Updating mental representations, 1 RUNNING HEAD: UPDATING MENTAL REPRESENTATIONS Psychological Science (in press) Thinking of Things Unseen: Infants’ Use of Language to Update Mental Representations

One of the most distinctive characteristics of humans is our capacity to learn from what other people tell us. Often new information is provided about an entity that is not present, requiring incorporation of that information into one’s mental representation of the absent object. Here we present evidence regarding the emergence of this vital ability. Nineteenand 22-month-old infants first learned a name for a toy and later were told that the toy had undergone a change in state (it had become wet) while out of view. The 22month-olds (but not the 19-month-olds) subsequently identified the toy based solely on the property that they were told about but had never seen. Thus, before the end of their second year, infants can use verbally communicated information to update their representation of an absent object. This developmental advance inaugurates a uniquely human and immensely powerful form of learning about the world. Updating mental representations, 3 Thinking of Things Unseen: Infants’ Use of Language to Update Mental Representations Human knowledge is based on a combination of direct and indirect experience. A large proportion of our knowledge comes indirectly, much of it from another person telling us something new-what Paul Harris (2002) refers to as testimony. Sometimes, testimony informs us about objects or events that are present and directly perceivable. (“Look at that dog. Its paws are all muddy!”) Often, however, the subject of the information we receive is absent at the time we hear it. (“Matt’s mom gave him his first haircut, and it looks terrible.”) The information about Matt’s new hair style can enter our knowledge base, because our mental representation of Matt is activated and then updated by this communication. The ability to learn through indirect experience rests on our capacity to understand references to absent entities and events (Hockett, 1960). Although nonhuman animals may share this ability to limited degrees (e.g., monkeys may be informed of the presence of unseen predators by other animals’ communicative calls; Hauser, 1996), most cases of updating in humans are made possible by language. Language is an immensely flexible system that liberates us from the here and now. It allows us to acquire new information about any entity to which we can refer by a word or expression, and to update our knowledge of that entity in the absence of any direct encounters with it. In this research, we investigate the emergence of this ability in very young children. Infants begin to talk about absent objects in the second half of their second year (Lewis, 1936; Sachs, 1983; Scollon, 1979; Shimpi & Huttenlocher, 2004; Veneziano & Updating mental representations, 4 Sinclair, 1995), but they understand someone else’s reference to an absent object substantially earlier. Naturalistic observations in the home have documented comprehension of references to absent familiar objects as early as 13 months (Huttenlocher, 1974; Lewis, 1936). For example, an infant who hears a favorite toy referred to might go search for it, indicating that hearing the name of the toy had activated the child’s mental representation of it. Recent laboratory studies have confirmed that references to absent objects are comprehended early in the second year of life, and they have provided additional information about the emergence of this important ability (Ganea, 2005; Saylor, 2004; Saylor & Baldwin, 2004). For example, Ganea (2005) reported that 13and 14-montholds can understand references to a relatively novel absent object. A group of infants first learned a proper name for a stuffed animal, which was then removed from view. When the infants subsequently heard the name of the toy, they spontaneously attempted to reestablish contact with it, providing evidence that this verbal reference had brought the toy to mind. This research also revealed that infants’ response to hearing someone refer to an absent object is initially fragile. The infants were less likely to search for the toy when it was less accessible, so slightly more effort was required to regain contact with it, or when its name was pronounced after a delay. Thus, for children as young as 13 months, hearing the name of an absent object can bring it to mind. Although infants’ response to the mention of an absent object is initially fragile and context dependent, their mental representation of the object evidently can be activated by what someone else says. Updating mental representations, 5 This ability sets the stage for the development of the crucial ability to acquire new information about absent entities. If infants are able to activate a representation of an unseen object on hearing its name, then they may be able to update that representation when they hear of a new event that has befallen the object or a new property that it has gained. The present research tests for this latter ability. It is the first we know of to examine the emergence of the ability to learn something new from everyday conversations about absent objects and unseen events. To investigate this ability, we first taught 19and 22-month-olds a proper name for