Text genres and mental models: What readers get from text and how they use it. Keis Ohtsuka The University of Western Australia Nedlands, WA 6009 AUSTRALIA Running head: Text genres and mental models This research was funded by a grant from the Australian Research Council. Paper presented at the Australian Association for Research in Education (AARE) Annual Conference, Fremantle, Western Australia, on November 23, 1993. Abstract This study examined: (1) how and to what extent text genres influence the characteristics of mental models readers build from text; (2) whether the characteristics of mental models change if the subjects know the nature of questions that will follow the reading session. Seventy undergraduates, read one of the three text genres, descriptive, procedural, or narrative that were based on the same underlying information and later answered three types of inference questions about spatial information. The three types of inference questions were constructed so that each question type would be more similar than the other two question types to the characteristics of a model derived from the text. Accuracy scores and response time data showed that the mental models derived from the descriptive text showed two- dimensional characteristics, whereas the models built from the other two genres were one-dimensional. The results were similar to those from the previous study (Ohtsuka, 1990). When the subjects were asked to read a different text genre for the second time, however, the effect of text genre became statistically nonsignificant. The results indicate that the subjects were able to build optimal mental models for answering questions when they could anticipate the content of questions. The findings suggest that mental models built from text are flexible representations that readers actively manipulate rather than static, rigid structures that simply hold the textual information for retrieval. Text genres and mental models: What readers get from text and how they use it. How do readers choose an optimal mental model when they read? Are the characteristics of the mental models derived from text different depending on the text genre that readers process? Or are the mental models derived from text similar regardless of the input text genres so long as the content information is similar? To address these questions, this study examines readersÕ mental models that they create when reading different text genres. The aims of the study are to examine: (a) how accurately readers build a spatial mental model from text written in different text genres; and (b) whether the characteristics of mental models change if the subjects know the nature of questions that will follow the reading session. The differences among text genre types and their underlying structures have recently attracted the attention of researchers studying the text comprehension processes. These researchers claim that the characteristics of representations that underlie a text differ for each of the three text genres: descriptive, narrative, and exposition (e.g., Brewer, 1980; Johnson-Laird, 1983). This claim suggests that the characteristics of mental models that readers build from one text genre would be different from the mental models that they form from other genres. According to these researchers, descriptive text written from a birdÕs eye perspective embodies a two-dimensional spatial array. The mental models constructed from this type of text will be based on an external coordinate system (north, south, east, west) where landmarks are distributed across the landscape. Narrative text embodies a series of events that occur in time. A typical subclass of narrative text includes stories whose discourse force is to entertain readers (Brewer, 1980; Brewer & Lichtenstein, 1981, 1982). The mental models built from narrative text will be sequential with an internal coordinate system built with respect to the characterÕs point of view. Expository text represents the underlying abstract logical processes in the linguistic form. The underlying mental processes involved are presumably induction, deduction, classification, and comparison. This theory postulates different underlying representations for genres and clearly suggests that different types of text genre will produce different types of mental models. So far, the theoretical claim on the effect of text genres to the characteristics of mental models has not been fully substantiated by empirical research as little experimental research has been carried out with the explicit purpose of comparing different mental models derived from different text genres. However, the results from several previous studies appear to suggest that readersÕ mental models are indeed shaped partly by their assumption about text genres and what is relevant to the comprehension of a given text. Using two types of text describing a small town, Perrig and Kintsch (1985) tested subjects for recognition of the verbatim experimental text and for information that the text did not present explicitly. Spatial information was given in geographic terms in the survey text, whereas the equivalent information was given in the form of driving instructions in the route text. Subjects were faster and more accurate in verifying the statements about text when the statements were given in the same form as the original text than when they were given in the other form. These results suggest a correlation between different text types and differen mental models. However, evidence appears somewhat inconsistent as to whether a readerÕs perspective is retained in a mental model. Reporting a series of similar studies, Taylor and Tversky (1992) claim that a perspective of text is not preserved in mental models. Fundamentally, the results in both studies were not entirely conclusive regarding the effect of text type because an internal coordinate system (front, back, left, right of the narratorÕs position) was used in the route text and an external coordinate system (north, south, east, and west of already mentioned locations) was used in the survey text. Therefore, the effect due to the difference of coordinate systems was confounded with the effect due to the different text types. Morrow (1985, 1986) investigated how readers infer referents of ambiguous pronouns in descriptive and narrative texts. He found that readers use the thematic role of character to determine the referents of ambiguous pronouns in narrative text, whereas they use the information of thematic places in descriptive text. MorrowÕs results suggest that readers use substantially different information when they integrate sentences into a coherent mental representation depending on the text genre they are processing. Readers rely on the knowledge of places or characters depending on the text genre of experimental texts when they determine ambiguous referents. If different information is used to integrate sentence-level information to a mental model, it follows logically that the prominent characteristics of mental models being generated might well be different depending on text genres. Ohtsuka (1990) has shown that readers of procedural and narrative texts build one-dimensional mental models, whereas readers of descriptive texts build two-dimensional mental representations from the experimental texts. In addition, he found that the direction of order inference questions does not influence accuracy in the descriptive text group. In the other two text groups, however, accuracy for the inference questions stated