Some years ago I happened to look up shellfish in The Macquarie Dictionary and was astonished to find it defined as "an aquatic animal (not a fish in the ordinary sense) having a shell, as the oyster and other molluscs and the lobster and other crustaceans". It was the last six words of the definition that were astonishing. Until then it had not occurred to me that anyone regarded crustaceans as shellfish. A comparison of definitions in eight English dictionaries, from the UK, USA and NZ, revealed that all agree, in essentials, with Macquarie. They give the same the defining characters (aquatic animal with a shell) and the same two main types (molluscs and crustaceans) and generally cite a similar selection of typical members (usually lobsters, crabs and shrimps, oysters, mussels, and often whelks and winkles). The main point of variation is that half of the dictionaries say that shellfish is used especially of a category of food animals, while the other half (including the Oxford English Dictionary, Webster's Third and Random House) make no reference to this.

I smelt a rat, or at least the scandalous whiff of crustaceans in places where they don't belong. In New Zealand, where I spent most of my formative years, shellfish is a much-used term. I was pretty sure that my understanding of it - as a generic for molluscs with external shells that live in water, especially when spoken of as food - is standard among New Zealand English speakers and is shared by a good many Australians. Could it be that the definers of shellfish in the various dictionaries, instead of checking local usage, had simply assumed the correctness of the definitions given in one or more earlier dictionaries? I proceeded to carry out an informal survey, badgering more than 100 friends and strangers from four English-speaking regions: Australia, New Zealand, North America (the USA and Canada), and the British Isles. The main questions asked were (1) what kinds of things do and do not count as shellfish? (2) Does shellfish refer exclusively, or chiefly, to animals as a category of food? I consulted written texts too but most uses of shellfish in such texts were uninformative as to the scope of reference and did not give essential biographical information about the writer.

By smoothing out rough edges (most importantly, by considering only "typical" members of the class) it is possible to assign 90 percent of informants' definitions to one or another of the following: (a) edible water molluscs with shells, (b) edible crustaceans, (c) edible molluscs plus edible crustaceans. Only two of the four regional groups showed a high degree of internal consistency. North Americans were the only regional group who agreed fairly consistently (82%) with the standard dictionary definition.

Table: Definitions of shellfish in three regions

For 21 of the 23 New Zealand informants, shellfish referred exclusively to water molluscs with external shells. The two exceptions were people working in the food trade where international naming conventions have presumably influenced usage: both included crustaceans as well as molluscs.

Over 50 Australian-raised informants were interviewed, mainly from NSW, Victoria and Tasmania. They were divided between definitions (a), (b) and (c). No clear regional differences were discerned.

Australians in general are less familiar with the term shellfish than New Zealanders. For all the New Zealand informants shellfish is an everyday term; for many of the Australians, it is not. Most New Zealanders said they sometimes gather them and often eat shellfish whereas most Australians said that they had never gathered any sort of shellfish and a fair number said they had never or rarely eaten them.

Informants from the British Isles were fairly evenly divided between two principal definitions.

The Table above disguises further differences among informants about what counts as typical and what is marginal or uncertain. The contexts in which English speakers (other than marine biologists) normally speak of shellfish are food-related. We gather shellfish (to eat), we like (to eat) shellfish, we are allergic to shellfish. So the most typical members of the category are preferred foods. Nevertheless, some people include as marginal members other molluscs, such as cowries and cones, that are not customarily eaten in English-speaking communities. One Australian informant, a young woman who grew up mainly in Melbourne and Sydney, said "For me, the most typical shellfish are things like pipis - bivalves that live in the sand. But molluscs that live on rocks - like oysters and mussels and periwinkles - are also shellfish." "How about prawns?" - "Hmm, maybe, but doubtful." "Crayfish?" - "Even more marginal." "Crabs?" - "Definitely not." "What about molluscs that Australians don't eat, like cowries and trumpet shells?" - "Yes, they're shellfish, too, the animals in the shells."

The uniformity of dictionary definitions of shellfish raises some troubling questions about the way commercial dictionaries of English are produced and about their accuracy as records of current usage. How carefully do English lexicographers test for variability across the speech community in the meanings of the words they define? In the case of familiar terms ("general vocabulary") is it normal practice for lexicographers to copy or accept published definitions without doing original research? Is it reasonable to expect compilers of large dictionaries to check meanings of words with a range of informants selected to reflect the usual sociolinguistic variables: age, area, social background, and so on? Surely it is, at least for words in those semantic fields most likely to be show variability. Terms for plant and animal categories, especially generic terms like shellfish, fish and tree, notoriously fall into this class.

Given practical limits on the size of entries how are lexicographers to handle cases where they find considerable variation in definitions within the speech community? Should all recorded variants be listed or only the the most common ones? Should informants whose notions of what a term means are vague or in conflict with the experts be discounted? How to deal with the fact that many categories have fuzzy margins? Let me at least put my own definition where my mouth is. The following summary is far from adequate but it does indicate the main variations found in the survey.