lecture 10 - speech production and speech errors Flashcards

Question

experimental Freudian slips?

Answer 1

* Motley & Baars (1979) * Hypothesis: Spoonerisms more likely when the resulting content is congruous with the situational context. * Method: 90 males, SLIP procedure. ○ 3 Conditions: □ “Electricity” - expecting to get shocked □ “Sex” - researcher provocatively attired female □ Neutral Experimental Freudian slips? * Same word pairs in all conditions * spoonerism targets were non-words (e.g. goxi furl -> foxy girl) * targets preceded by 3 phonologically biasing word pairs not semantically related to target words * Some resulting errors were sexually related (S), some were electrically related (E) * Bine foddy -> “fine body” * Had bock -> “bad shock” Experimental Freudian slips? * Results (number of errors, by type): * Electricity set: 69 E, 31 S * Sex set: 36 E, 76 S * Neutral set: 44 E, 41 S Speech errors sensitive to cognitive set

Answer 2

* Hypothesis * high levels of sex anxiety will make more “sex” spoonerisms than those with low sex anxiety. * Method * 36 males selected on the basis of high, medium, & low sex anxiety (Mosher Sex-Guilt Inventory). SLIP task same as previous experiment

Answer 3

* High sex anxiety > medium > low. * Conclusion: supports Freud’s view of sexual anxiety being revealed in Slips of the Tongue * BUT: any type of anxiety, not just sexual produced similar results. SO: anxiety was at play but it was more general, so the priming was more global.

Answer 4

* Semantic errors * Salient words intrude * Freudian slip * Disturbing intentions intrude Many non-Freudian errors

Answer 5

▪ Production levels (reverse comprehension) ▪ Speech errors reveal structure ▪ Planning ▪ Semantic association

Answer 6

Many animals are clearly able to solve some problems without language, suggesting that language cannot be essential for problem solving and thought. Although this may seem obvious, it has not always been considered so. Among the early approaches to examining the relation between language and thought, the behaviorists believed that thought was nothing more than speech. Young children speak their thoughts aloud; this becomes internalized, with the result that thought is covert speech— thought is just small motor movements of the vocal apparatus. Watson (1913) argued that thought processes are nothing more than motor habits in the larynx. Jacobsen (1932) found some evidence for this belief because thinking often is accompanied by covert speech. He detected electrical activity in the throat muscles when participants were asked to think. But is thought possible without these small motor movements? Smith, Brown, Thomas, and Goodman (1947) used curare to temporarily paralyze all the voluntary muscles of a volunteer (Smith, who clearly deserved to be first author on this paper). Despite being unable to make any motor movement of the speech apparatus, Smith later reported that he had been able to think and solve problems. Hence there is more to thought than moving the vocal apparatus. Perhaps language sets us apart from animals because it enables new and more advanced forms of thought? We need to distinguish how language and thought might affect each other developmentally, and in the fully developed adult state. We can list the possible alternatives; each of them has been championed at some time. First, cognitive development determines the course of language development. This viewpoint was adopted by Piaget and his followers. Second, language and cognition are independent faculties (Chomsky’s position). Third, language and cognition originate independently but become interdependent; the relation is complex (Vygotsky’s position). Fourth, the idea that language determines cognition is known as the Sapir– Whorf hypothesis. The final two of these approaches both emphasize the influence of language in cognition.

Answer 7

The Russian psychologist Vygotsky (1934/1962) argued that the relation between language and thought was a complex one. He studied inner speech, egocentric speech, and child monologues. He proposed that speech and thought have different ontogenetic roots (that is, different origins within an individual). Early on, in particular, speech has a pre-intellectual stage. In this stage, words are not symbols for the objects they denote, but are actual properties of the objects. Speech sounds are not attached to thought. At the same time early thought is non-verbal, so up to some point in development, when the child is about 3 years of age, speech and thought are independent; after this, they become connected. At this point speech and thought become interdependent: thought becomes verbal, and speech becomes representational. When this happens, children’s monologues are internalized to become inner speech. Vygotsky contrasted his theory with that of Piaget, using experiments that manipulated the strength of social constraints (see Figure 3.11). Unlike Piaget, Vygotsky considered that later cognitive development was determined in part by language. Piaget argued that egocentric speech arises because the child has not yet become fully socialized, and withers away as the child learns to communicate by taking into account the point of view of the listener. For Vygotsky the reverse was the case. Egocentric speech serves the function of self-guidance that eventually becomes internalized as inner speech, and is only vocalized because the child has not yet learned how to internalize it. The boundaries between child and listener are confused, so that self-guiding speech can only be produced in a social context. Vygotsky found that the amount of egocentric speech decreased when the child’s feeling of being understood lessened (such as when the listener was at another table). He claimed that this was the reverse of what Piaget would predict. However, these experiments are difficult to evaluate because Vygotsky omitted many procedural details and measurements from his reports that are necessary for a full evaluation. It is surprising that the studies have not been repeated under more stringent conditions. Until then, and until this type of theory is more fully specified, it is difficult to evaluate the significance of Vygotsky’s ideas.