in the direction opposite to the text was significantly lower than those stated in the direction of the text. Overall, OhtsukaÕs results suggest that the main characteristic of the mental model derived from descriptive texts is that it is two-dimensional. On the other hand, both procedural and narrative texts produce primarily one-dimensional mental models. He also postulated that the initial characteristic of the mental model derived from a text during the reading session can be modified in other forms as they make inferences afterwards. The present study examines: (1) whether similar data trends can be obtained by presenting a new set of experimental passages to a new group of English native speakers; (2) whether the characteristics of mental models change when subjects are instructed to re- read the experimental passage knowing what types of questions they answer after the reading session. The Current Study The emphasis in the present study is to allow readers to read experimental texts as they usually would in natural settings until they feel comfortable answering questions regarding the content of the text. This procedure ensures readers use whatever knowledge they have about text genres to build mental models about the text during the natural reading process. The experimental texts are designed to represent the essential characteristics of text genres and serve as their plausible representatives. The study specifically examines the mental models that readers build from text information without any pre-learning of an external spatial model before reading. Rarely in our daily reading, do we have an opportunity to study the spatial background before we start reading the text. In many cases, the mental model that a reader builds from text is not necessarily complete, nor is it the only possible interpretation of the text. Taking into account the complex nature of reading as it is conducted in the natural context, the current study focuses on the nature of spatial mental models constructed from textual information only. Furthermore, the current study attempts to separate any effects due to the protagonistÕs perspective from those due to text genres. Previous theoretical work on text genres and underlying mental models (e.g., Brewer, 1980) as well as experimental research (Perrig & Kintsch, 1985) confounded the perspective difference (omniscient birdÕs eye point of view versus first person) and the coordinate system difference (internal versus external). To separate the two effects, the experimental passages in this study use a common (external) coordinate system. The current study uses a spatial model paradigm, which was originally used in Ohtsuka (1990). Subjects read a text from one of the three types of text genres, descriptive, procedural, or narrative. These texts were designed to contain identical underlying information about the spatial layout of a fictitious town. By comparing the different mental models that readers constructed from each genre type, the study examined how genre could influence the characteristics of mental models of text information. The descriptive text described the global configuration of a hypothetical town and the locations of various landmarks from a birdÕs eye perspective. This text used an external coordinate system (north, south, east, west) and landmarks were distributed neutrally in the text. It was hypothesized that readers of the descriptive text build a global spatial model as the mental representation is generated from the text. The procedural text described a series of instructions for driving through the same hypothetical town. Although the procedural text in this study used an external coordinate system, it was hypothesized that readers of the procedural text build a sequential, one-dimensional route model that contains a series of instructions and orientation information along the route. The narrative text described the trip of a woman and her dog through the same hypothetical town and a series of events that happened along the way. The narrative text used in this study was written as a story that produces affective response in readers in compliance with the structural-affect theory of story comprehension (Brewer & Lichtenstein, 1981, 1982). The narrative text embodies a series of events organized along a time line. It was hypothesized that readers of the narrative text build an event model from the characterÕs point of view. This type of model derived from the narrative text is also a sequential, one-dimensional model but is constructed from the protagonistÕs point of view. The differences in the characteristics of the mental models generated from the three text genres were measured by asking questions about information from the underlying models that was only available from inference. To test readersÕ success at constructing mental models, three types of inference questions were used: global perspective questions, spatial orientation questions, and order inference questions. Global perspective questions and spatial orientation questions represent two different ways of testing spatial information.1 Global perspective questions are questions regarding the spatial configuration among pairs of landmarks viewed from a global perspective, for example, "Is Landmark A north of Landmark B?" One can answer global perspective questions directly by examining a two-dimensional spatial model and making spatial inferences from a global perspective. Spatial orientation questions test the spatial orientation inferences about landmarks with respect to a given ground-level location, for example, "If youÕre at Location A and look south, would you see Landmark B?" One can answer this type of question by imagining oneself to be at the indicated location and then making a spatial orientation inference. Order inference questions assess the event order inference with respect to the underlying spatial model. For this study, these questions were generated by translating the relation between a pair of landmarks into a hypothetical trip along a specified route, for example, "If you are heading south, would you see Landmark A first and then Landmark B?" By reversing the trip direction, the relations between Landmark A and Landmark B can be tested in yet another question: "If you are heading north, would you see Landmark A first and then Landmark B?" Subjects can answer both types of inference questions by imagining a hypothetical trip along a specified route using the spatial mental model built from the text. Response time and accuracy are the dependent variables in this study. The basic underlying methodological assumption is that any extra cognitive processing effort needed to manipulate a mental model constructed from the text requires additional processing resources. Consequently, the more steps the readers need to transform the mental model built from the text in order to answer inference questions, the longer it should take for them to answer questions and the less accurate should be the answers. According to this processing model, the closer the form of the constructed mental model from the text and the mental model required to answer the inference question, the faster and more accurate readers will be in answering the question. Therefore, the present study addresses the following specific hypotheses: 1. Readers of the descriptive text will be faster and more accurate in answering global perspective questions than will readers of the procedural text or narrative texts. For readers of the descriptive text, the answers for global perspective questions concerning spatial information can be obtained directly from the original mental model. On the other hand, readers of the procedural or narrative texts will have to transform a route model or an event model to a global spatial model to answer this type of question. This additional processing will increase both response time and error for the procedural and narrative text readers. 