I just received a fourth edition (1981) copy of the Dunn and Wakefield Exposure Manual in the mail today. There is some mis-reading of their methods in this thread and elsewhere. The duplex method is shown in the book as a correct method with a flat sensor incident meter (or perhaps an invercone), not with a domed meter. So for duplex readings, a flat sensor is used to meter toward the light source, then toward the camera, and the two readings are averaged.

They have this to say about using a domed incident receptor under the heading:

Three-Dimensional Incident-Light Meters

"As will be seen, the salient feature of these meters is the hemispherical type of translucent receptor employed, whose object is to effect automatically and with a single (camera direction) reading the necessary correction for most conditions of lighting (the major categories of which are given on page 126).

The claims made for this meter were investigated by practical testing under carefully controlled conditions, and by comparison with the Duplex method using a flat-receptor meter. These comparative tests confirmed that under all lighting conditions except backlighting beyond about 130 degrees from the subject-to-camera line the exposure indications for a given film speed setting agree within one-third of a stop with those given by the flat-receptor Duplex method.

The application of the Norwood-type meter is quite simple, and consists of merely pointing the meter's hemispherical receptor directly towards the camera from the subject position, irrespective of the type of lighting employed or its direction up to a lighting angle of about 130 degrees from the subject to camera line."
Italics from the original have been set bold in my quote because vbulletin automatically italicizes quotes.

So Dunn and Wakefield recommend the duplex method for flat-sensor incident meters, and a single reading with the hemisphere pointed toward the camera for domed-sensor incident meters.

The Janusphere meter attachment mentioned somewhere on APUG recently took an incident reading but combined it with a reflected reading only from the bottom half of the scene when used on a Weston meter.

The series of the woman on Kodachrome mentioned in this thread or another was taken with an experimental meter using a half dome of a table tennis ball that leaked light from "behind" the meter where the dome rose slightly above the body of the meter.

Lee