voice in English

The English passive is a form of be plus the past participle. It promotes the internal argument to subject and either drops the agent or demotes it into a phrase headed by by, and the agent phrase is genuinely optional, which is much of the reason the construction is used: it allows a clause to be built about the affected participant without naming who acted. A second passive built on get exists alongside it, is more restricted, and carries an implication that the subject was affected or somehow responsible. Both are analytic, and neither the participle nor any other verb form carries a voice value on its own, since the same participle appears in the perfect where no voice is at issue. That is why the state is periphrastic.

There is no middle voice. The construction sometimes given that name, in which a patient subject combines with an active verb and an adverbial to describe a property of the patient, has no marking at all: the verb is in its ordinary active form and nothing signals that the subject is not an agent. A language with middle morphology marks that clause; English does not, and the reading is available only for certain verbs with certain adverbials, which makes it a construction with lexical conditions rather than a voice.

The obligatoriness is recorded as optional rather than conditioned, and the distinction is deliberate. Availability is conditioned, since only a verb with an internal argument can be passivised and a small set of transitive verbs resists it anyway. But within that domain nothing requires the marking: the active is the unmarked option, every passive clause has an active counterpart, and the choice between them is the speaker's. No English clause is ever ungrammatical for failing to be passive, which is the test this field is asking about.