Memory for Prose: The Influence of Relational and Proposition-Specific Processing
散文的记忆：关系和命题特定处理的影响

In recent years there has been a growing body of literature in the verbal learning area that manifests a recurring theme: Manipulations that increase the difficulty, complexity, and/or novelty of verbal material produce better memory for that material. For example, sentences that are presented in transformed typography (upside down or backward) are better remembered than sentences that are presented in normal typography (Kolers, 1973; Kolers & Ostry, 1974; Masson & Sala, 1978), and syntactically complex sentences are better recognized than the same sentences presented
近年来，语言学习领域的文献越来越多，它们表现出一个反复出现的主题：增加语言材料的难度、复杂性和/或新颖性的操作会产生更好的材料记忆。例如，以转换排版（倒置或倒置）呈现的句子比以正常排版呈现的句子更容易记住（Kolers，1973 年;Kolers & Ostry， 1974;Masson & Sala， 1978），并且句法复杂的句子比呈现的相同句子更容易识别

in a simpler syntactic form (McDaniel, 1981). At the level of individual words, Jacoby (1978) increased the difficulty of identifying the second word of a related word pair by deleting letters from the second word. He found that if letters are deleted from the word, it is recalled better than if the word is presented intact. Also, a word presented as a difficult anagram is remembered better than if the word is presented as an easy anagram (Tyler, Hertel, McCallum, & Ellis, 1979). Furthermore, retention for intervals longer than a month appears to be enhanced by difficulty manipulations (Einstein, 1976; Kolers & Ostry, 1974). To explain these results, Lockhart, Craik, and Jacoby (1976) have postulated that if material is uncommon, unfamiliar, or difficult to process, many features will be analyzed, and this will result in a well remembered, rich encoding. Jacoby and Craik (1979) echoed this theme: “Difficulty necessitates more extensive processing, which then results in the formation of a more distinctive trace” (p. 12), with distinctiveness being positively related to memory.
以更简单的句法形式 （McDaniel， 1981）。在单个单词的层面上，Jacoby （1978） 通过删除第二个单词中的字母，增加了识别相关单词对的第二个单词的难度。他发现，如果从单词中删除字母，则比单词完整呈现时更容易被回忆起来。此外，一个词被呈现为困难的字谜比被呈现为简单的字谜更易被记住（Tyler， Hertel， McCallum， & Ellis， 1979）。此外，超过一个月的间隔保留似乎会因困难操作而得到增强（Einstein， 1976;Kolers & Ostry，1974年）。为了解释这些结果，Lockhart、Craik 和 Jacoby （1976） 假设，如果材料不常见、不熟悉或难以处理，将分析许多特征，这将产生记忆深刻、丰富的编码。Jacoby 和 Craik （1979） 回应了这一主题：“困难需要更广泛的处理，然后导致形成更独特的痕迹”（第 12 页），独特性与记忆呈正相关。

Research in memory for texts, however, has produced findings not easily integrated into
然而，对文本记忆的研究产生了不容易整合到
either the findings cited for individual words and sentences or the degree-of-elaboration theory that ties these findings together. Manipulations that increase the difficulty of comprehending a text do not seem to enhance memory for the text. For instance, Carter (cf. Carter, 1977) has reported that subjects who read a scrambled text recalled the text no better than subjects who read a normal version of the text. Similarly, Kintsch (1977) found that summarizations of stories with randomly ordered paragraphs were no better than summarizations of normally ordered stories, and for children, recall of scrambled stories was worse than recall of normal stories. More recently, Schwarz and Flammer (1981) found that for adults, memory for a fairy tale with scrambled sentence order was worse than memory for the fairy tale with normal sentence order. These results are puzzling when viewed within the context described initially. The empirical findings with words and sentences and the recent levels-of-processing framework articulated by Craik and his colleagues (Jacoby & Craik, 1979; Lockhart et al., 1976) lead to predictions inconsistent with the findings obtained with prose. Specifically, scrambling text should make processing more difficult, inducing more elaborative processing and, as a result, better memory. That this has not been borne out in previous studies raises several possibilities that should be explored.
要么是针对单个单词和句子引用的发现，要么是将这些发现联系在一起的阐述程度理论。增加理解文本难度的操作似乎并不能增强对文本的记忆。例如，Carter（参见 Carter， 1977）报告说，阅读乱序文本的受试者对文本的回忆并不比阅读正常文本的受试者更好。同样，Kintsch （1977） 发现，具有随机排序段落的故事摘要并不比正常排序的故事摘要更好，对于儿童来说，对混乱故事的回忆比对正常故事的回忆更差。最近，Schwarz 和 Flammer （1981） 发现，对于成年人来说，对句子顺序混乱的童话故事的记忆比对正常句子顺序的童话故事的记忆更糟糕。在最初描述的上下文中查看时，这些结果令人费解。Craik 和他的同事们阐述了单词和句子的实证发现以及最近的处理层次框架（Jacoby & Craik，1979;Lockhart et al.， 1976）导致预测与散文获得的发现不一致。具体来说，乱序文本应该使处理更加困难，从而引起更详细的处理，从而获得更好的记忆力。这在以前的研究中尚未得到证实，这提出了应该探索的几种可能性。

