Steve
There has been only limited formal work done on development of vocabularies for coffee (versus those for other products such as whiskey and beer which have a longer history in terms of sensory analysis, certainly in the literature). The ICO have carried out some studies in the past (e.g. their work on a vocabulary for East African coffees dating back to the early 80’s, author was Hal Macfie off the top of my head) but most of this has been defect based and I think the industry and the market is now realising the value in an approach which looks to better define a coffee from a positive standpoint or one that allows comparison on non-defect attributes to in some way allow comparison from a qualitative standpoint as per what CofE is trying to achieve.
Enough rambling, there are a couple of ways to develop a vocabulary and it depends on what you are looking for but for a start point:
1. Is it to be consumer based? i.e. should it have relevance to consumers by that I mean are the terms user friendly and will it be used to review coffees in consumer format, e.g. brewed as filter, espresso, latte given the fact that cupping extractions have artefacts and profiles that are somewhat alien to certainly the UK consumer.
2. Is it to be industry based? i.e. probable use of some defect vocabulary and terms such as acidity (which don’t translate to most consumers)
To illicit a consumer based vocabulary one way to go about it is to have consumers into structured focus groups present them with drinks in the format for which you want to develop the vocabulary and then basically look and see what you get. The important element here is the structure to the management of the group, for instance while being careful not to lead them you could ask the consumers the relevance of certain terms in the description of products which you may find more commonly in more technical vocabularies. You can also run a free choice profiling experiment where you allow consumers to use their own vocabulary to describe products; you can then actually derive quantitative data from this using a statistical tool called generalised procrustes analysis (developed in the 70’s by Gower). The latter is not for the faint hearted or numerically challenged, requires dedicated statistical software and ultimately is perhaps of limited value in terms of true vocabulary development. What would you expect to get out of all of this, perhaps 20 -25 vocabulary terms after you have distilled them down.
On the industry based side, while I would question the relevance of such a vocabulary (aside from one cupper talking to another cupper, but then again they would never agree to use the same terms!) you could simply pull together references from various cupping forms / vocabularies around the world. I once saw Ken Davids present a vocabulary at the SCAA of > 100 terms and suggest that they could be used quantitatively, now you could never use these statistically as it would be impossible to train to use so many terms consistently. Again you would have the choice as to whether to include defect terms or not depending on what you planned to use the vocabulary for.
Having carried out either of the above, at this point you could just stop, in that you would have generated terms to describe your coffee and for some that would be enough. To actually start sticking numbers against these to get quantitative values from cupping or sensory sessions that are actually statistically relevant and meaningful is a whole different ball game and would require as a minimum the following:
- panellist / cupper selection, e.g. screening for anosmia on relevant compounds and sensitivity test as well discriminatory test using triangle tests (as a guide for statistical relevance you would need typically 6 - 8 good panellist / cupper on every session)
- training, training and more training on vocabulary terms using training compounds and actual samples with statistical analysis and interpretation of panellist / cupper performance
- experimental design to ensure no session effects (e.g. randomised sample presentation using latin squares)
- ability to statistically analyse and interpret results from sessions (from a roasters standpoint that could mean also correlating process data or blend data to sensory data) for comparison of products. There are a bunch of sensory statistical methods you can use to normalise panellist data (if required), identify vocabulary terms where the variance is too high and therefore they are significant in that particular experiment etc, etc
Seeing the detail, effort and time that goes into getting truly relevant statistical data this does make me slightly sceptical of the scoring / quantitative systems employed by CofE and similar programmes in a number of areas but I still don’t think that detracts from the fact that they
do highlight good coffees and quality within the industry. However to date I don’t think they have been overly successful in reaching consumers but I guess this will come in time.
Lastly (and perhaps I should have put this at the start
) to answer your question more succinctly, you might want to check out: Guidelines for Sensory Analysis in Food Product Development and Quality Control, (Carpenter, Lyon & Hasdell, Aspen, ISBN – 0-8342-1642-6). This is a pretty good intro text without getting too much into the statistical requirements.