Undertaking Academic Research


C.ALEXANDER : Doctor of Education Student (Applied Linguistics and TESOL) Bristol University.


The purpose of this paper is to give UAM English-philology students guidance on conducting academic research. The range of research options discussed provide an overview of the main data elicitation techniques used in modern research. Even though only basic research methods are described, the reader is referred to other literature that can help in carrying out research projects. A key text which encompasses research methods for Applied Linguistics is :

Nunan, D. (1992) Research Methods in Language Learning , Cambridge: CUP

Nunan’s first chapter provides a general introduction to research methods.


1. First steps in research.

‘Research is cumulative in the sense that researchers attempt to build on and improve upon previous work’ (Johnson 1992)

Initial planning is highly important. Answer the following fundamental questions:

What do you want to investigate and why?

Is there a group of people/colleagues that would be interested in your findings? Who?

How do you propose to undertake the investigation ?

Brain-storming ideas and eliminating impracticable ones could be a useful way of narrowing down your options. During the ‘exploratory stage’, it may help to clarify your aims and objectives by talking to colleagues/tutors and by reading around the subject. See Lewis and Munn (1987) for discussion on research questions, Nunan (1992), or chapter three of Seliger and Shohamy’s SL Research Methods (1989, 43-64).


Lewis and Munn (1987) discuss the following exercise:

What information do you need? How much information do you need? Why do you want this information? When do you need it? How are you going to get the information? From whom are you going to get the information? Focussing on a question requires thinking carefully about the above questions.


Reviewing the literature

It is extremely important to read widely in the area you wish to research; referring to what others have done in your field will help you present your findings. It may be helpful to consult a tutor for guidance on the relevant literature in the area of your interest before you start. Citing other people’s work should be done appropriately i.e. according to academic conventions.


2. Educational Research : some common approaches


The researcher should choose the most appropriate research paradigm for his or her context. Approaches can either be quantitative (i.e. scientific) or qualitative. Research approaches could be placed on a continuum with experimental style at one end (quantitative) to ethnographic (qualitative) at the other; surveys, case studies and action research would be somewhere in the middle.

  1. The experimental approach is a quantitative method of data collection and analysis, and as the term suggests uses experiments to collect measurable data. Examples of experimental research could be finding out: (1) whether teaching students to use successful L2 reading strategies helps them improve their reading comprehension; (2) how the content of a task affects the processes in writing. In an experiment the researcher should try to establish the cause and effect between two phenomena i.e. whether an independent variable affects a dependent variable (also called outcome or criterion) usually in a simulated environment. Experimenters are less involved with their subjects than is the case with ethnographic research. An experiment has internal validity if the cause-effect relationship can be attributable to the independent variable and not to other causes. External validity (generalisation) concerns to what other people/settings the results would apply NB ‘Research findings cannot necessarily be generalised across settings (noted in Johnson 1992). For an overview of this approach see chapter two, Nunan (1992).
  2. Survey research: the purpose of a survey is to identify characteristics of a ‘whole’ group of interest (population) by looking at a subset of that group (a sample). Surveys are used to obtain demographic information about language learners. Surveys are a good way of collecting ‘facts’ rather than ‘opinions’. If survey questions are worded carefully and the survey reaches a large population, generalisations can be made. In survey design it is important: (1) to determine the purpose of the study i.e. research questions; (2) specify the population (i.e. is the survey feasible); (3) decide on methods of data collection (NB questionnaires are the most prevalent methods); (4) train data collectors or interviewers; (5) collect and analyse data (NB descriptive statistics are numbers that summarise the data). Correlation analysis can explore further important issues e.g. address the issue of non-response (NB analysis of precision of estimates and of non-response). Precision refers to how representative the survey results are of the whole population. It is important to assess the degree to which non-response affects/biases the sample. Surveys are not supposed to yield cause-effect conclusions as is the case with experiments. As no survey is value-free or theory-free, considering what motivated the study, how results are used, question wording is therefore crucial (Johnson 1992). So as to get a ‘rich picture’ of the complexities of social, cultural, linguistic and cognitive factors, surveys should not only be representative, but should also be thorough and in-depth. See Nunan (1992, 136-158) chapter 7 for more information on surveys.
  3. Case study: This approach provides individual researchers with an opportunity to study one aspect of an issue in depth i.e. usually as it occurs in its natural environment. The main reason for carrying out an intrinsic case study is interest in an issue, i.e. it is not undertaken to illustrate a particular trait or problem. An instrumental case study aims primarily to provide insights into an issue or refine a theory (NB the case itself is of secondary interest). Collective case studies look at more than one extreme (cases) with the aim of showing, despite their differences, they have some common theoretical characteristics. One weakness of case studies is data cannot easily be generalised beyond the case itself.