One possibility is that in previous studies comprehension may have been reduced as difficulty was increased. This appears to have been the case in Schwarz and Flammer (1981, Experiment 1) in which the difficult (scrambled) version of the story was rated as less comprehensible than the normal version. Since the comprehensibility of text influences recall (Dooling & Lachman, 1971); it is possible that in previous work inequivalent comprehension levels between difficult and “easy” passages have masked the mnemonic effects of increased difficulty. Tentative support for this idea is provided by the finding that increased memory resulting from the difficulty of a sentence depends on the sentence being understood (Auble & Franks, 1978).
一种可能性是，在以前的研究中，理解能力可能随着难度的增加而降低。在 Schwarz 和 Flammer（1981 年，实验 1）中似乎就是这种情况，其中故事的困难（混乱）版本被评为比正常版本更难理解。由于文本的可理解性会影响回忆（Dooling & Lachman， 1971）;在以前的工作中，困难和“容易”段落之间的不等价理解水平可能掩盖了难度增加的助记效应。这个观点的初步支持是由一个发现提供的，即句子的难度导致的记忆增加取决于被理解的句子（Auble & Franks， 1978）。

Perhaps theoretically more interesting are two other hypotheses for why scrambled prose has not produced mnemonic effects that par-
也许从理论上讲更有趣的是另外两个假设，说明为什么乱序散文没有产生 par-
allel those found for words and sentences. One hypothesis is that prose memory is dependent on different processes than those that operate for isolated sentences and words. The difficulty of a text may not influence the degree to which it is distinctively encoded; for example, content may play a more prominent role in this regard. Inasmuch as retrieval of information from texts is top-down, that is, guided by schemata (Anderson & Pichert, 1978), scripts (Bower, Black, & Turner, 1979), or other macrostructures (Kintsch, 1977), the distinctiveness with which particular ideas are encoded may not be an important consideration for text recall. If these conjectures were supported, it would imply that the domain of applicability for levels of processing principles is restricted to simpler kinds of verbal material. Such a finding would be consistent with Jenkins’s (1979) position that mechanisms of memory are affected by characteristics of the learner, the material, the test task, and the orienting task. In the present context, changing the verbal material would be seen as producing a change in the characteristics of memorial functioning.
allel 那些为单词和句子找到的。一种假设是，散文记忆依赖于与孤立句子和单词不同的过程。文本的难易程度可能不会影响其独特编码的程度;例如，内容在这方面可能发挥更突出的作用。由于从文本中检索信息是自上而下的，也就是说，由图式（Anderson & Pichert，1978）、脚本（Bower、Black和Turner，1979）或其他宏观结构（Kintsch，1977）指导，特定思想被编码的独特性可能不是文本回忆的重要考虑因素。如果这些猜想得到支持，则意味着处理原则层次的适用范围仅限于更简单的口头材料。这样的发现与 Jenkins （1979） 的立场一致，即记忆机制受学习者、材料、测试任务和定向任务的特征的影响。在目前的情况下，更改口头材料会被视为导致纪念活动功能特征的变化。

An alternative to the above hypothesis rests on the observation that scrambling the paragraphs or sentences in a story is a different manipulation from that used with words and sentences (deleting letters or inverting typography). Recently, McDaniel (in press) found that deleting letters from words expressing certain events in a story enhances the recall for those events. Thus, despite the relatedness of the various difficulty manipulations at an abstract level, it appears that not all manipulations that increase difficulty necessarily produce elaborative processing that is quantitatively or qualitatively similar (e.g., Wilkinson, Guminski, Stanovich, & West, 1981). The differences in the kind of elaboration induced by various difficulty manipulations may prove critical for the kind of memorial effect that is observed (Wilkinson et al., 1981). To the extent that the information elaborated as a function of a particular difficulty manipulation tends to be obligatorily encoded, the difficulty manipulation may have minimal impact on memory performance. The difficulty manipulation will have a positive effect on memory performance only to the extent that it produces more extensive processing of
上述假设的另一种选择是基于这样一个观察，即对故事中的段落或句子进行打乱是与用于单词和句子的操作（删除字母或反转排版）不同的操作。最近，McDaniel （正在出版）发现，从表达故事中某些事件的单词中删除字母可以提高对这些事件的回忆。因此，尽管在抽象层面上各种难度操作具有相关性，但似乎并非所有增加难度的操作都必然会产生数量或质量相似的精细处理（例如，Wilkinson， Guminski， Stanovich， & West， 1981）。由各种困难操作引起的阐述类型的差异可能被证明对观察到的纪念效果类型至关重要（Wilkinson et al.， 1981）。在某种程度上，作为特定难度操作的函数精心编制的信息往往是强制性编码的，因此难度操作对内存性能的影响可能很小。难度操作只会对内存性能产生积极影响，前提是它会产生更广泛的
information that is ordinarily only minimally processed. Two experiments are reported that investigate these hypotheses.
通常只进行最低限度处理的信息。据报道，有两个实验调查了这些假设。

