Date: Thu, 4 Dec 2003 10:18:46 -0500 From: Katharine Beals Subject: Language In Mind: Advances in the Study of Language and Thought
Gentner, Dedre and Susan Goldin-Meadow, ed. (2003) Language in Mind: Advances in the Study of Language and Thought, MIT Press.
Katharine Beals, Ph.D., Autism Language Therapies
SUMMARY "Language in Mind", edited by Dedre Gentner and Susan Goldin-Meadow, collects 16 recent essays by premier scholars exploring the influence of language on thought. Here Whorf's hypothesis that linguistic differences may cause differences in thinking, widely embraced in the 1950s and 1960s, lately the object of so much scorn and skepticism (Pinker 1994; Devitt and Sterelny 1987), receives renewed attention. Today's better understanding of linguistic distinctions and psychological processes affords more precise hypotheses about language and cognition and more scientific methods of testing them.
In the introductory chapter, Gentner and Goldin-Meadow lay out the essays' broad thematic interconnections ("Whither Whorf"). Next, in a section entitled "Position Statements", three opinion pieces introduce some of the underlying theoretical issues.
In "Languages and Representations", Eve V. Clark recalls Slobin's (1996) distinction between "thinking for speaking" (what we do when we select and combine words) and other types of thought. Thinking for understanding, remembering, and categorizing may involve less linguistic encoding and therefore less language-specific variation. Even a given event can induce multiple mental representations, some of which aren't motivated by the intent to use language; these, too, are potentially universal.
In "Language and Mind", Stephen C. Levinson adduces the huge variation among the semantic structures of the world's languages as evidence against the "nativist" emphasis on linguistic universals and the "nativist" claim that universal, inborn concepts operate at the level of mental computation. Decomposing the semantics of all languages would produce a set of primitives larger than our notoriously short short- term memories can handle. Rather, concepts are language-specific constructs. Also language specific, given the ease and rapidity with which we encode and decode propositions, is the conceptual representation system closest to semantic representation.
In "Social Cognition", Michael Tomasello stresses the importance of joint attention. Occurring whenever one attends to what the speaker is attending to, joint attention does more than enable language learning. Since there are many ways in which a speaker may attend to something, attending to his attention teaches children multiple ways of construing objects or events, vastly enlarging their cognitive capacities. Language, Tomasello concludes, does not affect cognition, but is rather a form of it (an essentially social one).
The next section, "Language as Lens: Does the Language We Acquire Influence How We See the World?", addresses the question most akin to Whorf's original hypothesis.
In "Sex, Syntax and Semantics," Lera Boroditsky, Lauren Schmidt, and Webb Phillips compare Spanish and German students, and find significant effects of grammatical gender on memory, on object descriptions, and on judgments of picture similarity. They do so even when these tasks are as non-linguistic as possible, and when the potential effects of culture are eliminated in parallel experiments using a fictional language.
In "Speaking versus Thinking about Objects and Actions", Barbara C. Malt, Steven A. Sloman, and Silvia P. Gennari analyze the naming and sorting of artifacts and the memory and similarity judgments for events involving motion and manner by speakers of languages that categorize those artifacts and events differently. Only experiments involving explicit linguistic encoding (where subjects were asked to name the object or event) evinced language-specific differences.
Similar experiments, as well as experiments involving color and frames of reference (Brown and Levinson 1993), are reviewed by Edward Munnich and Barbara Landau in "The Effects of Spatial Language on Spatial Representation: Setting some Boundaries." Observing that not just explicit linguistic encoding, but also prior or covert linguistic encoding, can influence empirical results, they argue that none of the experiments with language-specific outcomes have ruled out covert encoding. Linguistic experience reorganizes linguistic representations, they conclude, but not nonlinguistic ones.
In "Language and Thought Online: Cognitive Consequences of Linguistic Relativity", Dan Slobin points out that the effects of linguistic experience and representations, or what he calls "thinking for speaking," are nonetheless non-obvious and nontrivial. Comparing languages whose verbs readily encode manner of motion (many Germanic languages) with those that don't (many Romance ones), he discusses the relative prevalence of manner verbs in the lexicons, conversations, and narratives of the former. Noting that listeners and readers of manner languages attend more to fine distinctions of manner of motion and form more detailed mental images of manner in reported events, he proposes that on-line linguistic experiences affect off-line memories for these events and for events recorded for later reporting.
The next section, "Language as Tool Kit: Does the Language We Acquire Augment Our Capacity for Higher-Order Representation and Reasoning?", harkens back to Vygotsky (1962) more than to Whorf. More important than cross-linguistic comparisons, here, are comparisons within a language of speakers with different linguistic skills.
