You are on page 1of 58

THE TEXT OF ARISTOTLE'S DE SENSU AND

DE MEMORIA

The present article is part of a larger project. In 2000 I conducted


some palaeographical studies of a few mss. containing the De sensu, and
I found that, even though much work had been done on the manu script tradition, scholars had still not taken full advantage of all the
relevant mss. I also found that the previous editors had not consulted
the available secondary textual evidence, and the most thorough edi 1

tion in this respect, that is, the edition of P. Siwek , does not have clear
editorial principles on how to treat the mss. and the secondary evi dence. This gave rise to the thought that a new edition of the text
could be useful, and seeing that Aristotle's philosophy of mind is one of
the most hotly debated issues in contemporary Aristotelian scholar ship, I began work on new critical editions of the De sensu and the De
memoria.
In this article I will present the results of my work on the manuscript
tradition. Thus, I will re-examine the textual tradition using all the
mss. written between the 10th and the 14th centuries ; I have collated
all these mss. with the use of microfilms. The investigations have,
2

I believe, produced some hitherto overlooked conclusions .


The editor of the De sensu and the De memoria has to consider the textual contributions of 21 different mss., written between the 10th and
the 14th centuries. The mss. written later than the 14th century were
never likely to provide interesting material, and, although I often dis agree with Siwek's stemmatic conclusions, his investigations seem to
justify ignoring them. Most of the earlier mss. have been physically
described several times in printed editions of Aristotle's writings as

(1) P. Siwek, ed., Aristotelis Parva naturalia, Graece et Latine, Roma, 1963. Prior to
Siwek's edition, the De sensu and the De memoria had been edited three times in the
20th century : A. Fo rster, ed., Aristotelis De sensu et De memoria Libri, Budapestini,
1942 ; R. Mugnier, ed., Aristote. Petits traite s d'histoire naturelle, Paris, 1953 ; W. D.
Ross, ed., Aristotle. Parva naturalia, Oxford, 1955.
(2) I have generally tried to provide all the evidence, but the reader should note
that a complete apparatus for the De sensu and the De memoria, recording all ms.
readings, can be found in D. Bloch, The Manuscripts of the De sensu and the De memoria.

Preliminary Texts and Full Collations, in Cahiers de l'Institut du Moyen -Age grec et latin,
t. 75, 2004, p. 7 -119.

Revue d'histoire des textes, n.s., t. III, 2008, 1-58

david bloch

well as in articles
and monographs on the subject, and I shall not restate
these facts 3.
The relevant mss. are the following:
E
1853, s. .
Y
261, s. - .
Vc
266, s. - .
314, s. .
C
M
37, s. .
i
2032, s. .
b
1859, s. (Only
).
La
253, s. .
214, s. - .
H
X
435, H 50 Sup, s. - .
y
2034, s. xiii (Only
).
Paris, Bibliotheque nationale, gr.

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

xiii xiv

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

xiii xiv

Paris, Bibliotheque nationale, Suppl. gr.

xiv

Vatican, Biblioteca Apostolica Vaticana, Urb. gr.

Paris, Bibliotheque nationale, gr.

xiv

Paris, Bibliotheque nationale, gr.

xiv

xiv

De memoria

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

xiv

Venice, Biblioteca Nazionale Marciana, Marc. gr. Z

xiii xiv

Milan, Biblioteca Ambrosiana, Ambr.

xii xiii

Paris, Bibliotheque nationale, gr.

De sensu

(3) For complete lists of surviving mss. containing the


, cf.
,
, in
, Paris, 1937, p. 327-333;
,
, Roma, 1961, p. 23-24;
,
. (n. 1), p. - . For descriptions and discussions of all or some of the mss. that are treated in this article, cf., in
chronological order, F. H.
,
, in
, t. 10, 1915, p. 188-214; M.
,
, in
, t. 7, 1933, p. 141-160, 261-281, 355-367; ,
. (n. 1); I.
,
, in
, 6. Fo ljden, Ser. A, Bd. 2, No. 1, Go teborg, 1943; H. J.
, ed.,
, Templum Salomonis, 1943; H. J.
, ed.,
, Leiden, 1947; R.
,
, in
, t. 26, 1952, p. 36-46;
,
... (
);
,
. (n. 1); A.
,
, Paris, 1963; P.
, ed.,
,
Paris, 1965, p.
; P.
,
, in
, t. 21, 1967, p. 17-41; D.
,
Peri a to` mwn grammwn, Amsterdam, 1971; P.
& D.
& D.
& J.
, eds.,
, Berlin, 1975; M. C.
,
, in
, t. 80,
1976, p. 111-159;
,
, in
t. 7, 1983, p. 161-76; A.
,
Peri Enupni` wn.
, Berlin, 1990 (unpublished diss.); M.
,
, in
, t. 12, Wiesbaden, 2001; F.
,
, in
, t. 21, Wiesbaden, 2005. In addition, Paraskevi Gatsioufa is currently
working on the textual tradition of the
. Unfortunately, it has not been possible to arrange for any collaboration between us.

R.

Parva

Mugnier

Les manuscrits des Parva naturalia d'Aristote

Desrousseaux

P. Siwek Les manuscrits grecs des Parva naturalia

d'Aristote

Siwek

Fobes

op. cit

xvii xviii

Textual Problems in Aristotle's Meteorology

cal Philology

de Corte

l'ame d'Aristote
ster

naturalia

Melanges offerts a A. M.

Revue de philologie

op. cit

Du ring

Fo r

Aristotle's De Partibus Animalium, Critical and Literary

Goteborgs Kungl. Vetenskaps

Commentaries

Classi

tudes sur les manuscrits du traite de


E

och Vitterhets Samhalles Handlingar


Drossaart Lulofs

Aristotelis

De Somno et Vigilia liber, adiectis veteribus translationibus et Theodori Metochitae commen


tario

Drossaart Lulofs

niis et De Divinatione Per Somnum


des Parva naturalia d'Aristote
manuscrits

cf. supra

Siwek

La filiation des manuscrits

Revue de philologie
op. cit

Siwek

Wartelle

grecs d'Aristote et de ses commentateurs


clviii clxxxvi

tote

Aristotelis De Insom

Mugnier

Moraux

Moraux

Aristote. Du ciel

Le Parisinus graecus 1853 (ms. E) d'Aris

Scriptorium

Harlfinger

Die Textgeschichte der

pseudo aristotelischen Schrift


Harlfinger

Les

Inventaire des manuscrits

Moraux

Reinsch

Wiesner

Aristoteles Graecus. Die griechi

schen Manuskripte des Aristoteles, I : Alexandrien London


The Text of Aristotle's De Motu Animalium

Nussbaum

Harvard Studies in Classical Philology

N. G. Wilson A Mysterious Byzantine Scriptorium : Ioannikios and his

Colleagues

Scrittura e civilta,

der aristotelischen Schrift

Escobar

Parva naturalia

Rashed

schichte der aristotelischen Schrift De Generatione et Corruptione


Berger

Die Textgeschichte

berlieferungsgeschichte der
Ein Beitrag zur U
berlieferungsge
Die U

Serta Graeca

Die Textgeschichte der Historia Animalium des Aristoteles

Serta Graeca

De sensu

aristotle's de sensu and de memoria

Ud
260, s. - .
209, s. - .
O
S
81.1, s. - .
W
1026, s. (Only
).
N
258, s. - .
ma
1921, s.
87.4, s.
C
4).
(Only
va
87.20, s .
87.21, s. .
Z
P
1339, s.
These mss. divide into two well-defined families 5. P alone of the Greek
mss. contains some traces of a third textual tradition. I designate the
two families a and b, in accordance with the practise of most editors.
The tradition sometimes represented by P will be designated g. First,
I will briefly set out the stemmatic relationships of all the mss. Afterwards I will present elaborate arguments in favour of this stemma.
The a family is represented by E, V (442b27-453b11), Y, b, Cc, M, i.
E is the ancestor of the entire family. V stems from a ms. copied
directly from E, and Y was copied
directly from E. Subsequently b
they
was copied directly from Y. CcMi are also descended from E, but
have been contaminated by readings from the b family. Cc is the
ancestor of M and i.
The b family is represented by L, Ha, X, y, U, Od, S, W, V (436a1442b26), N, m, v, Ca, Za, P. None
of these mss. stem from each other.
b divides into three branches 6. The least complicated branch is represented by LHaXy. These four mss. all stem from r. L and Ha have
Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

xii xiii

Venice, Biblioteca Nazionale Marciana, Marc. gr. Z

xii xiii

Florence, Biblioteca Medicea Laurenziana, Laur. Plut.

xii xiii

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

xiii xiv

De sensu

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.


Paris, Bibliotheque nationale, gr.

xiii xiv

xiv.

Florence, Biblioteca Medicea Laurenziana, Laur. Plut.

xii

De sensu

Florence, Biblioteca Medicea Laurenziana, Laur. Plut.

. xiv

Florence, Biblioteca Medicea Laurenziana, Laur. Plut.


Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

(4) For the date of this ms., cf.

xiv

xiv.

,
(n. 3);
,
, in
, t. 70, 2003, p. 100, note 92.
&
&
&
,
. (n. 3), p. 291a-293 (D. Harlfinger), dates it to
the 13th century. At the time of Siwek's edition C was still believed to be a 14th
century ms. On Ca as part of a more or less complete edition of Aristotle's works, cf.
G.
& M.
,
, in
, t. 64, 1997,
p. 177.
(5) This has been accepted at least since G.
, ed.,
,
Lipsiae, 1898.
(6) Scholars have not agreed upon sigla for the b family. Since my reconstruction of the stemmatic relationships differs from that of other scholars, I have been
forced to introduce new ones.
,
(n. 3) and M. C.
, ed.,
Wilson

art. cit.

G. Vuillemin Diem

Aristoteles Latinus in den Fragmenten der Quaternuli des David von Dinant
toire doctrinale et litte raire du moyen age
finger

Reinsch

Zum

Archives d'his

Moraux

Harl

Wiesner op. cit

Vuillemin Diem

Laur. 87.7 et Laur. 81.18

Rashed

Burgundio de Pise et ses manuscrits grecs d'Aristote :

Recherches de theologie et philosophie me dievales


Biehl

Nussbaum

art. cit.

Aristotelis Parva naturalia

Nussbaum

david bloch

been copied from the same ms. (

f), and X and y have both been copied


h, also

from a ms. ( ) closely related. The second branch, designated

divides into two sub-branches. In the first sub-branch a single lost ms.

m) gave rise to U and to x, and x in turn produced O

SW. Od and S

were copied from the same ms., and W is closely related. The second

i
l (from
which V and N were produced). The same ms. i, together with s and a
ms. of the a family, also constituted the basis of a contaminated copy of
sub-branch has a more complicated structure. A single lost ms. ( ) was
a

the ancestor of C ,

k (from which v and Z

were produced) and

the texts which produced m ; for the De sensu, which is found twice in m,

i
b is represented

both copies have been made from the same ms. The entire sub-branch
is very contaminated and corrupt. The third branch of

only by P, but this ms. is corrupt and contaminated, and therefore it is


difficult to use. Thus, when the first two branches agree in reading, I use
the siglum

b, even if P has another reading.

The resulting stemma can be seen in the illustration at the end of


this article. This stemma is, at least partly, controversial. In the follow ing I will state the arguments on which it is based.

The division of the textual tradition into

and

is based on the

large number of passages in which the readings of two particular


groups of mss. disagree. The following passages present just a few
examples from the two texts, and many others could be produced
436a13.
436b17.
438a10.
440a24.
440b19.
442a28.
442b23.
443a8.
444a23.
444b11.
446b14.
448a15.
448a16.
449b17.

ousai b : om. a
mori` ou b : om. a
autw b : om. a
aki` nyton b : kinyton a
kata pollouq lo`gouq b : polloiq lo`goiq a
katakaome` nwn/katakaiome` nwn b : kaome` nwn a
twn jumwn ai shysin poiy`sei b : an poiy`seien ai shysin a
ap' e gju`mou b : apan jumou a
e n autaiq b : peri autaq a
poiousi proq to me` li b : om. a
o ayr kai to udwr b : pan udwr a
le` gw d' b : legome` nwn a
kalw b : all' wq a
ote ... tugja`nei a : om. b

Aristotle's De Motu Animalium, Princeton, N. J., 1978, used b


LH X, while

Forster, op. cit. (n. 1), p. x, used b

for LX and b

for O S and b

for

for SW.

(7) For some further examples from the first few Bekker pages of the De sensu, cf.
436a1, 436a17, 436a18, 436a19, 436b17, 436b20, 437a6, 437a9, 437a20, 437a32,
437a32, 437b5, 437b9, 437b24, 437b29, 438a18, 438a19, 438b16, 438b17.

aristotle's de sensu and de memoria


450a4.
451b15.

o nown b : onoma`zomen a
e te` rouq b : allouq a

Since this division is not controversial, there is no need to argue the


case any further.

The

a family

E, M and Y were part of Bekker's textual foundation, and every edi tor of the text has considered the contribution of these mss. V was used
by Forster, but only for the first part of the De sensu, and only because
this part of V does not agree with the
who has used C

a family. Siwek is the only editor

and i, and he also collated b, which he dismissed as a

textual witness, but it is sometimes cited in his edition.

1. Paris, Bibliothe que nationale, gr. 1853 (E)


E is by far the oldest of the mss. containing the Parva naturalia, and
its value as a textual witness has never been disputed. It is one of the
most studied Aristotelian mss., and there is no need to restate the gen 8

eral facts here . The texts of the De sensu and the De memoria in this ms.
both contain a number of spelling errors, but they seem to be careful

a tradition often presents a more elliptib tradition, but this is in complete accordance

copies of their exemplar. The


cal Greek text than the

with the nature of Aristotle's writings, as viewed by modern scholars.


That E is the ancestor of the entire

a family is not generally acknowl-

edged and requires some argumentation. This I will provide in the fol lowing sections on the other mss. of

a.

2. Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 261 (Y)


I have argued in another article that Y was copied directly from E
as regards the De sensu, and my general argument can be gathered
9

from that article . My subsequent examination of the De memoria has


confirmed that the same relationship holds between E and Y for this
text. Therefore I will not reargue the case extensively, but merely state
the most important facts.

(8) For discussion of E, cf. in particular Siwek, Les manuscrits... (n. 3), p. 107109, 121-122 ; Moraux, art. cit. (n. 3), and the introductions to the published edi tions of the Parva naturalia.
(9) D. Bloch, A Note on the Textual Transmission of Aristotle's De sensu, in B. Amden
et al., eds., Noctes Atticae, Copenhagen, 2002, p. 24 -35.

david bloch

The most compelling evidence is two substantial omissions in Y 10,


and a strange error that Y has in 446a27 11, which are all best
explained by the physical appearance of E. Otherwise, E and Y regu larly agree in reading ^ often in clearly wrong readings 12. They differ
only in obvious errors and in a number of revealing instances where Y
has tried, unsuccessfully, to correct a spelling error in E 13.
The reader should also note the arguments presented by Forster and
the passages from the De partibus animalium cited by During, both lending clear support to the claim that Y was copied from E 14.
3.

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

266 (V)

There are 32 peculiar readings in V :


(Sens.) 443a1, 443a19, 443b3-4, 443b16, 443b19, 444a1, 444a8, 444a17,
444a18, 445a9, 445a28, 445b11, 446a1, 446b14, 446b18, 447a7,
447a15, 448a30, 448b3, 449a30, (Mem.) 449b20, 450a17, 450a18,
450b28, 451a11, 452a5, 452a7, 452a9, 452a14, 452b25, 453b1, 453b9.

There are four possibly correct readings in V which are not shared
by any other ms. of the a family :
444a3.
447b1.
447b19.
448a25.

e n bV : an a
e nanti` a bV : e n a
y ins. bV : om. a
y om. bV : ins. a
15

There are 12 passages in which V has corrected (or tried to correct)


the readings found in EY :
444a14.
444a23.
446a1.
447a13.
447a14.
447a25.
447b6.

to toiouton eidoq V : ton toiouton eidoq EY


ugieinote` rwq V : ugieinwte` rwq EY
tyq V : tou EY
duoin V : duein EY
e la`ttw V : e la`ttwn EY
oudete` raq V : oudete` traq EY
duoin V : duein EY

(10) 446b3-4 and 448a20-21.


(11) Reading ton pedoklei q metaxu. The scribe of Y did, however, correct his
own mistake by erasing pedoklei q.
(12) Cf. e.g. 441a25 (lisjro`tyta) and 447a25 (oudete` traq).
(13) Cf. e.g. 438b12 (pole` mw
Arist. : pole` moi E1 : pole` moiq E3Y), 439a1 (ti
Arist. : ty E1 : tyq E3Y) and 447a8 (e stin Arist. : e styn E : e sty Y).
(14) Cf.
, op. cit. (n. 1), p.
, and
, op. cit. (n. 3), p. 50.
(15) Bekker accepted the reading of bV, but Mugnier, Ross, and Siwek all preferred the reading of a.

Fo rster

xiii xv

Du ring

aristotle's de sensu and de memoria


447b20.
448b21.
451a11.
452a17.
453a2.

duoin V : duein EY
ato`mw
V : atomon
tyn V : to EY
y EY : om. V
kai EY : om. V

EY

V is a complicated ms. For approximately the first four chapters of


the De sensu (436a1-442b26), that is, the section referred to by the
Byzantines as Book I, it follows a tradition closely related to the

b fam-

ily, and in particular it is related to N. But from 442b27 and to the end
of the De memoria it adheres even more closely to the

family and in

particular to the readings presented by E and Y, deviating only in the


passages mentioned above. The entire ms. has been thoroughly cor 2

rected by another scribe (V ), who used a ms. of the

b family. The pasb family

sage 436a1-442b26 will be treated in the discussion of the


below.
The peculiar readings of V in the

a-section are almost all trivial scrib are coinciden-

bal errors and omissions, and the readings shared with

tal if not the result of contamination. The corrupt passages of E and Y


in which V contains a more likely variant reading must be considered
(attempts at) easy emendations.
These being the only readings that differentiate V from E and Y the
figures suggest a ms. presenting no independent evidence of the texts :
blatant errors are plenty, but in general they are clearly errors, not
attempts at emendation, and the extremely strict adherence to the tra dition of E and Y suggests dependence on one of these ms. It has usu ally been thought closer to Y than to E ; scholars have sometimes con sidered the possibility that it was copied directly from Y, but several
features tell against this

16

EV alone agree in error against Y in 5 passages, whereas YV alone


agree in error against E in only one insignificant instance

17

. Nussbaum

has quite different statistics for the De Motu Animalium, and she believes
that V and Y are closely related and both independent mss., but her
statistics are certainly at odds with those obtained from the De sensu
and the De memoria.

(16) Nussbaum, art. cit. (n. 3), p. 121 -122, considered but rejected this view,
which she wrongly attributes to Drossaart Lulofs. He (Drossaart Lulofs, Aristotelis De somno ... (n. 3), p. xlix -l) claimed that It is highly probable that we possess a
faithful copy of y in the ms. V ... , but y is not, as Nussbaum seems to think, a mis take for Y ; on the contrary, it is the supposed exemplar from which Y, according to
Drossaart Lulofs, was made.

ai shytai EV1 : om. Y), 446b15 (meme`ristai d' EV : meme`risauty y auty EV : y auty Y), 447b29 (ete`rw
EV : ete`rwq Y),
EV). YV : 451a20 ( y E : y YV).

