Accordance 3.5 Biblical Software: Is It Reliable ?
Université Catholique de Louvain
Département d'Études grecques, latines et orientales
Place Pascal 1
1.1. Accordance (or acCordance: both writings seem to be used) software may be described as an integrated tool for Biblical studies: it gives access not only to the original Hebrew and Greek texts of the Bible, the Septuagint (here after: LXX) and of the New Testament (here after: GNT), but also to an impressive panel of supporting devices: translations (in English — several ones —, but also in Latin and various modern languages), dictionaries, lexicons, commentaries, parallel texts, etc.
One of the major features of Accordance is that every word of the Hebrew and Greek texts is grammatically analysed and translated. So, the user is immediately told that, for instance, the Greek form GR epoiEsen (on this transcription, see here under, § 2) comes from the verb GR poieO, meaning "to do, make", and is its third person of the singular of the indicative aorist active. Thus, one becomes able to search, count and handle the text of the Scriptures in a theoretically infinite number of ways. And this is made possible by a very powerful and cleverly imagined software.
Moreover, one may ask Accordance to find words (or expressions) sharing some characteristics — for instance, nouns at the dative plural. More refined searches are possible, such as associations of some particularities — nouns at the dative plural preceded by a verb, or by a preposition, etc.
Accordance’s price depends on the configuration the customer wants : it varies from 69 (which is reasonably cheap) to 1400 US $ (which seems to me rather expensive). The software is edited by the Gramcord Institute (2218 NE Brookview Drive; Vancouver; WA 98686, USA). It is made by OakTree Software (498 Palm Springs Drive, Suite 100; Altamonte Springs; FL 32701, USA).
I am a brand new user of Accordance: its version 3.5 for Macintosh was sent to me at the very end of the year 1998. However, before I allowed myself to be delighted by its numerous features, I felt it necessary to put its accuracy to the test to see if it really is the reference tool it claims to be.
By "putting its accuracy to the test", I simply meant to see if the program’s results were truly reliable. Since my main focus was the Greek texts of the LXX and of the GNT, I did not check the Accordance’s numerous other modules or tools (translations, dictionaries, lexicons, parallel texts, etc.).
1.2. The points I wanted to check were totally factual and, I believe, basic.
The first one concerns the text itself: is it correct and without mistakes or discrepancies when compared with the reference paper editions it is supposed to reproduce?
The second one concerns the grammatical analysis: is it satisfactory? In this examination, I do not intend to discuss some subtle details of classification or interpretation, but primary features like: are the nouns properly labelled as nouns; verbs as verbs; is the translation of every word appropriate; and so on.
The last point goes over the lists and the counts: are they reliable? If one asks for all the examples of a given lemma, form or expression, are they all listed and duly counted?
1.3. What are the results of my check? There is a lot of good things to say about Accordance: everybody will easily find them for instance in the forms issued by the Gramcord Institute itself as traditional sheets of paper or in its Web site (http://www.gramcord.org). So, I feel there is no need to repeat them here.
I shall rather concentrate on the disillusions I experienced: I hope that the errors and problems I will describe may help, on the one hand, the Accordance’s editors to improve their product, and, on the other hand, the Accordance’s users to be aware that, marvellous as it may seem, this software should be treated with caution: it is by no means perfect, and one should carefully control its results before using them.
2. Some conventions
Here after, LXX and GNT will conventionally always refer to the Greek texts of the Septuagint and of the New Testament as edited in Accordance 3.5 for Macintosh.
Accordance uses the Rahlfs edition for its LXX, and the Nestle-Aland 27th edition for its GNT.
All the English translations of Greek words will be taken from Accordance (when available!), and so will the grammatical labels given to the Greek words ("noun", "verb",etc.).
Every Greek form will be transcribed by Latin letters, each time preceded by GR, and will be taken from Accordance’s lists of lexical or inflected forms, which use neither accents nor breathings. I will conventionally use here the following Latin letters to transcribe some Greek ones:
E = Eta
h = aspiration in thEta, phi and chi
O = Omega
Oi = Omega with iota subscript
ph = phi
th = thEta.
3. Results of the check
In the LXX, Accordance gives the right spelling of the conjunction GR kai ("and, also, even, and yet, but") when contracted with its following word (this phenomenon is called "crasis" in Greek grammars). The standard writing shows a breathing (customarily called a "coronis") above this contraction.
If one looks at the GNT, there is no sign of crasis at all. Instead of that, one has in some cases just an acute or a grave accent above the a of GR kai.
On the whole, there are 147 examples of incorrect writing of the GNT crasis.
I have also regularly found false punctuations or accents.
