Psycholinguistics James Myers March 12, 2004 Language and cognition OVERVIEW: 1. Cognitive influences on language 2. Linguistic influences on cognition: the Sapir-Whorf Hypothesis 3. Lexical influences on cognition 4. Grammatical influences on cognition ============================================================= 1. Cognitive influences on language 1.1 A culture's experience can affect language: The old false story about the "many" Eskimo words for "snow": This is not in fact true; depending on how you count (Eskimo words are morphologically very complex), Eskimo only has about 12 words, which is not that different from English (Pullum 1991): "snow", "sleet", "slush", "hardpack", "powder", "flurry", "avalanche", "blizzard", etc. As Pullum (1991) also comments, even if it were true, it's not theoretically interesting: horse-breeders have various names for kinds of horses; botanists have names for shapes of leaves; neurologists have names for parts of the brain; etc. 1.2 Innate nonlinguistic cognition can also affect language 1.2.1 Color term hierarchy: Berlin and Kay (1969) BLACK, WHITE RED, GREEN YELLOW BLUE BROWN, PINK PURPLE, ORANGE, GREY Where does this hierarchy come from? Perhaps the color processing subsystem of visual cognition (see e.g. Hubel 1988): Most basic light-sensitive cells in retina are "rods": sensitive to light and dark (BLACK vs. WHITE). Three types of color-sensitive "cones": blue, green and red. Cones are linked to two types of "bipolar cells": blue/yellow and green/red. Red cones inhibit both types of bipolar cells, so RED is most "powerful" color. GREEN is opposite of red; YELLOW is also generated by red cones. BLUE requires red cones to be inactivated. Other colors require more complex mixtures. 1.2.2 The effect of spatial cognition on language: Landau and Jackendoff (1993). "Design of Spatial Representation Hypothesis": due to innate properties of spatial cognition, language ends up having characteristics that parallel spatial cognition. Spatial cognition is divided into two basic subsystems: shape ("what") or location ("where") Evidence: a monkey may be trained to find food in a distinctive object regardless of location, or to find food in an object located near a particular reference point. Performance in the "what" task is impaired by damage to the inferior (=lower) temporal cortex, whereas performance in the "where" task is impaired by damage to the posterior (=back) parietal cortex. Similar patterns from (naturally) brain-damaged humans. A parallel what/where difference is found in language: Shape is encoded by nouns: e.g. "plate" vs. "bowl"; "jar" vs. "bottle" vs. "jug" (Also recall the "Shape Bias") Location is encoded by prepositions, which mostly ignore shape: "in" can be used with any of the above nouns (except plates, which require "on") Nouns and prepositions are quite different grammatically, just as the "what" and "where" cognitive systems are quite different. 1.3 Summary: Nonlinguistic cognition gives us certain categories; language must be constructed within these categories. 2. Linguistic influences on cognition: the Sapir-Whorf Hypothesis 2.1 Some biographical background: Edward Sapir was one of the great American linguists of the 20th century (he wrote classic papers such as "The psychological reality of the phoneme") Benjamin Lee Whorf was a very talented amateur linguist who became Sapir's student. Perhaps because he was an amateur, Whorf made some pretty serious mistakes which hurt his credibility as a theorist: Whorf is responsible for popularizing the "Eskimo snow" myth. Whorf supported his theories by pointing to Hopi (spoken in the American Southwest): [quotations from Pinker 1994:63] He claimed that the Hopi language contains "no words, grammatical forms, constructions or expressions that refer directly to what we call 'time', or to past, or future, or to enduring or lasting." This supposedly paralleled Hopi culture, which has "no general notion or intuition of TIME as a smooth flowing continuum in which everything in the universe proceeds at an equal rate, out of a future, through a present, into a past." Reality: Hopi speech contains tense, metaphors for time, units of time (e.g. days, parts of day, yesterday and tomorrow, weeks, months, seasons), ways to quantify units of time, and words like "ancient", "long time", "finished". Their culture keeps records with sophisticated dating methods, including a sun calendar, ceremonial day sequences, sundials, etc. 2.2 The Whorf Hypothesis (Whorfian Hypothesis, Sapir-Whorf Hypothesis): one's language influences how one thinks. Linguistic determinism: language influences thought. Linguistic relativism: different languages influence different thoughts. Does this hypothesis conflict with the view of language as a module? Maybe, but it depends on what you mean by "influence". 