(17) EV : 446b3 (ei

tai di'

Y), 447b28 (y

r Y : om.
453a22 ( ga

david bloch

Additional support of the claim that V is descended from E can be


found in a few revealing passages :
444a9.
444b33.
450a18.

vu`xin V : e xin E(gra`fetai vu`xin in marg.)Y


tyq tou hei` ou duna`mewq] farma`kou in marg. EV
hnytwn E(twn alo`gwn dylady in marg.)Y : hnytwn twn
alo`gwn dylady V

Considering the respective texts of E and V, it seems possible that V was


copied directly from E, but the number of errors is uncomfortably high.
Therefore, V was probably copied from a ms. now lost that had in turn
been copied from E

18

As regards V , the readings stem from a manuscript related to

19

but the scribe also had some independence of mind, and Harlfinger

De lineis inseDe sensu made by the scribe, one

praised his corrections, adducing seven examples from the

cabilibus

20

. From the corrections in the

could point to the following three passages :


443a7.
438a1.
446a6.

plu`siq V : fu`siq cett.


g' V : t' H. Diels : om. cett.
jwri q V : my jwri q a : jwrish b
2

21

22

As a text witness V

is not interesting, but the readings must be consid -

ered as possible conjectures and emendations on a par with suggestions


made by modern scholars.

4.

Paris, Bibliotheque nationale, gr. 1859 (b)


There are eight peculiar readings in b :
450a8, 450a9, 450b9, 450b23, 451a30, 452b10, 452b26, 453a31.

The peculiar readings apart, b departs from the readings of Y in


only three passages :

(18) Cf. also

Escobar

op. cit. (n. 3), p.

98-101, 205.

(19) Cf. e.g. 437b3, 442a2, 442a21, 442b2, 443b6, 446a3, 446a28, 447a19,
448b26, 451b2, 453a26.
(20)

Harlfinger

op. cit.

(n. 3), p. 138-139. Cf. also

Escobar

op. cit.

(n. 3),

p. 100.

(21) J. C. Cook Wilson , Conjectural Emendations in the Text of Aristotle and Theo phrastus, in Journal of Philology, t. 11, 1882, p. 119, who had not seen V, proposed the

same conjecture.
(22)

Fo rster

op. cit. (n. 1), records the reading in V as the reading of V

, but,

even with the use of microfilm, it is clear that this correction was made by V . Forster's collations of V are generally less reliable than his other collations.

aristotle's de sensu and de memoria

451a7.
452a9.
452b29.

mataball Y : metaba` ll b et plerique : metaba` l


metaba` llei U
toiau` ty Y : toiau` tyn b et cett.
y Y : y b

CcMi :

There are five peculiar errors shared by Y and b :


449b10.

e xapatatai Yb : e xapata[.]tai E(littera illegibilis) : e xapata

450a9.
452b8.
453a20.
453a25.

gnwri` zei Yb : gnwri` zein cett.


w Yb : varia cett.
to Yb : tou cett.
eu hupory` soi Yb : eu hupory` sei EP : eu hupory` s cett.

cett.

De memoria

De sensu

b contains the
but not the
. The ms. has been corrected by another scribe who used a ms. of the b family, providing the
Aristotelian editor with no readings of interest. The corrections have
often been made with such care that the original reading cannot be dis cerned. A few passages suggest that the corrections were made using a r
ms., but the evidence is not sufficient to establish this conclusively 23.
When uncorrected, b adheres closely to the a family and in particular to
Y.
All the peculiar readings in b are trivial errors constituted by different word-order, omissions, plain scribal errors, etc. Apart from these
errors, two of the three passages where b departs from Y are blatantly
corrupt in Y, and the third might be thought to be so. In all other
cases, b shares the readings of Y, including the five specific errors
found only in b and Y. It is also noteworthy that b always agrees with
the reading of Y whenever E presents more than one reading (E1 and
E3). A rough count produced 21 examples 24. In such cases, Y agrees
sometimes with E1, sometimes with E3.
Considering the extreme similarity of b and Y, there can be no
doubt that b is a direct copy made from Y. This is confirmed beyond
any doubt by the fact that Y is itself copied from E. The peculiar errors
found only in b and Y cannot be descended from a common ancestor
now lost, since we know for a fact that the errors first arose in Y.
This conclusion agrees with During, who claimed that b was copied
directly from Y for the
, and Escobar reached the
same conclusion as regards the
. P. Siwek reached similar

De partibus animalium
De insomniis

(23) Cf. 452a20 (

e pi tou EH b XH : varia cett.), 452a23 (e pezy` tei b LXH m :


2

om. C Mi : e pizytein Z : e pizytei cett.).


c

(24) Cf. 449b7, 449b8, 449b20, 449b22, 449b25, 449b29, 450a6, 450a8, 451a2,
451a5, 451a19, 451a29 (twice), 451b9, 451b14, 451b21, 451b23, 451b30, 452a4,
453a14, 453a20.

david bloch

10

conclusions, and in his edition this particular ms. is not part of the tex tual basis, but it is still cited at times in his apparatus (e.g. on
452a7) 25.
5.

Paris, Bibliotheque nationale, Suppl. gr.

Apostolica Vaticana, Urb. gr.

2032 (i)

37 (M) ;

314 (Cc) ;

Vatican, Biblioteca

Paris, Bibliotheque nationale, gr.

Cc, M and i are the only serious candidates for representing a tradition independent of E although belonging to the a family, but still
they share a great many readings with EYVb and some other features
linking them in particular to E 26. That they are part of a is not controversial.
However, if the mss. constitute independent evidence, we would
expect at least some valuable readings. For this purpose, there are
only five relevant passages, the remaining being obviously inferior
readings (see the list of peculiar readings below) :
438b27.
445a12.
450a13.
451b2.
453a1.

hermon CcMi : hermy cett.


kai tw te CcMi : kai to` te E : o kai tw b
tou nou me n CcMi : tou noytikou P : tou nooume` nou cett.
maho`nta ti CcMi : maho`nta OdSV2 : paho`nta cett.
ody`pote e poi` ysen CcMi : oti me` ntoi pote e poi` ysen rmNm :
ody`pote poiysai EYV1 : oti me` ntoi pote poiysai b2 : oti
me` ntoi e poi` yse pote V2(in marg.) : ote me` ntoi pote e poi` ysen
vZa : otoi me` ntoi pote e poi` ysen P.

As explained below (n. 32), 438b27 is probably not the right read ing. Even though it does not trivialize the passage, the mistake is easily
explained.
The reading of 445a12 in CcMi was accepted by Forster and Ross,
and may be right (though I doubt it), but the correction of to` into tw

would be possible for most attentive scribes. Similarly, these mss. may
well be right in 453a1, but the reading is almost certainly derived
partly from the b tradition.

Du ring
Siwek

Escobar
Siwek

(25)
, op. cit. (n. 3), p. 39, 52, 55 ;
, op. cit. (n. 3), p. 110-111,
205 ;
, Les manuscrits ... (n. 3), p. 118-119, 146 ;
, op. cit. (n. 1), p.
.
(26) For instance, the shared marginal comments in 439a31 (ECc) and 442a25
(ECcM).

xxii

xxi

11

aristotle's de sensu and de memoria

The reading in 451b2 is interesting, but not only was this passage
always likely to be corrupted
ing in some
Ross,

27

; there is also evidence of the same read-

b family mss., viz. O

using

only

M,

singled

and S.
out

446a1

orwme` nyq

CcMimvZa :

orwme` nou EY : orw`menon cett.) and 450a13 to prove that M has sometimes preserved a reading which is plainly right
reading

orwme` nyq

28

. But, as seen, the

in 446a1 is shared by other mss. Ross also claims

the support of Alexander (In Sens. 116.21-22 :

y ke` gjroq ai shyty).


orwme` nou).

Mugnier is the only recent editor to print another reading (


But

the

correction

(or

mistake)

orwme` nyq

would

not

be

beyond

scribes, and the alternative readings are not obviously inferior. Thus,
the passage cannot prove the independence of C cMi.
This leaves only 450a13, a passage in which the reading of C cMi is

tou noytikou), Mugnier (tou noytikou) and


tou nooume` nou ?) ; Forster and Ross adopt it. This is a plausible reading, and it would explain the corrupt tou nooume` nou. The
reading of P, tou noytikou, can be explained as an attempt to make
` tou ai shytikou in the following
the passage correspond with tou prw
clause. The corrupt reading of E (touto noomen) is explained as a scrinot accepted by Biehl (

29

Siwek (

bal corruption. Thus, I find the reading of CcMi to be the best transmitted ms. reading.
However, one of the most peculiar characteristics of these three mss.
when they differ from the rest of the tradition is the tendency to struc turalize the text using singular words such as particles, conjunctions,
etc., and therefore this would be a natural emendation (see further
below) : Not only did the scribe give the meaning required by com mon-sense to the passage

30

, but he also created a correspondence

between this and the following clause, using

me n ... de .

The result of this investigation of CcMi is a single correct reading


against the rest of the textual tradition with some uncertainty as to the
c

origin of the reading. This suggests that C Mi do indeed constitute a


group of mss. that are directly descended from E, but contamination

(27) A similar confusion has occurred in 450a21 (

e pahe P : e mahe cett.).

(28) Ross, op. cit. (n. 1), p. 67.


(29) Siwek prints

tou nooume` nou,

but in a note (Siwek, op. cit. (n. 1), p. 154,

note 26) he, apparently, explains that he reads

nou me n (mistake for tou nou me n ?)

with Ross. He translates ad intellectum pertinere .


(30) He may have been inspired by the commentators, cf. Mich. Ephes., In

ei de touto, fanero`n, oti y mny`my tou nooume` nou, toute` sti tou
noytou, kata sumbebykoq an ei y, kah' auto de tou prw`tou ai shytikou, toute` sti
tyq prw`tyq kai koinyq ai shy`sewq. ei d' yn y mny`my pa`hoq tou nou, yn an kah' auto
kai twn noytwn.
Mem. 13.1-5 :

david bloch

12
from the

b tradition must then have occurred at some point in the tra-

dition.
c

As regards peculiar readings, C Mi disagree with the rest of

a in the

following 68 passages :
(Sens.)

436a21,

443b23,
448b5,

443b25,
448b11,

450a13,

450a21,

436b5,

437a27,

445a12,

438b27,

445b26,

449a13,

449a24,

450a22,

438b29,

446a12,
(Mem.)

450a25,

439b31,

446b3,
449b4,

450a26,

440b8,

447a15,

447b11,

449b21,

449b29,

450a28,

450a31,

450b8,

450b9, 450b22, 450b31, 451a4, 451a5, 451a7, 451a10, 451a17, 451b2,


451b4, 451b17, 451b18, 451b30, 452a2, 452a3, 452a9, 452a16, 452a17,
452a18,

452a23,

452b21,

452b22,

452a29,

452b2,

452b24,

452b4,

452b27,

452b13,

453a1,

452b16,

452b19,

453a3,

453a17,

453a2,

453a20, 453a23, 453a29, 453b4.

Most of these peculiar readings are clearly wrong and seem to be the
result of scribal attempts earlier in the tradition to clarify the text or
the structure of the text. Thus, a single, explanative word has often
been added

31

. A few peculiar readings seem to have originated unin -

tentionally in the mind of the scribe (e.g. 438b27


450a31

34

32

, 449b4

33

, 450a22,

). And sometimes the errors are just plain mistakes and easily

understood when applying the general principles of textual criticism

35

It is clear, however, that some of the specific readings of this group


cannot have been derived exclusively from E since they show manifest
affinities with

36

. There is also a number of passages, in which these

mss. share the reading with


tain another reading

37

b or part of the b family, while EYV conc

. Since it has already been determined that C ,

(31) For examples, cf. 436b5, 437a27, 438b29, 440b8, 443b23, 443b25, 446a12,
446b3, 447a15, 448b5, 449b21, 450a25, 450a28, 450b9, 450b22, 450b31, 451a4,
451a5, 451a17, 451b17, 451b18, 452a3, 452a16, 452a18, 452b13, 453a3.

hermon C Mi : hermy cett. The scribe may have expected that the subject
to ai shyty`rion of the preceding sentence, or, less likely, he unintentionally focused on the important word tou vujrou, producing the right gender but
c

(32)

would still be

the wrong case. It must, however, be admitted that the reading is not implausible.
(33) The scribe, thinking that Aristotle is about to treat two subjects, wrote

ti` naq ai ti` aq instead of ti` n' ai ti` an.

(34) Perhaps the scribe thought that we still need an object of


since

oion tu`pon tina could be considered strictly parenthetical

(35) Cf.

e.g.

439b31,

445b26,

448b11,

449a13,

449a24,

e nsymai` netai,

449b29,

450a21,

450a26, 450b8 (a scholion mistakenly incorporated), 451a7, 451b2, 451b4, 451b30,


452a2, 452a9, 452a17, 452a23 (twice), 452a29 (a scholion, originally probably

yt(oi) duna`mei, mistakenly incorporated instead of e sti` n), 452b2, 452b16, 452b20,
452b24, 453a17, 453a20, 453b4.

autoq e pei C Mi : auto`se p b : auto`q E), 452b27 (ama


alla b : om. E), 453a23 (e nojlei C Mi : e nojlountai b : e nupa`rjei E), 453a29
(ge` nytai` ti autw
n sfo`dra C Mi : ti autwn ge` nytai sfo`dra b : ge` nytai` tiq fo`ra E).
c

(36) Cf. e.g. 452b4 (

C Mi :

13

aristotle's de sensu and de memoria


M and i stem from E, contamination from a ms. of the

tradition

must have occurred earlier in the tradition.


As regards the source(s) of this contamination, there are some differ ences between the De sensu and the De memoria. For the De sensu,
I believe that the source must be the exemplar of the two texts in m, the
c

only ms. with which C Mi consistently agree in this text ; and in quite a
c

few passages it is the only ms. that agrees with C Mi


c

passages in which C Mi agree with only v or m

38

. This explains

39

, because the textual

tradition of v has been contaminated from the same source, that is, the
ancestor of m. Thus, the contamination worked both ways. Still, the
c

contamination found in C Mi is to be considered the result of a minor


process of correcting, not a substantial contamination. In the De memoria
CcMi still regularly agree with m, but frequently they side with
against m

40

. If the source of m held readings of

m,

we would have

expected more than one, insignificant agreement (452b24) of these mss.


against the rest of the tradition. Therefore, I would suggest that

m was

used by a scribe in the tradition of C Mi to correct his copy of the De


memoria. Again the process of correction did not substantially alter the
c

C Mi tradition.
Some signs that the CcMi branch arose from a very early copy of E
can also be detected, since these mss. sometimes give the reading of E 1
against E

and the rest of the tradition

41

. However, the evidence is not

sufficient to prove this.


c

Thus, in these stemmatic conclusions concerning C Mi I generally


agree with Escobar, although I differ on some minor points, and
I reject Drossaart Lulofs' view that assigns M to neither of the two fami -

(37) Cf. e.g. 436a6, 444a27, 445a12, 445b8, 445b19, 446b7, 446b16, 447a17,
447a21, 447b21, 448a12, 448b28, 449a6, 449a8, 449b10, 449b15, 449b17, 449b26,
449b28, 450a2, 450a8, 450a13, 450a19, 450b8, 450b13, 450b22, 450b25, 450b28,
451a7, 451a25, 451b2, 451b4, 451b6, 451b11, 451b16, 451b22, 451b27, 451b31,
452a27, 452b4, 452b8, 452b14, 452b15, 452b21, 452b22, 452b30, 453a22, 453b9.

oti d'
ai sha`noit' C Mim : ai shoit' cett.), 447b28 (y auty y auty
y auty varia cett.), 449a20 (to auto C Mim : tw autw cett.).
C Mim : y auty
` menon cett.), 448b9 (to
(39) E.g. 446a1 ( orwme` nyq C Mim v : orwme` nou EY : orw
C Miv : tw
cett.).
c

(38) For agreements of C Mi and m against the other mss., cf. 443b6 (

C Mim : varia cett.), 447b6 (


c

(40) For some examples, cf. 449b5, 451a8, 451b11. This fact agrees perfectly
c

with Escobar, op. cit. (n. 3), p. 205, who thinks that the contamination in C Mi in
the De insomniis originated from a ms. of the

m tradition.

` rxei
(41) E.g. in 451a5 (ei E YbVP : om. E C Mi), 451a29 ( upa
upa`rjei cett.), 451b14 (hatton E C MiV : mallon cett.).
3

E C Mi :

david bloch

14

lies, and the views of Forster and Siwek 42, who assign respectively M
and CcMi to the a family, but claim their independence 43.
The internal relationship of these three mss. is more easily decided.
Cc contains two peculiar errors : 440a1, 448b24. M contains 60 peculiar errors :
436a6, 436a13, 438b8, 436b18, 437a1, 437a10, 437a16, 437a28,
437b21, 438a4, 438a10, 438a11, 438a20, 438b8, 439a8, 439b7, 439b8,
439b13, 439b24, 440a6, 440a20, 441a4, 441b4, 441b8, 441b20, 441b24,
441b26, 442a14, 442b8, 442b23, 443a28, 443a30, 443b13, 443b27,
444a21, 444b14, 445a3, 445a11, 446a1, 446a6, 446a17, 446b1, 446b17,
449b29, 447a13, 447a14, 447b17, 447b19, 448a29, 448b21, 450a17,
450a23, 450a26, 450b11, 450b19, 451a17, 451b31, 452a16, 452b17,
453b5.

i contains 28 peculiar errors :


436b1, 436b12, 437b7, 437b13, 437b25, 438b19, 439b28, 441a15,
441b8, 442a21, 442b8, 445a2, 445a24, 445b12, 447a9, 447a23, 448b6,
448b17, 448b28, 449a28, 449b18, 450b5, 450b11, 452a23, 452a25,
452b7, 452b8, 453a25.

Considering first the relationship between M and i, the number of


clear corruptions not found in i makes it impossible that i originates
directly from M. Thus, the ms. i almost never shares the peculiar errors
of M against Cc. On the other hand, the 28 errors in i not found in M or
Cc make it equally impossible that M stems directly from i. The only
disturbing passage is 446b19 where both M and i have originally omit ted part of the text ; the lacuna has afterwards been confusingly filled in i.
Considering the number of separative differences between the two mss.,
the best explanation is, however, that they both stem independently
from the same exemplar in which this omission had occurred, or in
which a difficult text was found at this point in the ms.

Escobar
ix Fo rster
Siwek
xx Ross

Drossaart Lulofs
x Siwek

(42)
, op. cit. (n. 3), p. 103-109, 205) ;
, Aristotelis De
... (n. 3), p.
;
, op. cit. (n. 1), p. ;
, Les manuscrits... (n. 3),
p. 146 ;
, op. cit. (n. 1), p.
.
, op. cit. (n. 1), p. 61-68, does not state
clearly his position on M, but since he sometimes makes use of M as sole carrier of
the correct reading, he would probably agree with Fo rster and Siwek.
(43) A minor peculiarity : CcMi alone share the reading in 19 passages of the De
sensu, but the same phenomenon applies to no less than 49 passages of the De memoria. Since the De sensu is approximately three times longer than the De memoria this
seems strange. I have no certain explanation, but perhaps the phenomenon should
be considered in relation to the fact that the contaminating sources of the De sensu
and the De memoria were different.
Somno

15

aristotle's de sensu and de memoria


c

As regards the relationship between C and M, the case is equally simple. As shown above, M regularly has individual errors differentiating it
from the rest of

a, while this is almost never the case with C

44

. Thus, the

two errors found in C could easily have been corrected by most scribes.
Since both mss. are descended from E, and since C c adheres much more
c

closely to this ms. than M does, it is obvious that M stems from C .


The close connection between the two mss. can be further estab lished through their respective scholia that are almost completely iden 45

tical. Some of these scholia are of considerable length


lion referring to 439a31 (cf.

supra

n. 26) shows that C

. But the scho-

has kept some of

the E scholia that M does not have. A few of the other peculiarities
establish that Cc is in every respect closer to E than M is. Cc always

duein with E, while M writes duoin with most of the tradition.


osmw`terai, which has been corretains E's corrupt a
osmo`terai in M and in i. Similarly in 447a18, C has kekrected into a
ramme` nou with E, corrected into kekrame` nou in M and in i. Interestingly all four mss. have kekramme` nou in 447a19 .
retains

In 443a19 C

46

The relationship between C

and i is similar to the relationship of C

and M. The many peculiar errors in i contra the few in C

clearly indi-

cate the relationship, but i sometimes shares a reading with part or all
of the tradition against M and Cc

47

. The reason for these discrepancies

is certainly that i is a contaminated and corrected ms.

