CHAPTER 9 - Errors and Pitfalls

Academic writing involves a complex and comprehensive structure, requiring not only adherence to formal conventions but also a strong emphasis on comprehension and language proficiency. This study aimed to analyze language errors in academic writing produced by language students at Universitas Terbuka using a descriptive qualitative design. The analysis was based on the surface strategy taxonomy and key linguistic components, including phonology, morphology, semantics, syntax, and discourse. These linguistic aspects provided a comprehensive framework for identifying and categorizing the errors. The results revealed a total of 314 language errors. Within the surface strategy taxonomy, omission was the most frequent error type (50 instances), while misordering was the least frequent (15 instances). In terms of linguistic categories, syntax errors were the most common (90 instances), whereas phonological errors were the least frequent (15 instances). The findings indicate that students often omit essential linguistic elements, leading to contextual ambiguities. Moreover, a limited understanding of syntax significantly contributes to confusion and reduced clarity for readers.

Block programming has been suggested as a way of engaging young learners with the foundations of programming and computational thinking in a syntax-free manner. Indeed, syntax errors—which form one of two broad categories of errors in programming, the other one being logic errors—are omitted while block programming. However, this does not mean that errors are omitted at large in such environments. In this exploratory case study of a learning environment for early programming (Kodetu), we explored errors in block programming of middle school students ( N = 123), using log files drawn from a block-based online. Analyzing 1033 failed executions, we found that errors may be driven by either learners’ knowledge and behavior, or by the learning environment design. The rate of error types was not associated with the learners’ and contextual variables examined, with the exception of task complexity (as defined by SOLO taxonomy). Our findings highlight the importance of learning from errors and of learning environment design.

This study was aimed at finding out: (1) the types of grammatical errors in narrative writing made by eleventh-grade students of one private Islamic senior high school Palembang, South Sumatera, Indonesia and (2) the most frequent type of grammatical errors in narrative writing made by eleventh-grade students of private Islamic senior high school Palembang, South Sumatera, Indonesia. This study was a descriptive qualitative research. The data were collected from narrative writing written by the 32 eleventh grade students. In this study, grammatical items analyzed were covering noun, pronoun, verb, conjunction, adjective, adverb and preposition. In analyzing students’ errors, some procedures of error analysis were applied. They were identification, description, and explanation. The frequencies of errors were calculated in percentages. Based on the result of the analysis, there were 621 grammatical errors found in students’ narrative writing. Errors in using verb became the most frequent type of errors with 308 errors (49.59%), followed by errors in pronoun with 136 errors (21.90%), in adjective with 87 errors (14.00%), in preposition with 61 errors (9.82%), in noun with 13 errors (2.09%), in conjunction with 10 errors (1.61%) and the last errors in adverb with 6 errors (0.96%).

Background: Mild cognitive impairment (MCI) was originally used to describe a transitional state between normal condition and dementia. Revised and extended definition of MCI has been proposed that covers a broader range of cognitive impairment, distinct from normal ageing and from Alzheimer’s disease. Despite the existence of reports regarding analyses of the Clock Drawing test (CDT) in Alzheimer’s disease and vascular dementia, those focusing on MCI subjects are still lacking. The purpose of this study was to assess the characteristics of CDT and compare the results of quantitative and qualitative analyses of CDT performance in patients with MCI. Materials and methods: Five hundred four consecutive patients with MCI (Korean versions of the Mini-Mental State Examination (K-MMSE)>23) were recruited for the study. All participants were examined by the K-MMSE, modified-MMSE (3MS) and the CDT. Quantitative scoring of the CDT was done by the Manos and Wu’s method. Qualitative error types of the CDT were classified as stimulus-bound response (SBR), conceptual deficit (CD), spatial and/or planning deficit (SPD), and perseveration error (PE) by Rouleau’s classification. We divided the subjects into two subgroups by the MMSE scores (lower cognitive function group, MMSE=24~26 vs. higher cognitive function group, MMSE=27~30) and compared the CDT scores and frequency of the error types between them. Results: The total scores in the CDT significantly correlated with the total scores of the K-MMSE, 3MS and the level of education. Of the errors in the CDT, SPD was the most frequent type of error (45.3%) in the total samples. The scores in the K-MMSE and 3MS in the patients having CD errors were lower compared with those of other types of error. The lower cognitive group made more CD error (32.9%) than that of the higher cognitive group (25.5%). Conclusion: Although the CDT cannot be used solely for clinical diagnosis of dementia, it provides useful cognitive information quantitative as well as qualitative ways, estimating the characteristics of MCI patients as a simple screening test.

