Discussion Forum—A Way with Words, a fun radio show and podcast about language

Table of Contents

1 Introduction 2
1.2 Thesis statement 2
1.3 Method, Outline & Operationalisation 2
2 Literature Review 3
2.1 Attentional Learning Account 3
2.2 Essentialist Account 3
3 Analysis 4
3.1 Evidence for the Attentional Learning Account 4
3.2 Supporting the Essentialist Account 6
3.3 Recapitulation 8
4 Discussion 9
4.1 Explanatory force – Attentional Learning Account 9
4.2 Explanatory force – Essentialist Account 10
4.3 Recapitulation 11
5 Conclusion 11
6 Bibliography 13

 
1 Introduction
According to Steven Pinker, language is a miracle, because it gives us the ability to communicate and bind informative-sharing networks (Pinker, 1994; 15-16). Therefore, language is a crucial tool, which children have to acquire no matter where in the World they are born. Normally, children learn their first word at the age of 12 months, and at 16-18 months of age they are relative skillful at word-learning (Bloom et al., 1998; 67)
This paper will provide a comparative analysis of how children learn the meanings of words. Two competing approaches will be contrasted: The attentional learning account (also called the “brute-shape” account) put forward by Smith and colleagues and the essentialist account (“the “shape-as-cue” account) offered by Bloom and by Booth and Waxman. It might be utopia to believe that this particular paper will give the long sought answer, but it is however plausible to expect that it will at least point in a direction that will give support to one of the two approaches. It is important to note that both approaches are widely acknowledged in academia, which makes it interesting to see if one turns out to be more convincing. This paper is composed on the basis of the thesis stated below.
1.2 Thesis statement
How do children learn the meaning of words? What is the evidence in favor of a) the attentional learning account and b) the essentialist account? Given the available evidence, which of these two accounts is most convincing?

Even though language is a complex systems which consists of many rules, infants are able to acquire their native languages in natural and rapid ways (Waxman et al., 2000; 34-35), but why is this? This is a very interesting question, which is indeed worth studying further.
1.3 Method, Outline & Operationalisation
The outline of the paper is as follows: to begin with the section called: Literature Review, will briefly introduce and explain the two approaches. Hereafter the analyzing part follows in which the intension is to test and analyze how and with what arguments the two different approaches explain how children learn the meanings of words. Next step is then to discuss and later evaluate which of these approaches seem to have the most convincing explanation as to how children learn the meanings of words. Finally, the conclusion will answer the paper's thesis by stating which approach has the most convincing arguments as to how children learn the meanings of words.
The thesis of the paper will be answered by conducting an inductive analysis of how children learn the meanings of words. More precisely this will be carried out by analyzing the arguments different scholars have given to support one of the two previously mentioned approaches.

Obviously there are also many other acknowledged approaches as to how children learn the meanings of words. The reason why these abovementioned theories in particular were chosen was due to a wish to select two very different approaches – although both with apparent explanatory forces. This inspires for an exciting investigating of a controversial topic, which has been much discussed in academia for centuries – and probably will for many years to come.
2 Literature Review
As mentioned earlier, the intention of this section is to introduce and explain the two approaches included in this paper; that is at first the attentional learning account and secondly the essentialist account.
2.1 Attentional Learning Account
According to Markson et al. the attentional learning account prescribes that a child will extend the name of an object (e.g. ball) on the basis of shape, because the child has learned that when people say ‘this a ball', they also have a tendency to point toward an object which is shaped like a ball (Markson et al., 2008; 204). When children reach the age of 3 years, they easily learn to name new objects, because they extend new names to unfamiliar objects on the basis of similarities in shape. Shape is the perceptual property that matters most when a child has to determine what category an object falls into (Smith et al., 2002; 13).
In continuation hereof, it is worth mentioning what is called ‘Shape Bias'. This phenomenon emerges once the child has reached a point where he or she makes the second-order generalization, which implies that words in contexts such as ‘This is a ball' are generalized on the basis of shape. According to the attentional learning account, this phenomenon is a fundamentally linguistic issue, which is learned through exposure to speaking adults (Markson et al., 2008; 204-205).
2.2 Essentialist Account
The essentialist account is also known as the pragmatic cuing account or shape-as-cue account. Children use categorization in the process of word learning, which means that once a child recognized that something is a part of a certain category, the child already knows relevant things about this entity. According to Bloom, these categories are products of psychical laws, biological evolution, and intentional design (Bloom, 2001; 1101).
