The specific threats that damage speaking safety
Public humiliation. Being corrected in front of the class in a way that draws attention to the error produces a specific emotional response that can last for years. Students who have been publicly humiliated while speaking a foreign language often describe the experience as a primary reason they stopped trying. This is not overreaction - it's a proportionate response to genuine social threat.
Unpredictable cold-calling. When students don't know if they might be called on at any moment to speak without preparation, a significant portion of their cognitive resources goes to monitoring this risk. This is resources not available for processing language.
Comparative failure. When students hear a peer speak noticeably more fluently than themselves in whole-class formats, the comparison can be demoralising rather than motivating - particularly for students who already have low confidence.
Teacher impatience. A teacher who fills pauses, speeds through a struggling student's answer, or visibly loses interest during a hesitant response communicates that slow, imperfect speech is not acceptable. Students respond by producing less of it.
Building safety: the structural interventions
Make pair work the default
The single most impactful structural change is moving the default speaking format from whole-class to pair work. In pair work, the audience is one person. Public humiliation is structurally impossible because the performance is private. Comparative failure doesn't operate because the comparison pool has one person in it.
This doesn't mean never doing whole-class speaking - but whole-class performance should come after students have already practised successfully in the lower-stakes pair format.
Build predictability into classroom procedures
Students who know exactly what to expect in a speaking lesson experience less anxiety than students who face unpredictable formats. Establish consistent routines: the lesson always starts with a 2-minute pair warm-up, pair work always uses a visible timer, whole-class sharing always happens the same way.
Predictability is safety. Unpredictability is threat.
Never cold-call for extended speaking
Cold-calling is appropriate for yes/no comprehension checks. It is rarely appropriate for requesting extended spoken production, particularly from students who haven't had preparation time. A random student picker used after pair discussion is very different from cold-calling: students have already formulated their response in the pair phase and are being invited to share it, not required to generate it in real time.
Error correction timing and style
Correcting students during fluency activities breaks their production flow and signals that being monitored for errors is more important than communicating. This is exactly the wrong message for a speaking class.
The safe approach: note errors during observation, address patterns in the debrief after the activity, and use reformulation ("Yes, it's true that you were at the library" when the student said "I was in the library yesterday") rather than direct correction.
Tool tip: YapYapGo handles pair rotation and question delivery automatically, which means you never have to interrupt a speaking activity to give instructions or manage transitions. Uninterrupted pair work time without teacher interruption is one of the most reliable ways to signal that the pair conversation is the real activity - not a warm-up for something you're about to say. The optional Fun Mode further deflects individual scrutiny: scores attach to teams (not students), and attention rests on Fun Run or Bubble Blast rather than any one speaker, which keeps the safety properties of pair work even when a game layer is added.