    Correlation studies provide information about the relationship between variables for student groups, case studies can provide ‘rich’ information about a single learner. Case studies can range from large to small scale, though it is at the ‘micro’-end of the spectrum that they are arguably most appropriate for teacher generated research (McDonough and McDonough 1997). Case studies should generally be naturalistic i.e. the individual or entity should be studied in its naturally occurring state and environment. Even though case studies involve some quantification of information, they are on the whole qualitative. They may be longitudinal i.e. conducted over a lengthy period of time and should involve five key issues (noted in Johnson 1992): (1) initial problem formulation (NB researchers start with research questions but also develop and refine them as the study progresses); (2) defining the unit of study and its boundaries (i.e. it focuses holistically on an entity and ‘it tells a story about a bounded system -- Stake 1994 p 256); (3) data-collection techniques and researcher roles (NB a wide range of techniques can be used e.g. naturalistic observations—observing natural communication in a school setting; elicitation techniques—a large number of techniques exist, it is important to consider how the information obtained could have varied depending on the task and task context; interviewing- effective informal interviewing is a main data-collection technique for case studies that are ethnographic in origin (noted in Johnson 1992); verbal reports i.e. obtaining students’ verbal reports of their own thinking (NB think-aloud methods, introspective/retrospective accounts); collecting existing information—collecting a range of other types of data in order to help clarify the research question; data triangulation; (4) analysis—the search for patterns in the data; (5) communicating the experience in a report. See Nunan 1992 chapter 4 for examples of case studies.
  4. Action research is essentially practical and problem solving in nature with the teacher as ‘researcher’. A problem is identified (by practitioners) in situation, the idea is to continually evaluate and improve teaching practice while investigating a problem. A variety of methods can be used to analyse a problem, feedback is immediate (this is an advantage of action research). McDonough and McDonough (1997, 26) discuss a ‘self-reflective spiral’ i.e. initial idea—fact finding—action plan—implementation—monitoring—revision—amended plan—and so on through the cycle. See Nunan (1992, 17-20, Nixon (1981), Elliott (1991), Winter (1989), Richards (1990 i.e. The Language Teaching Matrix chapter 7). Also Nunan’s 1989 work Understanding Second Language Classrooms, could be useful for classroom-based research of this type.
  5. The ethnographic approach is qualitative and was developed by anthropologists wanting to study entire or particular aspects of societies. The researcher has to integrate himself/herself with the group under study, it should be noted that the presence of the researcher may have an influence on the data being collected. A typical ethnographic research question could be how culture affects learning patterns (e.g. interaction and communication) in different contexts. Generalising about ethnographic research may be problematic. Ethnographic researchers pose broad questions initially and then refine or refocus them as the study progresses (Johnson 1992). Watson-Gegeo (1988, 579) argues that theory is important for helping ethnographers decide what kinds of evidence are likely to be significant in answering research questions posed at the beginning of the study and developed while in the field’. It is held that the most important aim of ethnographic research is to learn about the insider’s view of reality—the ‘emic’ view. Whether ethnographic researchers can actually identify the ‘true’ insider view of reality is a controversial question. There are a number of data elicitation techniques (noted in Johnson 1992): (1) field techniques—watching and asking i.e. participant observation and interviewing + non-written sources; (2) participant observation—a primary data-collection technique, the researcher needs to spend a lot of time ‘on-site’; (3) role options and conflicts the ethnographer both observes and participates in the cultural setting; (4) interviewing i.e. non-naturalistic; (5) holism—attending to context. Context is seen as a crucial factor in ethnographic study. Microethnographic study involves analysis of small-scale events and processes e.g. dyadic communication in classroom lessons, whereas macroethnographic study addresses large-scale events and processes in relation to classroom communication and L2 learning e.g. culture of the home. Ethnographic analysis can be (noted in Johnson 1992): (a) recursive (i.e. researchers make decisions about what is important to investigate after data has been analysed); (b) ground in data i.e. categories and concepts need to be developed that have functional relevance to the participants in the setting; (c) comprehensive to interpretative i.e. does the data involve stereotyping or generalising? See Nunan (1992 chapter 5 pages 159-183).
  6. Correlation approaches refer not to how data are collected, but to the kinds of research questions that are addressed, how data are represented, and the kinds of analysis that are undertaken to answer the question posed (Johnson 1992). In experiments a researcher may try to assess that one variable causes another, while in correlation studies a researcher seeks to ascertain the relationship between variables e.g. what is the relationship between x and y? The correlation coefficient is a quantitative measure that represents the degree of relationship between two variables (Johnson 1992). A bivariate study assesses the relationship between two variables whereas a multivariate correlation study accounts for relationships among a number of variables. Numerical analyses are used to describe and interpret data of many kinds; numerical methods can only be useful if the data can be expressed as numbers.
  7. Multisite, multimethod, and large-scale research refers to large studies in which a team of researchesr collect data from a number of sites and employ a variety of qualitative and quantitative data collection and analysis strategies (Johnson 1992). Studies are usually longitudinal lasting several years. Example research questions could be : What are some impacts of technology on L2 learning? What is the impact of immigration and migration on technology?