Answer 8

In George Orwell’s novel Nineteen Eighty-Four , language restricted the way in which people thought. The rulers of the state deliberately used “Newspeak,” the official language of Oceania, so that the people thought what they were required to think. “This statement … could not have been sustained by reasoned argument, because thenecessary words were not available” (Orwell, 1949, p. 249, in the appendix, “The principles of Newspeak”). Orwell’s idea is a version of the Sapir– Whorf hypothesis. The central idea of the Sapir– Whorf hypothesis is that the form of our language determines the structure of our thought processes. Language affects the way we remember things and the way in which we perceive the world. It was originally proposed by a linguist, Edward Sapir, and a fire insurance engineer and amateur linguist, Benjamin Lee Whorf (see Whorf, 1956a, 1956b). Although Whorf is most closely associated with anthropological evidence based on the study of American Indian languages, the idea came to him from his work in fire insurance. He noted that accidents sometimes happened because, he thought, people were misled by words— as in the case of a worker who threw a cigarette end into what he considered to be an “empty” drum of petrol. Far from being empty, the drum was full of petrol vapor, with explosive results. The Sapir– Whorf hypothesis comprises two related ideas. First, linguistic determinism is the idea that the form and characteristics of our language determine the way in which we think, remember, and perceive. Second, linguistic relativism is the idea that as different languages map onto the world in different ways, different languages will generate different cognitive structures. Miller and McNeill (1969) distinguished between three versions of the Sapir– Whorf hypothesis. In the strong version, language determines thought. In a weaker version, language affects only perception. In the weakest version, language differences affect processing on certain tasks where linguistic encoding is important. It is the weakest version that has proved easiest to test, and for which there is the most support. It is important to consider what is meant by “perception” here. It is often unclear whether what is being talked about is low-level sensory processing or classification.

Answer 9

The anthropological evidence concerns the inter-translatability of languages. Whorf analyzed Native American Indian languages such as Hopi, Nootka, Apache, and Aztec. He argued that each language imposes its own “world view” on its speakers. For example, he concluded that as Hopi contains no words or grammatical constructions that refer to time, Hopi speakers must have a different conception of time from us. Whorf’s data are now considered highly unreliable (Malotki, 1983). Furthermore, translation can be very misleading. In fact, Whorf’s translation was very idiosyncratic, so it is far from clear that speakers of Apache actually dissect the world in different ways (Clark & Clark, 1977; Pinker, 1994). For example, both languages have separate elements for “clear,” “spring,” and “moving downwards.” Why should the expression not have been translated “It is a clear dripping spring”? The appeal of such translations is further diminished when it is realized that Whorf based his claim not on interviews with Apache speakers, but on an analysis of their recorded grammar. Lenneberg and Roberts (1956) pointed out the circularity in the reasoning that, because languages differ, thought patterns differ because of the differences in the languages. An independent measure of thought patterns is necessary before a causal conclusion can be drawn.