2.2.1 How might language influence thought? Answer 1: Thought is ALWAYS determined by language. ("Strong" version -- TOO strong!) It predicts: Thought should be impossible for prelinguistic infants, animals, aphasics, etc. (??!) Images, music, etc should be encoded as sentences. (???!) Speakers of a language with a single word for two distinct entities (e.g. English "uncle" is used both for father's brother and for mother's brother) should never be able to perceive these entities as different. (?????!!!!) Answer 2: Thought is SOMETIMES determined by language. ("Weak" version) Two versions of this: Answer 2a: Thought is sometimes RESTRUCTURED by language. This predicts that speakers of a language with a single word for two distinct entities should tend to perceive these objects as the same, no matter what tasks are used to test this. Answer 2b: Thought is sometimes ENCODED by language. When? During off-line tasks when subjects have enough time to "talk" to themselves: i.e. use internal speech. Speakers of a language with a single word for two distinct entities will tend to treat them the same in tasks where they are allowed to name the objects to themselves... But there will be no direct effect on non-linguistic cognition (e.g. as shown in a task where labels cannot be used). 2.2.2 How can one test Answer 2b (which I could call the "Boring Version of the Whorf Hypothesis")? Well, one major prediction is: On-line measures of the influence of language on thought (e.g. reaction times in categorization tasks) will show little if any effect. This contrasts with Carroll (p. 381): "...cognitive measures such as reaction time have not been used very often in studies of the Whorf hypothesis. This is unfortunate because effects of the sort that Whorf had in mind may be rather subtle." On the contrary, I think the effects may disappear with speeded tasks, because subjects would then have less time to "talk to themselves". 2.3 Another methodological problem: how can you separate linguistic relativism from cultural relativism? E.g. Schwanenflugel and Rey (1986) find that English speakers and Spanish speakers give quite different typicality ratings to birds: English speakers consider the robin most prototypical, while Spanish speakers consider the canary most prototypical. But is this due to a linguistic or cultural difference? So one must be careful to show that any purported difference in cognition is coming from the language itself (see below for problematic examples like this). 3. Lexical influences on cognition 3.1 Color terms again: Kay and Kempton (1984) Purpose: to provide evidence for the "Boring Version of the Whorf Hypothesis", against the other, more interesting versions Subjects: Speakers of English and speakers of Tarahumara, a Mexican language with no separate words for "blue" and "green" (somewhat like classical Chinese, by the way!) Materials: 3 colored chips: 1 "blue", 1 "green", and 1 in between "blue" and "green". Task: decide if the middle color is more similar to the first or to the second color. Results: English speakers behaved "categorically": they could always clearly decide if the middle color was closer to "blue" or to "green". But Tarahumara speakers did not show this "categorical" effect. Wrong interpretation (Carroll, p. 372): "Thus, the perception of colors appears to be dependent upon the terms we use to refer to them." Perception??? Right interpretation (according to the original authors of this study -- Kay and Kempton 1984): English speakers did well because they used what K&K called the "name strategy": they were explicitly encoding the task in linguistic terms -- they were "talking to themselves". The Tarahumara speakers couldn't use this strategy (due to the lack of separate words for "green" and "blue"), so they did worse. In a second experiment, K&K encouraged English subjects to label the middle chip "both blue and green." This means that the "naming strategy" now allows English subjects to use three categories in the task, and the "categorical" effect disappeared. Thus language influenced the English subjects' BEHAVIOR on the task, but not their PERCEPTION: the structure of thought was NOT affected by language. 3.2 The representation of number: Miura (1987), Miura and Okamoto (1989). Subjects: first-graders from the US and Japan (also tested Korean and Chinese children) Task: must represent numbers using colored blocks which included white "ones" blocks and purple "tens" blocks. Results: Asian kids much more likely to use the "tens" blocks most efficiently (e.g. representing "42" with 4 "tens" and two "ones" rather than 42 "ones" or some weird combination). Wrong interpretation: (Carroll 1999:375) "...the way one thinks about numbers is fundamentally different in Chinese versus English." Is THINKING really deeply affected? Right interpretation (according to me): During the task, Asian kids could use linguistic labels for the blocks: e.g. a purple block can be called "the  block", so 42 is literally 4  blocks plus 2 white blocks. Thus they had an extra TOOL available that the American kids didn't have: the Chinese words. Using a linguistic tool is NOT the same as having a different cognitive structure!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 4. Grammatical influences on cognition 4.1 The inverse of Landau and Jackendoff (1993): linguistic influences on spatial cognition? Pederson, et al. (1998) say YES. 4.1.1 There are two ways that languages can indicate locations: "Relative": relative to observer's orientation (e.g. "left, right"). Like English, Dutch uses relative terms for spatial relations at a small scale: "left", "right", "front", "back" "Absolute": relative to geographical directions (e.g. "east, west"). Tzeltal (Mayan language) makes extensive use of absolute terms even at small scales: "uphill" (roughly "south"), "downhill" (north), and "across" (both east and west). 