Concepts in the Light of Evolution Session 5 Reza Maleeh Institute of Cognitive Science University...

Concepts in the Light of EvolutionSession 5

Reza MaleehInstitute of Cognitive Science

University of Osnabrü[email protected]

University of Osnabrueck 1

Object Permanence

For us, and for all higher animals, Objects exist.

There are objects within objects. Human ideas of what are objects change

with the cultures. Language is not a good guide. Nor is perception a reliable giude.


Object Permanence

Perhaps for some primitive organisms there are no objects.

Three different schemes of reference relevant to understanding systems of perceptual representation:

1. A location system-based system dubbed “feature placing”.

2. A system of “visual indices” referring to things called “proto objects”.

3. The full sortal-based individuation allowed by a natural language.


Object Permanence


Particular objects exist for even so lowly a creature as an earthworm, as Charles Darwin noticed. In his book The Formation of Vegetable Mould, through the Action of Worms (Darwin 1881), he makes it very clear that worms are able to size up objects, check out their shapes, and drag them into their burrows in the most efficient way. The objects they drag are leaves and petioles of various kinds, pine needles, and little round mini-pebbles. (Hurford, 2007, p. 37)

Object Permanence



Object Permanence


It seems likely that earthworms can only manage to represent one object at a time, unlike many mammals and birds, who are capable of representing scenes or events with several objects (Hurford, 2007, p. 38).

Object Permanence: Visible and Invisible Displacement




Naturally enough, every animal that can succeed at the invisible displacement task can also succeed at the visible displacement task.



Visible Invisible

Some Birds no no

Domestic Chicks no no

Parrots yes yes

Apes yes yes

Dogs yes yes

Children over 3 yes yes



Logical Deduction Less Logical Association

Object Permanence


Knowing that an object is permanent takes a cognitive step beyond simple perception of the object. It involves knowing something about an object even when it is inaccessible to perception. During perception, the brain tracks perceived objects by attending to them, concurrently absorbing some information about them. When objects disappear behind a screen, they can no longer be attended to, and only the absorbed information remains, as a representation of the no-longer-present object.

Object Permanence


Even if an attended-to object does not disappear, but the eyegaze is quickly distracted away to another object, this distraction does not impede processing of the information taken in about the first object focused on and used to recognize what it is (Irwin and Brockmole 2004).

Knowledge Based Control


The ability to solve the displacement tasks, especially the invisible displacement task, shows that apes, humans (over age 3.0), and surprisingly parrots can maintain a representation of an object in a mental space controlled by knowledge, rather than perception. The capacity to know something about an object, even when ‘it isn’t there’, is a first step along the road to the impressive characteristic of human languages: their capacity for displaced reference.

Biological Motion and Animacy

‘Biological motion’ is a label attached to a kind of motion typical of an animal;

It is distinct from trees waving in the wind, rocks tumbling down a cliff, waves in the sea, or eddies in a stream.

Recognizing biological motion is not just a matter of certain sensors being excited.


Biological Motion

In many animals, including humans, a disposition to recognize biological motion is hardwired

The biological motion studies take off from a paper by Johansson (1973), who showed that human babies attend preferentially to mobile patterns generated by lights attached to the joints of moving animals.


Biological Motion


Biological Motion


Biological motion is a basic component of higher-level perceptual judgements.

Animacy


Hauser (1998)

Animacy


‘Some feature other than self-propelled motion accounts for the tamarins’ looking time responses and at least one candidate feature is whether the object is animate or inanimate’ (Hauser 1998b, p. 31)

Evolutionary Succession: From Proto-Concepts to Linguistic Concepts

Regular and systematic behaviour in connection with a thing

Sufficientcondition

Possession of Language

Sufficientcondition

GeneralizationFree access and control over mental states

24

Necessarycondition

Necessarycondition Proto-Concepts Pre-Linguistic

ConceptsLinguistic Concepts

No Clear Border

Animacy


Biological motion and animacy differ in degree of abstraction. Biological motion involves an immediate perceptual (possibly multimodal) response to an experience, whereas animacy reflects a more permanent, and less perception-dependent, judgement. A sleeping dog is animate, but (except for dreaming twitches) does not exhibit biological motion. Animacy is a generalization from biological motion. Anything which could exhibit biological motion, though it may be inert at the time of referring to it, is credited with animacy. Animacy is potential biological motion.