Answer 10

The way in which different languages have different vocabularies has been used to support the Whorfian hypothesis, in that researchers believe that cultures must view the world differently because some cultures have single words available for concepts that others may take many words to describe. For example, Boas (1911) reported that Eskimo (or Inuit) language has four different words for snow; there are 13 Filipino words for rice. An amusing debunking of some of these claims can be found in Pullum (1989): Whorf (1940/1956b) inflated the number of words for snow to seven, and drew a comparison with English, which he said has only one word for snow regardless of whether it is falling, on the ground, slushy, dry or wet, and so on. The number of types of snow the Inuit were supposed to have then varied with subsequent indirect reporting, apparently reaching its maximum in an editorial in the New York Times on February 9, 1984, with “one hundred” to “two hundred” in a Cleveland television weather forecast. In fact, it is unclear how many words Inuit has for snow, but it is certainly not that many. It probably only has two words or roots for types of snow: “qanik” for “snow in the air” or “snowflake”; and “aput” for “snow on the ground.” It is unclear whether you should count the words derived from these roots as separate. This story reinforces the importance of knowing how you define a “word,” and also of always checking sources! Speakers of English also in fact have several words for different types of snow (snow, slush, sleet, and blizzard). Vocabulary differences are unlikely to have any significant effects on perception— although again it is important to bear in mind what perception might cover. We can learn new words for snow: people learning to ski readily do so, and while this does not apparently change the quality of the skiers’ perception of the world, it certainly changes the way in which they classify snow types and respond to them. For example, you might choose not to go skiing on certain types of snow. Vocabulary differences reflect differences in experience and expertise. They do not seem to cause significant differences in bperception, but do aid classification and other cognitive processes. Not having words available for certain concepts does seem to have a detrimental effect. Members of the Piraha tribe from the Amazon basin have words for the numbers “one” and “two,” and then just “many.” Their performance on a range of numerical tasks was very poor for quantities greater than three (Gordon, 2004). Whereas we can count above two and assign precise numbers to quantities, members of the Piraha tribe just seem to be able to estimate. Not having a word available for a concept does appear to limit their cognitive abilities.

Answer 11

Carroll and Casagrande (1958) examined the cognitive consequences of grammatical differences in the English and Navaho languages. The form of the class of verbs concerning handling used in Navaho depends on the shape and rigidity of the object being handled. Endings for the verb corresponding to “carry,” for example, vary depending on whether a rope or a stick is being carried. Carroll and Casagrande therefore argued that speakers of Navaho should pay more attention to the properties of objects that determine the endings than do English speakers, and in particular they should group instances of objects according to their form. As all the children in the study were bilingual, the comparison was made between more Navahodominant and more English-dominant Navaho children. The more Navaho-dominant children did indeed group objects more by form than by color, compared to the English-dominant group. However, a control group of non-Native American English-speaking children grouped even more strongly according to form, behaving as the Navaho children were predicted to behave! It is therefore not clear what conclusions about the relation between language and thought can be drawn from this study. A second example is that English speakers use the subjunctive mood to easily encode counter-factuals such as “If I had gone to the library, I would have met Dirk.” Chinese does not have a subjunctive mood. Bloom (1981, 1984) found that Chinese speakers find it harder to reason counter-factually, and attributed this to their lack of a subjunctive construction. Their memories are more easily overloaded than those of speakers of languages that support these forms. There has been some dispute about the extent to which sentences used by Bloom were good idiomatic Chinese. It is also possible to argue counter-factually in Chinese using more complex constructions, such as (translated into English) “Mrs. Wong does not know English; if Mrs. Wong knew English, she would be able to read the New York Times ” (Au, 1983, 1984; Liu, 1985). Nevertheless, Chinese speakers do seem to find counter-factual reasoning more difficult than English speakers. If this is because the form of the construction needed for counter-factual reasoning is longer than the English subjunctive, then this is evidence of a subtle effect of linguistic form on reasoning abilities. A third example is that of grammatical gender. Although English does not mark grammatical gender, many languages do. Italian, for example, marks nouns as masculine or feminine, and German marks them as masculine, feminine, or neuter. Vigliocco, Vinson, Paganelli, and Dworzynski (2005) found that effects of gender on thought were highly constrained. They were found in Italian (a twogender language), but only with tasks that require verbalization and only with certain semantic categories (animals) and not others (artifacts). For example, when participants are asked to judge which two of three words are most similar (e.g., donkey– elephant– giraffe), grammatical gender affected similarity judgments for animals but not for artifacts. There were no effects at all in German, a language with an additional neuter gender. The likely reason for this difference is that in two-gender languages gender is a reliable cue to sex— but of course this rule is inapplicable with artifacts. The conclusion is consistent with a weak version of the Sapir– Whorf hypothesis— language can affect performance on some tasks that use language.