2. Readers of the procedural text will be faster and more accurate in answering spatial orientation questions than will readers of the descriptive or narrative texts. Readers of the procedural text should construct a route model in which a series of instructions and orientation information are represented. Therefore, they will be able to answer spatial orientation questions directly from their original mental model built from the text. Readers of the descriptive text will need to move perspectives in their two-dimensional spatial model to obtain orientation information. Similarly, readers of the narrative text will need additional processing to compute orientation information because, to them, only the salient information, a story plot and the event order, is readily available from their mental model. 3. Readers of the procedural text and the narrative text will be faster and more accurate in answering order inference questions than will readers of the descriptive text. Readers of the procedural or narrative texts will build one-dimensional sequential models and therefore be readily able to retrieve order inference information. Readers of the descriptive text, on the other hand, will need extra inference steps to answer questions of this type from their two-dimensional global model. 4. One-dimensional mental models will be sensitive to the directionality of order inference. Readers of the procedural text and the narrative text will be faster and more accurate in verifying order inference statements when a statement is presented in the order consistent with the direction of underlying representation than when presented in the opposite direction. For example, if the text describes a trip from an airport to hotel, the statement "If you drive from the airport to the hotel, you would see Landmark A then Landmark B" (in the order consistent with the underlying representation) should be verified faster and more accurately than the statement "If you drive from the hotel to the airport, you see Landmark B then Landmark A" (in the order opposite to the underlying representation). 5. When the reading session is repeated, presenting a text of a different text genre, subjects will no longer show any effect of text genre on their mental model. Readers will build a spatial model regardless of the text genre because they are already familiar with the types of questions that will follow the reading session. Method Subjects Seventy undergraduate psychology students participated in the study to fulfill a course requirement. They were all required to be native speakers of English to participate in the study. Materials Three types of passages (descriptive, procedural, and narrative), each written in two forms (north to south and south to north), were constructed from the same underlying information. Each passage was designed to represent information about the same imaginary town whose map is given in Figure 1 (no subjects saw this map in the experiment). ÑÑÑÑÑÑÑÑÑÑÑÑ Insert Figure 1 about here ÑÑÑÑÑÑÑÑÑÑÑÑ Descriptive passages. The descriptive passages described a birdÕs eye view of an imaginary town. The global layout of the town was given first, followed by its large scale structure. After the global information, a series of statements about specific landmarks in the town was introduced. The location of each landmark was described in terms of the relevant street names (i.e., on Third and Orange Street), but the exact location of the landmark within a street block was not explicitly specified. The descriptive passages described the landmarks in an orderly manner, from one end of the town toward the other end (either from North to South or South to North), so that the reader could form a coherent mental model when each statement was processed without violating the immediate integration principle. This principle states that discourse comprehension is facilitated by introducing new information in discourse in such a way that it can be integrated immediately into an already constructed underlying structure (Ehrlich & Johnson-Laird, 1982; Ohtsuka & Brewer, 1992). Procedural passages. The procedural passages consisted of instructions on how to navigate from Landmark A to Landmark B in an imaginary town. The procedural passages addressed the reader as "you." First, the passages introduced the setting of the town. Then, the purpose of the passages, to instruct readers about a route to a destination yet to be described, was explicitly given to readers. The procedural passages were designed so that if they were read from the beginning they would guide the reader to the destination with no need to backtrack. Narrative passages. The narrative passages described a walk by a woman bringing her dog to a veterinary hospital from her home (or taking her dog from the veterinary hospital to her home). On the way, the character loses her dog and finds it again after an anxious search. The route in the narrative passages was the same route described in the procedural passages. Examples of the passages are given in Appendix A. Inference questions. As noted earlier, three types of questions were used to assess readersÕ inferences: global perspective questions, spatial orientation questions, and order inference questions. The answers for all questions were not available verbatim from the experimental passages. To answer questions, subjects had to make inferences by building a mental model. There were two different randomizations of question item order for each of the three types of questions. The global perspective questions consisted of 18 true-false statements testing the global spatial configuration among pairs of landmarks in the passage. The spatial orientation questions consisted of 18 true-false statements testing the spatial orientation inferences about the locations of landmarks with respect to a given ground level location. The question items were carefully examined to check their validity using world knowledge regarding the line of sight between the observer and the object, the relative height of the observer's position and the object, and the visibility of the object from the distance. Items that yielded multiple answers were eliminated. The order inference questions consisted of 18 true-false statements testing order inferences when a hypothetical trip was carried out across the town described in the passage. Half of the order inference questions described the hypothetical trip in one direction while the other half described the trip in the opposite direction. Because there is more than one way to navigate from Landmark A to Landmark B, the route information has to be provided in an order inference question to determine a particular route. Most distant landmark pairs were eliminated because it was not possible to determine a route without giving more than two street names. Appendix B presents examples of the three question types. Apparatus An Apple Macintosh personal computer controlled the presentation of the experimental texts, the display of inference questions and the measurement of response time to answer each question. The computer programs for the experiment