute frequencies of IN and OUT words 208 8.11 Observed and expected frequencies for two-word types 209 8.12 Observed and expected frequencies for two domains 209 8.13 Two treatment and one control group 211 8.14 Interaction/task type 214
- 10. figures and tables xi 8.15 Excel-ready format of Doughty and Pica’s (1986) data 215 8.16 Excel ANOVA results 216 9.1 Practice in years and proficiency scores for 12 speakers 222 9.2 Ranks for proficiency and practice scores 223 9.3 Simple and complex verb phrases in percentage 227 9.4 Relative frequencies, difference (d) and absolute difference d(abs) 228 9.5 Ranked differences 229 9.6 Ranked scores for Sadighi and Zare (2006) 231 10.1 Summary of studies used for AoO meta-analysis 244 10.2 Multiply sample size N with effect size r 245 10.3 Calculate confidence intervalls – step 3 246 10.4 Meta-analysis – fictive example 248
- 11. ACKNOWLEDGEMENTS As with any other book, countless people were involved in the process of this one coming into existence, to all of whom I am eternally grateful. Many thanks to, in no particular order, my former and current students and colleagues at Anglia Ruskin University, for their input, patience, and for serving as guinea pigs in my attempts to bring quantitative methods closer to them; the readers and reviewers of the first edition for their kind and useful feedback, and suggestions for improvements for this second edition; and Gurdeep Mattu and the entire team at Bloomsbury for bringing this book to life. Any faults and shortcomings are, of course, my very own. Sebastian M. Rasinger Cambridge, March 2013
- 12. CHAPTER ONE Introduction Some things never change: the sun rises in the morning. The sun sets in the evening. A lot of linguists don’t like statistics. I have spent a decade teaching English language and linguistics at various universities, and, invariably, there seems to be a widespread aversion to all things numerical. The most common complaint is that statistics is ‘too difficult’. The fact is, it isn’t. Or at least, it doesn’t need to be. This book is several things at once. It is, to no surprise, an introduction to quantitative research methods. As such, it is a primer: a book that is primarily aimed at those with a minimum of prior knowledge. In fact, most readers should be able to follow this book without any prior knowledge or familiarity with the contents. It is also a guidebook – one that leads its readers through the sometimes daunting process of setting up and carrying out a quantitative study from start to finish. And finally, it is a manual, as it provides its readers with hands-on practical advice on quantitative research in general and statistical methods in particular. The original idea for the first edition of this book, published in 2008, followed an argument with some of my MA students. Having discussed the factors influencing second-language acquisition success or failure, we had started to look at the issue of learner motivation, and I came to ask (a) how we could empirically prove how motivated a learner was and (b) how we could draw sound conclusions that motivation and outcome are causally related. The result was grim. Several students, all of them language teachers with several years experience, argued that ‘any good teacher is able to see that this is the case’. And, finally, one of them accused me of ‘trying to measure everything – you just have a too scientific mind’. It took some not- so-gentle encouragement to persuade them that any academically sound piece work would inevitably have to use methodologically proper methods and techniques, or else. . . . Five years down the line, I am still having these discussions, but at least I have a book to refer them to. Quantitative research need not to be difficult!
- 13. QUANTITATIVE RESEARCH IN LINGUISTICS 2 The subject benchmarks for linguistics undergraduate degrees by the UK Quality Assurance Agency – the organization monitoring academic standards in higher education – state that: 6.13 On graduating with an honours degree in linguistics, students will typically be able to: demonstrate an understanding of the issues involved in the basic l l techniques of data analysis, and evaluate and choose appropriate techniques such as distributional criteria, spectrographic analysis, the use of IT tools for the investigation of electronic databases, the use of computer packages for the analysis of acoustic phenomena, the use of laboratory techniques for the investigation of articulatory phenomena, relevant statistical tests, the use of video and audio material in the analysis of spoken interaction . . . l l demonstrate understanding of data and analyses presented by means of graphs (including tree diagrams), tables, matrices and other diagrams and present data appropriately by these means with minimum supervision. (The Quality Assurance Agency for Higher Education. 2007. Linguistics. Subject benchmarks. www.qaa.ac.uk/Publications/ InformationAndGuidance/Documents/Linguistics07.pdf, pp. 14–15. My emphasis.) It might be rather odd to start an introductory textbook with quotes from a policy document; yet, it shows just how important a thorough understanding of research methods is. Over the years, I have marked a number of student assignments at various universities and at various levels, which started off as excellent pieces of work, but once it came to the data analysis part it went downhill: all too often, the author just did not know how to properly handle the data they collected, analyses are flawed and basic mathematical operations simply wrong. Even worse, some continued to make the argument they had started, whether the empirical findings support it or not. Hardly anything is worse than getting (or giving!) a low mark for something that could have been solved so easily – I do speak from experience. There is another side to this, too. Some of us may not want to conduct a quantitative study ourselves, for whatever reason; however, this does not mean that we do not need to understand the basics of quantitative research. The majority of studies in psycholinguistics and sociolinguistics – particularly those dealing with linguistics variation – are based on quantitative data and analysis; and if we want to fully understand these studies, we need to have at least some idea what all those terms, indices, figures and graphs mean. It may be unnecessary to know all the mathematical intricacies surrounding
- 14. Introduction 3 a t-test or Analysis of Variance (ANOVA) (see Chapter Eight), but it is for obvious reasons no good if we read in a paper that ‘a t-test showed statistically significant differences between the 2 groups’ and we do not know what either a t-test or statistical significance is. There is, undoubtedly, a myriad of books on quantitative methods and statistics on the market. Predominantly, these inhabit the bookshelves of disciplines such as social and natural sciences, psychology or economics. Indeed, there are statistics books which cater specifically for linguists, too, for example Butler (1985) or Woods et al. (1986), or more recently the texts by Johnson (2008) or Gries (2009), to name but a few. All of them comprehensive pieces of work which leave little to add to. Yet, experience in teaching quantitative methods and statistics to students and professional linguists alike has shown that what many people read is not only an introduction to statistics (the number crunching bit) but also the general approach that is quantitative research: frequently it is not only the maths that is the problem, but the entire methodological setup. And hence, this book is not one on statistics alone, but an introduction to quantitative methods in general, specifically written for students and researchers in languages and linguistics. As such, this book is written for three particular types of people, which, in some form or other, all feature in the life of its author: Colleagues and students who get panic attacks when asked to l l perform simple statistical analyses such as calculating averages and hence simply and safely refuse to approach anything vaguely related to numbers. The linguist who, despite being well-versed in statistical methods, l l insists on doing complicated calculations by hand with pen and paper and refuses to use readily available software, hence making statistics more painful. The undergraduate or postgraduate student or any other researcher l l in linguistics and language related disciplines who realizes at some point that they need a comprehensible introduction to quantitative methods in their field, in order to make their project methodologically sound. This book has no prerequisites other than very basic numeracy skills and some basic knowledge of using a computer and standard software: being able to do basic mathematic operations such as addition, subtraction, multiplication and division, by hand or with the use of a calculator is all that is needed. And, of course, being a linguist helps! This book does not intend, nor claim, to make its reader love statistics. However, ideally, having read the book, any reader should be able to set up a simple but well-developed quantitative study, collect good data using appropriate methods, perform
- 15. QUANTITATIVE RESEARCH IN LINGUISTICS 4 the most important statistic analyses, and interpret and present the results adequately; in other words, it should stop you from falling into the ‘stats trap’. Alternatively, you may want to use the book or individual chapters to understand particular concepts and tools of quantitative research when reading studies using these tools. This book does not aim at elaborating or explaining the mathematical intricacies of the methods discussed. Rather, it has been written as a practical hands-on guide and should enable its readers to get going and actually do quantitative research from the very first pages onwards. Throughout the book, we try and look at examples or real linguistic research, and will discuss the methods and tools employed in research such as Labov’s Language in the Inner City (1977), Trudgill’s work on sociolinguistic variation in Norwich (1974), or Johnson and Newport’s studies on the ‘Critical Period’ in second-language acquisition (1989, 1991), all of which is aimed to help readers understanding the concepts introduced. This has the advantage that readers can read the original studies simultaneously with the relevant chapter, and increase understanding of the issues discussed. Reviews of the previous edition have criticized that the examples focus somewhat heavily on sociolinguistic and applied linguistic examples – something I entirely agree with. This, in part, reflects my own research interests, but, more importantly, a lot of the examples are comparatively easy to understand and often part of a teaching curriculum. Trying to explain statistics while simultaneously explaining complicated linguistic concepts is beyond the scope of this book. Furthermore, with computers being omnipresent, we will introduce the appropriate steps for performing a statistical analysis using spreadsheet software, specifically Microsoft Excel. Again, this is to avoid complicated manual calculations and will give readers the opportunity to actually do statistics as quickly as possible. While this is not a textbook on using Excel, readers will need no more than basic knowledge of how to enter data into a spreadsheet software – the title of Harvey’s (2003) Excel 2003 for Dummies might scratch people’s egos a bit but it is a good starting point for those who have not previously used the programme. The task of writing a second edition of an introductory textbook is not an easy one: on the one hand, it ought to be different, but on the other, it ought to retain its scope and depth, too. I have hence both expanded and gently updated the previous edition. The book is divided into three parts: Part One (Chapters Two to Four) introduces readers to the basics of quantitative research, by discussing fundamental concepts and common methods. Chapter Two introduces the key concepts of quantitative research and takes a close look at different types of variables and measurement as well as reliability and validity. The third chapter introduces readers to various different research designs and provides information on how to choose an appropriate design for a particular research questions. I have expanded this chapter to include a brief section on project management: knowing how to
- 16. Introduction 5 set up a project that is methodologically sound on paper is not enough – one has to actually do it. Chapter Four is dedicated to an in-depth discussion of questionnaires, their design and use in quantitative research. Part Two (Chapters Five to Nine) covers statistical tools most commonly used in linguistic research. Specifically, we will discuss how to describe data properly (Chapters Five and Six), how to test relationships between variables using standard statistical tools (Chapter Seven), and how to check whether different sets of data are significantly different from each other, for example in pre- and post-test situations (Chapter Eight). Following popular demand, Chapter Eight now includes a discussion of ANOVAs, too. Chapter Nine provides an overview of statistical tools for the analysis of non-normal data. Part Two has been written with the motto in mind, ‘get as much as possible out of your data with as little effort as possible’, and readers should be able to perform simple but thorough analyses of any data set, using Microsoft Excel or similar software. The second part contains many tasks and exercises readers are encouraged to do. Many of these tasks are discussed in detail on the spot to ensure understanding – nothing is worse than having a generic answer somewhere on the back pages but no one really knows where the solution comes from, let alone how to interpret it! Solutions for some additional exercises can be found in Chapter Eleven, together with the statistical tables necessary for the interpretation of results. The final part, consisting of Chapter Ten, is new: it provides an overview of more advanced methods. The first section discusses Multivariate Analysis of Variances (MANOVAs), followed by a discussion of statistical meta- analysis – a method to combine (not to say, recycle) previous results in a meaningful manner. The final section introduces two pieces on statistical software: IBM SPSS (formerly ‘Statistical Package for the Social Sciences’) and R, with pointers to as where readers can find more information. In Part Two, I have included ‘Quick Fix’ boxes which summarize the main statistical tools discussed in the chapter. You can also find cheat sheets for these tools on the companion website, together with Excel templates and short video clips demonstrating how to conduct certain statistical tools in Excel. All parts now also include a ‘further reading’ section at the end. Before you start reading the book, a word of advice from my experience as both a student and a teacher of quantitative methods: do try and actually apply what you read as soon as possible. I have written the book in such a way that it should be – hopefully – relatively easy to simultaneously read and practically follow the issues discussed. So get pen and paper, or your computer, ready. As with many things in life, you will improve with practice – so start as early as you can.
- 18. CHAPTER TWO Quantitative research – some basic issues Keywords: causality–concept–dependentvariable–hypothesis–independent variable – latent variable – law – level of measurement – measurement – qualitative research – quantitative research – reliability – theory – type of variable – validity This chapter discusses some of the basics of quantitative research – the main focus of this book. After an outline of the differences between qualitative and quantitative research, we will look in some detail at the concepts of measurement and different types of variables, followed by a discussion of causal relationships. Lastly, the concepts of validity and reliability are introduced. 2.1. Qualitative and quantitative data While most people have some idea that there is a difference between quantitative and qualitative research, the general concept they have often seems fundamentally flawed. Having taught research methods courses at several universities and at several levels, the most common answer to my question ‘what’s the difference between qualitative and quantitative data’ has usually been somewhere along the lines of ‘quantitative data means much data, and qualitative data is good data’. In the worst-case scenario, this may lead to statements such as ‘I want to do qualitative research because I want it to be really good’ or, alternatively, ‘I can’t be bothered to collect all that data’. It hence seems a good idea to begin with a brief outline of the differences between qualitative and quantitative data, providing specific examples
- 19. QUANTITATIVE RESEARCH IN LINGUISTICS 10 of applications for both types, and examples for data sets that can be interpreted both quantitatively and qualitatively. The idea that quantitative data refers to large amounts of data is only partially true. In fact, the vast majority of quantitative data analysis tools which will be discussed in this book require a data set of a decent size in order to work properly. Small amounts of data can often lead to insignificant, inconclusive or flawed results because the mathematical procedures involved in quantitative (or statistical) analyses require a certain amount of data in order to work properly – we discuss this in more detail from Chapter Five onwards. Nevertheless, the main characteristics of quantitative data is that it consists of information that is, in some way or other, quantifiably. In other words, we can put quantitative data into numbers, figures and graphs, and process it using statistical (i.e. a particular type of mathematical) procedures. When using quantitative analyses, we are usually interested in how much or how many there is/are of whatever we are interested in. A typical quantitative variable (i.e. a variable that can be put into numbers) in linguistic research is the occurrence of a particular phonological or syntactic feature in a person’s speech. Assume we are interested whether speakers of group A are more likely to drop the /h/ than speakers of group B. When analysing our data, we would hence count all the instances in which /h/ is produced by the speakers (such as in hotel or house), as well as all the instances in which /h/ is omitted (‘ouse, ‘otel). What we will get at the very end is four numbers: /h/-used in group A, /h/-drop in group A, /h/-used in group B, and /h/-drop in group B – a nice and basic set of quantitative data, which could look like the one in Table 2.1. In fact, counting occurrences of particular features (or, technically correct, the outcome of a particular variable) is the simplest form of quantitative analysis – yes, quantitative analysis can be as easy as that! Other popular quantitative variables are people’s age (again, a number), people’s weight (probably less important for linguistics, but quantifiable and hence quantitative), and even people’s sex (how many men/women are in a particular group?). We have a closer look at the different types of quantitative data further down in this chapter. Qualitative data, on the other hand, deals with the questions of how something is, as opposed to how much/many. When using qualitative data, TABLE 2.1 Basic set of quantitative data Group /h/ produced /h/-drop A 5 18 B 36 1
- 20. Quantitative research – some basic issues 11 we are talking about texts, patterns and qualities. Qualitative analysis in linguistics is most commonly found in discourse analytic research. By definition, the two approaches of analysis also differ in another respect: qualitative research is inductive, that is, theory is derived from the research results. Qualitative analysis is often used in preliminary studies in order to evaluate the research area. For example, we may want to conduct focus groups and interviews and analyse them qualitatively in order to build a picture of what is going on. We can then use these findings to explore these issues on a larger scale, for example by using a survey, and conduct a quantitative analysis. However, it has to be emphasized that qualitative research is not merely an auxiliary tool to data analysis, but a valuable approach in itself! Quantitative research, on the other hand, is deductive: based on already known theory we develop hypotheses, which we then try to prove (or disprove) in the course of our empirical investigation. Accordingly, the decision between qualitative and quantitative methodology will have a dramatic impact on how we go about our research. Figures 2.1 and 2.2 outline the deductive and inductive processes graphically. At the beginning of the quantitative-deductive process stands the hypothesis (or theory). As outlined below, a hypothesis is a statement about a particular aspect of reality, such as ‘the lower the socio-economic class, the more non-standard features a speaker’s language shows’. The hypothesis is Hypothesis true Result Hypothesis wrong Hypothesis/ theory Methodology Generate Analyse Develop Deduct Data FIGURE 2.1 Quantitative-deductive approach.