However, similarities are not enough. Children also need to learn that looking alike do not necessarily mean that things share the same qualities. Subsequently, children need to understand that the category which an object falls into is also determined by deeper properties of the object (i.e. internal structure). What separates an essentialist notion is hence that an essentialist views the possibility that something might look like an object, but not actually be this given object. Moreover, the essentialist knows that just because a thing does not resemble a specific thing, it might be this thing nonetheless. That is, children know that category membership is not solely determined by appearance, it also depends on the quality and internal structure (Ibid; 1101-1102).
3 Analysis
As mentioned earlier, the intention of this analysis is to test and analyze how and with what arguments the two different approaches explain how children learn the meanings of words.
3.1 Evidence for the Attentional Learning Account
According to the attentional learning account children acquire the meanings of words by turning to the shape of objects (Markson et al., 2008; 204). When children at the age of 3-4 years are taught a new noun for a new object, they have a strong tendency to extend the name to other objects that are similar in shape to the target objects, rather than to objects that are similar in size, texture, or color (Ibid; 168).
In an important study, Smith et al., found that children at the age of 17 months repeatedly showed a tendency to generalize by shape. This meant that when the unfamiliar objects had to be given a name, this way done a notion that…: “only objects with similar shape have the same name” (Smith et al., 2002; 13). Eight boys and eight girls were recruited, and they were randomly assigned to either a trained- or a baseline condition. For 9 weeks children from the trained condition repeatedly heard names of objects from unfamiliar object categories. These categories were well organized so that the objects differed in all properties except shape. The children learned that only objects with the similar shape had the same names. Results showed that the children in the training condition extended the trained names to new objects by shape 88 % of the time, while children from the baseline condition selected a shape-match only 36 % of the time, which was approximately at chance level. This meant that children in the trained condition made the first-order generalization that the taught names did not only refer to the trained objects, but also to other objects that were similar in shape. Moreover, results showed that children from the trained condition generalized new names for new objects on the basis of shape 70 % of the time, while children from the baseline condition only generalized on the basis of shape 34 % of the time. This meant that children in the trained condition also made the second-order generalization. After, the children had learned four lexical categories organized by shape, they showed a strong tendency to extend each of the trained names to new objects by shape, and in addition they also showed a general tendency to extend on the basis of shape when they named new objects (Ibid; 13-15). Later, these findings were used in the second part of the experiment. Experiment 2 included three between-object conditions, where the replication condition replicated the training from experiment 1. The next condition was called varied-category-structure condition, and here the children were taught four names for four categories like in experiment 1, but instead, two of the categories were organized by similarities in color, and the last two were organized by similarities in texture. This meant that there was no basis for higher-order generalization, because the categories did not only rely on one property. In the last condition, the no-name condition, the children were taught about four categories, but none of these were provided with any names (Ibid; 16-). Children were randomly assigned to the three conditions with 4 boys and 4 girls under each. The tests for first-order and second-order generalization were structured as in experiment 1. During the test for first-order generalization, the exemplars were the same as it had been during training for the children in the replication- and varied-category-structure conditions, but the names were new to the children in the no-name condition. When testing for second-order generalization both the exemplars, the names, and the choice objects were new to all of the children. Results showed that children in the replication- and no-name conditions clearly had learned that the categories were organized according to shape. They extended the name to same-shaped objects respectively 66 % and 62 % of the time. It was interesting to see that even though the children in the no-name condition had never heard the name, they still generalized the names by shape, which shows that these children had learned to attend to shape of objects. The children in the varied-category-structure condition did not generalize on the basis of shape, but attended to the texture or color of the trained objects 67 % of the time. This meant, that children in all of the three conditions made first-order generalizations, as they generalized the name for a trained category to new objects on the basis of the property which had organized the trained category. It was only children in the replication condition who made the second-order generalization, and extended new objects names to new objects by shape, 66 % of the time. Children in the other two conditions did not select matches in either: shape, texture, or color that differed from change level. The fact that children from the no-name condition made the first-order generalization but not the second-order suggests that learning names, and not just the shape-based categories, is crucial in regard to making the second-order generalization. Altogehter, this study indicates that shape determines what name a new word is given, which gives support to the attentional learning account.
