项目说明

本文的研究对象，（即源语语篇“《波普尔的证伪探试启发法是否有利于发展批判性思维？》”），选自马克·梅森编著的论文集《批判性思维与学习》（~~Chi-Ming~~ Lam，2008）。这是一篇学术论文，探讨了波普尔的证伪探试启发法是否能应用于教学，以来发展学生的批判性思维。文章的观点新颖，内容很合且有道理，笔者对此文很感兴趣，因此选择将它作为本论文翻译部分的源语语篇。

原文属非文学文本，全文以逻辑论证的方式展开~~阐述，属于非文学文本~~。~~因此翻译过程中~~译者须完成的最主要的任务是准确理解原文的逻辑信息并及按照汉语叙事与语篇规范传递这些信息~~逻辑进行表达~~。

在完成此任务~~翻译这篇文章~~的过程中，译者遇到了许多长句，这些句子的组成成分较多，结构复杂，~~成分较多，~~对理解原文理解和表达译文表达都构造成了一定的困难。~~，因此~~如何解决这些困难是翻译项目实施过程中译者面临的最大问题。

译者（本文作者）用以解决此问题的理论与方法是美国著名语言学家和翻译理论家尤金·奈达（Eugene Nida）的核心句理论和核心句分析法。~~奈达的~~核心句理论认为：，每一个表层句法上复杂的句子均由底层结构中数量有限、被称为“核心句”（kernel 或 kernel sentence）的基本结构成分转换而来；找出原文的核心句才是准确理解原文及表达译文的根本途捷径。，它核心句分析法强调主张通过分析核心句及核心句之间的逻辑关系，来把握原~~分析法分析~~句子的底表层结构，弄清句子的主要信息。这种方法正好可以很好地帮助译者解决英语长句翻译中遇到的困难。本文在借鉴~~介绍了~~其他译者利用核心句分析法来翻译长难句~~的策略，在借鉴前人~~经验的基础上，着重~~本文将从准确理解原文及合理表达译文两方面~~探讨了本人用在核心句分析法解决翻译项目中准确理解原文及合理表达译文方面的问题的过程~~指导下的英语长句汉译策略~~。

本文共分为三部分，第一部分为项目说明，第二部分为源语/译语对照语篇，第三部分为翻译评注。

源语/译语对照语篇

Note (KP):

(1) To align globally the original and the translation over the left and right columns in a table, edit your Word document in “Web” instead of “Page” or “Normal” viewing mode; set the display zoom level at 150%; pull down the dialog box “Unit of Measure” in “Table > Attribute > Table” and select “Percentage”, tick the box “Specified Width” and type in the dialog box “100%”; click on “Options” in “Table > Attribute > Table” and tick the box “Automatically Readjust Size to Suit the Content”; patiently adjust the width of the left and right columns so that the original and the translation are best aligned with each other. (The most appropriate width of the left column as displayed in the following table, e.g. is found ― after repeated trials ― to be approximately 7 cm for a Word document [.doc] and 7.75 cm for a web page [.htm].) {* These operations will also ensure that the table you created in MS Word will be automatically extended across the full page width when you save the Word document as a web page and view it in a web browser.}

(2) To tweak the alignment of specific paragraphs in the left and right columns, you may increase the line space before a blank line before a paragraph which you need to align with a corresponding paragraph in the adjacent column [Examples: 1, 2]. {Such tweaks, however, should be used only sporadically.}

Is Popper’s Falsificationist Heuristic a Helpful Resource for Developing Critical Thinking?

CHI-MING LAM

The University of Hong Kong

Three Core Concepts of Critical Rationalism

Formulated fundamentally by Popper as an attitude of admitting that ‘I may be wrong and you may be right, and by an effort, we may get nearer to the truth’ (1966, p. 225), critical rationalism is an attitude of readiness to listen to critical arguments and to learn from our mistakes. Near the end of his life, Popper revealed that he owes the idea of this formulation to what a young Carinthian member of the National Socialist Party, not long before the year in which Hitler came to power in Germany (1933), said to him: ‘What, you want to argue? I don’t argue: I shoot!’ (1996, p. xiii). The young man’s readiness to shoot rather than to argue may indeed have planted the seeds of three core concepts of Popper’s critical rationalism, viz. fallibilism (‘I may be wrong’), criticism (the required ‘effort’), and verisimilitude (‘we may get nearer to the truth’).

By fallibilism Popper (1966) intends the view that we are fallible and that the quest for certainty is mistaken. While the former view can be substantiated historically by the fact that what we once thought to be well-established may later turn out to be false, the latter can be understood theoretically by the problem that what we can explain or know is limited. One such limitation concerns the power of our brain to explain: according to Hayek (1952), any apparatus of classification must possess a structure of a higher degree of complexity than that possessed by the objects which it classifies; it implies that no explaining agent can ever explain objects of its own kind or own degree of complexity, and thus that the human brain can never fully explain its own operations. Another limitation arises from our inability to predict the future course of history, not least because of our inability to predict the future growth of human knowledge: as Popper puts it, ‘if there is such a thing as growing human knowledge, then we cannot anticipate today what we shall know only tomorrow’ (2002a, p. xii). Accordingly, his fallibilism denies the possibility of certain knowledge and of authoritative sources of knowledge. Instead, he asserts that nothing is secure and that our knowledge remains conjectural and fallible.

However, because we can learn from our mistakes, fallibilism need not cause any keptical or relativist conclusions. And criticism, he claims, ‘is the only way we have of detecting our mistakes, and of learning from them in a systematic way’ (1966, p. 376). It includes criticizing the theories or conjectures not only of others but also of our own. Since, for Popper (1989), criticism invariably consists in pointing out some contradiction (within the theory criticized, or between the theory and another theory which we have some reason to accept, or between the theory and certain statements of facts), deductive logical reasoning is suggested as the method of criticism: only by purely deductive reasoning can we discover what our theories imply, and thus where contradictions lie. More specifically, the importance of deductive or formal logic to criticism lies in the fact that it adopts the rules by which truth is transmitted from premises to conclusions while falsity is re-transmitted from conclusions to premises. It is this re-transmission of falsity that ‘makes formal logic the Organon of rational criticism—that is, of refutation’ (ibid., p. 64). In fact, rejecting all attempts at the justification of theories, Popper replaces justification with criticism in his non-justificationist or falsificationist view of rationality: ‘Previously, most philosophers had thought that any claim to rationality meant rational justification (of one’s beliefs); my thesis was, at least since my Open Society, that rationality meant rational criticism (of one’s own theory and of competing theories)’ (2002b, p. 173). However, considering a theory may stand up to criticism better than its competitors, he concedes that we can sometimes ‘justify’ our preference for a theory in the negative sense that a theory receives some kind of support if it has, rather than secured positive evidence, withstood severe criticism.

The idea of getting nearer to the truth or achieving greater verisimilitude is crucial to Popper’s concept of critical rationalism, for it is only the idea of truth that allows us to speak sensibly of fallibilism and criticism: the purpose of searching for mistakes and eliminating as many of them as we possibly can through critical discussion is to get nearer to the truth. Criticizing subjective theories of truth for conceiving truth as something we are justified in believing or in accepting in accordance with some criterion of well-foundedness, Popper (1989) adopts Tarski’s correspondence theory of objective truth that a statement is true if and only if it corresponds to the facts. For one thing, Tarski’s objective theory of truth allows us to make certain assertions that appear obviously correct to Popper but selfcontradictory within those subjective theories of truth: for example, a theory may be true even if nobody believes it, and even if we have no reason to think it true; another theory may be false even if we have comparatively good reasons for accepting it; we search for truth, but may not know when we have found it; and we have no criterion of truth, but are guided by the idea of truth as a regulative principle. To allay suspicions about the idea of getting nearer to the truth, or of the growth of knowledge, Popper (1979) introduces a logical idea of verisimilitude by combining two notions from Tarski, viz. truth and content. Defining the class of all true statements and false statements following from a statement p as the truth content and falsity content of p respectively, Popper explain that:

Intuitively speaking, a theory T₁ has less verisimilitude than a theory T₂ if and only if (a) their truth contents and falsity contents (or their measures) are comparable, and either (b) the truth content, but not the falsity content, of T₁ is smaller than that of T₂, or else (c) the truth content of T₁ is not greater than that of T₂, but its falsity content is greater. (Ibid., p. 52)

He accordingly regards the search for verisimilitude rather than truth as a more realistic aim of science in that while we can never have sufficiently good arguments for claiming that we have actually attained the truth, we can have reasonably good arguments for claiming that we may have made progress towards the truth (i.e. That the theory T₂ is nearer to the truth and thus preferable to its predecessor T₁).

Stratagems Opposed to Criticism

Yet, to put such a falsificationist theory into practice, it is necessary to identify and combat a nest of philosophical presuppositions that work against criticism and help to confine individuals to the justificationist framework. As the Chinese proverb cautions, ‘It is easy to dodge an open spear thrust but difficult to guard against an arrow shot from behind’, one is unlikely to circumvent or eliminate the effects of these anti-criticism presuppositions unless various hidden stratagems that reduce and eschew criticism are themselves exposed to criticism. Popper, as an advocate of falsificationism, spares no pains to reveal such protective or evasive stratagems. To begin with, he (1989) points out that the doctrine that truth is manifest runs counter to the doctrine of fallibility and thus of tolerance: if truth were manifest, we would be unlikely to make mistakes, and thus would not need to tolerate or pardon others for their mistakes committed as a result of their prejudices. Since criticism involves searching for errors of our own and of others, which assumes that we are prone to errors and consequently should be tolerant of others, the doctrine that truth is manifest is diametrically opposed to it. Another stratagem Popper combats is the demand for precision in concepts as a prerequisite for criticism or problem-solving. Affirming the non-existence of ‘precise’ concepts, or concepts with ‘sharp boundary lines’, Popper (ibid.) emphasizes that words are significant only as tools for formulating theories and don’t need to be more precise than our problems demand. To deal with the problem that our problems may sometimes demand that we make new distinctions for the sake of clarity or precision, he suggests an ad hoc approach:

If because of lack of clarity a misunderstanding arises, do not try to lay new and more solid foundations on which to build a more precise ‘conceptual framework’, but reformulate your formulations ad hoc, with a view to avoiding those misunderstandings which have arisen or which you can foresee. And always remember that it is impossible to speak in such a way that you cannot be misunderstood: there will always be some who misunderstand you. (Popper, 2002b, p. 29)

Popper identifies further three approaches that work against criticism, namely essentialism, instrumentalism, and conventionalism. Essentialism assumes that science aims at ultimate explanations that describe the ‘essences’ of things—the realities that lie behind appearances—and therefore are neither in need nor susceptible of further explanation. Popper (1989) criticizes essentialism as obscurantist in the sense that it prevents fruitful questions or further criticisms from being raised. He (ibid.) also condemns as obscurantist the instrumentalist view of theories as mere instruments for prediction, because it stresses application but neglects falsification or criticism: for instrumental purposes of practical application, a theory may continue to be used within the limits of its applicability even after its refutation—in other words, a theory cannot be falsified insofar as it is interpreted as a simple instrument, for it can always be said that different theories have different ranges of application. And with respect to conventionalist philosophy, which regards laws of nature as our own creations and arbitrary conventions rather than representations of nature, although Popper admits that it deserves credit for clarifying the relations between theory and experiment, or rather for recognizing ‘the importance … of the part played by our actions and operations, planned in accordance with conventions and deductive reasoning, in conducting and interpreting our scientific experiments’ (1980, p. 80), he rejects its methods of protecting the theoretical systems of the natural sciences against criticism. He asserts that there are at least four conventionalist stratagems—introducing ad hoc hypotheses, modifying ostensive definitions, adopting a skeptical attitude to the reliability of the experimenter, and casting doubt on the acumen of the theoretician—which make it impossible to falsify these systems.

A Bias Towards Confirmation

The Pervasiveness and Various Guises of the Confirmation Bias

Apart from exposing to criticism the various hidden stratagems that work against it, it is also important to combat what appears a common psychological tendency of humans to be biased towards confirmation, or against disconfirmation, a tendency that reflects a conflict between falsificationism and apparently deep-rooted psychological mechanisms. Unfortunately, Popper did not give much attention to this. According to Nickerson (1998), confirmation bias connotes an unwitting process of seeking or interpreting evidence in ways that are partial to existing beliefs or hypotheses. A great deal of empirical evidence supports the view that the confirmation bias not only is extensive and strong but also appears in various guises: reflected in the tendency of people, for example, to demand less hypothesis-consistent evidence for accepting a hypothesis than hypothesis-inconsistent information for rejecting a hypothesis (Pyszczynski & Greenberg, 1987); to recall or produce reasons supporting the side they favour rather than the other side on a controversial issue (Baron, 1995); and, when assessing the validity of a conditional ‘if p then q’, to seek for the presence of p and q so as to confirm the conditional rather than for the presence of p and not-q so as to disconfirm the conditional (Wason, 1966).