3.2. Grammatical analysis
a) I ask the examples of the lemma GR Omos in the LXX. This lemma is supposed to be a noun translated as "shoulder". What I actually get, is a mix of 52 examples of the noun GR Omos (accent: properispomenon) and the adjective GR Omos (accent: oxytone). The noun means really "shoulder", but the adjective has the meaning of "raw, crude; cruel". Notice that even in its superlative form (GR Omotate: 4Mac. 9: 30), this adjective is nevertheless translated as "shoulder"… Accordance has no lemma known for the biblical adjective GR Omos.
b) In the LXX, forms like GR eidOs, which are perfect participles of the verb GR oida, "to know", are analysed by Accordance as forms of an almost incredible panel of verbal lemmas: the about fifty items I checked are attributed not only to GR oida, but also to: GR eideO, GR eidotes, GR eidO and — pieces of pure (grammatical) poetry — GR eidon (sic!) and GR oraO, "to see".
c) In the LXX, the five examples of GR errethE, which come from the verb GR legO, "to say", are analysed by Accordance as forms of a nice variety of verbal lemmas: GR eipon, GR eirO, GR reO ("to flow" : sic!). In theGNT, the ten same forms are consistently (and falsely) said to come from GR eipon…
d) In the GNT, whenever one has an example of crasis (always, alas, without any typographic sign of crasis: see above, § 3.1), Accordance considers only the first element of the crasis, without any allusion to its second part (this is shown in the Instant Details window, which gives the grammatical analysis of every word the cursor is posited on). For instance, GR kagO, crasis of GR kai, "and, also, even, and yet, but", and GR egO, "I, me, my", is translated by Accordance as "and, also, and yet, but" (that’s all). There are 147 examples of this mistake. It is, however, possible to get a better analysis : one has to select each crasis, then, go to the Amplify Menu, choose Selection, and ask for Parsing. The window which appears gives a description of both elements of the crasis.
In the LXX, the four examples of the crasis of GR kai with GR ekei, "there", are headed under two different lemmas: GR kaiekei (three forms) and GR kakei (one form).
e) In Greek, when a infinitive is qualified by an article, the article’s gender is always the neuter (no exception known). This is rightly acknowledged by Accordance in most cases. But there are about 50 examples where Accordance analyses this neuter article (GR tou or GR tOi) as a masculine in the LXX.
f) In 1Mac. 3: 12, 1Mac. 14: 13 and 4Mac. 4: 22, the verbal form GR polemOn (accent: perispomenon) is the present participle of the verb GR polemeO, "to wage war". It is analysed by Accordance as the genitive plural of the noun GR polemos, "war".
g) In Matt. 26: 45, the imperatives of the two verbs GR katheudete and GR anapauesthe are analysed by Accordance as indicatives. Notice that in the parallel text Mark 14: 41, the Instant Details window correctly analyses the same two forms as imperatives — but the Parsing window (see § 3.2d above) gives two analyses of these very same forms (as imperatives and as indicatives…).
h) I ask for GR E (as "particle") in the LXX. The result is a list of five examples.
Actually, one of these five supposed particles (Prov. 5: 3) is the feminine relative pronoun GR E (with rough breathing, while the particle has a smooth one). Accordance analyses it as a conjunction meaning "or, than, either/or" or as an adverb meaning "surely".
i) In 1Mac. 5: 30, one has a verbal form GR Eran, which is the third person of the plural of the indicative aorist active of the verb GR airO, "to take up". Accordance analyses it as the singular accusative of the feminine noun GR ara, meaning "cursing, oath"…
j) In the forms of GR stadion, meaning "a stade (length), stadium", I ask for the genitive plural GR stadiOn in the LXX. One of them (Dan. 4: 12) is analysed as a neuter. Another one (with exactly the same form: 2Mac. 12: 9) as a masculine form. Thus, the same form of the same word may be labelled by Accordance either as masculine or as neuter.
k) In the Acts 13: 46, the form GR strephometha means "we turn ourselves", with the typical meaning of the medium voice of the verb GR strephO, "to turn". It is analysed by Accordance as a passive.
l) In 1Sam. 24: 10, one reads GR ina (with rough breathing), the conjunction meaning "that, in order that, so that". But Accordance analyses it grammatically as a form of a purely imaginary conjunction GR una (sic: with initial Greek upsilon; meaning: "that, in order that, so that"). Needless to say, one finds a GR una (sic) in the list of conjunctions in the LXX.
m) The LXX has, quite naturally, the lemma of the article, GR o, in its lists. This lemma is conspicuously absent in the GNT.
3.3. Lists and counts
a) Obviously, every bit of error like those listed above (§ 3.1-2) is duly embedded and counted in the lists and counts Accordance realises…
b) There is a structural counting problem in the LXX for every word used in a crasis.
Let’s take the example of a first element GR kai ("and, also, even, and yet, but") used in a crasis. This GR kai is of course part of the total number of GR kai. Now, if one asks for the number of forms of the lemma GR kai in the LXX, one gets only the number of GR kai without crasis : the result is thus only a partial one. Happily, in the GNT, the answer includes the number of crasis of GR kai with another word.
