Language learners who take an online language test usually expect to receive an evaluation of their speaking ability in the results. But online tests don’t do a very good job of assessing speaking ability because they lack construct validity: they cannot create the type of conditions the learner will be speaking the language in, such as a conversation or presentation. The iBT TOEFL has speaking components, but the test taker has no interlocutor, creating a highly unrealistic speaking situation – a monologue spoken into a microphone with no audience – on which speaking ability will be evaluated. Some online tests contain no speaking component at all; claims about the test taker’s speaking ability is even more inferential than those of the iBT. None of this prevents test makers from making confident claims about their test’s ability to measure learners’ speaking ability.
Speaking is a particularly difficult skill to test properly, especially the ‘spoken interaction’ described in the Common European Framework of Reference. Research has shown that learners perform differently under different conditions. For example, a test taker scored more highly when paired with another learner in a conversation than when assessed by interview with an examiner (Swain, Kinnear, & Steinman, 2011). Conversation is co-constructed by participants, who build on and scaffold each other’s utterances. Conversation requires cooperation, the successful negotiation of meaning, strategies to understand the other person, asking questions, requesting clarification, affirming, and paraphrasing. Is it likely that any of this can be evaluated by an assessment that does not require the learner to do any of these things?
Online tests have emerged from the psychometric testing tradition, which assumes that an ability is stable in an individual, and therefore requires isolation of the individual in order to avoid extraneous influences. This is the opposite of most spoken language in use. We should call into question the usefulness of tests that make claims based on a lack of validity.
The best way for spoken language to be assessed is by an expert interlocutor interacting with and observing learners in interactions with others over a period of time. Language teachers – trained and experienced in assessment and evaluation techniques, and in many cases able to assess learners over the course of a session or semester – are best placed to offer this kind of assessment.
Swain, M, Kinnear, P., & Steinman, L., Sociocultural Theory in Second Language Education, Multilingual Matters 2011
Picture credit https://tc.iupui.edu/