Oral Reading Fluency Practices with Automatic Speech Recognition

Dr. Gerald Tindal

Published on: September 5, 2023

In this release, we describe instructional practices for the development of fluent readers. Before describing these various practices, however, it’s important to address the critical components of fluent reading and distinguish instructional practices for developing fluent readers from that of using fluency to monitor progress. We adopt the 2018 NAEP study definition of oral reading fluency (ORF): When reading “text with speed, accuracy, and proper expression…fluency is demonstrated during oral reading of passages through ease of word recognition and appropriate pacing, phrasing, and intonation (i.e., expressiveness), which indicates the extent to which these prosodic features match the structure and meaning of the text”[1].

Instructional practices for the development of fluent readers comes close to confusing dependent variables and independent variables. It’s important to remember that reading fluency assessment is designed to serve as dependent variable like a thermometer, to gauge the overall temperature of the reading health of the student. As such, it should be measured using passages that are unpracticed and read ‘cold’ from the beginning, much like those used in the original research at the University of Minnesota and in NAEP. This strategy allows ORF to serve as the dependent variable to monitor performance and progress.

When using passages as a part of an independent variable for instructional purposes, they can be manipulated in any way that provides students active engagement. The instructional practices described below, are designed to increase the opportunity to learn, particularly by increasing engaged, active learning with guidance and feedback. Research findings from the past several decades, indicate that it is active responding that is critical in learning. The degree to which these strategies can be used to engage student is, therefore, important. When trying to improve oral reading fluency as an instructional technique (an independent variable), it is important to measure its effectiveness with passages that are not practiced. For example, select passages from easyCBM© to see if fluency instruction is effective.

Important Note: Most researchers, consider the relation between fluency and comprehension to be correlational (e.g., they are highly related to each other). Some researchers, however, think of it in a slightly more causal way, with fluent reading allowing attention (cognitive load) to be allocated to the meaning of words, rather than the decoding of words. CBMSkills is neutral on this issue, and as noted earlier, ORF can be used as an independent variable. Either way, it is possible that with greater comprehension, fluency will increase, or with greater fluency, comprehension will increase.

Finally, the list below has been adapted from other practices described in the literature that are particularly suited for our use of automatic speech recognition (ASR) as part of CBMSkills. For a more comprehensive list of practices, see the book written by Timothy Rasinski (2003): The Fluent Readers: Oral Reading Strategies for Building Word Recognition, Fluency, and Comprehension. We  do not cover specific curricula, like Read Naturally, which is an excellent program designed to improve students’ reading fluency and should be used in addition to the following practices that incorporate ASR into classroom activities.

Repeated reading on independent passages. Find a passage that is within the students reading range. That is, the students should be reading few words incorrectly (fewer than two or three). Tell the student that the passage is to be read repeatedly, but not to read it as fast as they can: They should be told to read at a comfortable rate. In CBMSkills, passages are automatically timed for one minute and students can move at their own pace. Then, the student repeats this process with the same passage, again reading at a comfortable rate. Each reading is recorded, so both students and teachers can review their readings in succession. One quick strategy for this review is to play the first recording with the student and discuss their fluency (how phrases are ‘chunked’ and how punctuation marks are incorporated into their reading) and then analyze the last recording to see if improvement was made.

Paired student reading. Identify a pair of students who can work together in a compatible, collaborative fashion. Either they can be paired based on their reading fluency, or other behavioral characteristics. Direct students to read a passage to each other. With CBMSkills, the students can select different passages then share the screen, with one student reading and the other observing. At the end of the reading, the two students can share the results (which are immediately available) and then switch roles. It also might be useful to have an older student and a younger student reading as pairs Because of our use of ASR, different passages can be used for reading and reviewing, which may be aspirational for the younger student and instructional for the older student.

Use guided reading and model word reading ahead of time. The teacher basically reads with students and guides them through various aspects of the passage. This can be done with CBMSkills by printing the passage out ahead of time and guiding students through the content, before, during, and after the lesson: Highlighting various word types, noting structural characteristics with various words, word meanings, making comments on the meaning of key vocabulary words, or identifying chunks of the passage that can be read fluently together. At the end of the passage, the teacher can have students read the passage independently using our ASR, and then review their readings.

Pre-read passages (title and questions ahead of time). Like guided reading, teachers can preview the passage for students, going over the title and highlighting any words with nuanced meaning or complex structures. The passages in CBMSkills can be printed, one copy for the student and another for the teacher in this preview. These passages vary in terms of the story grammar, allowing teachers to emphasize different elements like characters, settings, events, resolution to anticipate and comprehend the meaning of the passage. Following passage reading, students can read the passage on the computer and have their performance recorded.

Underline important words. The students can be provided a printed copy of the passage that has specific words underlined. These words can reflect broad linguistic structures or different semantic nuances of the passage. While reading, the student can be directed to use the passage as a crib sheet when they come to the underlined words, but otherwise continue to read with prosody. In CBMSkills, this strategy is likely to interfere with fluency, but the results of words read incorrectly can be compared with the printed copy of underlined words.