Answer 12

There is more evidence that language can have an indirect effect on cognition, particularly on tasks where linguistic encoding is important. Carmichael, Hogan, and Walter (1932) looked at the effects of learning a verbal label on participants’ memory for nonsense pictures (see Figure 3.12). They found that the label that the participants associated with the pictures affected the recall of the pictures. Santa and Ranken (1972) showed that having an arbitrary verbal label available aided the recall of nonsense shapes. Duncker (1945) explored the phenomenon known as functional fixedness, using the “box and candle” problem (see Figure 3.13) where participants have to construct a device using a collection of commonplace materials so that a candle can burn down to its bottom while attached to the wall. The easiest solution is to use the box containing the materials as a support; however, participants take a long time to think of this, because they fixate on the box’s function as container. Glucksberg and Weisberg (1966) showed that the explicit labeling of objects could strengthen or weaken the functional fixedness effect depending on the appropriateness of the label. This demonstrates a linguistic influence on reasoning. In an experiment by Hoffman, Lau, and Johnson (1986), Chinese– English bilinguals read descriptions of people, and were later asked to provide descriptions of the people they’d read FIGURE 3.13 The objects presented to participants in the “box and candle” problem. about. The descriptions had been prepared so as to conform to Chinese or English personality stereotypes. Bilingual people thinking in Chinese used the Chinese stereotype, whereas bilingual people thinking in English used the English stereotype. The language used influenced the stereotypes used, and therefore the inferences made and what was remembered. Hence work on memory and problem solving supports the weakest version of the Whorfian hypothesis. Language can facilitate or hinder performance on some cognitive tasks, particularly those where linguistic encoding is routinely important.

Answer 13

Hunt and Agnoli (1991) examined how different languages impose different memory burdens on their speakers. English has a complex system for naming numbers: we have 13 primitive terms (0– 12), then special complex names for the “teens,” then more general rule-based names for the numbers between 20 and 100, and then more special names beyond that. On the other hand, the number naming system in Chinese is much more simple, necessitating only that the child has to remember 11 basic terms (0– 10), and three special terms for 100, 1,000, and 10,000. For example, “eleven” is simply “ten plus one.” English-speaking children have difficulty learning to count in the teen range, whereas Chinesespeaking children do not (Miller & Stigler, 1987). Hence the form of the language has subtle influences on arithmetical ability, a clear example of language influencing cognition. Although Welsh numbers have the same number of syllables as their English counterparts, the vowel sounds are longer and so they take longer to say (Ellis & Hennelly, 1980). Hence bilingual participants had worse performance on digit-span tests in Welsh compared with English digit names, and also slightly worse performance and higher error rates in mental arithmetic tasks when using Welsh digit names. Key evidence comes from the Piraha people of the Amazon (Everett & Madora, 2011; Gordon, 2004). The Piraha lack precise numerical terms, and seem to have great difficulty on tasks involving numbers greater than three. It appears that in order to count accurately we need to have linguistic number terms available.

