Vous êtes sur la page 1sur 17

Language and Reasoning 1

Reasoning in Ideographic and Analytic Languages Renee Dale Louisiana State University

Language and Reasoning 2 Abstract Chinese speakers are thought to be at a disadvantage in logical thinking because of a lack of counterfactual indicators. In this study, the performance of English, Chinese, and bilingual subjects on reasoning problems with different levels of abstractness was compared. It has been found that subjects ability to reason correctly is decreased with more abstract reasoning problems, such as classical logic problems, such as, if p, then q. Such classical abstract problems were compared with less abstract problems. In one manipulation, numbers were used instead of the abstract references, mimicking mathematical problems. In the least abstract manipulation, easily identifiable words were substituted for abstract references. Chinese speakers were given problems in Chinese, English speakers in English, and bili ngual speakers constructions with the counterfactual indicators in either Chinese or English, and the subject matter in the alternate language. It was found that the highest rate of subject judgment of validity was highest with words, and lower with numbers and letters. Chinese and bilingual subjects outperformed English subjects on word problems. No significant difference was found between the letter and word problem types. The bilingual subjects performed significantly better on the word problems when the logical structure was in English, and the subject matter in Chinese.

Language and Reasoning 3 Reasoning in Ideographic and Analytic Languages Problems with counterfactuality are the source of many reasoning errors. Counterfactual reasoning problems reflect possibilities in alternate worlds , such as, If I had a car, then I would have a house. This statement could be true in any possible world, even if false in the present one. True inferences from this statement are, I have a car and a house, I do not have a car or a house, and I do not have car and I have a house. A false inference would be I have a car and do not have a house. Santamaria, Espina, and Byrne (2005) found that individuals inferred the not-p, not-q statements easier when presented with counterfactuals, with p and q statements slightly more difficult to identify, and not-p and q statements the most difficult. With increasing symbolism of the logical problem, the ease of inference would likely increase. The reasoning process which individuals utilize when analyzing logical problems is under much debate. One theory is that people translate the logical problem into the logical format, such as if p, then q. This theory seems to be invalidated by the fact that the Chinese language contains no letters, and is the basis for the condition wherein logical problems contain letters. Another theory is that people construct mental models of the statements within the logical problem. Laird, Legrenzi, Girotto and Legrenzi (2000) found consistent reasoning errors on the part of the subjects in their experiments which supported the theory of mental models . Subjects were fairly accurate in determining true statements, but consistently failed at detecting falsities. It was thought that when subjects made internal truth tables, they only contained statements which modeled truth, in addition to the minimal number of statements . This caused validity to be misjudged, as inconsistent false statements seemed to be consistent and true,

Language and Reasoning 4 based on the truth-centered mental picture. These findings were best explained by the mental models theory, wherein individuals model possibilities in separate premises, consisting primarily of premises verifying logical statements, to simplify the reasoning process. It is often thought that the Chinese language does not easily allow for logical reasoning because of a lack of ability to indicate counterfactuals (Laing, Cherian & Liu, 2010; Laird et al, 2000; Feng & Yi,2006). Feng and Yi (2006) directly compared the ability of English and Chinese speakers to identify counterfactuals. Both English and Chinese have indicators for the counterfactual state. The subjuncti ve mood strongly indicates a counterfactual in English, with various markers indicating the possibility condition in Chinese. The rate of accurate judgment of counterfactuality was directly proportional to the strength of the relative indicator for both Chinese and English subjects. Laing et al (2010) found that Chinese speakers required more time to process counterfactuals, as opposed to English speakers. It was thought that this was due to the counterfactual indicators being not obvious. English contains a structure dedicated to logical reasoning; specifically, the if-then construction. Chinese lacks such constructions, and must rely on time indicators to denote futurity and causality. This requires extra cognitive processing to discern the temporal situation of the logical problem. It might thus be inferred that, though possible, extra time is needed for Chinese speakers to create mental models from counterfactual statements. Symbolism within a logical problem increases the rate of subjects correct judgments. Even in cases of mathematical logic, subjects have been found to utilize the visual structure in judging validity (Laney & Goldstone, 2007). Spacing was manipulated in

