The untranslatable refrain: radif, rhyme and the limits of the English ghazal

It has become conventional to say that the ghazal came into English in the 1960s, was misunderstood for thirty years, and was corrected by Agha Shahid Ali. The conventional account is accurate about the sequence and, I will argue, wrong about what was at stake in it.

The disagreement was not about whether English poets should keep the qafia and radif. It was about what those elements do. The prevailing view — that they are formal constraints, difficulty for its own sake, a discipline the poem submits to — misdescribes their function badly enough that keeping them, as Ali urged, does not by itself recover what was lost.

A ghazal's couplets are famously autonomous: each sher is a complete poem, and the sequence has no narrative or argumentative through-line. This autonomy is the form's most-cited feature and, taken alone, it makes the ghazal sound like an anthology rather than a poem.

The radif is what makes it a poem. Because the same word or phrase closes every couplet, each new sher arrives as a fresh context for a phrase the reader has already met several times. The repetition is not refrain in the English sense — it does not return to a stable meaning — but a kind of semantic accumulation, in which the phrase is progressively loaded and each couplet is read partly through the residue of the ones before it. The couplets are autonomous in their statements and not autonomous at all in their effect.

Consider the opening of Ghalib's most anthologised ghazal in the versions by Rich and Ghia, by Pritchett, and by Ali. Rich, working from a trot, drops the radif entirely and produces a sequence of aphorisms of considerable beauty and no cumulative pressure whatsoever. Pritchett, whose scholarly apparatus is the best in English, preserves the radif in gloss and abandons it in the verse, on the explicit ground that English cannot bear that much repetition without bathos.

Pritchett's objection is the serious one, and Ali's practice is the reply to it. Ali chooses radifs that are syntactically flexible in English — prepositions, short verbs — so that repetition changes grammatical role between couplets. The strategy works, and it works by substituting a different mechanism: English gets its accumulation from grammatical shift where Urdu got it from semantic loading.

If the mechanism differs, the form differs, whatever the name on it. This is not a complaint. The English ghazal as Ali left it is a real form with a real internal logic, and it has produced work of the first order in the hands of poets who learned it from him.

The critical cost of calling it a translation of the Urdu form is that it makes the Urdu form's actual mechanism invisible to English readers, who are told they now have access to something they do not have access to. Naming the two as cousins rather than as original and copy costs the English poem nothing and returns the Urdu one its specificity.