Although Polya (1954) has argued that what distinguishes scientists from ordinary people is their disposition to seek disconfirmatory evidence for their hypotheses, instances of confirmation bias still abound in the history of science. This can be illustrated at two—personal and institutional—levels. At the personal level, Michael Faraday advocated ignoring disconfirmatory evidence when dealing with a novel hypothesis until the hypothesis was well-confirmed (Tweney, 1989), while Robert Millikan reported only those observations that fitted his hypothesis when publishing experimental work on determining the electric charge of a single electron (Henrion & Fischhoff, 1986). At the institutional level, just as Newton’s concept of universal gravity was rejected by Huygens and Leibniz due to their resistance to the idea of a force not reducible to matter and motion extending throughout space, scientific discoveries have often met with resistance from scientists themselves, especially from those whose theoretical positions were challenged by the discoveries. The typical reaction of scientists to the challenge of anomalous data to an existing theory is in fact to challenge the data first and, if the data prove reliable, then to complicate the theory just enough to accommodate the anomalous result (Nickerson, 1998). Perhaps Polya’s characterization of individual scientists as being inclined to disconfirm their own hypotheses is half correct at most: they appear eager to criticize or disconfirm other scientists’ hypotheses rather than theirs.

Theoretical Explanations for the Confirmation Bias

With regard to the question of how to account for the confirmation bias, apart from what Matlin and Stang (1978) dub the Pollyanna principle, which explains in a commonsensical way that people tend to be partial towards pleasant thoughts and memories rather than unpleasant ones and thus to believe propositions they would like to be true rather than those they would prefer to be false, there are at least four theoretical explanations that various researchers have proposed. First, according to Nickerson, people are basically limited to consideration of only one thing—and inclined to gather information about only one hypothesis—at a time. However, restricting attention to a single hypothesis might strengthen that hypothesis even if it is false:

An incorrect hypothesis can be sufficiently close to being correct that it receives a considerable amount of positive reinforcement, which may be taken as further evidence of the correctness of the hypothesis in hand and inhibit continued search for an alternative. (Nickerson, 1998, p. 198)

Hence the confirmation bias.

Second, discounting the possibility that people seek deliberately to confirm rather than disconfirm their hypotheses, Evans (1989) attributes the confirmation bias not to their motivation to confirm but to their failure to think in explicitly disconfirmatory terms. His argument accords with much evidence that people find it more difficult to deal with negative than positive information. For example, it is more difficult to decide the truth or falsity of negative sentences than of positive ones (Wason, 1961); and inferences from negative premises need more time to evaluate and are more likely to be evaluated wrongly than those from positive premises (Fodor, Fodor, & Garrett, 1975).

Third, just as Friedrich asserts that ‘our inference processes are first and foremost pragmatic, survival mechanisms and only secondarily truth detection strategies’ (1993, p. 298), the judgements people make in many real-life situations are motivated more by a desire to achieve success and survival—and thus to balance potential rewards against perceived risks—than by the objective of determining the truth or falsity of hypotheses. This explains why confirmation bias may result when the undesirable consequences of considering a true hypothesis as false are greater than those of considering a false hypothesis as true.

Last, stressing the importance of being able to justify what one believes at all levels of education can establish or strengthen a tendency to seek confirmatory evidence selectively: if one is always stimulated to adduce reasons for opinions that one holds and is not urged also to articulate reasons that could be given against them, one is being trained to exercise a confirmation bias (Nickerson, 1998). To make matters worse, some educational practices fail to distinguish explicitly between case-building (i.e. seeking selectively or giving undue weight to evidence that supports one’s position while neglecting to seek or discounting evidence that would tell against it) and evidence-weighing (i.e. seeking evidence on all sides and evaluating it as objectively as one can) so that what is in reality case-building passes for the impartial evaluation of evidence: hence the ubiquity and strength of the confirmation bias among educated adults. A typical example of such case-building educational practices is debate, in which debaters give their primary attention to arguments that support the positions they are defending—even if they might advance potential counter-arguments, their intention is only to reveal the shortcomings of these counter-arguments. After all, debaters aim to win, and the way to do so is to make the strongest possible case for their own position while countering, discounting, or simply ignoring any evidence that might be brought against it.

The Teacher’s Role in Undermining the Strength and Spread of the Confirmation Bias

Although it can be argued that the confirmation bias helps both to protect our sense of self by rendering our preferred beliefs less vulnerable than they otherwise would be (Greenwald, 1980) and to guard science against indiscriminate acceptance of alleged new discoveries that fail to stand the test of time (Price, 1963), the bias is still generally regarded as a human failing: it can contribute to the formation of various delusions, the development and survival of superstitions, and the perpetuation of hostility and strife between people with conflicting views of the world (Nickerson, 1998). It is probably a good idea to start with the education of children if the strength and spread of the confirmation bias are to be undermined and checked. What then are the implications for educational practice? First, teachers themselves should be aware of the confirmation bias—its pervasiveness and the various guises in which it appears. Such awareness could help students to be more cautious in making decisions about important issues and more open to opinions that differ from their own.

Considering, moreover, that the confirmation bias is partly attributed to the tendency of people to consider only one hypothesis at a time, teachers should encourage their students to think of several alternative hypotheses simultaneously in attempting to explain a phenomenon. The discovery by Tweney et al. that individuals seldom employ this thinking strategy successfully—for they prefer ‘to evaluate several pieces of data against a single hypothesis, rather than one datum against several hypotheses’ (1980, p. 119)—demonstrates the superiority of working in groups in learning to avoid the bias: having each individual work on a different hypothesis, groups can keep track of several hypotheses at the same time.

Teachers should also realize the significance of making explicit the distinction between case-building and evidence-weighing, and encourage their students to evaluate evidence objectively in the formation and evaluation of hypotheses. Here it is vital to cultivate in students a critical mindset that prompts them to think of reasons both for and (especially) against any judgement that is to be made. And they should be made aware that the motivation to find support for preferred beliefs ‘often leads a person to overlook even glaring faults in the data, because it is difficult to find what is not sought’ (Dawson, Gilovich, & Regan, 2002, p. 1386).

Despite the inclination of scientists to discount data inconsistent with their theory, Fugelsang et al. (2004) found that scientists began to modify their original theory when repeated observations of inconsistent data occurred. Indeed, the initial reluctance of scientists to accept inconsistent data and their subsequent re-theorization through repeated experimentation can be considered as a practical heuristic device: it prevents them from prematurely accepting findings that may be spurious while permitting the revision of theories and thus the growth of knowledge. In the realm of science teaching, this heuristic device should be introduced to students, particularly for fostering an appropriate attitude towards inconsistent data.

Can Students Be Taught to Falsify?

Influential as Popper is in the philosophy and practice of science, a question can still be raised about the effectiveness of his methodology, for there is much controversy in the psychological literature over the feasibility and utility of falsification as a strategy for solving scientific problems. To start with, many psychological studies show that many scientists have difficulty in disconfirmatory reasoning. For instance, in a survey conducted by Mahoney and Kimper (1976), a sample of physicists, biologists, sociologists and psychologists were asked to rate the validity of four forms of material implication (i.e. to judge whether it is valid, assuming that p materially implies q, to infer q from p, not-q from not-p, p from q , and not-p from, not-q ) and to identify the logically critical experiments that could test the validity of a hypothesis of the form ‘if p then q’. It was found that over half of these scientists failed to recognize modus tollens (i.e. the inference from not-q to not-p) as logically valid, and that fewer than 10% of them were able to select correctly the experiments that had the critical potential of falsifying the sample hypothesis. Perhaps more surprisingly, similar difficulty in recognizing the logical validity of falsification was found in a sample of statisticians who had been formally trained in testing statistical (null) hypotheses and thus in examining possible disconfirming evidence (Einhorn & Hogarth, 1978).

However, having difficulty in using disconfirmatory reasoning does not mean a lack of ability to do so. Indeed, some researchers have successfully taught college students to employ disconfirmatory strategies to solve such reasoning problems as Wason’s (1960) 2-4-6 problem and Gardner’s (1977) ‘New Eleusis’. Wason advanced the 2-4-6 problem as a test of inductive reasoning: subjects were told that the sequence of three numbers ‘2-4-6’ was an instance of a rule that the experimenter had in mind (the rule was ‘any three numbers in ascending order’); they were required to discover the rule by generating their own test sequences of three numbers which the experimenter would describe to them as correct or incorrect instances of the rule. Considering Wason’s subjects displayed a strong confirmation bias and tended to generate test sequences consistent with their tentative hypotheses, Tweney et al. (1980), using the same 2-4-6 task, made an attempt to teach disconfirmatory strategies to their subjects, that is, to ask their subjects to try generating disconfirmatory instances. They found that the mean number of confirmatory and disconfirmatory instances generated by subjects in the disconfirmatory group was 1.5 and 6.6 respectively (in Experiment 1). This indicates that Tweney et al. were successful in eliminating most attempts at confirmation and thus in changing the inquiry strategy of those subjects in the disconfirmatory group.

‘New Eleusis’ is a card game designed to simulate the inductive search for truth. Gorman, Gorman, Latta, and Cunningham (1984) adapted it to create a task for studying scientific reasoning: subjects were asked to guess what the underlying rule behind a sequence of cards was by playing cards one at a time (one of the rules, for instance, was ‘a difference of 1 must separate adjacent cards’); they would be informed by the experimenter whether their cards were right or wrong but would, not receive any feedback from the experimenter on whether their guesses were right or wrong until the end of the experiment. Using this task to study how confirmatory, disconfirmatory, and combined strategies affected group problem solving (in Experiment 2), they instructed their subjects to concentrate respectively on getting as many cards right as possible, on getting as many cards wrong as possible, and on getting cards right until they had a guess which was then tested by playing cards that would be wrong. They found that disconfirmatory groups played incorrect cards 41% of the time, combined groups 33% of the time, and confirmatory groups only 20% of the time. Again, the result shows that the instructional manipulation was successful; hence the feasibility of inducing the use of disconfirmation.

Two Contributory Factors in Eliciting Disconfirmation

Here, two contributory factors in the higher use of disconfirmation—collaborative reasoning and lower normativity—need attention if disconfirmatory strategies are to be promoted in the classroom. To illustrate how group processes often facilitated disconfirmation, Gorman et al. (1984, p. 75) provided the following brief exchange between two subjects in one disconfirmatory group:

One subject complained to the other group members: ‘I have a hard time guessing wrong’. Another subject tried to tell her how to disconfirm: ‘If you think the series goes like this (pointing to a sequence of cards ascending by ones), try to prove it wrong by putting down a card that doesn’t go with the series’.

The second subject soon induced not only the first subject but also other group members to falsify more and more guesses. Such beneficial effects of peer interaction are echoed in the study of Moshman and Geil (1998), who found that while 75% of the subjects in interactive groups could apply correctly a disconfirmatory strategy in testing a hypothesis, only 9% of the individual subjects working in isolation could do so. As close examination of the videotaped group discussions revealed little evidence of passive conformity to majority views or to the views of an apparent expert but a usual attempt to co-construct a consensus solution—a structure of arguments qualitatively more sophisticated than that generated by most individuals—by sharing perspectives and reasons, they attributed the superior performance of the groups to collaborative reasoning rather than to peer pressure or imitation. Insight into the logic of falsification appears to be more readily achieved in collaborative reasoning than in individual reasoning.

The question whether disconfirmation is used during collaborative hypothesis testing might, however, depend upon the type of relation reasoners have with their partners and with the experimenter: the study of Butera, Caverni and Rossi (2005) showed that while confrontation with a low-competence partner rendered subjects able to learn to use disconfirmation, confrontation with a high-competence partner induced them to use confirmation, even when the partner used disconfirmation. A possible explanation is that confrontation with a high-competence partner could threaten subjects’ sense of competence, thereby leading them to test their own hypotheses through confirmation as a defensive strategy that seems to support their hypotheses and thus to protect their competence; in contrast, a low-competence partner is less likely to threaten the subjects’ sense of competence, thereby allowing them the opportunity to test the limits of the validity of their own hypotheses through disconfirmation (ibid.). Moreover, Butera et al. (2005) showed that subjects who were confronted with the violation of a conversational rule—i.e. were told by the experimenter in solving Wason’s (1960) 2-4-6 problem that 2-4-6 was not a good example of the rule and had been chosen only to show them what was a number triad—used a high proportion of disconfirmation, whatever the competence of the partner. They explained that disconfirmation stemmed from the possibility of diverging from not only social norms in the case of interaction and social influence (e.g. the constraining power of competence), but also conversational norms in the case of language (e.g. the constraining power of the example given by the experimenter): considering the high-status experimenter might lead subjects through conversation to focus on the given triad, to formulate a hypothesis that captured all the salient features of the triad, and to try to confirm it, telling them that 2-4-6 was not a good example of the rule might break the focused processing of the task and lead them to use disconfirmation. It appears therefore that people use confirmation in constraining reasoning situations but that ‘the use of disconfirmation can be increased by lowering the normativity of the situation, either by a less threatening source or by less constraining conversational rules’ (ibid., p. 186).