Answer 14

The most fruitful way of investigating the strong version of the Sapir– Whorf hypothesis has proved to be analysis of the way in which we name and remember colors. Brown and Lenneberg (1954) examined memory for “color chips” differing in hue, brightness, and saturation. Codable colors, which correspond to simple color names, are remembered more easily (e.g., an ideal red is remembered more easily than a poor example of red). Lantz and Stefflre (1964) argued that the similar notion of communication accuracy best determines success: People best remember colors that are easy to describe. This early work seemed to support the Sapir– Whorf hypothesis, but there is a basic assumption that the division of the color spectrum into labeled colors is completely arbitrary. This means that, but for historical accident, we might have developed other color names, like “bled” for a name of a color between red and blue, and “grue” for a name of a color between green and blue, rather than red, blue, and green. Is this assumption correct? Berlin and Kay (1969) compared the basic color terms used by different languages. Basic color terms are defined by being made up from only one morpheme (so “red,” but not “blood red”), not being contained within another color (so “red,” but not “scarlet”), not having restricted usage (hence “blond” is excluded), and being common and generally known and not usually derived from the name of an object (hence “yellow” but not “saffron”). Languages differ in the number of color terms they have available. For example, Gleason (1961) compared the division of color hues by speakers of English with that of the languages Shona and Bassa (see Figure 3.14). Berlin and Kay found that across languages basic color terms were present in a hierarchy (see Figure 3.15). If a language only has two basic color terms available, they must correspond to “black” and “white”; if they have three then they must be these two plus “red”; if they have four then they must be the first three plus one of the next group, and so on. English has names for all 11 basic color terms (black, white, red, yellow, green, blue, brown, purple, pink, orange, and gray). Berlin and Kay also showed that the typical colors referred to by the basic color terms, called the focal colors, tend to be constant across languages. Heider (1972) examined people’s memory for focal colors in more detail. Focal colors are the best examples of colors corresponding to basic color terms: they can be thought of as the best example of a color such as red, green, or blue. The Dani tribe of New Guinea have just two basic color terms, “mili” (for black and dark colors) and “mola” (for white and light colors), although subsequently there has been some doubt as to whether this really is the case. Heider taught the Dani made-up names for other colors. They learned names more easily for other focal colors than for non-focal colors, even though they had no name for those focal colors. They could also Berlin & Kay, 1969). remember focal colors more easily than non-focal colors, again even those for which they did not have a name. Three-year-old children also prefer focal colors: they match them more accurately, attend to them more, and are more likely to choose them as exemplars of a color than non-focal colors (Heider, 1971). In a similar way, English speakers attend to differences between light and dark blue in exactly the same way as Russian speakers, even though the latter have names for these regions of the color spectrum while English speakers do not (Davies et al., 1991; Laws, Davies, & Andrews, 1995; note that there has been considerable debate about whether these are basic color names). At first sight then, the division of the color spectrum is not arbitrary, but is based on the physiology of the color vision system. The six most sensitive points of the visual system correspond to the first six focal colors of the Berlin and Kay hierarchy. Further evidence that differences are biological and have nothing to do with language comes from work on prelinguistic children by Bornstein (1985). Children aged 4 months habituate more readily to colors that lie centrally in the red and blue categories than to colors that lie at the boundaries. Bornstein (1973) found an environmental influence on the take-up of these color terms. He noted that with increasing proximities of societies to the equator, color names for short wavelengths (blue and green) become increasingly identified with each other and, in the extreme, with black. He argued that the eyes of peoples in equatorial regions have evolved to have protection from excessive ultraviolet light. In particular, there is greater yellow pigmentation in the eyes, which protects the eye from short-wave radiation, at a cost of decreased sensitivity to blue and green. Brown (1976) discussed the revised interpretation of these color-naming data and their consequences for the Sapir– Whorf hypothesis. He concluded that these later studies show that color naming does not tell us very much about the Sapir– Whorf hypothesis. If anything, it appeared to emphasize the importance of biological factors in language development. There are some problems with some of these studies, however. Of the 20 languages originally described in the Berlin and Kay (1969) study, 19 were in fact obtained from bilingual speakers living in San Francisco, and the use of color categories by bilingual speakers differs systematically from that of monolingual speakers. In particular, the color categorization of bilingual people comes to resemble that of monolingual speakers of their second language, whatever their first language. This in itself would give rise to an artifactual universality in color categorization. There are also methodological problems with the expanded set of 98 languages studied later by Berlin and Kay (Cromer, 1991; Hickerson, 1971). The criteria Berlin and Kay (1969) used for naming basic color terms are suspect (Michaels, 1977). The criteria seem to have been inconsistently applied, and it is possible that the basic color terms of many languages were omitted (Hickerson, 1971). There were also problems with the materials used in the original studies by Heider. The focal colors are perceptually more discriminable than the non-focal colors used in Berlin and Kay’s array in that they were perceptually more distant from their neighbors. When the materials are corrected for this artifact, Lucy and Shweder (1979) found that focal colors were not remembered any better than non-focal colors. On the other hand, a measure of communication accuracy did predict memory performance. This finding suggests that having a convenient color label can indeed assist color memory. Kay and Kempton (1984) showed that although English speakers display categorical perception of colors that lie on either side of a color name boundary, such as blue and green, speakers of the Mexican Indian language Tarahumara, who do not have names for blue and green, do not. Hence having an available name can at least accentuate the difference between two categories. These more recent findings suggest that there are indeed linguistic effects on color perception. There are limitations on the extent to which biological factors constrain color categorization, and it is likely that there is some linguistic influence. The Berinmo, a hunter-gatherer tribe also from New Guinea, have five basic color terms. The Berinmo do not mark the distinction between blue and green, but instead have a color boundary between the colors they call “nol” and “wor,” which does not have any correspondence in the English color-naming scheme. English speakers show a memory advantage across the blue– green category boundary but not across the nol– wor one, whereas Berinmo speakers showed the reverse pattern (Davidoff, Davies, & Roberson, 1999a, 1999b). In a further series of experiments using more sensitive statistical techniques, Roberson, Davies, and Davidoff (2000) were unable to replicate Heider’s earlier results with the Dani with the Berinmo. They found no recognition advantage for focal stimuli, no facilitation of learning focal colors, and a relation between color recognition was affected by color vocabulary. It is now also apparent that even within English not all basic color terms are equal. “Brown” and “gray” are acquired later than other basic color terms, are the two least preferred colors, and are used less frequently in adult speech to children than other color terms (Pitchford & Mullen, 2005). In summary there appear to be effects of biological and linguistic constraints on memory for colors. Perhaps color naming is not such a good test of the Sapir– Whorf hypothesis after all. First, the task is clearly very sensitive to the details of the experimental procedures and materials used. Second, the more basic the cognitive or perceptual process, the less scope there is likely to be for the top-down influence of language, and color perception, a mechanism shared with many nonhuman species, is pretty low level. As Pinker (1994) observes, no matter how influential language might be, it is preposterous to think that it could rewire the ganglion cells. Third, in any case, there do appear to be effects of language on color perception: Roberson et al. found effects of categorical perception for colors, but aligned with linguistic categories rather than more biologically based categories.