48

Thus, when it

differs from M and C , and the reason is not merely a scribal error, it

almost always adheres to the


related to m

(44) Siwek,

49

family, and here it is, apparently,

. However, the evidence for this particular affiliation

Les manuscrits

... (n. 3), p. 102, note 23, based his argument on 11

passages, in which he claims that C

differs from M and the rest of the mss., and

which M could not have corrected. I have examined the three readings mentioned
from the

De sensu

and the

De memoria

(448a30, 449b4, 452b7), and in all passages


c

Siwek is wrong as to the disagreement between M and C .


(45) Ignoring the singular glosses written between the lines (even though, as far
as I can tell, they are also all found in both mss.), I have counted 55 scholia in both
c

M and C . Of these 51 are identical in the two mss. The remaining four in M are
written by another scribe and therefore not interesting, while the remaining four in
C

are not found in M.


(46) Mugnier,

La filiation

... (n. 3), p. 40 -41, thought that C

(= his

d)

stems

directly from M, but the argument presented here clearly shows that this is wrong ;
especially since Mugnier would agree that they are both descended from E.
(47) E.g.

437b32,

438a2,

440a23,

442b6,

444b31,

445b8,

445b10,

445b27,

446a7, 447a7.
(48) Physical evidence of contamination can be found e.g. in 437b30, 438a1,
438a18, 442b6, 444a17, 445b8, 445b10, 446a1.

(49) Cf. 441b26 (oude n autw


n rim : varia cett.), 453a23 (tu`joi im : an tu`joi
tu`j cett.). However, in the latter passage both i and m show signs of correction.

v:

david bloch

16

with m is not conclusive. No other existing ms. has such a blend of tex tual features, and contamination must be the explanation. Moreover,
M clearly stems from Cc, and a relationship between M and i is established by the aforementioned omission in 446b19. Therefore, as part of
this tradition, i can only contain readings from b as a result of contamination.
Thus, it can be stated with a fair amount of certainty that CcMi all
stem from E, but some contamination has occurred in the textual
transmission. M and i are independent of each other, but they both
stem from an exemplar now lost, which was copied from Cc.
Some scholars disagree. Thus, Siwek claimed that M is the ancestor
of i (even if most conspicuously seen in the biological treatises of the
Parva naturalia), and that Cc is a better ms. than M, but that M is
nevertheless independent of Cc 50. I believe that it has been shown
above that both claims are manifestly false.

II ^ T

he b family

L, U, S and P were used by Bekker and by all subsequent editors of


the De sensu and the De memoria. Forster added X, W and V (the last
for the passage 436a1-442b26 only). The general good quality of Ha
and Od was recognized by Siwek 51, who also examined y, N, m, Ca, v
and Za.
1. Group

Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 253 (L) ;


214 (Ha) ; Milan, Bi-

Venice, Biblioteca Nazionale Marciana, Marc. gr. Z


blioteca Ambrosiana, Ambr.
nale, gr.

435, H 50 Sup (X) ;

2034 (y).

Paris, Bibliotheque natio-

These four mss. generally agree with each other, and in 56 passages

r presents peculiar readings against the rest of the tradition :

(Sens.) 436a8, 436a10, 436b7, 436b9, 436b12, 436b14, 436b15, 437a6,


437a9, 437a11, 437a16, 437a24, 437b21, 438a7, 438a23, 438a29, 438b9,
439a3, 439a12, 439a17, 439b8, 439b20, 439b21, 439b23, 440a2, 440a6,
440a17, 440a20, 440a28, 440a30, 440b3, 440b6, 440b12, 440b17,
440b20, 440b21, 440b23, 441a17, 441a24, 441b2, 441b6, 441b13,
441b14, 442a15, 442a21, 443b18, 447b5, 448a7, 449a6, 449a24, ( Mem.)
450b14 (v.l. m), 450b17, 451a15, 451b18, 452b11 ( v.l. m), 453a3 (v.l. m).
(50)
(51)
and Od.

Siwek,
Nussbaum,

Les manuscrits

... (n. 3), p. 97-98, 100-102.


... (n. 6), also made good use of H a

Aristotle's De Motu Animalium

aristotle's de sensu and de memoria

17

Most of these passages constitute omissions, words inserted, different


word-order and trivializing of the text, but at least some of the variant
readings are interesting 52. That r has far more peculiar readings in the
first part of the De sensu than in the second part and in the De memoria
is explained by the fact that other ms. traditions tend to use this
branch more extensively as a contaminating source from approximately 442a/b.
In 26 passages (Sens. 22 ; Mem. 4) L has a peculiar error against the
rest of the tradition :
(Sens.) 436a18, 436b16, 436b21, 437b20, 437b24, 438b9, 438b21,
439a6, 439a23, 439a24, 440a14, 440b15, 442a9, 442b8 -9, 443a20,
444b29, 445a3, 447a12, 448a30, 448b14, 448b15, 449a4, ( Mem.) 451a4,
451b12, 452a3, 452b8.

In 27 passages (Sens. 19 ;
rest of the tradition :

Mem.

8) Ha has a peculiar error against the

(Sens) 436a1, 437a4, 438a19 (corr.), 438a21, 439a7, 440a13 ( corr.),


441a3, 441a15, 441a25, 442b27 53, 444a6, 444b9, 444b28, 444b31,
445a21, 445b16, 446a30, 447a29, 447b19, ( Mem.) 449b4, 449b23,
450a24, 451a12, 451a30, 451b12, 451b30, 452a25.

In 35 passages (Sens. 18 ;
the rest of the tradition :

Mem.

17) X has a peculiar error against

(Sens.) 436b12, 437a21, 437b16, 438a1, 438a15, 439b20, 439b24, 440a2,


443a26, 444b9, 445a15, 445a30, 445b25, 446a27, 447b24, 448b11,
449a3, 449a14-7 (omission), (Mem.) 449b14, 450a5, 450a7, 450a27,
451a1, 451a9, 451a25, 452a6, 452a12, 452a18, 452a22, 452a23, 452b27,
453a11, 453a27, 453a31, 453b7.

In 42 passages y has a peculiar error against the rest of the tradition 54 :


(Sens.) 436a10, 437a2 (also as a variant reading in v), 437a7, 437a10,
437b7, 438a2, 438a4, 438a13, 438b19, 439a6, 439a7, 439a29, 439b26,
439b31, 439b31, 440a1, 440b12, 440b15, 441a13, 441a27, 441b18,

(52) Omissions : e.g. 436a8, 436a10, 436b7, 439a17, 440b6, 441a24. Words
inserted : e.g. 436b12, 437a9, 439b20, 443b18, 448a7. Word -order : e.g. 436b15,
437a6, 440a6, 440b12, 441b2, 447b5. Trivializing : e.g. 438a23, 439a12, 439b23,
440b21, 449a24. For an interesting reading, cf. 438b9 ( sta`sin r : om. cett.).
(53) This reading is the result of incorporating a subtitle also found in other
mss.
(54) Siwek , Les manuscrits ... (n. 3), p. 42, has counted only 25 peculiar errors in
y, but since he mentions just a few passages, I have no way of comparing our results.

david bloch

18
441b21,

442a2,

445a16,

445a28,

442b3,

443a16,

445b21,

443b15,

446b29,

444b12,

447a17,

447a20,

444b23,

445a11,

447b5,

447b22,

448a15, 448b15, 449a2, 449a11, 449a28.

Even though LHaX seem to be almost statistically equal, it should


be noticed that LHa often agree in reading against Xy. The mss. Xy
do not agree equally often, but they do share some readings not in
a

LH , and furthermore X often agrees in error with m. This is espe cially true for the

De sensu

, and it explains the low proportion of pecu-

liar errors noted in the statistics compared to the statistics of the

memoria

De

. Therefore, the numbers of X are actually somewhat higher

than it is possible to discern from the statistics above

55

. The numbers

of LH y are reliable guides, as regards the value of the mss.

L and H
Their

are the most reliable witnesses as to the readings of

peculiar

readings

easily be corrected

56

are

often

obvious

scribal

errors

that

r.

can

. This is particularly true of L which is therefore


a

slightly to be preferred over H . Thus, L never inserts text not found


in

other
a

57

mss.,

while

this

sometimes,

although

rarely,

happens

in

. All mistakes committed in L are, apparently, omissions and

mistakes in copying.
L and Ha have not suffered much from contamination, although
there is some evidence that Alexander's commentary may have been
present earlier in the tradition (cf. below). In the few passages where
they contain variant readings, as opposed to corruptions, the variants
are presented as such, either with
ing written above the line
mss.

59

58

gra`fetai

or with the variant read-

. A few corrections have been made in both

Even when they disagree, it is often obvious why they do so. In

particular, it seems that


439a20.
445a2.

r contained some abbreviations :

cett

parou' H X : parousi` a
.
dusw`douq L : dusw`d' Ha : duswdi` aq
a

cett

(55) For some examples of errors in X shared by m (and/or other inferior mss.),
cf. 442b2, 443b12, 447a24, 447b3, 448b2, 448b10. In a few passages, X and other
mss. contain the same errors, probably through contamination, since they are not
a

likely to have made them independently, e.g. 444b10 (with NC ), 444b20 (with

corr
koiny cett
1

S [

. S ]WNC ), 446b8 (with WN).

cett
cett

fhartika
.), 439a23 (kainy L :
a
apatwntai
.). H : 441a15 (lua
(jlisjro`tyta H : glisjro`tyta
.), 449b23

(56) A few examples. L : 436b21 (fartika L :


.), and 442b8-9 (apaitw
ntai L (

cett
cett
oion ai shytwn
Ton auton cett

ke`wn Ha : gluke`wn
a
(enno`ysen H : eno`ysen

.), 441a25

):

.).

(57) Cf. e.g. 439a7 (

auton Ha :

bis

H :

oion

cett

.), 442b27 (Peri

.).

(58) L : e.g. in 449a2. H : e.g. in 453a2, 453a15.


a

cett

(59) L : e.g. in. 448b2, 448b16. H : e.g. in 438a19, 440a13.

osmwn ton

19

aristotle's de sensu and de memoria

afiknoun' H : afiknountai cett.


sullo` H X : sullogismo`q cett.
a

446a20.

453a10.

The plain scribal errors having been dismissed as irrelevant, the two
mss. regularly have the same readings, and in 14 passages (Sens. 10 ;
Mem. 4) they share a reading against all the other textual witnesses

60

Almost all these peculiar readings are unacceptable. The passages con cerned with word-order may be considered by an editor, but since the
rest of the textual tradition stands united, they should probably be dis missed. Thus, the only interesting passage is 448a17 (
a

LH y(v.l.) :

to leukon cett.). The reading of LH

tou leukou

is the reading wanted

by Alexander (In Sens. 145.21-23), but which he did not find in his ms.
I think it likely that an Alexandrian gloss, present in the exemplar of
a

LH , has intruded into their text.

The scribe who made X, even though he is copying in a nice and


readable hand, has been a little less careful than the scribes of L and
a

H .

He

has

made

some

simple

troubled by the similar looking

mistakes
and

eu

particular

he

was

^ and a few characteristic

mental errors, but the remaining mistakes are of the kind that all copy ists will make (different word-order, omissions, etc.)

61

. The scribe (or

his source) does not seem to tamper deliberately with the text : there
are only a few examples of inserted words, and they could well be
explained as mental errors, and, apart from these examples, he does
not often trivialize the text either

62

. Thus, X is generally a good copy

of the text.
X also seems to be relatively free from contamination. It does share
a few readings with

against the rest of

b,

but they are too few and

insignificant to suggest anything but independent corruption

63

. Some

influence from Alexander's commentary can, however, be discerned


(cf. below).

(60) (Sens.) 437a7, 438a1, 448b16 (om.

jro`noq,

sed corr. L), 440b12, 441b20,

442b29, 444a16, 445b30, 448a17 (v.l. y), 448b2, (Mem.) 450a19, 450b29, 451a26,
453a27.
(61)
453b7.
445a15,

a/ eu confusion : e.g. 437a21, 438a15. Simple mistakes : e.g. 438a1, 443a26,


Mental

errors :

453a11,

e.g.

453a31.

445a30,

450a27.

Omissions :

e.g.

Word -order :

440a2,

453a11.

446a27,

e.g.

437b16,

449a14 -17,

444b9,
452a18,

r cett.) and 452a12 (labe` shai e x arjyq X :


(62) Cf. 447b24 (me n gar X : ga
labe` shai arjyn Z : labe` shai arjyq cett.).
` matoq EMYC X : jrw`mata cett.), 440a1 (alourgoun
(63) Cf. e.g. 439b32 (jrw
EMYiX : alourgoun C : alourgon cett.), 449b28 (meta jro`non aX : meta jro`nou
cett.).
cett.), 450b28 ( y aX :
a

david bloch

20

X has been corrected by another scribe (X ) dating from approximately the same time as the original scribe. The sources of X

cult to determine, but, an ancestor of m is a likely candidate


is a chronological problem if, as Harlfinger suggests
dated

at the time

65

are diffi64

. There

, this source is

of Nikephoros Gregoras (1290/1 -1359-60), and

therefore the suggestion that contamination arose from a ms. of this


textual tradition is only tentative. It may, however, be strengthened
by the fact that Alexander of Aphrodisias' commentary also seems to
have played a part in the contamination

66

, suggesting perhaps that the

text and the commentary were present side by side. m is the only
known early ms. in which this is the case.
The ms. X is closely related to L and H a. It is, however, somewhat
closer to y than to L and Ha. Thus, it never shares the reading of L
and H

against y, except in the passages where y has made obvious,

peculiar mistakes. Usually, in these cases, y stands against a united


textual tradition. In addition, as stated above, X shares a number of
readings and scholia with m.
Xy only share a reading against the rest of the tradition in two pas sages
a

67

. However, Xy not only share readings in all the passages where

have individual readings, but they also share a few significant

LH

readings with one of the m versions (and sometimes with v and Z a)


A few more passages confirm the relationship between X and y

69

68

Of the four mss. y has by far the worst text. The number of errors
exceeds reasonable limits, and interesting readings not found also in
the other

mss. are few. The scribe has made a number of very

obvious mistakes, often producing sheer nonsense or just non -Greek

(64) Suggested e.g. by 448b1, 448b10.


(65)

Harlfinger

, op. cit. (n. 3), p. 57.

(66) Cf. 436b9, 436b17.

e k tou Xy : e k cett.), 446b29 (fhoraq X(corr.)y : foraq cett.).


e kaston Xyvm : e ka`sty cett.), 444a8 (einai Xym : einai i dion
VNZ : i dion einai cett.), 448b2 ( ora
y oti tou`tou ti Xym : y oti tou`tou ti a : ti`
tou`tou x : om. LH ).
(69) In 439a21 X has erased kei` shw after sko`toq, and in y the scribe has written
above the line : lei` pw to kei` shw. I have found this confusion in no other ms. written
before the 15th century. The erroneous kei` shw has its origin in Alexander of Aphro disias (In Sens. 43.10-11) : ei y d' an ako`louhon to `kei` shw touto kai mene` tw'... In
441b17 X has incorporated the Alexander -scholion (In Sens. 74.8-9) gew
deq
e napoplu`nei tw ugrw . y has the reading above the line preceded by lei` pw. How(67) 437b13 (

(68) E.g. 442b2 (

ever, v and Z
agree with y.

have also incorporated the passage, while U and both versions of m

aristotle's de sensu and de memoria

21

words 70. At least some of these errors suggest that he misread his exemplar. This may also be the explanation of a number of the remaining
errors, but the scribe seems also to have tried his hand at correcting the
ms. in a number of passages. The results are never very fortunate. With
a single exception, they are, at best, trivializing 71.
Thus, the only variant reading of any interest is the following :
438a2 :

fe` ggoq y : be` nhoq cett.

This reading is textually interesting, although no editor will accept


it. I have not found this suggestion in any other ms., neither in the text
nor as a scholion. I conclude that it constitutes a conjecture made by a
Byzantine scholar.
449a11 (ai shyty`rion kai organon y : e stin cett.) is also interesting,
not so much as a textual variant but rather as an error. The comment
must have been meant to explain e n ti immediately preceding e stin,
but somehow it displaced e stin instead. This proves at the very least
that y was copied from a source now lost since the variant reading is
not found in the text or as a scholion in any of the extant mss. written
before the 15th century. Finally, in 436a17 y is, apparently, the only
ms. to present all three variant readings considered by Alexander of
Aphrodisias (hreptikou, geustikou, geustikou hreptikou). One is
given in the text, two written above the line with gra`fetai. The scribe
may, of course, have excerpted the readings directly from the commentary. The same may be true of the variant readings in 438a19
(gra`fetai vujro`tyti) and in 439a28 (gra`fetai an ti), but it is not
unlikely that these readings were present already in the exemplar as
variant readings since the two related versions of m agree with y (and
some other mss.) in these passages (cf. below). The variant readings in
436a17 and in 446b27 may also have been in the exemplar. Similarly,
in 446b27 P and y (in the latter as a variant reading : gra`fetai ...)
and Alexander in the lemma are the only mss. to preserve e neinai
which is probably the right reading.
y has been corrected, but it was apparently done rather casually,
and no new mss. seem to have been used in the process.

(70) For the most obvious examples, cf. 437a7 ( jrw`mamatoq y : jrw`matoq cett.),
438a4 (apori` aiq y : aporroai q hmamb : aporroi` aiq cett.), 438a13 (alyhyq y : alyhe q
a
cett.), 439a7 (kai kai y : vo`fou kai vZ : kai vo`fou kai cett.), 444b12 (markw
ny :
makrwn mV2LHaXN : {mi/ma} krwn mamb : om. Za : mikrwn cett.), 444b23 (ptwma`
ti y : pwma` ti cett.).
(71) Cf. e.g. 437a10 (dy y : de cett.), 439a29 (sumbebyko`twn y : sumbaino`ntwn
` lista pe` fuke gi` gneshai y : ma`lista gi` gneshai LHa : ma`lista
cett.), 440b12-13 ( ma
mi` gnushai X : mi` gnushai ma : gi` gneshai cett.), 446b29 (fhorai y : forai cett.).

david bloch

22

Defining the internal relationship of the


seems certain that L and H

mss. more precisely, it

were either copied from the same source,

or that one of the mss. was copied from the other. However, even
though very few of the errors in L and H a could be called separative,
Byzantine scribes could not have emended passages such as 448b15
and 451b12 in L, and probably not passages such as 439a7 and 450a24
a

in H , without the use of a third ms. But correction with the use of
another ms. would have left more traces in the relevant ms., and it
should also have eliminated more of the blatant errors that are still
present in both mss. Therefore, L and H

are independent mss. both

originating from the same lost ms. Since there is no extensive contami nation in these mss., and because of the great similarity between them
as well as the easily determinable peculiar errors, the readings of this
sub-archetypus can almost always be determined. X does not originate
from the same ms. but from a closely related one. In these conclusions
I agree with Escobar

72

The more difficult question is the position of y. It certainly does not


stem, as Siwek suggested, from L

73

; it cannot stem from H

either, and

it also seems unlikely that X was its ancestor. The variants in 436a17
and 449a11 mentioned above seem to prove that the direct origin was
not X. Furthermore, one could point to the large omission of 449a14 17 in X but not in y. On the other hand, y is related to X, and it seems,
then, that they stem from the same original source, although not
directly, as L and H
2. Group

do.

= Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 260 (U) ;


d

Venice, Biblioteca Nazionale Marciana, Marc. gr. Z 209 (O ) ; Florence,


Biblioteca Medicea Laurenziana, Laur. Plut. 81.1 (S) ; Vatican, Biblioteca
Apostolica Vaticana, Vat. gr. 1026 (W) ; Vatican, Biblioteca Apostolica Va ticana, Vat. gr. 266 (V) ; Vatican, Biblioteca Apostolica Vaticana, Vat. gr.
a

258 (N) ; Florence, Biblioteca Medicea Laurenziana, Laur. Plut. 87.4 (C ) ;


Florence, Biblioteca Medicea Laurenziana, Laur. Plut. 87.20 (v) ; Florence,
a

Biblioteca Medicea Laurenziana, Laur. Plut. 87.21 (Z ) ; Paris, Bibliotheque


nationale, gr. 1921 (m)
The

h-family

comprises two sub-branches : one of good quality (

m)

and one very corrupt ( ). Still, the unity of the family is established by

(72)
(73)

Escobar, op. cit. (n. 3), p. 205.


