Content uploaded by Robert Jackson
Author content
All content in this area was uploaded by Robert Jackson on Apr 23, 2024
Content may be subject to copyright.
Spider Cognition
Robert R. Jackson*
,†
and Fiona R. Cross*
,†
*School of Biological Sciences, University of Canterbury, Christchurch,
New Zealand
†
International Centre of Insect Physiology and Ecology (ICIPE),
Thomas Odhiambo Campus, Mbita Point, Kenya
1 Crossing the cognitive line 116
1.1 A day in the life of a spider in the field 116
1.2 A day in the life of a spider in the laboratory 118
1.3 Representation 120
1.4 Specialization and preferences 121
2 Paying attention 124
2.1 Selective attention 124
2.2 Perceptual learning 126
2.3 Innate search images 127
2.4 Capacity limitations and trade-offs 128
3 Perception 129
3.1 Classification and objects 129
3.2 Spiders that see things 132
3.3 The unseen world of the spider 136
4 Communication 137
4.1 Mind games in spider webs 137
4.2 Within-species mind games 143
4.3 Mate choice goes cognitive—Again 147
5 Cognition via chemistry 148
5.1 Flexible living through chemistry 148
5.2 Olfactory search images 149
5.3 Cross-modality priming of selective attention 152
6 What is it like to be a spider? 154
Acknowledgements 156
References 156
Abstract
Spiders, having minute brains, were once considered simple, instinct-driven automatons,
but research on spider biology is revealing increasing evidence of their cognitive
abilities. In this review, we discuss the complex, flexible behaviour of spiders, especially
salticids, and highlight how sometimes the cognitive character of spider behaviour
closely parallels that of much bigger animals. This includes the use of selective attention
ADVANCES IN INSECT PHYSIOLOGY VOL. 41 Copyright #2011 by Elsevier Ltd
ISBN 978-0-12-415919-8 All rights of reproduction in any form reserved
DOI: 10.1016/B978-0-12-415919-8.00003-3
(both visual and olfactory) and the use of planned detours. The implications of these
findings, and how they relate to bigger issues traditionally associated with big-brain
animals, such as ‘representation’ and ‘mind’, are discussed. Also discussed are issues
relating to animals, including spiders, having a preference, instead of a search image, for
a particular type of prey, and issues relating to spiders classifying different types of prey.
Some of these issues are illustrated by exploring how spiders communicate and play
mind games with their prey, as well as with potential mates. We also discuss how much
about cognition can be revealed by exploring the perceptual systems of spiders.
1 Crossing the cognitive line
‘Cognition’ in everyday language is a word for something to do with thinking—
something to do with minds. Our common sense may tell us that mental things
go with big brains, mammals or even people alone. How would Descartes
(1637/1994) have reacted seeing the title ‘Spider Cognition’ in a scientific
journal? For many people, a first reaction might be to ask whether we are
being facetious. Do we have anything to review?
There are, however, many ways in which our understanding of spider biology
interfaces with topics related to cognition. We could begin by saying precisely
what we mean by ‘cognition’, but we prefer not to start with a definition at all.
In fact, the style of writing we adopt in this review may seem unconventional.
If you are expecting something like a textbook style, you will be disappointed.
Our goal is more like a casual guided tour through topics related to cognition
drawn from things we have learned about spider biology.
1.1 A DAY IN THE LIFE OF A SPIDER IN THE FIELD
Let us start by comparing Portia, a spider with a brain that would sit comfort-
ably on a pinhead, with an African lion. Next to a lion’s brain, Portia’s brain
seems to vanish into insignificance. Members of a lion pride work together
when hunting for their prey. Let us suppose we are watching a pride closing in
on a zebra herd under the cover of tall grass. One of the lions separates itself
from the rest of the pride and wanders off in another direction, going around
some boulders, completely losing sight of the zebras and then finally coming out
on the other side of the herd. Having reached this vantage point by a circuitous
route (a detour), the lion then roars and rushes towards the unwary zebras that
now panic and gallop towards the rest of the pride still hiding in the grass
(Schaller, 1972). What is going on here? Did the lions act on a plan they made
ahead of time? In particular, was the circuitous path one of the lions took around
the herd a pre-planned detour? Planning sounds like thinking, but a conclusion
like that cries out for experimental evidence from carefully controlled experi-
ments based on good sample sizes. However, for a lion researcher, tightly
controlled experiments and large sample sizes sound like dreaming.
116 ROBERT R. JACKSON AND FIONA R. CROSS
Now we can look at what happens when Portia goes hunting. Portia is a
genus of about 20 species from the spider family Salticidae. ‘Jumping spider’ is
the accepted common name in English for salticids. There are more than 5000
described species in this family (Platnick, 2011; Pro
´szynski, 2011), and jumping
is something many of them do exceptionally well, but what makes the jump of a
jumping spider particularly distinctive is that it is usually aimed with remark-
able precision at a target (Hill, 2010; Parry and Brown, 1959). Salticids have big
forward-facing eyes, and their eyesight is based on a level of spatial acuity that
would rival a cat’s (Harland and Jackson, in press). When we look at a salticid, it
stares back with those big front eyes. These eyes, coupled with an alert and
inquisitive demeanour, encourage us to envisage a salticid as a miniature cat
(Harland and Jackson, 2000a; Land, 1974). Even the salticid’s prey-capture
method often seems feline. The typical prey of a salticid is an insect, and most
salticids, instead of using a web, adopt a routine based on stealth and stalking
(Forster, 1982a,b; Richman and Jackson, 1992). With its body low to the
ground, a salticid slowly stalks a fly or another insect, pauses when close and
then, with a well-aimed leap, pounces on its unwary prey.
However, Portia is different, being a salticid that actively prefers other
spiders as prey (Jackson and Wilcox, 1998). Another spider is also a predator
and often fully competent at making a meal of an arthropod of Portia’s size. For
Portia, planning what to do ahead of time might seem advisable, but can a
spider plan ahead?
As with the lion, let us suppose we are watching Portia in the field. For this,
we go to an Australian rainforest where Portia fimbriata has encountered an orb
web on a tree trunk. The spider in the web is Argiope appensa (Fig. 1). We may
think it would be easy enough for Portia to simply walk up the tree trunk and
step on to the web, but A.appensa has a particularly effective defence. Upon
detecting an intruder, A.appensa violently shakes its web, sometimes with
enough force to sling Portia away to the forest floor (Jackson, 1992a; Jackson
et al., 1993). However, as we watch, Portia stops on the tree trunk below the
web, faces one way and then another and then walks away. At first, we think
Portia has given up. However, we then notice Portia taking a circuitous route
around the tree trunk and through the vegetation, losing sight altogether of the
prey spider and its web (Jackson and Wilcox, 1993a). Eventually, Portia moves
out on to a vine directly above A.appensa’s web. Nothing happens in a hurry,
but Portia finally fastens a silk line and slowly eases down suspended by this
thread alongside, but not touching, A.appensa’s web. When level with A.
appensa on the web’s hub, Portia begins rocking the suspension thread, even-
tually swinging in and grabbing the resident spider (Jackson, 1992a). As Portia
avoided contact with the web until the moment of attack, A.appensa never saw
the predator coming. However, as A.appensa, like most web-building spiders,
has only rudimentary eyesight (Homann, 1971; Land, 1985; Land and Nilsson,
2002), perhaps we should say ‘‘felt’’, not ‘‘saw’’.
SPIDER COGNITION 117
What is going on here? The same question we asked about the mighty lion
can now be asked about Portia, a mere spider. Did we see a predator act on a
plan made ahead of time? Now we come to an important difference. It is a lot
easier to take Portia to the laboratory, standardize each individual’s prior
experience, carry out experiments that are designed to demand planning ahead
and base each experiment on large numbers of individuals, each tested only
once (Tarsitano, 2006; Tarsitano and Andrew, 1999; Tarsitano and Jackson,
1992, 1994, 1997; also see Heil, 1936; Hill, 1979).
1.2 A DAY IN THE LIFE OF A SPIDER IN THE LABORATORY
Here is an example of an experiment designed to determine whether Portia can
plan ahead (Tarsitano and Jackson, 1997). We put Portia on top of a central pole
(Fig. 2) with a surrounding array of ramps and other poles creating two
convoluted routes that Portia can see from its elevated perch. One route leads
to a target (a lure made by mounting a dead spider in life-like posture on a cork
disc) and the other route is a dead end. Distances between Portia and the
surrounding array are too far for Portia to cross simply by leaping. The only
way Portia can reach its prey is to walk down to the floor and over to one of the
two poles, climb it and then follow the path to the prey. However, once on the
floor, Portia can no longer see the prey. This has important implications, as it
FIG. 1 Argiope appensa, a prey spider frequently encountered by Portia fimbriata in
Australian rainforests. A. appensa is exceptionally capable of defending itself when it
detects an intruder on its web. P. fimbriata’s solution is to avoid moving directly on to
A. appensa’s web and instead to take a detour by which it reaches a vantage point above
the web. From above the web, P.fimbriata then lowers itself down on a line of silk and
attacks A.appensa at the hub without first touching the web.
118 ROBERT R. JACKSON AND FIONA R. CROSS
implies that, in order to reach its prey, Portia must use a plan made while
viewing the two paths from on top of the central pole.
One of the important things to realize is that, in the field, the details of the
detour route leading to prey will vary, being virtually unique to the particular
time and place of the encounter (Jackson and Wilcox, 1993a). The situation in
the laboratory was similar because the paths were configured in different ways
and, as the same individual Portia was never tested more than once, every
individual used in an experiment was confronted with what was for it a novel
path-finding problem (Tarsitano and Jackson, 1997). Portia’s task always came
down to choosing which one of two paths would lead to the prey. Sometimes
Portia had to walk past the wrong pole before reaching the correct pole, and
sometimes Portia had to head directly away from the prey before accessing the
correct pole. Yet, regardless of the details and despite having no prior experi-
ence of taking or even seeing the paths available in the experiments, Portia
chose the correct pole significantly more often than the wrong pole.
Using casual language, we might say that Portia identifies the problem (how
to reach its prey), thinks about it, understands the solution and acts. If that is not
provocative enough, we might say Portia makes up its mind. Regardless of how
Lure on
ramp
Incorrect
pathway
No lure
on ramp
Central pole
Correct pathway
to prey
FIG. 2 Apparatus used for testing Portia fimbriata’s detour-planning ability. Portia
was on top of the central pole before each test began. The prey item (lure made by
mounting a dead spider in life-like posture on a cork disc) (not shown) was on one of the
two ramps (whether on the left or the right ramp decided at random). Portia viewed
the prey while on top of the central pole but could not see the prey when it went down the
pole. By consistently taking the route that leads to the prey, Portia demonstrates ability to
plan ahead.
SPIDER COGNITION 119
we say it, Portia’s behaviour in these experiments seems to lie squarely in the
domain of cognition, but maybe it is time to define ‘cognition’.
1.3 REPRESENTATION
Defining cognition is notoriously difficult, but representation has often been
envisaged as a key attribute at the boundary between what does and does not
qualify as cognitive (Damasio, 1994; Maunsell, 1995; Markman and Dietrich,
2000; but see Epstein, 1982). Although this can look like substituting one hard-
to-define term for another, implications about representation make it hard to
deny that Portia’s detour-taking behaviour is cognitive. After leaving the
central pole, Portia seemed to be guided by a representation of what it could
no longer see.
It is tempting to think about representation primarily in the context of vision,
and this, in turn, may tempt us to equate representation with something like a
picture in the animal’s head—a mental picture, or imagery, in an animal’s mind
(Kosslyn et al., 2003; Neiworth and Rilling, 1987; but see Pylyshyn, 2003a,b).