General texts

Cohen, L and Manion, L. (1989. 1994) chapter 1

Bell, J. (1993) Chapter 1,2, 3, and 4




Good research


McDonough and McDonough (1997, 57-73) maintain that good research should be: interesting; original; use all kinds of observations of specific events to uncover general facts; published; sensitive (i.e. research aims to discover both broad generalisations and subtle differences); objective (i.e. eliminating any biases); reliable (will the measure give the same answer given the same thing to measure?); falsifiable (i.e. how easy is it to falsify the research?); replicable (i.e. research will be valued if it shows that somebody else can do the same thing again and get the same result); generalisable (i.e. what kind of generalisations can be made about the research?); utility (i.e. can the findings be used?); ethics (how ethical will the researcher be in terms of collection, interpretation, and publication of research findings?)

Below are presented some definitions used in research literature:

  1. Basic and Applied Research. Basic research is often described as research without immediate practical utility, driven only by the advancement of theory, whereas applied research involves some kind of applicability’ McDonough and McDonough (1997).
  2. Description and intervention research. McDonough and McDonough (1997, 44-45) note that descriptive research aims to describe the significant events within the context itself (i.e. it provides a ‘rich account’ of the whole situation rather than minimising it) and that interventionist research intervenes by manipulating variables that can be identified and attempting to isolate the influence of one or more on the process.
  3. McDonough and McDonough (1997) state that (a) ontology deals with philosophical issues about the nature of reality. Holders of a realist ontological view assert that there is an objective reality that can be discovered within a particular perspective or discipline (i.e. the researcher has to uncover the causal laws); (b) epistemology deals with the relationship of the known to the unknown. An objectivist epistemology holds that inquiry can be objective and free of values i.e. a state of subject-object dualism is assumed in which the observer remains detached and distant from the object of study; (c) a hermeneutic methodology aims not to discover cause-and-effect relationships, but to make sense of a case, to understand the situation. Empirical studies rely on practical experience rather than theory.


3. Methods of data collections


a) The questionnaire

The questionnaire is probably the most frequently used data research technique used by students undertaking research at BA and MA level.

The questionnaire has four main advantages: (1) time is used efficiently i.e. questionnaires can be drafted at home, respondents can complete questionnaires in their own time, information can be collected from a large group of people, if questions are mainly ‘closed’ analysis is fairly straightforward; (2) anonymity i.e. very few research techniques offer such security of anonymity; (3) possible high return rates i.e. if presented appropriately return rates can be high; (4) questions are standardised i.e. all respondents have the same questions; there is no third person between the respondent and question. As there is no opportunity to clarify what a question means, it is important to draft questions and pilot a questionnaire.

