concepts & categories in animal cognition Flashcards
lecture 2 of Peter!
what are theory as to why categories are useful?
examplar theory
prototype theory
what is the examplar theory?
A theory of categorization which assumes instances are classified either by remembering the category (exemplar) to which they belong, or by their similarity to instances that have already been classified.(Pearce,2008)
-example seen in Watanabe, 2001
what is the prototype theory?
A representation that corresponds to the average of the members of a category.
Pearce, 2008
-activated whenever an exemplar is presented and elicit the response that is appropriate for the category
-likelihood of theresponse is determined by the extent to which the prototype is activated.
what did Herrnstein et al 1976 do?
investigated category learning in pigeons
-showed the pigeons sets of 80 different pictures in training (half containing trees which were subtle),
-pecks at a response key were rewarded w food only in the presence of pictures w trees
found successful discrimination of novel pictures at test
-even learnt cats of water and people, managed to learn specific people.
what did Watanabe et al 1995 investigate?
the styles of paintings and whether animals can categorize new paintings
what did Bhatt et al 1988 do?
investigated discrimination among multiple categories
-four cats learnt and novel pictures were classified successfully
-performance was maintained when training pictures were never repeated
what did Bhatt 1988 find?
larger sets make for better categories
-produced poorer learning for training items but better performance on test items
what did Huber & Lenz 1993 find?
categorisation is determined by features
- Four stimulus dimensions, each with three possible values (-1, 0, +1)
-Rate of responding determined by feature sum
-62 faces used in the training stage, pigeons rewarded for pecking a response key in the presense of any face for which sums of feature values greater than 0
they found
what did D’Amato et al 1988 find in capuchins?
categorisation is successful
-but of people from non-people,
the monkeys had to touch the photos
photgraphs classified as people when they looked nothing like people,
-as in the training set, only the people slides contained red and non people test slides containing red were misclassified
what did Aust & Huber 2006 find?
perhaps non-human animals dont know what the pictures stand for and just respond on based on low level features
how can you summarise Aust & Huber 2006,2010?
non-human animals can learn categories, these categories allow classification of new items
this appears to be based on features, with little evidence of concept learning.
what are concrete categories?
even when they are very different poctures of people have common concrete attributes
-there are identifiable physical features that are present in both cases
-simple associative learning can explain acquisition of concrete categories
what are abstract categories?
the category that has no concrete representation, based on the presence or absence of one or more physical features
-based on a relationship
-the evidence that non-humans can learn relational categories is mixed
how do human infants categorise?
by using perceptual features rather than abstract category
-evidence reviewed earlier suggests non human animals use perceptual features
-perhaps non-human animals will struggle to learn relational categories
what did Wright et al 1988 find?
same/different learing in pigeons
pigeons were rewarded for pecking at the flanker that matched the sample image
-Group 2-samples: Two sample pictures, fast learning, four possible combinations for G2 SAMPLES
-may have remembered the correct rewponse for each configuration
-novel pictures showed no evidence of a same/different rule
-Group 152-samples: 18 months to learn, but generalised to novel pictures (45,000 combinations possible)
-pigeons found matching to sample difficult but other species master it more easily.
how do pigeons find same/different judgements so difficult?
Wilson, 1985
pigeons can perceive sameness, but lacks salience
-based on the relationship of sameness. They suggested that for species like corvids and apes, the relational cue is of relatively high salience and subjects readily learn to select the comparison stimulus that is the same as the sample. On the other hand, they further suggested that for pigeons the salience of the relational cue is relatively low, so that they will tend to solve matching problems by learning how to respond in the presence of individual training configurations.
may prefer to use individual stimuli and training with lots of instances should help
what did Katz & Wright 2006 investigate?
Trials commenced with the presentation to pigeons of a single photograph, which they were required to peck.
A second photograph and a white rectangle were then displayed below the first photograph
A peck to the second photograph resulted in food if it was the same as the first photograph, but if the two photographs were different then a peck to the white rectangle resulted in food.
-The same effect was seen for rhesus monkeys and capuchins, although their performance was generally better
what is the abstract rule of opposites?