Before leaving the central pole, did Portia really render a mental picture of
where to go in order to reach the prey, and then use this mental picture like a
map? Surely, this interpretation is excessively literal. We need a concept of
representation that will be more basic and not unique to vision. Representation
might be something more like an internal state that carries information and is
then put to use during decision making. A key idea is that representations are
used for processing that happens several steps removed from simple stimulus–
response chains (Terrace, 1985).
Defining ‘mind’ may be even harder than defining ‘cognition’. Minsky’s
(1986) simple way out of this impasse was his famous saying, ‘‘Minds are
simply what brains do’’ (p. 287). However, these days, cognition is often
equated with ‘information processing’ (Dukas, 2004; Shettleworth, 2009), a
definition with simplicity to match Minsky’s definition of ‘mind’. These defini-
tions may often be all we need but only if we are careful to think of mind and
cognition as being regions on a continuum, rather than being qualitatively
distinct things. At one end of the continuum, there may be animals that are
restricted to behaviour that corresponds roughly to Descartes’ (1637/1994)
‘automatons’, or what we might nowadays call ‘stimulus–response machines’.
These animals may be of little interest to us in the context of cognition or minds.
Nor do we have much need for the term ‘representation’ when an animal’s
behaviour is more or less automatic, like a simple, direct response to a stimulus.
Moving towards the other end of the continuum, however, we encounter
information processing that seems more and more elaborate. Imperceptibly,
we cross into a region on the continuum where many of us will accept that the
words ‘cognitive’ and ‘mind’ are clearly appropriate (see Dennett, 1996).
Worrying about where to draw a line on an automaton-cognition continuum
may be a waste of time, but the idea of a spider, with a minute brain, lying
120 ROBERT R. JACKSON AND FIONA R. CROSS
anywhere near the cognition end seems to defy common sense. Instead, com-
mon sense tells us that advanced cognition requires a big brain (Eberhard,
Chapter 4 of Volume 40). ‘Advanced cognition’ may be unfamiliar territory
when discussing spiders. Perhaps we should start with more familiar topics that,
despite their familiarity, relate to cognition.
1.4 SPECIALIZATION AND PREFERENCES
At first glance, discussing predatory specialization and preferences may seem a
lot less contentious than discussing spider minds, but maybe it appears that way
because we too often forget what these topics are about. For illustrating this, we
can return to Portia.
Portia’s preference for spiders as prey is unusual, and preference for spiders
is part of what makes Portia the most striking and thoroughly studied example
of something we call ‘specialized araneophagy’. We can shorten this to ‘ara-
neophagy’, with ‘specialized’ being implicit, but we need to emphasize that
‘araneophagy’ does not mean simply that Portia eats other spiders. Nor does
‘specialized’ mean anything related to trade-offs or restrictions. This is very
important. Portia’s prey-capture behaviour (tactics) and Portia’s preference are
both ‘specialized’ in an everyday sense of the meaning of the word ‘specialized’
(i.e. traits that are well designed for a specified task). We can say that, as a
predator, Portia is a spider specialist, meaning that this predator is especially
well adapted with respect to exploiting this kind of prey. Whether specialization
with respect to preying on any one prey type is accompanied by adaptive trade-
offs that somehow restrict a predator’s proficiency with respect to preying on
any other prey type is a completely different issue. Trade-off hypotheses must
be tested on a case-by-case basis (Harland and Jackson, in press; Nelson and
Jackson, 2011). These hypotheses are not foregone conclusions, and ‘restricted
diet’ is certainly not part of the definition of ‘specialized’ (Fry, 1996; Huseynov
et al., 2008). It is also not just automatically true that the jack-of-all-trades is the
master of none (Whitlock, 1996). Often the evidence is clearly contrary to any
such hypothesis (Jackson and Hallas, 1986a), and there are many examples of
predators being poly-specialists (West-Eberhard, 2003).
The terms ‘specialized’ and ‘generalized’ are of especially limited use, and
often seriously misleading, when used solely for specifying the range of prey in a
predator’s natural diet (Berenbaum, 1996; Futuyma and Moreno, 1988; Wilson
and Yoshimura, 1994). For considering natural diet, ‘monophagy’ and ‘polyph-
agy’ are more appropriate terms, but it is better to avoid the way these terms
suggest a dichotomy and instead consider a stenophagy–euryphagy continuum,
where ‘stenophagy’ refers to a narrow range and ‘euryphagy’ refers to a wide
range of prey types in a natural diet. These terms encourage a more realistic
expectation of variation in diet breadth being on a continuum. However, the
concept of ‘diet breadth’ raises serious but rarely acknowledged issues, and some
of these issues are related to cognition. Suppose we want to decide where a
SPIDER COGNITION 121
particular predator lies on a euryphagy–stenophagy continuum. Whose classifi-
cation scheme should we use? When and why should we refer to prey species,
genera, families and so forth? Using formal scientific taxonomy may tell us
something interesting in the context of community ecology, but the predator’s
own classification scheme is something cognitive that we can know only on the
basis of appropriate experiments (e.g. Harland and Jackson, 2000b, 2001).
The problems only get worse when data on natural diet are used as a basis for
statements about preferences and choices. The cognitive implications of pre-
dators expressing preferences for particular types of prey have often been
overlooked and sometimes actively deflected. For example, it has become
commonplace in ecology to conflate three distinctively different issues: a
predator’s diet, its choices and its preferences (e.g. Hughes, 1993; Manly,
1974; Roa, 1992). Lockwood (1998), for instance, said ‘‘the relative consump-
tions of different food types’’ correspond closely ‘‘with our intuitive definition
of ‘preference’’’ (p. 476). Perhaps what is ‘‘intuitive’’ in ecology is different,
but our intuition is that an animal’s preference is what it would like to eat and
that this allows for the possibility of an animal’s diet (what it actually does eat)
being different from its preferences (‘you can’t always get what you want’).
By using the term ‘preference’, we acknowledge that certain kinds of prey are
identified by the predator as being especially salient. ‘Choice’ is the appropriate
word for a type of behaviour that is driven by preference (Huseynov et al., 2008;
Morse, 1980).
Although data on a predator’s natural diet may suggest hypotheses about
preferences and although these hypotheses can be used for predicting the
choices a predator will make in experiments, data on diet alone cannot reveal
what a predator chooses or what it prefers. For example, a predator’s preferred
prey may be difficult to locate or to capture. Being a cognitive attribute inherent
to the predator, preference can be demonstrated only by carrying out appropriate
experiments (Huseynov et al., 2008; Nelson and Jackson, 2011) and, when we
demonstrate a predator’s preferences, we learn something about the predator
that is distinctively different from learning about its natural diet.
Owing to their ability to see in fine detail (Harland and Jackson, in press) and
the intricate vision-guided predatory tactics (Forster, 1982a; Gardner, 1964;
Jackson and Pollard, 1996) by which they can demonstrate their prey-choice
decisions from a distance, jumping spiders (Salticidae) have been especially
cooperative subjects in experimental studies designed to determine predatory
preferences. Another advantage of using salticids in research on preferences is
that these are predators that will respond readily to lures made by mounting dead
prey in life-like posture on cork discs (Jackson and Tarsitano, 1993; Jackson
et al., 2005) and to virtual prey generated by computer 3D animation software
(Harland and Jackson, 2002; Nelson and Jackson, 2006) in much the same way
as they respond to living prey.
When experiments are based on staging encounters between predators and
living prey, the prey’s response to the predator introduces potentially confounding
122 ROBERT R. JACKSON AND FIONA R. CROSS
variables that make interpreting results difficult. By using lures or virtual prey
instead of living prey, much tighter control of these variables can be achieved,
including the elimination of any differences in how effectively different prey types
defend themselves. A variety of testing protocols have consistently shown that
Portia has a strong preference for spiders as prey (Li and Jackson, 1996a). Not only
does Portia choose a spider significantly more often when it is presented alongside
an insect, but Portia also accepts spiders more often than insects when prey is
presented one at a time on successive days. Moreover, Portia will often drop an
already captured insect in order to grab hold of a spider, but almost never drop a
spider to grab an insect (Li and Jackson, 1996b; Li et al.,1997). Having a
convergence of findings from different experimental methods gives us confidence
that an underlying cognitive attribute of these spiders is being detected, this being
something that data on diet alone could never reveal.
Although their eyes do not achieve the spatial acuity known for salticids, some
lycosid spiders make considerable use of vision, and a few species have been tested
successfully with virtual prey generated by computer animation (Persons and
Uetz, 1997; Uetz and Roberts, 2002). However, most spiders have only rudimen-
tary eyesight and do not respond to lures and computer animation at the level of
accuracy achieved by salticids. This means that, for most spiders, it will be
especially difficult to control for the effects of prey behaviour in preference
tests. This drawback notwithstanding, ample evidence suggests that distinctive
preferences for particular prey are widespread across spider families, despite the
way a routine characterization of spiders as ‘generalist predators’ (Nentwig, 1987;
Riechert and Bishop, 1990; Wise, 1993) encourages a contrary expectation.
For example, araneophagy is far from unique to Portia. Besides having been
demonstrated experimentally for species from the salticid genera Brettus,
Cocalus,Cyrba,Gelotia,Myrmarachne and Phaeacius (Jackson et al., 2008;
Li et al., 2003; Su et al., 2007), araneophagy has been shown at least to be likely
in species from the spider families Archaeaidae, Gnaphosidae, Lamponidae,
Mimetidae, Oxyopidae, Palpimanidae, Pholcidae, Scytodidae and Theridiidae
(Cerveira and Jackson, 2005; Cutler, 1972; Forster and Blest, 1979; Gonzaga
et al., 1998; Jackson, 1986a, 1992b; Jackson and Brassington, 1987; Jackson
and Whitehouse, 1986; Jarman and Jackson, 1986; Legendre, 1961; Li et al.,
1999; Smith Trail, 1980; Whitehouse, 1986, 1987; Wood, 2008).
Distinctive preference for ants as prey (myrmecophagy) may also be espe-
cially common among spiders. Examples from the families Zodariidae (Allan
et al., 1996; Peka
´r, 2004, 2005; Peka
´ret al., 2005, 2008) and Salticidae (Allan
and Elgar, 2001; Cutler, 1980; Huseynov et al., 2005; 2008; Jackson and Li,
2001; Li and Jackson, 1996a) are especially well known, but there may also be
numerous examples from the families Aphantochilidae, Clubionidae, Thomisi-
dae and Theridiidae (Castanho and Oliveira, 1997; Fowler, 1984; Lubin, 1983;
Oliveira and Sazima, 1984, 1985; Porter and Eastmond, 1982). Other spiders
appear to have distinctive preferences for termites (Jocque
´and Dippenaar-
Schoeman, 1992; Wesolowska and Cumming, 1999, 2002; Wesolowska and
SPIDER COGNITION 123
Haddad, 2002), woodlice (Bristowe, 1941; R
ˇeza
´c
ˇand Peka
´r, 2007; R
ˇeza
´c
ˇet al.,
2008), moths (Haynes et al., 2002; Stowe, 1986; Stowe et al., 1987) or flies
(Yeargan, 1994; Yeargan and Quate, 1996, 1997).
However, as a predator that expresses extreme specificity, Evarcha culici-
vora would have few rivals. This East African salticid feeds indirectly on
vertebrate blood by actively choosing, as preferred prey, blood-carrying female
mosquitoes (Cross and Jackson, 2010a). Prey-choice experiments show, for
example, that E.culicivora can discriminate by sight between blood-carrying
female mosquitoes and male mosquitoes, between blood-carrying female mos-
quitoes and female mosquitoes that are not carrying blood and between mos-
quitoes and midges that are similar in size and appearance (Jackson et al., 2005).
