Proposed Sources of Cognitive Complexity in PARCC Items and Tasks: ELA/Literacy August 31, 2012 Text Complexity Explanation and Justification We expect text complexity to be one of several factors that will contribute to the determination of the cognitive complexity of individual items. For example, an item that may be considered moderately complex on the basis of factors such as Response Mode or Processing Demands may become highly complex when paired with a highly complex text. For that reason, we include text complexity as a source of item and task cognitive complexity. We will determine the complexity of reading passages by following the process outlined in 1. Text complexity 08-31-12.pdf. As a result of this process, a text will be assigned to one of three categories of test complexity: Readily Accessible, Moderately Complex, or Very Complex. These categories correspond with the categories of low, moderate, and high complexity used in the cognitive complexity framework. Command of Textual Evidence Explanation and Justification We define this source of cognitive complexity as the amount of text that an examinee must process (i.e., select and understand) in order to respond correctly to an assessment item. The amount of text to be processed is not a reference to the length of text or volume of reading that is required. Instead, this category focuses on the numbers of details in one or more texts that must be processed in order to respond to the requirements of Close Analytic Reading and Comparison and Synthesis of Ideas items. The amount of text processed is influenced by both the cognitive complexity of items and tasks and the complexity of the text or texts, so that ordinarily low complexity cognitive tasks can be adjusted upward when paired with very complex text and vice versa. This source of complexity is drawn directly from and is deeply grounded in the Common Core State Standards. The ELA/Literacy standards require both Close Analytical Reading and Comparison and Synthesis of Ideas, both within and across texts, and require examinees to command evidence from various places in one or more texts. This source of cognitive complexity measures the nature and amount of text that an examinee must process (i.e., select and understand) from one or more texts. It encompasses but is not fully explained by the length of text or volume of reading that is required. Instead, this category focuses on the numbers of ideas that must be processed in order to respond to the requirements of Close Analytical Reading and Comparison and Synthesis of Ideas. This source adjusts the interaction of the cognitive complexity of items and tasks with text complexity, so that ordinarily low complexity cognitive tasks can be adjusted upward when paired with very complex text and vice versa. This source of cognitive complexity was proposed originally in the June 25 Achieve memo as Amount of Text Required to Answer the Question Asked (in Figure 2). The definitions of this source of cognitive complexity are taken from that memo. Cognitive Complexity: ELA/Literacy Page 1 Low Complexity Items at this level require examinees to identify a single idea or detail in a text. For example, identifying the main idea in a text may be as simple as locating the explicit statement of that main idea. Moderate Complexity Items at this level require synthesis of ideas and details across multiple sections a single text. For example, identifying the main idea or theme of a text may require inferring the main or theme or integrating ideas and details from several locations in the text. Other types of items at this level that require synthesis of ideas and details from two texts will be only moderately complex if the two texts are closely related in theme or genre. High Complexity Items at this level require synthesis of ideas and details across multiple texts. High complexity items may require examinees to construct the main idea or theme that is common across multiple texts, especially multiple texts that are not closely related in theme and/or genre. Response Mode Explanation and Justification The way in which examinees are required to complete assessment activities influences an item’s cognitive complexity. We propose that, in general, selecting a response from among given choices often is less cognitively complex than generating an original response. This difference is due in part to the response scaffolding (i.e., response choices) in selected response items that is absent from constructed response items. Selected response items can be highly cognitively complex due to the influence of other sources of complexity in test items. Response Mode interacts with other sources of complexity to influence the level of complexity: in ELA/Literacy, with Text Complexity and Command of Textual Evidence; in Mathematics, with Mathematical Content and Processing Demands. Further, the degree to which response choices may be easily distinguishable or highly similar can be influenced by other sources of complexity, such as Text Complexity and Mathematical Content. Low Complexity Items at this level primarily require the examinee to select a correct response rather than generate a response. For example, an evidence-based selected response item, in which both parts A and B require examinees to select the correct answer from a set number of options, is a low complexity item. Similarly, a technology enhanced constructed response item in which part A requires examinees to select the correct answer from a series of options provided, and part B requires a simple response construction such as highlighting a single section of text, would also be considered low complexity. Moderate Complexity Moderate complexity requires examinees to construct responses or pair selected responses appropriately and correctly. Multiple selection multiple choice items (i.e., evidence based selected Cognitive Complexity: ELA/Literacy Page 2 response items) might require examinees to choose one of two correct responses in part A and the only corresponding correct