Language and Reasoning 5 each experiment. In experiment 1 the spacing between mathematical arguments was increased, and in experiment 4 the arguments were directly associated with one another by the addition of parentheses. Subjects accuracy decreased when arguments were deceptively presented as inconsistent, but increased when the arguments associated with one another were consistent. Such findings are consistent with a reliance upon visual perception, even in logical reasoning problems. The relative symbolism of a logical problem would be related to the ability of the subject to conceptua lize and visualize the problem, thereby easing the construction of a mental model. Ideographic languages represent ideas and concepts with their characters, as opposed to a particular word with a particular pronunciation. Often one character can be pronounced multiple ways, and mean multiple things, regardless of relevance. By virtue of ideographic languages increased symbolism and conceptuality, it is theorized that subjects would be more easily able to translate written problems i nto concepts and mental models of those concepts. Reasoning problems presented in narrative format further increase subject accuracy. Wason and Shapiro (1971) found that abstract material in reasoning problems caused subjects to err more in their judgements. They separated subjects into either a group with reasoning problems which were based upon a story, or a group with abstract reasoning problems. The group with realistic situations was able to judge validity more accurately than the group with abstract problems. Situations which subjects are able to visualize ease the subjects creation of a mental model of the reasoning problem. It is hypothesized that symbolic representation of arguments will facilitate mental modeling of true inferences. The utilization of a conceptual language will further ease

Language and Reasoning 6 this process. Subjects judgment of the validity of problems containing abstract concepts will be the least accurate, with no significant difference across participants . When the abstract concepts are substituted with numbers, the subjects accuracy will increase. Finally, when words referring to real world phenomena are used as the conceptual information in problems, the subjects will be most accurate in judging validity. This will be relatively increased if the material is in Chinese, as opposed to English. The bilingual participants are expected to perform best on less abstract material when the logical structure is in English, with more abstract material being approximately equal for any construction. The worst performance is expected when the logical structure is in Chinese and the subject matter is in English. Method Participants All participants were pre-screened for educational experience, and had obtained a minimum of 3 years of schooling after secondary school. Participants were rejected if they were highly educated in the subject of reasoning or logic. Participants were also required to have taken at least one mathematics class which involved variables. Forty Mandarin Chinese speakers were selected from Macau, due to the high population of Chinese speakers and only one percent English speaking population. Subjects ranged in age from 29 to 32, with 20 females and 20 males participating. Forty English speakers from Raleigh, North Carolina participated in the experiment. These subjects also ranged in age from 30 to 33, and were half female and half male. Eighty bilingual speakers were selected from Hong Kong, China, because of the educational systems

Language and Reasoning 7 emphasis on educating students in multiple languages. Subjects ranged in age from 30 to 32, with 40 females and 40 males participating. Stimuli Each reasoning problem concerns either letters, numbers, or easily identifiable words. For the Chinese subjects, the problems are entirely in Chinese; for English subjects, the problems are entirely in English. To ensure proper and consistent grammar, three translators skilled in both Chinese and English from the Mandarin Training Center in Taipei, Taiwan created the question, answer and direction formats. In the letter condition, the letters were English, and were letters normally used for variable names. For bilingual subjects, the set of problems were transformed into half-Chinese, halfEnglish constructions. When the conditionals were in Chinese, the sub ject matter of the problem was in English; and when the conditionals were written in English, the subject matter was in Chinese. Sixty problems were constructed. Twenty problems contained letters as the argument type, 20 contained numbers, and the remaini ng 20 contained words. For the bilingual subjects, 30 problems were phrased with Chinese conditionals, and 30 with English conditionals. Of each construction, 10 problems involved letters, 10 problems involved numbers, and 10 problems involved words. The sequence and type of problem each subject received was randomized, with each subject receiving 15 problems. The subjects were instructed to choose which sets of possibilities were valid for the problem presented. Each problem was in the form If p then q. Each set of answers included possible inferences of four types: modus ponens (if the antecedent is valid, then the consequent is valid), modus tollens (if the consequent is not valid, then the