There are three primary limitations of the questionnaire: (1) questionnaires tend to be descriptive rather than explanatory i.e. there is no interviewer to explain or probe questions for the interviewee; (2) information can be superficial i.e. a questionnaire may need to be followed up with a structured interview; (3) lack of preparation i.e. drafting and piloting takes a lot of time. If a questionnaire is not drafted, questions may be ambiguous or even offensive for the respondent. A questionnaire assumes that a respondent possesses knowledge to answer a question and that he or she actually is willing to answer a question honestly; this is not always the case especially if questions are probing, threatening or technical.

Questionnaires can contain factual questions e.g. age, sex, occupation as well as questions about e.g. experience or training. Some researchers believe it is better to start a questionnaire with items that are related to the stated purpose and leave classificatory items (e.g. age, sex, race etc) to the end so as not to put off the respondent with the seeming irrelevance of such items. When designing a questionnaire it is important to consider who the respondents will be, e.g. will they be highly educated teachers or trainee teachers? Will the questionnaire be too specialised for the respondent audience?

Specific questions provide more standardisation than general questions which tend to lead to a range of possible respondent interpretations and may be a poorer predictor of order. A general question could be: list the TV programmes you watched yesterday; a specific question would be: which of the following TV programmes did you watch yesterday? Specific questions aid respondent recall and enable more precise communication.

Closed questions are easier to analyse and more communicative than open, though some people criticise them for forcing respondents to choose from a fixed set of options. In some cases an open form is preferred for the measurement of attitudes or disapproved behaviour.

Offer a ‘no’ option. Some respondents may not have an opinion on an issue therefore if no ‘no’ option is provided in a questionnaire some respondents may ‘think up’ an option. Measuring intensity and omitting the middle option is maybe a better technique NB up to 20% of respondents may choose a middle option if it is offered, though would not choose it if it were not offered. Yes-no questions are informative though a little blunt; a ‘it depends’ or ‘sometimes’ options should be included. Multiple choice questions are also used in questionnaires though the questions should not constrain the choice thereby compromising the quality of the data. Ranked questions ask respondents to rank alternatives like e.g. ‘indicate your preference from 1-least favoured to 7 most favoured’. It should be noted that some candidates tend to do this the other way round i.e. they misread the question. Scaled questions (e.g. the Likert scale) present statements and ask for degrees of agreement e.g. (strongly disagree, disagree, no opinion, agree, strongly disagree). Data elicited using the Likert scale can be analysed numerically; though it is not clear whether the midpoint (no opinion) means the respondent is not interested in the question or the question is not relevant).

Munn and Drever (1995, 19) suggest that a questionnaire should be: attractive to look at, brief, and easy and quick to understand. Providing a cover letter explaining what you are doing and thanking the respondent for filling in the questionnaire is important. Questionnaires can be delivered by post, email/web or personally. Once mail has initially been sent, it may be sensible to send a follow-up letter re-emphasising the importance of the survey.

When drafting a questionnaire remember: (1) to eliminate non-essential items; (2) not to use sophisticated meta-language (high-brow) on an ‘unsophisticated’ audience; (3) not to include ambiguous questions e.g. double-barrelled questions; (4) to use clear response categories; (5) to be careful with opinion questions; respondents may not have an opinion on a given question and might ‘make one up’ for the sake of the questionnaire; (5) with factual questions do not ask the respondents to collect information, as they may not respond to the question or just guess or estimate something. This obviously would affect data reliability; (6) avoid leading questions i.e. one that points the respondent to the desired answer e.g. What do you like best about the course? (The respondent may not like the course at all); (7) to avoid complex questions; (8) to avoid irritating questions e.g. Have you ever attended a workshop in your whole career? (9) to avoid negative questions e.g. How strongly do you believe that no teacher should be allowed to teach English if he/she has not got an MA in English Philology?


The following bibliography may be useful for additional reading on questionnaires

Munn and Drever (1995, 10)

Cohen, L and Manion, L. (1989, 1994), Chapter 4

Bell, J. (1993), Chapter 7

Lloyd-Jones, R. (1985), Chapter 10

Oppenheim, A. N. (1992), Chapters 2, 3, 4, 7, 8

Nunan, D. (1992), Chapter 2


Whether an interview will be effective or not depends on whether questions are clear and unambiguous. There are three common types: structured (i.e. a set of fixed, sequenced questions); semi-structured (i.e. more open than structured, keys issues are listed as an ‘aide memoire’); unstructured (i.e. more conversational style, difficult to analyse data). One-to-one interviews are by far the most popular because they are easy to carry out, confidential, and straightforward to analyse. Group interviews are useful where people have been working together over a period of time and where group awareness is an important factor, NB people may be more relaxed/open in a group. It is however more difficult to record individual answers in group interviews and probing personal issues is also problematic e.g. because some interviewees may dominate the interview. There also might be difficulties in undertaking an interview in one’s own institution, e.g. probing personal areas or ascertaining whether the interviewer is being totally honest.

