Animal learning and intelligence define how non-human beings solve their living problems based on their individual and social experience. Learning performs adaptive tuning to a changeable environment, and intelligence helps animals to use their learned experiences in new situations. Individual adaptive behavior involves different kinds of learning together with innate behavioral patterns. Classification of learning classes involves basic forms of learning. The modern schema for ordering learning classes makes it easier to work with different forms of learning in animals in comparison with humans and artificial agents.
The rise of scientific study of animal intelligence may be portrayed as progressive changes in experimental methods. The development of objective methods of analysis of animal intelligence is attributed to researches studying animal mind in the nineteenth century, based on Darwin’s evolutionary ideas. In 1870, D. Spalding (1873) experimentally investigated innate and learned behaviors in birds and mammals, and J. Lubbock (1882) was one of the first to introduce apparatus and quantification into the study of animal intelligence. Apart from being a powerful stimulus to the development of experimental investigations, Darwin’s ideas of succession in animal and human thinking gave new arguments for anthropomorphic approach to animal intelligence, and the most known example is the G. Romanes’ book “Animal Intelligence” (1881). The predominance of anecdotal evidence of animal intelligence led one of the pioneers of comparative psychology, L. Morgan (1896), to construct the idea of animal intelligence based on quantitative studies of animals’ reactions to different stimuli. Morgan’s lecture on habit and instinct in animals prompted E.L. Thorndike (1911) to elaborate a novel experimental approach based on the study of animals escaping from puzzle boxes. At the beginning of the twentieth century, two scientific schools that approached learning basing on insight (Gestaltism) and on conditioning (behaviorism, with its Pavlovian and Skinnerian branches) had started almost simultaneously on their efforts to describe learning quantitatively and objectively. After half a century of battles, W. Köhler (1959) invited students of animal intelligence to “forget about schools” and proceed in another direction. The coherent development of ethology and experimental comparative psychology has resulted in cognitive ethology, that is, the comparative, evolutionary, and ecological study of animal minds, including rationality, information processing, and consciousness. Revolutionary experimental paradigms have been developed for studying animal “linguistic” capacities, numerical competence, abilities for rule extraction, sophisticated tool use, complex forms of communication, social learning, and social navigation (for a detailed review see: Reznikova 2007).
Animal intelligence has been experimentally studied for not much longer than a century and controversial ideas still exist about how animals learn and to what limits they understand relations between things and their properties, as well as relations between members of their social groups. Many elegant experimental schemes have been elaborated for investigating how complex are the problems that animals are able to solve. However, there is no common metric for measuring animal intelligence. There is a growing body of evidence that members of different species can solve very complex problems but their cognitive abilities lie within a narrow domain of “species genius.” For example, ants appeared to be more competent (Reznikova and Ryabko 2011) than chimpanzees (Beran 2009) in numerosities, and New Caledonian crows (Kacelnik et al. 2004) are even more advanced than chimpanzees (McGrew 2004) in tool manufacture. There is much work to be done to extend our understanding of whether at least some species share advanced characteristics of intelligence with human beings, or whether all animals think about the world in a way radically different from our own.
To complete the multifaceted panorama of animal intelligence, the working schema of learning classes is needed that involves recent discoveries in the field.
|Level 1 – Habituation or Sensitization.|
|Level 2 – Signal Learning (Classical or Pavlovian Conditioning).|
|Level 3 – Stimulus–response Learning (Instrumental or Operant Conditioning).|
|Level 4 – Chaining (Learning Sequences of Stimulus–response Learning Units).|
|Level 5 – Multiple Discrimination Learning: Concurrent Discrimination Learning (CDL) or Learning Set Formation (LS).|
|Level 6 – Absolute and Relative Class Concept Learning.|
|Level 7 – Using Class Concepts in Conjunctive, Disjunctive, or Conditional Relationships.|
|Level 8 – Using Class Concepts in Biconditional Relationships.|
|3.||Catalog learning (Stimulus–Pattern)|
|6.||Latent learning and exploration|
|7.||Learning set formation|
It should be noted that whereas latent learning, learning set formation, rule extraction, and social learning can be attributed to cognitive abilities, catalog learning, guided learning (Gould and Marler 1987), and imprinting (Lorenz 1935) are based on innate predisposition to build up one set of associations more readily than another. Among these more or less “pre-programmed” forms of learning, “catalog learning” has been described only recently and means animals’ ability to select quickly and to manipulate readily with innate behavioral patterns. Animals look like “cataloging” their repertoire of innate patterns in order to optimize their response to a certain repetitive event (Reznikova 2007). This is a relatively simple, universal, and quite “natural” form of learning that possibly underlies cognition.
The schema for ordering learning classes aims at completing the picture of interactions between different forms of learning in human and nonhuman mentality, and can be applied in cognitive ethology, comparative psychology, and robotics.
Abstract Concept Learning in Animals
Accounting and Arithmetic Competence in Animals
Conditional Reasoning by Nonhuman Animals
Learning Set Formation and Conceptualization
Reinforcement Learning in Animals