Siwek, Les manuscrits ... (n. 3), p. 42. In support, he cites only 439a6 and

446a30, in which passages he claims that Ly have a peculiar reading against the rest

ai shyty`rion
ai shytyri` wn, and in 446a30 there is nothing separating L and y from

of the tradition. In both cases he is wrong. In 439a6 L does not read


(with y) but

the tradition. Thus, Siwek's claim is unsubstantiated.

23

aristotle's de sensu and de memoria


a number of passages, in which the reading of

is generally easy to

establish, and in which this reading agrees with all or most of the
mss :
436b11.
436b16.

a nagkaion m
a na` gky
feu` gein
fugein
ti`q-udati
ti` ga` r e stin h
kai tou to e stin h tou to de
y
h
to toiou to` n m
toiou to` n
jro` an m
jroia n
jria n
o ri`zetai h wristai
tw m
to
t a o
ta
to m
dia h
dio
a ll o ti h a ma de` tiq a lla` tiq o ti
to gewdeq
gewdeq m
stery` seiq m
ste` rysiq
ou j e n mo` non
ga r to me n m
me n ga r to ga r to me n ga r
y h y kai kai
memigme` nwq
memigme` nw m
hermo n kai hermo n
hermo n o m
a panta
pa` nta
tw u grw to xyro` n
tw xyrw to u gro` n
u gro` tyq tiq
u gro` tyq m
o ti deper
o ti de` per e ti d
e ti d eiper m
ei r o ti d
polla` kiq nosw`deiq eisi`n m
polla` kiq noswd` e sti`n
nosw`dyq polla` kiq e sti`n polla` kiq nosw`dyq e sti`n
dusjeranei h dusjerai`nei
aisha` neshai a o moi`wq aisheshai
aisheshai m
duoin
dusi m
ou j a ma me n
ou j m
e nde` jetai aisheshai i aisheshai e nde` jetai e nde` jetai
aisha` neshai
a faireishai a fry` shw
a fairei`shw m
o ra
o ran m
aisheshai
aisha` neshai AGB aisha` neshai
a

C VN(p.c.)P :

437b15 -21.
438a17.
438b5.
439b2.
439b4.
439b6.
440a7.
440a14.
440a26.
440b13 -4.
441b17.
441b25.
441b26.
441b29.
441b30.
442a2.
442a4.
442b18.
442b28.
443a27.
443b6.

O SWC vZ

cett.

cett.

plerique codd. :
P:

cett.

cett : om.

VNvZ :

cett.

C VvP :
:

C NvZ :

cett.

/ } m

(incertum Z ) :

'

corr. Z )

cett.

cett.(ex

yC vZ : om. Y :

cett.

cett.

P:

cett.(

s.l. m

et m ).

C NP :

cett.(ins. Z )

C VNvP :
d

cett.

O SWC NvZ : varia cett.


a

C VN :

P:

v:

Z :

cett.

cett.

C V vP :

cett.

C Nv :

Z :

SWC NvZ :

cett.

cett.

O SWC VN :

C NZ :
2

m :

cett.

cett.

V C NZ P :

'

444a17.

' C Mivm

EV :

Y:

C NP :

Z :

V:

444b29.
447a27.
447b9.
448a20.
448a25-6.

C vZ :
a

cett.

cett.

C vZ P :

'

N:

cett.

2 cett.

C vZ :

cett.

SW :

U:

cett.

448b5.
448b7.
448b12.

C NZ :

yC Nv :
d

v:

cett.

cett.

O SWC NZ :

P:

cett.

Complete agreement cannot be expected in all cases since there are


many mss. in this tradition and some of them very corrupt, but the
coherence of

h is proved. In particular, the reading in 437b15-21 lays

a solid foundation of the family.

david bloch

24
(a)

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

260 (U)

The ms. U, dating from the 12th or 13th century, is one of the oldest
extant mss. containing the Parva naturalia, but, even though editors
have used it since Bekker, it has always been somewhat underestimated. It is clearly a member of the b family, and it is closer to x than
to r and i, but at times it preserves readings with one of these against
x, and sometimes even with a against the rest of h and r 74.
In the following 29 passages (Sens. 22 ; Mem. 7) U contains a reading
against the rest of the textual tradition :
(Sens.) 436b11, 438a19, 439b13, 440a10, 440a28, 441a29, 441b4,
441b26, 442a25, 443a17, 443b11, 444b6, 445b17, 445b19, 445b20,
446a30, 446b1, 446b4, 446b21, 447a25, 447b22, 448a25, ( Mem.)
450a27, 450b22, 451a7, 451a12, 452b2, 453a16, 453a29.

Most of these passages are clearly individual scribal errors 75. The
scribe does not deliberately alter the wording of his exemplar, and, in
general, he has copied the text carefully. The number of individual errors
is acceptable, and some of them may well have been present in the exemplar. Thus, U compares favourably with the best mss. of both r and x.
In passages that are generally confused in the transmitted mss. the
contribution of U may always be considered as a possibly independent
variant reading 76. U is a ms. of some age, and more than any other
major ms. of b it occupies an intermediate position. The readings that
agree with a may, however, have been caused by contamination. The
following statistics provide the agreements of U with the different
groups of mss.
To the 56 peculiar readings of r, presented above, can be added the
following 28 passages (Sens. 14 ; Mem. 14) giving a total of 84 passages
in which U agrees with a and/or x against r :
(Sens.) 436a13, 436a17, 436b1, 437b7, 441a30, 441b27, 442a13, 442a18,
443a26, 443b17, 444a33, 445a28, 445b5, 447b24, ( Mem.) 449b9,
450a12, 450b14, 450b17, 451b5, 451b8, 451b10, 451b15, 452a1, 452b8,
452b11, 453a11, 453a13, 453b7 77.

Escobar

(74)
, op. cit. (n. 3), p. 118-123, has noticed the peculiar status of U,
and he believes that its exemplar must have been contaminated.
(75) E.g. 441a28-29 (jumoi U : jumoi pa`nteq cett.), 443b11 (ekei na amaP : ekei nai U : om. W : ekei noi cett.), 445b17 (wn U : om. ambP : onta cett.), 445b19 (adu`naton
CcMiS2vZaP : adu`natai U : adu`nt{a/on} ma : adu`nata cett.), 446b21 (an an U : om. a :
an cett.), 452b2 (mi U : om. a : my cett.).
(76) E.g. in 438a19 and 439b13.
(77) Omission : e.g. 443a26, 445a28, 450b17. Words inserted : e.g. 441a30,
453a13. Word-order : e.g. 442a13, 442a18, 444a33, 445b5, 451b8, 451b15, 453a11.

aristotle's de sensu and de memoria

25

To the 24 peculiar readings of x that will be presented below, can be


added the following 8 passages (
. 7;
. 1) giving a total of 32
passages in which U agrees with a and/or r against x :

Sens Mem

Sens

(
.) 437b1, 438a4, 439a20, 440a10, 441a18, 442a25, 446a9, (
453a26 78.

Mem

.)

The agreements of U and the rest of b against a are many, but in


the following 23 passages (
. 15 ;
. 8) U agrees with a against
the reading of b. I disregard individual misreadings and the readings
of the problematic mss. m and P.

Sens

441b5.
442b17.
443a28.
443b13.
443b21.
444a4.
444b5.
444b16.
445b3.
446b21.
446b28.
448a12.
448a13.
448b2.
449a24.
449b19.
450a23.
450a30.
451a3.
452a22.
452b2.
452b19.
453a5.

Mem

jumo n aU : jumo n kai ta dia twn a llwn o moi`wq b


twn aU : twn a llwn b
koino n aU : koiny b
o aU : om. cett.
pa` hy einai aU : varia cett.
twn a nhrw`pwn aU : a nhrw`pou b
wsper aU : wsper e n cett.
aisha` netai aU : aisha` nontai b
a peira aU : a peiron b
einai aU : yn b
tiq b : om. aU
polu aU : polu n b
e ti a pe` jei aU : a pe` jei e ti b
y o ti tou` tou ti aU : varia cett.
de aU : dy b
e nergeiwn aU : e rgwn b
y aU : om. b
mny` myn einai aU : einai mny` myn b
e ggignome` nwn aU : gignome` nwn b
e xy` tei aU : e pizytei b
ge aU : om. b
pro q ty n AB aU : pro q AB b
e n toiq pro` teron aU : pro` teron b

The statistics substantiate the view that U is a ms. that is clearly


part of b but also contains peculiar but intended readings, as well as
readings shared with the different ms.-groups. The disagreements that
U has with a and the rest of the b mss. respectively are differing only
in quantity, not in quality, some being obvious corruptions, some
being clearly right. Differences in omissions, words inserted, wordorder, and the like are the most common variations. Seeing, then, that
(78) Omission : 438a4, 439a20, 442a25. Words inserted : 437b1, 440a10,
453a26. Word-order : 441a18.

david bloch

26

U has both good and bad variant readings from all traditions, extensive contamination is less likely, and U must be considered an important textual witness. One should not expect many correct readings in
U not found elsewhere in the tradition, but for h it is certainly the single most important ms., and, as I will show below, U does preserve at
least one independent reading of some interest.
As regards the relationship between U and the other mss., there is a
clear relation between U and W (see further below), and generally U
is closer to x than to any other group of mss.; in particular, it may be
noted that it shares a large number of scholia with W (cf. note 98).
Besides the regular readings, the substantial omission in 437b15-21 is
also strong evidence that U stems from h.
From the scholia it can also be discerned that the scribe, like many
other scribes, was influenced by Alexander's commentary. There is a
conspicuous example in 444a15 that will be discussed below in relation
to W, and another example can be found in 441b26 where prosupakou`steto jumoq kai trofy` has been written in the margin. The origin
of this is Alexander (
. 77.25-26) 79.
Two passages constitute particularly strong evidence that U holds
some independent readings. In 441b27 U has a text portion written in
the margin (oude au toiq futoiq), referring to toiq zw` oiq. Of the other
mss. this text is found only in P, m and N, and, interestingly, the text is
not part of Alexander's commentary.
By far the most interesting passage is, however,
451b1415. The text in U and in the rest of b is:
In Sens

De memoria

sumbai` nei d' e ni` ouq apax e hishynai


[allouq a] polla`kiq kinoume` nouq.

mallon [hatton a] y e te` rouq

Written in the margin of U by the original scribe we find:

gra`fetai sumbai` nei d' e nia apax e hishynai mallon y e tera polla`kiq
kinoume` nouq.

The variant e nia ... e tera is not found in any other ms. It is not
found in Michael of Ephesus either, but his commentary (
.
25.23-29) shows that the Byzantines, like modern scholars, were uncertain as to the precise meaning of this sentence.
Ross, partly following Freudenthal, rightly considered the discrepancy between e nia and e teroi in the following line (451b15-16) too
In Mem

(79) It should perhaps be mentioned that not all the scholia in the ms. are by
the original scribe. Cf. Escobar,
. (n. 3), p. 123. For a description (by J. Wiesner) of another ms. copied by the same scribe, cf. Moraux & Harlfinger &
Reinsch & Wiesner,
. (n. 3), p. 103-104.
op. cit

op. cit

27

aristotle's de sensu and de memoria

great, and since e nioi is impossible, he read e tera with a few mss.
instead of e teroi 80. The resulting reading makes Aristotle contrast the
things remembered instead of the persons remembering, and, as a con sequence of this, e ni` ouq ... e te` rouq (allouq) cannot (pace Siwek 81) be
read in the preceding sentence. The variant reading in U may origi nally have been an attempt at correction along the same lines as Ross,
but while Ross was focusing on the previous part of the text and therefore wanted to read e ni` aq ... e te` raq (referring to kiny`seiq), the originator of the reading in U focused on the following part of the text and
wrote e nia ... e tera.
The most remarkable thing about this is that it is a true variant
reading. This is clear from the use of gra`fetai. Taken with the rest of
the evidence, it proves beyond any doubt that U must generally be
considered by modern editors of the texts.
(b) Group x =
(Od) ; Florence,

Z 209
81.1 (S) ;

Venice, Biblioteca Nazionale Marciana, Marc. gr.


Biblioteca Medicea Laurenziana, Laur. Plut.

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

1026 (W)

S has been corrected (S2), but most often it is still possible to discern
the reading of S1. Od is carefully written but badly damaged at the top
of most pages. Thus, quite often it is not possible to discern its read ings. In the following statistics I disregard the passages in which Od
has suffered physical damage.
The mss. of x regularly agree in their readings, and they are alone in
their reading against the rest of the textual tradition in the following
24 passages (Sens. 16 ; Mem. 8) :
(Sens.) 437a21, 437b32, 438a22 82, 438b23, 438b25, 439b11, 439b14,
439b26, 441b2, 442a13, 442b20-1, 445b30, 446a12, 446a24, 447a3,
447b18, (Mem.) 449b16, 450a21, 450a28, 451a15, 451a19, 451b16,
452a2, 453a27.

In 63 passages (Sens. 42 ;

Mem.

22) S has a peculiar error :

(Sens.) 436b1, 437a6, 437b24, 437b26, 437b32, 438a28, 439a6, 439a29,


440b9, 440b28-9, 441a6, 441a22, 441b5, 441b13, 441b14, 442a11,
442b14, 443a16, 443b5, 443b14, 443b20, 443b29, 444a16, 444a23,
444b2, 444b3, 444b5, 444b18, 444b20, 444b27, 445a3, 445a5, 445b14,
446b2, 446b16, 447a21, 447a27, 447b19, 448a14, 448b2, 448b7,
(80) Ross , op. cit. (n. 1), p. 245.
(81) Cf. Siwek , op. cit. (n. 1), p. 138, 159, note 64. Siwek does not record the
reading of U.
(82) S1 and Od have arrigwto`taton. The reading has been corrected by S 2 and W.

david bloch

28

448b25, (Mem.) 449b13, 450a20, 450b2, 450b22, 450b27, 451a5, 451b4,


451b28, 452a8, 452a16, 452a19, 452a20, 452a23, 452b4, 452b8, 452b9,
452b22, 452b28, 453a3, 453a4, 453a31.

In 60 passages W has a peculiar error :


(Sens.) 436a3, 436a6, 436b15, 436b20, 437b11, 437b13, 438a18, 438a30,
438b15,

439a32,

439b15,

440a29,

440b5,

441a14,

441a26,

441b5,

442a3, 442a6, 442a10, 442a15, 442a17, 442b4, 443a3, 443a7, 443b11,


444a4, 444a7, 444a12, 444a16, 444a21, 444b6, 444b23, 444b30, 445a1,
445a11,

445a17,

445a30,

445b9,

445b15,

445b29,

446a4,

446a12,

446b11, 446b18, 446b30, 447a2 447a8, 447a17, 447b8, 447b16, 447b22,


448a4,

448a14,

448a15

83

448a20,

448b21

449a8,

449a16,

449a19,

449a24.

In 20 passages (Sens. 16 ; Mem. 4) O

has a peculiar error :

(Sens.) 437a29, 437b1, 438b2, 438b26, 439b17, 440a17, 440b30, 442a10


(shared, but committed independently, by N), 446a29, 446b21, 447b29,
448a28, 448b5, 448b12, 449a4, 449b1, (Mem.) 450a6, 452b9, 453a27,
453b1.

Even if only the statistics were used, it is clear that W is by far the
worst of the three mss., S occupies a middle position, and O
superior of both S and W

84

is the

. This is substantiated by the types of error.

The errors in S are often serious mistakes, ruining the text. For
instance, it presents a number of substantial omissions not found in W
and O

85

, and a number of mistakes in reading. Some variant readings

are attempts at correction


of them rather obvious

87

86

. Ordinary scribal errors are plenty, some


2

. A later scribe (S ) has corrected the ms.,

probably in the 13th century, but his sources are difficult to determine.
It was certainly a

b ms., and, as in the case of X (see above), the exem-

plar of m would be a likely candidate were it not for the fact that S

(83)
to gluku kai pikron leukon W : to gluku kai leukon cett. I doubt that pikro`n was in the exemplar of W. Rather, the scribe of W unconsciously produced the
contrary of to gluku`. This reading is meaningless since Aristotle is talking about
su`stoija. The text in W was later corrected.
(84) This agrees with

Nussbaum

, art. cit. (n. 3), p. 123 -125, but

Siwek

, Les manu-

scrits ... (n. 3), p. 65, claims that [l]e manuscrit S est sans aucun doute le plus
important de la famille.
(85) Cf. e.g. 436b1, 438a28, 439a29, 443a16.
(86) Cf. e.g. 437b32 (

ote S : to`t' cett.), where the scribe probably recalled what

he thought was a similar beginning in 437a26, even if he made a plain mistake when
he copied this verse the first time.
(87) Cf. e.g. 441b5 (

alwn S : om. a : allwn cett.).

aristotle's de sensu and de memoria

29

was probably in Italy as early as the beginning of the 14th century 88. If
Harlfinger is right in assuming that the ms. from which m was made
was an edition of the Corpus Aristotelicum connected with Nikephoros
Gregoras 89, it cannot have been the basis of the corrections in S. Gregoras (or whoever edited the works) may, of course, himself have based his
edition on a previous edition unknown to us, but that is mere speculation. The readings of S2 are very diverse, showing no clear connection
with any particular ms. or group of mss. Therefore, the ms. on which S2
based his corrections seems also to have been contaminated, and the
only known textual tradition to present such variety is the one connected with m. I tentatively connect the contamination of S2 with this
tradition. As a textual witness S2 is irrelevant.
In addition to the high percentage of omissions, trivializations (especially by insertion of articles, conjunctions or prepositions) and pure
scribal errors 90, W seems to have suffered further from some contamination 91. Thus, there is a number of passages in which W sides significantly with a, r or some of the remaining mss. against S and Od in
readings that could not have been produced accidentally 92. It is, however, possible that the variant readings were present in the ms. from
which W was copied. This seems especially likely since O dS also contain some agreements with a that are best explained as contaminations. Furthermore, the errors being simple mistakes 93, it seems that,
even though W is a badly corrupted ms., the scribe had no intention of
tampering with the text. Thus, the trivializations were probably
unconscious additions.
The errors committed independently in Od are almost always wrong
transscriptions or minor omissions. The most serious mistake is found
in 442a10, where the scribe has trivialized ydu`smatoq into udatoq.
(88) Cf. Escobar , op. cit. (n. 3), p. 160-161.
(89) Harlfinger , op. cit. (n. 3), p. 57.
(90) Omissions : e.g. 437b11, 438a30, 438b15, 440a29, 442a10, 443b11, 445b29,
446b11, 448a14, 449a19. Trivializing : e.g. 436a3, 436a6, 441b5, 442a17, 443a3,
444a4, 448a4. Scribal error : e.g. 437b13, 439a32, 439b15, 442a3, 442a15, 444b6,
445a1, 447b16, 448a20, 449a16.
(91) For some examples selected from the peculiar readings, cf. 438a18 ( to |dvr
to e kre` on W : to e kre` on udwr OdS et al. : udwr to e kre` on arUP), 440b5 (tauta
mo`nwq W : mo`nwq OdS et al. : tauta mo`non arP).
(92) Cf. e.g. 438a17 (dylon W et cett. : om. iUOdS), 438b12 (ydy ga`r tisi arWP :
tisi gar OdS et cett.), 440a10 (o W et cett. : om. UOdSvZa), 440b5 (tauta mo`non arP :
tauta mo`nwq W : mo`nwq UOdSCaVNvma), 441b1 (outoi upa`rjonteq arWP :
upa`rjonteq outoi UOdSCaVNZamamb : upa`rjousi outoi v), 446a25 (usteron W et
d
cett. : usteroq O S).
(93) Cf. e.g. 436b15 and 447b8 for some strange accentuations, and the scribal/
phonetic error in 447a8 : hy`- for hi` -.

