Mutual exclusivity bias and word learning in children

Advocates of the role of lexical biases in word-learning tend to emphasise the learnability problem posed by the arbitrary relationship between words and their referents. This is commonly illustrated using Quine’s (1960) ‘Gavagai’ paradox, in which a linguist in a foreign country sees a native speaker point to a white rabbit running past, and hears them shout ‘gavagai!’ The linguist is then faced with the task of ascertaining precisely what the speaker is referring to. ‘Rabbit!’, ‘Running!’ and ‘Look, dinner!’ are just three of the infinite possibilities. This is known as the problem of referential indeterminacy. As much of children’s early linguistic input is comprised of ostensive definition, it is claimed that this is precisely the problem that wordlearners face as they acquire their native lexicon (Markman, 1987). In light of this, researchers have proposed a number of lexical biases which constrain the hypothesis space and reduce the potential referents of a novel label. For example, it is suggested that Mutual Exclusivity (ME hereafter) ‘leads children to prefer that each object have only one category label’ (Markman, Wasow and Hansen, 2003:242). This bias is believed to underlie the so-called disambiguation effect (Merriman and Bowman, 1989), whereby children initially apply novel labels to novel objects (see Davidson, Jergovic, Imami and Theodos, 1997; Merriman and Stevenson, 1997; Au and Glusman, 1990; Markman and Wachtel, 1988). ME therefore reduces the number of hypotheses a child is required to entertain by biasing them against assigning a novel label to an already labelled object.

Lexical biases are understood as representing ‘violable default assumptions’ (Jaswal and Hansen, 2006:163), which give children ‘good first guesses about the meaning of a novel term’ (Markman, 1992:69-70). As such, this account predicts that word-learners should initially behave in accordance with lexical biases, and, in principle, should thus subordinate socio-cognitive cues to lexical cues. Specifically, if early lexical acquisition is constrained by lexical biases, then children should honour ME and assume that novel labels should be assigned to novel objects, even when a socio-cognitive cue indicates that the intended referent of a novel label is, in fact, a familiar object. The lexical bias account therefore predicts that children will be ‘unwilling’ to violate ME in this context.

Another bias, mutual-exclusivity, is overridden very early on (Markman, 1990). This bias decrees that there should be one, and only one, label for each object (see Table 6.3). But at the age of two years, my son Alex called our pet both by its name (Michelangelo) as well as by the basic-level term, cat. (Incidentally, Michelangelo was named after a ninja turtle, not a Renaissance artist.) Mutual-exclusivity can be overridden, but, to complicate matters it does not entirely disappear. Children can be seen adhering to mutual-exclusivity years later, at the age of six (Davidson & Tell, 2005). The example of different names for our pet could also be seen as a case of overriding the basic-level assumption (see Box 6.1). While cat is the basic-level term, the child also acquires both superordinate (animal) and subordinate terms (Michelangelo). As learning progresses, many other terms could be recruited for our cat, including feline, pet, mammal and moggy. In a similar vein, if we turn to the taxonomic assumption, we find that by the age of four years, children are able to shift from taxonomic to thematic assignments for unfamiliar words; depending on the choices they are offered (Ralli & Dockrell, 2008). (Saxton, 2010, 154)

“Words are mutually exclusive…. [E]ach object will have only one label” (Markman, 1994, 163)

Mutual-exclusivity is the assumption that there is a “one-to-one” relationship between objects and words. One word refers to one object and likewise one object has only one label; for example a “cat” cannot also be a “dog”. This assumption helps young children learn to develop an association between words and whole objects. It is when children start to develop a more substantial vocabulary that they begin to realise two seemingly different objects can be labelled with the same categorical word, for example, “pet” can refer to both “cat” and “dog”. (Guasti, M, 2004)

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!
Find out more about our Essay Writing Service

Mutual-exclusivity is used by young children when they required only to learn ‘basic-level category labels’ (Imai, M., & Haryu, E., 2004). It has been suggested that mutual-exclusivity can help foster word learning in young children. Where a child is offered a new word to associate with an object that they have already attach a label to, the child is forced to understand the new word as a label to a separate part of the object. For example ‘tail’ said when pointing at a cats tail when the child has already accepted “cat” as the label for whole object may force the child into accepting the ‘tail’ as a separate part of the object and therefore having its own name. Markman and Wachtel (1988) suggest that this is significant in the child’s understanding of substance names such as ‘plastic’ or ‘wood’.

