Page 1 NATURE D. J. D . D. G. KENDALL

1118
,
NATURE
It was observed. that chicks which had boon reared with
the foam object often pecked at and pressed against it
to the accompaniment of contentment chirps, while the
others generally showed less positive activity when con·
tacting the box. This suggested that the foam provided
an intrinsically more B&ti.sfying object of atteohment
than the box. We tested this by comparing tho propor-'
tiona of the two groups which approached and which
contacted their original object in the test (for this analysis
the chick which approached both objoots was oxcluded as
furnishing ambiguous data). Of the 15 box-reared chicks,
6 approached tho box and 4 OOuched it; while of 16 fcamreared chicks, 12 approached and 11 contacted the foam.
The proportions test by Brown' shows theso differences
both in approach and in contact 00 be significant beyond
the 0·05 level.
It could be argued that this result was due 00 chance
differences in the general liveliness of the chicks, rather
than 00 differenoes in the strength of attachments to the
two objects, To test this we compared the mean latenoy
times of the two groups, excluding those aubjecte which
did not move at all. For box-reared chicks this was 2-7
min (range 0-8-4'7), for foam-reared it was 2·1 min (range
0'6--4.'8). Clearly there is no significant difference between
these means nor, in 80 far as may be inferred from latency
times, in the reootivitt'of the two groups..
"
Our main result supports the hypothesis that. ohicks
can form attachments of the kind classified as imprinting
to statio objects forming a part of their environment from
the beginning:' as well as to objects introduced into. and
so disturbing. their familiar world. Although this
attachment was demonstrated by our subjects in a novel
situation it must have been established previously in
their home environment. Nor can this be explained. in
terms of conventional reinforcement. since food and water
were availahlti" in the comers of the pens away from the
objects and temperature was oonstant in all areas.
It was noted, however, that tho chioks appeared. to
seek contact with the box: or the foam during the two-d.a.y
period. of rearing. They wore often disoovered sitting
pressed against the object or oven, in the case of the
foam, actually inside it. Consequently the normally
stationary objects did move when pushed. and it is possible
that both contact and the movement which this produood
had reinforcement value whioh increasod the likelihood
or the strength of the attachment. But chicks of this
.age are not particularly mobile, 80 that while movement
.of an object might occasionally have been produced.
by random floundering against it, systematic or frequent
movement would have been unlikely unloss the bird had
specifically sought contact with it. Movement would
therefore have been a concomitant rather than a cause of
imprinting. The fact that the foam elicited stronger
attachments thaD the box suggests that some variable
akin to Harlow's 'contact comfort" may have boon of
importance in our experimont. Further work on this
question is now in progress in our laboratory I. We
~hank Prof. S. G. Lee and Dr. W. Sluckin for edvice.
A...'!'iN TAYLOR
K. F. TAYLOR*
Department of Psychology.
University of Leicester.
• K. F. Taylor Is no..... in the Department of Psychology, Vnlvenitr of
Melbourne, Australia.
t Ram5ay, A. 0., .Auk, 88, 1 (1951).
• Thorpe, W. H., Lt.Gming and Indinet in .Animall (Methuen, London,
2nd ed., 1968).
• HellS, E. H •• ln Nebrad:a SIfmPorium on Motimtioll, edit. by Joncs, M. :R.
(Unlv. Nebraska ProIlS, Lincoln. Neb.• 1959).
'Oray, P. H .• Scienol:,lS2, 1884 (1060).
I Abercromble, B •• and lames, H., .Anim. BduJ"., 9, 205 (1961).
• Brown, 1., OocuP. Prvchol., 30. 169 (1956).
'Harlow, H. P., and Zimmermann, R. R., Scien«, 130, '21 (1959).
• Taylor. K. P., and Sluckln, W., Bull. Brit. P'IIthol. Sot., 17. No.5', IDA
(1964).
Epidemics and Rumollrl
Goffman and Newill 1 have directed. a
s.na.logy bstween the spreeding of an inf
and the dissemjnation of information. Wo
examined the spreading of a rumour from t1ii
of mathematical epidemiology and wish
briefly here on work to be published in detail
In particular. we must emphasize that a ~
model for the spreading of rumours can be
a number of different ways, depending on the
postulated to describe the growth and decay'
spreeding proceas. In all these models the
techniques familiar in mathematical epidemio
applied, but even the qualitative results so 0
not necesaarily be 88 expected. on the basis ,.
s.na.logy with epidemics.
•
To illustrate this point (but not to ad .
per se), let us divide the (finite, olosed,
mixing) population inoo three mutually
exhaustive classes:
-'
CIuI
X
y
Z
No.
z
~
ll'
Rumonr interpretation
Baa not heard. rumOW',
Aotlvely Ipreadlna: rumour
No IOn(~er 1Pread1Da rumour
If we B88ume that the only transitions are:
tiona' {x. Y. z) ..... (z-l. y+ 1. z), at a rate'
to XV; (2) 'removals' (z. y, z) __ (x, Y -1, z+ I),
proportional to Y t. then we obtain the
Karma.ck-McKendrick epidemic' in which an iJiitiil
of infection will generate an appreciable epidenihi 8DIl
only if, the initial number ",(0) of susceptibl.. ~'.
certain 'threshold' value p. In this formulation tbii
of spreading is due to a process of 'forgetting' ~
plausible hypothesis is that an active ' s p ~
telling the rumour because ho learns that it hl1!i
'news value'; if this happens as soon 88 he meets anotheir
individual knowing the rumour (that is, a member?Gf's or
another member of Y). then transitions from fY
.J
occur as a result of YZ and YY encounters (with a-double
transition in the latter case), and hence at a rate P-f'OIQ'.
tionaltoyz+2[}y(y-l)J = y(y+z-I).
~,
If this assumption is made, however. the behaviour of
the system is strikingly different from that of a KflnIWk_
McKendrick epidemic, and could not have been precIicW
from familiarity with the latter. The fraction! of tha]>Opo.
lo.tion which ultimately learns the rumour (correspondinc
to the fraction of the population which succumbs to the
disease) is always about 0,80, almost ~ctive of the
value of ",(0), so that the familiar threshold effeot hal
disappeared. A stochastic analysis yields very similar
results; we find that J (now a random variable) his an
expectation of about 0·80 and essentially a variance of
approximat<>ly 0'31/N, where ",(0) = N.
'Reluctance to tell stale news' can be incorporated into
the model in other ways. some of which we have investi·
gated. by similar methods. and the results are then cor·
respondingly modified.
However. the object of'thir
communication is not to urge the appropriateness of any
one model of rumour.tolling j but to stress the danger of
transferring formulm designed for epidemiological use
without a. thorough re·examination of the hypotheBeSt
which may not always be appropriate in the new situation.
0)0,
"*•
D. J. D
.
D. G. KENDALL
Statistical La.boratory,
University of Cambridge.
Goffman, W., and NewlD, V. A., NGlur~,204, 225 (1004).
Daley. D. 1., and Kendall, D. O. (unpublbhed).
• Bailey, N. T.1., TAt MIJt.'lnMticdl TMO"1I of BpidnniQ (GriftlD.IMM,;
I
t
--
.