- 21. QUANTITATIVE RESEARCH IN LINGUISTICS 12 based on findings of previous research, and the aim of our study is to prove or disprove it. Based on a precisely formulated hypothesis, or research question, we develop a methodology, that is, a set of instruments which will allow us to measure reality in such a way that the results allow us to prove the hypothesis right or wrong. This also includes the development of adequate analytic tools, which we will use to analyse our data once we have collected it. Throughout this book, I will keep reminding readers that it is paramount that hypothesis/research question and methodological- analytical framework must be developed together and form a cohesive and coherent unit. In blunt terms, our data collection methods must enable us to collect data which actually fits our research question, as do our data analysis tools. This will become clearer in the course of this and subsequent chapters. The development of the methodological-analytical framework can be a time-consuming process; especially in complex studies we need to spend considerable time and effort developing a set of reliable and valid (see below) tools. The development of a thorough and well-founded methodology is followed by the actual data-collection process, that is, the generating of data which we will base our study on. Ideally, the data-collection stage is preceded by a pilot stage, where we test our tools for reliability and validity; in Chapter Four, we discuss piloting with regard to questionnaire-based surveys. Depending on the topic, the data-collection stage may be the most time and resource intensive of the entire process, particularly when data (or people we can get data from) is not readily available. This emphasizes the need for thorough planning in advance: imagine you are to conduct a study involving a tribe in the Amazonian rainforest. The last thing you would want is to stand in the middle of the jungle and realize that your data-collection tools are not working! In Chapter Three, I provide a very Hypothesis/ theory Patterns, structures Methodology Generate Analyse Induce Data FIGURE 2.2 Qualitative-inductive approach.
- 22. Quantitative research – some basic issues 13 basic introduction to project management, which will give you an idea of how to balance project scope, time and resources in order to achieve the intended outcome. Once collected, our data is then analysed using the analytic methods developed previously, ultimately providing us with our results. In quantitative research, our results will be a set (or a whole bunch) of numbers and numerical indices which describe in great detail what is going on in our data. Now comes the crucial part: based on our own results we look back at our hypothesis and compare how well or badly our results fit the hypothesis; then, we deduce whether the hypothesis is right or wrong. In our simple example of socio-economic class and the use of non- standard language, we may find in our data that lower social classes show a high number of non-standard language use, middle classes show slightly less non-standard use, and upper social classes show no use of non-standard language at all. Hence, our results consist of three general observations, from which we deduce that there is a relationship between social class and language use in such a way that the lower the class, the more non-standard features occur; our original hypothesis is proven correct. However, in real life, we would usually have many more factors involved, and the relationships between them are often not as obvious as we would like them to be. In this case, the deduction process is one of carefully deriving our conclusions based on and carefully balancing the available evidence. Frequently, we may not be able to conclusively say whether a hypothesis is right or wrong, simply because circumstances are so complex. A side note: most of us (and I do not exclude me) would like to prove whatever hypotheses we come up with correct, and frequently people perceive it as a failure if they prove their own hypothesis to be wrong. Yet, given that our original hypothesis made sense, that our methodological- analytical framework was good and thorough and that our data was of good quality (see reliability and validity discussion at the end of this chapter), there is nothing inherently bad about proving a hypothesis wrong. It might ruffle our ego a bit, but even if we show that something is not the case, we still contribute to the accumulation of knowledge. So, we may find that in our particular study, using our particular respondents, there is no relationship between social class and language use; if our methodological framework is reliable and valid, this is still a permissible result, even though other researchers before us came to different conclusions: it is simply that in this particular context it is not the case. Much worse, however, is a situation where the results we produce simply do not make sense and appear at random. If we take the United Kingdom, it is generally true that social class and use of Standard English are linked to a certain extent, as proposed in our hypothesis (exceptions just confirm the rule).However,wemightfindthatourresultsinvertthehypothesis,andhigher social classes use more non-standard language while lower socio-economic classes use more standard language. While we cannot immediately discard this as ‘wrong’, we should have a very thorough look at our methodology,
- 23. QUANTITATIVE RESEARCH IN LINGUISTICS 14 our analysis and our results and try to account for this pattern. As before, it might be that our result is just a phenomenon of the particular context we looked at; however, it could also be that something has gone horribly wrong somewhere along the lines, either in the way we collected or analysed our data. Statistics using modern software can be comparatively easy – a few mouse-clicks and we get the required results – but the number on the computer screen is really just the result of certain mathematical operations. The final interpretation is still down to us, and as a general guideline, if we feed out software rubbish, we will get rubbish out. The qualitative-inductive process, on the other hand, is significantly different. We shall just sketch it here. At the beginning is not a definite hypothesis, but an ‘idea’. As before, we develop a methodology that allows us to generate and analyse data. However, whereas the results in a quantitative study inevitably revolve around the hypothesis, the qualitative analysis reveals patterns or structures which are present in our data. And it is from these patterns that we induce a hypothesis or theory. Quantitative research is probably the approach most commonly found in traditional sociolinguistic, psychological and psycholinguistic and considerable parts of social research in general. Discourse analysis, and especially Critical Discourse Analysis (CDA) in the Faircloughian tradition (Fairclough 1988, 1992, 1995) is usually qualitative in nature, although a recent development towards using methods derived from corpus linguistics to support CDA has given it quantitative aspects, too (see Hardt-Mautner 1995 or Baker 2006 for an overview). Let’s look at a piece of data to illustrate the difference between quantitative and qualitative analysis. The following example is taken from Fairclough’s ‘Language and Power’ (Fairclough 1988: 18) and represents an interview between a police officer P and a witness W. 1 P: Did you get a look at the one in the car? 2 W: I saw his face, yeah. 3 P: What sort of age was he? 4 W: About 45. He was wearing a . . . 5 P: And how tall? 6 W: Six foot one. 7 P: Six foot one. Hair? 8 W: Dark and curly. Is this going to take long? I’ve got to collect the kids from school. 9 P: Not much longer, no. What about his clothes? 10 W: He was a bit scruffy-looking, blue trousers, black . . . 11 P: Jeans? 12 W: Yeah.
- 24. Quantitative research – some basic issues 15 It does not take a trained linguist to realize that the dialogue is characterized by a power imbalance, with P being clearly in control of the interaction. A quantitative analyst would try to support this hypothesis with numbers. We could, for example, look at the average turn length, the numbers of questions asked and answers given by each participant, or we could look at interruptions made by each participant. Our result would look similar to Table 2.2. From Table 2.2, we could conclude that P indeed seems to control the interaction, as he (or she?) asks the majority of questions, while giving fewer responses, and is the only one who interrupts. We could also employ various statistical methods to strengthen this argument. All in all, we have a rather good indication that P controls the situation to a certain extent. More conclusive in this case, however, is looking at what is going on in the conversation in terms of content and patterns – that is, a qualitative analysis (for a full qualitative analysis, see Fairclough 1988). P’s questions seem to follow a particular predefined sequence, trying to elicit key information. W’s elaboration on clothing in line 4 is interrupted by P, but the issue taken up later by P in line 9. W’s attempt to leave the interview in 8 is, although acknowledged, but more or less ignored (line 9) – indicating rather strongly that this is not a chat between two friends. In summary, our brief (and extremely superficial) analysis has shown how quantitative and qualitative approaches differ. Both can be applied to the same set of data, but in this case the analysis of how is much more useful than the analysis of how much. The choice of whether to use a qualitative or a quantitative approach is inseparably linked to the kind of research question we ask. Research questions and methods used to answer it must be developed together in one coherent framework – I will keep stressing this throughout the book. We have said above that while qualitative research looks at how something is, quantitative research is more interested in how much/many there is/are of the issue we are interested in. For any successful research project, we need to specify from the outset what we would like to TABLE 2.2 Quantitative analysis of police interview Police officer Witness Number of turns 6 6 Average turn length in words 5.5 7 Number of questions asked 6 1 Number of responses given to questions 2 6 Numbers of interruptions made 2 0
- 25. QUANTITATIVE RESEARCH IN LINGUISTICS 16 find out and how we are going to find this out. In particular new researchers such as undergraduate students often come up with the idea that ‘I want to look at X and I will use Y’ – even though Y is an inappropriate tool to lead to any reasonable conclusions about X. This is often linked to a strong fear (or hatred) of anything related to maths, leading to a fierce rejection of anything statistical and hence quantitative – even if the only reasonable way of answering the question is a quantitative one. So let’s have a look at some situations where a quantitative analysis is inevitable. Any study whose main argument is based on the mere counting of l l things. This may sound daft, but it does constitute the most basic form of a quantitative analysis. We may, for example, be interested in the following: ‘How many transitive verbs are there in “Beowulf”?’ m m ‘How many students have taken English Language A-levels this m m year?’ ‘How often does a politician contradict himself within one m m speech?’ None of these require any kind of sophisticated mathematical procedures; nevertheless, they are all quantitative in nature, as they are based on working with numbers (of words, students, contradictions). If we are choosing a research question that is essentially based on numbers, we need to work with numbers, and this has not to be difficult. Any study that aims at l l proving that two or more groups of people (or objects) are distinctively different. Typical research questions could include: ‘Men use more swear words than women.’ m m ‘Lower social class members use more non-standard forms than m m members of higher social classes.’ ‘This year’s cohort has obtained much better exam marks than m m last year’s cohort.’ All of these cases are, in essence, based on simple counting, followed by a comparison: in the first case, we would count the number of swear words used by men (within the period of an hour, for example), and compare the result to the number of swear words used by women. The second example works identically: we count and eventually compare the use of non-standard forms between 2 groups. The third example is, again, based on the comparison of 2 values: exam results from this year’s cohort with that of last year’s.
- 26. Quantitative research – some basic issues 17 In Chapters Five and Six we discuss in more detail how to describe these differences, while Chapter Eight explains how these comparisons can be made in a statistically sound way. Any study that aims at showing that two variables are related, l l that is, co-occur in a particular pattern, often in the form of ‘the higher/lower X, the higher/lower Y’. These studies are based on the assumption that a change in one variable results in a change of another variable: ‘The older a learner at the start of language acquisition, the m m slower the progress.’ ‘The lower the social class, the more non-standard forms occur.’ m m ‘The number of loan words has steadily increased after 1300.’ m m In very basic terms, these questions are again based on m m comparisons of different numerical values, but unlike the above examples, we look at more than two groups and/or more than two points in time. There are, of course, many more situations where we will have to use a quantitative methodology; however, to a certain extent they are only modifications of the above situations. So, in what situations can (and should) we not use a quantitative but a qualitative method? Remember from above that quantitative research is deductive: we base our question and hypotheses on already existing theories, while qualitative research is deductive and is aimed at developing theories. Qualitative research might come with the advantage of not requiring any maths, but the advantage is only a superficial one: not only do we not have a profound theoretical basis, but we also very often do not know what we get – for inexperienced researchers this can turn into a nightmare. Prime examples of qualitative research in languages are studies in ethnography or anthropological linguistics (see, for example Duranti 1997, or Foley 1997, for introductions to the topic). Anthropology in general is interested in, broadly speaking, human behavioural patterns, and anthropological linguistics’ focus is on communicative behaviour in general. Baquedano-Lopez (2006) analysed literacy practices across learning contexts, looking how literacy develops and emerges in different contexts, such as in educational settings, outside school and so on. Baquedano- Lopez’s work is qualitative as the focus of her work was, essentially, on how literacy is, as opposed to how much literacy there is (i.e. the extent or proficiency – clearly a quantitative question), as well as on patterns and structures, as opposed to, for example, proportion of literacy in a given population (which again would be quantitative). Qualitative analyses can provide astonishing insights, but require the analyst to deeply engage with
- 27. QUANTITATIVE RESEARCH IN LINGUISTICS 18 the data in order to develop some kind of theoretical framework for it – a mere description of what is happening is not a proper qualitative analysis. And so, in qualitative research, we often find ourselves confronted with large amounts of data which we need to make sense of, by ordering it and developing categories into which we can put different bits of our data. For those less versed, this can be difficult, especially since qualitative data can often be interpreted in different ways with no ‘right’ or ‘wrong’ answer. Quantitative methods, on the other hand, give us at least a mathematically sound result (if applied correctly) which often acts as a starting point for our interpretation: 2 times 2 is always 4, and ‘all’ we need is to interpret the result within our theoretical framework. Yet, in recent years, we have seen a proliferation of research that uses mixed-methods designs, that is, use both quantitative and qualitative methods to investigate a topic; this is particularly common when dealing with topics that are so complex that a categorical separation into qualitative/ quantitative would not work. Bergman (2012), Cresswell and Plano Clark (2010) and Dörnyei (2007), inter alia, all provide good overviews on how to conduct mixed-methods studies. 2.2. Variables and measurement In the centre of quantitative research is the concept of a variable. As a general definition, a variable is a measurable feature of a particular case: a case is one particular ‘thing’ or unit that shows this particular measurable feature. In linguistics, a case may be a person, or a group of people, or a linguistic unit. As the term ‘variable’ implies, the values (or outcomes) of this feature can vary considerably between cases. Each case can only show one value for a particular feature. For example, the variable ‘gender’ when applied to human beings is generally considered to have two potential values or outcomes, ‘male’ and ‘female’, and a person can only be one or the other, that is, either male or female but not both at the same time – exceptions apply. The variable ‘age’ can, in theory, vary from zero years of age to indefinite (assuming that there is no naturally predefined age by which humans drop dead), and again, a single individual can only show one outcome: you obviously only have one age at a time. In a group of, for example, ten people, we may get ten different values for the variable ‘age’, namely when none of ten people in this group are of the same age. Linguistic variables work with the same principle: in a study on syntactic variation in spoken language we may want to look at the presence or absence of copular verbs in subject-predicate constructions, such as ‘The girl is pretty’ or ‘He Ø fat’. For convenience’s sake, we may want to define our variable ‘copular presence’ in such a way that it can only have two outcomes: copula present ({+Cop}) and copula absent ({−Cop}). In our examples, the first sentence,