were written by the author in HyperTalk and incorporated the timer and video stimulus synchronization routines by Pugh (1991). The measurement was accurate to the nearest millisecond2. Design and Procedure There were three types of text genre and three orders of question block presentation as between-subject factors, and three types of inference questions as a within-subject factor in a factorial design. The three types of questions were given in block format and were counterbalanced.3 Subjects were randomly assigned to one of the nine experimental conditions. Reading of text on the computer screen. Subjects participated in the experiment individually. They were told that they would be asked to read a passage on a computer and to answer questions about its content later. The subjects were allowed to read the text for as long as they wished until they felt ready for the questions. The subjects were not allowed to take notes or draw maps during the reading session. Once they started responding to the questions, the subjects were not allowed to look back at the text. Reading time was recorded from the onset of the experimental text presentation until the subjects started to answer the questions. Key press practice trial. Before the question session, the subjects read brief instructions on how to press keys to respond to questions on the computer display. After the instructions, five practice questions were presented. The instruction display and the key press trials also served as the intervening task to reduce effects due to possible rehearsal of the experimental texts. The practice question items were unrelated to the content of the experimental passages or the inference questions. Inference question session. After key press trials, the three types of 18 questions each were administered in a block format on a computer display. The order of question item presentation within each question type was randomized for each subject. Subjects were asked to verify a series of statements regarding the hypothetical town. All subjects verified all three types of inference statements. The statements were displayed on the computer screen one at a time. The subjects responded by pressing a colour-coded yes key or no key on the keyboard. They pressed the yes key with their dominant hand and the no key with their nondominant hand. The key press as well as the response time from the onset of the presentation of the statement until the key press was recorded. Subjects did not receive feedback concerning the accuracy of their answers, nor about the speed of their responses. Post-experimental questionnaire. After the question session, subjects completed a brief questionnaire assessing their introspective observations regarding reading strategies, perspectives taken during the reading session, and the inference question session. The second reading session and question assessment. After the completion of Part 1, subjects were randomly assigned to one of the other two text genre conditions and were instructed to read a second passage on the computer display. They were told that experimental procedures were exactly the same. No explicit information was given that the second experimental passage was based on the same spatial model. Subjects completed the tasks in the same order as described, including the post- experimental questionnaire. Results and Discussion Passage Reading Time The average reading times for the descriptive text group, the procedural text group, and the narrative text groups were, in order, 9 min 34 s, 8 min 31 s, and 9 min 16 s. There were no statistically significant differences among the three genre groups. Accuracy Scores The mean accuracy scores for each of the three question types were calculated for each subject by adding correct responses. The maximum accuracy score was 18 and a chance score was 9.4 Percentages of correct responses were used for analysis. Response Time Data A logarithmic transformation was applied to normalize raw response time data. One hundred twenty-nine miss key presses, 2.1% of the total data of 5,950 responses were removed from the data set. Means and standard deviations were calculated for each type of question. Outliers exceeding the means plus or minus three standard deviations were removed from the analysis. Thirty-nine outliers, 0.67% of the total data of 5,821 responses, were excluded. All response time data given in the article are transformed data. Table 1 shows the mean accuracy scores and mean response times for each of the three types of inference questions by genre. ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ Insert Table 1 about here ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ Overall results. On the accuracy scores, repeated-measure analysis of variance (ANOVA) indicated a statistically nonsignificant interaction of text genre and question type, F(4, 122) = 1.16, MSe = 0.017, p = .33. The effects of text genre and question block were not statistically significant. There was a statistically significant two-way interaction of text genre and question block, F(4, 61) = 2.54, MSe = 0.096, p < .05, indicating a trivial effect of the overall reduced comprehension in two question types in one of the three question block presentation orders. On the response times, the two-way interaction of genre and question type was significant, F(4, 122) =3.22, MSe = 0.007, p < .05. This interaction, critical data for this research, will be explained in full detail in the following section. In addition, there was a significant effect for question type, F(2, 122) = 222.26, MSe = 0.007, p < .001, indicating a trivial effect of the average question length for each question type. There was also a significant two-way interaction of question block order and question type, F(4, 122) = 11.26, MSe = 0.007, p < .001, indicating a trivial effect of question block permutation and subjectsÕ learning effect within a question block. There was no evidence of speed-accuracy trade off in the data. Two-way interaction of genre and question type. The two-way interaction of genre and question type on response time data is given in Figure 2. The results indicate that response times to the three question types varied according to the text genre. The underlying processing hypothesis predicts this interaction because the characteristics of mental models derived from one text genre are more compatible with one type of question but less with the other types. ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ Insert Figure 2 about here ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ On the order inference questions, as predicted in Hypothesis 3, both the procedural text group, F(1, 61) = 6.66, MSe = 0.021, p < .05 and the narrative text group, F(1, 61) = 6.55, MSe = 0.021, p < .05 were faster than the descriptive text group (3.99 vs. 4.10 for both). The results suggest that the mental models derived from the narrative and procedural texts were more compatible with the order inference questions. The differences between the narrative text group and the procedural text group were not statistically significant. Directionality in order inference. To test the hypothesis concerning the directionality in order inference, order inference question data were analyzed further by contrasting the direction of the order inference with those of mental models built from the text. For this sub-analysis, there were three types of text genre and three orders of question block presentation as between-subject factors, and two directions of order inference question as a within-subject factor in a factorial design. All responses, both correct and incorrect ones, were used in this analysis. Directionality on accuracy. As predicted in Hypothesis 4, a two-way interaction of the direction of order inference and genre was statistically significant, F(2, 61) = 7.13, MSe = 0.04, p < .001. Both the procedural, F(1, 61) = 27.47, MSe = 0.07, p < .001, and the narrative text groups, F(1, 61) = 6.79, MSe = 0.07, p < .05, were more accurate in answering the order inference questions that were consistent with the direction of the order of presentation in the passage (85.4% and 77.3% respectively) than those that were inconsistent (56.4% and 63.3%, respectively). In contrast, the descriptive text readers did not show statistically significant differences between the order inference questions that were consistent with the direction of the text (69.9%) and those that were inconsistent (70.3%). When the order inference questions were formulated in a direction consistent with the mental model built from the text, both procedural and narrative text readers were better able to answer them. However, when the order inference questions were formulated in a direction inconsistent with the mental model built from the text, the procedural and narrative readers made more errors.5 Directionality on response times. The two-way interaction of genre and the direction of order inference did not reach a statistical significance level, F(2, 61) = 0.31, MSe = 0.006, p = .735. The Second Reading Session Table 2 shows the data obtained after the second reading session. The accuracy data showed a statistically significant effect of question type, F(2, 118) = 3.53, MSe = 0.014, p < .05. Although the overall average accuracy of spatial orientation questions appeared lower than the average counterparts of the other two question types, a priori contrasts failed to detect statistically significant differences between question types. Likewise, contrary to a statistically significant univariate F-statistic for a two-way interaction of text genre and question type, F(4, 118) = 2.47, MSe = 0.014, p < .05, the corresponding multivariate statistic for the interaction remained statistically nonsignificant, Wilks' Lambda = 0.88, F(4,116) = 1.94, p = .11. Therefore, it was concluded that the statistical significance of the two-way interaction of text genre and question type was statistically nonsignificant6. ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ Insert Table 2 about here ÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐÐ The response time data showed the statistically significant main effect of question type indicating a trivial effect of the average question length for each question type, F(2, 118) = 240.16, MSe = 0.008, p < .001. A two- way interaction of question type and question block was also statistically significant, F(4, 118) = 4.41, MSe = 0.008, p < .01, indicating a trivial effect of question block permutation and subjectsÕ learning effect within a question block. There was no evidence of a speed-accuracy trade off in the data. Directionality on accuracy and response times. Neither accuracy nor response time data, as predicted, showed a statistically significant two- way interaction of text genre and question type , F(2, 59) = 2.6, MSe = 0.033, p = .08. The absence of inferential directionality suggests that the mental models constructed during the second reading session were two- dimensional spatial models regardless of text genres. Since the readers were already familiar with the three types of inference questions, the readers could build a two-dimensional spatial model regardless of the input text genre. Therefore, the data support OhtsukaÕs (1990) postulation on the plasticity of the mental model. That is, the characteristics of mental models derived from text can be modified into other forms depending on task expectation and the purpose of reading. General Discussion The overall results from this study indicate that the characteristics of mental models derived from the descriptive text are different from the characteristics of the models built from the other two text genres. The mental model derived from the descriptive text appears to be a two- dimensional global model, whereas the models derived from both procedural and narrative texts are one-dimensional. On the global perspective questions, the descriptive text group was not significantly more accurate than were the other two text genre groups. There was evidence that the spatial orientation questions were easier for the procedural text groups. There were no differences in the response times for the spatial orientation questions among the three text groups. On the spatial orientation questions, the descriptive text group was significantly more accurate than the other two text genre groups. On the order inference questions, procedural text readers and narrative text readers were faster than the descriptive text group. In this study, the procedural text used external coordinates (e.g., north, south, east, west) rather than the self-reference internal coordinates (e.g., right, left, ahead, back) used in the previous research (Perrig & Kintsch, 1985). Because the use of self-reference was shown to increase cohesiveness of the text, its use may have helped to increase the differences between the survey text and the route text observed in previous research. However, the results of this study show that the use of the external coordinate system in the procedural texts does not appear to affect the basic characteristics of the mental model built from it. Despite the increased similarity between the descriptive texts and the procedural texts, both using external coordinates, the one-dimensional nature of the mental model derived from the procedural text appears to be substantial. Regarding the hypothesis of the directionality in order inference, the accuracy scores showed a statistically significant effect of the directionality in both the procedural and narrative text groups. Both groups were more accurate in answering the order inference questions that were consistent with the direction of the mental model than those that were inconsistent. As expected, the readers of the descriptive text did not show any differences due to the direction of order inference questions. Since there were two versions of the descriptive text (from North to South and from South to North) following the same route described in the procedural and the narrative text versions, there is at least an implicit order in the presentation of information even in the descriptive texts. However, the absence of directionality in order inference indicates that the mental model derived from the descriptive text was not affected by the order of information presentation in the discourse. Thus, the results suggest that descriptive text subjects built a two-dimensional analogue of the space described in the text. The findings of this study are consistent with the claims that two-dimensional spatial mental models underlie descriptive discourse (e.g., Brewer, 1980) as well as with a previous study using a similar paradigm (Ohtsuka, 1990). In contrast to the previous Ohtsuka study, the present study has several new aspects: (1) The experimental texts used in the present study were shortened to equate a number of words per genre as much as possible without losing the distinct ÒflavourÓ of each representative text genre; (2) A distinctively different group of native speakers of English, Australian English