30

david bloch

The latter reading is also found in N, but it does not seem stemmati cally possible that they took it from the same source.
Examining the internal relationship of these three mss., it seems
clear from the passages listed in the statistics above that they are all
independent textual witnesses, even if closely related.
Considering the number of omissions and the other corruptions,
S cannot be the exemplar from which O

(or W) was made, and

I doubt that the scribe of S could have corrected all the passages in
which O

made mistakes, wherefore it is not possible either that O

was the exemplar of S. Some of the mistakes in O


been corrected

94

could indeed have

, but passages such as 442a10 and 446b21

95

need

another ms. to correct, and generally the mistakes made by O d are


rarely noticeable if one does not consult other mss. But S 1 does not
show any signs of working from more than one ms., and since S and
d

also share a number of evident errors

96

, making it improbable that

there were other mss. in between the two, they must have been copied
independently from the same source. This source was probably not the
source from which W was copied, but even ignoring the overwhelming
number of agreements in readings it is also clear from several other
passages that W is closely related to S and Od

97

In addition, several features relate W to the tradition of U. Thus,


the number of identical scholia shared by these two mss. is too high to
be

accidental

98

Furthermore,

is

the

only

ms.

omitting

ou

in

438a30, but there was certainly a similar corruption in U since the


word has been inserted. Considering the stemma that I have drawn, a
connection between U and W is to be expected.

tou to
tou t ou
e narge` steron
ou k a nhi` stantai
e nerge` steron
ou kahi` stantai
a mnymomone` steroi a mnymone` steron a mnymone` steroi
a mnymone` nesteroi
y du` smatoq
a n au tou
u datoq
(94) E.g.

437a29

'

cett.),

cett.), 453a27 (

440b30

cett.), 453b1

V:

v:

cett.).

O N:

(95) 442a10 (

cett.), 446b21 (

(96) Most conspicuously in 438a22, 443b12, 446a9. S

: varia cett.).

has attempted to correct

all of them.
(97) In 436b11 S and W are the only mss. with the wrong reading
although W also has the right

ou

tou` tw

as a variant above the line ; in 437a29 O

tou to

omits

, and it is clear that it was also omitted at first in both S and W. It seems certain

that the exemplar of these three mss. did not have

ou

, but a moment of reflection

made the scribes of S and W insert it ; in 444a2 S and W are the only mss. with the
abbreviation
reading

BG

sunyh
GB

for

' for

sunyhei` a

in 448b5 O

and W agree against all other mss. in

(98) E.g. scholia on 443a31, 446a1, 446b11, 447b6, 447b9, 448a9, 449a12. Cf.
also J. Wiesner, Zu den Scholien der Parva naturalia des Aristoteles, in J. N. Theodora copoulos, ed., Proceedings of the World Congress on Aristotle, Thessaloniki August 7 -14,
1978, vol. I, Athens 1981, 233 -237.

31

aristotle's de sensu and de memoria


d

Siwek thinks that W stems from O , but neither of the two passages
cited by Siwek as evidence for readings shared particularly by W and
d

is to the point. In 446a24 Siwek fails to notice that S has been cor -

rected, and that S1 clearly read

po`teron

with W and Od ; in 448b2 the

reading is also shared by v, and S made a simple mistake :

tou`tou.

tou`toiq

for
d

Otherwise it would have had the same reading as W and O .

However, Siwek has also argued from 444a15 that W stems, directly
or indirectly, from O

99

. The argument is based on a reading taken

from an explanatory note on the Aristotelian

y touto

written by

Alexander of Aphrodisias (In Sens. 98.23-25), occurring with a slight


variation in the text of both W and O

444a15.
y touto riUSm m : om. aP : y touto. to hermai` nein kai diajein touq peri ton egke`falon vujrote`rouq ontaq to`pouq kai ei q to
su`mmetron agein W : y to hermai` nein kai diajein touq peri ton egke`falon vujrote`rouq ontaq to`pouq kai ei q to su`mmetron agein Od.
a

But Siwek's argument can be dismissed for three reasons :


(1) The reading is not difficult to explain as a variant, since it is also
found as a scholion in U, and thus it was present in the tradition. (2) In
d

the gloss has completely replaced

there

100

touto, but in W the word is still


wq sjo`lion ekeito touto, refer-

. (3) In the margin of W we read

ring to this text. Surely this means that the exemplar of W had the text
d

as a scholion, but in O

it is part of the regular text. Since the ms. U

shows that the text was indeed present in a related textual tradition as a
scholion, Siwek's argument fails.

(c) Group

i = Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 266 (V)

(436al-442b26) ; Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 258


a

(N) ; Florence, Biblioteca Medicea Laurenziana, Laur. Plut. 87.4 (C ) ;


Florence, Biblioteca Medicea Laurenziana, Laur. Plut. 87.20 (v) ; Florence,
a

Biblioteca Medicea Laurenziana, Laur. Plut. 87.21 (Z ) ; Paris, Bibliotheque nationale, gr. 1921 (m)

The ms.

i was the ancestor of the most corrupt mss. of the entire tex-

tual tradition. Furthermore, some mss. that do not really belong to


this branch of the tradition, viz. m and P, have been influenced by it.
P will be treated separately, but seeing that most scholars treat m and
v as closely related, I will examine m in the context of the

branch,

even though it turns out that m should be differently categorized.

(99) Siwek, Les manuscrits ... (n. 3), p. 63, 144.


(100) Siwek crucially fails to notice this difference in reading, cf. Siwek, Les manuscrits ... (n. 3), p. 63, and Siwek, op. cit. (n. 1), p. 39 (the critical apparatus).

david bloch

32

Some of the agreements constituting i can be seen in the following list,


and further arguments will be presented below in the sections on the
individual mss.
437a1.
437a28.
437b25.
438b11.
439a26.
441a21.
445a18.

kai a CaVN : kai cett.


orwnta ta orw`mena CaVN : orwnta orw`mena v : orwn me n
anti` a P1 : orwnta m : orwnta orw`meno`n ti cett.
pro`teron ei rytai CaVNv : ei rytai pro`teron cett.
kai 3 cett. : om. i.
me n VNvZa : me n oun cett.
oun CaVN : toi` nun cett.
einai dei i : dei einai cett.

Again, it must be noted that complete agreement among the i mss. is


not often found against all other mss. because of the generally very corrupt and contaminated nature of the former. Still, I believe that the
above list does establish a number of i readings against the rest of the
tradition.
Subgroup l =

258 (N) ;
266 (V) (436a1-442b26)

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

Vatican, Biblioteca Apostolica Vaticana, Vat. gr.

In the following 26 passages of the De sensu 436a1-442b26 the two


mss. have peculiar readings against the rest of the textual tradition :
(Sens.) 436a12, 436b2, 436b5, 437a27, 437b5, 437b13, 437b14, 437b21,
437b24, 437b25, 437b28, 437b29, 437b31, 438a2, 438a16, 438b11,
438b23, 439a22, 439a25, 439a26, 439b29, 440a20, 440b11, 440b19,
440b20, 440b31.

These readings consist primarily of omissions, words inserted, different word-order, and scribal mistakes, proving that the source was a
rather corrupt ms., but the errors do not clearly indicate the particular
cause of corruption 101.
From 442b27 to the end of the De memoria N holds a peculiar reading
in the following 147 passages (Sens. 91 ; Mem. 56), not counting multiple errors in the same Bekker-line :
(Sens.) 443a1, 443a7, 443a9, 443a15, 443a16, 443a19, 443a20, 443a21,
443a24, 443a28, 443b2, 443b6, 443b8, 443b9, 443b11, 443b14, 443b21,
(101) Omissions : e.g. 437b25, 437b29, 438a2, 439a25, 440a20, 440b20. Words
inserted : e.g. 437b21, 438b11, 440b31. Word-order : e.g. 436b5, 437a27, 439a26.
Scribal mistakes : e.g. 436a12 (tugja`nousai VN : tugja`nousi cett.), 437b28 (aiva
VN : av S1 : av OdWZa : avaq cett.), 437b31 (apeire` sin VN : atyre` ssin ma :
atyre` sin mb : ateire` sia Za : ateire` sin cett.).

33

aristotle's de sensu and de memoria


443b22,

443b25,

443b26,

443b28,

443b31,

444a6,

444a7,

444a16,

444a17, 444a18, 444a25, 444a26, 444a28, 444b2, 444b15, 445a2, 445a3,


445a5,

445a6,

445a10,

445a13,

445a14,

445a15,

445a17,

445a22,

445a30, 445b6, 445b11, 445b12, 446a2, 446a3, 446a8, 446a14, 446a30,


446b1,

446b17,

446b24,

446b30,

447a6,

447a12,

447a17,

447a22,

447b3, 447b7, 447b20, 447b22, 447b25, 447b27, 448a5, 448a8, 448a10,


448a13, 448a14,

448a15, 448a16, 448a17,

448a18, 448a21, 448a23,

Mem

448a26, 448a30, 448b2, 448b5, 448b6, 448b7, 448b16, 448b21, 448b30,


449a3, 449a5, 449a12, 449a20, 449a29, 449b2, (
449b19,
450a12,
450b2,
451a13,
451b5,
452a26,

449b20,
450a14,
450b5,

452b8,

450a15,

450b12,

451a16,
451b10,

449b21,

449b25,
450a15,

450b15,

449b30,
450a16,

.) 449b9, 449b18,

450a1,

450a18,

450b16,

450b30,

451a17,

451a18,

451a20,

451a23,

451b12,

451b15,

451b23,

452a2,

452b13,

452b22,

452b23,

452b24,

450a3,

450a6,

450a19,

450b1,

451a4,

turalia

451b4,

452a8,

452a23,

453a13,

453a17,

453a18, 453a23, 453a28, 453a29, 453a31, 453b6.

With the exception of v, Z

451a12,

451a26,

and P no other early ms. of the

Parva na

can produce comparable numbers of peculiar readings, and

almost all blatant errors. Since the proportion of peculiar readings is


significantly higher in comparison with 436a1 -442b26, where V and N
can be compared, it cannot be assumed that the readings of N gener ally reproduce its exemplar. Having no other ms. in this section of the
text (442b27-453a11) from this part of the tradition, the readings of

are difficult to determine.


The errors of N are of a wide variety. The common errors, such as
omissions,
found

102

insertions

of

text

and

word-order,

different

are

often

, but the dominating ones are more unusual. Thus, by far the

most striking errors are mistakes in endings

103

, regular misreadings

104

and some very peculiar readings, making no sense at all. These will be
discussed below.

(102) Omissions : e.g. 443a19, 443b8, 444b2, 447b7, 451a17. Insertions of text :
e.g. 443b21, 445a15, 449a3, 451a20. Word -order : e.g. 445b11, 447a12, 447b27,

cett

451a26.

.),
fusikwq NZ : fusikou
zwyn N : zw on .), 439a1 (tyn kardi` an N : t kardi` a
.), 443a7 (jhr
xyrou
.), 443b2 (kaly N : kalw
q .), 448a16 (kala N : all' wq a : kalw

cett

cett
a

(103) For a few significant examples, cf. 436a17 (

cett

436b11-12 (

cett

N:

.).

cett

cett
cett

he` shai N : he` sei .), 440b27prwton N : pro`teron


.), 442b17 (swma`twn N : sjyma`twn
.), 444a25
(hermainon N : hermy
.), 445a14 (pausiq N : plu`siq
.), 446a2 (diahe` sei N : diai` sei EYV : de` sei Z : die` sei .), 449b9 (pro`teron oun N : prwton me n Z : prwton
me n oun .).

cett
cett
cett

(104) For a few significant examples, cf. 440b8 (

28 (

cett
1

cett

david bloch

34

The following statistics concern the individual readings of V and N


respectively for the De sensu 436a1-442b26.
In the following 41 passages V has a peculiar reading :
436a3, 436a6, 436a14, 436b4, 436b16, 436b19, 436b21, 437a4, 437a21,
437a23, 437a26, 437b12 -13, 438a4, 438a8, 438b3, 438b4, 438b7, 438b8,
438b15,

438b16,

439a33,

439b7,

441a27,

441b13,

438b21,
439b8,

438b23,

439b12,

441b24,

439a5,

439a6,

440b23,

441b26,

439a8,

440b29,

442a29 -b3

439a10,

440b31,
2

V ),

(corr.

441a4,
442b10,

442b23.

In the following 65 passages N has a peculiar reading :


436a6 (sed

gra`fetai

v.l. in marg.), 436a13, 436a17, 436a18, 436a19,

436b7, 436b11, 436b12, 436b18, 437a2, 437a4, 437a21, 437a28, 437a32


437b7,

437b12,

437b30,

437b32,

438a12,

438a13,

438a18,

438a21,

438a25, 438b2, 438b3, 438b4, 438b22, 438b23, 438b25, 439a1, 439a2,


439a9,

439a13,

439b22,
440b8,

439a21,

440a24,
440b10,

441b18,

439a24,

440a25,
440b19,

440b27,

442a5,

442a11,

441b28,

439a25,

440a27,

440a29,

441a13,

439a28,

439b5,

439b7,

440a30,

440b6,

440b7,

441a17,

442a16,

441a22,

441a27,

442b1,

442b14,

442a21,

442b17, 442b20.

Even if we only used statistics it is clear that both mss. are very unre liable guides to the Aristotelian text. Neither copyist has made an
accurate copy of his exemplar. N contains significantly more peculiar
readings than V, but the errors in both mss. are of the type that
I noted above in the remarks on N.
It is clear from the statistics that V is not only independent of N ; it
is also a better ms. On the other hand, the shared errors are so numer ous and conspicuous that the mss. must have been copied from the
same source. Thus, I find my results in complete agreement with Harl finger who made similar claims for the De lineis insecabilibus

105

This is contrary to Siwek, who calls N le plus important de la


`branche' (his familia

d) and thinks it possible, despite the enormous

number of peculiar errors and omissions, that N was the archetypus of


the familia

106

. On V, he is somewhat ambiguous. In his edition he

seems to include V in the textual basis. It is, however, clear from his
stemma that he does not think that it presents independent evidence,
and it is supposed to be indirectly descended from N

(105)
(106)
(107)

Harlfinger
Siwek
Siwek

107

, op. cit. (n. 3), p. 127 -139.

, Les manuscrits ... (n. 3), p. 67 -82.


, Les manuscrits ... (n. 3), p. 145 ;

Siwek

, op. cit. (n. 1), p.

xxi

aristotle's de sensu and de memoria

35

It is striking that two closely related mss. not only agree in so many
errors but also individually contain an incredibly high number of errors
of a kind not shared by any other Aristotelian ms. And it becomes even
more striking when it is considered that the number of individual errors
in V decreases dramatically, as soon as the copyist starts to follow the a
tradition. Thus, compared with the 41 errors in the passage 436a1442b26, V has only 32 peculiar errors in the passage 442b27-453b11 (cf.
the 108description of V above). The explanation is provided by Harlfinger : the copyists of N and V had great difficulties in reading their
exemplar, and in particular they made a number of mistakes in reading
the abbreviations. My statistics support this conclusion.
V has done better than N, but the consistent mistake of the scribe in
reading ga`r for ara betrays him. This phenomenon is found 4 times in
a short passage of the
(438b7, 438b8, 439b7, 439b8), and it is
noted by Harlfinger for the
as a characteristic mistake of this scribe. A further example is the confusion in V between the
endings -iko`q and -to`q (436b21, 439a6), found even more extensively
in N. In addition, there is a number of ordinary misreadings 109 that
are most likely to have been caused by the difficulty in reading the
exemplar, and the many omissions should probably be explained in
the same way.
The errors in N are far more extensive. The following is a list of
some of the most conspicuous:
De sensu

De lineis insecabilibus

448a23: peri Nv: para . & 438b25: para N: peri . 110


438b4: meta N: dia .
439a28: all' oti N: an ti .
450a19: ta N: otan .
449b20 & 450a3: trigw`nou . N 111 112
: Overlooking the abbreviation .
: oude N: oude n . 113 114
: -iko`q/-to`q confusions .
: Wrong ending of a word.
cett

cett

cett

cett

cett

om

Passim
Passim

cett

Passim
Passim

(108)
,
. (n. 3), p. 136-138.
(109) E.g. in 437a4, 437a26, 438a4, 438b15, 439a10, 439a33, 442b10.
(110) In the first example the case of the following pronoun was accordingly
adjusted.
(111) In the first passage the scribe has written de` instead, and in the second he
has made room for a word. This suggests that he did not know what to do with the
abbreviation.
(112) Cf. e.g. 437a32 (me n N: me` ntoi .).
(113) E.g.
in 440a29, 440a30, 443b28, 445a2, 451a13, 441a27. See also the comments on Za below.
(114) E.g. 438b22, 439a9, 449a20.
Harlfinger op. cit

cett

david bloch

36

Thus, both scribes had difficulties in the first part of the De sensu, but

the scribe of V improved when he switched to the


approximately De sensu 5

115

tradition at

, while the scribe of N was not at all capa-

ble of reading the ms. before him ; therefore he made a great number
of mistakes, some of them very peculiar

116

. I also suspect that he some-

times simply guessed as to the reading, when the general sense was
clear to him but the reading was not, thus producing understandable
but inferior readings. At least some passages may well be taken as
examples of such a procedure

117

. Also, it is psychologically understand-

able if the scribe grew tired of trying to read a ms. that he was not
really capable of reading. The large number of primitive errors, such
as wrong word-order, iotacism, dittographies and some omissions were
more likely to be committed by such a scribe than by a scribe who
could easily and carefully copy his exemplar. And finally, if the diffi cult exemplar also had scholia, these would not make it easier on the
scribes. A single passage suggests that this was the case. In 451a20 N is
the only ms. with

mori` ou in the text, but in X the word is there as part

of a scholion. Some variant readings in N may also indicate the pres ence of scholia in the exemplar
sages

119

118

, and, on the basis of a few pas-

, it may be suspected that Alexander's commentary has had

some influence on the textual tradition of

l.

Aristotelian scholars have hitherto regarded the members of


closer to

r than to any other group of mss.

120

as

However, this conclusion

is only true for the De memoria and for part of the De sensu (442a/b-

(115) I believe that the scribe of V actually began a partial use of the

ms.

before he reached 442b26, but did not discard the original exemplar until he
reached what he considered book II of the De sensu. This is suggested by the readings
in 442a17, 442a22 (twice), 442a24, 442a28, 442b7, 442b9, 442b11, 442b23, 442b25
where

a and V agree against the other mss. In the text until 442a17 there is only a

single passage (440a9) where this is the case, and here V has been corrected. The
readings in 444a8, 444a15, 446a25, 446b16, 449b1 also indicates that the copyist

b ms. after 442b27 but regarded it as a secondary source.


myni goun N : my`nixin UO S WP : my`nisin v :
my`nigxin cett.), 443a16 (ogjw`dy N : osmw`dy cett.), 443b26 (liparo`n NZ : lupyro`n
cett.), 444a16 (y didousa N : ydei a LH : ydei a ousa cett.).
(117) Cf. e.g. 445a5 (ai sha`neshai N : osfrai` neshai cett.), 445a17 ( le` gonteq N :
ga`r fasin am : om. Z : fa`skonteq cett.), 452b13 (kiny`sei N : dioi` sei cett.).
(118) Cf. 446b23 ( e te` r{w
/ou} arihm{w /ou} N), 451a10 (g{ig/e} no`mena N).
(119) Cf. 436b2 (te cett. : om. VN Alex ), 436b9 (ti` te` e sti X VN Alex : om. r :
ti` e sti cett.), 441a23 (pa`ntwn twn outwq ugrwn to udwr VN Alex : varia cett.).
still possessed the

(116) Cf. in particular 437b32 (

Mugnier, La filiation ... (n. 3), p. 42-45 (only N) ; Nussbaum, art. cit. (n.
Nussbaum, Aristotle's De Motu Animalium ... (n. 6), p. 15 ; Escobar, op. cit. (n. 3), p. 141, 205. Siwek, op. cit. (n. 1), p. xviii, is unclear on this
(120)

3), p. 131-132, 134 ;

issue : Plures codices familiae

d transcripti

erant a N, reliqui ab ignoto stipite.