Further investigation into the power of the mutual-exclusivity bias is done by Haryu (1991) and Haryu and Imai (1999) who did research with three- and five-year-old Japanese children. The children were allocated to one of two groups; either the ‘word-only’ condition or the ‘word and pragmatic context’ condition. Similar to the Markman et al. (1988) research, the children were offered two objects, one familiar (e.g. apple) and one unfamiliar (e.g. lipstick holder). The difference in conditions were that the ‘word and pragmatic context’ children were presented with the novel word in a context predispose the novel word to refer to the familiar item. An example is given by Imai, M., and Haryu, E. (2004)

“The experimenter introduced a puppet named Mary, and said to the child, ‘Mary is hungry now. I would like to give Mary (the) heku,” where heku was a nonsense word. Then the experimenter placed two objects, a familiar object (e.g., an apple) and an unfamiliar object (e.g., a lipstick holder), in front of the child, and asked her to select the heku.” (pg424)

In the child’s mind there is a contradiction between the pragmatic context and the mutual-exclusivity bias. The pragmatic context suggests that ‘Mary’ wants the known object (the apple to eat), however if the child is influenced by the mutual-exclusivity bias then the known object has already been labelled ‘apple’ and therefore cannot be also labelled ‘heku’. Haryu and Imai (1999) found that how much a child was influenced by the mutual-exclusivity bias was determined by their age. The three-year-olds were still massively led by the mutual-exclusivity whereas the five-year-olds were more influenced by the pragmatic context.

In this two studies by Markman and Wachtel (1988) and Haryu and Imai (1999), at a certain age the mutual-exclusivity bias overrides any other syntactical or contextual cue. Some studies go one step further to suggest that this bias can be countermanded at a certain age by more complex and subtle cues (Merriman et al., 1989). There has been a lot of research done to determine at what age the bias can be overruled (Clark and Svaib, 1997).

Merriman (1989) states that the mutual-exclusivity bias may influence word learning in young children in four ways.

Firstly, a child’s decision about the reference of a new word could be affected. This can be seen in Markman and Wachtel’s (1988) study. Markman and Wachtel (1988) showed evidence for the mutual-exclusivity bias through their ‘Dax’ experiment. Two objects were presented to 3yr old children, one familiar to them (e.g. cup) and the other unfamiliar. The children were then asked to identify the ‘dax’. Due to the mutual-exclusivity rule the child would assume the unfamiliar object was the ‘dax’ because the familiar object already had a label (i.e. it was already known to them as ‘cup’, and therefore by process of elimination the unfamiliar object must be the ‘dax’). Using the novel word ‘dax’ hopes to eradicate the problem of previous knowledge, so it is a label that the child will be unlikely to have labelled to any other object. The experiment was extended to observe whether changing the syntax may affect the child’s choice. When the child was asked to distinguish the property of a familiar object (e.g. “show me pewter”, Imai, M., & Haryu, E., 2004). ‘Pewter’ still being an unfamiliar word to the child meant that despite the change in syntax that would give the child a clue that ‘pewter’ is not a noun, the children still obeyed the mutual-exclusivity rule, labelling the unfamiliar object with the unfamiliar word. Markman and Wachtel (1988) suggest that the child decides upon the unfamiliar object in spite of the contradict evidence given by the change in syntax because “selecting the object whose name was already known as the referent of the novel word [i.e. ‘dax’ and ‘pewter’] would violate the mutual-exclusivity bias.” (Imai, M. and Haryu, E., 2004, 423). Merriman (1989) calls this the ‘disambiguation effect’.

The second way the bias may influence a child’s word learning is through the ‘correction effect’ (Merriman, 1989). In this way, mutual-exclusivity is also helpful for aiding the self-correction of overextension; when a child generalises one word to many similar objects. An example of this is when a child visits the zoo and calls a tiger a ‘cat’. In one way the child is correctly signifying that the tiger belongs to the cat family, however in their own situational knowledge, a ‘cat’ is a pet. When the child is corrected by being told that the animal is called a ‘tiger’ the child due to mutual-exclusivity has to reject one of the two labels offered either ‘cat’ or ‘tiger’. Therefore the overextension of the category ‘cat’ becomes smaller with each correction.

The ‘rejection effect’ is Merriman’s (1989) third influence of the mutual-exclusivity bias. This being where the child is told a new name for a familiar object, for example a ‘coat’ can also be called a ‘jacket’. If the child is told to ‘get their jacket’ they may respond by rejecting the new word and correcting the person; “I have got my coat”. This rejection can be seen to be an effect of the mutual-exclusivity bias. The child rejects the new terminology because the ‘coat’ can only have one label and therefore a ‘jacket’ must not be correct and must refer to something different. This can be considered an obstruction to word learning in child development as the child becomes very constrained with their vocabulary.