To top it off, E.culicivora even has a preference for a particular mosquito
genus—Anopheles (Nelson and Jackson, 2006).
2 Paying attention
2.1 SELECTIVE ATTENTION
Preference might be a cognitive topic, but we will now review something that
lies deeper within the realms of cognition—spiders that make use of search
images. Use of the term ‘searching image’ in biology, nowadays usually
shortened to ‘search image’, can be traced back to von Uexku
¨ll (1934) (see
Bond, 2007). However, it was especially with Lukas (Luuk) Tinbergen (1960)
that the topic of search images began its rather peculiar history in biology. The
name Tinbergen is, of course, strongly associated with research on animal
behaviour, with Nikolaas (Niko) Tinbergen being widely regarded as one of
the two principal founders of ethology (Kruuk, 2003), Lorenz being the other.
Luuk and Niko Tinbergen were brothers, but Luuk was primarily an ecologist,
not an ethologist. He did some remarkable field-based research in the Nether-
lands on insectivorous birds beginning in 1946 but ending abruptly in 1955 with
his untimely death at the age of 39 (Baerends and de Ruiter, 1960). His work,
published posthumously 5 years later (Tinbergen, 1960), included his hypothe-
sis on search images. Tinbergen envisaged search images as perceptual changes,
the idea being that a predator, after discovering a particular type of prey, ‘‘gets
an eye for’’ or ‘‘learns to see’’ this particular type of prey.
Tinbergen (1960) also suggested that predators ‘‘perform a highly selective
sieving operation on the visual stimuli reaching their retina’’ (p. 332). ‘Sieving’,
or ‘filtering’, implies that certain features of the prey are more or less ignored,
whereas other, more salient features are attended to. It may also imply that the
predator ignores distractors in the environment, such as features of non-prey.
There is parallel evidence that sieving is important in the visual search para-
digms adopted by people, where a particular target with a certain configuration
of features is attended to within a crowd of distractors lacking in this
124 ROBERT R. JACKSON AND FIONA R. CROSS
configuration (Pashler, 1998; Treisman, 1986; Treisman and Gelade, 1980). In
biology, however, Tinbergen’s search-image hypothesis began as a topic rife
with controversy, and maybe it is easy to see why. ‘Search images’ are funda-
mentally about predators paying attention to prey, and attention is fundamen-
tally about cognition.
‘Attention’ joins ‘cognition’ and ‘mind’ as another hard-to-define term.
James’ (1890) solution was to say ‘‘everybody knows what attention is’’, but
modern psychologists are more likely to say the opposite, that ‘‘no one knows’’
(Pashler, 1998). As with ‘cognition’ and ‘mind’, it helps to envisage a contin-
uum instead of absolutes. For example, Cyclosa octotuberculata, an orb-web
spider, has been shown to be more responsive to prey that land on specific parts
of its web, with these parts of the web being connected to threads the spider’s
legs are holding under high tension (Nakata, 2010). We can say the spider is
selectively attentive to the sectors of the web under high tension, and perhaps
this is attention in its most basic manifestation. However, Tinbergen’s search-
image hypothesis pertains to an expression of selective attention that goes
beyond what we see with Cyclosa.
Reading Tinbergen’s paper now, more than 50 years later, we get an interest-
ing glimpse of a field biologist coming to grips with the cognitive implications
of animal behaviour while writing for what appears to be primarily an audience
of ecologists. ‘Attentional priming’ and other terms now in common use would
have been unfamiliar to Tinbergen. However, what Tinbergen called ‘‘learning
to see’’ can be rephrased now as previous experience by the predator with a
particular type of prey priming the predator to be selectively attentive to specific
features of this particular prey (see Blough, 1989, 1991, 1992; Brodbeck, 1997;
Dawkins, 1971a,b; Langley, 1996; Langley et al., 1996; Reid and Shettleworth,
1992; Shettleworth, 2009).
After Tinbergen, the term ‘search image’ came into widespread use in
ecology, but with the accepted meaning of the term being shifted away from
selective attention and with the distinction between hypotheses and findings
being blurred. Search-image use was Tinbergen’s hypothesis, not his findings.
Based on sampling in the field, Tinbergen showed that the diet of birds deviated
in particular ways from the relative abundance of the different types of potential
prey in the field. Tinbergen’s data set was extraordinary and his hypothesis was
innovative, but field sampling can never simply demonstrate that animals adopt
search images. Testing this hypothesis about cognition depends on designing
experimental studies for demonstrating specifically selective attention.
Tinbergen’s search-image hypothesis has been the impetus for many studies
over the past five decades, and some of these studies (e.g. Bond and Kamil,
1998, 2002) have been clearly about selective attention. At the same time, it is
as though, in ecology, a tradition has been established of misconstruing search
images as being about preferences instead of selective attention. This basic
confusion is then compounded when the meaning of ‘preference’ is also mis-
construed (e.g. Ishii and Shimada, 2010).
SPIDER COGNITION 125
With search images, it seems particularly easy to miss the point. The point is
that, as a determinant of diet, an animal deploying a search image is different
from an animal expressing a preference. ‘Preference’ refers to something an
animal would like to eat, and what drives an animal’s choice behaviour. ‘Search
image’, however, refers to something an animal has become cognitively
prepared to detect and identify. Neither search images nor preferences can be
determined from data on diet alone. Search images and preferences are two
different cognitive processes, and discriminating between these two cognitive
processes depends on having data from appropriately designed experiments.
2.2 PERCEPTUAL LEARNING
In conventional search-image studies, the test subject (usually a bird or a
mammal) is exposed repeatedly to a particular prey type, usually accompanied
by food as reinforcement. The rationale for repeated exposure is to train the
predator to identify features of the prey (Gendron, 1986; Gendron and Staddon,
1983; Royama, 1970). This training is based on a hypothesis, often unstated,
that perceptual learning (e.g. Goldstone, 1998; Yotsumoto and Watanabe, 2008)
for these particular prey features takes place simultaneously with the predator
becoming selectively attentive to this type of prey.
‘Learning’ in general can be defined broadly as the modification of behaviour
by experience (Lorenz, 1965; Stephens, 1991), and by this definition, learning is
exceedingly widespread in the animal kingdom. On the whole, examples of more
complex learning are known for insects than for spiders (Dukas, 2008). However,
there is already abundant experimental evidence of spiders learning in the context
of foraging, web building, intraspecific interactions, navigation and avoidance of
aversive stimuli (Bays, 1962; Chmiel et al., 2000; Edwards and Jackson, 1994;
Grunbaum, 1927; Heiling and Herberstein, 1999; Hoefler and Jakob, 2006; Jakob
et al., 2007, 2011; Lahue, 1973; LeGuelte, 1969; Morse, 1999, 2000a,b; Nakata
and Ushimaru, 1999; Punzo, 1998, 2004; Sandoval, 1994; Sebrier and Krafft,
1993; Seyfarth et al., 1982; Skow and Jakob, 2006; Tso, 1999; VanderSal and
Hebets, 2007; Venner et al., 2000; Whitehouse, 1997).
It may be implicit most of the time that ‘learning’ pertains specifically to
modification of behaviour that is adaptive (Beecher, 1988; Beer, 1996;
Johnston, 1982; Kamil, 1988; Staddon, 1983). We can say that an animal
‘learns’ when it solves a problem and then remembers the solution. Sometimes
there may be an interesting interface between learning and cognition (Shanks,
2010), but we should avoid the habit that is prevalent in casual conversation of
using the terms ‘learning’ and ‘cognition’ interchangeably. Showing adaptive
experience-derived modification of behaviour does not necessarily imply any-
thing like the animal having gained an understanding of the solution to a
problem.
The cognitive character of search-image use is not strictly tied to learning.
When search images are derived by perceptual learning, we can say that the
126 ROBERT R. JACKSON AND FIONA R. CROSS
problem the predator solves is how to identify the prey type for which selective
attention is being developed. While using the search image, the predator
remembers the solution to this problem. However, as we shall see, the search
images used by spiders seem to be pre-existing solutions to the identification
problem.
2.3 INNATE SEARCH IMAGES
Evidence of search-image use by spiders has come from research on two salticid
species: Portia labiata (Jackson and Li, 2004) and E. culicivora (Cross and
Jackson, 2010b). Besides illustrating that search images (i.e. primed selective
attention) and perceptual learning are distinctly different processes, this
research can be used to illustrate why we say that search images are about
selective attention, not preferences. In the search-image studies using P.labiata
and E.culicivora (Cross and Jackson, 2010b; Jackson and Li, 2004), the critical
distinction was whether a stimulus was cryptic or conspicuous. Importantly, the
stimuli used in the search-image experiments were already (innately) salient to
the individual being tested. For P.labiata (Jackson and Li, 2004), the stimulus
was one of two prey spiders commonly found in its habitat in the Philippines.
For E.culicivora, one of the stimuli was its preferred prey, blood-carrying
female mosquitoes. The other stimulus for E.culicivora was a potential mate
(i.e. an opposite-sex conspecific). Considering search images in the context of
finding mates is unconventional, but this example from E.culicivora might be
especially useful for highlighting how selective attention is what defines search
images, regardless of the context in which selective attention is deployed.
The rationale for making a stimulus appear cryptic in search-image experi-
ments is that we want to make the task of finding the stimulus difficult (i.e. we
could say we want to ensure that finding the stimulus will demand a lot of
attention). When the stimulus is conspicuous, it is easy to find and, therefore, we
do not expect priming to have a particularly pronounced effect on the subject’s
success at finding the stimulus. Priming effects should be most evident when the
stimulus is cryptic. With preference, we expect the opposite—that the expres-
sion of preference should be most pronounced when the preferred stimulus is
easy to find (i.e. conspicuous).
In the first search-image study (Jackson and Li, 2004), each test spider
(P.labiata) was given an opportunity to capture and eat one of two spider
species or a house fly (Musca domestica). The two prey spiders are, at least for
people, very different in appearance, with one (Scytodes pallida, Scytodidae)
having a characteristic heavy-set appearance and with the other (Micromerys
sp., Pholcidae) having a slender, pencil-like body and characteristically long
legs. Depending on the experiment, the spiders and flies were either alive or
they were lures (dead prey mounted in life-like posture on cork discs). In the
experiments, there was no evidence of prior experience with a house fly (non-
preferred prey) calling up a search image. Nor was there any evidence of
SPIDER COGNITION 127
priming effects when the prey spider was conspicuous. However, as predicted
by the search-image hypothesis, effects of priming were evident only when the
prey spider was cryptic. When the prey was cryptic, but not when it was
conspicuous, more individuals of P.labiata found S.pallida after initially
being allowed to eat S.pallida, and more found Micromerys sp. after initially
being allowed to eat Micromerys sp. Evidently, in these experiments, experien-
cing a particular type of prey primed selective attention, not preference. This
finding is strikingly different from the more familiar search-image studies
because selective attention was primed after a single exposure to prey of a
type the tested salticid had never encountered before. Perhaps one-trial percep-
tual learning is a possibility, but other evidence we consider will show it is not a
necessity.