speakers, participated in the present study and read experimental texts written in standard Australian English; (3) A smaller number of subjects (N = 70) was used in this experiment compared to the previous study (N = 122). Taken together, the successful replication of data trends across the two studies suggests that text genres indeed influence the characteristics of mental models in a consistent fashion. Although the accuracy data support the hypothesis of the directionality in order inference, the response time data of order inference did not provide sufficient support for the hypothesis that consistent order inference requires less time than inconsistent order inference. The absence of a response time difference was again consistent with the results obtained in the previous Ohtsuka study (1990). The results suggest that the order inference questions that were inconsistent with the mental model could be processed with a relatively smaller cognitive effort. In summary, the results of the procedural and the narrative text groups show a clear contrast to the data from the descriptive text group. In these two genre groups, the subjectsÕ accuracy decreased when order inference questions were presented in a form inconsistent with the existing mental model. The increase of errors in order inference suggests that the mental models built from the procedural and the narrative texts have one- dimensional characteristics, which are sensitive to the direction of order inference, as opposed to the two-dimensional model built from the descriptive text. The findings were consistent with the claims that event sequences underlie narrative discourse (Brewer, 1980, Brewer & Lichtenstein, 1981, van Dijk, 1976). Evidence of the directionality in order inference for the narrative text group suggests that the mental model built from the narrative text is an event model in which narrative events are placed along a time line. Finally, the disappearance of the effect of text genre after the second reading session supports OhtsukaÕs postulation on plasticity of mental models (Ohtsuka, 1990). According to this view, the mental models built from text are not static, rigid representations. Rather readers are able to manipulate the mental models derived from text actively to make inferences based on them. After the subjects were exposed to the questions, they seemed to be able to construct predominantly two- dimensional models during the second reading session. As mentioned earlier, research evidence has not been entirely conclusive whether the perspective of the reader is retained in one's mental model (cf. Perrig and Kintsch, 1985 vs. Taylor & Tversky, 1992). The findings from the first reading session of this study suggest that the perspective of the reader was retained and influenced the nature of the mental model constructed from the text. It appears that there is a crucial difference in the experimental procedures that might be responsible for the different results in the Perrig-Kintsch study and the Taylor-Tversky study. Before being tested for spatial information, Perrig and KintschÕs subjects read only one of two text types (where text type was a between-subject factor) whereas, Taylor and TverskyÕs subjects read both types of texts (where text type was a within-subject factor). The repeated presentation of different text types and the repeated testing of spatial information perhaps helped to focus readersÕ reading strategy in order to extract spatial information from the text. As a consequence, the differences between the initial mental models derived from the two text types might be minimized under such condition. Indeed, the results from the second reading session of the present study indicate that readers, with sufficient knowledge of what types of questions to expect, were able to build mental models that were not influenced by the input text genre. In conclusion, the present study has shown the extent of the influence of text genres on mental models. It has also provided some insight into whether perspectives in the original text are retained in subsequent mental representation. The outcome appears to partly depend on readersÕ expectations and self-adopted reading strategies. The view that readersÕ perspectives influence the nature of text representation can be traced back to earlier schema theories. For example, research on the effect of schema on the retrieval of text information has shown that the readers are able to retrieve previously unrecallable information after alternative perspectives are given (Anderson & Pichert, 1978; Pichert & Anderson, 1977). This evidence is interpreted as the role of schema in the memory retrieval process (e.g., Brewer & Nakamura, 1984). Namely, a schema functions in memory recall processes by providing: (a) a search path in memory retrieval, (b) an editing plan for the output of memory recall, and (c) best guesses for unavailable information. However, what is crucial in discourse comprehension processes appears to be the relationship that makes a particular piece of new information relevant to an entire mental model of text. Therefore, a perspective adopted by readers influences not only memory recall processes but also the processes of mental model construction. This study examined the mental models which readers construct when reading the text. These models are formed at the time of information encoding and readers can then actively manipulate them to represent textual information. The present study has shown how and to what extent readersÕ expectation of the text genre, which after all determines the relevance of particular incoming information to the mental model, influences the characteristics of the mental models constructed from the text as well as subsequent inference processes. References Anderson, R. C., & Pichert, J. W. (1978). Recall of previously unrecallable information following a shift in perspective. Journal of Verbal Learning and Verbal Behavior, 17, 1-12. Brewer, W. F. (1980). Literary theory, rhetoric, and stylistics: Implications for psychology. In R. J. Spiro, B. C. Bruce, & W. F. Brewer (Eds.), Theoretical issues in reading comprehension (pp. 221-239). Hillsdale, NJ: Erlbaum. Brewer, W. F. (1987). Schemas versus mental models in human memory. In P. Morris (Ed.), Modelling cognition (pp. 187-197). Chichester, UK: Wiley. Brewer, W. F., & Lichtenstein, E. H. (1981). Event schema, story schemas, and story grammars. In J. Long & A. Baddeley (Eds.), Attention and performance IX (pp. 363-379). Hillsdale, NJ: Erlbaum. Brewer, W. F., & Lichtenstein, E. H. (1982). Stories are to entertain: A structural-affect theory of stories. Journal of Pragmatics, 6, 473-486. Brewer, W. F., & Nakamura, G. V. (1984). The nature and function of schemas. In R. S. Wyer, Jr., & T. K. Srull (Eds.), Handbook of social cognition vol. 1 (pp. 119-160). Hillsdale, NJ: Erlbaum. Brewer, W. F., & Ohtsuka, K. (1988a). Story structure and reader affect in American and Hungarian short stories. In C. Martindale (Ed.), Psychological Approaches to the Study of Literary Narratives (pp. 133-158). Hamburg: Buske. Brewer, W. F., & Ohtsuka, K. (1988b). Story structure, characterization, just world organization, and reader affect in American and Hungarian short stories. Poetics, 17, 395-415. Bryant, D. J., Tversky, B., & Franklin, N. (1992). Internal and external spatial frameworks for representing described scenes. Journal of Memory and Language, 31, 74-98. Byrne, R. M. J., & Johnson-Laird, P. N. (1989). Spatial reasoning. Journal of Memory and Language, 