a stuffed animal. Then—with the toy out of sight in another room—the infants heard that the toy had undergone a change in state. The question was whether they would incorporate this information into their mental representation of the toy. When asked for the animal by name, would the children select from an array of toys the one that was identical to the original but whose appearance was inconsistent with the new information, or the one that looked different from the original in a way that was consistent with the new information? Study 1 In the first study, infants were shown 3 stuffed animals—two of them identical (e.g., 2 green frogs and 1 pink pig). So that we could later refer to one of the toys in its absence, the infants were taught a proper name –“Lucy”– for the target animal (one of the two identical toys). Then, the children went into another room, leaving the toys behind. While they were in the other room (with the toys out of sight), a different adult informed them that something had happened to the target toy—that she had “spilled water on Updating mental representations, 6 Lucy.” Finally, the children returned to the first room and were asked to identify which of the 3 toys was Lucy. Of the 3 toys displayed, two were wet—the target (the original Lucy) and the distractor toy. The toy identical to Lucy (the non-target toy) was dry. Thus, if the children were simply attracted to wetness, they should choose randomly between the two wet toys. If they remembered the kind of toy that Lucy was, but had not updated their representation of Lucy, they should either choose the dry toy (because it was more similar to what they had originally seen) or they should choose randomly between the target and non-target toys. Finally, if the children had updated their representation of Lucy based on what they had heard, they should choose the wet target toy, ignoring both the identical but dry toy and the wet distractor. Method Participants The participants were 40 children, with 20 in each of two groups: 19-month-olds (range: 17.6 to 20.6, M = 18.9, 12 girls, 8 boys) and 22-month-olds (range: 21.2. to 24.0, M = 22.3, 11 girls, 9 boys). Sixteen additional infants were excluded for being fussy, 2 for experimenter error, and 2 for parental interference. Children for all studies reported here were recruited through a database of published birth records. The majority of participants were white and middle-class. Materials Materials included 2 sets of 3 stuffed animals (2 identical green frogs and 1 pink pig or 2 identical pink pigs and 1 green frog), a bucket with water, and 3 aluminum trays. Half the children at each age were randomly assigned to one animal set. Updating mental representations, 7 Procedure The procedure included 3 phases – familiarization, attribution of new information, and test. Familiarization. The purpose of this phase was to teach the child a proper name (Lucy) for one of the 3 stuffed animals, as well as to provide the child with equal experience with the target and the distractor toy. The named animal – the target – was always one of the 2 identical toys. The experimenter first showed the child a basket containing the 3 animals and then took out the target animal (e.g., one of the frogs) and labeled it with a proper name—“Lucy.” Then she showed the child the non-target animal, the identical toy from the pair (e.g., the other frog), saying that it was “Lucy’s friend” who would sit on a cabinet to watch them play. Then the experimenter talked about the distractor (e.g., pig) without naming it (“Look at the piggy! It’s a nice piggy.”). The experimenter engaged the infant in various play activities with the 2 animals (e.g., talking about their body parts, playing peek-a-boo), spending equal amounts of time with each. Thus, the child played equally with the target (“Lucy the frog”) and the distractor (pig), while the nontarget (frog) was visible but inaccessible. To ensure that the child had learned the name for the target animal, the child had to identify it twice during the familiarization. After the first play activity (e.g., peek-aboo) in which the child had interacted with the 2 animals, the experimenter placed the target and distractor together and asked the child to get “Lucy.” If the child chose correctly, the experimenter continued with another game and then later asked the child to identify Lucy again. If the child correctly answered these 2 questions, the experimenter Updating mental representations, 8 went on to the next phase. If the child responded incorrectly to either question, the experimenter told the child which toy was Lucy. After another play activity, the identification question was repeated. The children had to correctly identify the target toy twice before going on to the next phase. The children were corrected from 0 to 4 times, with an average of 1.45 (SD = 1.54) for the 19-month-olds and .84 (SD = 1.26) for the 22-month-olds. After the child had correctly answered 2 comprehension questions, the experimenter said that the animals were tired, and she put all 3 animals in the basket. Then she and the child went into the adjoining room to read a story. The average length of the familiarization phase was 8 minutes. Attribution of new information. The purpose of this phase was to provide the child with new information about the target