The sentence is still the main object
Marks are not intended to replace reading. They are deliberately small because the sentence should remain the thing the learner is processing. A cue can identify a subject, verb function, complement or another reviewed role, but the learner still sees the surrounding vocabulary, word order and meaning.
That matters for a simple reason: structure is useful because it occurs in language, not because it looks tidy in a chart. Metkagram’s original idea was to put the cue where the evidence already is — inside the example — and keep a machine-readable representation behind the visual rendering.
Salience can direct attention, but louder is not always better
Research on textual enhancement asks whether making target forms visually salient can influence noticing and learning. A 2025 meta-analysis reported a modest aggregate effect across the studies it reviewed, with important limits and variation. That is a useful signal for design, not permission to turn every sentence into a fluorescent grammar Christmas tree.
Metkagram therefore aims for selective cues. The Mark should answer a real question about the current example. If every token competes for attention, the annotation has stopped being a cue and started being another layer the learner must decode.
Marks are a bridge to a Frame, not the destination
A visual cue becomes educationally interesting when it helps the learner extract something reusable. After inspecting a marked sentence, Metkagram moves toward a Frame: the stable structure that can survive changes in topic, person or tense. The learner should eventually be able to work with the Frame without depending on the Mark.
This also explains why annotation is a capability, not a requirement for every future language. A language can enter Metkagram through reviewed Frames and examples first. Marks can be added later when the annotation profile is reliable enough to deserve trust.
A visual system is useful only if it remains inspectable
Metkagram stores annotation as structured token or span data rather than only as coloured HTML. That makes a cue traceable: researchers and tools can inspect what was marked, which role was assigned and how the rendering was produced. The same principle helps quality control because an attractive highlight cannot quietly substitute for a reviewed linguistic decision.
The practical test is simple. Remove the colour and the animation. Can the learner still identify what the cue means, connect it to the sentence and carry a useful Frame away? If not, the decoration is doing more work than the method.