In this experiment we used a Grimm Brothers’ fairy tale as our prose passage. Two features of the design were central to examining the three positions outlined above. First, in addition to increasing the difficulty of the story by scrambling the sentences, we used a difficulty manipulation designed to increase processing of information that is usually only minimally processed: individual lexical items. To produce more extensive processing of individual lexical items, some of the letters from each word in the story were deleted. Assuming that the letter-deletion manipulation encourages processing of information only minimally processed ordinarily, while scrambling sentences encourages the type of schema-based processing usually employed in story comprehension (Kintsch, 1977), letter deletion was expected to enhance retention more than scrambling sentences. Second, an attempt was made to promote approximate equal levels of comprehension for the different versions of the fairy tale. This was done by requiring subjects in the scrambled-text condition to rearrange randomly ordered slips of paper containing the sentences of the text into a meaningful order. Subjects in the letter-deletion condition were required to fill in the letters that had been deleted.
在这个实验中，我们使用格林兄弟的童话故事作为我们的散文段落。该设计的两个特征是检查上述三个位置的核心。首先，除了通过打乱句子来增加故事的难度外，我们还使用了一种难度操作，旨在增加对通常只进行最低限度处理的信息的处理：单个词汇项目。为了对单个词汇项目进行更广泛的处理，删除了故事中每个单词的一些字母。假设字母删除操作鼓励对通常只进行最低限度处理的信息进行处理，而打乱句子鼓励故事理解中通常采用的基于图式的处理类型（Kintsch， 1977），字母删除预计比打乱句子更能提高保留率。其次，试图促进对童话故事的不同版本的理解水平大致相等。这是通过要求在乱序文本条件下的受试者将包含文本句子的随机排序的纸条重新排列成有意义的顺序来完成的。信件删除条件中的受试者需要填写已删除的信件。

Subjects. Subjects were 72 male and female undergraduate students, 32 from the University of Notre Dame and 40 from Furman University. All were in general psychology courses and received credit for their participation. Subjects were tested in groups of one to four.
科目。受试者为 72 名男女本科生，其中 32 名来自圣母大学，40 名来自福尔曼大学。他们都参加了普通心理学课程，并因参与而获得学分。受试者以 1 至 4 人为一组进行测试。

Materials. A Grimm Brothers’ fairy tale entitled “The Old Man and His Grandson” was used in the present experiment (taken from Rackham, 1979). This fairy tale contained 14 sentences. Informal observation found it to be unfamiliar to subjects. Four versions of the fairy tale were prepared by factorially combining two levels of a sentence-order variable (normal or random) with two levels of a word-integrity variable (letters intact or deleted). Thus, one version contained sentences in normal order and intact words. In a second version, the sentences were randomly ordered and the words remained intact. A third version was typed with sentences in normal order but approximately $30\mathrm{\%}$$30\mathrm{\%}$30%30 \% of the letters (of these, $42\mathrm{\%}$$42\mathrm{\%}$42%42 \% were vowels) were
材料。本实验中使用了格林兄弟的童话故事《老人和他的孙子》（摘自 Rackham，1979 年）。这个童话故事包含 14 个句子。非正式观察发现受试者不熟悉它。童话故事的四个版本是通过因子组合一个句子顺序变量的两个级别（正常或随机）和两个级别的单词完整性变量（字母完整或删除）来准备的。因此，一个版本包含正常顺序的句子和完整的单词。在第二个版本中，句子是随机排序的，单词保持不变。第三个版本是按正常顺序输入的句子，但大约 $30\mathrm{\%}$$30\mathrm{\%}$30%30 \% 有字母（其中是 $42\mathrm{\%}$$42\mathrm{\%}$42%42 \% 元音）是
replaced by blanks. The letters were deleted in haphazard fashion with at least one letter deleted from each word of two or more letters. Through pilot work wth 8 subjects, the letter deletions were adjusted so that on average, more than $98\mathrm{\%}$$98\mathrm{\%}$98%98 \% of the words were correctly identified by the subjects. A fourth version was typed so that the sentences were randomly ordered and the letters deleted as above. For the randomly ordered versions, two random orders were employed, and they were counterbalanced across the subjects.
替换为空格。这些字母被随意删除，两个或多个字母的每个单词至少删除了一个字母。通过与 8 名受试者的试点工作，调整了字母删除，以便平均而言，受试者正确识别了多个 $98\mathrm{\%}$$98\mathrm{\%}$98%98 \% 单词。输入第四个版本，以便句子随机排序并删除字母，如上所述。对于随机排序的版本，采用了两个随机顺序，它们在受试者之间进行了平衡。