toy that he or she had played with earlier. As the child and the experimenter were looking at the book, an assistant came into the room carrying a bucket of water. The assistant announced that she was going to wash the table in the room next door (where the toys were). To make sure the child was aware of the water, she invited the child to touch the water in the bucket. Then she repeated that she was going to wash the table and went into the other room, closing the connecting door. After approximately 2 minutes, the assistant returned and announced with a distressed expression: “I’m so sorry! I was washing the table, and I spilled water all over Lucy. Lucy is wet now! She’s covered with water.” The experimenter then told the child: “Oh no! Did you hear that? Lucy got wet! She’s all covered in water.” Then the experimenter invited the child to go see the animal: “Do you want to go see Lucy? Let’s go see Lucy!” Updating mental representations, 9 Test phase. The test phase was designed to see whether the children would identify the target animal as the one that had undergone the change they had heard about. Figure 1 depicts what the child saw when entering the room: a small table with the 3 animals sitting on trays. Two of the animals – the target and the distractor – were wet, whereas the non-target was dry. The distractor animal was wet to control for the possibility that a child might choose the wet target animal just because it was more interesting than a dry one. The dry non-target animal was always in the middle, with the wet target animal on the left for half the children and on the right for the others. The experimenter asked the child to show the target animal (“Which one is Lucy? Show me Lucy!”). Coding Films of the test sessions were coded to identify intentional choices of one of the animals. We counted as correct only responses that were judged to be intentional acts of indicating a test item – deliberately pointing to a toy or picking up a toy and showing or giving it to the experimenter. If a child simply touched or explored an animal without intentionally indicating it to the experimenter, then the experimenter repeated the question. On the few occasions that a child chose 2 animals, the experimenter asked the child to indicate which one was “Lucy.” The coding was the same for all three studies. One person coded all the tapes, and another coded 50% of the tapes for each study. The level of agreement between the coders for the three studies ranged from 89% to 100% (Cohen's kappa = .77 to 1, p’s < .05). The disagreements were easily resolved by a third coder. Results and Discussion Updating mental representations, 10 When presented with 3 choices—the wet target, the identical but dry non-target, and the wet distractor—all but one child selected either the altered target or the unaltered non-target. Thus, both age groups had learned the name “Lucy” for the target object— one of the two identical animals. The important question was whether they realized that Lucy was the wet one of the two identical animals. That is, had they updated their mental representation of the Lucy toy to include the fact that water had been spilled on it? To answer this question, we compared their performance against a chance level of 50%. Although there were 3 animals for the children to choose from, which would make chance 33%, the fact that only 2 children ever chose the distractor meant that the children knew which category of animal Lucy was. Hence, chance performance – failure to update their representation of Lucy – would involve choosing randomly between the 2 identical toys. The more conservative – and appropriate – chance level was therefore 50% (rather than 33%). As shown in Figure 2, most (16 of 20) of the 22-month-olds selected the correct toy when asked to show Lucy. This level of performance was significantly above the chance level of 50%, χ (1, 20) = 7.2, prep = .99, p < .05. The rest selected the unaltered non-target (3) or the other wet item (1). In contrast, fewer than half (9 of 20) of the 19month-olds chose the target object (a level of performance clearly not above chance). The rest selected the unaltered non-target (10) or the other wet toy (1). The number of children selecting the target toy in the 2 age groups differed significantly, χ (1, 40) = 5.22, prep = .98, Φ = .36, p < .05. These results indicate that 22-month-old children can generally use new information they hear about an absent object to subsequently distinguish that object from Updating mental representations, 11 an identical one. Thus, by 22-months of age, most children are capable of updating their representation of an object to incorporate new information about it. The chance level performance of the 19-month-olds indicates that the ability to update representations of objects based on information communicated in their absence may emerge between 19 and 22 months of age. However, it is also possible that the 3choice test procedure used in Study 1 was cognitively or linguistically taxing for such young children and may have masked some fragile ability that might be evident with a simpler test. Accordingly, in Study 2, we tested a new group of 19-month-olds, using a simpler, 2-alternative test in which they had to choose between 2 of the same type of toy-the wet target and the dry non-target.