In other words, if disconfirmation is to be taught effectively to students, merely creating the opportunity for them to collaborate with each other is not enough. The teacher should also attempt to lower the normativity of the learning environment by such means as ensuring that students interacting within the group are not threatening or dominating, and avoiding proposing as an authority a model solution to them during problem-solving. The latter is particularly noteworthy in that many teachers really see themselves as an authority in the classroom who, they think, should know the answer to every question. Such an authoritative image teachers have of their role is detrimental to the adoption of disconfirmation in two ways: first, it makes the interaction between the teacher and students more normative; and second, it makes the classroom less likely to satisfy the basic requirement for implementing falsificationism in education, that is, to become a place that values mistakes made by both teachers and students (Sankey, 1999).

The Influence of the Complexity of the Problem

Despite the foregoing evidence in support of the argument that people can be taught to falsify their hypotheses, some studies have shown that instructional manipulations might fail to elicit falsification when the inference problem turns complex. For example, to achieve a more realistic simulation of science in their study, Mynatt, Doherty and Tweney (1977) designed a rather complex inference task: after observing a set of computer displays made up of stationary geometric figures and moving particles whose motion was influenced by the figures, subjects were asked to discover the laws that governed the motion of particles by first generating a hypothesis and then choosing the appropriate experiments to test that hypothesis. They found that their manipulation failed to induce the disconfirmatory group to seek disconfirmation. In a follow-up study, Mynatt, Doherty and Tweney (1978) gave subjects more extensive instructions to falsify and rendered the task even more complex yet more realistic by allowing them to explore it in a less constrained manner (e.g., allowing them to design their own experiments instead of forcing them to choose from the potential ones). As in Mynatt et al. (1977), however, they found that instructions to disconfirm produced little or no effect on the disconfirmatory group; indeed, ‘there was … almost no indication whatsoever that they (both the disconfirmatory and control groups) intentionally sought disconfirmation’ (Mynatt et al., 1978, p. 400).

A possible explanation, they suggest, is that a disconfirmatory strategy might simply overload the cognitive capacity of most people—hence the difficulty in eliciting it from them—when they are groping for a means of dealing with complex inference problems. Accordingly, the feasibility of teaching people to falsify seems to depend on whether or not the task is complex.

To complicate matters still further, sometimes it is difficult to judge from the testing behaviour of people whether they have actually followed the falsificatory instruction, for it can be argued that the falsificatory instruction is not carried out successfully if people who are instructed to falsify perform what Wetherick (1962) calls negative tests—i.e. testing their hypothesis by means of test items that it predicts to be false—but at the same time expect the hypothesis to be confirmed rather than falsified by the test result. This argument is echoed and supported by the study of Poletiek (1996), who found that although 60% of subjects in the falsificatory group adopted negative tests, only 10% of them expected a hypothesis-falsifying result, concluding that:

… when subjects are asked to behave as falsifiers in a hypothesis-testing task, their behaviour expresses the paradoxical character of this requirement by showing a preference for negative tests on the one hand, but nonetheless expecting this strategy to fail with regard to the production of hypothesis-inconsistent data on the other. (Poletiek, 1996, p. 456)

In other words, it seems ‘paradoxical’ to regard those who simultaneously use negative tests and expect confirmation of their hypotheses as following the falsificatory instruction. Leaving aside the problem of how to deal with such paradoxical situations that may arise when people are instructed to falsify, an interesting question is: why don’t they think and act in the same way? That is, why don’t they expect to falsify their hypotheses when performing negative tests? Does it reveal a disbelief at heart in the utility of falsificatory strategies? Do these strategies work in reality?

Should Students Be Taught to Falsify?

Judging from the results of several studies conducted by Michael Gorman and his colleagues in the 1980s, there appear to be grounds for cautious optimism about the utility of falsification. To begin with, in the afore-mentioned study of how different strategies affect the performance of groups in the task adapted from ‘New Eleusis’ (in Experiment 2), Gorman et al. (1984) found that disconfirmatory groups solved significantly more rules (72%) than combined (50%) and confirmatory (25%) groups. Together with the findings that disconfirmatory groups played the highest percentage of incorrect cards (41%) while combined and confirmatory groups played the middle (33%) and the lowest (20%) percentage respectively, and that the percentage of incorrect cards played by these three different groups was highly correlated with their success in solving the rules, this would indicate that the strategy instructions were carried out successfully and indeed accounted for the differences in performance: hence the effectiveness of disconfirmation in problem-solving.

Considering that scientists do not work in error-free environments, in order to model the role of disconfirmation in scientific inference more authentically, Gorman (1986) added the possibility of error to the ‘New Eleusis’ experiment in another study (with a design virtually identical to the preceding study): subjects were told that 0–20% of the feedback on their trials from the experimenter might be in error, that is, a card that should be correct would be classified as incorrect and vice versa. He found that disconfirmatory groups did not perform significantly better than confirmatory or control (i.e. no-strategy) groups, because the possibility of error interfered with the ability of disconfirmatory groups to obtain and use disconfirmatory information in the sense that it allowed them to immunize their hypotheses against disconfirmation by classifying disconfirmatory information as error, and that it made them spend so much time checking potential errors that they failed to test their hypotheses adequately. However, this result does not mean the futility of disconfirmation under possible-error conditions. Given that the few successful groups used a strategy that combined disconfirmation with replication (i.e. replicating situations in which they thought an error might have occurred), it would imply that disconfirmation becomes even more important as a necessary, though not sufficient, strategy (ibid.).

A Favourable Condition for Disconfirmation to Be Effective

Gorman and Gorman (1984) showed further that the positive effect of disconfirmatory instructions found by Gorman et al. (1984) could be replicated on Wason’s (1960) 2-4-6 task with individual subjects. Specifically, they found that a significantly larger number of disconfirmatory subjects (95%) than confirmatory (48%) and control (53%) subjects solved the original rule (i.e., ‘any three numbers in ascending order’) of the task. Curiously enough, such positive effects of disconfirmation on performance did not appear in the afore-mentioned study by Tweney et al. (1980), the Experiment 1 of which used a design very similar to Gorman and Gorman’s and found that ‘while subjects did learn to seek disconfirmatory data, the possession of such strategies led neither to faster solutions, nor to a greater proportion of subjects with correct solutions’ (p. 112). Later on, Gorman and his colleagues discovered that the difference between their results and those of Tweney et al. was probably caused by the fact that their subjects were given no feedback on the correctness of their guesses until the experiment was over, but Tweney et al.’s subjects were informed whether or not each of their guesses about the rule was correct and thus could rely on the experimenter for confirmation or disconfirmation (Gorman, 1992). Therefore, it appears that disconfirmation might be an effective heuristic when people cannot appeal to an outside authority to ascertain whether they are making progress towards a discovery.

Yet, if disconfirmation is less useful when people can appeal to such an authority, then it has little value in the case of laboratory exercises done in many high school and college classes, for, according to Gorman (1995), the objective of most of these exercises is to get the correct answer rather than to explore a novel phenomenon, and frequent appeals to authority in the form of the laboratory assistant or the instructor are not only possible but likely to be helpful. The educational implication is that another, more open-ended and exploratory kind of exercise might provide better training in the use of disconfirmation for future scientists who typically cannot appeal to any authority to test their hypotheses.

The Limits of Disconfirmation

Disconfirmation seems, however, not to be a universally effective strategy for solving reasoning problems. This is substantiated by the results of some 2-4-6 studies (e.g. Gorman & Gorman, 1984; Gorman, Stafford & Gorman, 1987) indicating that it does not work on very general, or more difficult, rules such as ‘no two numbers can be the same’. Disconfirmation seems, moreover, not to be self-sufficient either, because sometimes its utility is dependent upon confirmation in two senses. First, strategically, confirmation acts as a necessary complement to disconfirmation, especially in the early stages of a complex inference task. Here are two illustrative examples: Mynatt, Doherty and Tweney (1978) found that although no subjects solved their demanding task, the most successful one initially concentrated on accumulating confirmatory evidence for his hypothesis without regard to disconfirmatory evidence and only sought to establish whether disconfirmatory instances could be found after a relatively well-confirmed hypothesis had been developed; and Karmiloff-Smith and Inhelder (1975) found that young children presented with difficult reasoning problems were incapable of using disconfirmatory evidence—i.e. recognizing counterexample—until after their hypotheses had been sufficiently confirmed. Echoing the findings of these two studies, Vartanian, Martindale and Kwiatkowski (2003) showed that reliance on a mixed strategy of confirmation and disconfirmation in the early and later stages of hypothesis-testing respectively appeared to be quite advantageous. In fact, in order to defend Faraday against the charge that his deliberate neglect of the disconfirmatory experiments (conducted in 1831 as part of his discovery of electromagnetic induction) reflected a confirmation bias on his part, Tweney and Chitwood (1995) argue instead that what Faraday had done simply manifested a sophisticated use of such a ‘confirm early, disconfirm late’ strategy, and explained in detail that:

Nature is chaotic in its character and will frequently provide false feedback to the inquirer. … [M]any of the experiments tried by Faraday were in fact producing the expected effects but the effects were small and could not be detected with [his] insensitive apparatus. The task of the scientist in such an environment is to impose order on the apparent disorder. … [O]ne of the necessary functions of a confirmation heuristic… [is that] it filters out some of the noise and may allow a signal to be detected. This is not a sure thing, which is why a disconfirmatory strategy is a necessary supplement later on. (Tweney & Chitwood, 1995, p. 255)

Second, essentially, a confirmatory strategy not only does not necessarily contradict the goal of seeking disconfirmation, but may be the only way to achieve it in some circumstances. To understand this, according to Klayman and Ha (1987), a confirmatory strategy is better interpreted as a positive test strategy, which means testing a hypothesis by examining instances where the target property is hypothesized to be present or is known to be present. Further, it is crucial to distinguish between two different senses of ‘seeking disconfirmation’. One sense, which is the focus of empirical investigations, is to examine instances that are predicted not to have the target property, or to conduct negative tests. The other sense, which is emphasized by Popper, is to examine instances that are most expected to falsify the hypothesis. Using Wason’s (1960) 2-4-6 task as an example, Klayman and Ha demonstrate graphically that although a positive test strategy cannot produce falsifications in the Popperian sense when the hypothesized rule (e.g. ‘increasing by 2’) is embedded within the correct rule (e.g. ‘increasing numbers’, as in Wason’s original task), it can do so when the hypothesized rule (e.g. ‘increasing by 2’) overlaps the correct rule (e.g. ‘three even numbers’). More importantly, indeed paradoxically, a positive test strategy is the sole strategy that can reveal conclusive falsifications—even negative tests cannot do so—when the hypothesized rule (e.g. ‘increasing by 2’) surrounds the correct rule (e.g. ‘consecutive even numbers’). We can thus conclude that it is impossible to judge the effectiveness of a confirmatory or positive test strategy in the absence of information about the nature of the task at hand.

Conclusion

To sum up: the implementation of Popper’s falsificationist epistemology means exposing to criticism various philosophical presuppositions that work against criticism, including the doctrine that truth is manifest, the demand for precision in concepts as a prerequisite for criticism, essentialism, instrumentalism, and conventionalism; it also means combating the confirmation bias through such educational means as helping teachers and students to acquire an awareness of its pervasiveness and various guises, teaching them to think of several alternative hypotheses simultaneously in seeking explanation of phenomena, encouraging them to assess evidence objectively in the formation and evaluation of hypotheses, and cultivating in them an appropriate attitude towards inconsistent data. With regard to the feasibility of teaching students to falsify, it appears high if teachers adopt relatively simple inference tasks, while creating an opportunity for students to collaborate with each other and lowering the normativity of the learning environment. As for the utility of doing so, although disconfirmation might be an effective heuristic when students cannot appeal to an outside authority to test their hypotheses, it appears not to be a universally effective strategy for solving reasoning problems. In contrast, confirmation seems not to be completely counterproductive and might be a useful heuristic, especially in the early stages of solving by hypothesis a complex inference problem. After all, whether disconfirmation or confirmation is better often depends on the characteristics of the specific task at hand.