For example, the findings from the search-image study using E.culicivora
(Cross and Jackson, 2010b) will be even harder to explain in the context of
perceptual learning. P. labiata ate the prey when being primed, but E.culicivora
had no physical contact with the prey or with the potential mates during
priming. Instead, E.culicivora began the experiment in a separate chamber
from which it could see lures made from blood-carrying mosquitoes or lures
made from potential mates. In control trials, E.culicivora began in a chamber
from which it could see neither. E. culicivora was then put in an arena where the
task was to find a lure made from a blood-carrying mosquito or a lure made from
a potential mate. Again, there was no evidence of priming effects when prey or
mates were conspicuous. However, after being primed with a view of blood-
carrying mosquitoes, significantly more E.culicivora individuals found a
blood-carrying mosquito that was cryptic. After being primed with a view of
potential mates, significantly more E.culicivora individuals found a potential
mate that was cryptic. None of the test spiders had any experience prior to the
experiment with potential mates or with mosquitoes. The only possible rein-
forcement in these experiments was from simply seeing the priming stimulus
because test spiders could not mate or eat during priming. Instead of being
trained to identify the visual features of prey or mates, E.culicivora apparently
called up an innate search image after one encounter of seeing prey or mates.
2.4 CAPACITY LIMITATIONS AND TRADE-OFFS
Tinbergen (1960) had suggested that birds might make use of more than one
search image at a time, but there has been little support for this hypothesis from
subsequent research. It is now widely appreciated that, even for birds and
mammals, being selectively attentive to one thing interferes with being selec-
tively attentive to other things (Bond, 1983; Dukas, 2002, 2004; Dukas and
Kamil, 2000, 2001; Kamil and Bond, 2006; Pietrewicz and Kamil, 1979). It is
easy to envisage a spider, a much smaller animal, being even more limited in
cognitive capacity when challenged by selective attention tasks, and perhaps it
is not surprising that there is evidence of capacity limits applying to spiders.
128 ROBERT R. JACKSON AND FIONA R. CROSS
After priming, E.culicivora and P.labiata became more effective at finding a
congruent stimulus and they also became less effective at finding an incongruent
stimulus (Cross and Jackson, 2010b; Jackson and Li, 2004). For example, signifi-
cantly fewer E.culicivora individuals found a cryptic mate after being primed
with mosquitoes than after no priming (i.e. being primed with an incongruent
stimulus impaired the spider relative to not being primed by anything at all).
We may have good evidence that capacity for selective attention is subject to
severe limitations, and these limitations may be especially pronounced for small
animals like spiders, but little is known about the cause of these limitations.
Dukas and Kamil (2000) referred to ‘computational resources’ being tied up
when an animal is being selectively attentive, with a consequence of this being a
severe reduction of the resources available for other tasks related to attention.
This may be a useful step towards rephrasing the question about why selective
attention is a demanding task. However, before we can go very far towards
finding an answer, we will need a clearer understanding of what a computational
resource might actually be.
Failing to detect an unexpected prey may seem like a minor price for a spider
to pay in exchange for being more effective at finding the expected prey, but
there may be mortal costs when selective attention impairs a spider’s capacity
for detecting its own predators. As a general hypothesis, Dukas and Kamil
(2000) suggested that predators may take advantage of times when the prey’s
attention is focussed on its own food and P. fimbriata seems to illustrate one
way of doing this (Jackson et al., 2002a). One of the prey species exploited by
P.fimbriata is Zosis genicularis, an orb-web spider that has no venom with
which to immobilize the insects that land on its web. Its solution is to wrap the
insect in enormous amounts of silk (Lubin, 1986). In experiments (Jackson
et al., 2002a), resident spiders that were busy wrapping their own prey became
inattentive to one of their predators, P.fimbriata, advancing through the web.
While stalking across the web, P.fimbriata’s footsteps corresponded to when
the resident spider was preoccupied with wrapping its own prey. In other words,
P.fimbriata took advantage of the resident spider’s inattentiveness.
3 Perception
3.1 CLASSIFICATION AND OBJECTS
The findings from research on visual attention, like the findings from the
research on prey-choice behaviour, imply that spiders see things—objects or,
more specifically, ‘visual objects’. A visual object is an extraordinary thing. In
less than a second, people can recognize different objects, and we do this even
when part of the object has been occluded (Biederman, 1987, 1995).
The word ‘recognize’ here is interesting. Often it seems that people agonize
over how to define cognition, all the while comfortably using its sister word
SPIDER COGNITION 129
‘recognition’. Object recognition, visual objects—these are cognitive topics.
Classification is also cognitive, and people are ever busy classifying objects.
People recognize categories, including categories of animals, but the way we do
this can be surprising. For example, most people more or less automatically
perceive robins and sparrows as being members of the category ‘bird’, whereas
penguins and ostriches are atypical for many people and not automatically
classified as ‘birds’ (Jolicoeur et al., 1984). The meaning of ‘bird’ (or avian
reptile) in formal scientific taxonomy is different, being something derived
intellectually and not necessarily corresponding to how we simply perceive
the category ‘bird’. This should be a warning when we consider whether a
predator is stenophagic or euryphagic. When our interest is in the prey cate-
gories recognized by the predator, formal scientific taxonomy is an exceedingly
dubious basis for conclusions related to stenophagy or euryphagy.
For example, there may be many salticids that feed on insects and spiders
indiscriminately (Huseynov, 2005, 2006; Jackson, 1977). Portia’s natural diet is
dominated by spiders (Jackson and Blest, 1982), but does this mean that Portia
is more stenophagic and these other salticids are more euryphagic? A question
like this comes down to whose perspective we mean when we say ‘stenophagy’.
It is behavioural evidence from experiments, not taxonomic data on prey in this
predator’s natural diet that reveals how richly populated Portia’s world is with
prey categories (Harland and Jackson, 2004). Formal scientific taxonomy
means nothing to Portia.
Behavioural data, in fact, tell us that Portia experiences life as a euryphagic
predator. In the field, Portia preys on many kinds of spiders and findings from
laboratory experiments reveal that Portia has a large repertoire of prey-specific
tactics, with different tactics corresponding to different types of spiders
(Jackson and Wilcox, 1998). For Portia, classifying prey goes considerably
beyond stopping with just ‘spider’ for a category, revealing that ‘spider’ is not,
for this predator, any one type of prey or even just a few. It is many.
Phidippus is another salticid that may do considerable classifying of prey (in
this case, classifying of different insects; Edwards and Jackson, 1994), but no
other spider has ever been shown to take prey classification as far as Portia
does. If we are interested in the classification schemes used by predators rather
than the classification schemes used by professional human taxonomists, Portia
stands out as one of the most extraordinarily euryphagic species ever studied.
A couple of examples might clarify what we mean when we say Portia adopts
prey-specific tactics. For the first, the prey we will consider is an unnamed
species of Scytodes from the Philippines. All species in the genus Scytodes are
spitting spiders, and the Philippines Scytodes has the distinction of being a
spitting spider that preys especially often on salticids (Li et al., 1999). P. labiata
in the Philippines is a salticid that frequently preys on this salticid-eating
spitting spider (Jackson et al., 1998). Spitting makes Scytodes particularly
dangerous as prey for P.labiata. However, Scytodes’ dangerous end is at the
front, as the gummy spit is fired from slits on Scytodes’ fangs, and one part of
130 ROBERT R. JACKSON AND FIONA R. CROSS
P.labiata’s prey-capture tactic is to take detours by which it can approach this
dangerous spider from its rear, keeping out of Scytodes’ line of fire (Fig. 3).
Scytodes lives in a web covering the top of a leaf. Once on the spitting spider’s
web, P.labiata concentrates on moving forward very slowly without provoking
spitting attacks.
However, prey-specific tactics reveal that, for P.labiata, all individuals of
Scytodes are not the same. For example, P.labiata foregoes the detour and
instead takes the shorter, faster head-on approach when the spider it sees in a
web is a Scytodes female that is carrying eggs (Li and Jackson, 2003). This
makes sense because Scytodes females carry their eggs around in their mouths.
Egg-carrying females can still spit, but only by first releasing their eggs
(Li et al., 1999). Being reluctant to release their eggs, egg-carrying females
are, for P.labiata, less dangerous as prey. The distinction between eggless and
egg-carrying Scytodes females, though irrelevant for formal scientific taxon-
omy, is highly relevant in P.labiata’s own classification system, and P.labiata
shows a pronounced preference for the safer, egg-carrying scytodids (Li and
Jackson, 2003).
For our other example, we can go to a rainforest in Queensland, Australia,
where P.fimbriata preys especially often on other (i.e. non-Portia) salticids
(Jackson and Blest, 1982). Something important to realize is that all species in
the genus Portia are very different in appearance from most other salticids.
Many people and, apparently, many salticids often mistakenly identify Portia as
simply being a piece of detritus. However, we know that P.fimbriata classifies
salticids as a type of prey distinctly different from other spiders because when,
and only when, stalking a salticid, P.fimbriata adopts a tactic known as ‘cryptic
stalking’ (Harland and Jackson, 2001; Jackson and Blest, 1982). This tactic is
especially effective at minimizing the other salticid’s chances of detecting and
FIG. 3 Portia labiata (left) from the Philippines stalking a spitting spider, Scytodes
pallida (right). Approaching head on would bring Portia into Scytodes’ line of fire.
Instead, Portia executes a planned detour by which it approaches this dangerous prey
from the rear.
SPIDER COGNITION 131
identifying P.fimbriata. A special posture is adopted (legs pulled back and
pedipalps retracted, hiding their outlines), and a special style of walking is
adopted (advancing very slowly while stepping with a choppy, stop-and-go
gait). While stalking, P.fimbriata manoeuvres about so that its approach is
from the salticid’s rear. Despite these precautions, the stalked salticid some-
times detects motion, pivots around and faces P.fimbriata. When this happens,
P.fimbriata’s response is to freeze until the salticid turns away. Eventually
P.fimbriata gets close, looms over the salticid from behind and attacks with a
rapid downward thrust. Preference is also tuned to salticids. Besides having a
specific tactic with which to target salticids, P.fimbriata also expresses prefer-
ence for salticids when the alternatives are spiders from other families (Li and
Jackson, 1996b).
If we envisage predators as being spread over a continuum from minimal to
extreme classifying, contrasting a toad with Portia might be revealing. Using
formal taxonomy and counting the number of prey species eaten, we might be
tempted to make a list of prey species and then use this list as justification for
saying that the toad’s diet includes many kinds of prey. However, the toad does
not appear to do a lot in the way of recognizing the individual species in this vast
list as being different kinds of prey. In fact, the toad seems to do minimal
classifying (Ewert, 2004; Lettvin et al., 1959; Wachowitz and Ewert, 1996).
Toads can quickly determine by sight alone whether something that moves is
food (i.e. whether it is a ‘bug’) or not: see a bug, out goes the tongue; not see a
bug, ignore it. Many salticids may resemble toads by eating prey from a
taxonomic range considerably wider than the taxonomic range of prey eaten
by Portia. However, from the predator’s perspective (i.e. a cognitive perspec-
tive), the toad and the conventional salticid appear to be distinctively stenopha-
gic if not monophagic and Portia appears to be, by a wide margin, the more
euryphagic predator.
3.2 SPIDERS THAT SEE THINGS
We should pause long enough to appreciate the cognitive implications of what
we now know about Portia’s prey-classification practices. Somehow, with its
minute eyes and brain, this salticid must be rendering visual objects and assign-
ing them to categories in an intricate classification scheme. We can only specu-
late about how the spider’s brain works, but a lot more is known about salticid
eyes (Cross and Jackson, 2006; Harland and Jackson, 2000a, 2004, in press).
Salticids have two big forward-facing (antero-medial) eyes, called the ‘principal
eyes’ (Homann, 1928; Land, 1969a,b). Salticids have another six eyes (antero-
lateral, postero-medial and postero-lateral), called the ‘secondary eyes’, which are
positioned around the side of the carapace and, at the risk of oversimplifying (see
Harland and Jackson, in press), we can just say these eyes function as motion
detectors (Land, 1971, 1972; Zurek et al.,2010). In this brief tour of eye function-
ing, it is the principal eyes that we most need to consider (Fig. 4).
