Since macrolanguages do not replace ordinary languages in E16/E17/E18 and are
relatively few in number, I do not discuss them further here.
2.3. Language classification. According to E16:
Language classification information comes from a variety of sources. Generally, the organization of lin-
guistic relationships outlined in the International Encyclopedia of Linguistics (Frawley 2003) is fol-
lowed for most language families. For Austronesian languages, the Comparative Austronesian
Dictionary (Tryon 1995) is followed most frequently. Departures from these primary sources are in-
cluded based on more recent comparative studies as they are reported to us.
As I pointed out in a review of the 15th edition (Hammarström 2005), the reference to
the International Encyclopedia of Linguistics, 2nd edn. (IEL, Frawley 2003), is an
empty self-reference since the IEL follows Ethnologue’s 14th edition in its classifica-
tion (Frawley 2003:xiv):
These lists [of language families and their members] were compiled by Barbara Grimes—not by authors
of the articles—using the Ethnologue … There remain great controversies in the field over which lan-
guages belong to which families, and, indeed, some of the groupings in the lists are at odds with the po-
sitions of the authors of the articles. The goal of including the lists was not to resolve controversies—or
promote them!—but to ensure that the user has maximum information.
The IEL adds no further substance to the classification, and surely one can provide the
user with better ‘maximum information’ than arbitrariness and contradiction, which the
passage boils down to. Similarly, the classification in the Comparative Austronesian
dictionary (Tryon 1995) says (Grimes et al. 1995:122) it follows the
Ethnologue 11th
edition (Grimes 1988) for all but the higher-level nodes, without adding or committing
any extra substance to this classification.
Furthermore, the E16 claim that ‘departures from these primary sources are included
based on more recent comparative studies as they are reported to us’ is not accurate. In
reality, SIL has a team of subarea editors who prepare reports to the general editor. The
present reviewer has seen such reports. These reports cover classification and combine
opinions from SIL area experts and advice actively solicited (by the subarea editors)
from non-SIL specialist linguists. The subarea editors find compromises for differing
opinions within their respective areas, but there is no evidence in the macrolevel classi-
fication of any attempt at unifying the (widely differing) principles for classification
current in the subareas. Beyond these subarea reports, according to testimonies from
many colleagues in linguistics, it appears that classification information submitted vol-
untarily by non-SIL linguists to the editor is set aside if not cosubmitted with a support-
ing SIL member.
While it would be inappropriate to ask that the SIL embark on a large-scale enterprise
of historical linguistics, it is perfectly appropriate to request that the procedure underly-
ing the E16 language classification should be described correctly, regardless of whether
this procedure is justified. A procedure that gives credence to SIL members over non-
members obviously could not survive scientific scrutiny, but it would nevertheless pre-
vent misunderstandings about the E16 classification, which is increasingly being cited
as ‘compromise’ classification.
The corresponding section in E17 (a similar passage is retained in E18) has improved
in its descriptive accuracy and no longer contains the circular justification:
Language classification information comes from a variety of sources. The Ethnologue attempts to report
the generally accepted consensus of scholars working in the language family based on published works
and scholarly review. For this edition, the language classifications for several major families have un-
dergone thorough review and revision. The sources on which the classifications are based are not overtly
cited in the language entry but may be included in the list of general references listed at the country level.
The sources used for classifications are available on request by contacting the Editor; see Contact us.
REVIEW ARTICLE
727
However, the actual procedure for the ‘attempts to report the generally accepted con-
sensus’ is still not declared. Whatever the attempts were (and were not) is highly rele-
vant information, and the failure to disclose it runs counter to scientific principles.
Similarly, if there are sources explaining the basis for classification, why not cite them
overtly? Lastly, the statement that sources are available on request appears to be nomi-
nally correct, but the underlying sources for various languages appear not to be in order.
I asked (Nov. 2013) for the source of the classification of five languages chosen for
their peculiar E16/E17 classification: Kamar [keq], Phimbi [phm], Santa Maria La Alta
Nahuatl [nhz], Enwan (Edu State) [env], and Eastern Ngad’a [nea]. For Kamar [keq]
and Enwan (Edu State) [env], the classification sources were not known. For Phimbi
[phm], the source was said to be Maho 2009, but this source actually follows E16 and
does not have any independent evidence for the language Phimbi [phm] or its classifi-
cation (Maho 2009 and p.c., November 2013). For Santa Maria La Alta Nahuatl [nhz],
the source was said to be Campbell 1997, but Campbell does not mention Santa Maria
La Alta Nahuatl [nhz] and makes no subdivisions of Nahuatl varieties at all (Campbell
1997:134), so this source gives no information on how to classify Santa Maria La Alta
Nahuatl [nhz] against the dozens of other Nahuatl entries in E16/E17. Nor does Camp-
bell (1997), in turn, cite any other sources that treat the classification of Santa Maria La
Alta Nahuatl [nhz] (Lastra de Suárez 1986 is cited but does not cover Santa Maria La
Alta Nahuatl [nhz], while Lastra 1990 does, but is not cited). For Eastern Ngad’a [nea],
the sources for classification were said to be Blust 2008 and Gray et al. 2009, but nei-
ther of these sources mentions or cites any work (beyond Ethnologue) that mentions
Eastern Ngad’a [nea]. Thus, out of the five queries for classification sources, none pro-
vided any noncircular information on the classification of the languages in question. In
§5 we review the actual outcome classification (rather than the description of it).
2.4. Sources and updates. A large number of sources for individual data items are
cited properly. A welcome novelty since the E16 edition is that ‘[c]itations of published
sources in the text of Ethnologue follow the conventional format of author surname fol-
lowed by publication year. Personal communications, unpublished, and more general
sources such as censuses, are identified by placing the year before the name of the
source’. For most items of data, however, no source is cited; in particular, most of the
time no source is cited to justify the entry itself, or to at least explain where the data
came from.
From a scientific perspective, the lack of systematic sourcing is the biggest weakness
of E16/E17/E18. The lack is somewhat puzzling. After all, no data is made up of thin
air—it all comes from somewhere
5
—so why not declare it? E16/E17/E18 gives only
one reason, namely, space: ‘Lamentably, space does not permit a listing of [every cor-
respondent who has communicated with us since [the fifteenth edition was released in
2005 (E16)/the sixteenth edition was released in 2009 (E17)]/every contributor since
Ethnologue came into existence (E18)]. Moreover, the list of contributors over the
nearly six decades of Ethnologue publication, whose contributions can still be seen, de-
fies documentation’ (E16). Possibly this is a valid reason for the book version, but for
the internet version there are no space limitations.
728
LANGUAGE, VOLUME 91, NUMBER 3 (2015)
5
Fortunately, it has not been the general practice of the E16/E17/E18 editorial team throughout the years to
discard the source or its name once the information from it has been integrated. As I have experienced myself,
it is occasionally possible to find out where a certain entry actually comes from via the help of a willing SIL
member.