- 28. Quantitative research – some basic issues 19 ‘The girl is pretty’ will take the value {+Cop}, while the second takes {−Cop}. In a fictional corpus of let’s say 1,000 subject-predicate constructions, we may find 750 {+Cop} and 250 {−Cop} – but as with gender, other values are not permissible (‘a bit of a copula’ does not make much sense, just as someone cannot be ‘a bit dead’). In a different project, we might simply be interested in the number of noun phrases (NP) per sentence in a corpus of 1,000 sentences. Here, the variable ‘number of NPs’ can – theoretically – take indefinitely many values: ‘ l l The girl is pretty’ contains 1 NP (‘the girl’). ‘ l l The cat chases the mouse’ contains 2 NPs (‘the cat’ and ‘the mouse’). ‘ l l The cat chases the mouse around the garden with the big trees and the green lawn’ has 5 NPs – ‘the cat’, ‘the mouse’, ‘the garden’, ‘the big trees’ and ‘the green lawn’. Closely related to the concept of variable is that of measurement. Measurement is the process of assigning a particular variable outcome to a particular case, using predefined criteria. The term and concept is well known: when measuring the size an object, we hold a ruler or a tape measure next to the object; depending on the units we work with (centimetres, metres, inches, feet), we can then assign the corresponding value to our object. If we are interested in its weight, we put our object on a scale and assign the corresponding value (in ounces, pounds or grams) to it. Measurement, however, can also be much simpler, as it generally means to put a particular object or person (or ‘case’) in a predefined category. So, strictly speaking, we can also measure someone’s gender: we have two clearly defined categories, male and female, and we put our case into either category, depending on how well our case fits the category. Crucially, the categories we assign to our objects must be well defined and must not change during the process of measurement: we can only make reliable statements about an object’s length in inches if, first, we have clearly defined the amount of length an inch refers to, and, second, one inch must always refer to exactly the same amount of length and does not vary during the measurement. In addition, our tools for measuring an object’s properties should be designed in such a way that it can measure all our cases adequately. For example, with human beings, we usually get away with two categories for measuring biological sex (it gets more tricky when we look at the more culturally defined category of ‘gender’): male and female, and, in the vast majority of cases, we should be able to put a human being in one of the two categories. However, if we wanted to use our two category measure for certain invertebras, such as earthworms, we would run into serious problems: earthworms are hermaphrodites, that
- 29. QUANTITATIVE RESEARCH IN LINGUISTICS 20 is, they are both male and female at the same time, and when mating, can perform both ‘functions’ simultaneously. We have said above that for any one case we can have only one value for a particular variable, that is, we cannot put a single earthworm in both the male and female category; at the same time, putting it into the male (or female) group is also an inaccurate reflection of reality. Hence, we will have to adjust our measuring tool appropriately – in the earthworm example, we would probably add a ‘both’ category. 2.3. Definitions, concepts and operationalization Some readers may argue that I pay undue interest to something straightforward. After all, most of us are able to measure someone’s age, gender or weight. However, there are cases where the process of allocating a variable value to a case is not that easy. A common problem, particularly for researchers new to a field, is that it is not always clear what is being measured, let alone how. A typical example for this is the concept of ‘motivation’. Over the years, I have met countless students, both undergraduate and postgraduate, who were all interested in how motivation influences second-language learners’ progress. The problem usually starts when I ask the question, ‘so how are you going to measure it?’ We all know that motivation plays a crucial part in second-language acquisition/learning; however, in order to come to a sound conclusion along the lines of ‘the higher learners’ motivation, the better the outcome’ we need to measure the different levels of motivation. But before we can even measure it, we have to clearly define what motivation actually is. A review of literature on the topic shows that motivation is a complex system of several interrelated factors, of which Stern (1983) identifies social attitudes, values and indeed other ‘learner factors’. More recently, motivation is often linked to the ‘learner’s self’ (see Dörnyei and Ushioda 2009 for an overview). As such, motivation is a concept rather than a variable; in fact, motivation consists of several variables, each of which has to be measured separately. In other words, we cannot reliably measure motivation as such, let alone by methods such as asking learners ‘how motivated are you’. When constructing concepts, one has to ensure that they are epistemologically sound – put more bluntly, they have to make sense. For example, a concept can only be described by definitions and terms that are already known and measurable. In the above example, we need several already known and measurable learner factors in order to define our concept ‘motivation’. A different example is the concept of ‘cohesion’: ‘Cohesion can be defined as a set of resources for constructing relations in discourse’ (Martin 2001: 35). The clue is in the definition: as it is a ‘set of resources’, it cannot
- 30. Quantitative research – some basic issues 21 be measured directly, but we can approach it by looking at the different components that make a text cohesive. Once we have defined the concept we want to investigate, we need to establish a set of operation definitions; we have to operationalize our concept. In other words, we need to decide how we measure the variable or variables describing our concept. To come back to our example, we would develop a set of questions (or variables) which refer to the different aspects of motivation, such as attitudes to the language, the learning process, the environment and so forth. We shall look at a real example to illustrate the idea of concepts and their measurability. In the late 1070s, Giles et al. (1977) developed the concept of ‘Ethnolinguistic Vitality’ in order to provide a framework that could account for what ‘makes a group likely to behave as a distinctive and active collective entity in intergroup relations’ (308). It does not take too long to understand that ‘group behaviour’ is not something that can be straightforwardly measured, such as a person’s age or weight, but is a construct that comprises various aspects and issues. Giles et al. identified three categories of variables, each to be measured independently and, more or less, objectively: a The group’s social status, based on objective and objectively measurable variables such as income (how much money do individual group members earn), economic activity (unemployment rates), employment patterns (which jobs do they have), but also including more difficult approaches to status, such as perception of the group or the group’s language (e.g. whether the group’s language is an official language, such as Catalan in Spain). b Demographic factors, taking into account the absolute and relative size of the group as well as its density (i.e. how many group members live in a clearly defined geographic space). Again, these variables are comparatively easy to measure. c Institutional support, measuring to what extent the group receives formal and informal support through official and/or community- intern institutions (government, schools, churches, local authorities etc). While this last category is a rather abstract concept in itself (for details, see original work), it ultimately provides us with a quantifiable result. Giles et al. argue that the higher a group fares in each of the categories, the higher its Ethnolinguistic Vitality is. So, the results (or scores) obtained in the individual categories allow us to fairly objectively quantify a group’s vitality. In other words, we can put a comparatively reliable number to describe a rather abstract concept. It is important that, for every quantitative study, we spend some time thinking thoroughly about our variables, how to measure them, and how to
- 31. QUANTITATIVE RESEARCH IN LINGUISTICS 22 operationalize this measurement. And as such, it is paramount that we have a good grasp as to what our variables could potentially look like, before we develop a method to capture as much of the variation in this variable as possible. Phonology is a good linguistic example to illustrate this. Traditionally, sociolinguists and dialectologists have focused on variation in pronunciation, and have tried to explain this variation along various dimensions such as social categories or geographic areas. Sebba (1993: 23) in his study on ‘London Jamaican’ discusses some of the phonological differences between Jamaican Creole and London English, and identifies for one particular speaker in his sample two distinct realizations for the pronunciation of the /u/ sound in ‘who’ and the initial consonant in ‘the’: [hu:] versus [hʊ:], and [di] versus [đə]. If, like Sebba, we consider this phonological difference to be important, we must use a measure that accounts for this. In other words, we need an operation definition that allows us – in practice – to clearly classify a particular vowel as [u:]/[ʊ:] and a particular consonant as [d]/[đ]. At the same time, if such differences are not of interest for our study, we do not need to account for them. The trick is to develop a methodology which allows us to measure as accurately as required but does not make things unnecessarily complex. A word of warning: some concepts, particularly abstract ones and those that comprise several variables and sub-concepts, are notoriously difficult to measure. Motivation, as outlined above, is one such concept, as are attitudes. A quantitative analysis of these takes time, effort and a lot of trial and error. Possibly more than for any other study, the thorough evaluation of previous work and in particular the methodologies used (including all the difficulties and pitfalls!) is vital. It is usually a good although resource- and time-consuming idea to support quantitative data on these issues with qualitative findings from interviews or focus groups. Often, the combination of qualitative and quantitative results provide a far better and reliable insight than one approach on its own. 2.4. Independent and dependent variables Following our discussion of variables in general, we now turn to the relationship variables can have to each other. More specifically, we focus on independent and dependent variables. The crucial difference between independent and dependent variables is that the latter can be influenced by the former, but not vice versa. A common independent variable in linguistic research is age. Research on second-language acquisition has shown that age influences – to some extent – acquisition success, with older learners being more likely to achieve lower proficiency in the second language than younger ones. In this case, language acquisition success depends on age, and is hence the dependent
- 32. Quantitative research – some basic issues 23 variable. At the same time, no one would argue that second-language proficiency influences a person’s age – you do not become older just because your language proficiency increases! In experimental setups, the independent variable is often deliberately manipulated by the researcher, while the dependent variable is the observed outcome of the experiment. For example, if we were interested in measuring whether word recognition decreases with increasing speed in which the word is displayed to a participant, the independent variable would be speed (as it can be manipulated by the researcher), while the dependent variable would be the outcome (i.e. the accuracy with which the words are recognized). While in many research questions the independent and dependent variables are clear from the outset, there might be constellations where it is more difficult to define which one is which. For example, sociolinguistic research has consistently shown that social class and the use of non-standard features correlate; often, non-standard features occur more frequently in the speech of lower socio-economic classes. It is arguable that social class is the independent variable, as it influences the way people talk. However, one may consider that attitudes against the use of particular non-standard features disadvantage non-standard speakers and prevent them from certain jobs and, ultimately, socio-economic success. At the very end, it is the chick or the egg question and needs careful consideration in the interpretation of our results. 2.5. Causality and latent variables The discussion of independent and dependent variables inevitably raises the issue of causality. As we have seen in the previous example, it is not always entirely clear which variable causes which outcome. Causality means that there is a causal relationship between A and B, that is, changes in A cause changes in B or vice versa. As we will see later on, several statistical methods can be used to measure the relationship between two variables; however, they do not tell us anything about whether the two variables are causally linked. Let’s go back to our age of acquisition onset example. I suggested above that age of acquisition onset influences eventual second- language proficiency; and a lot of empirical research has shown that the older learners are when they start learning a second language, the lower their eventual proficiency scores are. From here, it is a short step to arguing that age and proficiency are in a causal relationship to each other; in fact, many people argue exactly this way. However, strictly speaking, many studies only show one thing: values for the variable ‘age’ and values for the variable ‘proficiency’ co-occur in a particular pattern, that is, the higher the ‘age’ values, the lower the ‘proficiency’ values are. In order to be able
- 33. QUANTITATIVE RESEARCH IN LINGUISTICS 24 to speak of a proper causal relationship, our variables must show three characteristics: a They must correlate with each other, that is, their values must co-occur in a particular pattern: for example, the older a speaker, the more dialect features you find in their speech (see Chapter Seven). b There must be a temporal relationship between the two variables X and Y, that is, Y must occur after X. In our word-recognition example in Section 2.4, this would mean that for speed to have a causal effect on performance, speed must be increased first, and drop in participants’ performance occurs afterwards. If performance decreases before we increase the speed, it is highly unlikely that there is causality between the two variables. The two phenomena just co-occur by sheer coincidence. c The relationship between X and Y must not disappear when controlled for a third variable. While most people usually consider the first two points when analysing causality between two variables, less experienced researchers (such as undergraduate students) frequently make the mistake to ignore the effect third (and often latent) variables may have on the relationship between X and Y, and take any given outcome for granted. That this can result in serious problems for linguistic research becomes obvious when considering the very nature of language and its users. Any first year linguistics student learns that there are about a dozen sociolinguistic factors alone which influence they way we use language, among them age, gender, social, educational and regional background and so on. And that before we even start thinking about cognitive and psychological aspects. If we investigate the relationship between two variables, how can we be certain that there is not a third (or fourth of fifth) variable influencing whatever we are measuring? In the worst case, latent variables will affect our measure in such a way that it threatens its validity – see below. For our example, we will stay with the age of acquisition onset debate. It has generally been argued that the age at which people acquire a second language influences the acquisition success to a certain extent; many studies have shown that the older a learner at the start of acquisition, the less their progress will be. We do not have to be experts in the field to spot a considerable problem with this hypothesis: as most people are probably aware, second-language acquisition progress is subject to a whole range of factors, among them motivation or amount and quality of input they receive. Hence, arguing that age alone influences second-language development is highly problematic. Any study trying to establish such a link must hence pay considerable attention to other factors and eliminate them, either through methodological modifications or through fancy statistical methods.
- 34. Quantitative research – some basic issues 25 2.6. Levels of measurement When we, once again, consider some of the most frequent linguistic variables, it becomes clear that they differ with respect to the values they can take. As discussed, for humans, the variable sex, for example, can only take one of two values: male or female. Age, on the other hand, can take any value from zero onwards, as discussed above. The scores of standardized language proficiency tests, such as the International English Language Testing System (IELTS), usually put examinees into a particular class or category, which is described by a particular number: an IELTS score of 1 indicates that the examinee ‘essentially has no ability to use the language beyond possibly a few isolated words’, while a score of 9 means that he ‘has fully operational command of the language: appropriate, accurate and fluent with complete understanding’ (www.ielts.org 2013). These differences, which are explained in more detail in the following sections, refer to the level of measurement of a variable, and have a serious impact on the kind of statistical methods we can use, and it is critically that we, from the outset of our research, take the different levels a variable can take into consideration. Categorical scale data Generally, we distinguish between four different levels of measurement. At the bottom end are categorical variables. Categorical variables can have two or more distinct outcomes, that is, they allow us to put our case into one particular class or category only. When we look at the variable ‘gender’, the variable can only take one of two potential values, ‘male’ or ‘female’. Interim answers are impermissible – one cannot be ‘nought point seven male’. Similarly, the questions of whether a woman is pregnant or not can only elicit one out of two values: yes or no, as obviously you cannot be ‘a bit pregnant’. A typical Equal Opportunities Questionnaire asks people to state their ethnic background, and usually provides a list of categories of which respondents have to pick one, for example, White British, White Irish, Other White, Afro-Caribbean, South-Asian, Mixed Race and so on. Again, this allows us to put respondents in one particularly category, and again, interim answers cannot be accounted for (‘70% White British with 30% Afro-Carribbean’ – even though this is genetically possible). Most importantly, the outcomes of categorical variables do not reflect any kind of hierarchy: ‘male’ does not mean ‘better’ or ‘more’ than ‘female’, just as ‘pregnant’ and ‘non-pregnant’ cannot be brought into a meaningful order. We are merely labelling or categorizing our cases.