He does not say anything about its relations to other mss.

aristotle's de sensu and de memoria

37

De sensu

449b4). For the first part of the


(436a1-442a/b) l is closer to
than to r.
In the following 29 passages l/N agrees with r against am :

Sens

(
.) 436a13, 440b23, 442b24, 443a26, 443a31, 443b17, 444a7,
444a30, 445a22, 445a28, 445b5, 446a10, 446b1, 446b2, 446b21,
447a13, 447b10, 448a5, 448b26, 449a18, 449a23, (
.) 451a8, 451b5,
451b8, 451b11, 452a1, 452b1, 453a2, 453b7.

Mem

In the following 38 passages l/N agrees with

Sens

m against ar :

(
.) 437a5, 437a29, 437b6, 438a4, 438a19, 438a20, 438a24, 438b1,
438b5, 438b9, 439a11, 439a23, 439a31, 439b2, 439b6, 439b13, 440a14,
440a15, 440a26, 440b7, 440b13, 440b16, 441a15, 441a18, 441a26,
441b2, 441b4, 441b17, 441b26, 441b29, 441b30, 442a28, 442b28,
443a27, 444a17, 444b29, 448b5, 448b7.

These agreements establish connections with both m and r. The


agreements of r and l are mostly in questions of word-order and not
textually significant, but they are too numerous to be coincidental.
The agreements of m and l are textually more convincing, and the
existence of a relationship between the two traditions cannot be
denied. Their common ancestor is h 121.
The most remarkable feature of these statistics is the distribution of
agreements. It is striking that N has only three agreements with m
against ar in the second part of the
, whereas the main part of
the agreements with r is found in this second part. And further, N does
not have a single agreement (save a spelling variant in 451a4) with m
against ar in the
, whereas we find eight agreements of rN
against a and m. This fact acquires particular interest because V has
turned to a completely new ms. tradition at about the same point
where N makes a subtle change in preferences, that is, at about 442a
or 442b. This suggests that the physical appearance of i presented
scribes with problems that affected the subsequent textual tradition.
To sum up : V and N together (l) until approximately 442a/b contain a textual tradition related to m, almost without links to r. From
442a/b, and in particular from chapter five of the
(442b27),
V is no longer part of the b tradition, while N still preserves b readings
but more often readings that agree with r rather than with m. How-

De sensu

De memoria

De sensu

om

(121) Cf. in particular 438a4-5 (taiq apo rEYCciP : twn apo M :


. hmamb),
438a20 (lampron hmb : leukon arP), 438a24 (sklyro`teroi hmamb : sklyro`dermoi
.), 441b26 (ouk e n mo`non [ e n post
.), 441a15 (austyrote` rouq hmb : austyrouq
mo`non . V] h :
.). Cf. also the large omission in 437b15 -21 and the confusing textual variants in 437b28.

cett

ins

varia cett

cett

david bloch

38

ever, since the copyist of N still has numerous difficulties in reading his
exemplar, it must be conjectured that he kept his original all the way
through the texts, and that it was the scribe who made l that was faced
with some sort of problem at 442a/b, traces of which can be seen in V
and N.
1921 (m) ; Florence, Biblioteca Medicea
87.4 (Ca) ; Florence, Biblioteca Medicea Lauren ziana, Laur. Plut. 87.20 (v) ; Florence, Biblioteca Medicea Laurenziana,
a
Laur. Plut. 87.21 (Z )

Paris, Bibliotheque nationale, gr.


Laurenziana, Laur. Plut.

The mss. m, Ca, v and Za have never been used in an edition of the
and the De memoria. Siwek sometimes cites their readings in
the apparatus, but he makes no real use of them when constituting the
text.
A few preliminary notes on the ms. m are necessary, since the De
sensu is found twice in this ms. The passage 442a24-449b4 is found in
the beginning of the ms. (f. 5r-9v). The missing text 436a1-442a24 is
found later in the ms. (f. 142r-145v), immediately followed by another
copy of the entire De sensu (f. 146r-169v). I use the siglum ma to designate the text constituted by f. 5r-9v and f. 142r-145v and the siglum
mb to designate the text constituted by 146r-169v 122.
The two texts are closely related, and in 10 passages they hold a
peculiar reading against the rest of the tradition :
De sensu

(Sens.) 437a16, 437a26, 438b12, 439a11, 440a10, 443b24, 444b12,


447a27, 447b28, 448a27.

Some of these passages can be considered textually interesting, but


none of them is likely to preserve readings from mss. ; they are almost
certainly due to conjecture and mistakes. A single omission, a few
words inserted and some otherwise unknown variant readings are pre sented 123.
There can be no doubt that the two texts were copied from the
same, contaminated exemplar, and they were both in turn corrected/
contaminated with the use of other mss. They regularly agree in read-

(122) Siwek , Les manuscrits ... (n. 3), p. 84, used the sigla m and m1, but these
sigla produce a false sense of inferiority on the part of m 1. It seems that Siwek
regards my mb as constituting the text proper. He calls f. 5r-9v a fragment and does
not appear to include it under any of the sigla.
(123) Cf. also Siwek , Les manuscrits ... (n. 3), p. 84, who does not, however, find
the relationship between the two texts easy to determine.

39

aristotle's de sensu and de memoria

ing, and when they differ from each other, one or both of them often
contains the other reading as a variant or shows signs of corrections 124.
As regards the relationship between m and the other extant mss., it is
difficult to provide the relevant statistics, since m is a highly contaminated ms. I believe, however, that for the
442a24-449b4, and,
less obviously, for the
m is much dependent on r, and perhaps closer to X than to any other extant ms., and thus also closer to y
than to LHa. ma and mb often share the text of this group, ma adhering
somewhat strictly to it, mb a little more loosely. In the following 27 passages r and at least one version of m agree in reading against all other
manuscripts :

De memoria

De sensu

Sens

(
.) 437b16, 438a2, 439a16, 440a26, 441a1, 441a5, 441a20, 441b14,
441b18, 441b27, 442a3, 442b6, 443a4 -5, 443b30, 444a13, 446a9,
446b19, 446b26, 447a3, 447a27, 447b24, 448a4, 448b5, (
.) 450a12,
451b10, 452a23, 453a13.

Mem

For a contaminated ms. such as m, this is a very high number of


agreements, and some of the readings are clearly wrong 125. Admittedly, most of them are not conspicuous, but they do establish a relationship between r and m. This connection with r is further strengthened by 447a18-19. A passage is missing in r, and even though ma had
originally written the text, it has been erased to produce the omission
deliberately. The process of deliberately removing text is a rare phenomenon among copyists, and it certainly proves the presence of a
r ms. Incidentally, the passage proves (1) that the copyist valued this
r ms., and (2) that he was using several mss. when copying. The opposite, and more common, practise is found in 450b17-18, where a passage is missing in r. The text was also missing in the exemplar of m,
but the copyist has inserted it in the margin.
In the first part of the text (436a1-442a24) m presents some striking
agreements with h, but also some readings agreeing with r. The agreements with h are, however, so many that a ms. of this tradition must
be assumed to be the foundation of the text in this part of the
126
.

De

sensu

(124) Cf. e.g. 437a5, 438a20, 443a15.


(125) Cf. e.g. 441a1 (tyn osfrysin a hmbP : tyn ai shysin tau`tyn rma), 443a4-5
(osfraino`mena ahP : osfraino`menoi rma : osfraino`men{a/oi} mb), 443a5-6 (auta
anapne` onta amCaNvP : aut{oi / a} anapne` ont{eq/a} mb : autoi anapne` onteq r a :
auto`n Za), 453a13 (mo`noiq sumbe` byken rm : mo`noiq sumbe` byken fu`sei P : fu`sei
mo`noiq sumbe` byken .).
.),
(126) For some examples, cf. 436a19 (gene` shai CaVNZama : gi` gneshai
.), 441a13 (ouk VNvZama : outw E : outwq Y : ou tw
440b24 (ei si n VNma : e sti n

cett

cett

cett

david bloch

40

Thus, m presents different textual traditions, but no part of the text


really supports Siwek's suggestion that m and v are both descended
from S

127

, and the same is true of Mugnier's suggestion that U was the

ancestor of m

128

Harlfinger has argued that m was part of a collected, or at least a


partial, edition of the Corpus Aristotelicum, dating from the third
quarter of the 14th century, and he further argued that this edition
was based on an earlier edition

129

. It is not surprising, then, that m

shows several signs of contamination. It is actually more surprising


that conjectures in our text seem to be relatively few. Perhaps, the
number of readings available from different mss. tempered the Byzan tine editors. It is easily seen that the copyist had access to several ms.
traditions, and the same tradition was not used for all the texts ;
actually, the scribe even had the possibility to constitute the texts by
compiling readings from several manuscripts so that no firm basis was
discernable. This is partially what has happened to the De sensu.
The close relationship between X and m, in particular for the second
part of the De sensu, is suggested already by a number of passages in
which the mss. contain the same scholia, but these cannot, however,
be considered strong evidence, since scholia are abundant in both X
and m. Passages such as the following are more interesting
444a8.
447a24.
447b3.

448b2.

448b2.

450b5.
453b6.

130

einai Xym : einai i dion VNZ : i dion einai cett.


dio`per Xma(eras., ut v.)N : dei per v : aei per Za : ei per cett.
ende`jetai ai sha`neshai Xmamb : ai sha`neshai autwn ende`jetai N : ai sha`neshai ende`jetai cett.
ora y oti tou`tou ti Xmamby : om. LHa : ti` tou`tou OdSWv :
ti N : touto Za : y oti tou`tou ti cett.
me`gehoq, ei per esti ti me`gehoq Xmamb : ei per esti ti me`gehoq cett.
tou pa`houq Xm : tou dejome`nou to pa`hoq cett.
en aux Xm : en auxy`sei cett.
a

cett.), 441a26 ( en t
jeiri fula`xai CaVNvmamb : diafula`xai en t jeiri W : en t
jeiri fula`ttein Za : fula`xai en t jeiri cett.), 441b1 (oi polloi imamb : polloi` cett.),
a
a
b
441b4 (eido`q ti tyq gyq vm : eido`q ti gyq C VNm : eidoq gyq U : eidoq tyq gy
q Za :
gyq ti eido`q cett.).
(127)
(128)
(129)

Siwek
Mugnier
Harlfinger

, Les manuscrits ... (n. 3), p. 88.


, La filiation ... (n. 3), p. 43-45.
, op. cit. (n. 3), p. 55 -57.

(130) Further support is provided by the passage 445a30 in which X is the only
ms. to read

lon,

dylon,

all others having

fanero`n. However, mb also


fanero`n in the process of

but it was erased and replaced with

originally had
correcting.

dy-

41

aristotle's de sensu and de memoria


As regards y and m, it must be noticed that m

ote for otan at 443b31. In 445b7 m

has inserted

and y alone share

e kaston in the mar-

gin ; in y this word has been corrected in the text. Finally, in 447a17

` stw to the transmitted


has been corrected from e ka
e ka`stou, and the only other ms. reading e ka`stw is y.
ma

reading

The most difficult part in categorizing m is determining more accu -

rately the sources of contamination. The difficulty is twofold since con tamination has occurred both in the ancestor and in m itself.
That the exemplar was, indeed, contaminated, is proved by the fact
that m consistently follows two textual traditions, viz.

and

h. How-

ever, there are some other peculiar readings of m linking it to even


more textual traditions (unless otherwise specified m a and mb agree in
the reading) :

sjedon twn am (fort. sjedon eras.)m : sjedon twn te W : twn


PV : ta N : tw
n te cett.
438b7.
to e ntoq m Z P : e ntoq cett.
439a11. osmyn am (ti` s.l.) : osmy m : tyn osmyn rU : osmyn hP
441b14. e n autoiq e nantio`tyq e n e ka`stw
m (post corr.)P : varia cett.
q m (kai s.l.)P : om. amC Z : kai rNvm
443a26. w
443b6.
e ti d' ei per mV C NZ P : oti d' C Mivm : oti deper EV : oti
de` per Y : e ti d' ei cett.
n Z : e njuhei` q P :
446a9 -10. e kjuhei` q am : {diairehe n/e kjuhe n} y
e kjuhei q yn cett.
447b6.
ai sha`noit' C Mim m : ai shyto Z : ai shoit' cett.
auty y auty C Mim : y auty y auty EV : y auty cett.
447b28. y
449a20. to auto C MiZ m m : tw
autw cett.
449b16. to`de Z : om. C Mim : todi cett.
a

436a19.

From these statistics it appears that the exemplar of m had also been
contaminated

with

variant

readings

from

a.

The

contaminating

source was part of the CcMi branch, since EYV are never alone in
agreement

with

m,

while

this

is

sometimes

the

case

with

C cMi.

I believe that the contamination was probably reciprocal, thus explainc

ing, at least partly, the contamination in the C Mi tradition (cf. the


description of these mss. above).
Some of the agreements also suggest a connection between P and
m

131

. As will be shown below, P does not present a completely uniform

text either. Thus, it often shares the readings of the tradition around

in the first part of the De sensu but almost never in the second part and
in the De memoria. This suggests that P was also influenced by the ms.

(131) Cf. in particular 443a26, where the reading of m


accepted by modern editors.

i,

and P is generally

david bloch

42

which has left peculiar traces in NVm, and which is also relevant in
the discussion of Ca, v and Za (see further below) 132.
In addition to the contamination of the exemplar, m was directly
contaminated by other mss. than the exemplar. These readings also
stem from all branches of the textual tradition, and thus m is a highly
contaminated ms. Consequently, the individual readings of m carry no
manuscript authority, and must be regarded as errors or, occasionally,
emendations. They are few, which is understandable considering the
specific character and function of the ms. 133
The results may be summarized by saying that m is related to both h
(in 436a1-442a/b) and r (in 442a/b-453b11), but it also contains variants from the latter tradition in the first part of the De sensu. It is
extremely contaminated throughout, and the copyist (or editor) had
access to several mss. and many different ms. readings. Therefore, con clusions regarding the textual tradition of m can never be stated generally to cover all the texts in the ms. ; thus, my investigations applies
only to the De sensu and the De memoria, and I make no claims regarding the other texts in m 134. It seems, for instance, that m and P are
closely related in the textual tradition of the De partibus animalium 135.
The mss. Ca, v and Za are all part of the textual tradition descending
from h 136, but they are not as closely connected to each other as OdSW
and VN. The first two contain a number of scholia, and the scholia in
v are often identical with the ones found in UW (cf. above on W) 137.
In one passage of the De memoria (451b29-30) U and v seem to be the
only mss. to report a particular interpretation of the text.
(132) Perhaps it is also interesting that P is known, at some point, to have been
in the hands of Gennadios Scholarios, that is, the teacher of Matthaios Kamariotes
who was responsible for some marginal notes in m. Cf.
, op. cit. (n. 3),
p. 153-157 ;
, op. cit. (n. 3), p. 179-181.
(133) Cf. 437a16, 437a26, 440a26, 443b24, 447a26, 449b5, 450a11, 451a15,
451b5.
(134) This point must be stressed, since
, Les manuscrits ... (n. 3), p. 87 with
notes 8-9, has detected some difference in the text of m as regards the De sensu/De
memoria and the rest of the texts. This is an important discovery, and it is strange
that Siwek, apparently, does not think it has any significance as regards his stem matic considerations.
(135)
, op. cit. (n. 3), p. 39, 52, 55, and P.
, ed., Aristote. Les parties
des animaux, Paris, 1956, p. xxxii, have both noticed the relationship between P and
m. During thought that m was a direct copy of P, while Louis suggested a common
ancestor. Furthermore,
, op. cit. (n. 3), p. 261, note 1, has found an
equally close relationship between P and m for the Historia animalium.
(136) The omission in 437b15-21 has also occurred in these mss., and thus pro vides a strong basis for making them part of this textual tradition.
(137) E.g. in 438a14, 443a31, 445b11, 446a1, 448a9, 449a12.

Harlfinger

Escobar

Siwek

Du ring

Louis

Harlfinger

43

aristotle's de sensu and de memoria


a

It should be noted that, unlike VNm, C , v and Z

follow the

h tradi-

tion in the entire text. The sheer number of passages is convincing evi dence of a textual relation, and the character of the readings strength ens the evidence significantly

138

In addition to these readings there is a number of passages in which


either m

or m

also shares a reading, providing it either in the text or

as a variant reading ; this is explained by contamination.


In the following 59 passages C

has a peculiar reading against the

rest of the textual tradition :


(

Sens

.) 436a20, 437a5, 437a13, 437b1, 437b9, 437b12, 437b14, 438a4,

438a10,

438a15,

438b2 -3,

439b15,

439b20,

439b28,

438b12,
440a5,

439a1,

439a24,

439b3,

439b10,

440a23,

440b27,

441a2,

441a11,

441a13, 441a16, 441a30, 441b7, 441b10 -12 (omission), 441b21, 442a7,


442a13,

442a22,

443a20,

443a31,

443b3,

442b2,

443b18 -19,

442b4,

442b23,

444a26,

444b32,

445b8,

445b18,

442b30,

443b22,

445b25,

443a10,

444a3,

446b18,

443a12,

444a7,

444a16,

446b22,

446b28,

447a19, 447a27, 447b10, 448a23.

In addition to a number of wrong readings shared with other mss.,


there are 145 passages (

Sens

. 106 ;

Mem

. 39), in which v contains pecu-

liar readings :
(

Sens

.) 436a2, 436a5, 436b5, 436b13, 436b18, 437a17, 437a19, 437a21,

437a26,

437a27,

437a28,

437b4,

438b14,

438b23,

438b29,

439a3,

439a27,

440a13,

441a26,

441b1,

443a19,

443a20,

440a18,
441b22,

437b13,
439a13,

440b13,
442a4,

443a25,

440b21,

442a22,

443a26,

438a3,

438a15,

438a22,

439a14,

439a15,

439a17,

441a8,

442b17,

443b8,

441a18,

441a23,

442b20,

442b29,

444a18,

444a23,

443b19,

444a27, 444a29, 444b1, 444b2, 444b3, 444b4, 444b10, 444b13, 444b19,


444b29,

445a3,

445a29,

445b10,

445b18,

446a21,

446a30,

446b2,

447a1,
447b27,

447a16,
447b28,

445a6,

445a10,

446b3,

447a17,

445a11,

445b21,

445a22,

446a13,

446b15,

446b22,

446b24,

447a25,

447b3,

447b15,

447b16,

448a4,

448a5,

448a17,

448a18,

448a1,

445b26,

445a28,

446a1,

446b12,

447a24,

447b29,

445a19,

445b22,

448a23, 448a27, 448b2, 448b5, 448b8, 448b13, 448b20, 448b26, 449a8,


(

Mem

.) 449b4, 449b12, 449b21, 450a4, 450a8, 450a11, 450a31, 450b2,

450b8,

450b12,

451a22,

451b5,

452a10,

452a13,

450b25,
451b6,

450b29,

451a9,

451b29 -30,

452a19,

452a25,

451a11,

451b30,

452b4,

451a14,

452a3,

452b17,

452a6,

452b20,

451a19,
452a9,
452b23,

453a1, 453a19, 453a23, 453a27, 453a29, 453a31, 453b1.