In "Why We're So Smart," Dedre Gentner argues that much of human intelligence results from two species-specific faculties: analogical thinking and language. Experiments show that giving more than one object the same name invites children to extract similarities from specifics, and that naming a relationship between objects likewise helps them to extract relational concepts (e.g. "same" and "different"). Such experiences move the child beyond the immediate and concrete, bootstrapping her into ever higher levels of complexity and abstraction: "[I]f a pattern discovered by analogy is named, it becomes easier to see as part of yet another analogy." (p. 228).
In "Does Language Help Animals Think?" Stan A. Kuczaj, II and Jennifer L. Hendry observe that while animals show cognitive skills that don't require language-e.g. categorizing by perceptual properties- making similarity judgments and integrating sensory information from different modalities require symbols that abstract away from sensory details. Perhaps also facilitated by language are relational inferences about other points of view: language-enculturated dolphins have pointed at things to direct attention. Whether language actually creates new mental faculties, or merely facilitates existing ones (perhaps affording new ways to express an extant theory of mind), remains unclear. Of course, different animals are more or less capable of learning and exploiting language. Additionally, echoing Tomasello's paper, the authors remark that the most effective training involves "social interaction and clear indications of the referential nature of the symbols" as opposed to operant conditioning (p. 246).
In "What Makes Us Smart? Core Knowledge and Natural Language", Elizabeth S. Spelke argues that many core knowledge systems thought to be unique to human cognition-those relating to objects, number, geometry-are in fact surpassed by those of animals. What makes us smarter than them is our language faculty. The combinatoric properties of language allow us to connect these inherently encapsulated, domain and task-specific systems. Only those who've mastered spatial language, for example, can connect shape and color systems into novel concepts like "left of the blue wall." Similarly, language helps us relate our two systems of number awareness: the precise recognition of small numbers of distinct objects and of the effects of adding and subtracting them, and the sense of set size that for large numbers is only approximate. When we relate the counting routine ("one, two, three"...) to both systems, we construct, through language, a new knowledge system for large, exact numbers. Whether different languages construct different concepts and conceptual systems out of their universal building blocks (the core knowledge systems we all share) depends on how much they vary in their combinatoric properties.
In "Conceptual and Linguistic Factors in Inductive Projections: How do Young Children Recognize Commonalities between Animals and Plants", Kayoko Inagaki and Giyoo Hatano focus on two knowledge domains: plant and animal. Young children, who appear to lack the more general concept of "living things", treat these domains as distinct. Linguistic clues, however, can help them generalize across domains; how readily they do so depends on phrasing ("eats/drinks" versus "needs food/water").
In "Language for Thought: Coming to Understand False Beliefs", Jill G. de Villiers and Peter A. de Villiers focus on the uniquely human ability to understand and reason about false beliefs (as in "He did that because he thinks that the candle is an apple."). The acid test for the representational theory of mind, such reasoning reaches beyond external behavior into consciousness and differing points of view. Surprising evidence suggests that children, even older, language- delayed deaf children, cannot entertain false beliefs until they've acquired a specific linguistic structure: sentences with embedded complement clauses. Only children who can process such sentences pass false belief tests, even if all the specific structures they've learned contain verbs of communication rather than belief ("He said that the candle is an apple"), and even if the tests themselves avoid language. Only the embedded complement structure, acquired rather than innate, provides "the right degree of representational richness for capturing false beliefs." (p. 374).
The final section, "Language As Category Maker: Does the Language We Acquire Influence Where We Make Our Category Distinctions?", returns to cross-linguistic comparisons. It explores whether speakers of different languages draw boundaries between categories in different places, and whether concepts exist independently of language or are constructed by it.
In "Space under Construction: Language-Specific Spatial Categorization in First Language Acquisition", Melissa Bowerman and Soonja Choi argue that many spatial categories are not innate, but constructed by young children under linguistic guidance. Experiments show Korean and English-speaking children from as young as 18 months classifying spatial events more like the adult speakers of their language than like their same-age peers, and over-generalizing according to the language's semantic and statistical properties. Citing Gentner's observations about naming, they propose that children possess some innate categories (containment) and context-specific sensitivities (awareness of tightness of fit in specific situations). Hearing the same word for two different things prompts children to fill in the gap between them, partially mapping out a category; hearing different words for things prompts children to draw category boundaries somewhere between them. These effects extend into nonlinguistic cognition: nonverbal tasks show Korean-speaking adults exhibiting greater awareness than their English- speaking counterparts do of "tightness of fit", a distinction made only in Korean.
