PERSPECTIVES BEHAVIOR knows more than a 9-year-old , after all, and has a better memory, and a Can a Dog Learn a Word? better ability to understand the minds of adults. Rico’s limitations might reflect dif- Paul Bloom ferences in degree, not in kind. A more skeptical alternative is that hen making the emergence of a vibrant area Rico’s abilities have nothing to do with hu- point that of comparative cognition man word learning. For a child, words are Wlearning requires research. For psycholo- symbols that refer to categories and indi- more than just the right envi- gists, may be the new viduals in the external world (7). Even one- ronment, psychologists often . year-olds appreciate the referential nature point out that both a baby How does Rico’s learning of words. When children learn a word such and a dog are exposed to lan- compare with that of a as “sock,” they do not interpret it as “bring- guage, but only the baby child? Kaminski et al. point the-sock” or “go-to-the-sock,” and they do learns to talk. This example out just two differences: not merely associate it with socks. They may have to change. On Children have a more di- appreciate that the word refers to a catego- page 1682 of this issue, verse vocabulary, including ry, and thereby can be used to request a Kaminski et al. (1) report names for specific people, sock, or point out a sock, or comment on the impressive abilities of a properties, actions, and rela- the absence of one. named Rico, tions; Rico just knows words Does Rico understand reference? It is who might well be capable for fetchable things, mostly not clear (see the figure). In the experi- of learning words. toys and balls. And children can speak; ments, Rico’s abilities are limited to specif- When the experimenters place 10 items Rico cannot. ic routines. All new items are learned in the in another room, and Rico is asked by his But that is not all. Rico is 9 years old course of fetching, and Rico’s understand- owner to retrieve one, he is usually accu- and knows about rate, and repeated trials suggest that he has 200 words, whereas a vocabulary of more than 200 words. a human 9-year-old What is more impressive is that he can known tens of thou- learn in just a single trial, akin to the “fast sands of words, and Sock mapping” abilities of young children is learning more than (2–4). Kaminski et al. tested the collie’s 10 new words a day fast mapping abilities by placing a new ob- (2). Children’s word ject along with seven familiar ones and learning is highly ro- having the owner ask Rico to fetch, using a bust; they can learn …or… name Rico had never heard before. He usu- words from overheard ally retrieved the new item, apparently ap- speech, even if no- preciating, as young children do, that new body is trying to words tend to refer to objects that do not al- teach them (2). Rico, ready have names. A month later, Rico in contrast, learns Rico, where’s the sock? showed some retention of the name he had only though a specif- learned. His abilities are comparable to ic fetching game. those of children and adults who were test- Children can under- ed using similar designs (3–4). stand words used in a Dog owners often boast about the range of contexts; communicative and social abilities of Rico’s understanding their pets, and this study seems to vindi- is manifested in his cate them. Indeed, Rico’s word-learning fetching behavior. To

SCIENCE abilities surpass those of nonhuman pri- rephrase a remark mates such as chimpanzees, who have made by the psychol- never demonstrated this sort of fast map- ogist Lila Gleitman ping (5). As the researchers note, this (in reference to might be because dogs have been specifi- language): If any cally selected for attending to the com- child learned words Different ways to interpret a command. When Rico, a border collie, is municative intentions of people. More the way Rico did, the requested by his owner to fetch a sock, he might understand her in the generally, this study fits nicely with other parents would run same way a child would. That is, Rico might appreciate that the word research on the social capacities of do- screaming to the “sock” refers to a category of objects in the world and that the rest of mestic dogs (6), and might signal the nearest neurologist. the command means that he should act in a particular way (fetching) Perhaps Rico is toward a member of that category. Alternatively, he might not under- The author is in the Department of Psychology, Yale doing precisely what stand reference at all and might be limited to associating the word spo- University, Post Office Box 208205, New Haven, CT a child does, just not ken by his owner with a specific behavior such as approaching a sock or

CREDITS: BAUS-KRZESLOWSKI; (TOP) HUEY/ PRESTON (BOTTOM) 06520–8205, USA. E-mail: [email protected] as well. A 2-year-old fetching a sock.