As regards Za, a rough count produced no less than 424 ( !) peculiar


readings in the

De sensu

alone, making Za by far the most corrupt of

(138) Cf. in particular 437a28, 439b6, 441b26, 444b29, 447b9.

david bloch

44

the older mss. containing the 139 and the


. It would be
tedious to list all the readings here . I will comment in more detail on
this ms. below.
in
The peculiar
readings found in Ca differ in kind from those found
a
v and Z . Even though the number of clear errors is high, Ca is not
generally meaningless.
Some of them may indicate problems with
abbreviations 140. The general impression of Ca141is not, however, of an
incapable but of a careless yet thinking copyist . Most of his peculiar
mistakes are omissions, changed word-order, obvious errors142in trans
scription, mental errors and inclusions of scholia in the text . Ca was
further corrected using another ms., and a trace of this can be found in
445b13
in which almost the entire b tradition has omitted a passage.
Zaa, mb anda P are the only mss. of b with the text incorporated, but in
m and C the missing texta has been written in the margin.
The errors in v and Z , on the other hand, are very similar to the
errors found in N, which suggests that the textual tradition of these
mss. has also beena seriously corrupted by a difficult ms. This is telling,
because v and Z are closer to l than to 143any other group of mss. Some
interesting mistakes in v should be listed :
De sensu

De memoria

439a13-17: e ne` rgwq v: e nergei` a . 144


440a13: proeiryme` nw v: pro` teron ei ryme` nw .
437a21: e ou` saq v: pe` nt' ou saq . & 444b19: pe` nte . v 145
cett

cett

cett

om

(139) The readings can be gathered from my previously mentioned article, cf.
n. 2.
(a toma Ca : a ma .), 441b7
(140) Cf.
e.g. 440a5 (o te Ca : o tan .), 440a23
a
a
.), 442b23 (poiy` sei C : poiy` soien M: poiy` seien .),
(e jousi C : e jousai
446b28 (ou de n Ca : ou te Za : ou de .).
(141) In two passages, I even believe that Ca (apparently by conjecture) holds
the correct reading against a united textual
tradition: 442a22 (to axanho n Ca : to
a
xanho n me n .), 442b28-29 (e n a llw C : e n a llw ge` nei .). On C , cf. also Wilson,
(n. 3), p. 165.
(142) Omissions: e.g. 437b1, 438a15, 439b15, 444a7, 447b10. Word-order:
437a13, 438b2
-3, 439a24, 441a11, 442b30, 443b18 -19. Scribal error: e.g. 436a20
(gegrafe Ca : ge` graptai .), 444b32 (kari(i trwn Ca a: i atrwn .), 437b12
barou si C : karybariou si Za : karybar{ou /iw}si mb : karybarou si .). Mental
error: e.g.
442b2 (fanero n Ca : dy lon .). Scholia: e.g. 439b10 (tou` tou tou diafa
. 51.19-21: tou` tou .), 444a26 (proygoume` nwq Ca,
nou q Ca, . Alex.,
Alex.,
. 99.27-100.1: e pi .) and 445b8 (tw du` nashai ga r e stin e kaston
au twn e n toi q ai shytyri` oiq. tw du` nashai Ca : tou du` nashai CcM: to du` nashai ZaP:
tw du` nashai .).
(143) Even though I present the situation as if the scribe of v committed all the
errors, it is likely that his exemplar
was already very corrupt.
(144) The
abbreviation
e nerg ` (or e ne` rg) is regularly found in Za (see below).
(145) Za contains the abbreviation. Since Ca and v have some connections, it
may also be interesting that Ca is the only ms. to abbreviate te` ttara (D) in 439b28.
cett

cett

cett

cett

cett

cett

cett

art. cit.

cett

cett

cett

cett

cf

In Sens

In Sens

cett

cett

cett

cf.

45

aristotle's de sensu and de memoria

442b29: e teron v: ugro`n .


448a23: peri tou`twn v: peri tou`tou N: para touto .
445a6: me n v: me` son N: me` sy .
450b12: e pi v( peri ): peri .
451a11 & 451a14: wq ei ko`q v: wq ei ko`na . 146
cett

cett

cett

v.l.

cett

cett

These are all singular examples, but one of the most notable kinds of
error is the extreme frequency with which the ending of a word comes
out wrong in v.147This points towards an ancestor with many suspensions of endings . Also,148some of the mistakes are very peculiar, pro. The readings in 438a8 (ana`kla) and in
ducing sheer nonsense
453a25 (euhupory`) constitute traces of this original exemplar. The
scribe
has hesitated in dissolving the abbreviations.
Za shares these features. It is singularly corrupt, but ( Siwek) it is
relatively easy to incorporate it into the stemma. This has already
beena demonstrated above and will be further illustrated below.
Z is notable among the extant early mss. not least on account of the
number of abbreviations actually found in the ms. Since, as stated
above, abbreviations have caused a lot of problems in other mss. (in
particular,
in V, N and v) of the same tradition, it is very likely that
the
exemplar in this respect. However, contrary to the other
Za reflects
mss., Za still retains a large number of abbreviations, while the scribes
of mss. such as VNv have attempted to resolve most of them.
The phenomenon of suspension, that is, the omission of the ending of
the word, signalled at most by a stroke and often simply
by raising the
last letter from the line, is extremely frequent in Za. The following
table lists some of the most common suspensions, taken primarily from
the second part of
and from the
:
pace

De sensu

De memoria

a : 447a24, 448b12, 453a8.


y : 449b24, 449a28 (teh = teh), 449b28 (mnym ` ), 453a25.
-w : 447b25.
-ei : 448a21, 451a29 -30.
-ein : 448a3, 451a26.
-

(146) The abbreviation in the exemplar must have been ei ko ` (see below on Za).
(147) Cf. e.g.
436b5 (fulaky v: fulakai .), 437a19 (ai shyty`ria v:
.), 438a3 (tanaw`tera v: tanatw`teron P:
ai shytikon Za : ai shytyri` oiq
tanaw`teron .), 441a8 (me` roq v: me` rouq Za : me` rouq .), 441b1 (upa`rjousi v:
upa`rjonteq .), 442a22-23 (leukon v: leukou
.), 445a19 (aplwq v: apla .),
446b12 (i swq v: i soiq .), 447b3 (sumfwnousi v: sumfw`nou Za : sumfwni` a .),
450a11-12 (ai shytikon v: ai shytikw d .), 453b1 (amnymone` steron v:
amnymomone` steroi V: amnymone` nesteroi O : amnymone` steroi .).
(148) Cf. e.g. 436a2 (e ti v: e sti .), 436a5 (ei rytai v: ei ryme` na .), 438b14
(ai matoqa v: ommatoq .), 442b29 (e teron v: ugro`n .), 448b8 (tw dittw v: tw
.), 448b13, (tetra`pyjon v: tetra`pyju .).
dy tw Z : twdi tw aP: tw de tw
cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

cett

david bloch

46

en (or e ?) : 453a8 (oud ` = oude n or oude ).


-yn : 447a23.
-iq : (pollak ` = polla`kiq).
-oi : 450b22 (ment ` = me` ntoi).
-oq : 448b13, 449a20-1 (me` geh = me` gehoq).
-ou : 452a17 (kahol` = kaho`lou).
-sei : 452a29 (fu` = fu`sei).
n).
-wn : 452a22 (e lh ` = e lhw
-

Such an elaborate system of suspensions should, of course, only be


used with much care and when there can be no or little doubt as to
what has been left out ; this, however, is not the case in Za. Frequently,
the reader is left guessing what the ending must be. Some particular
words and some particular types of endings are especially difficult.
The following table contains some problematic examples :

lanhn ` : lanha`nei, lanha`nein, lanha`noi (448a21 = -ei ? : 448a27, 452b27 =


-ein ? : 448a27 = -oi) .
sunej : sunejei, sunejwq, sunejouq (448a28 = -ei ? : 448b4-5 = -wq ? :
450a8 = -ouq ?).
fu ` : = fu`sei, fu`sin, and fu`siq (452a29, 452b1 = -ei ? : 452a30 = sin ? :
453b8 = -iq ?).
e nanti| ` : -on, -a (448a4) .
le ` : le` gein, le` gei (449b23).
e nerg = e nerg, e nergei (450b27, 452b24).
ai shanomn ` : ai shanome` ny, ai shano`menoq (448b23) .
orh ` : = orhaiq ( ?) (449b20).
149

150

151

In addition, there are some abbreviations that will have been familiar to copyists but are still likely to corrupt the text if read (and copied) with speed ; for instance, by causing a problem distinguishing
between the conjugated -ei form and the infinitive -ein. The following
is a list of some occurrences :

_
diafe
` = diafe` rei (e.g. 448a14, 448a17, 452a4, 453a6).
_
o
ei k ` = ei ko`na (e.g. 450b23-24).
la ` ` = labein (451b30).
upola ` ` = upolabein (450b24).
ana ` = ana`gky and the other inflected forms (passim).

(149) Similar problems with manha`nein are found in 452a5.


(150) The general ms.-reading is -a, but v has -on. A similar case with
mnymoneut|| is found in 449b9.
(151) It should be ai shanome` ny, but v, reading from a ms. similar to Za, has
ai shano`menoq.

47

aristotle's de sensu and de memoria

There are even a few cases where the copyist is not consistent in his
abbreviation. Thus, for ene`rgeia and ene`rgeian he, apparently, uses
152
g
g
.) and ener ` (e.g. 447b17, 449a1,
both ene`r (e.g. 447b15, 449a12
j
449a11). Similarly, he writes both su`stoi (447b30) and sustoij ( !)
(448a16) for

su`stoija.

The above description of Z

shows that this is a difficult ms., but it

does not show its generally very corrupt character. I will not, however,
go into great detail at this point because a mere glance at the enor mous amount of peculiar readings and their character clearly reveals
the nature of Z
First, Z

153

. The following are some brief general observations.

is one of the few

mss. that does not omit the text in

445b13. This clearly points towards contamination, and the missing

estai

in the following line may hint that the contaminating source


b

was an ancestor of m . This is in accordance with the stemmatic relationships of v, Z

and m as reconstructed below.

The number of ordinary scribal errors is not higher than in some of


the other mss.

154

, and the mistake committed in 442b28 where the

scribe has inserted a commentators

fysi` n

in the Aristotelian text is not

startling either. A more peculiar fact is found in the omissions in


436b11-13 and 437a27-29 which have been indicated by the scribe
introducing a blank space. This suggests (perhaps) that the scribe
knew already when transcribing the text that there were some prob lems in these passages. Perhaps he could not read his exemplar and
intended to fill the lacunae later.
a

Concluding this brief examination of Z , it may be said that the state


of corruption makes it impossible to use as a textual witness. The edi tor might consider some of the readings, but they can at most be
treated as conjectural emendations, not as true textual variants

155

Noticing that v and Za, in particular, have gone through almost the
a

same process as N ; that C , v and Z

have a lot of obvious agreements

m ; and that the scholia in v are very similar to the UW scholia,


, v and Z must be considered part of h.

with
a

As regards the relationship between C , v and Z , they agree against


the rest of the tradition in only very few passages, but this is easily
explained by the amount of corruption in all of them. More impor tantly, they generally agree in substantial readings with each other

(152) This abbreviation is what caused the strange

ene`rgwq

in v.

(153) For the readings, cf. Bloch, art. cit. (n. 2).
(154) Cf. e.g. 438a8. (antana`klasiq), 438a22 ( aristo`tatoq), 438b21 (ai shysiq

osfrysiq).
(155) Siwek, Les manuscrits ... (n. 3), p. 127, treats 437b7 ( ke`faloi`

ne`q)

as an interesting variant, but he does not record it in his edition.

te, i jhueq ti-

david bloch

48
and with either

444b30.
a

156

The few agreements against the

udati C vZ : e n udati cett.


my C vZ : my` ti cett.

443a2.

l.

or, more often,

rest of the tradition are

and v agree in two readings against the rest of the tradition :

fu`sewn C v : fu`sewq cett.


on C v : onper cett.
a

436a20.

440b1.

Ca and Za agree against the rest of the tradition in 4 passages :

fysi n oran C Z : orai fysi` n V : oran fysi` n cett.


kassi` teroq C Z : kattite` rou v : katti` teroq cett.
me n C Z : me n oun cett.
toiq atoma C Z : toiq ato`moiq v : toiq ta atoma cett.
a

438a4.

444a3.

445b18.

v and Z

443a20.

agree against the rest of the tradition in 27 passages :

438b14.

sko`toq gene` shai vZ : gi` gneshai sko`toq UP : sko`toq


gi` gneshai m : gene` shai sko`toq cett.
vo`fou kai vZ : kai kai y : kai vo`fou kai cett.
wsper oun cett. : wsper vZ
einai ulyn to udwr vZ : einai to udwr ulyn arU : ulyn einai
to udwr cett.
oson avashai anagkaion vZ : o e n anagkai` w avashai Y :
oson anagkaion avashai cett.
kai e n ae` ri vZ : e n ae` ri cett.
ouji vZ : ouj cett.
anapne` onta vZ : varia cett.
e n vZ : e n tw cett.
e pei` per kai vZ : varia cett.
tyq trofyq vZ : om. XC N : tyq osmyq cett.
zw` wn twn toiou`twn vZ : twn toiou`twn zw` wn cett.
dio vZ : dio`per cett.
de vZ : me` ntoi cett.
e sjaton vZ : e sjata cett.
e stai vZ : e stin cett.
e sjaton vZ : e sjata cett.
e nanti` on vZ : e nanti` a mLH yC Nm : ta e nanti` a cett.
ai shyton e stai vZ : e stai ai shyto`n cett.
ei my` vZ : my` cett.
me n ga`r vZ : me` n cett.
d' ara vZ : ara cett.
a

439a7.
439a25.
441a18.

442a4.

442b29.
443a1.
444a27.
444b1.
444b5.
444b10.
444b14.
444b27.
445a29.
445b23.
445b26.
445b26.
445b26 -7.
446a13-4.
446b3.
446b15.
447b16.

(156) Cf. also 445b18 cited below.

49

aristotle's de sensu and de memoria

e nanti` on vZ : e nanti` a cett.


e nde` jetai vZ : e nde` joito mC
e nde` joito cett.
dylon vZ : dylon oti cett.
artion vZ : arti` ou cett.
tiq an vZ : tiq cett.
a

448a4.

448a5.

e nde` joito an rNm

an

448a5.
448a12.
448b26.

The number of agreements between v and Z , two very corrupt


mss., proves a close relationship between them. Considering that they
also follow the same tradition in general, they must stem from the
same ancestor

157

Concerning Ca, none of the readings are strong evidence of a close


relationship with the other mss. They are interesting, because they
present errors of a kind that may hint at a difficult exemplar. This
would certainly fit the description of the mss. given above, but, even
granting that both C

and vZ

are individually extremely corrupt, we

would have expected a lot more of this kind. Considering the relations
already established between these three mss. and the rest of

h, I find it

most likely that they are all indirectly descended from a common

ancestor ( ) that was difficult to read. This ancestor was also the source
of

l.

The C

line did much better than the vZ

been partly corrected. The vZ

line, but it has also

tradition has, however, been contami-

nated by another ms. related to m

158

A relationship between m and vZ

is established by the following 11

passages in which one of the m versions agrees with one or both of v


and Z

against the rest of the tradition :

438a10.

438b3.

dylon autw vZ m : dylon yn a : dylon autw yn UC m :


autw dylon yn P : dylon yn autw cett.
ei rytai pro`teron vZ m : ei rytai cett.(pro`teron expunctum
a

m )
439a20.
439a29.
439b13.

ste` rysiq vZ m : ste` rysi` q e sti cett.


e sti touto vZ m m : tout' e sti cett.
osoiq vm : oson Z : kai osoiq arP : kai oswn U : oswn cett.
a

o
(157) Some of the corruptions are also very telling. Cf. e.g. 441a8. ( me` r uq Z :
me` roq v : me` rouq cett.), 444b13 (poll ` Z : pollai v : polla cett.), 447b28 (e kei` noiq v :
e kei` ny Z : e kei` nyq cett.), 451b6 (e a`n tiq v : e a`n te Z : e an cett.), 451b16
` meha cett. (sed corr. ex anamimnysko`meha v).
(anamimnysko`meha Z : anamimnyskw
a

(158) These conclusions are contrary to Siwek, Les manuscrits ... (n. 3), p. 123124 and Siwek, op. cit. (n. 1), p. xxi, who believes that v and m stem directly from
the same source, and that C

and Z

cannot be categorized in any of the existing tra -

ditions. Escobar, op. cit. (n. 3), p. 205, thinks that v and m stem directly from the
same source, but he also believes that Z

is related to them.

david bloch

50
440a11.

fai`netai leuko q fai`neshai leuko q leuko q fai`netai


eido` q ti ty q gy q
eidoq ty q gy q eido` q ti gy q
eidoq gy q gy q ti eido` q
xyro` tytoq
xyro` tyq
Peri de
Tw n de loipw n prw ton skepte` on peri
Peri
e peidy pe` fuken y ki`nysiq yde gi`gneshai e peidy pe` fuken
e peidy pe` fuken
y ki`nysiq yde gene` shai gene` shai
yde y ki`nysiq m
yde e peidy pe` fuken y ki`nysiq yde r
e peidy
e peidy pe` fuken yde y ki`nysiq gene` shai
pe` fuken y ki`nysiq y` dy
a rjy n k i oi ny n labein a rjy n koiny n labein a rjy n
kiny` sewq labein
a rjy n kiny` sewq labein a rjy n
kiny` sewq
a

vm :

Z :

cett.
441b4.

445a13.
449b4.

vm :

C VNm

Z :

U:

vZ m m

cett.

cett.

vZ m :

C Mi :

cett.

451b11.

vm :

Z :

EYbV :

C Mi :

N:

V (in marg.) :

451b30 -31.

{ /

v:

m:

Z P:

cett.

H :

159

These readings are not very convincing, and often they are not
shared by both ma and mb. Still, the reading in 445a13 is interesting as
connecting the mss., and so, perhaps, are the readings in 438a10 and
438b3. However, these few variant readings are not equal to the read ings presented above which proved that m belongs to a textual tradi a

tion different from that of vZ . Therefore, the necessary explanation of


a

the relationship between m and vZ , given the general nature of the


three mss., is contamination.
Contamination in the sources of m and vZa explains why vZa somea

times agrees with m , sometimes with m . I conjecture that the scribe


who made the ancestor of v and Z

possessed a ms. of the

tradition

which was used extensively for the De sensu and the De memoria. However, he also possessed a copy of the ancestor of m, being, as stated
above, an earlier philological edition of the Corpus Aristotelicum. For
the De sensu and the De memoria, he used this to extract variant readings, perhaps correcting some mistakes, but he used the ms. of the

tradition as the basis. From the results of Siwek and Escobar it appears
that he used the ancestor of m as his textual basis from the De somno
and to the end of the Parva naturalia, but I have not examined the mss.
in these parts

160

. The copyist of v, but apparently not the copyist of Z ,

may also have introduced readings from the ancestor of v and Z

into

the ancestor of m, and thus again linking m to the tradition.

kiny n

(159) This last example is the only one cited for the De memoria by

manuscrits ... (n. 3), p. 87, note 9, but interestingly v actually has
has then been corrected above the line without erasing
(160)
181, 205.

Siwek, Les
( !), which

Siwek, Les manuscrits ... (n. 3), p. 123-124 ; Escobar, op. cit. (n. 3), p. 180-

51

aristotle's de sensu and de memoria

This explanation is conjectural, but contamination in the sources of


a

m and vZ , along with the other features of the mss., certainly provides
us with an understanding of almost every single variant reading in the
mss. It cannot be disputed that m and vZa, at least for the greater part
of the De sensu and for the entire De memoria, belong to different
branches of the textual tradition, and therefore solutions along the
proposed lines must be sought.

3. Vatican, Biblioteca Apostolica Vaticana, Vat. gr. 1339 (P)


The ms. P is probably the most hotly debated Aristotelian ms. I do
not want to repeat the entire discussion. Two problems have been par ticularly disputed : (1) the date of the ms., and (2) its value as a textual
witness. The first problem has, however, been solved by Harlfinger,
and P can now be securely dated to the 14th century

161

. This leaves us

with the equally interesting question of its textual value. The statistics
will prove the difficulty of categorizing this ms.
A rough count produced 283 passages (Sens. 212 ; Mem. 71), in
which P agrees with

b, or part of b, against a. It would be tedious to

list them all, but the mere number of passages shows that there must

b, and this is supported by the content

be a relationship between P and

of several passages. The agreements are of all kinds, but it is significant


that quite a few of them constitute factual agreements against
of them likely to be correct, some of them not

162

a, some

In the following 68 passages (Sens. 57 ; Mem. 11) P agrees with


against

b:

(161) Harlfinger, op. cit. (n. 3), p. 251 -255.