28, 564-575. Ehrlich, K., & Johnson-Laird, P. N. (1982). Spatial descriptions and referential continuity. Journal of Verbal Learning and Verbal Behavior, 21, 296-306. Franklin, N., & Tversky, B. (1990). Searching imagined environments. Journal of Experimental Psychology: General, 119, 63-76. Glenberg, A. M., Meyer, M., & Lindem, K. (1987). Mental models contribute to foregrounding during text comprehension. Journal of Memory and Language, 26, 69-83. Johnson-Laird, P. N. (1983). Mental models. Cambridge, MA: Harvard University Press. Morrow, D. G. (1985). Prominent characters and events organize narrative understanding. Journal of Memory and Language, 24, 304-319. Morrow, D. G. (1986). Places as referents in discourse. Journal of Memory and Language, 25, 676-690. Morrow, D. G., Greenspan, S. L., & Bower, G. H. (1987). Accessibility and situation models in narrative comprehension. Journal of Memory and Language, 26, 165-187. Ohtsuka, K. (1990). Variation in mental models of text as a function of genre (Doctoral dissertation, University of Illinois at Urbana-Champaign, 1990). Dissertation Abstracts International, 51, 1169A. Ohtsuka, K., & Brewer, W. F. (1992). Discourse organization in the comprehension of temporal order in narrative texts. Discourse Processes, 15, 317-336. Perrig, W., & Kintsch, W. (1985). Propositional and situational representations of text. Journal of Memory and Language, 24, 503-518. Pichert, J. W., & Anderson, R. C. (1977). Taking different perspectives on a story. Journal of Educational Psychology, 69, 309-315. Pugh, J. (1991). The timer XFCN Ver. 1.1.1 [Computer program]. Cupertino, CA: Apple Computer. (Available via anonymous ftp from mac.archive.umich.edu) Taylor, H. A., & Tversky, B. (1992). Spatial mental models derived from survey and route representations. Journal of Memory and Language, 31, 261- 292. Thorndyke, P. W., & Hayes-Roth, B. (1982). Differences in spatial knowledge acquired from maps and navigation. Cognitive Psychology, 14, 560-589. van Dijk, T. A. (1976). Philosophy of action and theory of narrative. Poetics, 5, 287-338. Author Notes This study was funded by a grant from the Australian Research Council. I am grateful to Uta Juniper for her assistance in running the experiments. In addition, I would like to thank Uta Juniper and Faye Lucas for their comments on an earlier draft of this article. Correspondence concerning this article should be addressed to Dr Keis Ohtsuka, Department of Psychology, Victoria University of Technology, St Albans Campus, PO Box 14428, MMC, Melbourne, Vic 3000, AUSTRALIA. E-mail: kohtsuka@essex.vut.edu.au APA style citation format of this paper is: Ohtsuka, K. (1993, November). Text genres and mental models: What readers get from text and how they use it. Paper presented at the Australian Association for Research in Education (AARE) Annual Conference, Fremantle, Western Australia, November 23, 1993. Footnotes 1 Thorndyke and Hayes-Roth (1982) reported that map-learning subjects were not as accurate as navigation subjects in indicating the direction of a hidden landmark. Their findings suggest that global spatial understanding and spatial orientation judgment might be two different kinds of spatial tasks in real-life spatial learning. 2 There was a random variance between 0 ms to 15 ms (average variance 5 ms) in measurement due to a polling rate of Apple keyboard input. There was no known method available to account for the variance. 3 The rationale for presenting questions in block format was that subjects may require a few question items to stabilize their responses. The question types were presented in a block format in the following order: global perspective--spatial orientation--order inference, spatial orientation--order inference--global perspective, and order inference-- global perspective--spatial orientation. 4 An error in describing the location of a landmark was found in the narrative texts. Question items regarding the misplaced landmark were discarded from the analysis. 5 There was a statistically significant two-way interaction of text genre and question block order, F(4,61) = 2.71, MSe = 0.071, p < .05, due to an aforementioned trivial effect of depressed comprehension in one of the three question block orders. 6 Indeed, none of the a priori contrasts reached a statistically significant level supporting the conclusion based on the multivariate measure. Appendix A: Examples of Experimental Passages Descriptive Passage (North to South) Rebecca Jones lives in the town of Newell. Newell is located in a valley near the seaside. Looking down on the town, you can see three major streets that run north-south. First Avenue is on the west side of the town, Second Avenue is in the middle, and Third Avenue is on the east. There are three major streets that run east-west. From the north end of town to the south end, they are: Apple Street, Cherry Street, and Orange Street. The Cairn River flows from east to west on the south side of the town. Rebecca's house is on the northeastern edge of town. Her house is located on the east side of Third Avenue up on the hill. Just to the south of Rebecca's house is a beautiful fountain. Apple Street crosses Third Avenue to the south of the fountain. There is a roundabout at the intersection of Apple Street and Third Avenue. The Greenwood Department Store is located on the west side of Third Avenue in the block between Apple Street and Cherry Street. The tall department store towers over the surrounding buildings. Cherry Street crosses Third Avenue to the south of the department store. Second Avenue runs parallel to Third Avenue one block to the west. Shoppers' Paradise, the major shopping centre in Newell, is on the west side of Second Avenue and occupies the whole block between Apple and Cherry Street. St. Mary's Cathedral is located on the east side of Second Avenue between Cherry and Orange Street. Directly across Second Avenue from St. Mary's Cathedral, on the west side of the street, Picasso's sculpture of a huge bird soars towards the sky. A big banner announcing the "Corn Festival" hangs over Second Avenue to the south of the sculpture. Orange Street crosses Second Avenue to the south of the festival banner. First Avenue runs parallel to Second Avenue one block to the west. Memorial Park is located on the west side of First Avenue. The park occupies the entire block between Cherry Street and Orange Street. To the south of Orange Street on the east side of First Avenue is a TV tower, the highest structure in Newell. There is a bridge to the south of the TV tower where First Avenue crosses over the Cairn River. The hospital is south of the river and located on the west side of First Avenue. Procedural Passage (North to South) Welcome to my home town, Newell. Newell is located in a valley near the seaside. My name is Rebecca Jones. Here are the directions about how to get to the hospital from my house. I live on the northeastern edge of town on the hill. You are standing in front of my house on Third Avenue. Go south on Third Avenue. You will immediately see a beautiful fountain on the east side of the street. The next street you will cross is Apple Street. The intersection you come to is Third Avenue and Apple Street, where there is a roundabout. Keep on going south on Third Avenue, you will pass the Greenwood Department Store on the west side of the street. The tall department store towers over the surrounding buildings. The next Street you come across is called Cherry Street. Turn west when you arrive at Cherry and go one