132 ROBERT R. JACKSON AND FIONA R. CROSS
The most striking characteristic of the principal eyes is their exceptional
spatial acuity (Su et al., 2007). The best spatial acuity known for a salticid is
of 0.04(Blest and Price, 1984; Blest et al., 1990; Williams and McIntyre,
1980) which is considerably better than the best spatial acuity (0.4) known for
an insect (Labhart and Nilsson, 1995). However, for putting this into perspec-
tive, we should point out that the insect that achieves spatial acuity of 0.4is a
dragonfly with eyes bigger than a salticid’s entire body. Another perspective
comes from appreciating that our own spatial acuity is 0.007(Kirschfeld,
1976), only five times better than a salticid’s.
Insects have compound eyes, but spiders, like vertebrates, have camera eyes
(Land and Nilsson, 2002). The basic difference is that each facet of a compound
eye acts as something like a pixel on a computer monitor, with the input from
pixels generating a picture. Technically, when we say this, we are considering
compound eyes that work as apposition eyes and what we have said is a
simplification, but this suffices for distinguishing between compound and camera
eyes (Land, 2005; Land and Fernald, 1992). A camera eye has a single lens
system that projects an image on to a retina. We can then think of seeing with a
camera eye as being something like extracting information from an image.
However, there is a massive difference in scale when, for example, we compare
the good spider eyes of salticids with the good vertebrate eyes of primates.
Secondary eyes
Principal eyes
FIG. 4 Adult male of Evarcha culicivora eating a blood-carrying Anopheles gambiae,
the mosquito species that is the primary vector of malaria in equatorial Africa. The
principal and secondary eyes of the male are indicated.
SPIDER COGNITION 133
In a salticid’s retina, you may find thousands ofphotoreceptors (Land, 1969a), but
there may be 100–200 million photoreceptors in a primate’s retina (Palmer, 1999).
The salticid eye, like the primate’s, has a fovea, a region where photoreceptor
packing is optimal for spatial acuity. Photoreceptor spacing in the salticid
fovea, like in the primate fovea, is optimal for sampling in visible light, but there
are only about 200 photoreceptors in the salticid’s fovea (Blest et al.,1990;
Gregory, 1998).
Salticids (Land, 1969a), like primates, have active eyes, but primates move
their eyeballs—the whole thing (lens system and retina). The salticid’s corneal
lens, however, is rigidly fixed in place on the front of the carapace. Behind the
cornea, there is a long eye tube with a second lens at the inner end and the retina
behind this lens (Williams and McIntyre, 1980). The salticid moves the eye tube
while the cornea remains stationary. ‘Scanning’ is the most distinctive kind of
eye movement by salticids (Jackson and Harland, 2009; Land, 1969b), this
being rather different from the way vertebrate eyes move.
While scanning, the salticid’s eye tube is moved from side to side while
simultaneously rotating clockwise or counter-clockwise. Scanning may have an
integral role in rendering visual objects. The cornea delivers into space an image
much bigger than the retina’s field of view. Land (1969a) proposed that
scanning may be a method by which the salticid actively searches for lines or
other salient features on the image generated by the lens system. If this hypoth-
esis is correct, then these active eyes, by scanning, may be generating a picture
piece by piece (Harland and Jackson, in press).
What is more, salticids see in colour (Harland and Jackson, in press), but we
often have to remind ourselves that seeing colour, like seeing objects, is
cognitive (Gregory, 1998). Good spatial acuity is critical for object recognition.
However, for colour vision, what we look for are different photoreceptors in a
retina having different spectral sensitivities. That salticids see colour is implied
by the retinas of salticid principal eyes having photoreceptors with different
specific spectral sensitivities (Blest et al., 1981; DeVoe, 1975; DeVoe and
Zvargulis, 1967; Peaslee and Wilson, 1989; Yamashita, 1985; Yamashita and
Tateda, 1976). There is also evidence from behavioural experiments supporting
the conclusion that salticids see colour (Kaestner, 1950; Li and Lim, 2005; Li
et al., 2008; Lim and Li, 2006; Nakamura and Yamashita, 2000). However, this
does not imply that colour seen by salticids is the same as colour seen by people.
For example, salticids have photoreceptors that are maximally sensitive to
ultraviolet light, which we do not see (Blest et al., 1981).
In general, what applies to colour vision applies to vision in general. We need
to guard against unexamined assumptions. Because we can see exceptionally
well, we may be tempted to think we intuitively understand what salticids see.
Yet, when we examine salticid eyesight more closely, our confidence melts
away. At first, it may seem disturbingly accurate to say that the objects we see
are also seen by a salticid. For example, once we are educated, we can distin-
guish mosquitoes from other small insects. We can also look at a mosquito and
134 ROBERT R. JACKSON AND FIONA R. CROSS
see that it is a female and that it is carrying blood. Moreover, we can see that one
mosquito belongs to the genus Anopheles and another mosquito belongs to the
genus Culex. How extraordinary that a mere spider, E.culicivora, can do all of
this, and do this well, without any training (Jackson et al., 2005).
Yet, it is not the same. Speed is one issue. Part of what we normally mean by
seeing well is the capacity for quickly identifying objects by sight. For example,
once trained, a person can make all these mosquito identifications so fast we
would say ‘instantly’, but it is normal for E.culicivora to spend many minutes
facing its prey before making a decision (Nelson et al., 2005). On the whole, we
should be cautious. Assuming especially close similarities between what seeing
means for us and what it means for the salticid may be hard to reconcile with the
hypothesis that, for salticids, the process of generating visual objects and
practising visual classification is based in a large part on scanning a large
image piece by piece with a tiny retina (Harland and Jackson, in press).
Intuition may be all the more misleading when we consider non-salticid spiders.
The spatial acuity of salticid principal eyes is extreme, but spatial acuity, like so
many of the things that interest us, lies on a continuum. Most spiders have eyes
situated in a region of this continuum far from where we expect eyes to function in
the identifying and classifying of visual objects (i.e. far from the region they need
to be in for the things we especially often associate with ‘seeing’). Exceptional
spatial acuity may be rare outside the Salticidae (Homann, 1971; Land, 1985), but
this is not the same as saying that, out of all the spiders, only salticids live in a
world of visual objects. Some lycosids, ctenids and other non-salticid spiders seem
to have eyes that function in ways that make the cognitive topics of visual objects
and visual classification meaningful (Clemente et al., 2010; Grusch et al.,1997;
Kaps and Schmid, 1996; Land, 1985; Land and Barth, 1992; Land and Nilsson,
2002; Schmid and Trischler, 2011). However, when considering non-salticid
spiders, we need to be even more cautious. We should be wary of assuming that
what they see is especially similar to what we see. For understanding the objects a
spider sees, we need clever, appropriate experiments. Intuition will not suffice and
sometimes it can be misleading.
Dinopis illustrates how seeing especially well can pertain to more than visual
objects and more than colour. This spider from the family Dinopidae shares with
the salticids a reputation for being a high achiever in the realm of seeing, but the
way Dinopis excels is different from the way salticids excel, this being a type of
seeing that makes sense for Dinopis’ own way of life. Neither is this a spider
that wanders around hunting for prey like a salticid, nor is it a spider that spins a
conventional web. Dinopis’ game is using a silken net when capturing prey
(Coddington, 1986). With net ready, Dinopis perches in the vegetation at night
close to the ground, facing downward. When an unwary ant walks by on the
ground below, Dinopis throws the net over it and then reels the ant in for a meal
(Austin and Blest, 1979; Getty and Coyle, 1996; Robinson and Robinson, 1971).
Like a salticid, Dinopis has a pair of big forward-facing eyes, but these are
Dionopis’ postero-medial eyes, not its antero-medial eyes (Fig. 5). With these
SPIDER COGNITION 135
eyes, Dinopis is extraordinarily good at detecting moving objects under dim
light (Blest and Land, 1977; Laughlin et al., 1980). However, we should think
about whether saying ‘object’ is appropriate. It is unlikely that this spider
perceives anything particularly similar to the kind of visual objects rendered
by a salticid’s visual system, or our own. For example, there is no evidence to
suggest that Dinopis has eyes that can discern detail of static appearance (Land,
1985). Perhaps it is more realistic to envisage the behaviour of this spider as
being based on cognitive constructs derived from discerning details related to
movement. For us, a movement-related construct may be intuitively inaccessi-
ble, but intuitive inaccessibility can be advantageous. Intuition and understand-
ing are different things, but intuition can give us an illusion of already
understanding something we should be investigating experimentally.
3.3 THE UNSEEN WORLD OF THE SPIDER
As intuition and imagery often go together, trying to think about spiders using
sensory modalities other than vision can feel like stepping into an alien world.
Yet other sensory modalities, ones that have nothing to do with seeing, are
important to all spiders. These sensory modalities are often exquisitely sensi-
tive, fine-tuned ‘masterpieces of engineering’ (Barth, 2001, 2002) that work
close to the limits of what is physically possible. For example, spiders have
sensors that register air flow at a level close to what appears to be the maximum
FIG. 5 Like a salticid, Dinopis (Dinopidae) has large eyes and good eyesight. However,
salticids use their eyes for seeing detail of static appearance, whereas Dinopis is
extremely proficient at detecting moving objects under dim light.
136 ROBERT R. JACKSON AND FIONA R. CROSS
possible sensitivity, as more sensitive sensors would be incapable of distin-
guishing between a meaningful signal and the meaningless noise from Brow-
nian motion of molecules (Barth, 1985, 2000, 2004).
Flow sensors are just the beginning. The spider’s body and appendages are
armed with a battery of tactile-like sensors that detect, among other things,
airborne sound and substrate vibration. We can say these sense organs are all
busy transmitting information to the spider, but the term ‘information’ can be
misleading. What matters to the spider is not information in the context of
abstract knowledge (see Barth, 2002). Although it may be beyond our intuitive
grasp, a spider using sense organs unrelated to seeing must be working with
cognitive constructs that have a role with functional similarities to the role of
visual objects.
Thinking about sensory systems becomes especially interesting when we
contemplate the web-building spider’s web. ‘Web’ seems to be another word
that stubbornly defies precise definition (Jackson, 1986a). However, when we
hear ‘spider web’, most of us probably first think of a silken device fixed in the
environment and considerably larger than the spider that built it. Most of us
probably also think of webs functioning as prey-capture devices, with the idea
being that the web intersects mobile insects and perhaps ensnares the prey long
enough for the spider to arrive.
That spider webs have sensory functions is widely acknowledged in the
literature, but often phrased as the web being an extension of the spider’s
tactile-like sensory systems (Landolfa and Barth, 1996; Masters, 1984;
Masters et al., 1986; Witt, 1975). However, saying ‘extension’ might be mis-
leading, as it is unlikely that the spider experiences the web as an extension of
anything. We may be predisposed to think of sense organs as being part of a
spider’s body, and therefore part of its phenotype. However, the web is also a
part of the spider’s phenotype or, to use Richard Dawkins’ (1982) expression, its
‘extended phenotype’. The web, used in conjunction with setae and slits on the
spider’s body and appendages, becomes a sense organ. It is an especially
interesting sense organ owing to how it is extended out into the environment.
Other spiders can walk directly into this sense organ and play mind games with
the web-building spider.