Language and Reasoning 8 antecedent is not valid), and two fallacious inferences, affirming the consequent (if the consequent is valid, then the antecedent is valid) and denial of the antecedent (if the antecedent is not valid, then the consequent is not valid). Each question was scored on a four point scale. Choosing no correct answers garnered a score of zero, and choosing all correct answers got a score of four. All 24 letters in the alphabet were used, and randomized for each trial. English speakers and Chinese speakers received randomized letters commonly used as variable names: x, y, z, i, t, m, s, w, a, b, c, j, k, and v. Arabic numerals from 1 to 9 were randomized for each trial for both English and Chinese speakers. Finally, for each word problem, a word was chosen out of a 40-word word bank. English and Chinese speakers received the same word in the appropriate language. Common words involved in everyday life were chosen. Each Chinese word contained no more than two characters for the corresponding single English word. Procedure Each subject was tested individually in a public location near to them. Each subject was given one minute for each problem. The problems were displayed upon a computer screen, in front of which the subject was seated. A stopwatch displaying the time the subject had left was displayed upon the screen. The possible inferences the subject could choose from were displayed at the same time as the problem itself. The subject was allowed paper to make notes on. Results Data from any participant who answered less than 10 problems was thrown out for presumed lack of effort. Two English participants and three bilingual participants were

Language and Reasoning 9 removed in this way, and to counterbalance, two Chinese participants and an additional bilingual participant data were removed by randomization. The accuracy of the participants from each language manipulation and condition was analyzed by six oneway ANOVAs (Figure 1). The ANOVAs showed variation between conditions (F>1, p<.05), and were then analyzed by five post-hoc tests (Figure 2). A main effect was found in each of the word conditions (p<.05). A main effect was found in the English condition, but no difference in overall performance between the Chinese and Bilingual groups (p>.05). In the English condition, word problem types differed significantly from the other problem types (p<.05), but there was no difference for English speakers in the letter and number problem accuracy (p=1). The Chinese group was found to perform significantly better in the word problem type as opposed to the other two types (p<.05), but there was no significant difference in performance between letter and number performances (p>.05). Bilinguals also performed significantly better in the word condition (p<.05), but there was no significance for the other two types of problems (p>.05). The bilingual manipulation was then analyzed with three paired t-tests (Figure 3), to test for the accuracy differences between the two types of problem compositions. Problem group A had its logical structure in English, and the subject matter in Chinese. Problem group B had its logical structure in Chinese, and the subject matter in English. The difference between problem groups was significant for the word condition (F=495.445, p=.003). The difference between the problem groups for the other two conditions was insignificant (p>.05). Conclusions

Language and Reasoning 10 The difference in counterfactual reasoning ability is compared between Chinese, English, and Bilingual participants, across three types of problems of varying levels symbolic and visual abstraction. The results supported the hypothesis that highly abstract problems will result in the least amount of subject accuracy for all language conditions. It was initially thought that there would be a significant difference in accuracy for letter and number problems, due to the difference in abstraction, but the difference was found to be insignificant. Number and letter problems were equally well performed on for all subjects, with a significant increase in accuracy for word problems overall. Chinese and bilingual subjects had an additional increase in accuracy with word problems over English subjects, but were not different from one another. Bilingual subjects performed equally for both groups of logical problems, regardless of linguistic constructions, with the exception of word problems. Bilinguals were significantly more accurate with word problems when the logical structure was in English, and the subject matter was in Chinese, as opposed to the logical structure being in Chinese, and the subject matter being in English. These findings supported the evidence that narrative format increases subjects reasoning accuracy over that of abstract logical problems. The increased accuracy of Chinese and bilingual subjects over English subjects in reasoning word problems supports the theory of mental models, as it is correspondent with the level of conceptuality of the problems. The lack of difference between number and letter problems was surprising, as it was thought that a more mathematical-type problem would be less abstract to subjects than a traditional format logical problem (Laney & Goldstone, 2007). Perhaps the format of the questions did not lend itself to

Language and Reasoning 11 mathematical association and thinking, and was consequently treated by subjects as equally abstract and unrelatable as the letter problems. The results from this study indicate that logical proble ms in narrative format are easier for subjects to reason from, as well as that the increased conceptuality of a problem, including a concept-based language, increases reasoning accuracy and the formation of mental models. However, the education level of the subjects is highly controlled. Further studies on other ideographic languages would further the theory that increased conceptualization eases reasoning. The manipulation of subjunctive and indicative tense for English problems, when comparing with ideographic language performance, would also be enlightening, as one of the issues in an ideographic language is the strength of the counterfactual indicator. Measure of the type of answers most often correctly chosen by English and ideographic subjects might prove interesting, as the reliance in ideographic languages on time-framing may ease the inference of correct answers not often inferred in English studies (such as affirmation of the consequent), or even falsities, in contrast to the study done by Laird et al.

