hi dear Zahra. as we know, two features of experimental research are random sampling and randomization. but, purposive sampling is a kind of non-random sampling. according to Ary, it is violating its feature if u do so. it is possible to conduct a quasi-experimental research in this way.

I believe it is reasonable to recruit sample purposively if you intend to generalize the results of your research to a certain type of samples. However, you should recruit all participants who meet the criteria. In this case, all participants who meet the criteria are now the population. If the population size is too big, you randomize them and take a number of participants. The number of participants you recruit shall be able to represent the population. I am so sorry for not having a luxury of time to provide references, but if you keep reading other previous research studies, I am sure you will come around.

Of course, we can. But the purposive is merely an alternative method for the limited number of subjects who serve as primary data sources based on the purposes and objectives of the study. As the result, your study has poor level of reliability and high of bias. So I suggest you to use the probability sampling technique.

In a true experimental study, I believe randomization is the key factor.

In a quasi-experiment, you may use the class you are teaching. ( no randomization)

In SLA research, the experimental study in this sense is of particular purpose. For instance, you want to see the relationship between the learners' proficiency and a developmental pattern of a particular grammatical feature; you may set a criterion to recruit them to a low and high proficiency groups by adopting a standard placement test (e.g. Oxford Placement Test, Allan 2004) to represent the population.

Hope this helps!

The main issue is what population your sample represents. The basic idea with random sampling is that it represents the population that you want to generalise your findings to. Does your purposive sample represent a population, e.g. females, university graduates or people with blue eyes? If it does and is randomly sampled from that (sub)population, then you can generalise your findings to that (sub)population.

I completely agree with Martin Willis.

Experimental studies require randomization; otherwise, you may incur bias if you use purposive sampling.

i won't recommend as it may incur bias. in quasi experimental design intact sections can be used for the experimental study when we equalized the group before experimentation it may be either on the basis of intelligence ,S.E.S or cognitive styles.

"However, sampling methods of any kind are insufficient to solve either problem of generalization. Formal probability sampling requires specifying a target population from which sampling then takes place, but defining such populations is difficult for some targets of generalization such as treatments." (p. 24)

"Purposive sampling of heterogeneous instances is much more frequently used in single experiments than is random sampling; that is, persons, settings, treatments, or outcomes are deliberately chosen to be diverse on variables that are presumed to be important to the causal relationship." (p. 92)

Shadish, W. R., Cook, T. D., & Campbell, D. T.. (2002) Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston, MA: Houghton Mifflin Company.

For true experimental design, it is normally not allowed because of having low reliability and experimenter bias, being randomization is the main tenet behind an experiment. So, go for quasi experimental design, if you want to use non-representative sample

Please allow me to raise an additinal question related to this. If two groups are selected from a number of intact classes (via a placement test, and let's say the two are not significantly different) and if this is a quasi experiment study, should we report or detail what the population is?

I ask this because often the school authority allows the researcher to do a placement test to two clasess in a particular grade that they believe are equal to one another (and let's say the t-test proves it)--perhaps the classes in the school are already arranged based on students' academic achievement.

If the sample was indeed selected from a population such as two classes of grade 2 were selected from all seven classes, then yes. The population to which the generalization is intended needs to be specified. You need to explain your population and how the sample was drawn from the population.

I putu ngurah wage Myartawan

"two clasess in a particular grade that they believe are equal to one another (and let's say the t-test proves it)"

Statistical tests can prove nothing. They are probabilistic. A p value of 0.05 means that there is greater than a 5% chance that you would have got this result if your null hypothesis is true. All the t-test can tell you is whether a difference between the groups is statistically significant or not. It cannot tell you if they are statistically significantly the same.