Non-experimental methods and design
Non-experimental (alternative) research methods differ from experiments in that they do not have an IV or a DV, are not conducted under controlled conditions and are therefore difficult to replicate, and do not show causality (cause and effect relationships). Each has strengths and weaknesses and is more appropriate to different types of research aims.
Correlational analysis involves assessing the degree of relationship between two or more co-variables, for example between the number of hours’ sleep and the score on a memory test. A positive correlation occurs when one co-variable increases as another co-variable increases, for example sales of umbrellas increase as the number of days it rains increases. A negative correlation occurs when one co-variable decreases while another increases, for example sales of bikinis decrease as the number of days it rains increases. Zero correlations occur when there is no association between co-variables. A correlational co-efficient is a numerical value expressing the degree to which co-variables are related. Measurements range between +1.0, a perfect positive correlation, and −1.0, a perfect negative correlation.
Correlations do not require manipulation and are used when experiments would be unethical.
Once correlations are established, predictions can be made, for example how many umbrellas will be sold on rainy days.
Correlations are not conducted under controlled conditions and therefore do not show causality.
Apparently low correlations can actually be statistically significant if the number of scores used is sufficiently high.
Naturalistic observations involve measuring naturally occurring behaviour in real-world situations, such as Festinger’s (1957) study where he infiltrated a cult that was predicting the end of the world, while controlled observations are conducted under controlled laboratory conditions, for example Zimbardo’s prison simulation (see page 8). Participant observations involve researchers being actively involved in the behaviour being assessed. Non-participant observations involve researchers not being actively involved in the behaviour being assessed. Overt observations involve the participants knowing they are being observed, while covert observations do not.
Observations have high external validity, as they involve natural behaviour in a real-life setting and so can be generalised to other settings.
As participants are usually unaware of being observed, there are few demand characteristics.
It can be difficult to remain unobserved and make accurate, full observations.
As observations are not conducted under controlled conditions, they are difficult to replicate to check the reliability and validity of findings.
Self-reports involve participants detailing information about themselves without researcher intervention.
Questionnaires are a self-report method where participants give answers to pre-set written questions, usually involving opinions, attitudes, beliefs and behaviour. Closed questions involve limited responses set by researchers, such as ’yes/no’ tick boxes. Answers are easy to quantify, but restrictive. Open questions allow participants to answer fully in their own words and therefore give a greater depth and freedom of expression, but are less easy to quantify and analyse.
Large samples can be generated by posting out questionnaires, which also means researchers do not have to be present when they are completed.
Questionnaires obtain lots of data in a relatively quick time.
There is a possibility of idealised and socially desirable answers, with participants answering how they think they should, rather than giving honest answers.
Questionnaires, especially those with closed questions, are not suitable for sensitive issues requiring careful and detailed understanding.
Interviews involve asking participants face-to-face questions. Structured interviews ask identical, simple, quantitative questions to all participants, while unstructured interviews involve an informal discussion on set topics, producing mainly qualitative data. Semi-structured interviews use a mixture of structured and unstructured questions.
Both quantitative and qualitative data are generated, producing a greater variety and depth of findings.
With unstructured and semi-structured interviews, follow-up questions can be asked to explore interesting answers.
Interviewers can bias responses through their appearance, age, gender, etc.
Some participants may not have the verbal skills to fully express themselves.
Case studies are detailed, in-depth investigations of one person or a small group, usually involving biographical details, behaviour and experiences of interest, for example Koluchova’s (1972) study of twins suffering privation.
Case studies allow ’difficult’ areas to be investigated where other methods would be unethical, such as sexual abuse.
Data relate specifically to one person, not an average produced from many people.
Findings relate to only one person and cannot be generalised to others.
Case studies are usually reliant on full and accurate memories, which can often be selective and affected by researcher bias.