The same mistake happens with the second element of a crasis : in the LXX, it is simply excluded from the count of its lemma, whereas it is duly included in GNT’s count…
Another discrepancy between LXX and GNT is that, in the GNT, you may easily get the total number of crasis: just ask for GR «"chrasis"» (on this curious spelling, see here under), or, if you want a specific word involved as first part of the crasis (let’s say GR kai), ask for GR «kai "chrasis"». But if you ask the same question in the LXX, you do not get the right answer, but a window asking for "Select inflected forms", which proves totally irrelevant. So, in the LXX, you get no answer at all to your question. There is no help to get from the Accordance Guide on line: the word "crasis" is unknown.
By the way, Accordance’s spelling of GR chrasis (sic: with an initial chi in Greek alphabet!) is puzzling, since the correct Greek writing has an initial k (Greek kappa).
c) Looking for individual examples of crasis of GR kai ("and, also, even, and yet, but") with several others words may prove complicated if you naïvely ask for the same forms in the GNT and in the LXX.
I take the example of GR kai plus the modal particle GR an. You may ask form them as a form or as a lemma.
If you ask for the form GR kan, there is no problem neither in the GNT nor in the LXX.
But if you look for the lemma involved in GR kan, the situation becomes complex. In the GNT, you should, quite normally (see § 3.3b above), ask for a crasis with either GR kai, or GR an. In the LXX, things are different: you must look for GR kaian (sic!). For the twofold lemmas given to the crasis GR kakei, see above § 3.2d.
4.1. The conclusions I draw from this by no means exhaustive check are:
a) text itself: some improvements should definitively be made, although the level of accuracy is not exceedingly bad;
b) grammatical analysis: quite unexpectedly, the accuracy of the Accordance software is much less satisfactory than one should wait for;
c) lists and counts: they suffer the same defects as the grammatical analysis, on the accuracy of which they basically rely;
d) to be complete, I may add that the way the software works is truly superb — although it is a pity that GNT and LXX have sometimes to be handled in distinctly different ways.
Of course, nobody is perfect, and I am willing to gladly admit Accordance should be allowed to suffer from a limited number of errors. But as regards the grammatical analysis and the lists and counts, I feel the mistakes are much too numerous to be acceptable — I regret to add that I did not stop my checking procedure because I discovered no more errors, but because I had no more time to devote to this search: everywhere, there are carefully hidden mistakes waiting to be detected…
My final conclusion is: as they are for the present being, the Accordance 3.5 for Macintosh data should definitely NOT be considered as reliable enough. In some cases, they may even prove severely inaccurate. Telling this, I do not underestimate the immense quantity of work and imagination which has been necessary to bring Accordance to its present level. I just observe its actual number of mistakes.
While using Accordance, nobody should thus forget to check its results (by cross-analyses, made by computer or with the aid of books), to be sure that what is given to him by Accordance is really right. This is, alas, especially true for people not familiar with the Greek language. Everybody should become aware that what comes from a software, even wonderfully wrapped as Accordance, is just what human creatures put into the program: so, a computer’s dictum has no reason to be more reliable than its source…
4.2. And now, what for the future? Since I am (happily) by no way in charge of Accordance, I may allow myself to dream about the best possible way to make the product better.
The first thing I would do, is to officially warn every Accordance’s user of the existence of an excessive number of inaccuracies.
The second one would be to promise that their number will be dramatically reduced. The first target I would adopt should be: no less than 99,97% accuracy (this means: no more than 0,03% words with one or several mistakes). Afterwards, the subsequent targets should of course change, with a constantly decreasing amount of errors…
The third one would be to make a thorough and complete check of every of the 587350 and 138167 Greek words (if Accordance’s counts are right…) of the LXX and of the GNT. This examination should be total: text, spelling, grammatical analysis, meaning.
The fourth thing would be to get the Accordance’s users constantly involved in the process of evaluating and bettering the product. Although Accordance is devoted to the Scriptures, it has nothing holy per se! So, users should be warmly encouraged to be extremely critical about Accordance’s data. One should never forget that, even after the first complete revision I am dreaming about (with its 99,97% accuracy…), we would be left with no less than about 176 and 41 wrongly written, analysed and/or translated words in the LXX and the GNT…So, Accordance’s users should be earnestly asked to let the editors know about every bit of imperfection they notice. And there should, of course, be some kind of reward for the help they give to the product.
Since I did not check Accordance’s Hebrew Bible, I am not able to judge its level of accuracy. But it goes without saying that it should be as carefully bettered than the Greek texts…
I am totally persuaded that the only way to get a reasonably tolerable number of mistakes is to strictly adhere to a zero-errors policy. Should this (admittedly dreamed) procedure become effective in the case of Accordance, its final result should be a much better and more reliable product: it would become a truly authoritative scholarly reference work, instead of just an interesting but problematic tool. Everybody should be the winner in this enterprise. Electronic Antiquity Vol. 5 Issue 1 - February 1999 Technical Editor, Terry Papillon: