The typicality effect generally effects to the phenomenon whereby individuals respond faster to more typical examples of a category across many different types of tasks. [1] If a person is asked to think of an example of a bird, it is more likely that they picture a robin than a penguin or bat. [2] These findings were presented in the work of Eleanor Rosch, one of the early researchers to identify the effects of typicality who sparked an extensive topic in categorization literature. [3] [4] Typicality in itself is a gradation in which items of a category can be either very typical to the prototype, moderately typical, or atypical. [5] Using this gradation researchers have conducted robust experiments on participants to tap into the way in which we store and represent categories in our mind. There are conflicting findings and explanations for this phenomenon, all of which demonstrate the influence that typicality gradient has on the way we categorize objects. [6]
Eleanor Rosch proposed that natural categories are continuous with items that are structurally organized based on the degree to which they are judged to be good examples of the category. [1]
This led her and the literature to conclude significant effects of typicality across five domains: [1]
- Subject ratings of the typicality of items
- Order in which category items are learned
- Verification times for category membership
- Probability of an item output
- Expectations generated by the category name
Across several tasks and categories people agree upon typical and atypical examples of a category. [2] Participants rate “chair” as a highly typical example of furniture, and agree that “stove” and “rug” are atypical examples of furniture; likewise, participants will rate cars and buses as more typical examples of vehicles than tractors and wagons. [2]
Empirical evidence of these findings extends across perceptual domains. [1] There are certain colours and forms that are better examples of colours and forms than others. [7] This in turn can affect learning, as Rosch suggested while teaching Dani people the names of colours and forms. Typicality is evident in the results as certain colours and forms are more attractive than the atypical ones, and are more easily remembered than the less salient instances of colour and shape. [1]
Other studies have replicated these findings. [8] Typicality effects have also been shown for categories thought to be well-defined categories; for example “odd” numbers, “even” numbers, and “female”. [8] Even the instances of these categories could be classified by participants in terms of their typicality; “3” is judged to be a better odd number than “501” and “mother” is deemed a better example of a female than “comedienne”. [8]
Typicality effects are also seen in the time it takes participants to verify semantic statements of category membership. [9] [10] The reaction time of participants to respond if a sentence is true is related to the type of noun used for the task priming. [9] That is, if “bird” is flashed before “hawk” as opposed to “animal” before “hawk” participants will respond faster. [9] However, if “mammal” is used for priming instead of “animal” there is an inverse effect; it takes longer to verify that “a dog is a mammal” than “a dog is an animal”. [9]
Furthermore, another study demonstrated the effects of typicality on category-based induction. [11] The degree to which the premises are typical of the conclusion’s category increases the perceived strength of an argument. [11] Consider the following two arguments:
“Robins have a higher potassium concentration in their blood than humans. All birds have a higher potassium concentration in their blood than humans.” [11]
“Penguins have a higher potassium concentration in their blood than humans. All birds have a higher potassium concentration in their blood than humans.” [11]
When Osherson et al. (1990) asked 80 participants which argument is stronger, 73 found the first argument to be stronger. Given that a robin is a more typical bird than a penguin [2] there is evidence of a relationship between the premise’s typicality to the conclusion’s category. [11]
Under the model proposed by Allan Collins and Ross Quillian, a concept is seen as a node in a network, with relational links between nodes and labelled properties associated at each level of node. [12] The links between the nodes are of different types, such as, superordinate (“isa”) links and subordinate links. [13] This forms a hierarchy involving property relations at the subordinate categories and superset relations at the superordinate categories. [12]
This hierarchy suggested by Allan Collins and M. Ross Quillian (1969) suggests that to verify the statement “a canary can sing” requires a person to start at the node “canary” and then at that specific level examine the properties of canary to verify that “can sing” is one of them. For the statement “a canary can fly” one would have to move up in the hierarchy to “birds” to retrieve the property of flying. [12] Moving up a level in the hierarchy requires more processing time during retrieval; hence, it takes longer for a person to verify that a canary can fly than it does to verify that a canary can sing. [12]
Consequently, this ties into notion of cognitive economy, which posits that properties are stored in memory only once, at more superordinate levels, such that “flying” will be stored with “birds” and not again with “robin”. [13] To test whether categories are structured in such a manner, Collins and Quillian (1969) measured the reaction time of participants to verifying statements of different levels. Their results verified their hypotheses but were later challenged by the findings of typicality effects. [13]