Design and procedure. Story version was varied between subjects, with 18 , subjects assigned to each of the four conditions. In order to maintain an incidental learning situation, subjects were told that the purpose of the experiment was to investigate story comprehension. Subjects were presented with the fairy tale and instructed to read it. Each of the 14 sentences was typed on a separate slip of paper for the subjects given the random orders. These subjects were informed that they would have to rearrange the sentences (i.e., the slips of paper) in order to make sense of the story. Subjects given the versions with missing letters were instructed to fill in the missing letters, All subjects were encouraged to take their time and use the title to help in the task of making the story understandable. Subjects were timed on the initial task, and the order in which they rearranged the random versions was recorded. After subjects finished processing the fairy tale, they were given a Likert-type comprehension rating scale with instructions to rate their comprehension from didn’t comprehend the story at all (1) to comprehended the story very well (5). When they completed this scale, the subjects were given math problems to solve for 5 min . They were then asked to write down as much of the exact story as possible. Once again, all subjects were urged to take their time and were allowed as long as they wished for the task.
设计和程序。故事版本因受试者而异，有 18 个受试者分配给四个条件中的每一种。为了保持偶然的学习情境，受试者被告知实验的目的是调查故事理解。向受试者展示童话故事并指示他们阅读它。这 14 个句子中的每一个都打在一张单独的纸条上，供给定随机顺序的受试者使用。这些受试者被告知，他们必须重新排列句子（即纸条）才能理解故事。给出缺失字母版本的受试者被指示填写缺失的字母，鼓励所有受试者花时间使用标题来帮助完成使故事易于理解的任务。受试者在初始任务上计时，并记录他们重新排列随机版本的顺序。在受试者完成童话故事的处理后，他们得到了一个李克特式的理解量表，并附有指示来评估他们的理解能力，从根本不理解这个故事 （1） 到很好地理解这个故事 （5）。当他们完成这个量表时，受试者被分配数学问题来解决 5 分钟。然后他们被要求尽可能多地写下确切的故事。再一次，所有受试者都被敦促慢慢来，只要他们愿意，就可以完成任务。

Comprehension. For all analyses the rejection level was set at .05 . The processing times (Table 1) revealed that the experimental manipulations were effective in producing versions of the fairy tale that were more difficult to process. An analysis of variance (ANOVA) with sentence order (normal or random) and word integrity (letters intact or deleted) as factors indicated that subjects took significantly more time to read the fairy tale when sentences were in random order than when the sentences were in normal order, $F(1$$F(1$F(1F(1, 68) $=228.56,M{S}_{\mathrm{e}}=23,614.11$$=228.56,M{S}_{\mathrm{e}}=23,614.11$=228.56,MS_(e)=23,614.11=228.56, M S_{\mathrm{e}}=23,614.11. (The analysis was carried out with seconds as the unit of measure.). Subjects also needed more time when letters were deleted than when letters were not deleted, $F(1,68)=457.80,M{S}_e=$$F(1,68)=457.80,M{S}_e=$F(1,68)=457.80,MS_(e)=F(1,68)=457.80, M S_{e}= 23,614.11. Finally, there was a reliable interaction between the two variables such that the effects of word integrity and sentence order
理解。对于所有分析，拒绝水平都设置为 .05 。处理时间（表 1）表明，实验操作在产生更难处理的童话版本方面是有效的。以句子顺序（正常或随机）和单词完整性（字母完整或删除）为因素的方差分析 （ANOVA） 表明，当句子按随机顺序排列时，受试者阅读童话故事的时间明显多于句子按正常顺序 $F(1$$F(1$F(1F(1 排列时，68）。 $=228.56,M{S}_{\mathrm{e}}=23,614.11$$=228.56,M{S}_{\mathrm{e}}=23,614.11$=228.56,MS_(e)=23,614.11=228.56, M S_{\mathrm{e}}=23,614.11 （分析以秒为计量单位进行。删除字母时，受试者也比未删除字母时需要更多的时间 （ $F(1,68)=457.80,M{S}_e=$$F(1,68)=457.80,M{S}_e=$F(1,68)=457.80,MS_(e)=F(1,68)=457.80, M S_{e}= 23,614.11）。最后，两个变量之间存在可靠的交互作用，因此单词完整性和句子顺序的影响