- 35. QUANTITATIVE RESEARCH IN LINGUISTICS 26 Ordinal scale data Similar to categorical data, ordinal variables allow us to put particular labels on each case. However, in addition, ordinal data can be put into some kind of order or ranking system. A popular example for illustrating ordinal data is the dullness of university lectures. Students may classify lectures as ‘exciting’, ‘ok’, ‘dull’ or ‘very dull’. It is obvious that there is an inherent semantic order to these labels, with dullness increasing from ‘exciting’ to ‘very dull’; in other words, lectures can be ranked according to dullness. However, we cannot make any statement about the differences between individual labels: ‘very dull’ does not mean ‘twice as dull as dull’, but is simply an indicator that one is worse than the other. Interval scale data One step up from ordinal data are variables on interval scales. Again, they allow us to label cases and to put them into a meaningful sequence. However, the differences between individual values are fixed. A typical example are grading systems that evaluate work from A to D, with A being the best and D being the worst mark. The differences between A and B are the same as the difference between B and C and between C and D. In the British university grading system, B, C and D grades (Upper Second, Lower Second and Third Class) all cover a stretch of 10 percentage points and would therefore be interval; however, with First Class grades stretching from 70 to 100 per cent and Fail from 30 downwards, this order is somewhat sabotaged. In their purest form, and in order to avoid problems in statistical analysis, all categories in an interval scale must have the same distance from each other. Interval data is commonly used in social and psychological (and hence sociolinguistic and psycholinguistic) research in the form of Likert scales, which are discussed in detail in Chapter Four. Ratio scale data In addition to indicating equal distances between two adjacent scores, as in interval scales, ratio scale data is characterized by having a natural zero point, with zero indicating that there is no measurable amount of whatever we are measuring. It is also possible for a ratio scale to be open-ended. For example, a weight of zero indicates that an object does not weigh anything; the amount of weight is hence zero. At the same time, an object can be indefinitely heavy. To illustrate the difference between interval and ratio data in more language-related terms, the IELTS scores described above are usually
- 36. Quantitative research – some basic issues 27 interpreted as having equal distances, indicating a gradual increase of learners’ proficiency; as such, we would interpret it as interval data. An IELTS score of 1 means that the examinee is a ‘non user’. A non-user may or may not produce ‘possibly a few isolated words’. A score of zero indicated the examinee has not attempted the test. It does not, however, indicate that the examinee has zero knowledge. We could, however, think of a simple vocabulary recognition test using a ratio scale, where learners are presented with a list of 100 lexical items and are asked how many items they know. At the end, we calculate the score by simply counting the number of recognized items. A score of zero indicated that the participant has failed to recognize a single vocabulary item, that is, has ‘zero’ vocabulary knowledge. A participant scoring 60 has recognized twice as many items as one with a score of 30. 2.7. Continuous and discrete data The final crucial distinction we can make is to look at whether our data is discrete or continuous. Discrete data consists of finite or countable values, while continuous data allows for infinitely many possible values on a continuous scale without gaps or interruptions. For example, the number is students in a class is a discrete variable, as the number must be an integer (‘whole number’), that is, 1, 2, 5, 62 and so on. We cannot have 6.5 or 12.7 students in a class. If we consider our students’ age, on the other hand, we look at continuous data, as a student might be 21.85482 years old – if we can be bothered to break it down to exactly how many years, days and hours he is old. Please note that the discrete-continuous distinction only applies to our raw, unprocessed data, not to the results any kind of statistical analysis may bring us: if we are interested in the average seminar class-size at a university, it is highly likely that on average we have something like 15.7 students – it does not mean that there is a class with 0.7 student in it. We discuss the calculation of the arithmetic mean in Chapter Six. 2.8. Reliability and validity Quantitative methods bear an advantage that should not be underestimated: natural sciences, such as chemistry or physics, require ‘proper’ experiments (i.e. certain research procedures or methods) to be objective and, crucially, have to be replicable by anyone who is following the instructions. In other words, anyone following this ‘recipe’ exactly, should get the same (or very near same) result. This is very much like baking a cake: if two people put exactly the same amount of ingredients together, mix them in exactly the
- 37. QUANTITATIVE RESEARCH IN LINGUISTICS 28 same way and then bake them under exactly the same conditions (i.e. time and temperature), we should get two identical (or very near identical) cakes. Even better, we should get two identical cakes independent from who the baker is: it should not matter whether one is more experienced, or in a foul mood, or tired, or quite simply hates cake. This works well in basic chemistry: if we put exactly the same amounts of A and B together, at exactly the same temperature and other environmental factors, we should reliably get C – over and over again. This is where the beauty of quantitative research lies: once we have established our methods, we probably want our methods to accurately measure the same thing again and again. Put briefly, reliability refers to a method repeatedly and consistently measuring whatever it is supposed to measure. In the most extreme case, if we took the same people and tested them again in exactly the same way and exactly the same environment, we should get exactly the same result if our method is reliable. In linguistic research, this is difficult to achieve: people change, the environment changes, and even if people are identical, they are unlikely to respond to the same test in exactly the same way again – be it because they have the experience of having done the test before or because they quite simply have a bad day. Nevertheless, we need to ensure that our methods work reliably, and that they remain reliable over time. At the same time, methods of measurements must be constructed in such a way that they work independently from the researcher using them: in whichever way we operationalize a variable, we need to make sure that all potential researchers using our operationalization will get same or similar data and results. That is why it is so important to define variables clearly and with as little ambiguity as possible: if we are interested in use the use of copular verbs like in the example above, we must define our variable, that is, what constitutes a copula and what does not, in such a way that any researcher carrying our the analysis knows exactly what they are looking for. Copulas are a good example: some people consider copular and auxiliary verbs as one and the same thing: English ‘be’ accordingly is the same in both sentences; after all, in both examples the copula has little to no independent meaning: a The girl is pretty. b I am working. In some cases, however, it might be useful to distinguish between the two types. Crystal (1999), for example, defines copulas’ primary function ‘to link elements of clause structure, typically the subject and the complement’ (73), while auxiliaries are ‘subordinate to the chief lexical verb in a verb phrase, helping to express such grammatical distinctions as tense, mood and aspect’ (30). According to these definitions, be in (a) is a copula, linking subject (‘the girl’) with the complement (‘pretty’), while in (b) it helps to indicate tense and aspect (present progressive). Whether we use a
- 38. Quantitative research – some basic issues 29 broad (no difference) or narrow (difference) approach does not matter per se; the important point is that, first, our definition is the one that best fits our research questions and, second, all researchers involved in our project use the same definition. Just imagine what would happen if two people analysing the same data using different definition – their results would be incomparable and simply useless. The ability of our methods to measure the same issue reliably independent from who is the researcher is sometimes also called inter-observer reliability. In some context, where accuracy is crucial but despite proper operationalization it can be difficult to measure things exactly, it is common to use two people to code the data, either independent from each other, or one spot-checking the coding of the other. But even if we are the sole researcher, it is important that we use the same definition every time throughout the same study. A quick and easy, even though by far not the most reliable way to check a method’s reliability (note the paradox) is the split-half method: we take a sample of, for example, 100 people (for sampling, see Chapter Three), and randomly split it into two groups of 50 people each. We then administer our method to both groups. If our method is reliable, we should get equal or near equal results for both groups. We discuss the practical issues of data being ‘equal’ in Chapters Eight and Nine. Another way to check reliability is the test-retest method: instead of testing two samples with the same method, we test the same sample with the same instrument at two different points in time. If our instrument is reliable, we should get very similar results for both test and retest. The test–retest method is problematic whenever we work with human beings: people learn from experience and tend to draw on their experience from the first test during the retest. In other words, they know what is coming, which can substantially impair the retest’s significance. When can obtain a good indicator for a measurement’s reliability using a correlation analysis based on the test-retest method – we will discuss this at the end of Chapter Seven. A second factor we have to consider in quantitative research is that of validity. Whenever we measure something we obviously want our result to be as accurate as possible. Validity comes in several different forms and shapes; here, we will focus on the one most important one for linguistic research: measurement validity. Measurement validity, often just called validity, is difficult to explain, let alone to achieve: it refers to the issue of whether our method actually measures what it is supposed to measure, allowing us to draw appropriate conclusions. The copula definition example above is also an issue of validity: if we have not clearly defined our variable, how can we make sure what we measure is what we actually want? Another typical example to illustrate validity is the design of questionnaires. Let’s assume we are interested in people’s attitudes towards the use of non-standard language or dialects. We are going to measure attitudes by means of a questionnaire comprising ten questions,
- 39. QUANTITATIVE RESEARCH IN LINGUISTICS 30 and respondents have the choice of three answers: ‘agree’, ‘neutral’ and ‘disagree’. Now imagine that we ask all our questions in a particular way or pattern, for example, ‘agree’ always indicates a positive opinion on dialects, while ‘disagree’ always indicates a negative opinion. Annoyingly, many people have the habit of answering questionnaires with a particular manner and tendency: they tend to either more agree or more disagree, or, in the worst case, have a ‘neutral’ opinion. If, as in our example, ‘agree’ always corresponds to a positive attitude towards dialects, all our questionnaire measures is respondents’ tendencies to either agree or disagree, but gives us certainly no idea about what respondents’ actual attitudes towards dialects are! In this case, our questionnaire is a highly invalid method. We will return to the issue of reliability and validity when we have a closer look at questionnaire design in Chapter Four. Validity, or lack therefore, can become are rather annoying issue in a situation where we have many third and latent variables, as discussed above. If our data is not actually influenced by what we think influences it, but by something else, how can it we validly measure it? The best option is to conduct a pure experiment (see Chapter Three), where we as researchers have full and total control over all variables. Certain psycholinguistic questions are well suited for an experimental inquiry, and experiments are frequently used. It does not take long, though, to realize that laboratory experiments are not an option for studies require linguistic data as it occurs in real life. In these cases, only thorough planning and preparation, based on extensive research of the theoretical background involved, can help us to minimize this problem. 2.9. Hypotheses, laws and theories Before we move on to the next chapter, we briefly need to define three core concepts in any kind of research, be it qualitative or quantitative. Hypotheses are statements about the potential and/or suggested relationship between at least two variables, such as ‘the older a learner, the less swear words they use’ (two variables) or ‘age and gender influence language use’ (three variables). Hypotheses can be proven right or wrong – the point of quantitative research is to do exactly that! In fact, a hypothesis must be proven right or wrong. For a hypothesis to be proven correct or incorrect, it is important for it to be well defined. In particular, hypotheses must be falsifiable and not be tautological: the hypothesis ‘age can either influence a person’s language use or not’ is tautological – independent from our findings, it will always be true (age either does or does not influence a person’s language use). A good hypothesis, however, must have the potential of being wrong. For quantitative research, it is useful to remember that our hypothesis should relate to something we can actually measure. As with all things, it is worth spending some time to think about the phrasing of our hypothesis
- 40. Quantitative research – some basic issues 31 (or hypotheses) – it will make things much easier once we get to the stage where we develop our methodology. As a general guideline, the more specific a hypothesis is, the better. Note that ‘specific’ does not imply the depth or scope of the study: for obvious reasons, a 3,000 word undergraduate research essay will be different in terms of depth and scope from a project that comes equipped with £500,000 of external funding. Yet, for both it is essential to be clear about what we want to find out. Based on the general focus of my department, many of my students are interested in issues related to language teaching and learning, and frequently, proposed topics for dissertations revolve around two aspects: learners ‘liking/hating’ something, and a particular teaching/learning method being ‘better’ or ‘worse’ than another. In the first stages, I usually come across hypothesis such as: Students like to learn in groups. l l Using electronic resources for language teaching is better than using l l traditional textbooks. There is nothing inherently ‘wrong’ with these two examples, and both outline what could be more or less interesting research projects. The problem lies in the detail, and, in our specific case, the measurability. Let’s have a closer look at the first example: while is gives us a general idea of what the researcher is after, the ‘hypothesis’ is far too vague and, as such, it seems impossible to find a reliable and valid measure to prove/disprove it. There is, for a start, the term ‘students’ – who exactly are they? What levels of study? What age? What cultural background? Everyone vaguely familiar with learning and teaching theory will know that these are three of a whole list of issues that are important; ‘students’ as a generic term is simple not precise enough, and a study, especially a small-scale one, is bound to fail if the group we are interested in is not defined properly. ‘Native speakers of English primary school children learning Japanese’, however, clearly defines the group and sets the scene for all methodological implications that come with it. Then, there is the word that regularly sends shivers down my back: ‘like’. We all like or hate things, but, unfortunately, ‘like’ is a rather difficult feature to measure. What exactly do we mean when someone ‘likes’ something? Do they have a preference in using one thing above the other? How strong is this preference? Where does ‘like’ end and ‘love’ (another favourite) end? A better phrasing is to work is ‘prefer’ or ‘show a preference for’ – quite simply because we can objectively measure preference. Thirdly, there is the generic ‘learn’ – learn what? Again, we need to specify this much more. And lastly, ‘in groups’ is equally vague: indications of group size is the least we should include into our hypothesis. Despite what it may sound (or read) like, I am not being nitpicking: how can we reliably and validly measure what we have not even clearly defined
- 41. QUANTITATIVE RESEARCH IN LINGUISTICS 32 beforehand? A project looking at ‘how students learn’ is predestined to go wrong – simply because it is too vague. For our example, a much better hypothesis could be: English native speaker primary school children show a preference for working in groups of 3–5 when acquiring new vocabulary items compared to working in traditional full-class environment. This hypothesis is clearly defined, it is measurable and, crucially, falsifiable: the kids may or may not prefer working in small groups, and we have quantitative evidence to prove it. Talking about falsifiability, most of us like our beautiful and carefully designed hypothesis to be proven right – it is an ego thing. However, it may be that your empirical evidence shows that your hypothesis is wrong. This might be disappointing, but is not inherently a bad thing: assuming that our hypothesis is a reasonable one, that is, emerged from what we know from previous studies, and assuming that our methodology is valid and reliable, proving a hypothesis wrong is still a noteworthy result. After all, we still contribute to knowledge by showing that a particular view on a topic does not work. And while we generally phrase hypotheses in a positive way (X does cause Y), if we are really too upset that our data has proven us wrong, we may want to change our entire argument in such a way that our hypothesis is that X does not cause Y – in other words, we would structure our study around an argument of exclusion. Laws are hypotheses or a set of hypotheses which have been proven correct repeatedly. We may think of the hypothesis: ‘Simple declarative sentences in English have subject-verb-object word order’. If we analyse a sufficient number of simple English declarative sentences, we will find that this hypothesis proves correct repeatedly, hence making it a law. Remember, however, based on the principle of falsifiability of hypotheses, if we are looking at a very large amount of data, we should find at least a few instances where the hypothesis is wrong – the exception to the rule. And here it becomes slightly tricky: since our hypothesis about declarative sentences is a definition, we cannot really prove it wrong. The definition says that declarative sentences must have subject-verb-object (SVO) word order. This implies that a sentence that does not conform to SVO is not a declarative; we cannot prove the hypothesis wrong, as if it is wrong we do not have a declarative sentence. In this sense it is almost tautological. Hence, we have to be very careful when including prescriptive definitions into our hypotheses. A much better example for laws are sound changes: if we start with the hypothesis that during the ‘Great Vowel Shift’ during the Early Modern English period all long vowels moved upwards into a higher position (e.g. /e:/ became /i:/, /o/ became /u:/), we will see that, when analysing a large enough number of examples, the hypothesis is repeatedly true. Yet, we will
- 42. Quantitative research – some basic issues 33 also see that there are exceptions, that is, situations where the vowels did not move their position – mainly in combinations with particular consonants (see Baugh and Cable 2002, for an overview). So, we have a law on sound changes, but as with every rule, there are exceptions. Lastly, theories are systems which combine several hypotheses and laws, all of which are in a logical relationship to each other. As such, we may want to consider the Great Vowel Shift as a theory, as it can only be fully explained as a complex interplay between various laws: most vowels moved upwards (law 1) but some did not (law 2) or moved into a lower position (law 3), based on the preceding/following consonant (laws 4 and 5) and so on. Exercise I: Variables 1 Identify whether the following outcomes are discrete or continuous, and explain your decision: a Total time a London cab driver spends stopped at traffic lights in any given day b Number of strikes in the car industry each year c Number of left-handed people in the population d Quantity of milk produced by a cow each day 2 Determine and explain which levels of measurement are most appropriate for the following data: a Ratings of excellent, very good, good, poor, unsatisfactory for students work b A person’s sex c British National Insurance Numbers d Percentage of all university students studying linguistics Exercise II: Hypothesis and theory building Design a set of hypotheses which have the following properties: at least one of your hypotheses should include two or more l l independent variables, your hypotheses should include variables at least two different levels l l of measurement, at least one hypothesis should contain a clear causal relationship, l l your hypotheses should be in a logical relationship to each other, l l that is, have the potential of forming a theory.