A revision of their spreading activation model was created in order to account for typicality effects in semantic verification tasks. [13] The revised model by Allan Collins and Elizabeth Loftus (1975) provided accounts for why atypical instances can take longer to verify or categorize, despite being at a closer level in the hierarchy. [13] Due to experiences, the strength of the links between nodes differ. [13] Frequency of exposure, for example, can strengthen the link between nodes, such that if a person often uses the link that a robin is a bird but rarely uses the link that a penguin is a bird, the former link will be stronger and more quickly verified than the latter. [13] Also addressed in the revised model was the cognitive economy. Collins and Loftus (1975) emphasized that properties did not necessarily have to be stored at the most superordinate category, in fact, they may be stored at several levels in the hierarchy, making them more easily accessible. [13]
Eleanor Rosch’s findings led to her posit the significant of prototypes in categorization. A category in which the prototype is central to the variations of items is easier to learn than a category in which the prototype is fuzzy and a peripheral member. [7] Consequently, when learning a category, the prototype tends to be learned first, regardless of its centrality to the category. [7] Furthermore, when defining the category, the prototype is operationally vital to the definition and is generally the best example of that category. [7]
Earlier research by Posner and Keele in 1968 exemplified the effects of prototypes using dot pattern arrangements. [14] After generating a prototype of a random arrangement of dots, they created random distortions of that pattern; some minor and some major. [3] When they tested participants using these different dot patterns, they found that participants more readily identified the prototype as a member of the category of dot patterns, even if it was not shown during test trials. [3] Furthermore, the patterns that were minor distortions were learned better than the patterns made by major distortions from the prototype. [3]
Some researchers posit that when participants were asked to classify items in artificial categories they appeared to inductively create a central tendency of the distribution which they use to classify the items. [7] This was suggested by Eleanor Rosch, who posited that the central tendency plays a significant role in categorization. For learning colour and forms, it was found that subjects could learn those concepts in which the central members were natural prototype faster than categories organized in different ways. [7]
One potential reason the central tendency is most likely to produce a prototype is because it is presumed that the central tendency minimizes the average distance of category members to the category standards. [6] Hence it is likely to yield a best exemplar, one that is closest to all other items in the category.
There are limitations to the central tendency approach. Central tendency may vary across contexts, based on perspectives. [6] A forest ranger and a pet store owner will have different central tendencies for animal size and ferocity based on their experiences. [6]. Likewise, when comparing experts and novices of a topic, such as trees, it is evident that tree experts do not perceive the central tendency to be the best category member, rather the most ideal tree as the best example. [15] Those experts considered specific dimensions, like weediness, and goals, in assessing which examples are the most typical trees. [15]
Furthermore, another limitation is the inherent induction strength of central tendency which in itself can yield a typicality effect. [4] That is, these central tendencies have a lot of overlap with the features of the category and more specifically, the characteristic features of the category; this in it of itself is enough to make them appear to be the most central to the category, even if they are not. [4]
Ultimately, the factors determining graded structure in one context are different in another context. At least two other determinants play a major role in categorization; ideals and frequency of instantiation.
[6]
Ideal characteristics are those that an exemplar must have if it is to be the best instance of a category and serve the goal of the category. [6] Ideals are not always the central tendency of a category; For example, zero calories would the ideal amount one would consume on a diet, but zero is not the central tendency for number of calories in food. [6] Barsalou (1985) specified that ideals were highly correlated with graded structure for what he called goal-derived categories. He distinguished these categories from common taxonomic categories that Rosch and others were using in their studies. [15]
Frequency may be the reason that it takes longer for people to verify statements involving animals [9] People are exposed to the label “mammal” much less frequently than “animal”. [9] This may explain why participants are quicker to verify that “a dog is an animal” than “a dog is a mammal” despite the fact that “animal” is superordinate or a superset and further away from “dog” in the semantic hierarchy. [9] [12]
Furthermore frequency can explain why penguins are judged to be less typical examples of birds. [5] Given that these studies are conducted in North America, it is presumed that participants have been exposed to robins and sparrows significantly more frequently than penguins. [5] There is a strong debate over the influence that the frequency one experiences the instances in the category plays in typicality effects. [6] While several researchers belittle the effects of frequency; [5] Rosch et al. go so far as to suggest that frequency is a product of typicality and is not in it of itself influential on typicality effects. [1] Subjects have been shown to overestimate the frequency of typical category members versus atypical category members. [1]
When judging how atypical a member is of a category, it has been demonstrated that participants do not rely on frequency. [10] Atypical members are not necessarily the items that people have come in contact with the least or know the least about. [10] For example, “lamp” and “stove” are atypical members of the category “furniture”, but it is assumed that people frequently see these items in their homes. [2] [10]
- ^ a b c d e f g Rosch, Eleanor; Simpson, Carol; Miller, R. Scott (Jan 1, 1976). "Structural bases of typicality effects". Journal of Experimental Psychology: Human Perception and Performance. 2 (4): 491–502. doi: 10.1037/0096-1523.2.4.491.
