Representation: A concept that fills no gaps

Cued and detached representations in animal cognition

This paper analyzes the function of certain aspects of cognition, like planning, deceiving, self-awareness, and communication. I distinguish between two kinds of representations of information. A cued representation stands for something that is present in the current situation. Detached representations stand for objects or events that are neither present in the situation nor triggered by some recent situation. The inner environment of an animal is defined as the collection of all detached representations. The fundamental difference between signals and symbols is that the reference of a symbol is a detached representation, while a signal refers to a cued representation. Detached representations make planning possible. I distinguish between immediate planning, where plans are made for present needs, and anticipatory planning, where future needs are predicted. The evolution of self-consciousness is outlined as a series of steps. The first is when other agents are seen as having an inner environment of their own. This is when deception becomes possible. A further step is when the agent realizes that the other agents' representations of the external world includes a representation of the inner environment of the agent itself. Then the agent can become self-conscious since it can form representations of its own representations.

Specialised use of working memory by Portia africana, a spider-eating salticid

Using expectancy-violation methods, we investigated the role of working memory in the predatory strategy of Portia africana, a salticid spider from Kenya that preys by preference on other spiders. One of this predator's tactics is to launch opportunistic leaping attacks on to other spiders in their webs. Focussing on this particular tactic, our experiments began with a test spider on a ramp facing a lure (dead prey spider mounted on a cork disc) that could be reached by leaping. After the test spider faced the lure for 30 s, we blocked the test spider's view of the lure by lowering an opaque shutter before the spider leapt. When the shutter was raised 90 s later, either the same lure came into view again (control) or a different lure came into view (experimental: different prey type in same orientation or same prey type in different orientation). We recorded attack frequency (number of test spiders that leapt at the lure) and attack latency (time elapsing between shutter being raised and spiders initiating a leap). Attack latencies in control trials were not significantly different from attack latencies in experimental trials, regardless of whether it was prey type or prey orientation that changed in the experimental trials. However, compared with test spiders in the no-change control trials, significantly fewer test spiders leapt when prey type changed. There was no significant effect on attack frequency when prey orientation changed. These findings suggest that this predator represents prey type independently of prey orientation.

A COGNITIVE PERSPECTIVE ON AGGRESSIVE MIMICRY

We use the term 'aggressive mimic' for predators that communicate with their prey by making signals to indirectly manipulate prey behaviour. For understanding why the aggressive mimic's signals work, it is important to appreciate that these signals interface with the prey's perceptual system, and that the aggressive mimic can be envisaged as playing mind games with its prey. Examples of aggressive mimicry vary from instances in which specifying a model is straight forward to instances where a concise characterisation of the model is difficult. However, the less straightforward examples of aggressive mimicry may be the more interesting examples in the context of animal cognition. In particular, there are spiders that prey on other spiders by entering their prey's web and making signals. Web invasion brings about especially intimate contact with their prey's perceptual system because the prey spider's web is an important component of the prey spider's sensory apparatus. For the web-invading spider, often there is also a large element of risk when practising aggressive mimicry because the intended prey is also a potential predator. This element of risk, combined with exceptionally intimate interfacing with prey perceptual systems, may have favoured the web-invading aggressive mimic's strategy becoming strikingly cognitive in character. Yet a high level of flexibility may be widespread among aggressive mimics in general and, on the whole, we propose that research on aggressive mimicry holds exceptional potential for advancing our understanding of animal cognition.

What connectionist models learn: Learning and representation in connectionist networks

Connectionist models provide a promising alternative to the traditional computational approach that has for several decades dominated cognitive science and artificial intelligence, although the nature of connectionist models and their relation to symbol processing remains controversial. Connectionist models can be characterized by three general computational features: distinct layers of interconnected units, recursive rules for updating the strengths of the connections during learning, and “simple” homogeneous computing elements. Using just these three features one can construct surprisingly elegant and powerful models of memory, perception, motor control, categorization, and reasoning. What makes the connectionist approach unique is not its variety of representational possibilities (including “distributed representations”) or its departure from explicit rule-based models, or even its preoccupation with the brain metaphor. Rather, it is that connectionist models can be used to explore systematically the complex interaction between learning and representation, as we try to demonstrate through the analysis of several large networks.