Responses to new (red) items can be based on similarity or an opposites rule
Humans: rule-based (Shanks & Darby, 1998)
Rats and pigeons: similarity-based (Maes et al., 2015)
Humans under cognitive load: similarity-based (Wills et al., 2011)
what did Oden et al 1990 find?
-test of second order relationships in chimpanzees
-Whether chimpanzees perceive the second-order relationship is unclear
-When required to choose the matching pair for food, they failed
-When allowed to choose which pair to handle, they chose the non-matching pair
-When the discrimination was augmented as shown below, they succeeded
what do all the studies mean for concepts and cats?
-lots of evidence that non-humans animals can form categories based on physical features
-stored using a concrete code like mental snapshots
-a moderate evidence of learning about relationships but from matching experiments
what is the exemplar effect?
The finding that familiar exemplars of a category are easier to classify correctly than unfamiliar ones.
Pearce, 2008
what is the feature theory?
A theory of categorization that assumes instances are classified on the basis of whether or not they contain a particular feature, or set of features.
Pearce, 2008.
what are categories as concepts?
-Schrier and Brady (1987, p. 142) made this point when they proposed that monkeys can categorize photographs of people because they possess a “concept ‘humans’ as we commonly understand it”.
Pearce, 2008.
what did Lee consider about categories as concepts?
Lee (1984) considered that it is difficult to specify concepts in animals, he concluded thatif animals acquire concepts then successful categorization need not depend on physical similarity of the members of the category but should also be successful when the exemplars bear no physical similarity to each other.
what is matching?
A test in which a sample stimulus is presented and then followed by a choice between two comparison stimuli, one of which is the same as the sample and one of which is different. The comparison stimulus that matches the sample is correct.
Pearce, 2008
what were the results of Katz & Wright(2006)?
different. Despite this large number, once pigeons were performing accurately on the task, test trials revealed that subjects would respond correctly on no better than 50% of the trials when they were shown novel photographs -> no evidence that the relational cue of sameness had acquired any influence
-later increased size of training sets
results show that It is apparent that as the set size increases so does the accuracy on the tests with the novel photographs.
Pearce, 2008
what does wright et al 2003 suggest?
if apes were tested with the task in Wright & Katz, their performance would be even better than that of the monkeys.
according to Wilson et al 1985 what is the difference between pigeons and apes?
because some can more readily use relational information than others.
-learning based on relational cues is more likely to be overshadowed by learning based on individual configurations in pigeons than in corvids.
what can be the main conclusion from Oden et al 1990?
that chimpanzees are unable to look at two relationships and make an inference about the relationship between them.
Pearce, 2008
what was an advantage in Huber & Lenz for using artifically created stimuli?
that the experimenter knows which features the subjects must use to solve the discrimination. It is also possible to study performance in the presence of different faces, to gain an understanding of the control exerted by these features.
adopted this strategy and found that there was an extremely orderly relationship between the sum of the feature values of a particular face and the rate of responding that it elicited.
Pearce, 2008
what did the stimli in Huber & Lenz 1993 show?
the presense or absence of a feature in a particular location that indicated the category to which an examplar belonged
-> easy to intepret in terms of theories that have just been considered, but possible that animals can acquire categories on a more abstract basis
What did D’Amato & Colombo 1988 find?
The monkeys mastered their problem more readily, but this might have occurred because there were five rather than eight response locations for the monkeys.
When they received their test trials with pairs of stimuli, the monkeys performed accurately with every possible test pair. suggests that these different species solved the serial-learning task in different ways.
shows the magnitude effect
what is the magnitude effect?
The finding that the latency to select the first item from a subset of a list is shorter when it is close to rather than far from the first item of the entire list.
Pearce, 2008
how did the magnitude effect of d’amato et al 1988 work?
Once the spatial representation has been constructed, assume, first, that it can only be accessed from the beginning; second, that the slots can be inspected in a fixed order; and, third, that it takes time to move from one slot to the next. If a test trial is conducted with two items from a list, then the first item will be selected more rapidly the closer it is to the start of the list, the magnitude effect; and the time taken to select the second item will be determined by the number of items that separate them