- ^ a b c d e Rosch, Eleanor (1975). "Cognitive representations of semantic categories". Journal of Experimental Psychology: General. 104 (3): 192–233. doi: 10.1037/0096-3445.104.3.192.
- ^ a b c d Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 22. ISBN 0262632993.
- ^ a b c Rein, J. R.; Goldwater, M. B.; Markman, A. B. (16 March 2010). "What is typical about the typicality effect in category-based induction?". Memory & Cognition. 38 (3): 377–388. doi: 10.3758/MC.38.3.377. PMID 20234027.
- ^ a b c d Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 31. ISBN 0262632993.
- ^ a b c d e f g h Barsalou, Lawrence W. (1 January 1985). "Ideals, central tendency, and frequency of instantiation as determinants of graded structure in categories". Journal of Experimental Psychology: Learning, Memory, and Cognition. 11 (4): 629–654. doi: 10.1037/0278-7393.11.1-4.629. PMID 2932520.
- ^ a b c d e f Rosch, Eleanor H. (1973). "Natural categories". Cognitive Psychology. 4 (3): 328–350. doi: 10.1016/0010-0285(73)90017-0.
- ^ a b c Armstrong, Sharon Lee; Gleitman, Lila R.; Gleitman, Henry (May 1983). "What some concepts might not be". Cognition. 13 (3): 263–308. doi: 10.1016/0010-0277(83)90012-4. PMID 6683139.
- ^
a
b
c
d
e
f
g Rips, Lance J.; Shoben, Edward J.; Smith, Edward E. (1). "Semantic distance and the verification of semantic relations". Journal of Verbal Learning and Verbal Behavior. 12 (1): 1–20.
doi:
10.1016/S0022-5371(73)80056-8.
{{ cite journal}}: Check date values in:|date=and|year=/|date=mismatch ( help); Unknown parameter|month=ignored ( help) - ^ a b c d Malt, Barbara C.; Smith, Edward E. (1982). "The role of familiarity in determining typicality". Memory & Cognition. 10 (1): 69–75. doi: 10.3758/BF03197627. PMID 7087771.
- ^ a b c d e Osherson, Daniel N.; Smith, Edward E.; Wilkie, Ormond; López, Alejandro; Shafir, Eldar (1 January 1990). "Category-based induction". Psychological Review. 97 (2): 185–200. doi: 10.1037/0033-295X.97.2.185.
- ^ a b c d e Collins, Allan M.; Quillian, M. Ross (1969). "Retrieval time from semantic memory". Journal of Verbal Learning and Verbal Behavior. 8 (2): 240–247. doi: 10.1016/S0022-5371(69)80069-1.
- ^ a b c d e f g h Collins, Allan M.; Loftus, Elizabeth F. (1 January 1975). "A spreading-activation theory of semantic processing". Psychological Review. 82 (6): 407–428. doi: 10.1037/0033-295X.82.6.407.
- ^ Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 28. ISBN 0262632993.
- ^ a b c Lynch, Elizabeth B.; Coley, John D.; Medin, Douglas L. (2000). "Tall is typical: Central tendency, ideal dimensions, and graded category structure among tree experts and novices". Memory & Cognition. 28 (1): 41–50. doi: 10.3758/BF03211575. PMID 10714138.
