JONATHAN POOL and BERNARD GROFMAN Linguistic artificiality and cognitive competence WARNING: This file exists for the sole purpose of exposing the content of the associated PDF page-image document (http://utilika.org/pubs/etc/lacc.pdf) to search engines. The text in this file is not entirely accurate and omits or misrepresents formulas, graphs, footnotes, etc. 1. Introduction Proponents of artificial and simplified natural languages for universal communication have faced numerous objections to the feasibility of such "planned" languages. We can fit most of these objections into two categories. The first category includes claims that the public support needed to secure the official or practical adoption of a planned language for common international use is impossible to obtain. The second category includes claims that no planned language could function adequately as the main medium of world-wide communication. In this research note, we are concerned with an objection belonging to the second category: the functional inadequacy of planned languages. The inadequacy of planned languages has been asserted on several grounds. The two most persistently offered grounds appear to be diversity and poverty. By "diversity" we refer to the assertion that the world-wide uniformity required for successful operation of the language would erode with time: the language would split into mutually incomprehensible dialects. By "poverty" we refer to the claim that a planned language would limit the ability of its users to formulate and communicate ideas. It is this second objection to the feasibility of a planned language that we focus on in this research note. We can summarize the "poverty" argument as follows: no person or group of persons knows enough about language to invent a new language (or even greatly simplify an existing language) with confidence that the product will succeed as a comprehensive medium of thought and communication. Natural languages have some known features that cannot be built into planned languages, such as high levels of redundancy, connections between terms and experiences, and meaningful stylistic variations. Moreover, so goes this argument, there 146 Jonathan Pool - Bernard GrofTnan are other essential, but at present unknown, features of languages used for centuries that are missing from planned languages such as Esperanto or Volapiik. Thus, attempts to use a planned language will result in unacceptable levels of unintentional miscommunication, frustrations caused by a perceived inability to communicate key concepts, and/or degradation of the tools of thought available to those who do their thinking in the new language. Arguments to this effect date back to at least 1930 (references in Large 1985: 188-189; Mead-Modley 1967; Orwell 1949; Tirnurtas 1969) but have never been experimentally investigated. Unspecified in the poverty argument are the mechanisms whereby the particular characteristics of planned languages produce the deficiencies in thought and communication. For example, if a planned language has low redundancy, why does this cause misunderstanding? And, if a planned language is grammatically simple, why does this limit the user's reasoning or judgment? There are numerous possible explanations for such purported relationships. We shall not attempt a conceptual or theoretical analysis of the poverty argument here. Rather, we shall ask whether a straightforward method for empirically testing its predictions can be developed. Our work on this problem win be confined to the prediction that the use of planned languages impairs their users' reasoning ability and thereby leads to observable biases in the judgments they make. One approach to gathering evidence on this proposition is to observe the judgments of persons who already speak a planned language. There are "a few" speakers of Volapiik (Blanke 1985: 213), about one-hundred speakers of Occidental (Blanke 1985: 167), about two-hundred speakers of Interlingua (Blanke 1985: 182), about threehundred speakers of Ido (Blanke 1985: 199; Large 1985: 154), and a half-million speakers of Esperanto (Blanke 1985: 289; but see Piron, "Who are the speakers in this volume - ed.). So the most obvious candidate for observation is the community of Esperanto speakers, which is about a thousand times the size of all other planned languages combined. The aggregate behavior of the speakers of Esperanto does not, however, clearly testify to any effects of the language on the reasoning ability of those who learn or use it. Were Esperanto speakers a linguistic cult, showing unanimous blind faith in the infallibility of the movement's leaders or doctrines, this judgmental inflexibility would at least be compatible with the hypothesis that the language impairs the reasoning ability of those who learn it. A degradation of reasoning ability as a result of the knowledge or use of Esperanto Linguistic artificiality and cognitive competence 147 would lead to the prediction that the judgments of fluent speakers would be more biased than the judgments of beginning speakers. The evidence does not clearly fit either of these patterns. Esperanto speakers differ greatly among themselves in political ideology (Blanke 1985: 288; Jordan 1987: 111-114), and their political opinions are generally not deviant (Forster 1982: 320-326). Impressionistic observation of the Esperanto movement reveals considerable factionalism, as well as acrimonious debate about principles, strategies, tactics, and linguistic theory. On the basis of what little evidence exists, it appears that Esperanto speakers are not, on the whole, unusually dogmatic, gullible, or