The interviewer should decide (1) what data he/she wants to elicit from the interviewees; (2) who are going to be interviewed and why; (3) how to explain the aims of the interview; (4) where the interview will take place; (5) whether to pilot the interview; (6) how to analyse and follow up (if necessary) the interview. In an interview, questions should be clear and non-threatening; the interviewee should speak more than the interviewer. Avoid complex, double-barrelled, leading questions. Dichotomous questions of the type ‘Do you? Are you? Have you? Was it? Is?’ should be replaced with ‘Who? Where? What? When? How? Questions should be neutral and non-confusing, and contain one idea.

The interviewer should start off by explaining the purpose of the interview (i.e. an introduction). This should be followed by a non-threatening ‘warm-up’ (i.e. easy questions). The main body of the interview would comprise the interview aim(s). Having a ‘cool-off’ (easy questions) and closure (thank you) is a good way of ending the interview. Data can be tape-recorded or taken down in note form.

Further reading

Cohen, L. and Manion, L. (1989, 1994), Chapter 13

Bell, J. (1993), Chapter 8

Oppenheim A.N. (1992), Chapter 5, 6.


Once you have decided that observation is the method you require you must negotiate access to teachers, learners, classrooms and gain permission to collect data. You should know in advance what you want to measure and so a checklist is better than relying on your memory. Nunan and Richards (1990, chapter 5) or Wajnryb (1992) have published some useful observation checklists/grids.

It is worth observing one or two classes to pilot your grid i.e. to practise observing and ticking grids at the same time. The teacher and pupils also have to get used to you because the presence of an observer in class may affect classroom interaction. An observer should also: (1) draw a plan of the seating arrangement; (2) take notes about the context in which events take place (i.e. this will help to analyse and interpret data; (3) thank the teacher and pupils for their co-operation. An observer may prefer to video an observation or use audio cassettes (NB data analysis of video or audio cassettes is time consuming). If you are however analysing your own practice, videoing can be useful.

What is ‘research observation’? (1) observation serves a research purpose; (2) it is planned in a systematic way; (3) it is recorded systematically and is related to planned tasks rather than being presented as an ‘interesting description’; (4) validity and reliability should be checked. There are two main forms of observation: structured (often used in classroom studies) and unstructured (the observer takes on different roles). In structured observation the researcher seeks to only observe the presence, absence and intensity of clearly specified types of behaviour. A researcher therefore needs to know a lot about the area under study, and should be in a position to know what types of behaviour will be monitored; in this way, data relevant to the research question will be collected. A research instrument must be designed by the observer and should be piloted a number of times. It must comprise: (1) whose behaviour (child/pupil/teacher/lecturer?); (2) categories of behaviour to be observed i.e. affective elements (attitudes), cognitive elements (intellectual components, learning strategies), (3) psychomotor elements (posture, gesture, movement), activity (repetition etc), content (lesson-related talk), sociological structure ( role, sex, race etc), physical environment ( room layout etc); (4) what counts as an act of behaviour; (5) time units i.e. unit sampling (every x seconds/minutes), natural sampling (no fixed time unit); (6) methods of classifying the behaviour (scales to all or none).

The most frequent problems in structured observations are : (a) inadequate definitions of what behaviour constitutes a given concept; (b) some observers doubt their own judgements (lack of confidence); (c) the observer effect (the presence of an observer may change the environment).

In unstructured observations the observer can participate in the activity. The main advantages are: (1) the observer can see the world of the subject group in its natural environment; (2) the researcher can record behaviour real-time, note critical incidents and can rely less on active co-operation of subjects. However, unstructured observations have some disadvantages: (a) they are less economical on time; (b) they produce masses of data; (c) they require a great deal from the observer (NB validity vs reliability).