The typicality effect generally effects to the phenomenon whereby individuals respond faster to more typical examples of a category across many different types of tasks. [1] If a person is asked to think of an example of a bird, it is more likely that they picture a robin than a penguin or bat. [2] These findings were presented in the work of Eleanor Rosch, one of the early researchers to identify the effects of typicality who sparked an extensive topic in categorization literature. [3] [4] Typicality in itself is a gradation in which items of a category can be either very typical to the prototype, moderately typical, or atypical. [5] Using this gradation researchers have conducted robust experiments on participants to tap into the way in which we store and represent categories in our mind. There are conflicting findings and explanations for this phenomenon, all of which demonstrate the influence that typicality gradient has on the way we categorize objects. [6]
Eleanor Rosch proposed that natural categories are continuous with items that are structurally organized based on the degree to which they are judged to be good examples of the category. [1]
This led her and the literature to conclude significant effects of typicality across five domains: [1]
- Subject ratings of the typicality of items
- Order in which category items are learned
- Verification times for category membership
- Probability of an item output
- Expectations generated by the category name
Across several tasks and categories people agree upon typical and atypical examples of a category. [2] Participants rate “chair” as a highly typical example of furniture, and agree that “stove” and “rug” are atypical examples of furniture; likewise, participants will rate cars and buses as more typical examples of vehicles than tractors and wagons. [2]
Empirical evidence of these findings extends across perceptual domains. [1] There are certain colours and forms that are better examples of colours and forms than others. [7] This in turn can affect learning, as Rosch suggested while teaching Dani people the names of colours and forms. Typicality is evident in the results as certain colours and forms are more attractive than the atypical ones, and are more easily remembered than the less salient instances of colour and shape. [1]
Other studies have replicated these findings. [8] Typicality effects have also been shown for categories thought to be well-defined categories; for example “odd” numbers, “even” numbers, and “female”. [8] Even the instances of these categories could be classified by participants in terms of their typicality; “3” is judged to be a better odd number than “501” and “mother” is deemed a better example of a female than “comedienne”. [8]
Typicality effects are also seen in the time it takes participants to verify semantic statements of category membership. [9] [10] The reaction time of participants to respond if a sentence is true is related to the type of noun used for the task priming. [9] That is, if “bird” is flashed before “hawk” as opposed to “animal” before “hawk” participants will respond faster. [9] However, if “mammal” is used for priming instead of “animal” there is an inverse effect; it takes longer to verify that “a dog is a mammal” than “a dog is an animal”. [9]
Furthermore, another study demonstrated the effects of typicality on category-based induction. [11] The degree to which the premises are typical of the conclusion’s category increases the perceived strength of an argument. [11] Consider the following two arguments:
“Robins have a higher potassium concentration in their blood than humans. All birds have a higher potassium concentration in their blood than humans.” [11]
“Penguins have a higher potassium concentration in their blood than humans. All birds have a higher potassium concentration in their blood than humans.” [11]
When Osherson et al. (1990) asked 80 participants which argument is stronger, 73 found the first argument to be stronger. Given that a robin is a more typical bird than a penguin [2] there is evidence of a relationship between the premise’s typicality to the conclusion’s category. [11]
Under the model proposed by Allan Collins and Ross Quillian, a concept is seen as a node in a network, with relational links between nodes and labelled properties associated at each level of node. [12] The links between the nodes are of different types, such as, superordinate (“isa”) links and subordinate links. [13] This forms a hierarchy involving property relations at the subordinate categories and superset relations at the superordinate categories. [12]
This hierarchy suggested by Allan Collins and M. Ross Quillian (1969) suggests that to verify the statement “a canary can sing” requires a person to start at the node “canary” and then at that specific level examine the properties of canary to verify that “can sing” is one of them. For the statement “a canary can fly” one would have to move up in the hierarchy to “birds” to retrieve the property of flying. [12] Moving up a level in the hierarchy requires more processing time during retrieval; hence, it takes longer for a person to verify that a canary can fly than it does to verify that a canary can sing. [12]
Consequently, this ties into notion of cognitive economy, which posits that properties are stored in memory only once, at more superordinate levels, such that “flying” will be stored with “birds” and not again with “robin”. [13] To test whether categories are structured in such a manner, Collins and Quillian (1969) measured the reaction time of participants to verifying statements of different levels. Their results verified their hypotheses but were later challenged by the findings of typicality effects. [13]
