The repression-dissent literature in the corpus rejects a one-direction model: coercion can delegitimize authority, deter participation, or produce both effects in different constituencies.
The coercion reports treat state force as conditional rather than mechanically radicalizing. Perceived proportionality, the behavior of demonstrators, victim identity, prior trust, verified evidence, elite response, institutional accountability, and the wider public demand for order all influence whether coercion is interpreted as legitimate enforcement, abuse, or proof of systemic failure.
Visible coercion can broaden opposition when previously neutral observers interpret it as disproportionate or unjust and when institutions appear unwilling to investigate or correct abuse. In the reports, accountability is a major moderator because it can preserve the claim that the system is capable of self-correction.
Coercion can also raise the perceived costs of participation, isolate highly committed factions, fragment networks, and increase public support for order. The corpus therefore rejects casualty counts or police presence as sufficient evidence of radicalization.
The reports find that public interpretation changes when political contention includes violence or serious disorder. This is presented as an empirical framing and legitimacy problem, not as tactical advice.
Repression and dissent are endogenous: governments choose coercion in response to threats, and movements adapt in response to government behavior. Observational correlations can therefore confuse cause and selection.
A short-term protest surge, anger after an incident, or increased justification of confrontation is not the same as durable preference for replacing the political order.
This brief synthesizes claims made in the supplied research corpus. It does not convert report conclusions into forecasts, endorsements, or operational guidance. Where the underlying reports disagree or rely on different definitions, the disagreement is preserved rather than silently resolved.