block to Second Avenue. Arriving at Second Avenue, you will see Shoppers' Paradise. It is on the north side of Cherry Street and the west side of Second Avenue. Shoppers' Paradise is the major shopping centre in Newell and occupies the entire block between Apple Street and Cherry Street. Turn south onto Second Avenue away from Shoppers' Paradise. You will see St. Mary's Cathedral on the east side of the street. Across the street from the Cathedral, Picasso's sculpture of a huge bird soars up to the sky. You will pass under a big banner announcing the "Corn Festival" directly to the south of the sculpture. Continue south until you come to Orange Street, then turn west onto Orange and go one more block. You will see Memorial Park. It is on the north side of Orange Street and the west side of First Avenue. Memorial Park occupies the entire block between Cherry Street and Orange Street. Turn south on First Avenue at First and Orange. You'll soon see the TV Tower, the highest structure in Newell, on the east side of the street. As you continue down First Avenue, you will cross the bridge over the Cairn River. Finally you will see the hospital on the west side of First Avenue. Narrative Passage (North to South) Rebecca and her dog, Fido, lived in Newell, a town located in a valley near the seaside. They lived on the northeastern side of town up on the hill. RebeccaÕs house was located on the east side of Third Avenue. Today Rebecca walked Fido to the Vet's for his vaccination. They started out walking south along Third Avenue. Immediately on the east side, they saw a beautiful fountain. They continued south and passed a roundabout at the intersection of Third Avenue and Apple Street. Then continued along Third until they came to the Greenwood Department Store. The tall department store towers over the surrounding buildings on the west side of the street. The two of them soon came to Cherry Street. They turned west on Cherry Street and walked on for one block to Second Avenue. Rebecca looked towards Shoppers' Paradise, the major shopping centre of Newell. It is on the north side of Cherry Street and the west side of Second Avenue. Shoppers' Paradise occupies the entire block between Apple Street and Cherry Street. Rebecca saw an old friend heading towards her. They became deeply involved in reminiscing about old times at the intersection of Cherry Street and Second Avenue. Suddenly Rebecca realised Fido was gone! She ran one block south where she came to St. Mary's Cathedral on the east side of Second Avenue. She walked across the street to Picasso's sculpture of a huge bird soaring towards the sky, hoping Fido might be hidden behind it. Then she passed under a big banner announcing the "Corn Festival" hanging over Second Avenue directly to the south of sculpture. She continued along Second Avenue until she reached Orange Street and went west for one block. Then she crossed over First Avenue to Memorial Park. It is on the north side of Orange Street and the west side of First Avenue. Memorial Park extends the entire block between Cherry Street and Orange Street. Arriving at the park, she called out her dogÕs name but Fido did not come. Rebecca was really becoming concerned and began to continue her search south on First Avenue. The TV tower, the highest structure in Newell, on the east side was blurred by her tears. Then she crossed the bridge over the Cairn River. She wondered if Fido could have fallen into the river. After a while, she found herself in front of the Veterinary Hospital on the west side of First Avenue. Standing at the front door was Dr. Jake with a small animal in his arms. It was Fido! "He's had his vaccination," said Dr. Jake, "but the next time, don't make Fido face this all alone!" Appendix B: List of Three Types of Inference Questions B.1 Examples of Global Perspective Questions 1. The fountain is north of the roundabout. 2. The Greenwood Department Store is to the northeast of Memorial Park. 3. Memorial Park is southwest of Shoppers' Paradise. 4. The Greenwood Department Store is to the south of the fountain. 5. Memorial Park is south of the Cairn River. 6. The roundabout is east of Shoppers' Paradise. B.2 Examples of Spatial Orientation Questions 1. If you're at Memorial Park and look west, you would see Picasso's sculpture of a huge bird. 2. If you're at Memorial Park and look northwest, you would see the TV tower. 3. If you're at the roundabout and look west, you would see the north side of Shoppers' Paradise. 4. If you're at St. Mary's Cathedral and look northwest, you would see Shoppers' Paradise. 5. If you're standing on the bridge over the Cairn River and look south, you would see the TV tower. B.3 Examples of Order Inference Questions 1. When you're going from Rebecca's house to the hospital via Second Avenue, you would pass Shoppers' Paradise then St. Mary's Cathedral. 2. When you're going from the hospital to Rebecca's house via First Avenue and Cherry Street, you would pass Memorial Park then the TV tower. 3. When you're going from Rebecca's house to the hospital via Cherry Street and First Avenue, you would pass Shoppers' Paradise then Memorial Park. 4. When you're going from Rebecca's house to the hospital via Third Avenue and Cherry Street, you would pass Shoppers' Paradise then the fountain. 5. When you're going from the hospital to Rebecca's house via Cherry Street and Third Avenue, you would pass Shoppers' Paradise then the Greenwood Department Store. 6. When you're going from Rebecca's house to the hospital via Second Avenue, you would pass St. Mary's Cathedral then the "Corn Festival" banner. Table 1 Accuracy Scores and Response Times for Each Genre ______________________________________________________________________ Text Genre Descriptive Procedural Narrative ______________________________________________________________________ M SD M SD M SD Global Perspective Questions Accuracy (%) 77.0 23.5 78.3 23.5 70.9 23.5 Response time 3.74 0.15 3.75 0.15 3.70 0.15 Spatial Orientation Questions Accuracy (%) 70.4 19.7 77.9 19.7 66.6 19.6 Response time 3.88 0.12 3.85 0.12 3.82 0.12 Order Inference Questions Accuracy (%) 70.2 19.2 70.8 19.2 69.8 19.1 Consistent 69.9 18.7 85.4 18.7 77.3 18.6 Inconsistent 70.3 26.9 56.4 26.9 63.3 26.5 Response time 4.10 0.14 3.99 0.14 3.99 0.14 Consistent 4.10 0.15 3.97 0.16 3.98 0.16 Inconsistent 4.10 0.15 4.00 0.15 4.00 0.16 ______________________________________________________________________ Note. Response time data are transformed data. Table 2 Accuracy Scores and Response Times for Each Genre After the Second Reading Session ______________________________________________________________________ Text Genre Descriptive Procedural Narrative ______________________________________________________________________ M SD M SD M SD Global Perspective Questions Accuracy (%) 79.5 24.5 90.6 24.9 77.8 24.5 Response time 3.65 0.15 3.67 0.12 3.70 0.14 Spatial Orientation Questions Accuracy (%) 76.2 22.1 81.2 22.0 75.0 21.6 Response time 3.77 0.14 3.78 0.15 3.80 0.14 Order Inference Questions Accuracy (%) 82.2 19.7 80.0 20.1 81.1 19.7 Consistent 85.4 23.0 76.8 22.9 79.1 23.0 Inconsistent 78.3 23.0 87.3 23.5 80.7 23.5 Response time 4.00 0.12 3.97 0.12 4.02 0.12 Consistent 4.05 0.14 3.98 0.14 3.98 0.14 Inconsistent 4.04 0.12 3.96 0.13 3.98 0.12 ______________________________________________________________________ Note. Response time data are transformed data. Figure Caption Figure 1. A map of Newell. Figure Caption Figure 2. Two-way interaction of genre and question type on the response time data.