波普尔的证伪探试启发法是否有利于发展批判性思维？

林志明

香港大学

批判理性主义的三个核心概念

波普尔将批判理性主义的本质阐释为一种承认‘我可能错，你可能对，经过努力，我们都更加接近了真理~~我可能错，你可能对，经过努力，我们都更加接近了真理~~’（1966，225页）的态度，实际上，批判理性主义是一种乐于听取批判性观点并从错误中学习的态度。波普尔晚年透露，批判理性主义的形成是受一位来自卡林西亚的年轻纳粹党成员的激启发。这位纳粹党成员在希特勒上台（1933）前不久对波普尔说；‘什么，你想辩论？我才不辩论：我开枪！’（1996，13页）。事实上，可能就是这位年轻人宁愿开枪而不愿辩论的行为埋下了激发波普尔~~提出了~~批判理性主义的三个核心概念的种子。，这三个核心概念就是即可错论（‘我可能是错的’）、，批判（必需的‘努力’），以及逼真性（‘我们可能会更加接近真理’）。

通过可错论，波普尔（1966）想要表达的观点是，我们是会犯错的，并且，追求确定性是错误的。前一种观点可从历史的角度得以证明，即我们曾经认为得到证实的事物可能后来发现是错误的；而后一种观点则可以从理论上来理解，比如我们能解释的或能了解到的事物是有限的。这种局限性一方面与我们大脑的解释能力有关：根据哈耶克（1952）的观点，任何用以分类的工具必须具有比被分类的物体更复杂的结构；这意味着用以解释的工具无法解释同类或者同等复杂程度的物体，因此人脑永远无法全面解释它自己的运作。局限的另一方面则是因为我们无法预测未来的历史进程，特别是无法预测未来人类知识的增长：就像波普尔说的，‘如果人类的知识确实在增长，那么我们现在就无法预测未来自己会知道什么，只有到了未来才知道’（2002a，, 12页）。因此，他的可错论否定了确凿的知识以及权威的知识来源的存在。相反，他坚持认为，没有知识是可靠的，我们的知识都是推测出来的，而且是会出错的。

但是，因为我们能从错误中吸取教训，所以，可错论未必导致怀疑主义或相对主义的结论。而且，他认为，批判“是我们发现错误并系统地从错误中学习的唯一方法”（1966，376页）。这不仅包括对他人的理论和推测的批判，也包括对自己的理论和推测的批判。在波普尔看来，批判总是包括指出矛盾（被批判的理论内部的矛盾，我们有理由认可的某一理论和另一理论之间的矛盾，或者理论与有关事实的某些陈述之间的矛盾），因此他建议将演绎逻辑推理作为批判的方法。原因是，只有依靠纯粹的演绎推理我们才能发现我们的理论隐含的意思，并据此找出矛盾。更具体地说，演绎逻辑或形式逻辑对于批判的重要性在于，它采用了一些规则，通过这些规则，真理通过前提得出结论，而证伪是从结论反传到前提。正是这种证伪的反传递‘使得形式逻辑成为批判理性主义的工具~~批判理性主义的工具~~，即反驳的工具’（同上，64页）。事实上，波普尔反对任何证实理论的尝试，在他的非证实或证伪的理性观中，他用批判取代了证实：‘过去，大多数哲学家都认为，宣称理性都意味着理性地证明（某人的观念）；但我的观点是，至少是在自我的《开放社会》出版以来我的观点是，理性意味着理性地批判（自己的理论或与自己的理论相矛盾的理论）’（2002b，, 173页）。但是，考虑到一个理论可能会比与之相矛盾的理论更能经得住批判，波普尔承认，有时我们可以‘证实’我们对某一理论的倾向性~~倾向性~~。从消极意义上说，一个理论会在经受住批判时，而不是获得正面的证据时得到一些支持。

更接近真理或者达到更高的逼真性这一想法对波普尔的批判理性主义的概念至关重要，因为只有通过真理我们才能理智地评说可错论和批判性：通过批判性的讨论找出错误并尽可能多地消除错误，目的是更加接近真理。主观真理理论认为，真理是我们按照某一有理有据的标准有理由相信或接受的事物。波普尔（1989）批判了这一理论，同时接受了塔斯基的客观真理符合论。客观真理符合论认为，当且仅当某一陈述与事实相符时它才是正确的。~~波普尔（1989~~）把真理看作是我们按照某一有理有据的标准有理由相信或接受的事物，因而他批判了主观的真理理论，接受了塔斯基的客观真理符合论，这一理论表示，当且仅当某一陈述与事实相符时它才是正确的。一方面，塔斯基的客观真理理论允许我们作出一些论断，尽管这些论断在波普尔看来是明显正确的，但在那些主观真理理论内部却是自相矛盾的：例如，一个理论，尽管没有人相信它，或者尽管我们没有理由相信它，但它仍然可能是真的；另一个理论，尽管我们有相对充分的理由认可它，但它仍然可能是假的；我们寻求真理，但当我们已经发现真理的时候可能并不知晓；我们没有真理的标准，却将真理本身作为规范准则。为了增强接近真理或者知识增长这一观点的可信度，波普尔（1979）将塔斯基的真理和内容两个概念合并，提出了逼真性这一逻辑观点。波普尔将由陈述p推断出的所有真实的陈述和所有虚假的陈述分别定义为真理性内容和虚假性内容，并解释如下：

直观地说，当且仅当（a）理论T1与理论T2的真理性内容与虚假性内容（或他们的方法措施）是相当相似的，并且，要么是（b）T1的真理性内容而非虚假性内容比T2的少，要么是（c）T1的真理性内容不比T2的少，但它的虚假性内容比T2的多时，理论T1比理论T2才具有更少的逼真性。（同上，52页）

因此，波普尔并不是把对真理的追求，而是把对逼真性的追求作为一个更加实际的科学目标。原因是，我们将永远无法拥有足够好的理由来宣告我们确实已经获得了真理，但是，我们可以有相当好的理由来宣告我们已经在走向真理的道路上取得了进步（即，理论T2更接近真理，因此比它的前一理论T1更可取）。

抵制反对批判的花招诡计

然而，若要将这样的证伪主义理论应用于实践，则有必要找出那些妨碍批判且易于将个体限制在证实框架中的哲学假设，并与之对作斗争~~之加以反对~~。就像中国谚语警示的那样：‘明枪易躲，暗箭难防’，躲避或消除由反对批判的假设所带来的影响是不太可能的，除非让那些旨在减少或回避批判的潜藏花招诡计自己也行接受批判。作为证伪主义的支持者，波普尔不遗余力地揭露这种保护自己、~~来庇护或~~回避批判的花招诡计。首先，他（1989）指出，认为真理显而易见的学说与证伪主义学说背道而驰，因此也与容忍学说背道而驰：如果真理是显而易见的，那么我们就不太可能犯错，因此也没有必要容忍或原谅他人因个人偏见而犯下的错误。而批判包括寻找自己及他人的错误，这本身就假定我们易于犯错并且因此应该宽容他人，所以认为真理是显而易见的学说与这一看法完全对立。波普尔抨击反对的另一个花招诡计是，要求概念的准确性并将此作为批判或解决问题的前提。波普尔坚称‘准确的’概念或具有‘明确界限’的概念不存在，他（同上）强调，语言之所以重要，只是因为它是形成理论的工具，其准确性只要能解决问题就足够了。我们遇到的问题有时可能要求我们对问题作进一步的加以区分，以便在表述这些问题时能够做~~别以达~~到清晰和准确。，面针对这一要求，波普尔建议采用具体问题具体解决专门的方办法：

如果因表达不清造成误解，不要试图通过新建或加固基础来构建一个更加准确的‘概念框架’，而要针对具体误解，重新表述设计你已表述过的东西~~的特定阐述方式~~，以便来消除已经造成的或者你能预见的误解。而且永远记住：没有不会引起误解的说话方式~~没有不会引起误解的说话方式~~。不论怎样，总会有人误解你。（~~波普尔~~Popper, 2002b，, 29页）

波普尔进一步发现了三种妨碍批判的方法，即本质主义，工具主义和传统主义。本质主义假定科学的目的是找到描述事物‘本质’——即位于现象之后的事实——的终极解释，因此不需要进一步解释，也不会受到深层解释的影响。作为反蒙昧主义者，波普尔（1989）批判本质主义，原因是，它阻止人们提出富有成效的问题或进行进一步的批判。同时，他（同上）也谴责工具主义理论是只能用来预测的工具，因为它只强调应用而忽略了证伪或批判：为了实际应用的工具性目的，一个理论在被驳斥之后，仍然可能在它的应用范围之内继续使用——换句话说，当一个理论被理解成简单的工具时，它就不能被证伪，因为无论何时都可以解释为，不同的理论具有不同的应用范围。对于传统哲学——它把自然规律看成是我们自己的创造及随意的约定，而不是自然的体现——尽管波普尔承认，弄清理论与实验之间的关系这一做法值得肯定，更确切地说，是认识到‘我们在实施或解释科学实验时，按照惯例约定或者演绎逻辑来计划的行动及操作的重要性’值得肯定，但他仍然抵制传统哲学中用来保护自然科学的理论系统免遭批判的方法。他认为，这种花招至少存在四种约定俗成的形式~~这种诡计~~——引入临时性~~进“特定~~的”假说，修改显而易见的定义，对实验者的可靠性采取质疑的态度，对理论学家的聪敏表示怀疑——这四种花招诡计使得证伪自然科学的理论系统无法实现。

确证实偏好见

确证实偏好见的广普性以及各种伪装形式

除了要批判各种潜藏的阻妨碍批判的花招诡计之外，反对与人类们偏好向确证实或忌讳反对否证的心理倾向作斗争也很重要——。这是一种~~常见的~~心理倾向十分普遍，它反映了证伪主义与~~据说根深蒂固的~~人类心理中显然是根深蒂固的一些机制之间的冲突矛盾。不幸的是，波普尔对此并没有给予太多关注。据尼克森（1998）说，确证实偏好见指的是无意地寻找或解释证据的过程，而且偏向于已有的观念或假设。大量经验主义的证据表明，确证实偏好见不仅广泛而强烈，而且以各种不同的形式存在：通过人们的倾向性反映出来，比如，人们在接受一个假说时所要求的与假说相吻合的证据较少，而在反对一个假说时则要求有较多的与假说相斥的信息（匹茨辛斯基、&格林伯格，, 1987）；面对一个有争议的问题时，人们回想或制造理由来支持他们赞同的一方而不是不赞同的一方（Baron巴伦，, 1995）；而且，在判断一个条件句‘若p则q’是否正确时，人们会寻找p及q的存在来确证实这一条件句，而不是通过寻找p及非q的存在来否证这一条件句（Wason，1966）。

尽管波利亚（1954）认为，科学家与普通人的不同之处在于，科学家倾向于为自己的假说寻找否证证据，但科学史上仍有大量确证实偏好向的例子。这可以从两个层面加以解释——个人层面和学院体制层面。在个人层面上，迈克尔·法拉第主张在处理一个新颖的假说时忽略否证证据，直到这一假说被很好地确证实（特温纳，, 1989），而罗伯特·密立根在公布判断单一电子的电荷的实验结果时，只报道与他的假说相吻合的观察结果（Henrion & Fischhoff，1986）。在体制层面，牛顿的万有引力概念之所以遭到惠更斯和莱布尼兹的反对，是因为后者不接受不随空间物质及空间运动的变化而减小的力存在这一观点。与此相似的是，科学发现常常遭到科学家自己的反对，特别是遭到理论地位受到科学发现挑战的科学家的反对。事实上，在面对反常数据对质疑现有理论提出的质疑~~反常数据~~时，科学家们的典型反应是，先质疑数据，若数据证明可靠信，然后就修改再现有升级理论，使其精致复杂化到正好足够迁就~~，使理论能够容纳这一~~反常数据结果的程度（Nickerson~~尼克森~~，1998）。波利亚认为个体科学家倾向于否证自己的假说，也许这一看法至多有一半正确：他们渴望批判或否证其他科学家的假说，而不是自己的。

确证实偏好见的理论解释

对于如何解释确证实偏好见的问题，除了迈特林和圣唐所称的波丽安娜效应（它以常识的方式解释了人们偏爱令人愉快的而不是令人不快的想法和记忆，并因此倾向于相信那些他们希望是真实的观点，而不愿相信那些他们希望是错误的观点），研究人员至少还提出了四种理论性解释。第一，尼克森认为，人们基本上都局限于一次只考虑一件事情，并且倾向于一次只收集一个假说的信息。但是将注意力集中于单个假说可能会增强它的说服力，即使它是错误的：

一个错误的假说可能会变成几乎正确的，因为它得到大量正面的强化，而这些强化可能会被认为是进一步证明这一假说正确的依据，同时会阻止证明它是错误的进一步探究。（Nickerson，1998，198页）

这就是确证实偏好见存在的原因。

第二，埃文斯并没有考虑人们有意试图确证实而不是否证自己的假说这一可能性，他（1989）认为，确证实偏好见并不是因为他们有确证实的动机，而是因为他们无法用清晰的否证术语来思考。他的观点与人们发现处理负面信息比处理正面信息更困难的许多实例相符。例如，判断负面信息的真伪比判断正面信息的真伪更加困难（Wason，~~沃森，~~1961）；相对于正面前提来说，由负面前提得出的推论需要更多的时间去评估，而且更有可能被评估错误（Fodor，Fodor，& Garrett，~~福多，福多、&加勒特，~~1975）。

第三，就像弗里德里希认为的那样，‘我们的推理过程首先是实用的生存机制，其次才是探测真理的策略’（1993，298页），人们在许多现实情景中做出的判断更多地是受成功与生存欲望的驱动——因此将潜在的奖赏与感知的风险保持平衡——而不是为了判断假说的真伪。这解释了为什么把真假设误以为是假的所导致的不良后果比把假假设误以为是真的所导致的不良后果严重时，就会导致确证实偏好见。