A revision of their spreading activation model was created in order to account for typicality effects in semantic verification tasks. [13] The revised model by Allan Collins and Elizabeth Loftus (1975) provided accounts for why atypical instances can take longer to verify or categorize, despite being at a closer level in the hierarchy. [13] Due to experiences, the strength of the links between nodes differ. [13] Frequency of exposure, for example, can strengthen the link between nodes, such that if a person often uses the link that a robin is a bird but rarely uses the link that a penguin is a bird, the former link will be stronger and more quickly verified than the latter. [13] Also addressed in the revised model was the cognitive economy. Collins and Loftus (1975) emphasized that properties did not necessarily have to be stored at the most superordinate category, in fact, they may be stored at several levels in the hierarchy, making them more easily accessible. [13]
Eleanor Rosch’s findings led to her posit the significant of prototypes in categorization. A category in which the prototype is central to the variations of items is easier to learn than a category in which the prototype is fuzzy and a peripheral member. [7] Consequently, when learning a category, the prototype tends to be learned first, regardless of its centrality to the category. [7] Furthermore, when defining the category, the prototype is operationally vital to the definition and is generally the best example of that category. [7]
Earlier research by Posner and Keele in 1968 exemplified the effects of prototypes using dot pattern arrangements. [14] After generating a prototype of a random arrangement of dots, they created random distortions of that pattern; some minor and some major. [3] When they tested participants using these different dot patterns, they found that participants more readily identified the prototype as a member of the category of dot patterns, even if it was not shown during test trials. [3] Furthermore, the patterns that were minor distortions were learned better than the patterns made by major distortions from the prototype. [3]
Some researchers posit that when participants were asked to classify items in artificial categories they appeared to inductively create a central tendency of the distribution which they use to classify the items. [7] This was suggested by Eleanor Rosch, who posited that the central tendency plays a significant role in categorization. For learning colour and forms, it was found that subjects could learn those concepts in which the central members were natural prototype faster than categories organized in different ways. [7]
One potential reason the central tendency is most likely to produce a prototype is because it is presumed that the central tendency minimizes the average distance of category members to the category standards. [6] Hence it is likely to yield a best exemplar, one that is closest to all other items in the category.
There are limitations to the central tendency approach. Central tendency may vary across contexts, based on perspectives. [6] A forest ranger and a pet store owner will have different central tendencies for animal size and ferocity based on their experiences. [6]. Likewise, when comparing experts and novices of a topic, such as trees, it is evident that tree experts do not perceive the central tendency to be the best category member, rather the most ideal tree as the best example. [15] Those experts considered specific dimensions, like weediness, and goals, in assessing which examples are the most typical trees. [15]
Furthermore, another limitation is the inherent induction strength of central tendency which in itself can yield a typicality effect. [4] That is, these central tendencies have a lot of overlap with the features of the category and more specifically, the characteristic features of the category; this in it of itself is enough to make them appear to be the most central to the category, even if they are not. [4]
Ultimately, the factors determining graded structure in one context are different in another context. At least two other determinants play a major role in categorization; ideals and frequency of instantiation.
[6]
Ideal characteristics are those that an exemplar must have if it is to be the best instance of a category and serve the goal of the category. [6] Ideals are not always the central tendency of a category; For example, zero calories would the ideal amount one would consume on a diet, but zero is not the central tendency for number of calories in food. [6] Barsalou (1985) specified that ideals were highly correlated with graded structure for what he called goal-derived categories. He distinguished these categories from common taxonomic categories that Rosch and others were using in their studies. [15]
Frequency may be the reason that it takes longer for people to verify statements involving animals [9] People are exposed to the label “mammal” much less frequently than “animal”. [9] This may explain why participants are quicker to verify that “a dog is an animal” than “a dog is a mammal” despite the fact that “animal” is superordinate or a superset and further away from “dog” in the semantic hierarchy. [9] [12]
Furthermore frequency can explain why penguins are judged to be less typical examples of birds. [5] Given that these studies are conducted in North America, it is presumed that participants have been exposed to robins and sparrows significantly more frequently than penguins. [5] There is a strong debate over the influence that the frequency one experiences the instances in the category plays in typicality effects. [6] While several researchers belittle the effects of frequency; [5] Rosch et al. go so far as to suggest that frequency is a product of typicality and is not in it of itself influential on typicality effects. [1] Subjects have been shown to overestimate the frequency of typical category members versus atypical category members. [1]
When judging how atypical a member is of a category, it has been demonstrated that participants do not rely on frequency. [10] Atypical members are not necessarily the items that people have come in contact with the least or know the least about. [10] For example, “lamp” and “stove” are atypical members of the category “furniture”, but it is assumed that people frequently see these items in their homes. [2] [10]
- ^ a b c d e f g Rosch, Eleanor; Simpson, Carol; Miller, R. Scott (Jan 1, 1976). "Structural bases of typicality effects". Journal of Experimental Psychology: Human Perception and Performance. 2 (4): 491–502. doi: 10.1037/0096-1523.2.4.491.
