The late Barry Fell was a marine biologist who spent the last twenty years of his life promoting the view that there had been extensive, unrecognized contact between the "Old World" and the Americas. He published three mass market books in addition to dozens of papers, mostly in his own journal. Most of his claims were based on what he claimed to be inscriptions in Old World languages and writing systems. Although regarded by most scholars as a crank, his work aroused considerable interest among lay people.

Fell published two claims about Carrier. One, in his book Saga America, is typical of his work. On pp. 270-276 he presents drawings of pictographs from the Fraser Valley, which, he claims, contain Carrier words written in Ogham, a writing system used for Old Irish. Ogham is particularly easy to "find" erroneously since it consists of groups of straight lines similar to tally marks. For example, most scholars consider the Ogham inscriptions that Fell claims to have found on stones in New England to be plough marks.

In this case, too, other scholars do not interpret the marks on the stone as Ogham. In any case, with the single exception of "hand", the words he thinks he reads bear no resemblance to actual Carrier words. Here are the words that Fell claims to read and the actual Carrier words.

The area from which the pictographs are said to have come, the Fraser Valley, is far south of Carrier territory. Fell apparently confused the Fraser Lake area, which is Carrier territory, with the Fraser Valley.

Fell's other claim is that the Carrier language is not an Athabaskan language but is Celtic. He published this in the papers "Takhelne, a Celtiberian language of North America" and "Takelhne, a North American Celtic Language. part 2".

Takhelne is Fell's distortion of Father Morice's transcription of dakelhne ``Carrier people'', which he mistakenly took to be the name of the language. All of Fell's information on the language comes from Father Morice's book The Carrier Language. Celtiberian refers to the Celtic language once spoken in the northern Iberian Penninsula, known to us primarily from a few inscriptions.

Although Fell claims that Carrier is not Athabaskan, he offers no evidence or argument in support of his claim. He does not offer any critique of the evidence for the relationship of Carrier to Athabaskan. He seems to believe that, if Carrier is related to Celtic, it cannot also be Athabaskan. This is false: Carrier could perfectly well be related both to the Celtic languages and to the Athabaskan languages.

Carrier is considered to be an Athabaskan language because there are extensive systematic correspondances between Carrier and the other Athabaskan languages in every part of the language. Individual words and morphemes, including the most basic vocabulary items, correspond according to regular sound laws. The order of the morphemes in the complex verb is very similar in Carrier to the other languages. For example, in Carrier as in the other languages there are two sets of subject markers, which occur in different positions in the verb. In Carrier as in the other languages, the object markers precede the leftmost subject markers.

The most convincing cases for genetic relationship include correspondances in the grammar as well as the lexicon. This kind of evidence is considered particularly convincing because grammar is harder to borrow than words. One piece of evidence of precisely this type involves the first person singular subject marker. In all of the Athabaskan languages, in most forms of the verb, the first person singular subject marker, which indicates that the subject of the verb is "I", is /s/ or a cognate sound. This is easily seen in Carrier verb paradigms. In some dialects of Carrier, /s/ is used everywhere, but in a number, /i/ is used instead in the Perfective Affirmative form of the verb. This can be seen in the paradigm of "to go around by boat" in the Lheidli dialect. Quite a few of the Athabaskan languages do the same thing. Such an odd alternation between /s/ and /i/ in a particular tense/mode is very unlikely to arise independently in different languages, and it is the kind of thing that is very difficult to borrow. The fact that Carrier does this is by itself a strong argument that Carrier is Athabaskan.

In contrast, Fell's argument that Carrier is Celtic is unconvincing. To begin with, his presentation of the data is, to put it charitably, sloppy. The Celtic forms that he gives are drawn apparently at random from different Celtic languages. (Indeed, in a few cases he gives forms from non-Celtic languages. It is hard to imagine what relevance these might have.) The Carrier forms that he gives are full of errors. Some are not identifiable and appear to be spurious. The glosses are often incorrect and he routinely omits diacritics that mark crucial distinctions. He does not understand Morice's somewhat idiosyncratic transcription. For example, he compares Carrier words containing the sounds that Morice represents by //r// and //rh// with Celtic words containing /r/, apparently unaware that these sounds are the voiced velar fricative (Carrier Linguistic Committee spelling //gh//) and the voiceless velar fricative (Carrier Linguistic Committee //kh//). The apparent similarity to the Celtic is greatly reduced once one is aware what sounds these are.

In some cases, Fell compares words whose analysis he does not understand. For example, in (1977/1992;200) he compares Carrier susti (Fell's sesthi) with Celtic swain and sofnio. The Carrier form actually means "I sleep". The components are /s/ "perfective", /s/ "first person singular subject", and /ti/ "to sleep". (The u is a meaningless epenthetic vowel, inserted by rule.) Once we remove the aspect and subject markers, we see that the part that actually means "to sleep" bears no resemblance to the Celtic comparanda.

Similarly, in (1977/1992; 202) Fell compares Carrier buye' (Fell's beye) with Welsh Bachgen "child". The Carrier form actually means "their son". bu is a possessive prefix; the stem meaning "son" is ye'. There is no resemblance between this stem and the Welsh form with which it is compared.

In other cases, Fell compares words that are not native to Carrier. For example, in (1977/1992; 201) he compares Carrier sooniya (Fell's sunya) with Celtic isarno "silver". The resemblance is not great, but in any case the Carrier word is clearly a loan from Cree.

The central problem with Fell's argument is the lack of systematic correspondance between the Carrier forms and the Celtic. He merely presents a long list of words and morphemes that resemble each other, in varying degrees, in sound and meaning. Over two hundred years of experience with historical linguistics have taught us that this technique is not reliable. The problem is that the probability of chance resemblance is too high. If we insist on regular sound laws, we greatly constrain the degrees of freedom, and thereby greatly reduce the probability of chance resemblance. Fell is actually aware of the issue of chance resemblance but does not appreciate the problem that it causes. In Fell (1977/1992) at p. 220 he gives an incorrect calculation of this probability.

As mentioned above, correspondances in submerged morphology are considered to be particularly strong evidence of genetic relationship. Since Carrier and Celtic languages are both highly inflected, if they were at all closely related we would expect to find correspondances in their morphology.

Fell gives only one grammatical argument. He claims that in Carrier as in Celtic the subject pronouns follow the verb. Even if this were true, it would not be convincing: since there are only two possible orders, the probability of such a similarity arising by chance is very high.

From the Carrier example that Fell gives, it turns out that he is not talking about the order of the independent subject pronouns and the verb but rather the order of the affixes that mark the subject and the verb stem within the word. His claim is incorrect; without exception, the subject marking prefixes precede the verb stem, as Father Morice explains in his grammar. Fell's claim is based on his misanalysis of a single verb form taken from the texts included in the grammar. This is ubuzni "we call them". Fell believes that the /z/ is the stem of the verb "to call" and that /ni/ is the first person plural subject. In point of fact, it is /ni/ that is the verb stem, as is easily seen from related forms such as huyulhni "they call him", nyuhutni "they call you", and sts'utni "one calls me".

In sum, Fell does nothing to rebut the well-founded received view that Carrier is Athabaskan. His argument that Carrier is Celtic is entirely unconvincing.

Yinka Déné Language Institute © 2006