最后，证明个人观点合理性的能力很重要，而从教育的各个层面去强调这一重要性，将会促进或者加强人们有选择地寻求确证实证据：如果一个人经常被鼓励举出理由来证明他所持的观点，而不是被敦促去清晰地表达反对这些观点的理由，那么这个人就是被训练实施确证实偏好见（Nickerson~~尼克森~~，1998）。更糟糕的是，一些教育实践无法明确区分理由构建（即选择性地寻求或过度关注支持观点的证据，而不寻找或者忽视反对观点的证据）和证据评估（即寻求各个方面的证据并尽可能客观地评价这些证据），因此实际上是理由构建的就会冒充公正的证据评估：这就是确证实偏好见在受教育的成人中普遍存在且根深蒂固的原因。这种理由构建教育实践的一个典型例子就是辩论。辩论者将精力主要集中于寻找支持他们所辩护的观点的论据，即使他们可能提出了潜在的反论点，他们的目的也只是揭露这些反观点的不足。归根结底，辩论者是为了赢得辩论，而赢的方法是将最有力的例证来支持自己的观点，同时反驳、不理会或者直接无视任何可能反对这一观点的证据。

在削弱确证实偏好向的力度及传播中教师所扮演的角色

有人们可以争辩说确证实偏好对我们向是有帮助好处的：一方面，它可以使我们偏喜爱的信念不像这种偏好不存在的情况下那样易受攻击~~观点变得更加有力~~，从而保护我们的自我意识感知（Greenwald，, 1980）；，另一方面，它也可以防止科学不加区辨别地接受未经受住~~禁不起~~时间考验的所谓新发现（Price，1963）。但尽管如此，这个偏好仍然它被普遍被认为是人类自身的一个缺陷~~的失败。~~：它导致~~有助于~~各种妄想的形成，迷信的发展及存在，以及拥有不一致世界观的人们之间长期存在的敌意和冲突（Nickerson~~尼克森~~，1998）。要想削弱和抑制确证实偏好见的力度及传播，从儿童教育着手可能是个不错的想法。那么教育实践的含义是什么？第一，教师自己要认识确证实偏好见——它的广泛性以及它存在的各种形式。这种认识能够让学生在对重要问题作出决定时更加谨慎，而且对与他们不一致的观点更加包容。

另外，考虑到引起确证实偏好见的部分原因是人们倾向于一次只考虑一个假说，教师应鼓励学生在试图解释一个现象时同时考虑几个可供选择的假说。特温纳等人发现，个体很少能成功运用这一思考策略，因为他们‘宁愿评估几组数据是否适用于某一假说，而不愿评估一组数据是否适用于多个假说’（1980，119页）。这一发现展示了在学习怎样避免确证实偏好差时小组形式的优越性：小组中的每个成员应对一个不同的假说，整个小组就可以同时了解几个假说的情况。

教师也应该认识到明确区分理由构建和证据评估的重要性，鼓励学生在建立和评估假说过程中客观地评估证据。在此过程中，培养学生的批判性思维模式至关重要，它可以促进学生在作出任何判断时都思考支持和（特别是）反对的理由。并且教师还应该使他们了解，为喜爱的观点寻求支持的动机‘经常会使一个人忽视数据中存在的即使是很明显的错误，因为我们很难发现自己不在寻找的东西’（(~~道森，季洛维奇&里根，2002,~~ Dawson，Gilovich，& Regan，2002，1386页）。

尽管科学家都有意忽视与他们的理论相斥的数据，福杰桑等人（ (2004) ）发现，当相斥的数据重复出现时，科学家就会修改自己原来的理论。事实上，科学家一开始不愿接受与理论相斥的数据以及随后通过重复试验重新构建理论都可以被认为是一种实用的探索性试手段~~的方法实际的启发式工具~~：它防止科学家过早地接受可能是错误的发现，并允许他们修改理论，进而增长知识。在科学教学领域，应该向学生介绍这种探索性试手段~~方法启发式工具~~，特别是要培养面对与理论相斥的数据时的合理态度。

学生们可以学会证伪吗？

尽管波普尔在哲学以及科学实践领域颇有影响，对于他的方法论是否的有效这点性仍然可以提出一个问题~~有待验证~~，原因为在是，心理学文献中，将证伪作为解决科学问题之解决策略的可行性及实用性还存在许多争议。首先，大量心理学研究表明，许多科学家在否证推理方面有感到困难。例如，马奥霍尼和金珀姆坡（1976）做了一项调查，要求一个样本组的~~科学家，包括~~物理学家，生物学家，社会学家和心理学家，评估实质蕴含的四种形式的合理性（即假设p实质上蕴含q，则判断从p 推出q，非p推出非q，q推出p，以及非q推出非p是否合理），并找出能够检验‘若p则q’形式的假说合理性的逻辑评上批判的实验。，结果发现：，超过半数的科学家都没有发现否定后件的假言推理（即从非qp推出非pq）在逻辑上是有效的，而只有不到10%的科学家能够正确选出具有证伪样本假说的批评判潜力的实验。也许更为令人惊讶的是，一组在检验统计学上的~~假设（~~零假设）方面受过正式训练、因而也就在检验可能的否证证据方面受过正式训练的一个样本组的统计学家们在辨识别证伪的逻辑合理性时也遇到了同样的困难~~，因此在检测可能的否证证据时也遇到了困难~~（~~艾因霍恩~~&~~贺加斯，~~Einhorn & Hogarth，1978）。

但是，在运用否证推理方面有困难并不意味着没有能力运用这一推理。事实上，一些研究者已经成功教会了大学生利用否证策略来解决这样的推理问题，如沃森（1960）的2-4-6问题以及加德纳（1977）的‘新艾留西斯’。沃森将2-4-6问题加以改善作为归纳推理的测试：研究对象被告知，‘2-4-6’这一顺序的三个数字是实验员所想规则的一个例子（规则是‘任意三个递增的数字’）；测试要求研究对象按照自己所想的测试顺序举出三个数字，实验员会告诉他们这三个数字是否合乎规则，研究对象则根据实验员的反馈找出规则。考虑到沃森的研究对象表现出了很强的确证实偏好见，而且常常举出与他们的尝试假说相符的数字顺序，特温纳等人（1980）运用同样的2-4-6任务，尝试向他的研究对象传授否证策略，即要求他的研究对象得出否证的例子。他们发现，否证小组的研究对象得出确证实例子和否证例子的平均数分别是1.5和6.6（在实验1中）。这表明，特温纳成功地消除了大部分的确证实尝试，由此也表明，他在改变否证小组中研究对象的探究策略方面也是成功的。

‘新艾留西斯’是一款模拟用归纳法寻找真理的纸牌游戏。格曼，拉塔以及坎宁安（1984）将此游戏进行改编，设计出一个研究科学推理的任务：各研究对象每次出一张牌，以此来推测大家所出牌背后的规则（例如，有一条规则是‘相邻的两张牌数字之差为1’）；实验员会告诉研究对象他们出的牌是对还是错，但直到实验结束后才会对他们对规则的猜测给予反馈。为了利用这一任务来研究确证实、否证以及合并策略对团队解决问题的影响（实验2），实验员让研究对象分别专注于尽量多地出正确的牌，尽量多地出错牌，以及尽量多地出正确的牌直到大家出的牌验证他们的推测是错误的。结果发现，否证小组出错牌的概率是41%，合并小组出错牌的概率是33%，而确证实小组的出错牌的概率只有20%。实验结果再一次表明教学操作是成功的，因此也就证实了劝说运用否证的可行性。

对诱导引出否证起作用的两个有利因素

如果要在课堂上推行否证策略，则需注意与否证的高级应用有关的两个促进有利因素——合作推理以及低规范性。通常，阐明小组的运作方式有利于否证的运用。以下是格曼等人（1984，75页）提供的两个研究对象在同一否证小组中的简短交流：

一位研究对象向其他小组成员抱怨：‘我很难猜错’。另一位研究对象试着告诉他如何运用否证：‘如果你认为牌的规则是这样的（指着一列数字依次加1的牌），那么出一张不符合此规则的牌来证明这一规则是错误的’。

很快，第二位研究对象不仅说服了第一位研究对象，也说服了其他小组成员来证伪更多的推测。同伴交流的有益效果在莫什曼和金邦（1998）的研究中也得到证实。他们发现，在测试一项假说时，相互交流的小组中有75%的研究对象能够正确运用否证策略，而独立工作的研究个体只有9%的人能够做到这一点。仔细研究小组讨论的录影带，并未发现小组成员被动附和大多数人的意见或者某一看似专家的意见，而是像往常一样，通过分享观点和理由，努力共同设计一致的解决方案——即比大多数个体想出的方案更加优质的观点架构。因此，实验员将小组的优秀表现归因于合作推理而不是同伴压力或同伴效仿。比起个体推理，在合作推理中更易洞察到证伪的逻辑。

但是，在假说的合作检验中是否用到否证取决于推理者与其同伴以及与实验员之间的关系类型：布特拉、凯文尼及罗西（2005）的研究表明，研究对象面对能力较低的同伴时能够学会运用否证，而面对能力较高的同伴时则倾向于运用确证实，即使他的同伴运用的是否证。一个可能的解释是，面对能力较高的同伴，研究对象的胜任感会受到威胁，因此他们会运用看似支持假说的确证实作为防御策略来检验自己的假说，以此来保护他们的能力；相反，面对能力较低的同伴，研究对象的胜任感则不太可能受到威胁，因此他们就有机会通过否证来检验自己假说的有效范围。另外，巴特拉等人（2005）表明，研究对象遇到违反谈话规则的现象时——也就是，在解决沃森（1960）的2-4-6问题时被实验者告知，2-4-6并不是规则的例子，之所以被选中只是想展示什么是一组三个数字——利用否证的机率较高，无论同伴的能力如何。巴特拉等人解释说，否证的产生不仅源于在互动及社会影响（如能力的约束力）中偏离社会规则的可能，而且源于在语言（如实验员所举例子的约束力）中偏离交谈规则的可能：鉴于处于引导地位的实验员可能会通过谈话引导研究对象专注于给出的三个数字，让他们形成体现这三个数字所有显著特点的假设，并试着对这一假设进行证实，那么，告诉研究对象2-4-6并不是规则的例子这一行为可能会打断他们对任务过程的专注，从而引导他们使用否证。因此，人们在限制推理的情况下会使用确证实，但是‘否证的运用可以通过降低这一情况的规范性来增加，而降低规范性可以通过一个不那么危险的资源或不那么约束的交谈规则来实现’（同上，186页）。

换句话说，要想有效地教授学生否证，仅仅为他们创造相互协作的机会是不够的。老师也应该试着通过以下方法降低学习环境的规范性：保证在小组中交流的学生不会威胁或左右其他成员，并且，在他们解决问题的过程中要避免以权威者的身份提供标准的解决方案。而后者尤其值得注意，因为许多老师在课堂上确实认为自己是权威者，并且，他们认为，权威者就应该知道每一个问题的答案。教师的权威形象对于否证的学习是不利的，这体现在两个方面：第一，它使得师生间的互动更拘束；第二，它使得课堂不太可能满足在教学中实施证伪主义的基本要求，即课堂应成为重视师生所犯错误的场所~~即使课堂不太可能变成评价师生错误的场所~~（Sankey，, 1999）。

该问题复杂性的影响

尽管上述证据支持人们可以学会证伪他们的假设这一观点，但一些研究表明，当推理问题变得复杂时，教学操作可能无法引出证伪。例如，迈纳特、多尔蒂及特温纳（1977）为了在他们的研究中得到更加真实的科学模拟，设计了一项相当复杂的推理任务：研究对象首先观察一组计算机显示图像（图像由静止的几何图形及运动的粒子构成，而且粒子的运动受图形的影响），然后提出控制粒子运动的规律假设，并选择恰当的实验加以检验，最后得出控制粒子运动的规律。结果发现，教学操作没能成功引导否证小组寻求否证。在接下来的一项研究中，迈纳特、多尔蒂及特温纳（1978）对研究对象的证伪给予更加广泛的指导，同时把任务变得更加复杂也更加真实，允许他们以更加自由的方式进行探索（例如，允许他们自己设计实验，而不是强迫他们必须从可能的实验中选择。）但是，在迈纳特等人的研究中（1977），他们发现，否证的教学指导对否证小组的影响很小，甚至没有影响；事实上，‘没有任何迹象表明他们（包括否证小组和确证实小组）有意寻求否证’（Mynatt~~迈纳特~~等人，1978，第400页）。

他们认为，一个可能的解释是，当人们在寻求一种解决复杂推理问题的方法时，否证策略可能超出了大多数人的认知能力——因此很难引导他们运用否证。由此看来，能否教会人们去证伪似乎取决于任务是否复杂。

让~~若更深一步看待~~问题进一步复杂化的是：，有时很难从人们在测试中的表现来判断他们是否真的遵从了证伪的指导，因为也许有人会说，如果被指导去证伪的人们开展~~表现出~~维斯瑞克（1962）所说的负检验~~面测试~~——即用预计结果为负值的检验项目~~被语言错误的测试项目~~来检验自己他们的假说——但同时又希望检验测试结果能够证实假说而不是证伪假说，那么证伪指导就并没有成功实施。这一看法与波利蒂克（1996）的研究结果相同并得到波利蒂克的支持。波利蒂克发现，尽管证伪小组中有60%的研究对象采用了负检验~~面测试~~，他们当中只有10%的人希望假说被证伪，包括：

