Just as there are different types of reliability, Seliger and Shohamy (1989: 102) suggest that there are different types of validity which provide ‘evidence’ for validity. Thus, their typology of ‘evidences’ of validity comprises

a) evidence on content validity which demonstrates appropriateness of data collection against the content to be measured;

b) criterion validity which provides an indication as to whether the instrument can be measured against some other criterion and compared with the previous results (concurrent validity), and whether the procedure is capable of foretelling certain behaviour (predictive validity);

c) construct validity which examines whether the data collection procedure is a good representation of and consistent with current theories underlying the variable being measured.

Chaudron (1988:24) gives another term to the content validity and suggests ‘treatment validity’ which relates to the process component of process-product study and demonstrates that the treatment was in fact implemented and that it was identifiable different from whatever it was being compared with.

For the results of the second language research Seliger and Shohamy (1989:104) identify internal and external validity. They propose that a study has internal validity if the outcomes of the observational data can be directly and unambiguously attributed to the treatment that is applied to the observed group, and that the interpretation of these data is not dependent on the subjective judgement of an individual researcher. Internal validity in this sense relates to three areas: ‘representativeness, retrievability, and confirmability of the data’ (Seliger and Shohamy 1989:104). External validity involves the extent to which the findings of a study can be generalized and applied to another situation and the categories of the study are treated as basic, applied, and practical.

To achieve evidences of validity items or questions of an instrument must be analyzed in the process of data collection. A researcher or observer should obtain information on whether the items are of ‘low-inference’ or ‘high-inference’ (Long 1980), too easy or too difficult, and whether the items are phrased and easily understood by the respondents. All these aspects are recommended to examine in the pilot phase of the research that is likely to be proved by evidences from a variety of sources, such as additional questionnaire data from pupils or teachers, interviews, surveys. Another way of examining the validity of observation is to ask colleagues to study the categories and to define the purpose of the observation. Simpson and Tuson (1995:65) treat this method as a useful check on face validity. Thus to achieve reliable and valid observation an evaluator should take into account the spatial location of an observer, engage more than one observer, involve ‘low-inference’ categories that do not require complex interpretation and check agreement of key aspects against independent studies.

Items of observation

The importance of items

In so far the language classroom observation ‘does not simply mean watching classes’ (Wallace 1991:123). An observer may record either very narrowly defined data such as a specific speech act, or more general kinds of language learning activity such as turn-taking, group work.

Any scientific research or observation is characterised by terms as ‘structured’, ‘organised’, ‘methodical’, and ‘systematic’. To follow these characteristics any data collection obtains a structure or format, and guided by some questions or variables. Croll (1986:55) notifies a variable as a basic unit that represents the process by which a concept of interest is turned into a set of working definitions whereby the results of observation or some other data collecting process can be categorized and measured.