(162) A

list

containing

different

examples

of

agreements :

Omissions :

e.g.

436a10, 436b1, 437b1, 438b17, 439a17, 440a20, 441a1, 442b8, 443a16, 445a2,
446b1, 447a6, 449b17, 450a11, 452b2. Words inserted : e.g. 436a13, 436b2, 437b9,
438a4, 438a18, 439a24, 441a21, 442a25, 444a12, 444b11, 446b21, 448a1, 451a5,
452a26,

452b27.

Word-order :

e.g.

438b25,

441a18,

441a23,

442a18,

444a33,

e nioi bP : nun
a), 438a1 (e jeu`ato hXym m P : e jei` ato LH : loja`zeto a), 438a20 (aima pi on
rhP : aima plei on EYC i : plei on aima M), 440a24 (aki` nyton bP : kinyton a),
441a14 (de kai kinoume` nouq iLH yO SWP : de kai keime` nouq EYC MX : om. i),
442a22 ( lei` petai bP : e petai a), 442b25 ( ske` vin bP : apo`dosin a), 445a17
(tre` feshai fa`skonteq rUO SC vm P : tre` feshai ga`r fasin a : tre` feshai gar
fa`skonteq W : tre` feshai le` gonteq N : tre` fonteq Z ), 446b3 (yshytai bP : ei
ai shytai EV : ei ai shyta C Mi), 446b14 (o ayr kai to udwr riUO Sm m P : pan
udwr EYC Mi : ayr kai to udwr V W : kai to udwr V ), 448a16 (kalw rmV C v q a), 450a4 (o nown bP : onoma`zomen a), 453a23 (e nojlountai
Z P : kala N : all' w
bP : e nupa`rjei EYbV : e nojlei C Mi).
445a27, 448a13, 450a30, 453a29. Substantial readings : e.g. 437a20 (
a

david bloch

52
a

(Sens.) 436b12 (with m m ), 437a11 (with m ), 437a26, 437a28, 437b7,


a

437b16, 437b29, 437b30, 438a15, 439a10 (one reading with m m v and


a

another with m ), 441b4, 442a23 (with Vm ), 442a24, 442a30 (with


a

m m ), 442b3 (with m ), 443a24, 443b1 (with m ), 443b11 (with m ),


b

443b23 (with m ), 444a15, 444a18 (with m m ), 444a29 (with m ),


a

444a33, 444b28, 445a1, 445a24 (with m m ), 445b9, 445b13, 445b15,


b

445b17 (with m ), 445b25, 446a5 one reading with

am

a and another with

), 446a9, 446a27, 447a3, 447a5, 447a8, 447a12, 447a20, 447a22,


a

447a25, 447a26, 447a28, 447b1 (with m m N), 447b4 (with v), 447b22,
447b26,

448a4,

448a5

(with

m ),

448a23,
a

448b16, 448b21 (one reading with m m


a

448a30,

448b6,

and another with

448b8,

a), 448b26,

449a23 (with m m v), (Mem.) 449b9, 449b19, 450a9, 451a24, 451b5,


451b27, 451b29, 452a23, 452b11, 453a9, 453b11 (with v).

The number of agreements between P and

a is too high to be coinci-

dental, but it should be noted that they are rarely alone in the agree ment. Very often m also has the reading, which makes it probable that
P had the readings from the tradition surrounding m. Looking, then,
only at the passages in which P and

a are alone in the agreement, there

are some that do not look coincidental, but they are too few to be
more than suggestive

amP

163

. It is also noteworthy that for the De memoria

are never alone in agreement against the rest of the tradition.

This points strongly towards contamination in the De sensu ; the remaining passages may, then, be either authentic readings in the textual tradi tions of P or contaminations from another source. Thus, the agreements
with

a are completely different from the agreements with b.

Finally, I have counted 215 passages (Sens. 168 ; Mem. 47) in which
P has a peculiar reading against the rest of the tradition, and even
though statistics may be misleading, the evidence thus shows that P is
to be considered part of the

family. However, it cannot simply be

placed in one of the existing traditions, and it probably constitutes an


independent tradition of

b.

A number of features must, however, be

thought through before editors consider using this ms., and thus I do
not include P under the siglum

b.

The enormous amount of peculiar readings, most of them plain


errors, is obvious. A partial explanation of the errors may be that they
are (unfortunate) attempts at correcting and clarifying the text, but

ai t' EYC MP : oi t' cett.), 437b30 (pur aP : fwq b),


pra`sion EYC MP : pra`sinon cett.), 444a15 (y touto b(ut v.) : om. aP),
444b28 (ontoq aP : ora
H : oran cett.), 445b9 (ty`n te aP : ei y du`namiq kai ty`n te
n Z : ei y du`namiq, kai tyn cett.), 449b9
O S : ei y du`namiq kai ty`n te W : kai ty
(lypte` on aP : skepte` on b), 451b29 (ally`laiq aP : om. b).
(163) Cf. e.g. 437b29 (

442a24 (

53

aristotle's de sensu and de memoria

this explanation can only be partial. Evidently, P is simply a very cor rupt ms. in a very corrupt tradition

164

The peculiar errors in P are not completely similar to the errors in


the other mss. containing many errors, viz. VNvZa. Even though P
does present some strange errors that are perhaps best explained by
scribal difficulties

165

, the peculiar readings are not generally meaning-

less as in VNvZ . Most of them are ordinary kinds of errors

166

Even more importantly, P's relationship with the different

tradi-

tions is not the same all the way through the De sensu and the De memoria. Whereas the agreements with

r are regular from the beginning to


h are

the end of both texts, the agreements with the tradition around

frequent in the beginning of the De sensu until approximately 442a or

a
r are also found in 436a1-442a/b, it seems certain that P is conta-

442b, but hereafter it is almost non-existent. Since readings of both


and

minated, at least in this part of the text. This coincides suspiciously


with the change of textual tradition, or, in the case of N, of preferences,
that is also apparent in NVm

167

. The best an editor can do is probably

to notice that the textual tradition of P is not uniform all the way
through the De sensu and the De memoria, but is, perhaps, even more
corrupt and contaminated in the first part of the De sensu.
As regards the value of P, it cannot be questioned that the ms. has a
huge amount of peculiar readings which are almost all plain errors.
First, it is obvious that P contains a text that has been contaminated
with explanatory notes

168

. Second, P shows some traces of attempted

(164) Escobar, op. cit. (n. 3), p. 176, has found marginal comments by Kamar iotes (15th century) in two related mss. (Berol. 1507 and Mosqu. 240), stating that
their exemplar was very deteriorated. Escobar does not relate this exemplar to the
exemplar of P, but a damaged ancestor would certainly explain a lot.

(165) Cf. e.g. 438b25 ( e mpuroq P : e k puro`q ab), 440a15 (apori` an P : aporroi` aq
a r : apo`rroian hm m ), 442b17 (e pta P : e ti ta ab), 446a8 (an kai diadu`ointo P :
an kai dialu`ointo mYVC Mi : an kai dialu`oionto E : anadialu`ointo N : an dialu`ointo cett.), 450b15 (to atopon P : to apo`n ab), 453a3 (ote P : otan ab).
a

(166) Omissions : e.g. 437b25, 438a17, 439a23, 441a19, 442b23, 443a9, 444a23,
445b30, 446a23, 448a18, 450b2, 451b13, 453a6 -7. Words inserted : e.g. 436b17,
439b13, 440b10, 443b13, 445a8, 446a24, 447a12, 447b4, 447b19, 448b10, 448b12,
448b27,

451a8.

Word -order :

437a22,

438a10,

438b1,

439a21,

440a8,

440b26,

441b22, 442b1, 444a10, 445a21, 447a6, 448a21, 449a23, 451a3, 451b31, 453a13.
(167) It should be mentioned that until now no scholar has noticed the change
in readings for any of these mss. except for V, in which it is rather obvious. Perhaps,
the relation established between the mss. involved will be significant for other texts
as well. For instance, the relationship between P and m for the De partibus animalium
and the Historia animalium mentioned in note 135 above, might be explained through
this ms. connection. I have not, however, investigated their relationship further,
since it is not a significant factor for the De sensu and the De memoria.
(168) Cf. e.g. 436b17 (

mori` ou tyq vujyq pa`hoq

P:

pa`hoq

EYC i :

pa`houq

M:

david bloch

54

corrections ; some more interesting than others


peculiar

readings

do,

however,

present

169

. The majority of the

blatantly

false

readings,

changes in word-order and insertions of (explanatory ?) words. There


are also traces of Alexander's commentary in the text of P

170

The really interesting question is whether or not any of the peculiar


readings

constitute

independent

and

relevant

textual

evidence.

I believe this to be the case, but the evidence is hard to interpret and
open for discussion.
By far the most interesting passage for our purpose is the quotation
from Empedocles in

De sensu 437b26-438a3 on the faculty of sight. In

437b30-31 the other mss., with only a few variations in spelling, read :

pu r fwq b d e xw diahrwskon o son tanaw`teron yen


la` mpesken kata bylo n a teire` sin a kti`nessin

'

However, P has included

e xw

dia` ntatai trei`ato hespesi`ysin o ho` nsi


o ho` nsi

, although the copyist has indicated that

after

should be erased.

Thus, he reads :

pu r d e xw dia` ntatai trei`ato hespesi`ysin diahrwskon o son


tanaw`teron yen la` mpesken kata bylo n a teire` sin a kti`nessin

'

This is certainly not a normal intrusion into the text. An, admit tedly, corrupt verse completely in line with the rest of the fragment,
and in style and choice of words it seems thoroughly Empedoclean. As
it stands, it does not make sense, but F. Blass used the words, along

joan si
hespesi`sin

with

after

( !) found for

438a1

171

with

the

o ho` nsi
ai joa` nsi di`anta tetry` ato
in 438a1 in P, to restore a verse

text :

<

>

. The Diels/Kranz edition of the fragments has adopted

Blass' conjecture, and so have Forster and Ross for the

De sensu. Siwek

does not mention the conjecture and does not inform the reader that P
has this reading in 437b30.
Either the text in P is a corrupt version of the original Empedoclean
quotation, or it is the work of a very gifted copyist. I think it can be

mori`ou pa` hoq rh mori`ou tou e n y min


mo` non par a llyla par
a llyla ab
aisha` neshai AB aisha` neshai ab
aisha` neshai
aisheshai cett
e n t vuj r yte` on
AGB aisha` neshai ar
e n t vuj ab
a neu lo` gou jro` nou a neu jro` nou ab
e n tw pe` rati` e stin pe` raq y e n tw pe` rati` e stin y
tou to noomen tou noope` raq cett
tou noytikou tou nou me n
me` nou tou nooume` nou cett
a nti`lyvin dia` tinoq metaxu e styn e sty e sti cett
:

X), 440b1 (

), 448b10 (

Uvm m

P:

P:

).

P:

.), 450a13 (

P:

(170) Cf. 447a8 (

(171)

C Mi :

E :

.).

P:

E:

Y:

.),

In Sens. 135.5-8.
F. Blass , Zu Empedokles, in Jahrbu cher fu r Classische Philologie , t. 29, 1883,

where the reading in P stems from Alex.,

p. 19-20.

'

.), 448b27 (

P:

(169) Cf. e.g. 439a30 (


3

P:

), 448b12 (

), 450a8 (

E :

'

P:

55

aristotle's de sensu and de memoria

stated with certainty that if the verse was written by a copyist it must
have originated much earlier than P itself. I have found nothing
remotely similar in the entire corpus of scholia on the De sensu and the
De memoria. On the contrary, in this particular Empedoclean fragment
the scholiasts are very busy explaining words and phrases ; they do not
at all seem capable of adding to the text. Quite a few of the scholia
have been taken from Alexander's commentary (In Sens. 23.5-24.9),
which, incidentally, shows that the need to explain the difficult words
was felt already in A.D. 200. Furthermore, copyists usually do not
deliberately
explanations

expand
of

on

the

difficult

text

of

passages

Aristotle.

and

They

often

interpretations

of

provide
unclear

phrases, but these are not meant to be incorporated into the Aristote lian text. When this happens, and it often does, it is, apparently,
caused by plain mistakes or uncertainty as to whether a particular
phrase is part of the text or part of the scholia. Admittedly, the num ber of passages elaborated on by P is somewhat alarming, but whether
it was done consciously or unconsciously we cannot know.
The most suspicious element in this otherwise unknown tradition
which is, then, allegedly used by the textual tradition of P is perhaps
the fact that, aside from a few interesting readings in the Empedoclean
fragment, P has very few variants worthy of consideration. We may
wonder if this particular part of the text, that is, Empedocles' poem,
was independently available to a scribe. This would, necessarily make
the evidence for this part of the text, very old

172

. If this is so, we would,

however, have to agree with the number of scholars arguing against


the value of P, since the rest of the De sensu and the De memoria does not
produce anything similar

173

. But, on the other hand, Nussbaum has

pointed to an almost equally interesting passage in the De motu animalium (700b23-24) in which P alone has the correct reading, and Ross'
Appendix II in his edition of the De anima is also suggestive

174

. It is not

actually strange that P holds so few correct readings against the rest of
the tradition. It must be remembered that P is primarily a member of

b,

and the interesting readings are contaminations. We should not

expect them to be found on every page of the ms.


Therefore, I conclude that P, in addition to being part of

b,

also

includes readings from a textual tradition not represented by any of

(172) It is, perhaps, important that another peculiar reading of interest is also
found in the Empedoclean fragment (438a3 :

dii` eskon P : diahrwskon cett.). It was

accepted by Fo rster and Ross but rejected (rightly, I think) by Siwek.


(173) Cf. Siwek, Les manuscrits ... (n. 3), p. 129 -136 ; Harlfinger, op. cit. (n. 3),
p. 259-261 ; Escobar, op. cit. (n. 3), p. 205.
(174) Nussbaum, art. cit. (n. 3), p. 130 [NB ! 700a23 -4 in her text should read
700b23-4.] ; W. D. Ross, ed., Aristotle. De Anima, Oxford, 1961.

56

david bloch

the other early mss. This tradition can be designated g. This makes P
interesting but extremely difficult to use.
In the use
of P, a hitherto neglected source, . the Latin
(G2), might sometimes throw light upon the exemplar, but since
corruption was already present in the exemplar, and since the translations are difficult sources, the benefits are slight. Still, these translations
by William of Moerbeke from the 13th century contain a surprisingly
large number of readings that agree with P 175against the rest of b and a ;
.
this is especially true for parts of the
The translations being older than P, William must have used an
ancestor of P. However, the foundation of William's
work was not this
ancestor, but the older
(G1) from the 12th century.
William only used the ancestor of P as an aid in his work with these,
and therefore he only sporadically sides with P. The following passages
prove their relationship:
viz

translationes

novae

De sensu

translationes veteres

( .) 439b10, 439b13, 439b33, 440a1, 440b1, 440b3, 440b13, 441a4,


441a17, 441a19, 441a21, 441b7, 441b26, 441b27, 441b29, 442b11,
442b12, 443b6, 443b24, 444b5, 444b6, 445b30, 446a7, 446a17, 446a19,
446a24, 446b15, 447a5, 447a6, 447a9, 447a14, 447a25, 447b4, 447b6,
447b11, 447b15, 447b19, 448a21, 448a25, 448a26, 448b8, 448b10,
448b12, 448b13, 448b23, 448b25, 448b26, 448b27, 448b29, 449a1,
449a23, 449a28, 449b3, ( .) 450a8, 450b8, 451b12, 452b5, 452b13.
Sens

Mem

In these passages we will know with a fair amount of certainty what


was in the exemplar of P and William.
C
The results that have now been obtained are partly orthodox partly
controversial. As regards the a family, the conclusions are at least
partly controversial. Mugnier has until now been alone in thinking
that the whole tradition is derived from E, and even then he still used
onclusion

(175) I have made a full collation of the Latin


of the
using
a still unpublished edition prepared for the Aristoteles Latinus by Dr. Ludwig Peeters, which he generously provided me with. For the
of the
I have used the
221 (f. 21v-24r). For the
of the
and the
, I have used
, ed.,
t. XLV, 2:
, Roma & Paris, 1985. I am grateful to the late Professor Jozef Brams, and to Dr. Pieter de Leemans, Dr. Ludwig Peeters and in
particular Dr. Griet Galle for advice on the Latin translations.
translatio vetus

translatio vetus

Avranches, Bibl. municipale

De sensu

De memoria

de Aquino, Opera Omnia,

tus est De memoria et Reminiscentia

De sensu

De memoria

translationes novae

R. A. Gauthier

Sancti Thomae

Sentencia Libri De sensu et Sensato cuius secundus tracta

57

aristotle's de sensu and de memoria

M in his edition. Forster and Ross thought that M was an independent


ms., and Siwek thought that Y, C
Escobar argued for the

De insomniis

and M were all independent of E.


that Y and b were independent of

E. I have argued that Mugnier's overall idea was correct, although he


did not, for instance, correctly determine the position of C c in the
stemma.

As regards the b family, things are much more complicated. For the
r branch my reconstruction agrees with most scholars that L and H

are closely related, and that X and y are of the same tradition. The
stemma of the

h branch, on the other hand, has not hitherto been well


m basically agree with those of

appreciated. My conclusions as regards

other scholars, although I would like to stress the relative importance

of U. The reconstruction of

is, however, somewhat different from


a

that of other scholars. I have argued that VNC vZ , a group of very

corrupt mss., are all descended from a lost ms. ( ) ; the ms. m also
belongs partly to this tradition but also partly to the

r tradition, and it

is extremely contaminated. A stemma collecting these mss. in a group


has never, to the best of my knowledge, been drawn before. Siwek
thought C

and Z

could not be placed in a stemmatic relationship

with any other ms., and he regarded v as more closely related to m


than to any other ms. On the latter point he was followed by Escobar,
who did, however, recognize that Z

could well be placed in a stemma.

Neither of them, in my opinion, paid sufficient attention to the conta minated character of m.
Compared generally to the other major investigation of the textual
tradition of the

De sensu

and the

De memoria

, that is, Paul Siwek's inves-

tigation, my interpretation differs substantially. Siwek did not work


out a single, complete stemma. He established seven minor textual tra ditions, which, he thought, could all in principle transmit the correct
reading, and therefore an archetypal reading is often almost impossi ble to obtain. Contrary to this procedure, I have established a single
stemma in which the position of every extant ms. can be seen and
explained. This provides the editor of Aristotle with a tool to reach the
archetypal reading in most cases.
The

investigations

carried

out

above

have

focused

on

the

mss.

almost exclusively from the Aristotelian editor's point of view. In the


more speculative parts of this article, the argument may still be refuted
by studies in the circulation of the texts and studies in their influence
on the Byzantine cultural elite. I am, however, convinced that the gen eral

stemmatic

conclusions

are

sound,

and

stemmatic

conclusions

based on specific ms.-readings have been the primary purpose of this


article.

david bloch

58

In conclusion, it is clear from the ms.-evidence that a coherent


stemma, comprising all mss. dating from the 10th to the 14th centuries, can be established. Thus, we are very often able to determine the
archetypal reading of the entire tradition. No editor has, so far, based
his edition on such firm ground 176.
David Bloch

temma Codicum
w

E
q

e
m
d

Y
c

W
O

y
a

L H

Y
b

C
C

V
k

V N
M

i
v

ma

mb

(176) I owe thanks to Prof. Sten Ebbesen, Prof. Birger Munk Olsen, Dr. Griet
Galle and the anonymous reader for the Revue d'histoire des textes for commenting on
this article.

You might also like