William Sharp.

International journal of comparative psychology (Volume v4no3) online

. (page 1 of 5)
Online LibraryWilliam SharpInternational journal of comparative psychology (Volume v4no3) → online text (page 1 of 5)
Font size
QR-code for this ebook


ISSN 0889-3667
yCP8 4(3)193-236(1991)




Sponsored by the International

Society for Comparative Psychology

and the University of Calabria



Ethel Tobach

The American Museum of Natural History

Central Park West at 79th Street

New York, New York 10024-5192 U.S.A.


Nancy K. Innis

University of Western Ontario

London, Ontario

Canada, N6A 5C2


Pier Augusto Bertacchini
Dept. of Science Education
University of Calabria
87030 Rende Cosenza, Italy


Ruben Ardila Marco Poli

University of Colombia University of Milan

Apartado 88754 Via Francesco Sforza 23

Bogota, Colombia, S.A. Milan, Italy 20122

M. Balakrishnan
University of Kerala
Kariavattom 695581
Trivandrum, India

Abelardo Brenes
University of Costa Rica
San Jose, Costa Rica

David Croft
P.O.B. 1

University of New South Wales
Kensington, N.S.W. 2033

J.D. Delius

Ruhr Universitat Bochum
Universitat sstr. 150
1 Bochum D4630

Gary Greenberg

The Wichita State University

Wichita, Kansas 67208 U.S.A.

Robert N. Hughes
University of Canterbury
Christchurch, 1
New Zealand

S. Koref-Santibanez
Museum f. Naturkunde
43 Invalidenstrasse
Berlin 104

Kiyoko Murofushi
Ritsumeikan University
56-1 Toujiin-kita machi
Kyoto City, 603 Japan

Anthony Nijssen
Universitiet van Amsterdam
Post bus 20218
Amsterdam, HE 1000
The Netherlands

J. Martin Ramirez
Universidad de Sevilla
Sevilla, 41080

A.F. Semiokhina
Moscow State University
Moscow, 119899

Jiao Shao

Department of Psychology

Peking University

Beijing 100871

People's Republic of China

L.C. Simbayi

Department of Psychology
University of Bophuthatswana
South Africa

Joseph Terkel
University of Tel Aviv
69 978 Ramat Aviv
Tel Aviv, Israel

Charles W. Tolman
University of Victoria
P.O. Box 1700
Victoria British Columbia
Canada V8W 2Y2

D.F. Ventura
Instituto de Psicologia
University de Sao Paulo
Cx. Postal, Sao Paulo
(SP) Brazil


Volume 4, Number 3

Wavelength Categorization by Goldfish
{Carassius auratus) 195

Marjorie Goldman, Robert Lanson,
and Gabriela Rivera

Dominance Behavior in Asexual Gecko,
Lepidodactylus lugubris, and its
Possible Relationship to Calcium 211

Susan G. Brown, Linda K. Osbourne,
and Maile A. Pauao

Boxing in Red Kangaroos, Macropus Rufus:
Aggression or Play? 221

David B. Croft and Fiona Snaith

sored by the International Society for Comparative Psychology, an affiliate of the Inter-
national Union of Psychological Sciences and the University of Calabria, Italy. In con-
sonance with the goals and activities of the Society, it publishes reports of studies in
the evolution and development of behavior of all species; on the use of the comparative
method for the understanding of behavioral processes; and the examination of the ade-
quacy of psychological and evolutionary theories. It accepts articles that deal with his-
torical and contemporary relationships between humans and other animals; that eluci-
date problems of ecological and behavioral processes in resource management; that
explicate fundamental concepts about human evolution; and that present research re-
sults and develop theories about the development and evolution of behavior. Articles on
the contemporary and historical organization of knowledge about comparative psychol-
ogy; promoting public education on the evolution and development of behavior; and
about the training of students in comparative psychology are also invited. Book re-
views; letters discussing published articles with replies by authors; notices of new
books, journals, meetings and competitions or other matters of interest to the readers
will be published at the discretion of the editors. The Editor and Co-Editor are elected
by the Operations Committee of the Society. The Editorial Advisory Board is appointed
by the Editors and the Operations Committee.

MANUSCRIPTS should be submitted in triplicate to Ethel Tobach, Editor. See inside
back cover for style requirements.