In a study by Gentner (1978) form was contrasted with function. The study contained three objects, two that differed both in form and function (called a jiggy and a zimbo), and a hybrid that was used as a test object. The children were taught the name of the jiggy in a natural way, so they would not think of it as an experimental device. Additionally, the children learned to use the jiggy, which was a square box with a face that would change expression when the child would move the box up and down. The children saw the zimbo in another room while they participated in an unrelated study, and were given the name of the zimbo when they asked what it was. The zimbo was a machine gun, which contained jelly beans. Afterwards, the children and the experimenter went back into the room with the jiggy to make sure that the children remembered the object. Finally, the children were lead to a third room where they were introduced to the hybrid which looked like the jiggy. Before the children could say anything, the experimenter would ask if the child could make it work. Yet, when the children played with the toy, the hybrid did not change facial expression like the jiggy, but instead it produced a handful of jelly beans, like the zimbo. The children were then asked to give the name of the hybrid. The results showed that the very youngest children would respond with one of the original names, and therefore they based their answer on form and not function. This meant that it was form (or shape) that determined which name was given to the new object. The older children and the adults responded with a combination, but when they were asked to choose one of the names, the middle-aged children based their' answer on function and named it zimbo, while older children mainly based their' answer, on the basis of form and named it jiggy (Gentner, 1978; 137-141). Smith's study chiefly supports the attentional learning account, but the finding revolving the middle-aged children is a problem as function rather than shape is decisive.
3.2 Supporting the Essentialist Account
According to an alternative account, called the essentialist account (or shape-as-cue-account) children learn the meaning of words, not by using shape as the determiner, but by using shape as a cue to the meanings of words (Diesendruck et al., 2003; 168).
A number of findings have shown that children from an early age have a sense of object kind (Ibid; 168), and in addition the youngest children have an impressive store of conceptual knowledge (Booth et al., 2005; 491). According to Bloom (2001), children are able to tell whether a new word e.g. dog refers only to this dog, or to dogs in general. This information is not present in a direct sense when the sentence is uttered, but children always get it right, because they turn to the syntax of the words (Bloom, 2001; 1096). This indicates that children have fundamental knowledge, which they use when they are searching for cues to determine objects kind. This fact gives basic support to the essentialist account.
According to Bloom, the relationship between count nouns and objects of the same shape is not due to a direct association. Instead, it exists because children assume that count nouns refer to object kind, and then consider shape to be a reliable cue to what kind an object belongs to. Because of their believe about object kind, children from the age of one year understand that only count nouns and not proper names extend to kinds (Diesendruck et al.,. 2003; 168). This supports the essentialist account, because it indicates that shape is not the only cue children use when they learn the meanings of new words. According to Diesendruck and Bloom (2003), shape is less important when additional information about the object is known. This means that the name an object gets is determined by the kind to which it belongs. It is important to note that shape is only a cue to kind membership. This means that something may be a certain object even though it does not look like it. The salient point is whether the object possesses the same intended function and internal structure as the object in question (Ibid; 169). The fact that deeper properties of an object allegedly override shape supports the essentialist account.
In an important study by Booth and Waxman (2008) the intension was to show that word-learning is not only determined by perceptual- but also by conceptual information. They showed 3-year-old children a novel target object, and labeled it with a novel count noun. The children were then asked to extend the noun to a series of test object that only matched the target object on one perceptual dimension either on: shape, size, or texture. In contrast to many previous studies, each of the target objects were introduced within the context of a short vignette. This design made it possible to hold constant the perceptual properties of the objects. There were two different classifications; an animate condition and an artifact condition. In one condition (animate condition), the children were asked to classify heard vignettes, which described the targets as animate objects with a sentence like: “e.g…. has a mommy or daddy who loves it very much” (Booth et al., 2008; 187). In another condition (artifact condition), the children were asked to classify heard vignettes that described the targets as artifacts, where they were introduced with a sentence like: “e.g…. was made by an astronaut to do a special job on her spaceship” (Booth et al., 2008; 187). Booth and Waxman predicted that if conceptual information would permeate word learning, the children that were assigned to the artifact condition would extend novel nouns primarily on the basis of shape, while children assigned to the animate condition would do so on the basis of both shape and texture. Their' prediction was confirmed (Booth et al., 2008; 187). They found that these results support the essentialist account, because they indicate that other properties than shape determine how children extend the meaning of a word. In this specific study, the conceptual information given in the vignettes, meant that the children's extensions differed systematically as a function of the vignettes (Ibid; 187). Thus, children learn the meanings of words on the basis of deeper properties of the objects, and not solely by paying attention to shape. In 2005 Booth and Waxman did another study, where they once again wanted to examine if conceptual information permeates word learning. This time a new object was initially labeled with a new word. As previously, results showed that conceptual information matter. When the target objects were artifacts, infants extended on the basis of shape, but when the targets were animates, children extended on the basis of both shape and texture. As earlier, these results challenge the notion that expectations in word learning emerge late, and moreover the notion that word learning rests entirely on perceptual information (Booth et al., 2005; 491-500), because shape is merely a cue to object kind rather than the salient point.