4 Communication
4.1 MIND GAMES IN SPIDER WEBS
Predators can exploit the web-building spider’s reliance on web signals by using
a special form of mimicry, called aggressive mimicry (Wickler, 1968), where
the predator deceives its prey by imitating something desirable (in this case, the
prey’s own prey). Sensory traps, sensory exploitation, sensory drive, receiver
psychology, exploitation of perceptual biases (Christy, 1995; Endler and
SPIDER COGNITION 137
Basolo, 1998; Guilford and Dawkins, 1991; Proctor, 1992; Schaefer and
Ruxton, 2009), along with aggressive mimicry, belong to the terminological
menagerie that has accumulated in the literature. All of these terms can be
applied to situations in which one organism interfaces the sensory system of
another organism. The distinctions that are made with these different terms are
often rather subtle—often too subtle. Maybe we have too many terms. Here, we
will step around these terminological issues and instead focus on how a pre-
dator’s strategy can sometimes be likened to playing mind games with its prey
(Jackson and Pollard, 1996, 1997; see Krebs and Dawkins 1984).
Many of the spiders that are known to be, or likely to be, araneophagic are
also known to be, or likely to be, web-invading aggressive mimics (Jackson,
1992b; Nelson and Jackson, 2011). Paralleling what we know about these
spiders, there are also insects that invade webs, prey on the resident spider
and practise signal-based aggressive mimicry, the most thoroughly studied of
these being Stenolemus bituberus, an emesine assassin bug (Hemiptera: Redu-
viidae) from Australia (Wignall and Taylor, 2008, 2009, 2010, 2011).
However, it is the intricacies of Portia’s strategy that have the most striking
cognitive implications. Finding a salticid in a web at all is unusual (Jackson,
1985a), but Portia’s most unusual characteristic is extreme predatory versatil-
ity. Besides making use of its own prey-capture web and besides invading the
webs of other spiders, Portia also hunts insect and spider prey away from webs.
These are like three different themes in Portia’s repertoire, but they have a way
of becoming thoroughly intertwined (Harland and Jackson, 2004). For example,
when salticids are walking on the ground below a web, P.fimbriata drops
slowly from the web on a silk line, suddenly grabs hold of the salticid when
close and then returns to the web to feed (Clark and Jackson, 2000). Moreover,
Portia’s own web is often spun connected to or embedded in the webs of other
spiders. Besides making predatory forays into its neighbours’ webs, Portia also
encourages its neighbours to attempt pilfering raids into its own web by leaving
insects there as bait (see Ruxton and Hansell, 2011). It then feasts on the would-
be thief that comes after the insect (Jackson and Blest, 1982).
Other layers of versatility become evident in abundance especially when we
examine what Portia does after entering other spiders’ webs. Portia does more
than feed on the resident spider. It may steal living insects it finds ensnared in
the web and the wrapped-up insects the resident spider has stowed away in the
web for a later meal. It may even feed on an insect alongside the resident spider,
only to later turn on its feeding partner and make a meal of it as well. The
resident spider’s eggs and juveniles are also food for Portia. Many spiders guard
their egg sacs by gripping them with their chelicerae, but to no avail when
Portia raids the web, as Portia will pull the eggs away from the mother spider
and eat them before eating the mother. Sometimes predator and prey are both in
an alien web, as Portia even eats other web-invading spider species that it finds
(Jackson and Blest, 1982; Jackson and Hallas, 1986b; Wilcox and Jackson,
1998, 2002).
138 ROBERT R. JACKSON AND FIONA R. CROSS
We could go on listing tactics from Portia’s repertoire, but we should pause
here and emphasize something especially important. With Portia, what we see
are not different individuals adopting different tactics. These are instead exam-
ples of individual versatility (Harland and Jackson, 2004), with each individual
predator deploying a large repertoire of distinctly different prey-specific tactics.
Something else is equally important. When Portia invades a web, it enters
into intimate and often dangerous contact with its prey’s sensory world—
dangerous because the tables may be turned, and Portia’s intended dinner
may become the diner (e.g. Jackson et al., 2002b). After entering the other
spider’s web, however, Portia’s behaviour is not simply to stalk or chase down
the resident spider. What happens next is more intricate and interactive. Portia
gains control over the resident spider’s behaviour. It does this by generating
signals to which the resident spider responds. The resident spider’s responses
are advantageous to Portia and are to the undoing of the resident spider
(Tarsitano et al., 2000).
There are similarities between Portia’s strategy and the strategies of some
other predators that are called ‘aggressive mimics’. Angler fish (Pietsch and
Grobecker, 1978), snakes (Catania, 2010; Hagman et al., 2008; Nelson et al.,
2010; Reiserer and Schuett, 2008), fireflies (Lloyd, 1965, 1975) and bolas
spiders (Eberhard, 1977; Haynes et al., 2002; Stowe, 1988) are among the better
known examples of predators that make aggressive-mimicry signals. However,
characterizing the signals of these other aggressive mimics may seem compara-
tively easy. It is with Portia’s aggressive-mimicry strategy that we see the
most distinctive departure from anything like simply making a few signals or
concentrating on just a few types of prey.
The way Portia makes web signals is by manipulating silken threads. It does
this by using one or any combination of its eight legs and two palps. Each
appendage can be moved in a variety of ways and each can be moved indepen-
dently of each other. Portia also makes signals by flicking its abdomen up and
down, with abdominal movements often being combined with appendage move-
ments. The net effect is that Portia seems to have at its disposal virtually an
unlimited array of different signals to use on the webs of other spiders (Jackson
and Blest, 1982; Jackson and Hallas, 1986a,b; Jackson and Wilcox, 1993b).
For understanding why proficiency at varying signals matters in Portia’s
strategy, we need to appreciate Portia’s prey range. Instead of targeting only
one or only a few web-building spiders, Portia appears to be ready to take on
virtually any spider found in a web as long as it is similar to its own size. We
also need to remember that Portia’s intended prey, the resident spider, has its
own refined ability to acquire and process sensory information (e.g. Suter,
1978), with the web being an inherent part of the primary sensory system it
uses. Whether a particular web signal is meaningful to a resident spider varies
among spider species, between the sexes, across age class, and in accordance
with previous experience, feeding state and more factors than we can easily
name (Jackson, 1986a; Masters et al., 1986; Witt et al., 1968).
SPIDER COGNITION 139
For Portia, having a large repertoire of web signals is not enough. The
problem is how to select the particular signals that will work for a particular
resident spider and Portia’s solution can be remarkably flexible. Portia is
innately predisposed to begin with particular signalling routines during encoun-
ters with some of its more common natural prey, but Portia uses trial and error
when the prey is any other spider (Harland and Jackson, 2004).
Here is the basic idea of how trial and error works. After going on to the web
of a spider for which it does not have a pre-programmed tactic, Portia begins by
generating a kaleidoscope of different web signals. When one of these signals
eventually elicits an appropriate response from the resident spider, Portia stops
varying signals and instead concentrates on making the signal that worked
(Jackson and Nelson, 2011; Jackson and Wilcox, 1993b).
At first sight, trial and error may appear clever and remarkably flexible, but it
may also appear not to be especially cognitive. A simple generate-and-test
algorithm (Simon, 1969) seems to underpin this signal-derivation procedure.
However, we have been grossly oversimplifying Portia’s strategy. It is now
time to consider the details that give Portia’s strategy its distinctively cognitive
character.
Trial-and-error sessions are considerably more dynamic than this brief sum-
mary might suggest. For example, although the resident spider may respond
appropriately and Portia may narrow in on signals that work, there is no
guarantee that the resident spider will continue responding appropriately long
enough for Portia to make a capture. However, when the resident spider
switches to behaving inappropriately, Portia reverts to trial and error until
again finding an effective signal (Jackson and Nelson, 2011; Jackson and
Wilcox, 1993b).
Another complication is that, for Portia, encountering common prey and
being innately predisposed to begin with a particular signal does not render
trial and error irrelevant. The function of starting with particular signals may be
primarily to get the sequence off to a good start, and it is rare that just these
initial signals suffice. Usually, Portia’s predatory sequence is finished using
trial and error (Harland and Jackson, 2004).
Portia’s use of aggressive mimicry also works beyond the realm of imitating
prey. Here is an example. In Queensland, Australia, Euryattus is one of the
salticid species preyed on by P.fimbriata. However, Euryattus females build an
unusual nest and this makes Euryattus a unique prey for P.fimbriata.
AEuryattus female uses silk guylines to suspend a rolled-up dead leaf from
the vegetation or from a boulder, rock wall or tree trunk. Euryattus males do not
build leaf nests, but they take an interest in the females’ nests. When a male
finds a nest, he slowly walks down one of the guylines to the leaf and then
adopts a distinctive posture with his body elevated. While in this posture, the
male signals to the female inside by suddenly and strongly flexing his legs,
making the leaf rock violently back and forth. The female responds to the male’s
courtship by coming out of the rolled-up leaf and either mating with the male or
140 ROBERT R. JACKSON AND FIONA R. CROSS
driving him away (Jackson, 1985b). Upon seeing one of these leaf nests,
P.fimbriata also goes slowly down a guyline and on to the leaf. Once on the
leaf, P.fimbriata settles next to an opening of the nest and then intermittently
makes signals resembling the courtship signals of Euryattus males. When the
Euryattus female begins to leave the nest, P.fimbriata secures this special prey
by executing a lunging attack (Jackson and Wilcox, 1990; Jackson et al., 1997).
We can use the term ‘algorithm’ to explain what Portia is doing in these
examples, and ‘algorithm’ is often perceived as almost antithetical to cognitive.
It may be tempting to say Portia is being intelligent but not particularly
cognitive. After all, we are familiar with how simple heuristics, or rules of
thumb, can be remarkably effective problem-solving routines (Cross and
Jackson, 2005; Hutchinson and Gigerenzer, 2005; Todd and Gigerenzer,
2000). However, we need to go back to a question we glossed over when first
describing how Portia uses trial and error. What do we mean by ‘appropriate’
when we say that Portia, by trial and error, derives a signal that elicits an
‘appropriate response’ from the resident spider? What makes a response appro-
priate? These questions are interesting because the answers are not always the
same.
When the resident spider is small and not especially dangerous, explaining
what happens may seem straight forward. From Portia’s perspective, an appro-
priate response may be the resident spider behaving as though the web signal is
coming from a small insect ensnared in the web. This seems simple. The
resident spider approaches and, when close, Portia lunges forward and makes
the kill (Jackson and Blest, 1982). Here, Portia seems to join the angler fish
(Pietsch and Grobecker, 1978) as a straight forward example of a predator
practising aggressive mimicry by simulating the prey’s own prey.
For illustrating why it is not always this simple, let us go back to the field and
suppose that this time we find Portia entering the web of a big, dangerous
spider. Simply simulating the resident spider’s own prey now seems like court-
ing disaster. Portia may be at risk of actually becoming, instead of simply
pretending to be, the resident spider’s prey. Portia’s apparent solution is to alter
its definition of ‘appropriate’. Now Portia’s goal when adjusting signals in
response to feedback from the resident spider appears to be almost the antithesis
of the goal when the resident spider was relatively harmless. When facing an
especially dangerous resident spider, Portia seems to be actively avoiding
repetition of signals that might initiate a full-scale attack (e.g. Harland and
Jackson, 2006). Sometimes success for Portia seems to be based on getting the
powerful resident to approach in slow, hesitating steps, as though the resident
spider were not certain of the identity of the signals it receives from Portia.
Alternatively, Portia may make signals that keep the victim calm and station-
ary, with Portia all the while moving in slowly for the kill. Calming effects
might be achieved by monotonous repetition of a habituating signal, as though
Portia were putting its victim to sleep with a vibratory lullaby derived by trial
and error (Harland and Jackson, 2004).