choice in part B. Technology enhanced constructed response items require examinees to construct one of several paths to a correct response, as is required in completing Venn diagrams, for example. Or examinees may be required to select a correct response in part A of an evidence based selected response item and use a technology tool in part B to bring together multiple pieces of evidence. Moderate complexity prose constructed response summary items focus on main ideas in text. While these ideas may be readily identifiable and in many ways of relatively low complexity, the prose constructed response mode requires examinees to generate an essay response. Such open ended responses must answer the assigned questions without the scaffolding provided by selected response items, and are therefore best categorized at the moderate complexity level. High Complexity Prose constructed response items, including analytic, narrative, and summary tasks, require examinees to generate an essay response and are equally unscaffolded. However, analytic and narrative prose constructed response items, unlike summary PCRs, typically address more challenging standards and evidence statements and often require synthesis of ideas or the use of more than one text. Such characteristics place these items in the high complexity category. Response Mode may be coded as high complexity for selected response items that correspond to high complexity for Text Complexity or Command of Textual Evidence. Processing Demands Explanation and Justification Though all items go through an intensive content and editorial review process, some level of linguistic demand and reading load remains within items. Linguistic demands and reading load in item stems, instructions for responding to an item, and response options contribute to the cognitive complexity of items. Linguistic demands include vocabulary choices, phrasing, and other grammatical structures. Item development and review processes are designed to remove any such demands that are likely to be construct irrelevant. The remaining linguistic demands may be construct relevant or at least construct neutral. That said, linguistic demands contribute to the complexity and cognitive load in processing, understanding, and formulating responses to test items and tasks. Research on the role of linguistic demands in complexity and difficulty in mathematical problem solving goes back as far as 1972 (e.g., Jerman & Rees, 1972). In their study of mathematical problem solving and cognitive level, Days, Wheatley, & Kulm (1979) identify linguistic demands that are related to the difficulty of mathematics problem solving items, which they refer to as the problem’s syntax. More recently, Ferrara, Svetina, Skucha, and Murphy (2011) and Shaftel, Belton-Kocher, Glasnapp, and Poggio (2006) have identified linguistic demands that are related to mathematics item difficulty and discriminations indices. We propose five linguistic demands, taken from these studies, to identify in PARCC items as sources of complexity: – – – Ambiguous, slang, multiple meaning, and idiomatic words or phrases Words that may be unusual or difficult and specific to English language arts (i.e., vocabulary) Complex verbs (i.e., verb forms of three words or more), such as had been going, would have gone Cognitive Complexity: ELA/Literacy Page 3 – – Relative pronouns, specifically, that, who, whom, whose, which (sometimes), why Prepositional phrases Similarly, lengthy item stems, instructions for responding to an item, and response choices also place reading and processing demands on examinees and may give rise to additional complexity. Ferrara et al. (2011) defined reading load and demonstrated its relationship to item difficulty and discrimination indices for grades 3-5 mathematics items. In research studies, linguistic demand and reading load have been identified by counting numbers of words, prepositional phrases, and so forth. That approach is not feasible for the thousands of PARCC items and tasks. We propose a holistic judgment approach to determining Processing Complexity. These holistic judgments will account for the details in the Reading Load and Linguistic Demands research frameworks. While research substantiates the impact of processing demands on item complexity, best practices in item development always seek to minimize construct irrelevant sources of complexity such as linguistic demands and reading load. Since the remaining features of the item should be construct relevant or construct neutral, this source of complexity may not be significant enough to account for more than 10% of the weighting in the overall complexity index. Low Complexity Low complexity is generally defined as a combination of low reading load and low linguistic demand. Compared to moderate reading demand and high reading demand, low reading demand is characterized by simple language with few words (approximately 25 words or fewer) in an item, including the item stem, response choices, and other directions for responding. Low complexity for this source also is characterized by low linguistic demands, generally, and low frequencies of all five linguistic demands (see above). Moderate Complexity Moderate complexity is defined as a combination of moderate reading load and moderate linguistic demand. Moderate reading load is characterized, generally, by a range of simple to grade appropriate language with items that are several sentences in length. Moderate complexity for this source also is characterized by moderate linguistic demands (i.e., generally, a few instances of some of the frequencies of all five linguistic demands; see above). High Complexity High complexity is defined as a combination of high reading demand and high linguistic demand. High reading load is characterized by grade appropriate language in prompts that are generally several sentences in length, with high linguistic demands (i.e., generally, instances of some of the frequencies of all five linguistic demands; see above). Cognitive Complexity: ELA/Literacy Page 4