……在检验假说的任务中，当研究对象被要求表现为证伪者时，他们的行为表现出自相矛盾的两方面特点：一方面偏爱负检验方式~~倾向于负面测试~~，另一方面又希望证伪策略无法得出与假说相矛盾的数据。（Poletiek，, 1996，, 456页）

换句话说，有些人既采用负检验方式，同时~~面测试~~又期望自己的假设会被证实假设，如果认为这样的些人是在遵行循证伪训条，那似乎就指导是在自相矛盾的了。人们被引导去证伪时就可能会出现这种自相矛盾的情况，如果我们暂且将怎样解决这一问题搁置一边，那么目前有一个有趣的问题是：为什么他们的思想与行动不一致？也就是，他们在进实行负检验~~面测试~~时，为什么不希望证伪他们的假说？是因为他们对证伪策略从心里感到怀疑吗？这些策略真的可用吗？

学生应该学习证伪吗？

由迈克尔·格曼和他的同事在20世纪80年代的研究结果来看，人们对于证伪的使用所持的谨慎乐观的态度是有原因的。首先，在上文提到的研究中，格曼等人（1984）研究了不同的策略如何影响各小组在由‘新艾留西斯’改编的任务中的表现，结果发现，否证小组找出规律的机率（72%）大大超过了否证与确证实合并小组（50%）以及确证实小组（25%）。研究还发现，否证小组出错牌的机率最高（41%），否证与确证实合并小组出错牌的机率居中（33%），而确证实小组出错牌的机率最低（20%）。并且，这三个小组出错牌的机率与他们能够成功找出规律有极大的关系，这表明策略指导的实施是成功的。实际上这也是他们表现不同的原因：这表明在解决问题时否证是有效的。

考虑到科学误差，为了使否证在科学推理中更具权威性，格曼（1986）在另一项有关‘新艾留西斯’实验的研究（实际上与之前的研究设计完全相同）中考虑到了误差的可能性：研究对象被告知，实验员对他们出牌的反馈中有0至20%可能是错误的，也就是，一张正确的牌可能被归类为错误的牌，或者一张错误的牌可能被归类为正确的牌。他发现，相比确证实小组和对照组（即没有使用策略的小组），否证小组的表现并没有特别好，因为可能的误差妨碍了否证小组获得及使用否证信息的能力：把否证信息归类为误差使得他们的假设不会被否证，而且他们花费了太多的时间检测误差，而没有充分检验他们的假设。但这一结果并不能说明在可能有误差的情况下否证就是无用的。考虑到实验成功的几个小组用的策略是将否证与重复结合（即重复他们认为可能会出现误差的情况），这意味着否证成为了一个不充分但更必要的策略（同上）。

使否证有效的有利条件

格曼（1984）进一步表明，他们发现，当个体研究对象完成沃森（1960）的2-4-6时，否证教学对他们同样有着积极的影响。更确切地说，他们发现，就研究对象发现最初规律（即‘升序的任意三个数字’）的人数而言，否证组（95%）大大多于确证实组（48%）和对照组（53%）。但奇怪的是，在上文提到的特温纳（1980）的研究（即实验1）中，并没有得出否证对于研究对象表现的积极影响。虽然该研究的设计与格曼的十分相似，但研究结果却是：‘尽管研究对象确实学习了怎样寻求否证资料，但拥有这种策略并没能让他们更快地找到解决方案，也没增加找到正确解决方案的人数’（112页）。随后，格曼和他的同事们发现，他们的研究结果与特温纳的之所以不同，可能是因为在他们的实验中，研究对象直到实验结束才被告知自己对规则猜测的对错，而在特温纳的实验中，研究对象对规则的每一个猜测都会被告知是正确的还是错误的，因此他们就可以依赖实验员来决定使用确证实还是否证（格曼，, 1992）。因此，当人们无法求助于外在权威来确定自己它们是否在朝着做出一项发现的方向中取得进步时，否证可能是一种项有效的探试启发式方法教学。

但是，当人们可以求助于这样的权威时，否证就没那么有用了，它在许多高中和大学课堂的实验练习中也就没那么有价值了。格曼（1995）认为其原因是，大多数情况下，这些练习的目的是得到正确答案，而不是发现新奇的现象，而且经常向实验室助教或导师这样的权威求助不仅是可能实现的，而且可能是有益的。这项研究的教育启示是，对于那些通常无法求助于权威来检验假说的未来科学家而言，另一个更加开放的探究式练习可能会更好地训练他们使用否证。

否证的局限性

但是，用于解决推理问题时，否证似乎并不是一个普遍有效的策略。一些2-4-6研究（如Gorman & Gorman，1984；Gorman，Stafford，& Gorman，1987~~格曼&~~格曼）的研究结果证实了这一点。这些研究结果表明，否证不能用于非常普遍或者更加困难的规则，如‘没有两个完全相同的数字’。而且，否证似乎也不能独立应用，原因是，它的使用有时需建立在确证实的基础之上。这体现在如下两个方面：第一，在策略上，确证实是对否证的必要补充，特别是在解决复杂推理任务的早期阶段。这里有两个说明性的例子：迈纳特、多尔蒂及特温纳发现，尽管研究对象都没能完成高要求的任务，但最成功的研究对象最初是致力于收集支持假说的确证实证据而不是否证证据，并且在实验中力求确认在假说被相对证实之后是否还存在否证的例子；另外，卡米洛夫·史密斯及英海尔德（1975）发现，儿童在面对复杂的推理问题时无法使用否证证据——即识别反例——直到他们的假说得到充分的证实。与这两项研究结果相同的是，瓦塔尼安、马丁代尔及科维亚特科夫斯基（2003）表示，在检验假说的早期阶段和晚期阶段都使用确证实和否证相结合的策略是非常有益的。有人说，法拉第故意忽视否证实验正是反映了他的证实偏好（指法拉第1831年所做的发现电磁感应的部分实验）。~~法拉第被指控：他的故意忽视否证实验（1831年实施的发现电磁感应实验的一部分）反映了他的确证实偏好见，~~事实上，为了替他辩护，特温纳（1995）论证说法拉第的实验只是表明使用了‘先确证实，后否证’这样一种复杂的策略，并作了如下的详细解释：

自然本是杂乱的，并且经常给询问者提供错误的反馈。……事实上，法拉第的（许）多实验都产生了期望的效果，但这效果太微弱，（他的）迟钝的仪器无法检测出来。在这样的情况下，科学家的任务是在明显的无序之上施加有序……确证实启发式探试法~~启发教学~~的必要作用之一……（是）它过滤掉了一些噪音，从而让信号可能被检测到。要做到这点并非是确定无疑~~不是必然发生~~的事情，这也就是为何之后~~为什么后来~~否证策略会成为了一个必要的补充的原因。（~~特温纳~~&~~奇特伍德，1995,~~ Tweney & Chitwood，1995，255页）

第二，从本质上说，确证实策略不仅不一定和寻求否证的目标相矛盾，而且在某些情况下还是达到这一目标的唯一途径。在克雷曼与哈耶克看来，为了理解这一点，确证实策略最好被解释为一种个正检测策略~~正面的检测策略~~，也就是，通过检验事例来检测假说，而在这些事例中，目标特性被假定存在或已知存在。另外，区分‘寻求否证’的两个不同的含义非常重要。一个含义是检验那些被预测没有目标特性或没有实施负检验~~面测试~~的事例，这是经验主义调查的重心。另一个含义是检验最有可能证伪假说的事例，这是波普尔所强调的。以沃森（1960）的2-4-6任务为例，克雷曼生动地展示了从波普尔式的意义上说，尽管当假定规则（如‘以2递增’）包含于正确规则（如沃森最初任务中的‘递增数列’）之中时，正面的检测策略不能得出证伪，但当假定规则（如‘以2递增’）与正确规则（如‘三个偶数’）重合时则可以得出。更重要的是，实际上也是自相矛盾之处，当假定规则（如‘以2递增’）与正确规则（‘连续的偶数’）密切相关时，一个正面的检测策略是唯一可以显示确凿证伪的策略——即使负检验~~面测试~~也无法显示。因此，我们可以断定，不了解当前任务的属性时，就无法判定确证实或正面检测策略的有效性。

结论

综上所述：波普的证伪认识论的实施意味着让各种反对批判的哲学假设接受批判，包括真理是显而易见的，精确的概念是批判的前提，本质主义，工具主义，以及传统主义这样的信条；它也意味着可以通过多种教育手段来反对证实偏好，包括：，通过帮助教师与学生获得对确证实偏好见的广泛性及不同形式的认识，教会他们在寻求对现象的解释时同时考虑几种不同的假设，鼓励他们在形成及评价假说时客观地评价证据，并且培养他们在面对与假说相矛盾的数据时拥有合理的态度~~等教育手段来反对确证实偏好见~~。至于教会学生去证伪的可行性，如果教师采用相对简单的推理任务，同时为学生创造相互合作的机会，并降低学习环境的规范性时，则可行性较高。至于它的实用性，当学生无法求助于外部权威来检测他们的假说时，否证可能是一个有效的探试启发法，但它可能不是一个解决推理问题的普遍有效的方法。与此相反，确证实似乎并不是完全产生相反的效果，它还可能是一个有用的探试启发法，特别是在通过假说解决复杂推理问题的早期阶段。归根结底，确证实与否证哪个更好常常取决于目前具体任务的特点。

翻译评注：核心句分析法在英语长句汉译中的应用——以“《波普尔的证伪探试启发法是否有利于发展批判性思维？》”的中译为例

一. 核心句分析法与长句翻译

1.1 核心句分析法

美国语言学家、翻译家、翻译理论家尤金·奈达（Eugene Nida）指出，一个有经验的译者往往不会被错综复杂的句子表层结构所遮住视线，而会透过表层结构，看到句子的深层意义。他认为，译者不应从句子的表层结构出发直接生成译入语的表层，而应拐弯抹角到达目的地。“他把翻译比作涉水过河，如果上游水流湍急，就不能硬闯，而要到下游找一个浅的地方过河，然后再折回上游的目的地”（叶子南，2001，p. 168页）。

为了找到这一“躲激流、，过浅滩”的方法，“奈达借引用了乔姆斯基早期转换生成语法中核心句（kernel sentence）的概念，并将之运用到他的翻译研究中”（Munday，, 2001, p. 39页）。“简单地说，核心句就是最简单的句子”（叶子南，2001，p. 168页）。奈达认为，“各种语言在核心结构上要比在表层结构上远为接近”（谭载喜，1984，p. 35页)。这就意味着，在翻译过程中，“可通过核心句层面的转换作为过渡而译出令人满意的译文”（赵护林，2009，p. 69页）。

奈达（2004）指出，“英语中最简单的核心句有七种：

John ran quickly. (subject + predicate + adverbial)

John hit Bill. (subject + predicate + object)

John gave Bill a ball. (subject + predicate + object + object)

John is in the house. (subject + be + subject complement)

John is sick. (subject + be + attributive)

John is a boy. (subject + be + indefinite article + noun)

John is my father. (subject + be + pronoun + noun)”

奈达他认为，有经验的译者在理解原文时会将原文复杂的句子分解成多个这样简单的句子。

~~而在对~~面对复杂的原文句子~~表层结构进行分析时~~，译者传统的做法一般是分析单词的词性和表层句法成分，并按照目标语语法规则输出译文。这种做法易使译者拘泥于原文的表层语法结构，造成译文意义上的含混不清和表达上的生硬死板。

为了改变这一现状，奈达按照语义将所有的词语~~进行重新分类，并按照语义~~分成为四个类，别，即：

实体、

事件、

抽象概念

及关系。

实体指具体的人和物等；事件指行动、过程等发生的事；抽象概念指对实体和事件等质量和程度的描写；关系指用来将实体、事件、抽象概念连接起来的手段（叶子南，2001，p. 166页）。在翻译过程中，通过在分辨对词语的语义类别的基础上分类分析出核心句及核心句之间的关系，进而彻底理解与正确传达译文。比如在

environmentally damaging waste

中，environmentally 是实体，damaging 是事件，waste 也是实体，这三个语义范畴词之间的底层逻辑修饰关系是：

(waste) [does] (damag[e]) [to] (environment)， {*存在于句子底层结构中而未出现在原文句法表层的成分用方括号标明；下同。}