www.sciencemag.org SCIENCE VOL 304 11 JUNE 2004 1605 P ERSPECTIVES ing of these items is tested in this context as word by being shown an object and hear- on the view that babies learn words and well. Also, it is always Rico’s owner who is ing a person name it? Can he learn a word dogs do not. communicating with him. These experi- for something other than a small fetchable ments are carefully designed, and so there object? Can he display knowledge of a References is no worry about problems akin to those of word in some way other than fetching? 1. J. Kaminski, J. Call, J. Fischer, Science 304, 1682 (2004). Clever Hans (a horse that seemed to have (Kaminski et al. note that there is anec- 2. P. Bloom, How Children Learn the Meanings of Words mastered arithmetic but was actually re- dotal evidence that he can—this issue is (MIT Press, Cambridge, MA, 2000). sponding to subtle cues by its owner). Yet, worth pursing experimentally.) Can Rico 3. L. Markson, P. Bloom, Nature 385, 813 (1997). 4. S. Carey, E. Bartlett, Child Lang. Dev. 15, 29 (1978). if Rico really is learning sound-meaning follow an instruction not to fetch an item, 5. M. S. Seidenberg, L. A. Petitto. J. Exp. Psych. Gen. 116, relations, as Kaminski et al. maintain, it just as one can tell a child not to touch 279 (1987). should not matter who the speaker is. something? Rico’s abilities are fascinat- 6. B. Hare, M. Brown, C. Williamson, M. Tomasello, Science 298, 1634 (2002). Further experiments can help to re- ing, but until we have answers to these 7. J. Macnamara, Names for Things (MIT Press, solve this issue. Can Rico learn a new sorts of questions, it is too early to give up Cambridge, MA, 1982).

MATERIALS SCIENCE of their spontaneous electric polarization, they can be used as binary data storage Fundamental Size Limits media in which opposite directions of po- larization represent the 1 or 0 data bits. In addition, because the electric polarization in Ferroelectricity is coupled to the structure of the materi- al, ferroelectrics can convert mechanical Nicola A. Spaldin energy to electrical energy and vice ver- sa. This leads to their widespread use in ith the continued demand for persists down to vanishingly small sizes. transducer applications such as piezo- portability in consumer elec- Ferroelectrics find three main techno- electric actuators and sonar detectors. Wtronics, it is becoming increas- logical niches based on three related Finally, they have very large dielectric ingly important to understand the effects physical characteristics. First, as a result permittivities leading to applications in of miniaturization on the properties of the capacitors. In all cases, it is crucial active components in electronic devices. to understand the size dependence In many cases, however, the basic physics AB of the ferroelectric behavior as ever of such size reduction is poorly under- smaller devices are produced. stood and can be difficult to characterize, Consider why the existence of a because competing effects such as surface critical size might be expected intu- properties, strain effects from substrates, itively. Most technologically impor- and fundamental size quantization com- tant ferroelectrics are perovskite- plicate the behavior. This is particularly Depolarizing field structure oxides (see the figure, pan- true in the case of ferroelectrics—materi- als that have a spontaneous electric polar- ization that can be switched by an applied C electric field. Indeed, it has long been be- lieved, on the basis of empirical evidence, that there is a critical size on the order of hundreds of angstroms below which a spontaneous electric polarization cannot be sustained in a material [for reviews of the literature, see (1)]. Such behavior would render ferroelectrics useless for ap- plications at sizes below this cutoff, there- by limiting their importance in future technologies. Recent first-principles theo- retical work (2–4), however, has indicated that the critical size is orders of magni- tude smaller than previously thought, and this view has been corroborated by some measurements (5). On page 1650 of this issue, Fong et al. (6) provide the first un- Domain boundary ambiguous experimental evidence that these theoretical predictions and recent Persistent polarization. (A) Schematic of the ideal cubic perovskite structure. A small cation (red) experimental indications are indeed cor- sits at the center of an octahedron of oxygen anions (white) with large cations at the corners of the rect by confirming that ferroelectricity unit cell. (B) In the distorted ferroelectric structure, the cation cage displaces (in this case down) rel- ative to the anion cage, creating an electric polarization and a consequent depolarizing field. (C). In The author is in the Materials Department, University the samples of Fong et al.(6), domains of oppositely oriented polarization form, so that the net de- of California, Santa Barbara, CA 93106–5050, USA. polarizing field is zero.The figure shows a schematic of two domains, separated by a domain bound- E-mail: [email protected] ary across which the polarization changes orientation by 180°.

1606 11 JUNE 2004 VOL 304 SCIENCE www.sciencemag.org