4.1.2 Five distinct spatial tasks, all involving 180 degree rotation of subject. E.g. "Animals in a row" task: Subjects saw on Table 1 three toy animals in a head-to-tail row, all facing the same way. When subjects "had seen enough", the animals were taken away for 30 seconds. Then subjects were rotated 180 degrees and asked to rebuild "the same" array on Table 2. Results: Dutch subjects gave mostly all "relative" responses, while Tzeltal subjects gave "absolute" responses. 4.1.3 Right interpretation (according to me again): As usual, subjects have to figure out what the experimenter means by "the same", and inner speech leads them to the language-influenced results seen here. The end. 4.1.4 My hypothesis has actually been tested. Li & Gleitman (2002) had English-speaking American subjects (college students) perform the "table turning" task, but varied the environment. When the task was performed in an environment with clear indication of absolute orientation (e.g. with fixed objects always on, say, the north end of both tables), the subjects behaved just like speakers of "absolute" languages! 4.2 Classification by morphology/syntax: 4.2.1 Carroll and Casagrande (1958): Navajo marks object shape on the verb; Navajo-speaking kids group objects on the basis of form earlier than their English-speaking counterparts. "Right" interpretation: the Navajo kids were "talking to themselves" in the task, as usual. 4.2.2 Lucy (1992): Yucatec Maya distinguishes between animate and inanimate nouns; it also has classifiers based on material rather than shape (more evidence that the Shape Bias is not extremely strong!) Results: the usual. "Right" interpretation: the usual. 4.2.3 Note 1: These researchers never try to block the subjects from "talking to themselves"! So as usual all these results are consistent with the "Boring Version" of Whorf! Note 2: somebody should replicate these studies using Chinese classifiers, but this time try to directly falsify the Boring Version by measuring RT, blocking internal speech, etc!!!! 4.3 Counterfactual reasoning: (my discussion here is stolen from Wu 1994) 4.3.1 Bloom (1981) claimed that Chinese does not have any way of indicating a counterfactual (i.e. marking when a statement is contrary to fact). English marks counterfactuals with morphology (the so-called subjunctive mood): "If all circles WERE large and this small triangle WERE a circle, WOULD it be large?" Chinese cannot mark counterfactuals morphologically: 安┮Τ蛾伴常狦硂 琌蛾伴ê或硂à琌ぃ琌 So Bloom (1981) gave the above sentences to English/Chinese speakers: 95% of English-speakers responded "Yes" to the above counterfactual question. Only 25% of Taiwanese subjects responded "Yes"; the remainder either questioned or rejected the premise: "How could triangles be circles?" "But triangles are not circles." Bloom's interpretation: Chinese are able to perform counterfactual reasoning, but their language makes it more difficult for them than for speakers of other languages. 4.3.2 Here my usual criticisms about the "Boring Version" of the Whorf Hypothesis are less relevant, since we assume that reasoning through these tricky logical puzzles always requires inner speech. Still, it's important to know if Bloom is right, since he and others have emphasized the social consequences of this supposed difference: In 1972-3, Bloom gave a questionnaire in Hong Kong: "If your government were to pass a law requiring that all citizens born outside of Hong Kong make weekly reports of their activities to the police, how would you react?" English-speaking citizens had no trouble answering this question. Chinese-speaking citizens simply replied: "But the government hasn't", or "It can't" or "It won't." Wu (1994:15) comments, regarding Elman's (1983) response: "Focussing on his methodological 'precision,' Elman considered Bloom's findings to provide hard empirical evidence for the long suspected 'cosmological gulf' between China and the rest of the world, and also for the 'characteristic mindset' of the Chinese." So is Bloom right or wrong??? Does the Chinese language make it difficult for Chinese people to imagine alternative worlds? 4.3.3 The first criticism is that the Chinese language CAN express counterfactuals (from Wu 1994): 璶琌и笵иΝ碞 Ν笵и碞ぃ 砯璶琌и碞硄 璶ぃ琌腊Ν碞斌 璶琌琌╧碞 4.3.4 Second, is it a difference in processing logic, as Bloom claimed, or merely a difference in dealing with the content of the sentences? Wu (1994) did her own experiments. Subjects: Taiwanese and American students. Materials: Bloom's "Triangle" question "XY" question: a logical equivalent of the triangle question, but using different content: "If all X's were large and this small Y were an X, would it be large?" 安砞┮Τヒ常安砞琌ヒ杠 ê或琌ぃ琌 Results: Americans did better than Taiwanese on the "triangle" question, just as Bloom found. But Taiwanese students did better on the "XY" question than the Americans!! This shows that whatever explains the results, Bloom's analysis cannot be right. 