As mentioned earlier, Bloom claims that objects fall into categories on the basis of physical law, biological evolution, and intentional design (Bloom, 2001; 1101). Intentional design means that a word is used in the intended context to express a certain notion. According to Bloom, these categories are rich, which means that once a person knows that an object is a member of a category, that person also knows relevant information about that object. Yet, there are also poor ‘unnatural' categories, which consist of objects that are not really similar. Perceptual similarity may be a good cue when forming concepts, but Bloom argues that perceptual similarity is enough. Instead, he claims that a child has to categorize maximally, which means that what category an object falls into is determined by deeper properties of the object; that is the internal structure (Ibid, 1101-1102), which explains the ‘unnatural' categories. Bloom's argument consists with the essentialist account, because it states that not only perceptual facts influence how children categorize, and hereby learn the meaning of words.
3.3 Recapitulation
This analysis has shown that both of the approaches have many supporting arguments. What supports the attentional learning account is the fact that several studies have concluded that children extend names to objects on the basis of shape. On the other hands, the essentialist account also has some convincing arguments, e.g. Booth and Waxman's (2008) findings.
All together this makes up a good foundation for an important and essential discussion of which approach then seem to have the most adequate and convincing arguments as how children learn the meaning of words.
4 Discussion
The analysis above has indicated that either of the two approaches have many explanatory elements as to explaining how children learn the meanings of words. But it is however worthwhile to compare and discuss which one seems to be the more convincing one, because the approaches also seem to have some shortcomings.
4.1 Explanatory force – Attentional Learning Account
The analysis above has indicated that the attentional learning account has many supporting argument, and some appears to be very convincing.
The most convincing piece of evidence comprised from the attentional learning account seems to be the fact that several studies have concluded that children extend names to new objects on the basis of shape (Smith et al., 2002; 13). The fact that several studies indicate the same tendency makes the attentional learning account very convincing. It is, however, problematic that other studies point towards a different tendency. As noted earlier, other studies have indicated that deeper properties of objects also influence how a child learns the meaning of new words (Booth et al., 2008; 187). This appears to be a problem, which on one hand weakens the explanatory force of the attentional learning account, but on the other hand it may be explained. As stated earlier, Gentner has shown that the age of a child is crucial, because it decides whether or not shape alone is determining. Young children up to 5 years of age generalize on the basis of shape; and the same thing goes for adults [15+]. While the 5-15 years-olds generalize on the basis of function (Gentner, 1978; 137-141). This could explain the contradictory results, which does not necessarily make the attentional learning account less convincing. However, it indicates that maybe scholars in favor of this account, should be more explicit by perhaps including a intermediate variable that ascertains that this shape-based approach applies to the youngest children.
Another crucial piece of evidence is Smith's study from 2002, which was introduced earlier. The experiment showed that children have a strong tendency to extend name to both familiar and new objects on the basis of shape (Smith et al., 2002; 13-15). It is worth noting that the study is well executed and the design of experiment 2 eliminated the basis of a higher-order generalization that objects were organized by only one property. It is difficult to discuss and question Smith's finding, because the results clearly showed that shape is very decisive when children extend names to new objects, because 88 % of the trained children did so in experiment 1, and in experiment 2 the children extended names to same-shaped objects 66 % and 62 % of the time (Ibid; 15, 17).
Altogether, the attentional learning account has many convincing argument, but is, however, a problem that studies are ambiguous. While many studies support the attentional learning account, there are also others that indicate that other factors than shape determine how a child learn the meaning of words.
4.2 Explanatory force – Essentialist Account
The analysis above has shown that the essentialist account also has many supporting argument, and some of them are also very convincing.
The most factual piece of evidence appears to be Booth and Waxman's study from 2008, in which they proved that not only perceptual- but also conceptual information influence the word-learning process (Booth et al., 2008; 187). This argument appears convincing, because the scientific works seems contemplated and creditable due to the consideration of intermediate variables etc. Moreover, the study supported not only the essentialist account but also Booth and Waxman's initial prediction, - a prediction which presumably builds on prior experience and knowledge from dealing with this topic.
