Asked whether a student going quiet in a seminar is a language problem, we said yes, and we said nothing after it. We would still say yes. We are less sure than we were about what the word is covering.
The case that makes the question hard is not the student who cannot follow the discussion. It is the student who has done the reading and could write a decent answer to the question in an hour, sitting through the whole seminar without saying any of it. We have been told this happens. We have not counted it and we do not know how often. There are at least two explanations for it, and we cannot tell them apart.
The first is practice. Our position, put plainly, is that these students have had nowhere to practise speaking. It is not the individual words. Reading a paper and following a lecture are mostly private activities, and a student can go back over a sentence they missed. Producing language out loud in front of people who are waiting for it is a different skill, and an education can leave somebody well supplied with the first of those and almost entirely unrehearsed in the second.
A student we sent an early version of the app to said something close to this without being asked. Her words were: "For Chinese students, we only remember and spell but don't know how to use it." That is one student after two days of use, and we are not going to inflate it into a finding. But she volunteered the distinction between knowing a word and being able to use one, and she is in the population under discussion, which is more than either of us can say.
What speaking costs in a seminar
The second explanation is risk. Speaking in a seminar is a social act and it has a price. You can be wrong in front of people whose opinion of you matters, in a language that makes you sound less capable than you are. Everyone else in the room appears to be managing.
The research tradition here is largely about that. Horwitz, Horwitz and Cope (1986), in the Modern Language Journal, volume 70, pages 125 to 132, argued that anxiety in a language classroom is specific to the situation rather than a general feature of the person, and that it comes from having to present yourself through a medium in which you cannot present yourself well. MacIntyre, Clément, Dörnyei and Noels (1998), in the same journal at volume 82, pages 545 to 562, set out a model of willingness to communicate in which the decision to speak at a particular moment sits on top of slower-moving things including anxiety and how competent a student believes themselves to be, which is not the same as how competent they are.
We are not in a position to survey that literature and have not tried to. What we take from the part of it we have read is that it treats speaking as a decision made under a perceived cost rather than as a capacity that is either present or absent.
The two explanations point at different treatments. If the problem is practice, then more rehearsal is the answer and the room can stay as it is. If the problem is risk, the useful intervention is inside the room: what happens when somebody is wrong, and whether the tutor calls on people or waits for a volunteer. Those are things the people in that room can change and an app cannot. We are not the people in that room.
The group that pointed instead of speaking
Now the part that goes against us.
The best single piece of evidence we have points at the second explanation. We were told about a seminar by the student it happened to. She is British, and as she understood it she was the only person in her group who had been speaking up in English. When a question came, the others settled among themselves who was going to answer it and pointed at her. Nobody said anything. Somebody pointed.
We keep coming back to that because of what it rules out. A group that can silently agree on a plan and carry it out is not a group that has failed to produce language. Those students had, at that moment, all the coordination they needed. What they declined to do was take the specific risk of being the one who speaks in English in front of everybody, and the arrangement they reached instead was social, and it was reached in silence. Nobody in that room was short of practice at the thing they actually did.
We should be careful about how much weight the story can hold. It comes from the person who was pointed at and not from the students doing the pointing, and we have never asked them what was going on. We do not have an account from the inside of what the silence is like. Neither of us has ever been the person in a room who could not follow. The two students who have written to us about the app wrote about what it was like to use, and neither of them mentioned speaking up. On the thing this essay is about, we have one story told from one side.
The explanation we favour is the one we sell
There is a further problem with us as judges of this, and it is better to name it than to have it named for us. We sell speaking practice. The explanation we favour is the one the app is built to address, and the explanation we do not favour would locate the work in a seminar room we have no access to and cannot sell into. That is not an argument against the practice explanation, because an argument is not refuted by pointing out who profits from it. It does mean the question should not be settled by the company with something to sell on one side of it.
We notice that the version of the world in which we are right is also the version in which we are needed. That kind of doubt is not new inside EAP. Moore and Morton (2004) compared the writing IELTS asks for with the writing universities actually set and found the two are not the same task, which is a problem for anybody who reports a student's readiness through that number. Turner (2004) described the philosophy behind a good deal of pre-sessional provision as "maximum throughput of students with minimum attainment levels in the language in the shortest possible time" (p. 97), a line Ding and Campion (2016) quoted as evidence that the quick fix attitude had persisted. A charge of that sort is easier to make against yourself in writing than to answer when somebody else makes it, which is why we have made ours here.
What follows practically is smaller than we would like it to be. If the practice explanation is right, rehearsal helps and the app is aimed at the cause of the problem. If the risk explanation is right, rehearsal in private might still do something, on the assumption that part of what makes a first attempt expensive is that it is the first. A student who has already said the sentence out loud, alone, is not saying it for the first time when they say it in a room. That is the honest version of the claim. It is conditional, and it is a claim about lowering a cost rather than about building a capacity.
Even that is weaker in the build than it sounds on paper. Our roleplay gives the least explicit hint that works and escalates only as far as it needs to, which is dynamic assessment in the sense Poehner and Lantolf use the term, and a description of a mechanism is not evidence that the mechanism achieves anything. The rehearsal itself is generic. Uploaded course material drives vocabulary extraction and does not reach the voice tutor, so a student practising with us is not practising the seminar they are actually going to sit in. We can build that. We have not built it. Nobody has used UniFluent for a hundred hours either, so there is no efficacy data behind any of the above. What we have is a mechanism and a plausible account of what it ought to do. We have no outcome to report.
Co-national friendship is not a failure
One thing we want to state carefully, because we have put it badly in conversation before. If a student spends most of their time with people from their own country and speaks their own language while doing it, that is not a failure. Bochner, McLeod and Lin (1977), in the International Journal of Psychology, described co-national friendship as doing a job: it is where somebody rehearses and holds on to who they already are. That is the only paper on this we have read. We have been told that the adjustment research since treats those ties as protective for wellbeing and for staying on a course, and we are taking that on trust.
Our job is not to hold an opinion about who anybody spends time with. It is to make the first English conversation cheaper than it would otherwise have been, for a student who wants to have it. Getting them ready for those interactions is the whole of the ambition, and it stops there.
The comparison we cannot run
So we are left holding two explanations and no way to choose between them. The practice account is the one we believe, and we believe it partly because it is the one we can act on. The risk account is the one our own best evidence supports, and it points at a room we will never be in. It is also possible that both are true for different students, or true of the same student in different weeks, in which case the question as posed does not have an answer and we have been arguing about the wrong thing.
What would separate them is a comparison we cannot run from inside the app: whether a student who has rehearsed with us speaks any sooner in a real seminar than a comparable student who has not. Answering that means somebody being in the seminar and counting, and that somebody is not us. Until it is answered, the position we are actually in is that we have built for one of two explanations, chosen partly on evidence and partly on what two undergraduates were in a position to build, and the strongest thing anybody has told us points the other way.