4.3.5 Is the "Bloom effect" due to language at all? Ramsey (1987): [quoted in Wu 1994:32] "The very complexity of a society makes it extremely difficult to isolate any one factor [e.g. language] ... as responsible for a particular set of results.... I suspect that if Bloom had shown his triangle to a group of working class Americans, he would have been met with looks as puzzled as any he saw in China." Lardiere (1992): (for me, the most convicing counterargument!) Subjects: Well-educated Arabs, fluent in English (note: Arabic marks counterfactuals quite explicitly, just like English) Materials: "Triangle" sentence, in ENGLISH!!! Results: Only 29% responded "Yes". The rest questioned or rejected the premise, just like the Chinese subjects. Interpretation: There are CULTURAL differences (unknown) that affect behavior on this task. It's not an issue of language processing at all! 4.4 So I don't like the Whorf hypothesis. However, I may be wrong. There are some experiments using new methods that have found some pretty interesting results (though still problematic). 4.4.1 One problem with almost all experiments on the Whorf hypothesis, as noted above, is that you can't be sure that subjects are really not "talking to themselves" during the nonlinguistic task. What if you made the task a linguistic task, but primed (i.e. mentally prepared) the subjects with nonlinguistic cognitive information? 4.4.2 Boroditsky (2001) used this new approach. Background: Chinese often represents time using a vertical spatial metaphor (so March is "above" April), while English represents time using a horizontal metaphor (March is "before" April). Subjects: Mandarin speakers and English speakers Subjects were first "primed" by looking at visual objects arranged vertically or horizontally -- nonlinguistic information. Task: After seeing these objects, they had to judge the truth of temporal statements in their native language, e.g. "March is before April." Results: Mandarin speakers were faster to confirm such statements if they first saw the vertical objects, and English speakers were faster if they first saw the horizontal objects. 4.4.3 But this type of experiment also has a problem. How do we know that it's not priming of the linguistic representation? That is, the Mandarin speakers saw pictures they can describe as "above" or "below" (in fact, in one experiment the pictures were accompanied by explicit descriptions!). This activates the semantics of the relevant words that then appear in the temporal statements. So is the Whorf hypothesis really supported by such results or not? REFERENCES Berlin, B., & Kay, P. (1969). Basic color terms: Their universality and evolution. University of California Press. Bloom, A. H. (1981). The linguistic shaping of thought: A study in the impact of language on thinking in China and the West. Erlbaum. Boroditsky, L. (2001). Does language shape thought? Mandarin and English speakers' conceptions of time. Cognitive Psychology, 43, 1-22. Carroll, J. B., & Casagrande, J. B. (1958). The function of language classifications in behavior. In E. E. Maccoby, T. M. Newcomb, & E. L. Hartley (Eds.) Readings in social psychology (pp. 18-31). Holt, Rinehart and Winston. Elman, B. (1983). Review of A. Bloom's The linguistic shaping of thought: A study in the impact of language on thinking in China and the West. Journal of Asian Studies, 42, 611-614. Hubel, D. (1988). Eye, brain and vision. Freeman. Kay, P., & Kempton, W. (1984). What is the Sapir-Whorf hypothesis? American Anthropologist, 86, 65-79. Landau, B., & Jackendoff, R. (1993). "What" and "where" in spatial language and spatial cognition. Behavioral and Brain Sciences, 16, 217-238 [target article], 262-265. Lardiere, D. (1992). On the linguistic shaping of thought: Another response to Alfred Bloom.. Language in Society, 221, 231-251. Li P., & Gleitman L. (2002). Turning the tables: Language and spatial reasoning. Cognition, 83(3), 265-294. Lucy, J. A. (1992). Grammatical categories and cognition: A case study of the linguistic relativity hypothesis. Cambridge University Press. Miura, I. T. (1987). Mathematics achievement as a function of language. Journal of Educational Psychology, 79, 79-82. Miura, I. T., & Okamoto, Y. (1989). Comparisons of U.S. and Japanese first graders' cognitive representation of number and understanding of place value. Journal of Educational Psychology, 81, 109-113. Pederson, E., Danziger, E., Wilkins, D., Levinson, S., Kita, S., & Senft, G. (1998). Semantic typology and spatial conceptualization. Language, 74(3), 557-589. Pinker, S. (1994). The language instinct. William Morrow. Pullum, G. K. (1991). The great Eskimo vocabulary hoax and other irreverent essays on the study of language. University of Chicago Press. Ramsey, S. R. (1987). The languages of China. Princeton University Press. Schwanenflugel, P. J., & Rey, M. (1986). The relationship between category typicality and concept familiarity: Evidence from Spanish- and English-speaking monolinguals. Memory and Cognition, 14, 150-163. Wu, C. H.-F. (1994). "If Triangles Were Circles,..." -- A Study of Counterfactuals in Chinese and in English. Taipei: Crane Publishing Company.