- 44. CHAPTER THREE Research design and sampling Key words: artefacts – ethics – longitudinal, cross-sectional, panel designs – population and sampling – real time and apparent time – sampling techniques – research design – representativeness – sample error – project planning and management Having discussed the very basics of quantitative research in the previous chapter, we now have to address the question of how we are going to get from our hypothesis to reliable and valid results. In this chapter, we will focus in some detail on research design and sampling. In Part One of this chapter, we will introduce different potential research designs, such as cross-sectional, longitudinal, experimental and quasi-experimental, and discuss advantages, disadvantages and pitfalls of each method. Part Two discusses various sampling techniques, explaining advantages and disadvantages of sampling methods. In particular, we address typical sampling problems encountered in linguistic research (e.g. lack of standardization of participants, etc.). This chapter also addresses ethical implications of linguistic research, with regard to privacy and anonymity, working with minors and minorities, and researchers’ safety. New in this edition is an overview of issues to do with managing your project and how to keep a check on time, resources and scope. Research design is best described as the actual structure according to which our study is organized. As such, together with the theoretical grounding, the design forms an important part of the overall methodological-analytical framework which we use to answer our research questions, and to prove or disprove our hypotheses. Research design does not, though, refer to the actual instruments we use in our investigation (such as questionnaires or interviews), although the relevance of good interplay between hypotheses, existing theory, methods and design cannot be emphasized enough; and there is a particularly strong connection between design and instruments chosen.
- 45. QUANTITATIVE RESEARCH IN LINGUISTICS 36 Research designs can be subsumed under two main categories: longitudinal, cross-sectional and panel designs structure our research in terms of a temporal order, while experimental and quasi-experimental designsallowustoinvestigateourquestionthroughtheexplicitanddeliberate manipulation of variables. We will discuss each type subsequently. 3.1. Cross-sectional designs Cross-sectional designs are probably the most frequently used in linguistic, social and psychological research. In cross-sectional designs, a comparatively large amount of data is acquired at one given point in time, providing an overview of how a particular variable (or variables) is distributed across a sample (see Section 3.6) at a particular moment in time. As such, cross-sectional research is useful for providing a snapshot of a status quo – it describes reality how it is right at this very moment. Cross-sectional research is characterized by two key features: first, we are dealing with a multitude of cases, that is, a number of cases (in linguistics, often people) that is significantly larger than one; the large amount of data will allow us to describe and explain the particular feature of reality we are interested in (i.e. our research question) in greater detail and accuracy. From Chapter Five onwards, we will see that many statistical tools provide us with increasingly accurate results with increasing amount of data, hence, a study based on 50 people will give us a more ‘true’ picture than a study based on only 5. Second, cross-sectional data is collected within a particularly short time frame – ideally, all our data is collected simultaneously, hence ruling out any changes which might occur over time. Needless to say that in practice this can be rather problematic. Labov’s study on the ‘social stratification of /r/ in New York City department stores’ (Labov 1972) is an excellent example of a cross-sectional study in linguistics. Using rapid anonymous interviews, Labov was able to collect data from 264 respondents in just 6.5 hours over one or two days. His results enabled him to make conclusions about the prestige of /r/ across different social classes at a particular point in time. The advantage should be obvious: collecting all data almost simultaneously, any changes in linguistic behaviour are unlikely to occur, simply because linguistic change and changes in the sociocultural environment surrounding it are much slower processes. Cross-sectional designs are commonly used in quantitative research, as the comparatively large amounts of data collected are particularly useful for studies which aim to either describe a status quo to an extent to which some generalizations are possible, or to detect relationships between independent and dependent variables while simultaneously ruling out that
- 46. Research design and sampling 37 those relationships change over time. In the Labovian example above, linguistic behaviour was strongly linked to socio-economic class and the prestige different classes assign to a particular feature. If, for whatever reason, prestige changes while we are conducting our data collection, we will inevitably receive skewed results. I am frequently asked by students about the time frame in which cross- sectional studies should take place; in other words, when does a cross- sectional design stop being a cross-sectional design? Unfortunately, a definite answer (‘two days and six hours’) does not exist. Yet, with a bit of common sense it should be relatively easy to determine a time frame. First, in linguistics, cross-sectional, unlike experimental studies, are usually used to investigate the variables linked to either the linguistic features or the language user (including exogenous sociocultural variables in sociolinguistics). I assume most readers agree that linguistic change is a comparatively slow process, with changes unlikely to happen overnight or even within a few days or months. Hence, a study investigating a particular linguistic feature is unlikely to be negatively affected by change, even if data collection takes place over a longer period of time. That is, although data is collected at various points within a particular time frame, it is unlikely that the results are affected by change. Milroy’s (1987) study on language use and social networks in Belfast was primarily a cross-sectional design, despite being conducted over a prolonged period of time: because the social networks she looked at were very stable, with very stable norms governing language use within those networks, language change did not occur, even though her data was not collected within a short period of time. It is important to remember that language is hardly ever looked at without due consideration of its user. As social beings, humans move around an environment that can, quite often, undergo rather rapid changes. In other words, the context in which language is used may change halfway through a cross-sectional study, which can cause a serious impact on the results. In this case we talk about a quasi-experimental design (see below). So, in short, while there is no definite time frame for a cross-sectional study, we should carefully monitor the environment in which we collect our data for sudden changes. Similarly, for research questions that involve, by their very nature, processes of change, cross-sectional designs are not suitable. For example, first language acquisition is a process in which children acquire are substantial linguistic competence within a very short time span, mainly between the ages of 18 months and 3 years. With change being that inherent to the subject matter, it is highly unlikely that we get reliable results with a cross-sectional design. Also, for topics where change in the defining criteria, it is more than likely that it is change and development we are interested in – again, this is something we cannot account for with a cross-sectional design, which can only provide us with a snapshot.
- 47. QUANTITATIVE RESEARCH IN LINGUISTICS 38 3.2. Longitudinal designs: Panels and cohorts Longitudinal designs are distinguishable from cross-sectional designs in two major respects: instead of looking at a large number of cases once, longitudinal studies look at a small(er) number of cases on several (at least two) occasions. In short, while cross-sectional studies allow us to draw conclusions about a status quo at a particular moment in time, longitudinal studies allow us to observe changes that occur over time. Also, longitudinal designs may, to a certain extent, shed more light on causal relationships than cross-sectional studies, because certain patterns may become more obvious when measured repeatedly. Please note that for longitudinal designs we still need a considerable amount of data for a proper quantitative analysis, for mathematical–statistical reasons. But unlike cross-sectional designs, where a large amount of data is created at one point in time by a large number of ‘cases’, in longitudinal designs, fewer cases produce data repeatedly. Since they are based on the repeated collection and analysis of data, longitudinal designs are rather resource intensive, and social sciences and psychological research often try to avoid them (see Bryman 2004). Yet, for linguistic studies, longitudinal designs offer the opportunity to observe changes in linguistic behaviour over time in great detail. In fact, longitudinal designs are the only way to observe linguistic change. As with cross-sectional designs, there is no definite answer as to what makes a longitudinal design longitudinal; as before, it depends on the research question we are trying to answer. We may, for example, be interested in the development of young children’s vocabulary size. As a general guideline, most children utter their first word at around the age of 12 months, have an average vocabulary size of 500 words at 18 months and by their second birthday, the average child has about 500 words at their disposal. Accordingly, if we base our longitudinal study on two data collections, one at the age or 12, the second at the age of 24 months, we will be able to see a dramatic increase in vocabulary size. However, this also means that we are losing a lot of detail, as we are unable to see what exactly happens within these 12 months; we cannot detect any major intermediate stages. Yet, some linguistic changes are much slower: the Great Vowel Shift, for example, during which all long vowels in English moved their position, is commonly agreed to have taken place somewhere between 1400 and 1700 (see, for example Baugh and Cable 2002) – that is, a massive 300- year span! As described below, Woods’ work on sound changes in New Zealand seems to occur from generation to generation – still much slower than a child’s vocabulary acquisition, but quicker than the Great Vowel Shift. Even though I am repeating myself, but this again emphasizes the importance of thorough research and a good familiarity with the theories
- 48. Research design and sampling 39 that are involved in our study, and to construct our methodological framework around them. Longitudinal studies come in two basic forms, differentiated only by their sample: panel and cohort designs. For panel designs, the sample (or panel) is drawn randomly from the population (see Section 3.6). For example, let’s assume we are interested in the progression of Spanish learners of English at a particular university (our population). From this population we randomly select a small number of students, maybe ten, and we monitor their progress at three different occasions through a standardized language proficiency test as well as reflective interviews: at the beginning of January, the beginning of April, and the beginning of July. Data collected at these three occasions will allow us to draw conclusions about (a) the progression of individual students over time, and (b) general patterns of progression for the whole sample. Cohort design work essentially similarly, but the sample we work with is slightly different. Cohorts are groups of people who share a particular characteristic which is often of temporal nature, for example year of birth, year of marriage, or time spent learning a particular language. In our above example, rather than looking at all Spanish learners of English at our university, we are only looking at students who started in the 2011/12 academic year. Cohort designs are arguably better for controlling third variables. According to Bryman (2004: 46), the crucial difference between panel and cohort designs is that while panel designs allow us to observe both cohort and aging effects, cohort studies can only identify aging effects. Whether this is an advantage or a disadvantage depends on the individual study. An example: we are interested whether a particular phonological variable changes over time. A panel study will allow us to analyse our data on two dimensions: first, we can draw conclusions as to what extent respondents’ progressing age influences their use of this particular variable (e.g. the older respondents get, the less frequently do they use the variable). Second, it also provides us with information about how respondents born in a particular year change their usage of this feature, hence allowing us to include information about generation differences. On the other hand, with all respondents born in the same year, cohort designs only allow us to observe change according to aging, as the year of birth is constant – generational differences, for example, cannot be observed. One of the main advantages of longitudinal designs, namely the ability to measure change over time, is also one of its major pitfalls. With the entire study depending on a particular group of people over a long period of time, the research is easily compromised once members of the sample start dropping out, due to lack of motivation, relocation or, in the most extreme case, death. In other words, we may start with a sample of 20 people in 2013 for a 5-year study, but by the time we arrive in 2017 only 7 respondents remain. This may substantially impair the significance of
- 49. QUANTITATIVE RESEARCH IN LINGUISTICS 40 any results we may get, and, in the worst case, endanger the entire study. Particularly in contexts where people frequently change due to the nature of the context (at British universities, for example, a typical undergraduate cohort is only here for three years), sample retention and hence longitudinal studies can be problematic. 3.3. Pseudo-longitudinal designs: Real-time and apparent time While Labov’s study used a cross-sectional design to detect relationships between linguistic and social variables at one particular point in time, cross-sectional studies may, to a certain extent, also be used to simulate longitudinal studies (Figure 3.1). In linguistic terms, we may want to design onesynchronicstudyinsuchawaythatwecaninferdiachronicdevelopment, without the hassle of having to run several repeated synchronic studies over a prolonged period of time (Bayley 2004). In particular in Labovian sociolinguistics, this is also known as real-time and apparent time. The method was first developed by Labov (1963) in his Martha’s Vineyard study. We shall look at a different example here. Woods (2000) in her study on sound changes in New Zealand in essence ‘simulated time’: by looking at the speech of respondents from three generations – grandmother, mother and daughter – she was able to pin down changes in vowel sounds over time, even though the majority of data was collected at the same point in time, that is, synchronically. Here, we have real-time data for the respondents, but based on the sociolinguistic theory that language is age-graded, looking at data from different generations allows us to draw conclusions about diachronic development, or apparent time, and hence language change. In other words, we are not looking at how language is changing, but how it has changed already. Note, though, that this type of design cannot offer the full explanatory power (e.g. regarding causality) that real longitudinal designs offer, as change can only be observed across the sample, but not at the level of an Cross-sectional at time a, b and c Longitudinal Time FIGURE 3.1 Longitudinal versus cross-sectional designs.
- 50. Research design and sampling 41 individual respondent. It also assumes the relative stability of all other factors – see Bayley (2004) for a comprehensive summary. 3.4. Experimental and quasi-experimental designs Experimental designs are fundamentally different from the ones discussed so far: unlike longitudinal and cross-sectional designs, which allow us to collect and observe ‘natural’ data, that is, data as it occurs in its natural environment, experimental studies are based on the systematic and deliberate manipulation of one or more variables by the researcher. In their purest form, experiments are conducted in laboratories – facilities that enable the researcher to have control over all, or almost all, variables involved in a study – both overt and latent ones. Traditionally, experimental designs are quantitative in nature and consist of the comparison between the experimental group (the one affected by the manipulation of variables) and a control group (not affected by manipulation). We shall start with a classic example from outside the field of linguistics: a pharmaceutical company, when developing a new drug against the common flu, at some point has to test the effects of this new drug. In essence, what they will do (with first animals and later humans), is to take a group of people suffering from flu and split them into two equal halves. Then, they administer the drug to the experimental group, and nothing (or a placebo) to the control group and measure the difference afterwards. If the new drug is any good, the experimental group should show some signs of improvement, while the control group does not. Statistical methods, which we will discuss in Chapters Six to Ten, help to determine whether the drug really makes a significant difference. It is important to note that in experimental designs, the assignment of any respondent into experimental or control group is completely random, that is, any respondent may end up in either group. Also, respondents usually do not know which group they belong to, as to avoid the impact of any psychological factors that this knowledge might create in a participant. In linguistic research, too, experimental designs are based on the deliberate modification of a variable for one group. Li’s (1988) study on the impact of modified input on advanced Chinese learners of English as a second language presented a set of sentences to his sample under two conditions: half of the respondents (experimental group 1, EG1) were given sentences for which meaning could be inferred from context (‘cue- adequate’), the other half (experimental group 2, EG2) received sentences which were cue-inadequate (meaning could not be inferred). That is, the variable ‘input’ is modified in two ways: adequate/inadequate. Figure 3.2 outlined the process graphically. In step 1, as outlined, the allocation into
- 51. QUANTITATIVE RESEARCH IN LINGUISTICS 42 EG1 or EG2 was conducted randomly, and this should ensure that EG1 and EG2 start with the same level of English language proficiency. Step 2 involved the introduction of the two stimuli – cue-adequate sentences for EG1, cue-inadequate sentences for EG2. In step 3, we test both groups for their proficiency after the introduction of the stimulus, followed by step 4, the analysis of the results: Li’s hypothesis at the outset was that cue- adequate input is more beneficial to learners that cue-inadequate input; accordingly, results on step 4 should be better for EG1 than for EG2. Note that in this example, we deal with two groups of people, divided by the two different ‘versions’ of the same variable ‘input’. This is often referred to as between-subject designs – each group receives a particular type of input, similar to the drug test example. As an alternative setup, in a within-subjects design the same group of people receive two types of stimulus, and we compare reactions to both stimuli. For either setting, given that all other factors are constant, for example, sentences should be of equal length and complexity, the comparison between the two sets of input should provide reliable information as to whether contextual cues help learners to understand the meaning of new words, and according to Li, it does. There are several advantages and disadvantages of both between- subject and within-subject designs. Between-subject studies require more participants, in order to create two groups of equal sizes. Also, it must be ensured that the two groups are equal in terms of participant characteristics. A random allocation of participants into groups should avoid any participant-based bias; nevertheless, it is always worth checking: it is all too annoying to realize afterwards that group A had significantly more men than group B, while group B was significantly younger than group A – both facts which would certainly substantially skew the result. Working with one group, that is a within-subject design, avoids this problem. On the other hand, within-subject designs put considerably more effort on participants. This can lead to boredom or fatigue – neither is a desirable feature for the reliability of our results. In addition, with increased length of the task, participant may gain practice and hence improve their results. In the worst case, this may lead to the creation of artefacts: that is, the results Step 1 Step 2 Step 3 Step 4 EG 1 Stimulus 1: cue adequate T E S T EG 1 ≠ EG 2 Stimulus 2: cue inadequate EG 2 FIGURE 3.2 Setup of Li’s study.
