Some research reports make statements like this:

A limitation of the research is that we asked respondents to self-report their levels of motivation and this affects the generalisability of the study.

I asked our team member, Dr Anthony Stacey, to comment on this statement. His response:
There is confusion here between generalisability, sample size, validity, and reliability.
Generalisability has to do with whether the sample (however large or small) is representative of the population. Probability sampling results in generalisable results by definition, but non-probability sampling can result in generalisable results if the sample is unbiased (which may depend on the topic of the study).
Sample size is irrelevant to generalisability. A small sample can be generalisable, albeit with relatively broad confidence intervals. The advantage of large samples is relatively small confidence intervals (although these may not be generalisable if the sample is not representative of the population).
Validity refers to various criteria, but commonly it is whether the respondents have interpreted survey or interview questions the way the researcher intended them to be understood. Obviously, if a study has poor validity, it would be unwise to generalise, but self-reporting is another matter.
Reliability has to do with accuracy and precision. Self-report studies can be unreliable because respondents may not give accurate responses. This may be due to a lack of self-awareness or wanting to put a positive spin on their behaviour or opinion. Therefore, unless there is reason to believe that respondents did not give accurate responses, there is no reason not to generalise the findings (provided the sample is representative and validity has been adequately addressed).