Answer 15

In a recent review, Boroditsky (2003) concludes that there are several instances where encoding differences between languages leads to differences in performance by speakers of those languages. For example, different languages encode spatial languages in different ways. Most languages (such as English) use relative terms (e.g., front of, back of, left of, right of) to encode relative spatial terms. Languages such as Tzeltal (a Mayan language) use an absolute system (similar to our system of describing compass points, e.g., to the north). Speakers of Dutch (which uses the relative system) and Tzeltal interpret and perform very differently on a non-linguistic orientation task. In this task, people see an arrow pointing in one direction, to the left or right. The viewpoint is then rotated 180 degrees, and people are asked which is most like the one they had originally seen— an arrow pointing in relatively the same way, or absolutely the same way. Preferences depend on whether the language uses an absolute or a relative coding system, with the Dutch speakers preferring the right-pointing arrow if they had seen that previously, but the Tzeltal speakers preferring the left-pointing arrow (Levinson, 1996a). This is because “what is north” does not vary with rotation, but “what is left” does. Different spatial frames of references are acquired with ease by children from different cultures using different languages— the absolute and relative systems are acquired equally easily (Majid, Bowerman, Kita, Haun, & Levinson, 2004). Different languages encode time in different ways: in English we mainly use a front– back metaphor (look ahead, falling behind, move meetings forward), while Mandarin speakers systematically use vertical metaphors (with up corresponding roughly to last and down to next). Mandarin speakers are more likely to construct vertical timelines to think about time, while English speakers are more likely to construct horizontal ones. For example, Mandarin speakers are faster to confirm that March comes before April if they have just seen a vertical array of objects than if they had seen a horizontal one. English speakers showed the reverse pattern (Boroditsky, 2001). Similar differences in performance can be found for the way in which languages encode object shape and grammatical gender (Boroditsky, 2003). Languages differ in the way in which they encode movement— do these linguistic differences lead to cognitive differences? English encodes the direction of motion with a modifier (“to,” “from”) and the manner of motion in the verb (“walk,” “run”), whereas in Greek the opposite is the case: the verb encodes the direction of motion, while the manner is encoded by a modifier. Papafragou, Massey, and Gleitman (2002) tested Greek and English children on two types of task involving motion: one involving non-linguistic tasks (remembering and categorizing motion in pictures of animals moving around), the other involving linguistic description. They only found a difference in performance on the linguistic tasks. There has recently been debate about whether these linguistic differences reflect the presence or absence of external cues, and whether they affect performance on all tasks, or just linguistic tasks. Li and Gleitman (2002) argued that the results of the studies by Levinson and colleagues on spatial frames of reference described above were artifactual. Li and Gleitman suggested that the results depend on the presence of environmental cues. They tested a group of native English speakers, and found that they could make them perform using relative or absolute frames of reference depending on the presence of landmark cues in the environment. When participants could not see the outside world (the blinds of the testing room were down), the speakers tended to use a relative frame; when they could see the outside world (the blinds were up), they were more likely to use an absolute frame of reference. On the other hand, Levinson, Kita, Haun, and Rasch (2002) were unable to replicate these results, arguing that the purpose of the task was too apparent to Li and Gleitman’s participants. They also pointed out that their groups were tested with equal amounts of environmental cues available, being tested equally indoors and out. In summary, there is evidence that the way in which different languages encode distinctions such as time, space, motion, shape, and gender influence the way in which speakers of those languages think. These differences suggest that our language may determine how we perform on tasks that at first sight do not seem to involve language at all, although this claim remains controversial.