- 52. Research design and sampling 43 we obtain are not based on respondents’ reaction to the actual content, but on their reaction to the task. In other words, artefacts are results that are based on how respondents react to a particular methodology, not on respondents’ knowledge or reaction to the change in variables. As such, the phenomenon is closely related to the issue of validity, as discussed in Chapter Two. Had Li used a within-subject design, we may imagine a situation whereby respondents indeed start to suffer from fatigue, leading to their performance decreasing across both stimuli. As a result, data we obtain may not necessarily tell us how respondents’ perform with regard to adequate or inadequate cues, but simply how well or badly they react to the task. Artefacts can occur in all research designs, and the risk of creating one should not be underestimated. Hand in hand with the thorough study of previous research and methodologies used, any researcher designing a study should hence continuously ask and critically evaluate whether the method used really measures what it is intended to measure. Quasi-experimental studies, too, are based on the change of one (or more) variables; however, there are two crucial differences to true experimental designs. First, the assignment of respondents into experimental and control group is not random, but to some extent ‘naturally given’. Second, instead of a deliberate manipulation through the researcher, in quasi-experimental studies an ‘external force’ changes a particular variable. And since this sounds rather abstract, we will look at an example. At the time of writing this book, there is a considerable debate in the United Kingdom about the improvement of sociocultural integration of migrants, and, inevitably, the use of the English language as a means of facilitating integration as well as the provision of English language/English for Speakers of Other Languages (ESOL) classes is a major topic for politicians, media, policymakers and the general public. Imagine, for ease of comprehension, that at this moment in time there was no ESOL provision available for migrants. Now, let’s also assume the government passes a law which, from next week onwards, significantly improves ESOL provision but also forces migrants to gain an English language proficiency equivalent to the Cambridge Advanced English Certificate (CAE). However, in this first instance, provision will only be provided for migrants of Pakistani descent. Hence, Pakistanis in Britain ‘automatically’ become our experimental group, that is, the group that will be subject to a change in a particular variable (namely ESOL provision), while all other migrants will not undergo any changes (i.e. are our control group). We would design our fictional study around two points of data collection: measuring proficiency of both experimental group and control group before the new law comes into effect, and afterwards, let’s say in six months time. Similar to real experimental designs, we would obtain
- 53. QUANTITATIVE RESEARCH IN LINGUISTICS 44 four scores which would allow then us to draw conclusion whether or not the new policy has any effect on migrants’ English language proficiency: Experimental group proficiency pre-stimulus l l Control group proficiency pre-stimulus l l Experimental group proficiency post-stimulus l l Control group proficiency post-stimulus l l The advantage of quasi-experimental designs is that they require relatively little effort from the side of the researcher, as the introduction of the change, or stimulus, is taken care of by other people. Yet, this is also our biggest problem: unlike in real experiments, we have no control over third variables and what is happening, and rely entirely on the observation of potential changes. In addition, since the allocation of participants into the experimental group is not random and participants will be aware of this (just imagine the media coverage!), it is likely to trigger some kind of psychological reaction (positive or negative) in them – a reaction that is not based on the ESOL provision, but the fact that they are made to participate. 3.5. Summary: Research designs The use of a particular research design depends on several factors, but our research question and hypotheses are the most crucial aspects. Table 3.1 summarizes the main characteristics. 3.6. Sampling Inseparably linked to the choice of a research design the question of who we would like to research, that is, who our participants are. On the surface, this may seem obvious and the discussion here obsolete; yet, in order to design a valid and reliable study we must have a look at the main concepts involved: the population, the sample and sampling techniques. The population defines the group of people we are generally interested in; however, it does not necessarily mean all people (as in ‘all people living in Britain’ or even ‘this universe’). More specifically, the population refers to a group of people who share certain characteristics. As such, population can equally refer to large or small groups: if we choose ‘mankind’ as the population we are interested in, we include all human beings living on this planet – currently around 6.6 billion individuals – with ‘human being- ness’ being the common feature. Similarly, we may define our population
- 54. Research design and sampling 45 as ‘all native speakers of the English language’ – compared to mankind a smaller population with just around 380 million individuals. However, populations can also be defined as much smaller groups. Over the last 20 years, sociolinguistics and linguistic anthropology has paid increasing interest in language use in minority communities. In my own research, I defined population as all first generation migrants of Bangladeshi origin who live in a particular London borough – explicitly excluding those second generation people who were already born in London. Yet again, I usually advise my undergraduate students for their final projects to go much smaller and to define their population as something along the lines of ‘all students of a BA(hons) English Language degree at Anglia Ruskin University’ – a population that comprises a few dozen people. The term population does not necessarily apply to people only (or animals if you are a zoologist, or plants if you are a botanist): strictly speaking, a ‘population is defined as a set of elements; an element is defined as the basic TABLE 3.1 Summary of different research designs Type of design Possible application Pros/cons Longitudinal Language change; development of a particular feature over time + real-time observation of change − but external factors difficult to control − time and risk of sample attrition Cross-sectional Description of current status of a particular linguistic feature + comparatively fast + very accurate with the right methods and framework (as no change occurs) − large sample − cannot account for any developments before time of data collection Experimental Influence of a particular external stimulus on a linguistic feature + total control of researcher − often difficult to set up − validity and reliability issues Quasi-experimental Influence of a particular external stimulus on a linguistic feature + observation of real change in comparatively short period of time (cf. longitudinal) − no control over change of variable − little control over external factors
- 55. Random documents with unrelated content Scribd suggests to you:
- 56. the big sticking-up collar gives a round-shouldered effect, and spoils what is one of their best points, a graceful set and carriage of the head and neck. They walk very straight, with all the motion from the hips, and their feet very much turned out, and generally wear no jewellery of any sort, except perhaps a pair of gold earrings, or a ring or two, or a rosary of European patterns. There is nothing characteristic in the way of native work or beads. The well-to-do Filipino women wear more trinkets, and the Mestizas (Eurasians) cover themselves with cheap and tawdry ornaments. The favourite material for the camisa (bodice) is a native muslin woven from the fibres of pine-apple leaves, called piña, an exclusive manufacture of the Islands of Panay and Negros, where the pine- apples grow wild in the jungles. This the Filipino women weave with or without silk stripes and checks, and dye all sorts of colours; but the lower classes and peasants hardly ever wear anything beyond the plain, undyed yellowish-white, which, after all, suits them far better than any other colour. They look well though, on great occasions, in crimson, purple, or yellow, and they are wise when they stick to those warm colours, for blues and greens are fatally unbecoming to their yellow-brown skins, making them look heavy and dirty. They seem to have no natural taste for colour though, as they use some appalling aniline dyes, and make mixtures which set one’s teeth on edge. They are only really safe when they stick to the red sarong and undyed camisa. The piña is woven on hand-looms, which can be seen and heard clicking in almost every hut, and it is sent all over the Islands, and fetches enormous prices, but then it is practically everlasting, and when washed and done up with rice-starch, it looks like new. They also have a muslin, much cheaper stuff, called Jusi (pronounced Hoosee), which is made from a fibre procured in China; and a third, and still cheaper one woven from hemp fibres and called sinamay—and the result of it all is that to the uninitiated the three materials all look exactly alike! On the piña the women do a very beautiful embroidery of graceful designs worked out in fine white
- 57. sewing-cotton and marvellously shaded, mixed with drawn threads, and some of the antique pieces are exquisite. This piña embroidery is the only characteristic Filipino work I have been able to see or hear of, except the decoration of some weapons, and the grass mats with patterns. The dress of the men I think I have already hinted at, and it, too, is the last word in simplicity (short of the loin-cloth, which costume is not allowed in the towns), for all the Filipinos wear in the house is tight drawers and a vest, and when they go out they draw on over those a pair of white or blue cotton trousers and a collarless shirt, rather like a Chinaman’s coat, which I described to you before, I think. This shirt hangs outside the trousers, really looking much better than it sounds, and on galas and occasions of state they turn out in an ordinary European shirt, with a starched front, all pleated and embroidered, such as Frenchmen and Germans sometimes wear, and they look so clean and smart in them. In fact they look quite nice in their native costume, but unfortunately many of them now affect the white man’s buttoned-up linen coat, with stand-up starched collar, and put on shoes and stockings, which subtly vulgarises the wearers at once. Like all coloured races and many white ones, as soon as they attempt modern European fashions the Filipino taste is villainous, and they look inexpressibly common and disheartening. They are so clean—so scrupulously clean—all their clothes, even those of the very poorest, being spotless and fresh. They are for ever washing their bodies, too, or at least it is certain that the poor people are, for they may be seen at the wells and outside their houses tubbing ingenuously, the men with a single fold of stuff retained for decency, the women struggling inside a wet sarong. We went yesterday evening for a walk along the beach, on the side of this spit where the view embraces the open sea and the end of the Island of Guimaras, the latter with a promontory of mountainous Negros jutting out behind and beyond it, and all the rest clear horizon. The tide was out, so we walked on the firm wet
- 58. sand at the edge of the waves, little, flat waves which did not run up very far, as the beach is steep and shelving. Over the mountains, inland, the sky was a deep glowing orange and crimson, but from where we were on the beach we could not see the mountains, only glimpses of the gorgeous colour through the high palms that fringe the shore; while on the other side, out to sea, was a reflection like a delicate wash of pinky gold, set above deep blue sea and purple islands. We walked a good long way, as far as the ends of the streets that come down on the beach, all dark with points of light, for the air was deliciously soft and the breeze almost fresh, and as the sunset faded, the stars came out and made quite a light upon the water, they looked so big and bright. We enjoyed the walk very much, and though we are too far this side of the town to be able to walk as far as the open country, we are very lucky not to be a long way from the beach, where we can always get a breath of fresh air and admire the lovely evenings.
- 59. LETTER VIII. SOCIAL AMUSEMENTS Iloilo, January 8, 1905. This is my first letter to you in the New Year, and it does seem so strange to be writing 1905 already. I wonder how you brought the year in. We were invited to a ball given by the Club Artistica, the Spanish Club, situated in a suite of very large rooms in the upper story of a big house in the Calle Real, the main street of the town, which I told you about when I was describing the amazing shops. The big basements are shops, but the long upper stories form large dwelling houses, very swagger ones, only the dust and noise are very disagreeable, and the rents about the same as flats in the best part of London, if not more. On these two accounts, most of them stand empty, displaying long rows of closed shutters, all the outside painted the prevailing bluey-grey. Some are used as clubs, however, one being this Artistica, and another, further down the street, the Filipino Club, which is called the Santa Cecilia—dedicated very appropriately to the patron saint of music, you see. These two clubs are very hospitable, and do nearly all the entertaining in the place, except for an occasional lecture at the Y.M.C.A., which, I daresay, is a wild revel, only I’ve never summoned up courage to go and see. The Swiss and Germans have a club, I believe, and the English Club has a beautiful house of its own, but neither of these institutions does anything towards the gaiety of nations, beyond playing billiards among their own members exclusively. It is a relief, however, to think that the poor fellows do not have such a very bad time as one might imagine, for they accept everything and go everywhere. The same comforting remark applies to the Americans, who have no club and don’t entertain privately, except tea or Bridge parties amongst each other. So, as I said
- 60. before, it falls to the Spaniards and Filipinos to keep the place alive, and very well they do it too, if the ball on New Year’s Eve was a specimen of their average entertainments. The Spaniards, Eurasians, and natives are all passionately fond of dancing, and really fond of it, for they do not make it a question of supper, as people do at home. All you have to do here is to clear the floor and get in some musicians (half the difficulty here is to keep groups of musicians out), and apparently your friends flow in. When we are coming home in the evenings, we often see the salas of quite little houses lighted up and full of people dancing, and I have seen small native huts having a baile of two couples jostling round in a space 10 feet square. The chief room of the Spanish Club is a large apartment, almost a hall, where, on ordinary evenings, the members can be seen through the big lighted window-spaces, sitting about at little tables, with glasses at their elbows, playing dominoes; but for the baile the club was cleared and hung with electric lights in paper flowers, and decorated with flags and palm branches, while in a large recess at one side was a numerous string band of Filipino performers. The music was excellent, but so slow that, as far as I was concerned, dancing was no pleasure, though that was not much of a grievance to me, as I was really far more anxious to look on than to dance. We were invited for ten o’clock, but when we arrived at eleven the entertainment was only just getting into full swing. We had missed the opening Rigodon, a dance without which no Filipino baile could get under weigh at all, but the second half of the programme began with one, and I was very much interested to see it. Everyone who wanted to dance the Rigodon, and there were only about three people who did not, sat round the room in an immense square, as for a cotillon, and the band struck up a very jolly old Spanish tune, to which the sides facing each other went through a few simple figures at a very slow walk. When they had done, they
- 61. sat down, and the other two sides took their turn; and that, to different tunes, was the whole dance, which went on for an incredible length of time. The figures were a mixture of lancers and quadrilles, but the dancers never went out of a dignified strut, and though the first tune was followed by the inevitable Sousa marches and “Hiawatha,” however lively the music became, the dancers continued to stroll and bow and shuffle about at the same slow pace. I am told that one becomes very fond of the Rigodon, but it seemed to me intolerably dull and listless as a dance, though as a spectacle it was vastly entertaining, and gave one a chance of really seeing the people, and they were well worth the trouble of turning out after dinner to look at. The men wore white suits, most of them buttoned-up white coats of the every day sort. There were three Englishmen in evening dress, one or two in white mess jackets, and several advanced young Filipinos in grey tweeds. The American women wore every sort of outfit, from the missionaries and schoolma’ams in blouses and boots to the more exalted personages in evening dress; while the Filipinas, Mestizas, and most of the Spaniards had on the native muslin camisa, some of them exquisitely embroidered and hand- painted, and always worn with European skirts of appalling colours and cut. One little brown woman had on a long train of scarlet plush, with huge white lace butterflies fixed across and down the front, which made one burst into perspiration merely to look at; and another was in emerald green velvet, with straggling bands of gold braid meandering over it in such a queer way that I could not resist walking round her to see if any point of view would make the lines come out as a pattern, but they refused to go by any rule of any art —even the “newest.” As to the waltzes, which formed the chief part of the programme, they were very amusing too, for the variety of styles was infinite, though the universal pace was so slow. The Spaniards and Mestizos dance very well, and by that, of course, I mean Filipinos in general, for it is very difficult to distinguish between them, and to say where one race begins and the other leaves off. They are slow and
- 62. graceful. The Americans are equally slow, but not very graceful, for they walk instead of dance, holding each other in such a peculiar way, sideways and very close, the man leaning very far back, with his partner falling towards him, and the hands that are clasped held very high, and swinging up and down. At twelve o’clock everyone began to cheer and shake hands as the New Year came in; while the band played the American National Anthem, which is a most magnificent air, and then the Spanish Anthem, and then a few bars of “God Save the King,” which did for us and the Germans equally well, and which we all thought a very nice little compliment. Filipino waiters came in, carrying trays covered with tall glasses full of some sort of champagne cup, and everyone drank healths, shook hands, and wished their friends a Happy New Year. We stayed on a little longer, and I danced a two- step with a very nice American, which was the best dance I had the whole evening, for it is one in which they excel, though they perform it quite differently to what we are told at home is “the real American way to dance it,” as they do not plunge down the room in straight lines in the English fashion, but turn round more and make more of a waltz of it. Suddenly, during an interval between dances in the middle of the programme, without a word of warning, a Mestiza sat down at the piano and played an accompaniment to which a young Eurasian, in a painfully blue satin dress, and with her face a ghastly grey-white with thick powder, sang a truly terrible song. She screamed in an awful manner, and I wondered that policemen did not rush up from the streets to see what was the matter, but she was perfectly self- possessed, and faced the audience with the aplomb and self- confidence of a prima-donna. I never heard such “singing” in my life —it was the sort of thing that is so bad that you feel all hot and ashamed, and sorry, and don’t want to catch the eye of any relation of the performer. This happened not once, but several times, and is, I am told, a custom in Filipino bailes.