SPIDER COGNITION 141
Calling Portia’s behaviour ‘aggressive mimicry’ draws us to the question of
what Portia mimics. When the resident spider’s response resembles the way the
resident spider normally attacks its own prey, there is no semantic discomfort.
However, when Portia’s signals achieve something like keeping the resident
spider calm, specifying the mimic’s model is less straight forward.
Here, we have another continuum. When we try to specify a particular model
that corresponds to a particular instance of aggressive mimicry, sometimes it is
easy and sometimes it is hard, and sometimes the difficulty is somewhere in
between. What this might be telling us is that envisaging Portia’s strategy as a
predator playing mind games with its prey (see Krebs and Dawkins, 1984) may
often be more instructive than the alternative of emphasizing questions about
models.
Even when we are comfortable with ‘aggressive mimicry’ and think the
model is clearly evident, we might be fooling ourselves. Perhaps we should
envisage aggressive mimicry as being examples of predators communicating
deceptively with their prey. Likewise, perhaps we should think about Batesian
mimicry as being examples of prey communicating deceptively with their
predators. When we switch to talking about communication, we seem to switch
from an emphasis on identifying models of a mimic to emphasis on identifying
the information conveyed by a signal. However, as with the stenophagy–
euryphagy distinction we discussed earlier, we need to consider whose classifi-
cation system we mean when we talk about models or information.
When Portia is in another spider’s web and making signals, there may be
instances when we can specify the model or the information taxonomically. For
example, perhaps we can say Portia’s signals mimic the struggles of a house fly
(M. domestica) in a web. This kind of precision may be very satisfying to us, and
yet it may not be particularly relevant to the resident spider. If we wish to
discuss the models or information that go with Portia’s signals, then we really
need to base what we discuss on the way the resident spider classifies the
different signals it detects in its web. The resident spider’s classification scheme
may, in fact, be less concerned with the taxonomic identity of an object in the
web and more concerned with the categories we think are imprecise (e.g.
‘something in a web that should be approached quickly’ versus ‘something to
approach cautiously’). When the model or the information seems imprecise to
us, Portia may actually be mimicking what makes more sense in the context of
the resident spider’s own way of classifying objects in the web. When we feel
like we are being precise (e.g. by identifying a resemblance to M. domestica),
we may, in fact, be identifying detail that has little relevance to the receiver of
the signal (i.e. to the resident spider).
Perhaps calling aggressive-mimicry ‘communication’ seems somewhat
unconventional, as ‘communication’ is more often used in discussing interac-
tions between conspecific individuals instead of between a predator and prey
belonging to different species. However, we should look more closely at how
conspecific spiders interact with each other.
142 ROBERT R. JACKSON AND FIONA R. CROSS
4.2 WITHIN-SPECIES MIND GAMES
There may be interesting parallels between the communication strategies spi-
ders adopt when interacting with members of their own species and the com-
munication strategies araneophagic spiders adopt when deploying aggressive
mimicry in the webs of other species. The early literature on spider intraspecific
interactions emphasized stereotypy and information transfer (Crane, 1949;
Drees, 1952; Robinson, 1982), but this has been largely supplanted by studies
showing that spiders engage in complex interactions in which they deploy a
large repertoire of signals in highly variable sequences (Cross et al., 2008;
Jackson, 1982a; Jackson and Pollard, 1997).
Two types of intraspecific communication, courtship and threat display, have
the longest tradition as topics in the spider literature (Aspey, 1977; Cross et al.,
2007a; Elias et al., 2008; Huber, 2005; Robinson, 1982). Courtship is a term for
the heterosexual communicatory behaviour that normally precedes mating, but
threat displays are often also understood most readily in the context of mating
strategies. Threat displays are communication by which rivals compete and
determine which individual secures access to limiting resources. With spiders, it
appears to be especially common for threat displays to be a means by which
males compete for opportunities to mate with females. However, caution is
sometimes needed when assigning behaviour to particular categories or strate-
gies. The problem is that, with spiders, mating strategies, predatory strategies,
anti-predator protection and other concerns in the spider’s life have a way of
running together (Jackson and Pollard, 1997).
We can begin by considering P. labiata from Sri Lanka, a salticid that seems
to blur the distinction between courtship signals and aggressive-mimicry signals
(Jackson and Hallas, 1986b). Salticid males typically approach conspecific
females in rapid stop-and-go spurts of activity, punctuated with intervals of
displaying with legs raised, all the while with the female scrutinizing the male’s
displays (Jackson, 1982a), but the courtship sequences of P.labiata are differ-
ent. One difference is that the P.labiata female is usually in a web when a male
comes along and another difference is that she is often the first to display, as
though she were inviting him into her web. The male usually obliges, although
his approach tends to be hesitant and even the slightest movement towards him
by the female will often send him running. When he runs away, he usually
comes back—slowly. Throughout the interaction, the female continues to dis-
play actively, her dominant displays being drumming (pounding on the silk with
her two palps) and tugging (sharp pulls on the silk with her forelegs). From time
to time, the female decamps higher up into the web, after which she turns, faces
the male and resumes displaying.
The male’s displays are visual (e.g. posturing and waving with his legs erect)
and vibratory (e.g. a distinctive stepping gait called jerky walking). When
within reach of the female, the male switches to tactile displays—tapping and
scraping on the female’s body with his legs and palps. These tactile displays are
SPIDER COGNITION 143
performed simultaneously with walking over the female (mounting). Either
while mounting or soon afterwards, the female drops on a dragline with the
male onboard. The pair then mates while suspended from a thread. Now it gets
interesting, and dangerous.
Besides being examples of courtship followed by mating, most of these male–
female encounters of P.labiata are also predatory sequences because, while
suspended on the thread, the female almost always makes a twisting lunge by
suddenly and violently swinging around with her fangs extended and with her
legs scooping towards the male. While mounted, the slightest twitch by a female
is enough to send the male running, but often the male’s reaction is too slow and
he ends up becoming the female’s next meal. For a person watching these grisly
proceedings, and presumably for the P.labiata male, precisely when one of
these twist-lunge attacks will take place is unpredictable, and twist-lunge
attacks are only part of the danger. Even when approaching a female before
mounting or approaching her after having survived a twist-lunge attack and now
returning for another attempt at mating, the male is at risk because a female,
despite appearing passive, may make a sudden, violent lunge forward, stabbing
him with her fangs. Sometimes the female holds on and eats the male immedi-
ately. Other times she stabs and immediately lets go, after which the male flees
but soon succumbs to the female’s venom while the female watches from the
web. The female then walks to the quiescent male, hauls him up into the web
and feeds (Jackson and Hallas, 1986b).
When we considered instances of P.fimbriata females encountering Euryat-
tus females, prey classification by P.fimbriata was an important part of how we
explained what happened. Likewise, when we considered male–female encoun-
ters of P.labiata, prey classification by the female was an important part of the
explanation of what happened. The basis for saying this in both cases is that a
predator deployed prey-capture behaviour that was specific to a particular type
of prey.
When a P.labiata female preys on a conspecific male, we also have an
example of ‘sexual cannibalism’ (Elgar, 1992; Schneider and Lubin, 1998). We
can say the female chooses the male as both a sperm donor and a meal when she
kills and eats him after he begins to copulate. When she kills and eats a courting
male before mating, we can say she rejects him as a sperm donor but accepts
him as a meal. However, what especially interests us here is that, with P.labiata
females, we have an example of a mating strategy and a predatory strategy
thoroughly intertwined with aggressive mimicry. Comparing P.labiata’s male–
female encounters with encounters between P.fimbriata and Euryattus illus-
trates what this means. We can say that P.fimbriata females deceive Euryattus
females by pretending to be Euryattus males. Likewise, we can say that unre-
ceptive P.labiata females deceive P.labiata males by pretending to be recep-
tive. When the P.labiata female is receptive, she may still be predatory and it
becomes ambiguous when we try to decide whether her displays are honest or
deceitful signals. However, there is something else that happens with P.labiata
144 ROBERT R. JACKSON AND FIONA R. CROSS
and this time feminine deceit is unambiguous. The large juveniles of P.labiata
(one moult short of maturity) are physically incapable of mating, and yet, like
adult females, they actively display at conspecific males. Apparently deceived,
males court, mount and perform pseudo-copulations when they encounter dis-
playing juvenile females. While the male seems to be fumbling around search-
ing for genital openings that are not there, the female almost always makes a
twisting lunge, and often the male becomes her prey (Jackson and Hallas,
1986b).
P. labiata males appear to be in a bind because, before they can mate, they
must approach a predator that has evolved prey-specific predatory behaviour
especially for them. Yet the P.labiata male’s predicament may be just an extreme
expression of the predicament of male spiders in general during courtship, and
even more generally to almost all intraspecific interactions of spiders.
On the whole, what we mean when we say animals communicate is that, by
generating a signal, one individual gains indirect control over the behaviour of
another individual (Dawkins and Krebs, 1978). We say control is indirect
because it is done with signals. From this perspective, communication in general
becomes a mind game (Krebs and Dawkins, 1984), but the mind games during
spider courtship and other intraspecific interactions may be particularly likely to
show similarities to Portia’s mind games with other spider species in the
context of predation.
Of course, saying that cannibalism influences male-spider courtship is nothing
new. In the early spider literature (see Robinson, 1982), it was routine to empha-
size cannibalism. The conventional portrayal (review: see Jackson, 1982a) was of
the female spider being a ravenous predator and of the male, when facing the clear
and present danger of being eaten, protecting himself by using courtship. Often
cannibalism was envisaged as a consequence of misidentification. The idea
seemed to be that, by courting, males identify themselves to conspecific females
and females oblige by toning down their predatory inclinations.
This portrayal of male–female encounters has not held up under close exami-
nation (Jackson, 1982a; Jackson and Pollard, 1982, 1990; Richman and Jackson,
1992; Starr, 1988), and it seems to be thoroughly irrelevant to P. labiata.Fora
P.labiata male, the problem is identification, not misidentification. A conspe-
cific female is dangerous to a P.labiata male because she can so readily identify
him and she identifies him not only as a suitor but also as a special kind of prey
against which a unique prey-capture tactic is deployed. Yet, the P.labiata male’s
predicament may be only an extreme illustration of how, for spiders in general,
mistaken identity usually has little to do with females killing males.
As when an aggressive mimic encounters another spider species in a web, it
may be common in encounters between conspecific spider individuals for com-
munication to take on the character of mind games between predators (Jackson
and Pollard, 1997). For example, lycosid and salticid courtship display often
includes posturing with the first pair of legs raised and extended forward (Aspey,
1977; Delaney et al., 2007; Hebets and Uetz, 2000; Richman, 1981, 1982).
SPIDER COGNITION 145
The traditional way to study communication would be to consider whether a
male, by posturing in this way, informs the female about his identity and his
intentions (see Smith, 1977). We do not dispute that considering information can
be an effective way to study communication, but sometimes the emphasis on
information may make it easy to overlook things that are more elementary. For
example, forward-extended legs may function partly as a physical obstruction
(i.e. this leg posture may be like a barrier should the female attack). Information
may also be important, as putting up this barrier may also function as a way in
which a male tells a female that, as prey, he will not be easy to capture. However,
‘telling’ does not mean delivering facts or anything like abstract knowledge
(Barth, 2002). We probably come closer to understanding what is happening
when we emphasize that the male is indirectly controlling the female’s beha-
viour, the idea being that he achieves control by providing stimuli to which the
female responds (Dawkins and Krebs, 1978).