A researcher wishing to undertake observational research should know: (1) what should be observed; (2) how observations should be recorded; (3) how to ensure the accuracy of observations; (4) what the relationship is between the observer and observed.


Further reading

Cohen, L. and Manion, L. (1989, 1994), Chapter 5

Bell, J. (1993), Chapter 10

Hopkins, D. (1993), Chapters 6 and 7

Nunan, D. and Richards, J. C. (1992)

Wajnryb, R. (1992)



As each research method has its own inherent weaknesses, researchers sometimes rely on a variety of research methods. There are four types of triangulation: (1) data triangulation i.e. involving time (i.e. research carried out at different points in time), space (i.e. more cross cultural—overcome limitations of undertaking research in one (sub)-culture, persons (i.e. using different subjects); (2) investigator triangulation i.e. uses multiple rather than single observers to record the event; (3) theory triangulation i.e. uses more than one theory to interpret the findings; (4) methodological triangulation uses more than one method.




4. Interpreting and reporting research findings

Once data has been collected it should be analysed and interpreted. Three headings may prove useful in a dissertation or research paper: (1) a classification of the findings i.e. tell the reader what you have discovered (use tables, charts, diagrams, bullet points); (2) a comparison of aspects of the data i.e. corroborate, modify/refine your findings (NB compare your findings/observations to those in established literature); (3) an interpretation of the findings (what tentative statements can you make?).

Further reading

Bell, J. (1993), Chapter 11 and 12

Lloyd-Jones, R. (1985), Chapter 12

Oppenheim, A. N. (1992), Chapter 14 and 15

Nunan, D. (1992), Chapter 10

Seliger, H. W. and Shohamy (1989), Chapter 10




In this paper I have described the main research techniques used in Applied Linguistic research at BA and MA level. I believe it should therefore be a useful reference source for UAM English philology students undertaking academic research. The bibliographical references may also prove useful for further reading.




Bell, J. (1993), Doing Your Research Project: A Guide for First-time Researchers in Education and Social Science, second edition, Milton Keynes: Open University Press

Cohen, L and Manion,L (1994), Research Methods in Education, fourth edition, London: Routledge

Cohen, L and Manion,L (1989), Research Methods in Education, third edition, London: Routledge

Denzin, N. K. and Lincoln, Y (eds) (1994), Handbook of qualitative Research, Thousand Oaks, CA: Sage

Elliott, J. (1991), Action Research for Educational Change, Oxford: OUP

Hopkins, D. (1993), A Teacher’s Guide to Classroom Research, 2nd edition, Oxford: OUP

Johnson, Donna. (1992), Approaches to Research in second Language Learning, London : Longman

Lewis, I and Munn, P. (1987), Doing Your Own Research, 3rd edition, London: Marion Boyars.

Lloyd-Jones, R. (1985), How to Produce Better Worksheets, London: Hutchinson

McDonough, J. and McDonough, S. (1997), Research methods for English language teachers, London: Arnold

Munn, P. and Drever, E. (1995), Using Questionnaires in small-Scale Research. A Teacher’s Guide, rev.ed. Edinburgh: Scottish Council for Research in Education

Nixon, J. ed. (1981), A teacher’s Guide to Action Research : evaluation, enquiry and development in the classroom, London: Grant McIntyre

Nunan, D. (1992), Research Methods in Language Learning, Cambridge: CUP

Nunan, D. and Richards, J. C. (1990), Second Language Teacher Education, Cambridge: CUP

Oppenheim, A. N. (1992), Questionnaire Design, Interviewing and Attitude Measurement, 2nd edition, London: Pinter Publications.

Richards, J. C. (1990), The Language Teaching Matrix, Cambridge: CUP

Seliger, H. W. and Shohamy, E. (1989), Second Language Research Methods, Oxford: OUP

Stake, R. (1994) Case studies. In Denzin and Lincoln (1994, 236-47)

Wajnryb, R. (1992), Classroom Observation Tasks: Resource Book for Language Teachers and Trainers, London: Prentice Hall

Watson-Gegeo, K. (1988) Ethnography in ESL: defining the essentials. TESOL Quarterly, 22: 575-592

Winter, R. (1989), Learning from Experience: Principles and Practice in action Research, London: Falmer Press