illogical. Furthermore, no tendency for students of Esperanto to become less "rational" as they learn more of the language has been reported. If the aggregate behavior of the speakers of the most frequently used planned language does not provide convincing evidence that use of Esperanto affects its users' reasoning ability, this is no surprise. "Although many have tried to do so, no one has successfully predicted and demonstrated a cognitive difference between two populations on the basis of the grammatical or other structural differences between their languages" (Fishman 1970: 92-94). Not finding such evidence is not the same as disproving the proposition, however. If previous experience is a guide (Fishman 1970: 97), the best evidence is likely to come from an experiment that controls for additional variables that are expected to influence subjects' reasoning ability. Any experimental comparison between persons who speak and persons who do not speak Esperanto would be compromised not only by the many incidental differences that can be assumed to characterize any two speech communities, but also by the self-selection that helps define the membership of the Esperanto (as of any second-language) community. The main reward from the experimental method in this study is the ability to control for such differences. To achieve such control, we can measure the behavior of artificial-natural bilinguals, i.e., persons who are competent in both an artificial language (here Esperanto) and a natural language. While such a design sacrifices the opportunity to compare unilinguals, this sacrifice is not as serious in the study of planned languages as it would be in comparing, say, English and Zuni, since unilingual speakers of planned languages are nonexistent. With this strategy in mind, a brief experiment was conducted at a site where numerous speakers of Esperanto could be found. In light of the poor regulation of the site conditions by the experimenters and the 148 Jonathan Pool - Bernard Grofinan impossibility of conducting a pretest to refine the stimuli, this experiment should be considered purely exploratory. 2. Subjects The subjects in our field experiment were 135 persons attending the 69th World Esperanto Congress, held in July, 1984, in Vancouver at the University of British Columbia. Persons attending the congress were approached unsysternatically in public places and asked (in Esperanto) whether they knew how to read both Esperanto and English. If so, they were invited to fill out an an anonymous questionnaire on the spot and return it immediately to the investigator. 3. Method The questionnaire contained two reasoning problems and a question on whether the respondent could more easily read Esperanto or English. The first item on the questionnaire was a test of deductive reasoning. Half the subjects received the item in English and half received it in Esperanto, the language being determined by randomization. The two versions read: English: Every friend of a friend of Country A is a friend of Country A. Every friend of an enemy of Country A is an enemy of Country A. Every enemy of a friend of Country A is an enemy of Country A. Country B is an enemy of a friend of an enemy of Country A. Therefore (choose only the answer that logically follows): [ I Country B is a friend of Country A. [ ] Country B is an enemy of Country A. [ I It is not certain whether Country B is a friend or an enemy of Country A. Esperanto: 6u arniko de amiko de Lando A estas amiko de Lando A. tiu arniko de malarniko de Lando A estas malamiko de Lando A. tiu malamiko de amiko de Lando A estas malamiko de Lando A. Lando B estas malarniko de amiko de malarniko de Lando A. Do (elektu nur la logike sekvan respondon): [ ] Lando B estas amiko de Lando A. [ ] Lando B estas malamiko de Lando A. [ ] Ne certas eu Lando B estas amiko A malarniko de Lando A. Linguistic artificiality and cognitive competence 149 The second item on the questionnaire was a test of consistency of judgment under uncertainty. Subjects receiving the first item in English received the second one in Esperanto and vice versa. This question, slightly adapted from one used by Tversky-Kahneman (1981: 453), began as follows: English: It is expected that a certain disease will kill 600 persons. The ministry of health is considering two methods for opposing this disease. which method do you prefer.? Esperanto: Oni atendas ke certa malsano mortigos 600 homojn. La ministerio pri saneco konsideras du metodojn por kontra0i tiun malsanon. Kiun metodon A preferas? The remainder of the question (always in the same language) took one of two forms, randomly selected: English, form 1: ] Method A: it would save 200 persons. I Method B: the chances are 1/3 that it would save 600 persons and 2/3 that it would save no one. English, form 2: Method A: 400 persons would die. Method B: the chances are 1/3 that no one would die and 2/3 that 600 persons would die. Esperanto, form 1: Metodo A: oni savus 200 homojn. Metodo B: la �ancoj estas 1/3 ke oni savus 600 homojn kaj 2/3 ke oni savus neniun. Esperanto, form 2: [ ] Metodo A: 400 homoj mortus. [ ] Metodo B: la gancoj estas 1/3 ke neniu mortus kaj 2/3 ke 600 homoj mortus. 