In "Reevaluating Linguistic Relativity: Language-Specific Categories and the Role of Universal Ontological Knowledge in the Construal of Individuation", Mutsumi Imai and Reiko Mazuka agree with Bowerman and Choi that language helps construct concepts from universal building blocks. Their paper compares languages that make grammatical distinctions between objects and substances (perhaps marking the former as count nouns that take plural endings and the latter as mass nouns that don't), and languages that make no such distinctions. Scrutinizing various experiments which purport to show different sensitivities to the distinction between objects and substances by speakers of different languages, they argue that people of all ages nonetheless share a fundamental awareness of the ontological differences between objects and substances. Specific differences in categorization arise gradually as children learn the norms of their culture.
In "Interaction of Language Type and Referent Type in the Development of Nonverbal Classifier Preferences", John A. Lucy and Suzanne Gaskins argue that other studies of linguistic influences on categorization, including Imai and Mazuka's, have been skewed by specific assumptions about innate, ontologically prior categories. It is both possible and more revealing, they argue, to dispense with these highly debatable questions entirely by starting with linguistic structure and deducing the characterization of reality "implicit in it." (p. 465). So long as linguistic differences (e.g. the grammatical encoding of the count-mass distinction in English vs. the indiscriminant use of classifiers in languages like Yucatec) yield corresponding differences in classification (e.g. the tendency of older English speakers to classify rigid artifacts by shape and of older Yucatec speakers to classify them by material), one can reasonably infer an impact of language on thought.
Finally, in "Thought before Language: Do We Think Ergative?", Susan Goldin-Meadow examines how people with little exposure to conventional language classify thematic roles-- specifically, whether language- delayed deaf children classify intransitive actors ("John" in "John ran home") like transitive actors or like transitive patients (respectively, "John" and "cat" in "John hit the cat"). Most languages are "accusative", classifying intransitive actors with transitive actors; ergative languages treat them like transitive patients (English equivalent: "Ran John home."). The spontaneous intentional gestures of language-delayed deaf children from two distinct "accusative" cultures (American and Chinese) follow the ergative pattern, despite the non- ergative gestural patterns with which their parents reciprocate. The gestures produced by adults to describe pictures without using speech, and the eye movements of people viewing pictures, also evince ergative classification and an ergative bias towards the patient. What in most languages outweighs this deeply-rooted, pre-linguistic bias, Goldin- Meadow surmises, are the competing demands for expressiveness, speedy communication, and easy processing, better met (she doesn't say why) by accusative structures.
DISUSSION Emerging from all this research is one over-arching question: how deep and broad are these various effects of language on thought? Potentially, extremely so. The more simply a language expresses a concept, for example, the easier it is to perform mental operations on it. A language's most basic concepts are its most handy building blocks for novel, complex concepts and innovative ideas in all sorts of domains. Languages that delineate and construct concepts differently may foster faster, clearer, and more creative thinking in certain areas than other languages do. Further magnifying these cross-linguistic conceptual differences are our metaphorical mappings from one domain to another, especially from the spatial domain from which so many metaphors extend, and in which so much cross-linguistic variation has been uncovered.
While most of the essays focus on linguistic effects on non-linguistic thought, even the less debatable purely linguistic effects may, as Slobin points out, be more significant than many of us realize. Much of our conscious mental activity is verbal-not just whenever we think for speaking, but often (consider all those extensive interior monologues) when we think for ourselves. The effects of language on memory, cognitive attention, and cognitive habits, dismissed by some as applying only with linguistic encoding, may in fact occur all the time.
How great all these cross-linguistic differences are depends on how broadly and deeply languages differ in their conceptual structures. These studies reveal a variety of specific cross-linguistic differences. But for all their conclusions about the classifications of color, space, gender, artifacts, manner of motion, objects/materials, actors/patients, we are far from a comprehensive inventory of the semantic structures of the world's languages. Many (if not most) basic domains, and most higher-order concepts, remain unexplored.
Along a number of dimensions, languages are similar: most if not all, for example, have embedded complement clause structures and delineate between animates and substances and transitive actors and intransitive patients. While they may draw different boundaries between categories, they often agree on the core exemplars. Furthermore, nearly all scholars concede that at some level lies a pool of primitive, universal concepts-the building blocks for all others. Perhaps at the other extreme as well, above a certain level of complexity, most concepts can be expressed with more or less equal facility in all languages. Even if the similarities between languages outweigh the differences, however, comparing those with language to those who lack it reveals a huge influence of Language on thought.
Deliberate attempts to manipulate thought through language show definite limits to the power of language. Societal prejudices survive euphemistic language and nudge euphemisms ("retarded") back towards the pejorative; gender-neutral language hasn't eliminated gender discrimination and neutral terms ("co-ed") eventually assume the very sorts of biases they were invented to circumvent.
As Levinson points out (p. 33), language may most affect habitual or non-reflexive thought in specific domains, rather than the more conscious, deliberate thinking that conceptualizes such notions as Freedom and Equality (which, as Pinker observes, are ridiculous to brand as unthinkable if nameless).