- ^ a b c d e Rosch, Eleanor (1975). "Cognitive representations of semantic categories". Journal of Experimental Psychology: General. 104 (3): 192–233. doi: 10.1037/0096-3445.104.3.192.
- ^ a b c d Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 22. ISBN 0262632993.
- ^ a b c Rein, J. R.; Goldwater, M. B.; Markman, A. B. (16 March 2010). "What is typical about the typicality effect in category-based induction?". Memory & Cognition. 38 (3): 377–388. doi: 10.3758/MC.38.3.377. PMID 20234027.
- ^ a b c d Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 31. ISBN 0262632993.
- ^ a b c d e f g h Barsalou, Lawrence W. (1 January 1985). "Ideals, central tendency, and frequency of instantiation as determinants of graded structure in categories". Journal of Experimental Psychology: Learning, Memory, and Cognition. 11 (4): 629–654. doi: 10.1037/0278-7393.11.1-4.629. PMID 2932520.
- ^ a b c d e f Rosch, Eleanor H. (1973). "Natural categories". Cognitive Psychology. 4 (3): 328–350. doi: 10.1016/0010-0285(73)90017-0.
- ^ a b c Armstrong, Sharon Lee; Gleitman, Lila R.; Gleitman, Henry (May 1983). "What some concepts might not be". Cognition. 13 (3): 263–308. doi: 10.1016/0010-0277(83)90012-4. PMID 6683139.
- ^
a
b
c
d
e
f
g Rips, Lance J.; Shoben, Edward J.; Smith, Edward E. (1). "Semantic distance and the verification of semantic relations". Journal of Verbal Learning and Verbal Behavior. 12 (1): 1–20.
doi:
10.1016/S0022-5371(73)80056-8.
{{ cite journal}}: Check date values in:|date=and|year=/|date=mismatch ( help); Unknown parameter|month=ignored ( help) - ^ a b c d Malt, Barbara C.; Smith, Edward E. (1982). "The role of familiarity in determining typicality". Memory & Cognition. 10 (1): 69–75. doi: 10.3758/BF03197627. PMID 7087771.
- ^ a b c d e Osherson, Daniel N.; Smith, Edward E.; Wilkie, Ormond; López, Alejandro; Shafir, Eldar (1 January 1990). "Category-based induction". Psychological Review. 97 (2): 185–200. doi: 10.1037/0033-295X.97.2.185.
- ^ a b c d e Collins, Allan M.; Quillian, M. Ross (1969). "Retrieval time from semantic memory". Journal of Verbal Learning and Verbal Behavior. 8 (2): 240–247. doi: 10.1016/S0022-5371(69)80069-1.
- ^ a b c d e f g h Collins, Allan M.; Loftus, Elizabeth F. (1 January 1975). "A spreading-activation theory of semantic processing". Psychological Review. 82 (6): 407–428. doi: 10.1037/0033-295X.82.6.407.
- ^ Murphy, Gregory L. (2004). The big book of concepts (1st MIT Press paperback ed.). Cambridge, Mass.: MIT Press. p. 28. ISBN 0262632993.
- ^ a b c Lynch, Elizabeth B.; Coley, John D.; Medin, Douglas L. (2000). "Tall is typical: Central tendency, ideal dimensions, and graded category structure among tree experts and novices". Memory & Cognition. 28 (1): 41–50. doi: 10.3758/BF03211575. PMID 10714138.