Answer 16

The weak version of the Sapir– Whorf hypothesis has enjoyed a resurgence. There is now a considerable amount of evidence suggesting that linguistic factors can affect cognitive processes. Even color perception and memory, once thought to be completely biologically determined, show some influence of language. Furthermore, research on perception and categorization has shown that high-level cognitive processes can influence the creation of low-level visual features early in visual processing (Schyns, Goldstone, & Thibaut, 1998). This is entirely consistent with the idea that, in at least some circumstances, language might be able to influence perception. Indeed, it is hardly surprising that if a thought expressible in one language cannot be expressed so easily in another, then that difference will have consequences for the ease with which cognitive processes can be acquired and carried out. Having one word for a concept instead of having to use a whole sentence might reduce memory load. The differences in number systems between languages form one example of how linguistic differences can lead to slight differences in cognitive style. We will see in later chapters that different languages exemplify different properties that are bound to have cognitive consequences. For example, the complete absence of words with irregular pronunciations in languages such as Serbo-Croat and Italian is reflected in differences between their reading systems and those of speakers of languages such as English. Furthermore, differences between languages can lead to differences in the effects of brain damage. The extent to which people find the Sapir– Whorf hypothesis plausible depends on the extent to which they view language as an evolutionarily late mechanism that merely translates our thoughts into a format suitable for communication, rather than a rich symbolic system that underlies most of cognition (Lucy, 1996). It is also more plausible in a cognitive system with extensive feedback from later to earlier levels of processing.

Answer 17

Perhaps the main conclusion about how language and thought are related is that there is a relationship, but it is a complex one. Environment and biology jointly determine our basic cognitive architecture. Within the constraints set by this architecture, languages are free to vary in how they dissect the world and in what they emphasize. These differences can then feed back to affect aspects of perception and cognition. We noted above that paralyzing overt speech does not stop us being able to think. Clearly language is an important medium of thought and conceptualization. Although there is a great deal of individual variation, a significant proportion of our mental life is conducted in language (Carruthers, 2002); we hear “inner speech,” which often seems to be expressing or guiding our thoughts, or which sometimes is the product of reading. The extent to which inner speech or language plays a real role in thinking is unclear and controversial (Carruthers, 2002). A strong view is that language is essential for conceptual thought and is the medium in which it is conducted; a weaker view is that language is the medium of conscious propositional (as opposed to visual) thought; an even weaker view is that language is necessary to acquire many concepts, and influences cognition in ways that we have seen above; yet another view is that there is essentially no relation at all (although language can clearly express thoughts). Carruthers presents evidence from a range of sources to justify his claim that inner speech is the glue that sticks cognition together, and enables the modules of the mind to communicate: that is, language is the medium of conscious thought. Even here we must note that there might be cultural differences. In the West, it is assumed that language and inner speech assist thinking; in the East, it is assumed that talking interferes with thinking. These cultural differences affect performance: thinking out loud helped European Americans to solve reasoning problems, but hindered Asian Americans (Kim, 2002; Nisbett, 2003). The influence of language on thought has some important consequences. For example, does sexist language really influence the way in which people think? Spender (1980) proposed some of the strongest arguments for non-sexist language. For example, that using the word “man” to refer to all humanity has the association that males are more important than females; or that using a word like “chairman” (rather than a more gender-neutral term such as “chair” or “chairperson”) encourages the expectation that the person will be a man. These expectations do have real effects. Gender-stereotyped nouns (e.g., “surgeon,” “nurse”) are those to which many people have a strong initial expectation of the gender of the person (surgeon as male, nurse as female). Readers take longer to read a subsequent pronoun referring to the noun if the pronoun is in conflict with the stereotyping (such as using “she” to refer to a surgeon rather than “he”; e.g., Kennison & Trofe, 2004). Such a theory is a form of the Sapir– Whorf hypothesis, although there has been surprisingly little empirical work in this area. As Gleitman and Papafragou (2005) conclude, clearly we can have thought without language— some animals clearly reason and solve problems; prelinguistic infants have rich cognitive abilities; people with brain damage destroying most of their language abilities display rich cognitive abilities. Yet there is also much evidence that language and culture can affect our ways of thinking. Language and thought are related, but in a complex way.