Schedule story reading time. Regularly schedule a time for teachers to read stories aloud to students, occasionally asking questions and piquing their curiosity and interest. With CBMSkills, the passages can be read aloud and then students read the passage independently to the computer. Because some students may not be good readers, reading aloud may be a point of embarrassment; this problem is avoided because their reading is completed privately on the computer.

Highlight words read incorrectly (with phonics strategies). Teachers can have students read a passage for a minute taking data on words read incorrectly. In CBMSkills, data are automatically collected on words read incorrectly and omitted. When the minute is up, teachers can review the reading and highlight both types of errors. This may also provide an opportunity to collect misread words across passages to provide a cumulative list and analyze them for consistent error patterns. In addition, other similar words with either semantic or structural components can be developed to highlight for students on how these words are similar or different.

Adjust complexity of passages (step down or step up). Teacher could vary the complexity of passages for students to read. In CBMSkills, 17 passages are presented at each grade level (1-5), so teachers have plenty of opportunity to stretch the range of passage complexity for students. Some passages might be quite simple and involve few unique words or unusual sight words, while others might be more complex, either in the types of words that appear in the passage or in the semantics or story grammar elements. Complexity also may be a function of readability, which is often calculated by a count of the number of words, words per phrase, dependent, clauses, etc. Caution should be exercised with this strategy to avoid having students read passages that are at their frustration level (with many incorrectly read words). See CBMSkills Technical Report.

Use Cloze technique. A passage is presented with selected words left blank intermittently throughout the passage. It is important not to have too many blanks, or the passage will become very difficult to read. Students can complete this printed copy with CBMSkills as a form of pre-reading. They can then read the passage on the computer and have their reading results printed. Teachers can then review both versions, highlighting consistencies and inconsistencies.

Assess word list reading fluency. CBMSkills has other measures in early reading that can be deployed in concert with the oral reading fluency passage. Rather than having students read passages in whichthe vocabulary and words can become constrained by the topic or story, reading a word list has no such constraints. Have students read from a word list or a pseudo word list to expose them to more and different words in isolation and not in to the context of a story. Both the original research at the University of Minnesota and the NAEP assessments used oral word and pseudo word reading as viable measures for documenting students’ reading skills.

Focus on phonemic awareness and word identification for young students. For young students,it may be important to analyze reading passages with  a focus on phonemic awareness and word identification skills. Because the CBMSkills early reading tasks are varied in a systematic manner, with near and far distractors, errors can provide clues to skills that are simply not present (the far distractor is selected) versus those that are emerging (the near distractor is selected). These errors can then be expanded by providing both examples and non-examples of different linguistic features, followed by having students practice them in isolation of the passage.

Use Readers Theatre. In this technique, students practice scripts as a part of a play much like that  in the performing arts. Because many of the CBMSkills passages include dialogue, students can be directed to act in the role of the characters and practice the text in quotation marks. Students basically study their scripts and then verbalize them at appropriate times when performing. This can be done ahead of time using a printed copy of the passage. Then, when they read into the computer, their reading can reflect prosodic quality. Note: This technique has not received widespread empirical support and should be used with guided practice to emphasize prosody.

Use sustained silent reading. In this technique, students read silently at their desks for an extended period. The passages from CBMSkills can be used and either students mumble (use low verbal levels to themselves) or read the entire passage silently. Then, after this pre-reading, students go to the computer and read the passage aloud. Like Readers Theater, this technique doesn’t have much empirical support. The big problem with silent reading is the inability to monitor mistakes being made or to model appropriate responses in their place. In summary, these practices are designed to ensure students are actively reading with explicit instruction and take advantage of an ASR system in place. They are not designed to replace a foundational curriculum, but can extend its application.

[1] White, S., Sabatini, J., Park, B. J., Chen, J., Bernstein, J., & Li, M. (2021). Highlights of the 2018 NAEP Oral Reading Fluency Study. Washington, DC: U. S. Department of Education. Page 3.

Dr. Gerald Tindal

Dr. Tindal is currently Professor Emeritus and the Director of Behavioral Research and Teaching (BRT) – University of Oregon. He is the former Castle-McIntosh-Knight Professor in the College of Education and past Department Head of Educational Methodology, Policy, and Leadership. His research focuses on alternate assessments, integrating students with disabilities in general education classrooms, curriculum-based measurement for screening students at risk, monitoring student progress, and evaluating instructional programs. Dr. Tindal conducts research on large scale testing and development of alternate assessments. This work includes investigations of teacher decision-making on test participation, test accommodations, and extended assessments of basic skills. He publishes and reviews articles in many special education journals and has written extensively on curriculum-based measurement and large-scale testing. He has also taught scores of courses on assessment systems, data driven decision-making, research design, and program evaluation.

Related Posts

Oral Reading Fluency Research

Currently, teachers recognize the significance of fostering fluent reading skills for students, blending appropriate speed, accuracy, and expression in their reading. Although the link between reading fluency and comprehension is widely acknowledged, it's particularly...

Oral Reading Fluency (ORF) Norms

The first study to be conducted establishing oral reading fluency (ORF) ‘norms’ occurred over 100 years ago. Specifically, in 1915, Starch asserted, “In every branch of instruction in the public schools we need a definite standard of attainment to be reached at the...