SUBSCRIPTION inquiries and subscription orders should be addressed to the pub-
lisher at Subscription Department, Human Sciences Press, Inc., 233 Spring Street,
New York, N.Y. 10013-1578 or faxed to the Subscription Department at its number
(212) 807-1047. Subscription rates:
Volume 4, 1990-1991 (4 issues) $95.00 (outside the U.S., $110.00). Price for individ-
ual subscribers certifying that the journal is for their personal use, $36.00 (outside
the U.S., $45.00). Members of the International Society for Comparative Psychology
receive their yearly subscription at the rate of $24.00 (sliding scale) as part of their
membership dues. For membership information, write to Dr. Jeanette Ward, Depart-
ment of Psychology, Memphis State University, Memphis, TN 38152, USA.

ADVERTISING inquiries should be addressed to the Advertising Sales Representa-
tive, Daniel S. Lipner, Weston Media Associates, P.O. Box 1110, Greens Farms, Con-
necticut 06436— telephone (203) 261-2500 and fax (203) 261-0101.

PHOTOCOPYING: Authorizations to photocopy items for internal or personal use of
specified clients is granted by Human Sciences Press for users registered with the
Copyright Clearance Center (CCC) Transactional Reporting Service, provided that the
flat fee of $6.50 per copy per article (no additional per-page fees) is paid directly to
CCC, 27 Congress St., Salem, MA 01970. For those organizations which have been
granted a photocopy license by CCC, a separate system of payment has been arranged.
The fee code for users of the Transactional Reporting Service is: 0889-3667/91/$6.50.

COPYRIGHT 1991 by Human Sciences Press, Inc. Published quarterly in the Fall,
Winter, Spring, and Summer. International Journal of Comparative Psychology is a
trademark of Human Sciences Press, Inc. Second-class postage pending at New York,
N.Y., and at additional mailing offices. Postmaster: Send address changes to Interna-
tional Journal of Comparative Psychology, Human Sciences Press, Inc., 233 Spring
Street, New York, N.Y. 10013-1578.

ISSN 0889-3667 IJCPE8 4(3)193-236(1991)

International Journal of Comparative Psychology, Vol. 4, No. 3, 1991


Marjorie Goldman

Robert Lanson

Gabriela Rivera

Queens College of the City University of New York

ABSTRACT: Goldfish pressed a paddle for intermittent food reinforcement in the
presence of one of seven different monochromatic wavelengths. Wavelengths in 20 nm
steps from 430 to 690 nm, matched for "brightness," were then presented for 20 days
during which food maintained responding to the training stimulus. Generalization gra-
dients calculated from the final four days were asymmetric. A long wavelength gradi-
ent showed maintained responding above 630 nm; at short wavelengths responding
generalized below 490 nm; four middle wavelength gradients could indicate two group-
ings having maximum responses at around 510 and 570 nm.

The physiology of the goldfish visual system has been extensively
described (Wheeler, 1982). The three cone pigments are maximally
sensitive to wavelengths of 455, 530, and 625 nm (Marks, 1965;
Harosi and MacNichol, 1974). In the subsequent stages of retinal
processing there are cells which respond in an opponent fashion.
Some horizontal cells hyperpolarize and depolarize to different wave-
lengths (MacNichol and Svaetichin, 1958; Tomita, 1965); many bipo-
lar and amacrine cells are color-coded (Kaneko, 1973); there are gan-
glion cells with double-opponent receptive fields (Daw, 1968;
Spekreijse, Wagner, and Wolbarst, 1972; Beauchamp and Lovasik,
1973; Mackintosh, Bilotta, and Abramov, 1987); and single cells in
the optic tectum of the goldfish respond in an opponent manner (Jac-
obson, 1964). In addition to these physiological descriptions, what is
needed is an understanding of how this information is integrated and
used by the fish.

This paper examines how the various wavelengths are grouped
together by the goldfish. This issue has been explored in nonhuman
animals in two ways: matching-to-sample and generalization gradi-
ents. Wright and Gumming (1971) described "color-naming" gradi-
ents for pigeons using a matching-to-sample technique. After the pi-
Address correspondence to Robert Lanson, Department of Psychology, Queens Col-
lege of the City University of New York, Flushing, New York 11367.

