Full rigidity, or the number is not comparable

A partial erection is not a smaller version of a full one that can be scaled up after the fact. It reads shorter and narrower by an amount that varies between people and attempts, which is why the protocol calls for full rigidity and treats anything less as a different measurement.

Why firmness is a state variable

Erectile tissue fills with blood in stages rather than all at once, and both length and girth continue to increase as rigidity increases toward its maximum. A reading taken partway through that process is a real length at that moment, but it is not the same kind of number as one taken at full rigidity - in the same way a flaccid reading and a stretched reading are not interchangeable even though both come from the same shaft. The full method for taking a comparable reading assumes the state is fixed before any of its other rules - landmark, line, angle, pressure - can mean anything at all.

Why "mostly there" does not repeat

The trouble with a partial erection specifically is that "mostly" is not a fixed point. Two attempts at a partial state can land at genuinely different levels of rigidity without either attempt being careless, so three readings taken at "close enough" firmness can disagree by more than three readings taken at genuine full rigidity would from pressure or angle alone. Home measurers see this too: in Herbenick et al. (2014), Journal of Sexual Medicine, 1,661 men who self-measured for a condom-fit study reported a mean erect length of 14.15 cm, and the authors concluded that how a man becomes erect "may be associated with his erect penile length and/or circumference." Full rigidity is closer to a fixed point than any partial state, which is the whole reason the protocol asks for it rather than for some intermediate level that might be easier to reach or hold.

Where it shows up in the readings

The clearest sign of a firmness problem is not a single low number but an odd pattern across a session. Pressure, angle and landmark errors tend to nudge a reading in one direction by a fairly consistent amount attempt to attempt. A firmness problem behaves differently: it tends to produce one or two readings that sit noticeably below the others, because rigidity at the moment of measurement, not technique, was the thing that varied. If three readings taken with identical pressure, landmark and angle still spread by more than the usual half-centimetre or so, firmness at the moment each reading was taken is worth suspecting before blaming the ruler or the hand holding it.

Girth carries the same problem, and it is easy to miss because girth gets less attention than length in most discussions of technique. A shaft short of full rigidity is narrower as well as shorter, so a girth reading taken at the same partial state as a length reading is affected the same way, by the same underlying cause, even though nothing about the tape or the wrap changed. Treating the two readings as independent checks on each other misses this - if firmness was the problem, it shows up in both numbers at once, not just in length.

Recording it honestly

If full rigidity was not reached, the honest move is not to take the reading and file it as erect anyway. A measurement is only worth as much as the conditions recorded next to it, and firmness belongs on that line the same way state, landmark and instrument do. Clinical research treats firmness as something worth recording in its own right: the one-item Erection Hardness Score, validated by Mulhall and colleagues in 2007 on data from 307 men, showed good test-retest reliability as a simple rating of hardness. A reading taken at partial firmness and labelled as such is still useful information about that attempt. The same reading filed as an ordinary erect figure is not a small inaccuracy - it is a different measurement wearing the wrong label, and it will sit in the wrong place against any published figure it gets compared to.

What this is not about

This entry is only about the firmness of the reading itself, not about how studies bring participants to that state in the first place - the three routes the literature has used for that are covered separately - and not about session length, which is its own question about how long a state can be expected to hold across three readings.

None of this is something a single photograph can settle either, because a still image cannot show whether the state in it was still increasing toward its maximum or had already peaked - an image model has no access to that information any more than it has access to scale. A rating built from that photograph, whether scored automatically or judged by a person, is answering what the image shows, not what state produced it, which is a different question from the one this measurement is trying to answer - the same distinction that separates a length in centimetres from a score like the ones Rate Cock produces.