Animacy



Animacy


The step from biological motion to animacy involves more than just perception; it involves forming a hypothesis about objects exhibiting biological motion. Barrett (2004) lists a number of criteria giving clues to agency (equivalent to animacy here). As he argues, there are strong evolutionary pressures on any animal to detect animacy in the objects around it, and clearly almost all species have solved the problem quite well.

Animacy


Inferring animacy in objects is the first step towards a fullblown ‘Theory of Mind’ necessary for communications of one’s inner representations to other members of one’s community.

Transitive Inference


Example: If Eve is older than Jack and Jack is older than Rosie, it follows that Eve is older than Rosie.



Humans can learn the transitivity of a relation without necessarily being aware in any explicit sense of what they are learning, as an experiment by Greene et al. (2001) has shown.



Can Animals do it? If yes then they are capable of:

1.Making generalizations that a relation is transitive: induction2.Doing the inference: deduction



Example:

Long Term Memory

Semantic Memory


There can be a kind of semantics, independent of language, as the set of representations in an organism’s brain corresponding to regularities experienced in external objects and situations.

Semantic Memory


Example: Kanzi signed fairly explicitly that he wanted the group room unlocked so that he could see his mother, Matata (Segerdahl et al. 2005, p. 59).

A Claim


A languageless animal can have small-scale models of

external reality

Two Hints

What is private is nevertheless real and can be intruded upon by science.

When we penetrate the barrier between public and private, we enter a new territory where familiar public instruments don’t help us to navigate.


Mental Representations


Outside the brain there are colours, smells, noises, touchable surfaces, hot and cold. All these external stimuli get transformed in the brain to patterns of firing of neurons. By carefully watching the firing of an animal’s sensory neurons while controlling its outside environment, clear correlations can be found between patterns of neural activity and classes of external stimuli. This is on the perception side , what is input to the brain.


Outside the brain there are colours, smells, noises, touchable surfaces, hot and cold. All these external stimuli get transformed in the brain to patterns of firing of neurons. By carefully watching the firing of an animal’s sensory neurons while controlling its outside environment, clear correlations can be found between patterns of neural activity and classes of external stimuli. This is on the perception side, what is input to the brain.



On the output side, in the public world, animals blink, swallow, reach, grasp, turn their heads,move their arms and legs. Inside the brain there are no arms and legs to move, but classes of an animal’s bodily movements can be correlated with patterns of motor neurons firing.


S: A sensor converting physical signals into neural impulses (type 1).A, B, C, …: Sequential stages, where one neural activity pattern P is converted into

another (type 2).

Hypothetical set of sensory information-processing stages in a nervous system.

Short-Term Memory: Persistence of neural activity patterns specific to a given

sensory stimulus for several seconds after the original stimulus has disappeared.

We further assume that there is a two-way information-driven relationship between the consecutive

neural levels A, B, C, … such that when for instance, pattern B is evoked in some independent way

at a later time, the original pattern A to which it corresponds, is elicited. If this is the case, we say

that information in pattern A, and hence also on the corresponding sensory input S, was stored in

Long-Term Memory.

Thinking or reasoning, and imagination are the act of information recall, alteration, and restorage without any external input. Anticipation also comes about when certain features of a frequently occurring sensory input trigger neural activity in relevant primary sensory areas of the cortex even before an actually occurring feature can elicit the corresponding response, if the expected feature is missing, the corresponding neural pattern appears anyway.

Memory recall: The above mentioned feedback process, through which an internal pattern (without any need to an external input) can elicit corresponding patterns at lower levels.



Hubel and Wiesel (1959) showed that different individual neurons in cats were responsive to diagonal lines at different angles. Information from such cells is fed forward to cells which combine input from both eyes (Hubel and Wiesel 1968); this is an early instance of the integration of simple information into a more complex assemblage of information. Again at a basic level, photoreceptorneurons with different pigments in the retina are sensitive to different wavelengths and different intensities of incoming light.

Mental Representations: an Example


But a given photoreceptor cannot distinguish a bright light with a little green in it (say) from a dimmer light with more green in it, so its firing response is ambiguous as to the ‘real’ colour input. Real-colour vision is achieved by higher levels of the visual system integrating whole patterns of signals from many different photoreceptors, sensitive to different wavelengths. The overall message is that the sensations that we can become consciously aware of, such as the colour or shape of an object, are the result of much transformation, via many integrating and differentiating way-stations.


Binding is the transformation of radically different patterns into single patterns that, for example in seeing an object, are in correspondence with the topological properties of the form of the object and with the spatial position of the object in the environment.