因此，那么这个短语就应该可译成：

“损害环境的废物”，

而非按照传统语法分析法所译成意义模糊、表达啰嗦的：

“在环境上有损害的废物”（叶子南，2001，p. 167页）。

奈达后来发现词语之间和词的关系十分复杂，就将词语的种类语义类别由四种增加到了七种：

实体（entities）、

活动（activities）、

状态（states）、

过程（processes）、

特征（characteristics）、

连接（links）和

指示（deictics）。

虽然词总体的语义种类增加了，但找出核心句的基本模式是一样的。

奈达在其与泰伯合著的《翻译理论与实践》（(2004)）中~~进一步~~总结出核心句分析法的五个步骤：

“1 . ）分析句子中每一个词语的语义种词类；

2. ）找出核心句，显化核心句中的内隐成分；

3. ）列出核心句；

4. ）找出各核心句之间的关系；

5. ）重新表达核心句，使之易于向译语的表层结构转换”。

在这五个步骤中，前三步是找出核心句，后两步是对核心句进行梳理，以便组成符合译语表达习惯的表层结构。

1.2 核心句分析法对英语长句汉译的启示

英语长句的特点是~~表现为~~“并列成分多，、修饰语多，、语言结构层次多，介词短语、分词结构、不定式、复合句等相互层叠”（谢莉，2009，p. 109页），长句中的多重修饰成分和层层逻辑关系既对译者看清~~掩盖了~~句子的主干结构，、条分缕析句义以~~对译者~~理解原文信息造成一定的困难~~。另外~~，也对译者按照汉语语法和惯用法规范重组原文信息、生成自然的译文构成相当~~在将英语长句汉译时，由于原语结构复杂，层次较多，译者易受原文语法结构~~的束缚~~，造成译文生硬的现象~~。

核心句分析法正是从找出句子的核心成分入手，层层分析出各核心句成分之间的逻辑关系，。这种方法的引入一方面使译者对原句的意义及意义分布子有清晰明了的理解，另一方面也有利于译者让自己的，“~~核心句的介入有利于使~~译文摆脱原文句法的束缚，不至于产生机械对等但风格上不可接受的句子”（赵护林，2009，p. 69页）~~，使译文较为灵活~~。因由此，核心句分析法便可以用来解决英语长句汉译中原文意思理解与~~不清，译文~~目标语表达两方面混乱的相关问题困难。

1.3 核心句分析法指导下的英语长句汉译策略

李长栓（~~2004~~）在其著作《非文学翻译理论与实践》（2004）中强调，核心句分析是克服句法障碍的有效手段，并提出了利用核心句分析和重新组的合方法来翻译英语长句的策略。封权初在于~~2007年发表的《~~“论英语长句‘核心句分析’汉译策略”》（2007）一文中也提出了英语长句“核心句分析”的翻译策略。彭长江在其著作《英汉-汉英翻译教程》（2002）中提到分析英语长句的一个方法是：把复杂句分解为核心句，分析各核心句之间的语义逻辑关系，然后用合乎译语表达习惯的排列方式对核心句进行重组。孙序、李小康和郑晶主编的《英汉互译基本技巧》（2014）中也借鉴了李长栓的核心句分析策略~~的方法~~，阐述了将复杂句拆分成包含“动作”的短句、，然后根据译者所理解的逻辑关系重写原文的翻译方法。龚雪萍在其著作《大学英语实用翻译教程》（2008）中提出了利用核心句分析进行长句翻译的策略——拆译，即先对英语长句意义层次进行分析，然后将其拆分为语义相对完整、的只包含单个主谓结构的核心句，分析各核心句间的语义逻辑关系，然后再用符合译语表达习惯的方式对核心句进行重组。

本论文翻译项目的原文选自一部有关批判性思维的学术著作，“学术著作主要运用逻辑思维，以客观事实、统计数字、逻辑的判断和推理说话”（李长栓，2004，~~，p.~~ 24页），,属于非文学文本。根据李长栓（2004）的总结，非文学文本翻译的首要任务是要做到信息传递的准确，另外，译文表达必须符合逻辑，前后意思连贯，语言朴实明白，不追求文采。~~因此在翻译本文时需特别注意原文信息的准确理解及译文表达的合乎逻辑。此外，~~本文作者选译的原文内容抽象，文中长句偏多且，结构繁复杂，连接词语使用频繁，~~内容抽象，使~~句子成分之间的关系~~有时显得更为~~复杂，译者在~~这对于准确理解~~原文理解与及译语表达两方面都感觉到造成一定的困难。为此，本人~~基于奈达提出的核心句分析法的五个步骤，~~借鉴前人从利用核心句角度~~分析法~~对英语长句汉译策略所做的研究，在翻译过程中有意识按照基于奈达提出的核心句分析法的五个步骤分析原句，解决长句理解与表达中的困难，并对解决这些困难的程序和方法进行了初步的归纳和思考~~并根据非文学文本翻译的特点，本文从两大方面来分析英语长句汉译策略：原文信息的准确理解和符合汉语习惯的译文表达~~。

二. “《波普尔的证伪探试启发法是否有利于发展批判性思维？》”翻译中的英语长句汉译案例

2.1 原文信息的准确理解

英语长句修饰成分多，结构复杂，~~修饰成分过长，~~“修饰语和中心词之间常常被其他词或词组隔开，以致译者对全句意义的理解受到破坏。”（蔡寿荣、朱要霞，2014，p. 155页）~~。另外，因~~英语长句中多有语义嵌套，多层语法修饰关系给~~层层句子重叠架构，本就对~~译者理清句子结构带来了造成困难，而~~有些长句~~为了避免成分重复，一些长句中有时还会省略掉一些成分词语，这使得译者在理解原句的努力~~子完整意义时~~更加艰巨：~~困难，~~译者会找错甚至找不出已省略的成分，最终造成~~错误理解~~原文信息理解失误的现象。

在本文翻译过程中，译者采用了两种基于奈达核心句理论的~~分析法，译者在翻译原文过程中采用了以下两种~~方法来分别解决这些问题。

2.1.1 找出核心句

英语长句中繁多杂的修饰成分以及层层嵌套的结构有时会使译者难以理解原文作者所要表达的信息。，针对这种情况，译者应仔细可先找出原文句子的核心句。每一个核心句都表达了原句文的部分信息，找出所有的核心句及其之间的关系便则可充分理解原句文的全部信息。

找核心句时，首先应可按照奈达的核心句分析法分析每一个词的语义词类别，确定其是实体、活动、状态、过程、特征、连接还是及指示~~中的哪一类~~。李长栓认为，“核心句分析的实质是找出‘谁做了什么’。”（2004，p. 223页）~~。因此，~~在分析核心句分析时，应先找出“活动”。“在英语中，“‘活动’”可能由一个动词表示，也可能由名词化动词表示。找出“‘活动’”后，再根据逻辑关系、百科知识或常识，找到这个动作的施事者及（如果有的话）受事者，”亦即两个实体。将一个“活动”与一或这两个“实体”按照“谁做了什么”的关系进行连接便可构成一个核心句的主要部分内容。，在此基础上，将原句中与该“实体”与该“活动”有关的其他词类按照逻辑关系连接到该核心句中，便可构成一个完整的核心句。按此方法连接各个“实体”、“活动”以及其他语义词类便可找出原句子的所有核心句。，所有核心句及其之间的关系均找出后，~~从而充分理解~~原句文的完整~~各部分~~信息便可以正确理解了。以下是几个例子：

例1： To deal with the problem that our problems may sometimes demand that we make new distinctions for the sake of clarity or precision, he suggests an ad hoc approach.

初读此看句子时，译者可以看出其~~句子的~~主干是“To deal with the problem, he suggests and ad hoc approach”，但因中间部分较长，且包含是由两个由 that 引导的定语从句嵌套结构以及~~而成，~~一个由短语 for the sake of 连接的目的状语，因此译者很难快速理解整句话所要表达的意思。

运用奈达的核心句分析法，先分析句子中每一个词语的语义词类别，并找出“活动”类词语（以下实例中用下划线标明）：

To deal with the problem that our problems may sometimes demand

连接活动实体连接实体状态活动

that we make new distinctions for the sake of clarity or precision,

连接实体活动实体连接特征

he suggests an ad hoc approach.

实体活动实体

该句中的“活动”有：deal with、，demand、，make、，suggest，分别找出作为每个“活动”施事和受事的两个“实体”，进行连接，便可辨识分析出四个核心句的主要内容。再将句子中表示“状态”、“连接”和“特征”的词类按照逻辑关系连接入到相关的核心句中，便可分析得到出四个完整的核心句。因为而第一个由 that 引导的定语从句是用来解释 the problem的，所以因此可将其看成是一句核心句，即：The problem is X。至因此，原句中共可找到~~分析出~~五个核心句：

1) ([We]) deal with a problem.

2) The problem [is X].

3) Our problems may sometimes demand [Y].

4) We make new distinctions (for [clear and precise characterizations of our problems]).

5) He suggests an ad hoc approach ([to the problem]).

从核心句中看出，原文中由两个 that 从句嵌套而成的结构被分析成 2)、）3)、）4)、）三个核心句，其中核心句 2)、）中的 X 是指核心句 3)），核心句 3)）中的 Y 是指核心句 4)），这样整这个嵌套结构的意思就一目了然~~清晰明~~了。根据上述各核心句间的关系将它们~~各核心句~~的信息加以进行整合，便可得到译出忠实通顺的译文：。

译文： 我们遇到的问题有时可能要求我们对问题作进一步的加以区分，以便在表述这些问题时能够做~~别以达~~到清晰和准确。，面针对这一要求，波普尔建议采用具体问题具体解决专门的方办法。

例2： In other words, it seems ‘paradoxical’ to regard those who simultaneously use negative tests and expect confirmation of their hypotheses as following the falsificatory instruction.

这句话中间插入了由 who 引导的长定语从句，该定语从句容易干扰乱译者对句子主干结构和句子整体信息的把握~~，使译者难以看出句子的主干~~。本文作者运用核心句分析法对句子中的每一个词进行了语义词类别分析~~，先找出“活动”~~：

In other words, it seems ‘paradoxical’ to regard those who simultaneously use

连接实体活动特征活动实体状态活动

negative tests and expect confirmation of their hypotheses as following the falsificatory instruction.

实体活动活动实体活动特征实体

该句中的“活动”有：seems、，regard、，use、，expect、，confirmation、，follow。~~，而句子开头的 in other words~~ ~~中还隐含了一个“活动”say~~，分别分析出作为各个“活动”的施事和受事的两个“实体”，进行连接，可构成五七个核心句的主要内容，再将“状态”、“特征”等连接到相关的核心句中，便可构成五七个完整的核心句：

~~(We say) in other words.~~

1) In other words, [X] seems ‘paradoxical’.

2) ([We]) regard ([some people]) as [Y].

3) ([Those Ppeople]) follow the falsificatory instruction.

4) ([Those Ppeople]) simultaneously use negative tests and expect [Z].

~~(People) expect Z.~~

5) ([~~Other~~ Those people’s own]) ~~confirm their~~ hypothesis [will be] confirm[ed].

~~(People) follow the falsificatory instruction.~~

经过分析可知：，句子的主干是核心句 1 2)），核心句 1 2) ）中的 X 代是指核心句 2)），核心句 2) ）中的 Y 代是指核心句 3)），核心句 4) 和）5) ）是对核心句 23) ）中 people 的修饰，核心句 4 5) ）中的 Z 是代指核心句 5)。~~），由~~至此，原文的各部分的信息已然便十分清晰。将各核心句的信息加以进行整合，便可得到如下~~译出忠实通顺的~~译文：。

译文： 换句话说，有些人既采用负检验方式，同时~~面测试~~又期望自己的假设会被证实假设，如果认为这样的些人是在遵行循证伪训条，那似乎就指导是在自相矛盾的了。

2.1.2 需要时显化核心句的内隐成分

谭载喜（2004，42页）指出：“在许多情况下，当几个核心句组合成一个复杂结构时，其中至少有一个核心句的部分成分在表层结构里得不到表明”。~~（谭载喜，2004，p. 42），这有时使~~译者或读者可能会因此难以看不出原句的完整意思或错误理解~~的信息，造成~~原文信息~~理解错误的现象~~。因此这时候，要求译者就在需要时对原句子中隐藏的语义成分进行辨识显化，看看~~。根据奈达的核心句分析法，运用词分析来确定每一个词语的词类，找出“实体”和“活动”，辨别出~~哪一个“实体”似乎“缺失”了“活动”，或哪一个“活动” 似乎“缺失”了“实体”等等~~，然后进而进行补充~~。经通过这一过程方法，译者便可会对原句文信息获得~~有一个~~清晰完整明了的认识，并在译文表达阶段对句中内隐、但译文上下文中需明白交代的语义成分进行合理兼必要的显化处理。以下是本论文翻译项目中的几个实例子：

例1： The typical reaction of scientists to the challenge of anomalous data to an existing theory is in fact to challenge the data first and, if the data prove reliable, then to complicate the theory just enough to accommodate the anomalous result (Nickerson, 1998).

此句主语部分为典型的嵌套结构~~层层嵌套~~，其中使得一些句法-语义子成分被隐藏。~~，译者难以充分准确理解原文信息。若要~~为正确理解原文，译者须在~~则应对这些隐藏信息加以显化。先~~对句子中的主要词语进行语义分类别分析的基础上，找出这些内隐的成分：

The typical reaction of scientists to the challenge [1] of anomalous data

活动连接实体连接活动连接实体

to an existing theory is in fact to challenge [2] the data first and,

连接实体连接状态连接活动实体状态

if the data prove reliable, then to complicate the theory just enough to

连接实体活动特征连接活动实体特征

accommodate the anomalous result (Nickerson, 1998).