During interactions between conspecific individuals, signals may often func-
tion as a means by which one spider achieves a careful balance between stimuli
that provoke and stimuli that inhibit predatory attack by the other spider. For
example, up-and-down flickering movement of pedipalps and forelegs is char-
acteristic of the initial displays made by the males of many lycosid and salticid
species during courtship (Hebets, 2005; Jackson, 1982a; Richman, 1981;
Rypstra et al., 2003). Flickering movement is also characteristic of stimuli
that alert hungry females to the presence of potential prey (Forster, 1985;
Heil, 1936; Persons and Uetz, 1996, 1997). By beginning courtship with dis-
plays that resemble predatory cues, males may be exploiting the female’s
predisposition to respond to movement in the context of predation (Clark and
Uetz, 1992, 1993; also see Proctor, 1992). We could say the male does this to get
the female’s attention, and much of the complexity and variability that charac-
terize spider courtship might make sense in the context of display by one spider
functioning to solicit and maintain the attention of another spider. For example,
males of some salticid species, like miniature peacocks, seem to dazzle females
during courtship with a garish display of colour (Hill, 2009; Otto and Hill, 2010;
Taylor and McGraw, 2007), with part of what we mean by ‘dazzle’ being that
the male keeps the female attentive.
Like bird song, spider display behaviour might often be explained in part as
each individual striving to avoid the other individual’s ‘monotony threshold’
(see Hartshorne, 1956, 1958). Another way of saying this is that a function of
the complexity and variability of male courtship may often be to overcome the
female’s tendency to habituate. This, in turn, suggests that a female’s level of
choosiness can be readily adjusted by lowering her threshold for habitation
(Jackson, 1982b).
Perhaps it sounds like we are saying females get bored when males become
monotonous and perhaps the idea of spider boredom is worth considering. There
is another context in which something like boredom might be relevant to
spiders. When keeping spiders in the laboratory, scientists often use small,
146 ROBERT R. JACKSON AND FIONA R. CROSS
bare cages. Yet, among people who make a living maintaining much larger
animals in captivity (e.g. zoo keepers), the importance of environmental enrich-
ment is well understood (Wells, 2009). In a remarkable study, Carducci and
Jakob (2000) showed that environmental enrichment matters even to spiders.
More specifically, they showed that salticids became more proficient at
performing cognitive tasks when simply given a spacious cage with a stick
inside as part of routine maintenance.
4.3 MATE CHOICE GOES COGNITIVE—AGAIN
At a time when Darwin’s theory about sexual selection was new (Darwin,
1871), Peckham and Peckham (1889, 1890) investigated the courtship beha-
viour of salticid spiders and related their findings to Darwin’s then controversial
theory. Wallace (1889) is especially well known for parting company with
Darwin over sexual selection and, on a smaller stage, Montgomery (1908,
1909) parted company with the Peckhams over the suggestion that sexual
selection applies to spiders.
Sexual selection went on to become a cornerstone of biology, and nowadays
it is easy to lose sight of one of the reasons for resistance to Darwin’s and the
Peckhams’ ideas. The word ‘selection’ in ‘sexual selection’ sounded like pref-
erence, and for many biologists, anything that sounded like cognition was
unpalatable. Of course, ‘selection’ in ‘sexual selection’ actually refers to an
algorithm (Dennett, 1995), sexual selection being a variation on natural selec-
tion. Yet, Darwin (1872) and the Peckhams (Peckham and Peckham, 1887) were
remarkably open to considering what we now call ‘animal cognition’ (Boakes,
1984). What is more, cognition becomes especially relevant when we consider
examples of sexual selection that are driven by mate-choice behaviour.
Mate choice can be envisaged as the flip side of courtship. Like prey choice,
mate choice implies that an animal has underlying preferences and, as we have
already emphasized, ‘preference’ refers to something cognitive. For much of the
twentieth century, biology and psychology were dominated by views that
actively discouraged talking about human cognition and especially animal
cognition (Morgan, 1896; Skinner, 1938; Watson, 1919). No wonder the Peck-
hams met resistance when they wrote about what sounded to their critics like
spider aesthetics. Beginning about 100 years after Darwin (e.g. Trivers, 1972),
sexual selection assumed its current status as a primary topic in biology, largely
because an interest in ultimate causation, not cognition, became the primary
focus in the literature on animal mating systems (Andersson, 1994). All the
same, a full understanding of intersexual selection requires that we address the
cognitive character of preferences (Dukas, 2004; Shettleworth, 2009).
Our understanding of mate-choice behaviour is rapidly advancing, with
spider research having an important contributing role. For example, Trivers’
(1972) parental-investment theory prepared a generation of biologists to expect
mate-choice behaviour to be primarily, if not exclusively, expressed by females
SPIDER COGNITION 147
and to expect active courtship to be expressed primarily by males. However,
Trivers’ theory was often interpreted too simplistically and research on spiders
is encouraging a shift to appreciating the prevalence of mating strategies based
on mutual mate choice and on active courtship by both sexes (e.g. Cross and
Jackson, 2009a; Hoefler, 2007; Rypstra et al., 2003; Sivalinghem et al., 2010).
Research on spiders is also contributing to a growing awareness of how
preferences are not always static. In species where female spiders mate more
than once (i.e. probably most spiders), expression of preference by females may
often become stronger after the first mating (e.g. Jackson, 1981). Something
different has been shown for E.culicivora (Cross et al., 2007b), as the females
of this salticid species reverse the kind of male they prefer after the first mating.
Both sexes of this species are especially variable in body size and, when mating
for the first time, both sexes prefer larger suitors. However, when given the
opportunity to re-mate, males continue to prefer larger females but females
switch their preference to smaller males.
An unusual feature of this species’ biology may help explain why E.culici-
vora females switch their preference after the first mating. The conventional
idea of how cannibalism is expressed in spiders is of females eating males, not
the other way around. Yet, for E.culicivora, males eat females more often than
females eat males, and we know that the females of this species are especially
vulnerable to being killed by a larger male (Cross et al., 2008). By switching
their preference to a smaller male, females can minimize the mortal risks that
accompany re-mating. It is also the case that, for males, larger females are more
dangerous than smaller females and yet the male’s preferences for larger
females remain intact after mating. Males and females seem to have different
priorities. After mating, females carry eggs, and it is understandable that issues
related to survival become more important for females than for males. However,
at this stage, we know next to nothing about the mechanisms by which these
shifts in preference are made.
5 Cognition via chemistry
5.1 FLEXIBLE LIVING THROUGH CHEMISTRY
The sensory systems of spiders vary considerably, but chemoreception seems to
be a baseline modality common to all spiders (Huber, 2005; Pollard et al., 1987;
Schulz and Toft, 1993; Stowe, 1988; Tietjen and Rovner, 1982) and probably to
all animals (Davis and Ludvigson, 1995; Wilson and Stevenson, 2006; Wyatt,
2003). In the spider literature, it is customary to distinguish between olfaction
and contact chemoreception (Foelix, 1996) with these corresponding roughly to
what we understand intuitively as being smell and taste.
Yet, spider chemoreception may seem particularly alien to us because spiders
use sensors on their appendages for smelling and tasting (Barth, 2001). Maybe it
148 ROBERT R. JACKSON AND FIONA R. CROSS
is especially here that we feel as though we are delving into a spider’s private
world, a world that is unintuitive to us, and one that we can only explore through
experimentation. It may stagger us, however, to learn how intricate and sensi-
tive spider chemoreception can be. We know, for example, that spiders can do
considerably more with chemoreception than just distinguish conspecific from
heterospecific individuals. There are also examples of spiders that can, even
when restricted to using silk- or olfaction-based cues alone, determine a con-
specific individual’s sex, determine whether an opposite-sex individual is a
virgin or has already mated, determine how close a juvenile female is to reach-
ing maturity, discriminate between familiar and unfamiliar rivals, discriminate
between its own and another spider’s eggs and draglines and even determine the
fighting ability of a rival (Blanke, 1972; Clark and Jackson, 1994a,b, 1995a,b;
Clark et al., 1999; Cross and Jackson, 2009b; Dor et al., 2008; Jackson, 1986b,
1987; Miyashita and Hayashi, 1996; Pollard et al., 1987; Roberts and Uetz,
2005; Rypstra et al., 2003; Searcy et al., 1999; Taylor, 1998; Watson, 1986).
Moreover, in the presence of odour from an unseen source, individuals from
some salticid species escalate conflict with same-sex rivals when the odour they
detect is from an opposite-sex conspecific instead of from an opposite-sex
heterospecific individual (Cross and Jackson, 2009a; Cross et al., 2007a).
Spiders are also remarkably good at using chemical cues for identifying their
preferred prey (Clark et al., 2000; Jackson et al., 2002c) as well as for identify-
ing the presence of specific predators (Bell et al., 2006; Eiben and Persons,
2007; Persons and Rypstra, 2001; Persons et al., 2001; 2002; Rypstra et al.,
2007).
With E. culicivora, there is an overlap between mate choice and prey choice
as contexts in which chemoreception is used. Besides identifying its preferred
prey, blood-carrying mosquitoes (Jackson et al., 2005), and potential mates
(Cross and Jackson, 2009b) by olfaction, E.culicivora expresses an olfaction-
based preference for potential mates that have recently fed on a blood-carrying
mosquito (Cross et al.,2009). Moreover, there is a remarkable convergence of
E.culicivora’s use of olfaction with the way Anopheles gambiae, one of the
primary mosquito species on which this spider feeds, uses olfaction. A. gambiae
feeds primarily on human blood, and it finds its blood meals partly by being
attracted to human odour (Njiru et al., 2006). Human odour is also salient to
A.gambiae’s predator, E.culicivora, this being the only known example of a
spider being anthropophilic (Cross and Jackson, 2011).
5.2 OLFACTORY SEARCH IMAGES
‘Image’ in ‘search image’ suggests imagery and, for many scientists during the
five decades following Tinbergen’s (1960) paper, literal interpretation of imag-
ery as pictures in an animal’s head must have made the search-image hypothesis
unpalatable (see Blough, 2006; Kennedy, 1992; Pylyshyn, 2003a,b). All the
while, other scientists have been comfortable with using ‘search image’ as a
SPIDER COGNITION 149
convenient way of talking about selective attention (Cross and Jackson, 2006;
Kamil and Bond, 2006; Langley, 1996). However, accepting that selective
attention is what search images are about means, we should be prepared to
find examples of search images that are independent of vision, as ‘image’ here
does not mean literally a picture. With chemoreception encompassing so many
aspects of spiders’ lives, it would seem particularly promising to investigate
whether spiders also use selective olfactory attention.
There is considerable evidence to suggest that olfactory search images might
be important for many animals and in a variety of contexts (Chittka and Raine,
2006; Schro
¨der and Hilker, 2008; Wyatt, 2003; see also Melcer and Chiszar,
1989), but there are remarkably few experimental studies in the literature that
clearly distinguish between selective olfactory attention and olfactory prefer-
ences. For example, after learning the odour of a particular type of food, skunks
show evidence of detecting this odour in a natural grassy area from greater
distances than before learning (Nams, 1991, 1997). In another study, two groups
of sniffer dogs were tested. Both groups were exposed to trinitrotoluene (TNT)
before tested, but the concentration of TNT varied between groups. When
tested, it was the dogs that had previously been exposed to highly concentrated
TNT that became significantly more effective at finding containers that held
TNT (Gazit et al., 2005).
The findings from these experiments might suggest that skunks and dogs
became selectively attentive to particular odours. However, in these studies, no
specific attempt was made to compare response to odour that was cryptic with
response to odour that was conspicuous. This matters because it is the cryptic–
conspicuous comparison that distinguishes between the effects of selective
attention and the effects of preference. Nams (1991) argued that the concept
of prey being ‘cryptic’ is clearly applicable when we are considering vision, but
not when we are considering other