Binding

Binding: an Example

Stein et al. (1998, p. 440) mention an example in which ‘the visual stimuli and the auditory stimulus are not very effective when presented individually within their respective receptive fields. However, when presented at the same time, their combined response is enhanced far beyond the sum of the two unimodal responses’.

Associative Learning


Binding and Evolution

Stein et al. (1998) also note the extremely ancient evolutionary foundations of such multisensory convergence and integration, which, they say, ‘are evolutionary ancient schemes that antedate mammals and even the evolution of the nervous system’ (p. 440).

Binding and Biological Motion

Detection of biological motion is very complex, involving tracking different shapes moving in rough, but not perfect, unison. In humans, it has been shown that different parts of the brain are actively involved in the processing of different types of biological motion. Bonda et al. (1996, p. 3737) showed that the interpretation of different types of biological motion engages brain systems in a differential manner.


Goal-directed hand action

Posterior part of the left hemisphere within the

intraparietal sulcus and the caudal superior temporal

sulcus.

Expressive body movement

Interaction between temporal neocortex and limbic areas


An ability to distinguish at a distance between a frog and a mouse, by its characteristic motion, shows that specific types of motion are components of our concepts of these creatures.

Animacy

At other times, other characteristic features come into play: we can tell a dead frog from a dead mouse. Our concept of a mouse, say, is a complex of many different features, its size, its shape, its colour, its way of moving, maybe the texture of its fur, each of which we can be separately aware of.

Biological Motion

At other times, other characteristic features come into play: we can tell a dead frog from a dead mouse. Our concept of a mouse, say, is a complex of many different features, its size, its shape, its colour, its way of moving, maybe the texture of its fur, each of which we can be separately aware of.

The point to emphasize here is that animals can make such categorial distinctions, too, and do so by using component features of the categories. An animal that can distinguish a mouse from a leaf has categories of MOUSE and LEAF and also has some category of biological motion which we can label MOUSY-MOTION, and it uses the category MOUSY-MOTION in its attribution of the category mouse to some perceived object.

Biological Motion

It is all done, in non-humans, without the help of language. And in fact for humans, much of this work is also done without the help of language. Many of the features involved in recognizing different classes of objects have no convenient linguistic labels, although we can concoct clumsy multi-word expressions which roughly capture the features we use.

Biological Motion & Language

Consider: walk, jog, limp, strut, and shuffle. It would be artificial in the extreme to try to invent distinct ‘semantic primitive’ verbal labels for what distinguishes these actions from each other. As Jackendoff says, ‘we can distinguish [the meanings of] these words by their appearance (and how they feel in the body)’.

Example

All such machinery can safely be taken to exist, in some form or other, in animals closely related to us.

Before Last Words

Animals just have not taken the human step of applying linguistic labels to some combinations of features regularly delivered by their senses.

Before Last Words

Details of such a system have been spelled out in neurological terms by Damasio (1989) in his ‘Convergence Zone’ (CZ) approach to storage and retrieval of concept-like entities in the brain.

Before Last Words

‘[T]he normal process of retrieving words that denote concrete entities depends in part on multiple regions of the left cerebral hemisphere, located outside the classic language areas’

Before Last WordsBickerton (1995, p. 24): The only holistic cat is the linguistic cat

Hurford (2007, p. 55): But unified cross-modal concepts do exist before language, in animals and in babies.

Perceptual Symbols: Barsalou (1999, p. 557)


During perceptual experience, association areas in the brain capture bottom up patterns of activation in sensory-motor areas. Later, in a top-down manner, association areas partially reactivate sensory-motor areas to implement perceptual symbols. The storage and reactivation of perceptual symbols operates at the level of perceptual components—not at the level of holistic perceptual experiences. Through the use of selective attention, schematic representations of perceptual components are extracted from experience and stored in memory (e.g., individual memories of green, purr, hot). As memories of the same component become organized around a common frame, they implement a simulator that produces limitless simulations of the component (e.g., simulations of purr). Not only do such simulators develop for aspects of sensory experience, they also develop for aspects of proprioception (e.g., lift, run) and for introspection (e.g., compare, memory, happy, hungry).

So:


The brain, having learned from perceptual and/or proprioceptive experiences, is able to recreate or rehearse versions of those experiences in the absence of the original perceptual or proprioceptive stimulus.

So:


2. Common neural mechanisms underlie imagery and perception.

1. Perception underlies cognition

Concepts in the Light of Evolution Session 5 Reza Maleeh Institute of Cognitive Science University...

Documents

Transcript of Concepts in the Light of Evolution Session 5 Reza Maleeh Institute of Cognitive Science University...