But some of these specific domains, e.g. gender, may influence many others. Recall the Spanish and German speakers of Boroditsky et al's study, who characterized keys and other objects (masculine in one language, feminine in the other) in gender-stereotypical ways: keys were "hard", "heavy", and "jagged" in German; "golden", "intricate", and "lovely" in Spanish (p. 70). Might our conception of other things more consequential than keys-- say, Freedom, and our choices between pacifistic vs. militaristic routes to it-- depend somewhat, even if ever-so-slightly, on what gender they assume in our language?
One of the most intriguing influences of Language on thought discussed in this book is de Villiers and de Villiers conclusion that learning embedded complement clauses enables the calculation of false beliefs-- the prerequisite, in turn, for a full-fledged theory of mind. Teaching my autistic son complement clauses is on my agenda (right now we're working on the syntax of yes-no questions). It's a tantalizing thought that once he masters embedded complements, as I'm sure he will, he might be able to reason about false beliefs. Given how unaware he is of what goes on in other people's minds, achieving this (as improbable as it seems) would strongly suggest that de Villiers and de Villiers prerequisite is not just a necessary condition, but a sufficient one.
Perhaps people with full-blown autism never learn true language, even when they appear to be verbal. Discussing the importance of joint attention and social context in language learning, Tomasello observes that "associating sounds with experience is not language" (p. 50). But then how do autistics like my son, socially aloof and unaware, engage in the sophisticated abstract and conditional thinking (as he does in his many scientific experiments and innovative solutions to problems) for which some have deemed language the sine qua non? At any rate, highly verbal, highly autistic persons would make intriguing subjects for future research on the influence of language on thought.
Finally, any conclusion on this topic must consider that sometimes what looks like linguistic influence may instead be cultural. As Levinson points out (p. 27), "[t]he contents of language, and much of its form, are ... largely the products of cultural tradition". Myriad cultural clues assist children in constructing language-specific concepts, he notes. For example, "every aspect of the environment", and "a thousand little details", help children internalize the frame of reference for their language, whether it's a relative frame (left-right-front-back) or an absolute one (North-South-East-West). In relative languages there's the consistent way that books and doors open, the arrangement of silverware or the sock drawer, the preferred side of the sidewalk, (always to the right or left). In absolute language there's the direction in which people point their heads when they sleep or build their windbreaks (always to the North). In both types, additional clues come from where people point when referencing absent entities. (p. 43).
But these observations suggest that the cognitive differences that Levinson observes between speakers of absolute vs. relative languages-- for instance that the former have a better sense of direction-- may stem from culture rather than language. Similarly, differences between Yucatec and American culture may explain the relative bias towards materials in the former: people in industrial societies are presumably less involved in and conscious of the materials that compose implements than those in societies where locals construct implements by hand. (Lucy and Gaskins claim that the fact that 7-year-old Yucatec speakers show the language contrast before they show the cognitive one rules out a cultural influence, but perhaps the former are simply less involved with and aware of materials are than the 9-year-olds the authors contrast them with). Lastly, Tomasello's observations about how social cognition is inextricably caught up in the effects of language on thought raises the possibility that some of the ultimate cognitive influences are social, not linguistic.
Besides raising these broad questions, "Language in Mind" proposes reasonable answers to many smaller ones, along with dozens of fascinating observations. Its contributors, with their more precise understanding of linguistic distinctions and psychological processes, and their carefully-constructed hypotheses and experiments, appear fully capable of starting to answer the big questions convincingly.
REFERENCES Bickerton, D. (1995). Language and Human Behavior. Seattle: University of Washington Press.
Devitt, M. and Sterelny, K. (1987). Language and Reality: An Introduction to the Philosophy of Language. Oxford: Blackwell.
Pinker, S. (1994). The Language Instinct. New York: Morrow.
Slobin, D. (1996). From "thought and language" to "thinking for speaking." In J. Gumperz and S. Levinson (Eds.), Rethinking Linguistic Relativity (pp. 70-96). Cambridge: Cambridge University Press.
Vygotsky, L. (1962). Thought and Language. Cambridge, MA: MIT Press.
Whorf, B. (1956). Language, thought, and reality: Selected Writings of Benjamin Lee Whorf (J. Carroll, Ed.). Cambridge, MA: MIT Press.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER Katharine Beals received her Ph.D. in linguistics from the University of Chicago in 1995. After working for five years as Senior Software Engineer with the Natural Language Group at Unisys, she founded Autism Language Therapies (http://autism-language-therapies.com), where she designs and creates linguistic software for children with autism. She's also working on a book about her deaf, autistic son, which explores such issues as language modality, cochlear implants, and language and consciousness in deaf and autistic people.