Finally, the mutual-exclusivity bias may influence word generalisation in children’s language development. When children learn a new word in reference to an object, the mutual-exclusivity bias limits them to this word only, not allowing them to use their own understanding of the object to make associations with similar words to use interchangeably. An example of this is an object which is a cross between a ‘cup’ and a ‘glass’, the object could easily be referred to as either but due to the mutual-exclusivity bias the child is restricted to only associating one word with the object. So the child will understand the object only as a ‘cup’ OR a ‘glass’ but not both. Merriman (1989) therefore describes this as the ‘restriction effect’.

Merriman’s (1989) theory suggests that if a child is presented with a new word, in particular a name of an object, one of three things can happen if they are obeying the mutual-exclusivity bias. Firstly they could ‘disambiguate’ it; so attach the word to an unfamiliar, unnamed object, or they could ‘self-correct’. Where the child has attached a label to an object they may choose to reject the old label and replace it with the new name. Otherwise the third instance would occur where they would have to reject the new word completely. In order to conform to the mutual-exclusivity bias the must chose one of these reactions and the restriction effect must occur simultaneously (Merriman, 1989). For example, if a child understands the word ‘can’ (as in can of beans) but the word ‘tin’ is presented to them as an alternative name, they must choose to accept or reject ‘tin’ as the new label for the object and at the same time believe that the object cannot be both a ‘can’ and a ‘tin’ (therefore applying the restriction effect).

The first two ways that mutual-exclusivity may influence word learning in children, specifically, the disambiguation effect and the correction effect, although they are not always

The mutual exclusivity bias in children’s word learning leads them to assume that new words do not refer to any object for which they already have a label . The mutual exclusivity bias also helps to explain how children can learn so many words at such a fast pace. Previous research has established the average age that children are able to use mutual exclusivity to increase language knowledge. Researchers have identified how children respond to both unfamiliar and familiar objects as well as unfamiliar and familiar words and how they connect the objects and words. Mutual exclusivity enables children to infer the meaning of unknown words with no information other than a photograph of an unknown object. Although the mutual exclusivity bias may sometimes lead to incorrect assumptions about word meaning, mutual exclusivity is beneficial to children because research has shown that children who are proficient in it have much larger vocabularies. Additionally, research has identified bilingual children as being more capable at using mutual exclusivity and thus more advanced in language development than monolingual children. This knowledge can be beneficial to parents and teachers to enable them to encourage language development to produce children with significant vocabularies. Also, this research can help explain to parents why their children may incorrectly identify novel objects. Possible further research should be done on children with special needs to help identify whether mutual exclusivity aids their language development. Additional research would be helpful to identify how children override mutual exclusivity and add hierarchies into their understanding of language.

Au & Glusman (1990),

for example, reported that children will accept multiple labels for a referent if they

believe the labels to be from different hierarchical levels, but will honor ME when

they believe two labels to be at the same hierarchical level.

Liittschwager & Markman (1994) argued strongly that an assumption of ME is a

bias that even very young children bring to the task of word learning. They emphasized

that ME is best characterized as a default assumption rather than an inviolable

principle – a bias that guides children’s initial assumptions about word meaning but

is flexible enough to be ignored in the face of other evidence. By this view, ME does

not prevent children from learning multiple labels for an object, it just makes it

harder for them to do so. In order to find support for this position, Liittschwager and

Markman attempted to show that 2-year-olds would be less successful at learning a

label for an object if they already knew a label for that object than if they did not.

In other words, the researchers hypothesized that learning a second label for

an object would be harder for children than learning a first label for the object. This

prediction was based on the idea that learning a second label would require children

to override the default ME bias, but learning a first label would not.

In fact, contrary to prediction, Liittschwager & Markman (1994) found that 2-yearolds

learned first and second labels for an object with equal facility. In other words, the

children showed no evidence of an ME bias at all. Speculating that the learning of a

single new word placed relatively light processing demands on the children, which left

them free to override the ME bias and accept two labels for the target object,

Liittschwager and Markman increased the demands by teaching the children two new

words for two different objects simultaneously in a subsequent study. The manipulation

was the same as in the previous study: half the children learned new labels for

familiar objects (for which they already knew a name), and half learned labels for novel

objects. In this case, there was a clear effect of condition, such that those children in

the first label condition successfully learned the labels for both objects, but those in the

second label condition showed no evidence of learning any new words.

The interpretation put forth by Liittschwager & Markman (1994) was that the

children were predisposed to assume mutual exclusivity, but the social demands of the

task caused them to override ME in the first study and accept two different labels for

the same object. However, they were unable to do the same in the second study,

because the increased processing demands prevented them from overriding ME.

Thus, they fell back on their default strategy, which was to accept only one label per

object.