活动实体

经词汇单位语义分类别分析后可发现：，“活动”词 ~~the~~ typical reaction、~~，the~~ challenge [1] 和 , prove 的施事者在句中都可直接找到~~有直接的体现~~，分别是 scientists、, anomalous data 和 the data，. 而“活动”词 challenge [2]、, complicate、，, accommodate 的施事者在句中并未没有直接体现，需要译者根据语境和及逻辑常识找判断出~~这些活动内隐的施事“实体”~~。经分析可知，译者判定 challenge [2]、complicate 和 accommodate内隐的内隐施事“实体”分别是 scientists、, scientists 和 , 及 the theory，并在此基础上将原句逆转换为~~显化在核心句中便可得~~下列出完整的核心句：

1) In fact Sscientists react to challenge [1] typically in [X and Y ways].

2) Anomalous data challenge [1] an existing theory.

3) ~~In fact (~~[sScientists]) challenge [2] the data first.

4) If the data prove reliable,. [scientists] complicate the [challenged] theory [to the Z extent]

~~(Scientists) complicate the theory.~~

5) ([The theory]) is just enough to accommodate the anomalous result.

核心句 2) 是对核心句 1) 中 challenge [1] 一词意义的命题展开或完整说明；核心句1) 中的 X 和 Y 分别代指核心句 3) 及核心句 4)、5)；核心句 4) 中的 Z 代指核心句 5)。句中方括号括起的部分中是内隐成分。在译文表达阶段，有些内隐语义成分（如核心句 4) 中的 [scientists]）在上下文里不言自明，因此并不需显化；但在作为核心句 4) 中 complicate 这一“动作”程度说明的核心句 5) 里，[The theory] 这一语义成分则需要显化（译成“其”），以使上下文意清楚且衔接自然显化的核心句的内隐成分。这样原文的意思就更加清楚，可避免译者对原文意思理解不清或扭曲。在表达译文时，有时也需要将内隐成分表达出来，使意思清晰明了。此句表达时便将6~~）中内隐的~~ ~~“理论”显化出来~~：。

译文： 事实上，在面对反常数据对质疑现有理论提出的质疑~~反常数据~~时，科学家们的典型反应是，先质疑数据，若数据证明可靠信，然后就修改再现有升级理论，使其精致复杂化到正好足够迁就~~，使理论能够容纳这一~~反常数据结果的程度（Nickerson~~尼克森~~，1998）。

比较： {译文中不显化核心句 5) 中内隐的 [The theory] 的译法} …… 若数据证明可靠，就精致复杂化现有理论到正好足够迁就反常数据结果的程度（Nickerson，1998）。

{KP: 此例原文理解与评注分析均错，建议完全删除。} 例2： Perhaps more surprisingly, similar difficulty in recognizing the logical validity of falsification was found in a sample of statisticians who had been formally trained in testing statistical (null) hypotheses and thus in examining possible disconfirming evidence.

~~此句句式层叠，内隐了许多主语，甚至同时内隐了主谓成分，造成译者难以准确把握原文信息。首先对原句中的各个词进行分类：~~

~~Perhaps more surprisingly, similar difficulty in recognizing the logical validity~~

状态实体连接活动特征实体

~~of falsification was found in a sample of statisticians who had been formally~~

连接实体活动连接实体状态

~~trained in testing statistical (null) hypotheses and thus in examining possible~~

活动连接活动实体连接活动

~~disconfirming evidence (Einhorn & Hogarth, 1978).~~

实体

~~对词汇分类后发现，“活动”的施事者在句中都没有直接体现，需要译者根据逻辑进行判断。而且，在句子的末尾 thus in examining possible disconfirming evidence~~ ~~中，thus~~ ~~后内隐了主谓成分 similar difficulty was found，显化后译者才能充分理解原文信息。经分析后可得出核心句：~~

~~1) ([People are]) more surprising.~~

~~2) ([A sample of statisticians]) found similar difficulty.~~

~~3) ([The sample of statisticians]) recognize the logical validity of falsification.~~

~~4) ([The sample of statisticians]) had been formally trained.~~

~~5) ([Statisticians]) test statistical ([null]) hypotheses.~~

~~6) Thus ([Sstatisticians found similar difficulty]).~~

~~7) ([Statisticians]) examine possible disconfirming evidence.~~

~~括号中都是原句内隐的成分，显化后便可对原文各部分的信息已十分清晰。特别是核心句 6)~~ ）~~中主谓成分的显化对译者充分理解原文信息至关重要。接下来对核心句之间的关系进行分析：核心句 2)~~ ）~~到核心句 6)~~ ）~~是核心句 1)~~ ）~~接下来要说的内容；核心句 2)~~ ）~~与核心句 6)~~ ）~~组成原句的主干信息，两者之间是因果关系；核心句 3)~~ ）~~是对核心句 2)~~ ）~~的时间限定，核心句 4)~~ ）~~是对核心句 2)~~ ）~~中 statisticians~~ ~~的修饰，核心句 5)~~ ）~~是对核心句 4)~~ ）~~的限定；核心句 7)~~ ）~~是对核心句 6)~~ ）~~的时间限定。至此，原文整体信息已十分清晰明了。在表达译文时，为使内容清晰，需将核心句 6)~~ ）~~中的内隐成分在译文中显化：。~~

~~译文：~~ 也许更令人惊讶的是，一组在检验统计学假设（零假设）方面受过正式训练的统计学家，在识别证伪的逻辑合理性时遇到了与此相似的困难，因此，他们在检测可能的否证证据时也同样会遇到这样的困难（Einhorn & Hogarth~~，1978~~）。

2.2 符合汉语习惯的译文表达

有些英语长句之所以长，是~~子偏长是~~因为其中包含了较多的并列成分或修饰成分。~~多而，连词使用频繁，但~~这些句子的其意思义并不复杂，译者所以理解起来~~原文并~~没有什么大的困难，。但是因~~有些长句会~~多次利用关系从句和词、非谓语结构等修饰连接手段而形成的，“~~一个套一个连用，形成~~迂回曲折、，层见叠出的长句结构”（柯平，1991，p. 102页）还是可能会在译文生成阶段对，译者产生无形~~在组织译文时易受原文句子结构~~的干扰影响，使得他们笔下的造成译文笨重冗长，背离~~逻辑表达不符合~~汉语的语篇常规或逻辑表达习惯习惯。

为避免发生这种情况，因此译者有必要应在分析原句核心句并彻底把握出核心句之间关系的基础上，~~将核心句~~按照汉语语篇常规与的逻辑表达习惯进行重组核心句的表层形式，以便生成译出自然地道的译文。~~而汉语逻辑表达的“主要手段就是各分句遵循客观世界时间顺序与空间顺序，以及主观世界的逻辑顺序”（彭长江、顾延龄，1999，p. 189）。，因本文是学术性论文，逻辑性较强，因此在对核心句进行重组时主要按照汉语的逻辑顺序组织译文。如~~以下便是本论文翻译项目中的几个实例子：

例1： Apart from exposing to criticism the various hidden stratagems that work against it, it is also important to combat what appears a common psychological tendency of humans to be biased towards confirmation, or against disconfirmation, a tendency that reflects a conflict between falsificationism and apparently deep-rooted psychological mechanisms.

此句的句子主干结构很清晰，即Apart from doing something, it is also important to do something，但次级结构~~附加成分~~较多，造成译者在重组织译文的表层形式时易受这些次级原文结构的影响。依照核心句分析法，译者首先~~按照上文找出核心句的方法~~找出此句的所有核心句：:

1) ([We]) expose the various hidden stratagems to criticism.

2) The various hidden stratagems work against ~~the~~ criticism.

3) Apart from [X], combating a common psychological tendency [Y] is also important.

4) Humans tend to be biased towards confirmation, or against disconfirmation.

5) The tendency reflects a conflict between falsificationism and apparently deep-rooted [human] psychological mechanisms.

该句~~是一个主从复合句，句子~~主干是核心句 3)）。X 代指~~复合句 X~~ ~~拆分成~~核心句 1) ）和 2)），Y 代指核心句 4) ）和 5) ~~）都是对 tendency~~ ~~的解释说明~~。~~由于修饰成分太长，~~为了使译文意思清晰明了，行文符合汉语表达习惯，译者应先将由核心句 3)、1) 和 2) 构成的句子主干部分即3~~）交代清楚，同时将~~以及作为核心句 3) 中动词宾语 a common psychological tendency 4~~）这个给出~~最重要信息的修饰成分的核心句 4) 译出。~~插入主句，~~核心句 5) 是核心句 3) 中动词宾语 a common psychological tendency 的同位语，作为次级修饰成分，其地位与核心句 4) 相当，但比核心句 4) 长许多，故译者将其独立出来，译成单独的~~其他修饰成分则再另起~~一句~~进行解释~~。如因此便得出了以下译文：

译文： 除了要批判各种潜藏的阻妨碍批判的花招诡计之外，反对与人类们偏好向确证实或忌讳反对否证的心理倾向作斗争也很重要——。这是一种~~常见的~~心理倾向十分普遍，它反映了证伪主义与~~据说根深蒂固的~~人类心理中显然是根深蒂固的一些机制之间的冲突矛盾。

例2： Although it can be argued that the confirmation bias helps both to protect our sense of self by rendering our preferred beliefs less vulnerable than they otherwise would be (Greenwald, 1980) and to guard science against indiscriminate acceptance of alleged new discoveries that fail to stand the test of time (Price, 1963), the bias is still generally regarded as a human failing. …

这是一个典型的主从复合句，~~此句的~~句子的主干脉络也很清晰，即：Although it can be argued that ... , the bias is still generally regarded as a human failing.，但由 although 引导的状语从句的实质主语为一个由 that 引导的主语从句，其谓语中间部分 ~~the confirmation bias~~ helps to ... and to ... 结构繁复，阵列很长，~~句子较长，~~译者如照此表层句法结构顺序译出，势必导致译文中从句过长（信息过度密集）而主句简短、整个句子头重脚轻、阵脚不稳的结果~~易受原文句子结构的影响，造成译文句子过长的现象~~。在实际翻译过程中，本文作者运用核心句分析法先找出全句中的所有核心句：

1) ([People]) can argue [X].

2) The Cconfirmation bias helps to protect our sense of self by [Y].

3) ([The Cconfirmation bias]) renders our preferred beliefs less vulnerable than they otherwise would be.

4) ([The Cconfirmation bias]) helps to guard science against indiscriminate acceptance of alleged new discoveries.

~~Science accepts alleged new discoveries.~~

5) ([The] alleged new discoveries) fail to stand the test of time.

6) [But the] Cconfirmation bias is still generally regarded as a human failing.

~~经分析可知，该句是主从复合句~~不难看出，以上核心句中，核心句 67) ）是主句，核心句 1) - 5) ）是从句；，从句命题和主句命题两者之间为是让步关系（句法表层上的连词 although 将主从两个分句连接在一起）。X 代指核心句 2) - 5)，Y 代指核心句 3)~~）到 6）是1）接下来要说的内容~~。核心句 2) - 5) 其中，2) ）和 4) ）是并列关系，核心句 3) ）是核心句 2) 所述活动）的方式，5~~）是 4）的受事，~~核心句 56) ）是对核心句 45) ）中 alleged new discoveries 性状的说明解释。核心句 1) 中活动词因 argue 的对象很多，涵盖包含了即2)、3)、4)、）到56) 核心句的内容。如将这五个核心句的内容都放在连词“尽管”后面顺序译出（“尽管 ……，……，……，这个偏好仍然……。”），整个句子会头重脚轻，有悖汉语句法常规，且让步状语从句和主句分别表达的两层意思间的逻辑关系也不够清楚直白。译者没有这样处理，而是根据核心句分析结果与~~）过长的全部，按照~~汉语的表达习惯，将原句中的状语从句（核心句 1) – 5)）译成一句话，将原句中的主句（核心句 6)）译成另一句话，状语从句和主句之间的让步逻辑关系用外位结构（独立于句外，同时又和句中某个成分所指相同的成分，也叫“外位语”）“但尽管如此”加以表达。由于状语从句较长，其中包含2) 和 4) 两个意思平行的并列核心句，译者先总说此句的~~可对其进行总结性的概括~~要旨从句，即“人们可以争辩说证实偏好对我们是有帮助~~有人说确证偏向是有好处~~的”，接下来再用“一方面，……；另一方面，……”的句式分别译出两个并列核心句所描述的两项具体的帮助形式。~~罗列其“好处”，~~整个译文如下：~~最后译出原文的主句，并用“尽管如此”表明与前一部分的逻辑关系。这样译文便会忠实通顺，符合汉语表达。~~

译文： 有人们可以争辩说确证实偏好对我们向是有帮助好处的：一方面，它可以使我们偏喜爱的信念不像这种偏好不存在的情况下那样易受攻击~~观点变得更加有力~~，从而保护我们的自我意识感知（Greenwald，, 1980）；，另一方面，它也可以防止科学不加区辨别地接受未经受住~~禁不起~~时间考验的所谓新发现（Price，1963）。但尽管如此，这个偏好仍然它被普遍被认为是人类自身的一个缺陷 ……~~的失败。~~

致 谢

摘 要