- 63. When we left at about half-past one, the ball was in full swing, and I afterwards heard that it went on till half-past four or five. Indefatigable people! I don’t know how they can keep it up so, for, of course, the heat was very great—a temperature in which no one would dream of dancing at home, and not a breath of cool air anywhere, but I suppose they become accustomed to it. One thing I have mentioned may strike you as odd, and that is the mixture of races and Eurasians, but there is socially no marked colour-distinction here as in every other country in the world, and this, I imagine, is because the natives of the civilised parts of the Philippines have been Christians for centuries, and intermarried with a Christian race. The fusion is not, however, really very complete, as one can see from a glance, at any gatherings, where the people of various shades of white and brown keep very much together. Some of the Eurasian women are quite pretty, but they spoil their little round faces with thick layers of powder over their nice brown skins, and use perfumes that nearly knock one down. The white men are friendly with many of the Mestizos, and dance with their pretty daughters, and are even occasionally foolish enough to marry the latter; but white women keep quite apart from the coloured folk, and it would be an unheard-of thing to dance with one; while as to marrying a Filipino, no woman one could speak to would ever dream of such a horrible fate. That is where the real impassable gulf is fixed. The Americans profess not to recognise any distinction, however, for, as I explained before, they announce that they consider the Filipino of any class as their social and every other equal, and have the expression “little brown brother” (invented by Mr. Taft), which is supposed to convey and establish this generous sentiment. The sentiment, apart from any political utility it may possess, is a noble one, but it does more credit to the heart of the Americans than to their wisdom. The Spaniards did not recognise the Filipinos as equals, but treated them with every courtesy, according to their degree, and I believe that whatever the political situation may have been in those days, society went peaceably enough, for every man knew his place
- 64. and kept it; a system admirably suited to an Oriental people. Now, however, the régime is quite different, and the sudden glare of ultra- equalising views is what the Filipinos can neither understand nor profit by. I wish I had been in the U.S.A to see many things for myself, but I have always read and heard much about the hard and fast line drawn in that country against “coloured” people and half-castes, and that the Americans have learned to adopt this custom from years of experience. This makes their professed attitude here very puzzling, and I can find no one who can even attempt to reconcile this extraordinary variation of opinion. Another unfathomable anomaly of American thought is that the “Equality,” Nobility of the Human Race —Rights as a human Being, and so on, are for the Filipinos, but all these grand schemes officially take no account of the fierce, naked savages; the Mahommedan tribes; the negritoes, and all the other wild natives of the Philippines; though how, or where, or when, or by whom the line is to be drawn and the distinction made is another unanswerable problem. New Year’s Day being a holiday, we thought we would treat ourselves to a drive. So we sent one of the boys out for a carromata, which is a sort of tiny gig, with the driver sitting on a small seat in front of his fare, in fact almost on one’s lap. Rain had been falling pretty well all day, and the carromata, when it arrived, was covered with mud, and looked such a disreputable turn-out that we burst out laughing when we saw it. However, there was no other to be had, and after all it was a very good specimen, so we climbed in over the wheel, and the driver, a boy of about twelve, gave the pony a chuck and a whack, and it turned round in the direction of the Plaza, and we stuck. Then the driver got down, and when he was out of the way and the pony became visible, we saw that we weighed the cart down so much at the back that as the little animal turned round he got his neck wedged under the shaft and was held in a rigid yoke. The youthful cochero shoved him down somehow, evidently both of them quite accustomed to the trouble, and, once righted, the little
- 65. beast tore along, and we had a delightful drive in the cool of the evening, enjoying the air, which was so fresh after the rain. We did not go far out of the town, as the sky was rather threatening, but kept more or less to the ever-amusing suburbs of native huts, which literally swarm with human beings, to every one of whom there is apparently an allowance of about six babies of under one year old, and on the roofs are cocks and hens clinging to the steep thatch; while under the hut lives the family carabao (a big grey water-buffalo) in his mudhole, along with stray dogs and wild pigs which eat up the refuse. The number of children, very young children, is something astounding, but, according to statistics, I learn that 60 per cent. of the children born in the Philippines die under one year old, so that must help to keep the numbers of grown-up people down a bit. They are miserable little languid scraps, thin and solemn, but so supremely fortunate as to wear no clothes whatever, till they are about six, when a short muslin jacket is put on, which is more for adornment than anything else. The tiny ones ride astride the mother’s hip, with little thin legs dangling, and round black head wobbling about, looking so uncomfortable, poor little souls. They are fed on rice, which they eat till their little bodies swell up to a certain tightness, when the food is taken away, and they are not allowed more till they have “gone down” again. This process results in a permanent “rice-tummy,” which makes the babies look like air- balloons set on drumsticks; but, somehow, they lose that as they get older, and if they live, are generally very slender and well made. There is a great fuss made now about this waste of infant life, much of which is ascribed to the horrible and unhuman practices and superstitions attending the birth of a Filipino child; but I imagine from the appearance of the children themselves, that the whole question is merely an example of the Survival of the Fittest, for of so many children born in such a delicate race there must be numbers who are unable and unfit to live. They are not a hardy people, these Filipinos, and the heat, fevers, and plagues of the country affect
- 66. them even more than they do the white races, oddly enough. I believe that in the wild parts the natives are stronger, and sometimes live to a great age; but there the life is simpler; the cross-breeding less frequent; in the absence of civilisation of any kind the great Darwinian Law operates even more rigorously; and the young who are sickly stand no chance at all of growing up and transmitting their weakness. The skin of these people is not a healthy skin, not a warm brown, but of a greeny-yellowy brown; their fingers are delicate and weak, and their eyes not clear or bright, but like little bits of dull plum-brown jelly.
- 67. LETTER IX. TARIFFS—INSECTS Iloilo, January 16, 1905. The day has come round for me to catch the mail, but I feel that I can hardly write calmly, as I am barely sane upon the subject I wish to tell you about, which is the Customs. I told you about the opening of our cases, and how we took them out of bond, as they were valued at £30? Well, a day or two ago the bill came in, and when we saw it we nearly fainted away, for the amount of duty came to 698 pesos—£70. Of course we thought some mistake had been made, so C—— went off to the Customs officer and asked him what it meant. All the consolation we got was that they were very sorry for us, but the Appraiser had made a mistake, and classed some of our things under Class B instead of Class A. So C—— said he could not afford this sum, which was far more than the whole of the contents of the cases were worth if they had been new. Of course it was impossible to send them back to Hong Kong, as we had taken them out of bond; but after a lot of talk, the officer said we could “abandon the goods” if we liked, which means refuse to pay the duty, when the things would be seized by the Customs and sold by auction to pay the Government; but we should be unable, by law, to buy them in ourselves. This seemed to be the only alternative open to us, and C—— came back and asked me what I thought of it, and asked the other Englishmen their opinion. They were full of sympathy and very kind, and at last one of them hit upon an excellent idea, which was to attend the sale and buy our things in for us as cheaply as possible. This, then, was arranged, but —“Oh no!” said the Customs, “you won’t gain anything by that,
- 68. because if goods, when put up for sale, do not fetch the price at which the Customs House has valued them, they are publicly burned.” So that is the end of our story. We have paid more than their value for our wedding-presents, which seems to me the meanest and cruellest imposition I ever heard of. But I won’t say any more, for the subject can only be as painful to you as it is to us. We must just grin and bear it, I suppose, but good-bye to a pony and trap for a longer time than ever, and good-bye to any little jaunts in the hot season. I must try instead to be more pleasant, and the only thing I can think of is a little lizard I have been looking at for the last ten minutes, while my thoughts roamed gloomily over each one of those seventy good golden sovereigns that have gone to help to teach the Filipino that he is my equal. A worthy cause, no doubt, but one that does not appeal to me—at any rate to the extent of 698 pesos. This little lizard, which lives in the cornice above my writing-desk, has just come down on to the window beside me and nipped up a fly in the smartest manner. This is his hunting-ground, for the windows in the house only have sliding shutters, such as I described to you, like all the houses here. Glass windows are almost unknown, but this house happens to have them along the S.-W. front, where some former occupant has put in doors on to the balcony, with glass in the upper panels, because in the rainy season the Monsoon drives in on this side. In all the houses here these little grey lizards abound, living in the cornices and corners of the ceilings, and feeding on flies, mosquitoes, and any little toothsome creature they can pick up. They must have plenty of supplies and wide variety, for one seems to come across some new sort of insect every hour of the day—and night. No fleas, however, I don’t mean that, for Filipinos are clean and fleas are rare; but all sorts of queer insects crawl and fly and sit about, all of which I suppose the lizards enjoy; and I imagine they,
- 69. in their turn, are having a good meal off some other still tinier creature. The ceilings are made of bulges of canvas or matting painted white, pale blue, or green; or, in some of the old houses, with patterns, as in Italy. In one house in Jaro, a big building with long, wide-open window-spaces, there is a ceiling that is covered with some sort of shiny oilcloth stuff, drawn up by buttons at intervals, so that it looks like the seat of some giant padded leather chair—a most fearful looking contrivance, but, no doubt, a source of much pride to the Filipino who owns it. There is a wide space above these ceilings, for the corrugated iron roofs are very deep, and here live rats, mice, cats, cockroaches, snakes, all sorts of beasts, which come down into the house for plunder. The nicest are these dear, clean, bright-eyed little lizards, which make a funny and very pretty note, a sort of clear, musical chuck-chuck. Sometimes, but very rarely, one of these lizards is found with a forked tail, and this the natives look upon as an emblem of the most extraordinary luck, and they do all they can to catch the lizard and try to take off his forked tail, which they dry and wear for anting-anting. Any kind of luck, or lucky emblem, is anting-anting, and the mystical emblems, observances, and relics of Roman Catholicism, which appeal to the Filipinos with irresistible force, have but added to their original stock of superstitions. In some of the houses there is a very anting-anting lizard, of a large size, which makes a loud, clear double note like a cuckoo, that can be heard a long way off. I have never seen a “Philippine cuckoo,” as they are called, but have often heard them, and the houses that have this anting-anting are well known. There is one in the old belfry at Jaro, another in a house the other side of the Plaza there, and one in a certain bamboo clump on the road to Molo, and so on, all over the place. A very general belief prevails that in the roof of each house there lives a big snake, which has a terrific meal of rats every now and then, and sleeps the rest of his time, coming down very rarely for water. I can quite credit this story, for the space between the roofs
- 70. must be the very place for a snake, and many people tell me they have seen these creatures, but I don’t suppose they are really in all the houses. Curiously enough, I thought there was a snake overhead before I had ever been told about such a thing, for one day, when I was sitting in the sala, I heard a most extraordinary noise in the roof overhead—a sort of heavy, dragging sound, and then a thump, and then the dragging sound again—and, somehow, the thought of a snake instantly came into my mind. When I spoke about it to some friends, half jokingly, they replied quite seriously that it probably was a snake I had heard, and then told me how they live in the roofs. Talking of noises, one of the most curious sounds here is made by the crickets, the cicadas, which shrill night and day, ceaselessly and for ever. The ear becomes accustomed to the aggregate sound of their high, thin note, though I, for one, never get to like it, and sometimes it gets horribly on my nerves, so that I feel I must go anywhere to get away from it. At first when I heard it I was always having a curious impression of being in a Swiss field in the summer; but now that has worn off, and I think if I ever go into the Swiss fields again I shall think of nothing but Iloilo. When one of these cicadas gets very near the house, it drives you nearly mad, and when, as happened a few evenings ago, one is actually in the house, everything must be searched for the beast before anyone can expect sanity or sleep. This one that got in, stowed itself away in the writing-table, and we had an awful time, standing almost on our heads and streaming from every pore, before we found it in a tiny corner where one of the drawers does not run quite into place. When we fished the cicada out at last, or rather when one of the servants came in and took up the hunt for us and caught it, we found the disturber of our peace to be an ugly little browny-black creature, with a narrow waist, and the silly thing refused to give a single chirrup to show us how it was done. Talking of insects, one of the things we are most fortunate about in this house is that we have very few of the black or red ants, which are a fearful plague in these Islands, so much so that one has to stand the furniture with its feet in small enamel bowls filled with
- 71. water or paraffine to prevent the ants crawling up, for they eat everything; and besides that, they look particularly nasty when dead in jam or butter, or floating in tea or coffee. Some of these ants are a good size, but the common sort are very small, and many of the most destructive are simply red specks that run like lightning. They are terrible destroyers, and I can’t think why ant-eaters don’t start living in the roof menageries, for they would get on splendidly if they did not die of over-eating. However, the ants do scavenge to a certain extent, and the way a busy little mob can carry off a huge dead cockroach is a lesson in natural history. The cockroaches, by-the-bye, are the size of mice. They are the most evil brutes I ever saw, besides being a constant source of terror and worry. You will hardly believe this, for you know that I never mind touching any animal—mice, worms, toads, slugs, earwigs —and how I have so often been laughed at, and even sniffed at, as rather an unpleasant young person, because I have no repugnance to taking them up in my bare hand, for, after all, they are only poor animals, and infinitely nicer to touch than many perfectly respectable human beings. Do you remember those people at Karnak who screamed when I brought them that lovely little toad with a speckled stomach? And the good folk at home who shudder if you pick up a poor slug out of a dusty road? Well, when it comes to these cockroaches, I confess that I have a genuine horror of the great red, evil-smelling brutes, with their horrible bulgy eyes and their long moving red antennæ. I can’t tell you what it is about them—but I am not alone in this, for everyone has a horror of them. They breed in the cesspits, and prefer manure to any other diet, but will gladly supplement their menu with any form of food, as well as leather, paper, books, or clothes. The houses, the shops, and the steamers are full of them, and in the evenings they come out of their holes and run about. Ugh! they make one shudder. And every now and then they take it into their heads to fly about or into the lighted rooms, and I have even seen men who have been here for years turn quite sick when a cockroach lights on them, and as for the
- 72. average woman, she screams outright, and many white women faint. These horrible brutes are the curse of housekeeping, necessitating everything being kept in glass jars or tins, and cupboards and drawers being overhauled and searched every week or so. I must say, though, that we have not had so much trouble with them as most people, and so far I have never had one amongst the linen or clothes, and I believe this is because I hang cakes of naphthaline in the rooms, and put balls of it in all boxes, drawers, and cupboards, and they don’t seem to like naphthaline, though they would come a thousand miles to eat ordinary insect powder, which is, apparently, just the very thing on which to bring up a nice little family of forty or fifty young cockroaches. There are some pleasing spiders too, one of which I saw the other day, with a body nearly the size of the palm of my hand, sitting in a huge, tough web like a hammock, and looking exactly like those in Doré’s picture of the Guest Chamber in the Castle Inn, in Croque Mitaine. I said there were very few fleas, but the mosquitoes make up for any biting that has to be done. I am beginning to get more accustomed to their venom now, but at first I was quite ill and feverish from it, and many people suffer so that it amounts to an illness, and white men frequently have to be invalided home for nothing but mosquitoes. Nothing I have ever seen in any place round the Mediterranean approaches the Philippine mosquito for venom or ferocity, and here, too, their efforts are not confined to the night-season when lucky mortals are stowed under nets with no rents in them, but they bite relentlessly all day as well. Well, I tried to leave harrowing subjects and tell you something more cheerful than the Customs woes, but I seem to have drifted into other griefs, and as my spirits are evidently damped beyond hope to-day, I had better leave off writing and end my letter.
