Page 109 - Encyclopedia of Philosophy of Language

Further work, however, has suggested that the effects of the innate differential discriminability of col- ors can be partialed out in appropriately designed experiments. A measure of 'communication accuracy' (how well a speaker can describe a color to enable it to be identified by a listener) then proves an effective predictor of memory performance. So effects of lan- guage on memory for colors are demonstrable, but they are effects of some subtlety, and are not the most direct examples of language influencing thought. A succinct review of this work is provided by D'Andrade (1989).
3. Codability
Codability is a concept that has appealed to many experimental psychologists working on short-term memory: the quantity of material individuals are cap- able of retaining accurately in short-term memory is limited, and different encodings of the same infor- mation can differ in how readily they can be 'squeezed in.' An early demonstration of this was by S. Smith (cited in Miller 1956), who trained subjects to recede a list of binary digits (Os and Is) into octal (the digits 0 to 7), so 000 is receded as 0, 001 as 1, 010 as 2, etc. Subjects so trained were able to recall accurately much longer sequences of binary digits than subjects who had not received this training.
Such a result points to one important general func- tion of language in thought: recoding material in a compact form enables us to retain more of it in short- term memory, and any thought processes that depend on manipulation of such material should benefit. The details of this idea have been worked out more fully recently: 'working memory' is the preferred term for manipulations of material on a short-term basis, and it has been established that immediate recall of verbal material is heavily dependent on the operations of an 'articulatory loop' in working memory, whose capacity is limited by how much the subject can say in 1.5-2 seconds. If the material takes longer than 2 seconds to say (because it contains many syllables or because the subject is not an agile articulator) then it will not always be accurately recalled (for a good review, see Baddeley 1986).
This property of the human memory system has curious implications for crosscultural intelligence test- ing. Many tests of intelligence include as a component a test of 'digit span' or some similar measure of immediate recall of unrelated words. Digit span (how many digits one can reliably recall immediately after one has heard them) depends on how fast they can be said. Compared with a monosyllabic digit speaker, subjects who speak languages with polysyllabic digits will be able to say fewer digits in two seconds and thus remember fewer of them. If this is not taken into account in comparing raw intelligence test scores across languages, the polysyllabic speaker will seem less intelligent. This effect was first demonstrated for
Welsh and English by Ellis and Hennelley in1980, and confirmed in a study of English, Spanish, Hebrew, and Arabic by Naveh-Benjamin and Ayres (1986). The results are quite substantial, with English speak- ers (mean number of syllables per digit 1.0, the digit 7 being excluded from the Naveh-Benjamin and Ayres study) having a mean span of 7.21 digits, and Arabic speakers (mean number of syllables per digit2.25) having a mean span of 5.77. Whether this early bottle- neck in processing has any implications for more com- plex thought processes is not clear: there are no reports of speakers of a particular language being particularly disadvantaged in calculation, and it would be fanciful to suppose that the Arabs developed algebra because they were having such difficulties with arithmetic. Per- haps the effects of the bottleneck exist, but are too subtle to have been recognized so far.
Thevalueoflabelinginsomeproblem-solving tasks has been demonstrated. Rommetweit (in Campbell and Smith 1978) asked 8-year old Norwegian children to solve a number of problems appearing in one of two linguistic forms: either they had to select an object with respect to two adjectival properties (e.g., in an array of circles of different colors and sizes, they were asked to select the second largest white circle), or one of the adjectives was combined with the noun into a single label (thephrase white circle was replaced by snowball}. The children performed better on the second version of the task.
Examples of language as a coding device are not restricted to material that is already in a verbal form. Labeling of nonverbal material (e.g., pictures) is a useful mnemonic strategy, particularly because words are more easily rehearsed than visual images—use of this strategy does not appear in children until they are of school age. But labels simplify or even distort the information they are summarizing. A study byCar- michael, et al. (1932) presented subjects with ambigu- ous figures (e.g., a crescent shape) for which different subjects were given different labels (crescent moon or Letter C). In reproducing these figures later, subjects made systematic distortions of the original figure in the direction of the label they had heard (the crescent was more moon-like or more C-like).
A further example of the way language can distort nonverbal memory is provided in a study by Loftus and Palmer (1974). They showed subjects a film of a car crash and afterwards asked them how fast the cars were traveling when the collision occurred. If the word smashed was used in the question, estimates for speed were higher than if the more neutral word hit was used. Moreover, when questioned later, subjects who had previously received the smashed question were more likely to report (erroneously) the presence of broken glass.
4. Presuppositions and Prejudice
The examples in Sect. 3 are small-scale and short- term: labeling may effect memory for isolated patterns
Thought and Language
87