As stated in the analysis, more scholars have argued that certain features of objects overrides the shape of an object. Bloom has claimed that deeper properties are more important than shape when a child has to decide what kind of category an object falls into (Diesendruck et al.,. 2003; 169), which later influence how a child learn the meaning of words. Bloom (2001) underpinned his argument by pointing towards findings that show that some objects are categorical adjoined with entities they do not look like. These objects are part of poor ‘unnatural' categories (Bloom, 2001; 1101-1102). On one hand, this arguments is therefore rather convincing, as it demonstrates that membership is sometimes determined by deeper properties than shape, as similarity is not decisive. Yet, on the other hand, it is worth noticing that Bloom does not explicitly state how these ‘unnatural' categories occur. Instead, he presumes that it depends on deeper properties of the objects. Obviously, it is difficult to determine, because the target group consists of young children, who presumably are not able to explain how these ‘unnatural' arise and what they build on. However, altogether the argument still appears convincing, as several scholars have made this contention, and even though it is not clear-cut how ‘unnatural' categories arise, it is more than likely that deeper properties of the objects are decisive when shape is not, because it is difficult to imagine what else would be.
Altogether, the essentialist account has many convincing arguments. However, it is worth noticing that some of them appear to be somewhat ambiguous.
4.3 Recapitulation
The discussion has shown that the attentional learning account is very convincing, as several studies confirm that shape plays a crucial role when children learn the meaning of words. Yet it worth noticing, that other studies show another tendency, which could indicate that scholars on favor of the attentional learning account ought to age-classify their argument.
Moreover, the discussion has shown that the essentialist account is also rather convincing, as several scholars have found that deeper properties of objects matters when a child has to learn the meaning of new words. Some of the supporting arguments are, however, ambiguous, but it does not mean that the account is not altogether convincing.
5 Conclusion
The aim of this paper was to determine how children learn the meanings of words. This was done by conducting a comparative analysis by contrasting two approaches: the attentional learning account and the essentialist account. The analysis showed that both of the approaches had many supporting arguments, and later the discussion also proved that both were rather convincing.
According to the attentional learning account children learn the meaning of new words by referring to the shape of new objects. Studies have shown that this is indeed the case. It is, however, worth noticing that there seems to be an age-bias, because other studies have shown that shape is primarily determining amongst children under the age of two. This means that the attentional learning account is convincing, but with reservation because it does not seem to apply to children of all ages. According to the essentialist account other things than shape also matter when children categorize new objects. Children also consider deeper properties of the object in question. Independently studies have supported this notion by proving that shape is not also pivotal. However, this account also does not exclude the possibility that shape does matter to some extent as it is a cue to object kind.
On this basis it is plausible to conclude that in general children learn the meanings of words on the basis of both shape and deeper properties of objects. Reliable studies have supported both of the approaches, and it is therefore difficult to rule out either of the two. Therefore it could be advantageous and exciting to combine the two approaches. They would complement each other, because this would decrease the number of outliers e.g. the fact that age matters, and the fact that shape appears to be of incontestable significance.

It is however important to mention that this inductive analysis do not eliminate the possibility that there might be another approach out there with better and more convincing arguments. Still, it is obvious that both perceptual- and conceptual information cannot be overlooked altogether when explaining how children learn the meaning of words.  
6 Bibliography
1. Bloom, Paul; Markson, Lori (1998): “Capacities underlying word learning”, Trends in Cognitive Sciences, Vol. 2, No. 2, February 1998.
2. Bloom, Paul (2001): “Précis of How Children Learn the Meanings of Words”, Behavioral and Brain Sciences, Vol. 24, pp. 1095-1103.
3. Booth, E., Amy; Waxman, R., Sandra (2005): “Conceptual Information Permeates Word Learning in Infancy”, Developmental Psychology. Vol. 41, No. 2, pp. 491-505.
4. Booth, E., Amy; Waxman, R., Sandra (2008): “Taking stock as theories of word learning take shape”, Developmental Science. Vol. 11, No. 2, pp. 185-194.
5. Diesendruck, Gil; Bloom, Paul (2003): “How Specific is the Shape Bias?”, Child Development, Vol. 74, No. 1, pp. 168-178.
6. Gentner, Dedre (1978): “A Study of Early Word Meaning using Artificial objects: What looks like a Jiggy but acts like a Zimbo?”, Papers and Reports on Child Language Development. No. 15, August 1978.
7. Markson, Lori; Diesendruck, Gil; Bloom, Paul (2008): “The shape of though”, Developmental Science. Vol. 11, No. 2, pp. 204-208.
8. Pinker, Steven (1994): “The language Instinct”, Penguin Books
9. Smith, B., Linda; Jones, S., Susan; Landau, Barbara; Gershkoff-Stowe, Lisa; Samuelson, Larissa (2002): “OBJECT NAME LEARNING PROVIDES ON-THE-JOB TRAINING FOR ATTENTION”, Psychological Science, Vol. 13, No. 1, pp. 13-19.
10. Waxman, R., Sandra; Booth, E., Amy (2000): “Principles that are invoked in the acquisition of words, but not facts”, Cognition, Vol. 77, pp. 33-43.