© 1991 Human Sciences Press, Inc. 1 95


geons were matching three different wavelengths with 90% accuracy,
sample probes of wavelengths in between the original training stim-
uli were presented. The pigeons had to respond to nonmatching hues
on the side keys as if they were a match. The three functions derived
from the data had transition points at 540 and 595 nm. These group-
ings are different from human color naming functions where transi-
tions occur around 492, 561, and 605 nm (Boynton and Gordon, 1965).
The application of this method to goldfish is complicated by the fact
that on matching-to-sample tasks goldfish do not reach the high
levels of accuracy needed for reliable assessment of "color-naming"
transition values (Goldman and Shapiro, 1979).

Generalization gradients are obtained by presenting a series of
unreinforced wavelengths after the animal has been trained to re-
spond to one wavelength. Usually the animal is reinforced on a vari-
able interval schedule during training to ensure a steady response
rate. If tested in extinction, the animal gradually stops responding
during a single session so that few responses are made to stimuli pre-
sented at the end. To maintain responding during the testing phase,
Blough (1961) obtained steady state generalization gradients from his
pigeons by reinforcing responses during four of the six presentations
of the training stimulus which were interposed among many other
unreinforced wavelengths. The reinforced stimulus was changed ev-
ery few days so that over the course of the study each bird was suc-
cessively tested for generalization to a number of different wave-
lengths. Although the results showed the effects of previously
reinforced wavelengths and individual idiosyncracies, Blough found
consistent asymmetries in the shapes of the gradients. A symmetrical
gradient shows responses tapering off equally at both higher and
lower wavelengths around the training stimulus, suggesting that the
training stimulus is located in the middle of a color category. An
asymmetrical gradient shows a rapid decline in responding on one
side of a specific training stimulus and a more gradual decline on the
other side, suggesting that the training stimulus is located closer to
one color boundary. For all of Blough's birds, a rapid decline occurred
at 540 nm, in agreement with the data from Wright and Gumming,
while two out of four birds showed a steep gradient at 590 nm. Em-
merton (1983) has summarized results from wavelength discrimina-
tion experiments on pigeons and has shown a high degree of corre-
spondence among data obtained with steady state generalization
gradients, matching-to-sample, and hue discrimination procedures.

Two wavelength generalization gradients have been obtained
from goldfish by Yarczower and Bitterman (1965) without controlling
for "brightness." One curve shows an almost symmetrical gradient
around a 550 nm training stimulus with responses slowing at 490 and
610 nm. The other curve is an asymmetrical gradient around a 580


nm training stimulus where the peak is shifted to 560 nm, responding
is still high at 520 nm, and there is a steep decline at 620 nm. The
small number of gradients and the narrow range of spectral values
explored make it difficult to assess color categorization in the gold-
fish. In the present experiment the fish were tested using a reinforced
generalization procedure similar to Blough's, but more training and
testing sessions were added to try to lessen the effects of previous
wavelength training. We used seven different training wavelengths
equated for "brightness" and the testing stimuli ranged from 430 to
690 nm to explore more of the spectrum.



Ten goldfish, 8-12 cm standard length, purchased from a local pet
store, were housed individually in 9.6 liter tanks (31 x 16.5 x 20.5
cm) continuously aerated through plastic filters. The room tempera-
ture was maintained at 21° C. Fluorescent room lights were on for 18
h, beginning at 7 am, and off for 6 h. The fish were fed once daily
during testing.


The fish tank, with filter removed and debris siphoned, was
placed into a black Plexiglas chamber through a hinged side door
along the 31 cm side. A single Plexiglas disc, 3.1 cm in diameter,
suspended on a steel rod could be lowered into the tank in front of a
single hole in one 16.5 cm wall by closing a top lid. The steel rod was
suspended from a mechanical relay contact. A sheet of painted black
metal extended 8.5 cm below the lid to prevent the fish from hitting
the rod. Food pellets were delivered from a tube by a Gerbrands
feeder through a 2.5 cm hole 4.75 cm from the black sheet. Mouth
press responses to the Plexiglas disc closed the relay contact and were
recorded on relay equipment.