4. Hypotheses 'Me first item on the questionnaire is a question with three possible answers, one of which is correct and two of which are incorrect. The fourth premise in this item implies, by virtue of the second premise, that Country B is an enemy of an enemy of Country A. No premise, however, implies anything about an enemy of an enemy of a country. To illustrate, suppose that X is an enemy of A and that B is an enemy of X. B is thus an enemy of an enemy of A. Given this information, B might be a friend of A, an enemy of A, or neither a friend nor an 150 Jonathan Pool - Bernard Grofman enemy of A. Even if a subject understood the text to imply that the relations of "friend" and "enemy" are symmetric (B is to A as A is to B) or exhaustive (A must be either a friend or an enemy of B), the correct conclusion would not change. Thus the third option ("It is not certain...") is the only correct answer. The prediction being investigated is that correct answers to this item win be more frequent among subjects reading the item in English than among subjects reading it in Esperanto. The second item requires the subject to express a preference between two policies with uncertain outcomes. Neither policy can be described as the correct one. However, the item was presented in two forms with the same implications for the numbers of deaths that each policy could cause and the probabilities thereof. It seems reasonable to assume that a person manifesting high reasoning ability would answer this question identically regardless of the form in which it is couched. A person who is reasoning poorly would, on the contrary, be likely to give an answer influenced by the differential presentation of the question in the two forms. In particular, form I tends to bias subjects in favor of Method A, and form 2 in favor of Method B. As argued by Tversky-Kahneman (1981), these biases result from the different "anchorings" suggested by the two forms. Form I suggests that 600 deaths is the status quo, while form 2 implies that 0 deaths is the status quo. Form 1 thus asks the subject to compare a certain gain with an uncertain gain. Since the value of an uncertain gain tends to be underweighted, this form biases the subject in favor of the certain gain. Form 2 asks the subject to compare a certain loss with an uncertain loss. Since an uncertain loss tends to be underweighted Oust as an uncertain gain is), this form biases the subject in favor of the uncertain loss. In light of previous research, we should expect the majority of subjects given form I to choose Method A and the majority of subjects given form 2 to choose Method B. Failing that, we should expect that the proportion choosing A will be greater with form 1 than with form 2. This should be true regardless of the language in which the item is administered. Further-more, the plannedlanguage poverty argument predicts that the difference between the proportion choosing Method A in forms I and 2, i.e., the extent of the presentation-induced bias, win be greater among subjects reading the item in Esperanto than among those reading it in English. Linguistic artificiality and cognitive competence 151 Several characteristics of the sampled population could be expected to interact with the language variable in strengthening or weakening the hypothesized effects. Most prominent among these is the subjects' competence in the two languages. Subjects more competent in either language should be expected to show greater reasoning ability when that language is used in formulating a problem. Thus we can offer a "strong" version of each of the above hypotheses. In their strong versions, the hypotheses predict differences that will hold true even among subjects who are more competent in Esperanto than in English. 5. Results The results for the first item are shown in Table 1. Table 1. Language and deductive reasoning Respondent better at English? Yes No All Language of question: Eng. Esp. Eng. Esp. Eng. Esp. Response (per cent): "friend" (incorrect) 18 32 21 39 19 35 "enemy" (incorrect) 9 19 12 13 10 16 "not certain" (correct) 73 49 68 48 70 49 (N) (33) (37) (34) (31) (67) (68) X 2 (incorrects pooled): 4.2 2.5 6.5 Significance: p < .05 p < .2 p < .02 Consistent with the hypothesis, the proportion of correct answers was higher among those receiving the question in English than among those getting it in Esperanto. As further predicted, the advantage held by those receiving the question in English was greater among those who claimed to read English more easily than Esperanto, but it was present even among those who claimed to read Esperanto at least as easily as English, thus supporting the "strong" version of the hypothesis. Given the number of respondents, however, we cannot place much confidence in the reliability of this latter result, since a difference between the English and Esperanto versions at least this great would occur by pure chance in from 10% to 20% of random samples of this size if there were no difference in the population as a whole. Also, as we shall discuss below, respondents may have tended to overestimate their Esperanto competence relative to their English (N) Signific responses being Method A and all form-2 responses being Method B, ifferences are small enough to occur often by chaj es drawn from a population exhibiting no reasoning ability in comparison with use of a natural language. A test of deductive reasoning was answered more successfully in English than in Esperanto. However, a test of resistance to presentationinduced bias was passed with greater success by those reading the problem in Esperanto than by those reading it in English. contradiction? One alish Then to overreport their relative E be exaggera speranto competen more than self-reported English y by reading the ques not easily undersumd.t die I stimulus would result in equally distributed answe leading to a finding of no presentati s the obvious method rs on the on Dias at an. Us attention to the epistemological ssessing compe reported Esperanto competence to compe nce in th omnrehension of the