Language and Reasoning 12 References Feng, G., & Yi, L. (2006). What if Chinese had linguistic markers for counterfactual conditionals? Language and thought revisited. Proceedings of the 28th Annual Conference of the Cognitive Science Society, 1281-1286. Landy, D., Goldstone, R. L. (2007). How abstract is symbolic thought? Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, 4, 720-733. Liang, B., Cherian, J., Liu, Y. (2010). Concrete thinking or ideographic language: which is the reason for Chinese peoples higher imagery-generation abilities? International Journal of Consumer Studies, 34, 52-60. Johnson-Laird, P. N., Legrenzi, P., Girotto, V., Legrenzi, M. S. (2000). Illusions in reasoning about consistency. Science, 288, 531-532. Santamaria, C. & Espino, O. & Byrne, R. (2005). Counterfactual and semifactual conditionals prime alternative possibilities. Journal of Experimental Psychology: Learning, Memory, and Cognition,31, 5, 1149-1154. Wason, P. C., Shapiro, D. (1971). Natural and contrived expe rience in a reasoning problem. Quarterly Journal of Experimental Psychology, 23, 63-71. Yeh, D., & Gentner, D. (2005). Reasoning counterfactually in Chinese: Picking up the pieces. Proceedings of the twenty-seventh Annual Meeting of the Cognitive Science Society, 2410-2415.

Language and Reasoning 13 Appendix Example problem, letter trial: If I had x, then I would have y. Which answers are consistent with the statement above? 1. If I have x, I have y. 2. If I do not have y, I do not have x. 3. If I have x, I do not have y. 4. If I do not have x, I do not have y. x, y

1. xy 2. yx 3. xy 4. xy

Example questions for each condition: Letter trials: English: If I had s, then I would have m. Chinese: sm Number trials: English: If I had 1, then I would have 7.

Language and Reasoning 14 Chinese: 17 Word trials: English: If I had an apple, then I would have an orange. Chinese:

Language and Reasoning 15 Accuracy


Chinese English Bilingual M F Sig.

Letter Number Word M F Sig.

2.4 2.7 3.8 3.2

2.6 2.6 3.6 3.05

2.5 2.6 3.9 3.2

2.5 2.6

0.384 0.659

.078 .150

3.8 496.382 .003

485.304 596.294 586.294 .002 .004 .001

Figure 1. Accuracy of the three language groups of participants with each problem type. Accuracy for each problem is rated on a scale from one through four, with four being the highest.

Language and Reasoning 16


Problem Type All Languages Letter Number Word Number Letter Word Word Number Letter English Condition Problem Type Letter Number Word Number Letter Word Word Number Letter Language Condition All Problem Types Chinese English Bilingual English Chinese Bilingual Bilingual English Chinese Bilingual Condition Problem Type Letter Number Word Number Letter Word Word Number Letter

M .05 1.3 .05 1.2 1.2 1.3

Sig. .40 .002 .40 .004 .004 .002

M Sig. .15 .05 0 1 .15 .05 .15 .05 .15 .05 0 1 Chinese Condition Problem Type Letter Number Word Number Letter Word Word Number Letter

M 0 1.0 0 1.0 1.0 1.0

Sig. 1 .005 1 .005 .005 .005

M .1 1.3 .1 1.2 1.2 1.3

Sig. .77 .002 .90 .004 .004 .002

M .3 1.4 .3 1.1 1.1 1.4

Sig. .84 .001 .84 .005 .005 .001

Figure 2. Post-hoc tests for significance of languages and problem types.

Language and Reasoning 17

Bilingual Accuracy
A B M F Sig. A B M F Sig.

Letter

2.4 2.5 2.45 .968 .07

Number 2.6 2.5 2.55 .856 .08

Sig.

Word

2.8 3.6 3.2 495.445 .003

Figure 3. Accuracy of bilingual groups. Group A had the logical structure in English, and the subject matter in Chinese. Group B had the logical structure in Chinese, and the subject matter in English. Accuracy is rated on a scale from one through four, with four being the highest.

Vous aimerez peut-être aussi