Light from a GE tungsten ribbon filament 6 volt bulb passed
through heat absorbing glass and was then projected through a Dif-
fraction Products Czerny-Turner grating monochrometer. The mono-
chromatic output could be intercepted by a shutter mounted on an
electromechanical positioning motor, which controlled the presenta-
tion of the visual stimuli with onset and offset of less than 100 ms.
The unblocked monochrometer output was brought to a focus. A Ko-
dak continuous 15 cm diameter circular neutral density absorbing
wedge with a mechanical compass along the outer edge was placed in


the focal plane so that the intensity values could be set manually.
Light passing through the filter was focused on a ground glass
mounted in front of the 3 cm circular hole at the end of the experi-
mental chamber. Fourteen wavelengths (half-band width = 16 nm)
ranging from 430 to 690 nm in 20 nm steps were used in testing.
These values were manually adjusted between stimulus presenta-


The wavelength vernier of the monochrometer was calibrated
several times during the course of the experiment with a mercury
vapor lamp at 546 and 579 nm. Stimulus intensity values were deter-
mined based on an average relative spectral sensitivity function as
determined for Yager's (1967) data for light adapted goldfish. The
points were connected and interpolated values for the wavelengths
used were determined. Stimulus value determinations and calibra-
tions were done with an EG and G Model 580/585 radiometer with a
photomultiplier head. A computer program determined the radio-
metric output inversely weighted by the fish sensitivity function (tak-
ing into account the spectral sensitivity of the photomultiplier as cali-
brated by EG and G against a National Bureau of Standards
standard). Given the relatively high fish sensitivity in the blue spec-
tral region and the lower output of the tungsten source in that region,
an attempt was made to achieve the highest intensity output possible
from the system with all stimuli equated for the fish sensitivity. A
maximum intensity output was determined for the lowest wavelength
value used and the computer determined the outputs required for all
other wavelengths. These values were manually set by positioning
the wedge to produce the closest approximation to the computer value
within accuracies of 3 percent. Radiometric energy levels at 450, 530,
and 630 nm were 11.5, 12.0, and 12.2 log quanta/s/square centimeter
respectively. These values are comparable to those used by Powers


Each fish was shaped to press the lighted disc with its mouth for
Noyes formula "J" 20 mg fish pellets. The disc was transilluminated
with one of the testing wavelengths at the appropriate intensity set-
ting. Three fish were initially trained in the presence of 450 nm,
three with 530 nm, three with 570 nm, and one with 630 nm. This
monochromatic light was the only illumination in the box. After
shaping, a random interval (RI) schedule of food reinforcement was
instituted and the mean time interval was gradually lengthened until


its final average value of RI 133 s. In this schedule, the first mouth
press in a 2 s repeating time period has a .015 probability of earning
a Noyes pellet. Subsequent responses in that period are not rein-
forced. Stimulus light on periods were gradually reduced from 10 min
to 2 min. A 15 s blackout followed all stimulus presentations. By the
end of approximately 40 days of training, each fish responded at a
steady rate throughout the hour and earned about 20 food pellets.
Generalization testing involved 2 min presentations of the 14 differ-
ent stimuli so that each stimulus was presented three times over two
days. Reinforcement was programmed to occur on an RI 27 s schedule
during six additional presentations of the training stimulus on each
day. This schedule permitted approximately the same daily number
of food pellets as in the final training conditions. Thus, an example of
one day's testing would be 27 light on periods which included 6 of the
training stimulus with reinforcement possible, one or two with the
training stimulus and no reinforcement possible, and the other 19
periods with the other 13 stimuli presented once or twice, without
reinforcement possible. The order of stimulus presentation was deter-
mined by random permutation. In the second half of the experiment,
a constraint was added that a given wavelength could not follow the
reinforcement period more than once. Testing continued for either 20
or 25 days, after which most of the fish were reinforced on an RI 133 s
schedule for responding to a new training stimulus. This second
training stimulus was presented alone for 10 or 20 days before the
generalization testing phase was begun and carried out in the same
manner as described for the first training stimulus. After generaliza-
tion testing, some of the fish were trained and tested on a third stim-
ulus. The following sequences of training stimuli were used for differ-
ent groups of fish: 450, 530, 630 nm; 530, 590, 490 nm; 570, 510 nm;
and one fish was trained to only 630 nm to replace an animal from
the first group that had died. For the groups begun with 530 and 570
nm, training time to the second and third stimulus was reduced from
20 days to 10 and testing time for all stimuli was reduced from 25
days to 20 because the shorter time was sufficient to obtain peaked


The first gradients obtained after single stimulus training to 450
nm, 530 nm, and 630 nm were flat across the wavelength spectrum.
The fish trained to 570 nm and one fish trained to 450 nm showed
peaked gradients from the first two generalization sessions. Many of
the first gradients for the second and third training stimuli showed
responses to the previous stimuli. By the 15th to 20th generalization




20 —

15 —


5 —

"1 1 I 1 I I I I I I I I — I — I — I — I — I — I — I — I — 1 — I — I — I — I — I — I — I I I
400 450 500 550 600 650 700

FIGURE 1. Generalization gradients after training to 450 nm
(dashed lines) and 490 nm (solid lines) for six different fish from test-
ing days 18-21 (450 nm) or 17-20 (490 nm).