- 74. LETTER X. A FILIPINO THEATRE—CARABAOS Iloilo, January 22, 1905. We went a night or two ago to a performance at the theatre—a Filipino performance in a Filipino theatre. I daresay it sounds strange to you to hear of a theatre in Iloilo, but you see this is really a very large town, and then all the people are musical, and they have plenty of time to rehearse. They get together little dramatic clubs, the chief one of which is not far from here, “as the crow flies,” though I think he would be a very keen crow for theatricals if he flew there as straight as he could. We heard this performance, an operetta, being rehearsed night and day before the performers considered it ready for the theatre. The rehearsals that went on until the early hours of the morning were those we cared least about; but we were really interested to hear them going on all day as well, for no one in the Dramatic Club apparently had any other occupation in life. At least, this seemed to me strange till I had become better acquainted with the Filipino character. To get to this show, we set off after dinner, driving in a hired quielez with a disturbing cockroach somewhere about it, and soon came to a squash of all sorts of carriages and carts in one of the broader streets of the town—and a squash of vehicles driven by Filipinos is something no human mind can imagine without experience. We escaped alive, and went in at a big gateway into a courtyard, passing several stalls lighted with flaring naphtha, where native women sat behind flat rush trays containing cakes and sweetmeats, tumblers of coloured drinks, and ordinary ginger-beer and lemonade bottles. This, though I did not know it at the time, was the buffet.
- 75. Inside the courtyard another high gate, decorated at the sides with palms and paper roses, and very dimly lighted, led to the door of the theatre, a big, crazy-looking building, and here stood two inconceivably stupid and self-satisfied natives bullying everyone, and making a hopeless and baffling muddle of the tickets. Why they did this I can’t think, as everyone passed into the place alike, whatever their ticket was, and scrambled up a broad wooden staircase, very steep and rickety, or else went about the ground-floor, every man looking for his own seat, and getting turned out of it by the next comer. The “boxes” were little pens railed off, containing six chairs with no room for your knees, and in and out of these and up and down the precipitous staircase jostled a crowd of Filipinos, Mestizos, Chinamen, and Spaniards, with little dark women in gaudy camisas, wearing flowers in their hair and diamond brooches. Here and there an American was patiently and persistently trying to gather information in his own language, while he took some female relation in a white cotton dress upstairs and then down again, to keep her quiet. I was so amused by these proceedings that I really felt as if it did not matter whether that was all we saw, but, nevertheless, we toiled up the staircase at the promptings of an obliging Filipino with one eye, very soon found our box, and settled down to wait for the friends who were to join us. In about two minutes, however, we were engaged in an endless discussion with a little mob of “brown brothers,” who declared quite politely that we had no right there, as the box was theirs. So we moved off and tried the ground-floor again; found another box with our number on it, empty; sat down again, put fans and programmes on the opposite chairs, and began to look about. But we were shifted again, so this time we tackled a native selling programmes, and asked him where our box was, and why the little pens all seemed to have the same number; and he, in very broken Spanish, at last made us understand that the numbers were
- 76. repeated six times, once on each side upstairs and down. This was a wonderful effort of lucidity for a Filipino, and really helped us a good deal. So we toiled upstairs again, feeling sure that we knew all about the theatre now, and determined on a shot at the sides. On the way there we were delighted to see that the people who had turned us out of our first box were being ousted in their turn, but by this time we had begun to giggle, and were too helpless with heat and laughter to take much notice of anything. At last we got into a box from which we were never evicted during the rest of the evening, though some people did come along with a programme-seller to back their claim, but we showed fight, and they went away again. The theatre, a long, wooden building, appeared even more ramshackle from the inside than it had from the outside, and infinitely more dangerous, for the electric light was supplemented by Japanese paper lanterns, which looked the last word in incendiarism; and, when one considered the packed mass of faces all round, it was wiser not to let the imagination dwell on that steep wooden stairway, which was all there was between us and the next world. The floor of the building was arranged with rows of chairs facing squarely, by way of stalls, surrounded by a row of the boxes I have described, where the chairs went sideways. Above jutted out a broad balcony with a similar row of boxes, and above that again, jammed under the ceiling, was a dense crowd of poor people, standing on what was really only a ledge with an iron rail; and they looked positively more like huge black and white flies clinging to the ceiling than anything else. Everything looked as if it must fall down or break up, but no one seemed to be worrying about their doom, in fact all the faces were remarkably pleasant and jolly. The stage was a fairly large one, with a row of electric footlights, which waxed and waned and waxed again at their own sweet will, and quite regardless of the needs of the performance. In front of the stage, on the floor-level, was an orchestra of natives who really played very well indeed, and they and all the men in the audience
- 77. were in white, which looks very quaint until one’s eye is accustomed to it. The piece performed was an operetta called “La Indiana,” a rather confused story about some old Mestizo with a white beard, whose son had secretly married an Indian, which is the word the Spaniards use for the Filipinos, and is employed by the Filipinos themselves as well, when talking Spanish. Well, the old father informed his son, an appalling, gawky, young Mestizo in a black morning coat, pepper and salt tweed trousers, and a very bright blue tie, that he must marry a white (Mestiza) girl of his, the father’s choosing. On hearing which, the hero sang a song to the effect that he would abandon the Indiana, and had a long duet with that personage to explain that they would just say nothing at all about being married. Then all the chorus came in again, the old father blessed the hero and the “white” girl, whereupon the Indiana, a frightfully ugly Filipina with a fine voice, sang a long and frenzied solo with her hair down—and then the curtain fell. I thought there must be another act, and was very much surprised to find that was the conclusion of the story. But evidently, to the native imagination, the plot was complete and the ends of poetic justice satisfied. They did not really act and sing as badly as I had expected, though, when one came to think of “La Indiana” as a public performance in a theatre, it really verged on audacity. No attempt at scenery or dress was made, the whole action taking place in a bare, worn, old “set” of a room, the usual stage room, unlike anything else on earth, and the only attempt at costume was the substitution of very ugly old European blouses for the camisa, which was a fatal mistake. We left after the first piece, though there were to be two more of the same sort, for it was very dull and depressing. There is nothing in these Filipinos, you see, for they have not the melodious voices of negroes, nor the faultless ear of Spaniards, nor the fine physique of Chinese, nor the taste of Japanese—they are simply dull, blunt, limited intelligences, with the ineffable conceit of such a character all
- 78. over the world, and when they break out into a display such as “La Indiana,” all these deplorable qualities show up in the glare of the white light that beats even upon an Iloilo stage. Yesterday we went for a delightful drive out along the Jaro road, off which we turned a little way beyond the town, and went down a rough, sandy track to the banks of a broad, half-dried-up river, not the Iloilo river, but another parallel to it, or a branch.
- 79. Riding a Carabao. To face page 78. There we got out and walked down the steep bank on to the sandy bed, where we strolled about for a long time, watching strings of carabaos coming up from being watered, each herd led by a small boy, riding on one of the big old grey cows with a calf running alongside. They looked very picturesque, with the shallow river all the colours of the sunset, and the tall palms on the opposite bank standing in black silhouette against an orange-crimson sky. The carabaos are big grey or reddish-grey water-buffaloes, with immense horns curving backwards, and a long, narrow, flat muzzle. They are used for every sort of purpose, the natives even riding and driving the great unwieldy creatures like horses, and guiding them by means of a single string passed through between the nostrils. If they want the carabao to go to the right they pull the string steadily,
- 80. if to the left, they give a sharp jerk. Sometimes when the master is angry he will pull the poor carabao’s nose, so that he tears the piece of flesh out altogether; not at all an uncommon occurrence, and nothing distressing to a Filipino.[3] In the days of the rebellion against Spain, a few years ago, when the Filipinos caught the hated Spanish friars, they ran a rope through the priests’ noses, tied their hands, and led them about like the carabaos, so that people might spit upon the hated tyrants, and insult them at their own pleasure. The carabaos are as gentle and amenable as horses with the natives; quite tiny children ride and bully the huge beasts, looking so comically small on the big backs, with their tiny brown legs hardly reaching to each side of the broad ribs, and driving whole herds with the most perfect independence and self-possession. The carabaos are not at all safe as regards white people, however, for they can smell and detect them at an immense distance; and they will occasionally charge them ferociously, so that they are very dangerous in the open country. I have heard some horrible stories of carabaos killing and trampling on white men in out-of-the-way places. They don’t gore, I suppose because their horns are so flat, but they trample to death, which does just as well. These great grey, lumbering animals are very picturesque, and redeem many a Philippine scene from utter dulness as they go shambling along, drawing the native two-wheeled cart, with its big hood of brown matting filled with bundles of emerald-green sacate grass. They can shamble at an amazing pace, and that is their usual gait; but they can gallop, too, as quickly as a horse. Besides the herds of carabaos, we saw several natives down in the bed of the river, going out to certain spots where the shelve of the sand was more abrupt for their supply of water. These were women, of course, for women do all the household tasks, even the most burdensome, their lords being busy standing about the roads or Plazas, or attending a cock-fight. These women had long bamboo poles, with the divisions knocked out and the end closed up, which they laid in the running stream to
- 81. fill with water, when they hoisted the long poles to their shoulders and carried them off like giants’ lances. The slender little figures looked quaint and pretty as they came up over the yellow, sandy, shallows in their bright red sarongs and white camisas, walking lightly and gracefully, with their thin brown feet well turned out, the fading light of the sky behind them, and the outline of dark, fretted palms. We walked through a little palm grove back to the place where we had left the carriage, driving back along the main road as the stars were coming out and the flaring naphtha lights appearing in the little mat-shed shops. There were a great many people about, and swarms of little children in fluttering muslin shirts, all enjoying the cool evening air, which was, as a matter of fact, the same temperature as an August mid-day at home. A lot of carriages and traps flew past, the little ponies tearing like the wind, amongst them the general’s wife in her victoria, drawn by ordinary Waler horses, looking like prehistoric monsters amongst the little Filipino ponies; and we met our pet aversion, three young Mestizo “mashers,” driving at a furious pace in a spidery buggy with huge acetylene lamps, and ringing a bicycle bell.
- 82. LETTER XI. SOME RESULTS OF THE AMERICAN OCCUPATION Iloilo, January 22, 1905. Mail-day has come round again, but I don’t feel as though I had much energy for writing, or anything else, as we are in the midst of a heat-wave, which means, in this part of the world, that the Monsoon has dropped unaccountably, and the heat is suffocating and appalling. Everyone is saying that such a temperature is quite unusual at this season, and some even go so far as to say they never felt it so hot here before; but this does not surprise me, as I have never yet come in for normal weather anywhere. This heat comes in the middle of a drought, too, as we have not had rain for about four weeks—another phenomenon. Our rain-tank is empty, so we now depend on the supply of brackish water from the wells, and even that is reported to be limited, which is alarming, as one would commit almost any crime to get enough water for a bath. Even at times of plenty, however, one does not rejoice in the European style of bath, but an arrangement of a tub, the acquaintance of which I first made at Singapore, and I can’t say I was much struck with it when I did see it. The tub, of wood or china, is placed in a small room with a sloping floor of concrete or tiles, and the bather stands on a wooden rack; first using what soap he sees fit, and then pouring water over himself as best he can with a tin dipper. It is an economical method in countries where water is scarce and valuable; but it was a terrible disillusion to me, after the grand ideas I had always formed, when I read how every one in the Far East has his or her own bathroom. Don’t you know how jolly it sounds in Anglo-Indian novels, or in
- 83. descriptions of the world beyond Port Saïd? A dreadful disenchantment! More than ever, in this heat, do we miss the dog-cart of our dreams, for we long to get out of the town on these hot evenings. Something to drive is a bare necessity of life out here, and even the humblest school-teachers and missionaries keep what the Americans call a “rig,” such a queer word, which is made to signify anything from a four-in-hand to a carabao-cart. The Americans all drive in a very strange fashion, holding a rein in each hand, which looks awkward at any time; but is most comical in the case of the swaggering negro who drives the military waggons, holding in a team about as fiery as a couple of old circus-horses, with a rein twisted round each of his hands, body thrown back, and the gestures of a Greek restraining an untamed pair round a stadium. The white man who drives the Government ice-cart amuses me too, for he is got up in full cow-boy pageantry—huge boots, loose shirt with broad leather belt, immense sombrero worn well over one eye, long moustaches standing out, and great gauntlets up to his elbows. All this to hawk ice about a dowdy little town. When a soldier rides one of these quiet old animals, he sits in an enormous Mexican saddle, with a very high peak back and front, and his feet, clad in big boots with huge spurs, thrust into roomy leather shoe-stirrups. To the casual observer these horsemen would certainly convey the impression that they were venturing great deeds in a wild country, and one can’t be anything but thankful to them for throwing a little picturesque relief into the humdrum life of the grey streets. We have tried hiring carriages, but besides the terrible discomfort of all hired vehicles, their prices are more uncomfortable still. Fancy, in a place like this, having to pay as much for a little carriage for two hours in the evening as one would for a brougham in London for the day! Yet such is the case, and it is only an indication of the cost of living here, which is really alarming; as you may imagine it must be when I tell you that all the Americans I have met complain bitterly,
- 84. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com