20 —

15 —

10 —

5 —

I I I I I I I I I I I I I I [ I
400 450 500 550


600 650

Y " T I Tf


FIGURE 2. Generalization gradients after training to 490 nm (solid
lines) and 510 nm (dashed lines) for six different fish from testing
days 17-20.




28 —


10 —

5 —

I I I I [ I I I I I I I I I I I I I I I I I '■ I '''' I
400 458 588 558 688 658 788


FIGURE 3. Generalization gradients after training to 530 nm from
testing days 17-20 for five fish. The dashed gradients are from two
fish for which 530 nm was the second training stimulus; the solid
gradients are from three fish for which 530 nm was the first training

testing session, the gradients peaked at the current training stim-

For consistency of comparison across the experimental condi-
tions, the data from sessions 17-20 (or 18-21) for stimulus presenta-
tions without reinforcement will be presented. The animals first
trained to 450 nm were given a single generalization session with a
restricted range of wavelengths, to see how flat the initial gradient
would be. Subsequent generalization sessions produced two day gradi-
ents, thus necessitating the use of days 18-21 for compatibility. Each
two day gradient was computed by dividing the sum of the responses
in the three unreinforced presentations of each wavelength by the
total number of responses to stimuli without reinforcement on those
days. The last two gradients, days 17/18 and 19/20, were then aver-
aged and are presented in Figures 1-5. Figures 1, 2, and 5 each show
individual generalization gradients for six different fish; Figure 4
shows gradients for the same three fish after training to one wave-
length, obtaining a gradient, and retraining to a second.

The data from all three fish trained to respond to a 450 nm light
transilluminating the Plexiglas disc and all three fish trained to a
490 nm light are presented in Figure 1. Although the functions peak



"1 — I — I — I — I — I — I — I — I — r
480 450 500


1 — I — I — I — I — I — r~i — I — I ¥"T I ^i"^!*
550 600 650 700

FIGURE 4. Generalization gradients after training to 510 nm
(dashed lines) and 570 nm (solid lines) for the same three fish from
testing days 17-20.

around their respective training stimuli, they have essentially the
same shape; the animals respond to wavelengths lower than 490 nm.
At wavelengths above 510 nm responding decreases and after 610 nm
stays below five percent. This similarity occurs even though 490 nm
was the third training stimulus after 530 nm and 590 nm for one
group of fish, while 450 nm was the first training stimulus for the
other fish.

Figure 2 shows the same three gradients to the 490 nm training
stimulus together with gradients from all three fish trained to re-
spond to a light of 510 nm. There is a clear separation of the 490 nm
and 510 nm functions. At wavelengths below 470 nm responding by
fish trained to 510 nm declines below five percent, while responding
by the fish trained to 490 nm remains around 10-15 percent of total.
Even the fish that generalized to 490 nm from the 510 nm training
stimulus shows a steep decline in responses below 490 nm. The re-
sponding of the 490 nm animals declines to wavelengths above 510
nm, including one fish that generalized to 510 nm. The animals
trained to 510 nm show generalization to all wavelengths between
530 and 590 nm. Two of these fish had secondary peaks around 570
and 590 nm, which could have been residual responding from pre-
vious 570 nm training.




20 —

15 —

10 —

5 —

I I I I [ I I I I I I I I I I I I I I I I I I I [ I I I I I
400 450 500 550 600 650 700

FIGURE 5. Generalization gradients after training to 590 nm (solid
lines) and 630 nm (dashed lines) for six different fish from testing
days 17-20.

Figure 3 shows the generalization gradients from all five fish
trained to respond to a light of 530 nm. The two fish trained to 530
nm after 450 nm (dashed lines) seemed to show some residual re-
sponding below 490 nm and had some responding to all wavelengths
lower than 630 nm, yielding a flat gradient; so a new group of three
fish were given 530 nm as their initial training stimulus. Their three
gradients (solid lines) also are rather broad and flat between 490 and

1 3 4 5

Online LibraryWilliam SharpInternational journal of comparative psychology (Volume v4no3) → online text (page 1 of 5)