262: Improving error identification with an inpatient maternal-fetal medicine pharmacist

This study aims to describe and compare the phonological characteristics of European Portuguese and Brazilian Portuguese preschool and school age children with Speech Sound Disorders (SSD). Speech samples for the European Portuguese Group (EPG) (n = 13) were collected using Subteste Fonético e Fonológico of Teste Fonético e Fonológico Avaliação da Linguagem Pré-Escolar. For the Brazilian Portuguese Group (BPG) (n = 13) Prova de Nomeação de Fonologia of Teste de Linguagem Infantil ABFW was applied. Different phonological measures were considered. Groups were matched according to sex, age, and percentage of correct consonants (revised). EPG presented more weak syllable deletion (p = .00); absolute index and relative index had a higher number of omissions (p = .003). BPG had more substitutions (p = .004). Intragroup analysis showed differences between groups in the occurrence of phonological processes (p ≤ 0.00). The most occurring was gliding of liquids, cluster reduction and devoicing in both groups; for the absolute index and relative index, the EPG presented differences in omission (p = .003), and the BPG in substitution (p = .002). Results suggested differences between groups in phonological processes occurrence and a relation with the most frequent type of error. These findings may occur due to the variation of phonetic and phonological characteristics between European Portuguese and Brazilian Portuguese in the two phonological tests. Linguistic variations had not directly influenced the measures studied, which characterized SSD. European Portuguese and Brazilian Portuguese children with SSD demonstrated similar characteristics as to the type of errors and phonological processes.

This study aims to analyse how translation experts from the German department of the European Commission’s Directorate-General for Translation (DGT) identify and correct different error categories in neural machine translated texts (NMT) and their post-edited versions (NMTPE). The term translation expert encompasses translator, post-editor as well as revisor. Even though we focus on neural machine-translated segments, translator and post-editor are used synonymously because of the combined workflow using CAT-Tools as well as machine translation. Only the distinction between post-editor, which refers to a DGT translation expert correcting the neural machine translation output, and revisor, which refers to a DGT translation expert correcting the post-edited version of the neural machine translation output, is important and made clear whenever relevant. Using an automatic error annotation tool and the more fine-grained manual error annotation framework to identify characteristic error categories in the DGT texts, a corpus analysis revealed that quality assurance measures by post-editors and revisors of the DGT are most often necessary for lexical errors. More specifically, the corpus analysis showed that, if post-editors correct mistranslations, terminology or stylistic errors in an NMT sentence, revisors are likely to correct the same error type in the same post-edited sentence, suggesting that the DGT experts were being primed by the NMT output. Subsequently, we designed a controlled eye-tracking and key-logging experiment to compare participants’ eye movements for test sentences containing the three identified error categories (mistranslations, terminology or stylistic errors) and for control sentences without errors. We examined the three error types’ effect on early (first fixation durations, first pass durations) and late eye movement measures (e.g., total reading time and regression path durations). Linear mixed-effects regression models predict what kind of behaviour of the DGT experts is associated with the correction of different error types during the post-editing process.

The objectives of this study were (1) to explore the types and most frequent type, and the sources and most frequent source of academic verb collocation problems of undergraduate students majoring in English at Walailak University, (2) to compare differences in the use of academic verb collocations among three groups of students: low, moderate, and high English language ability, and (3) to examine the relationship between the use of academic verb collocations and writing ability among three groups of students. The participants were 155 second- and third-year English majors who were studying at Walailak University in the second trimester of academic year 2009. The research instrument was the academic verb collocation writing ability test, which was a writing test consisting of 21 items. The data were analyzed by using descriptive statistics, one-way analysis of variance (one-way ANOVA), and Pearson’s Product-Moment Correlation Coefficient. The findings revealed that (1) the verb-noun collocation was the most frequent type of errors, and approximation was the most frequent source of errors, (2) students in the high English language ability group gained significantly higher average scores on academic verb collocations than the other two groups at the level of .05, and (3) there was a relationship between the use of academic verb collocations and writing ability of students at the moderate level at the significant level of .05. The findings indicated the importance of collocations in language learning and teaching in classroom.

This study investigated the level of grammatical accuracy in typically developing 3-year-olds and the types of errors they produce. Twenty-two 3-year-olds participated in a picture description task. The percentage of grammatical utterances was computed and error types were analyzed. The mean level of grammatical accuracy in typical 3-year-olds was ∼71%, with a wide range of variability. The current study revealed a variety of error types produced by 3-year-olds, most of which were produced by fewer than 5 children. The pattern observed for most of the children was to produce a scattering of errors with no more than a few of any 1 error type. The level of grammatical accuracy in 3-year-olds was skewed toward the high end. Although tense marking errors were the most frequent error type, they accounted for only 1/3 of the errors produced by 3-year-olds. A more general measure of grammaticality that considers additional aspects of language might, therefore, be useful in assessing language at this age.

Introduction: Clinical and population decision-making relies on the systematic evaluation of extensive regulatory evidence. The FDA drug reviews provide detailed information on clinical trial design, enrollment criteria, sample size, randomization, comparators, endpoints, and indications. However, extracting these data is resource-intensive and time-consuming. Generative Artificial Intelligence large language models (LLMs) may accelerate the extraction and synthesis of such information. This study compares the performance of three LLMs, ChatGPT-4o, Gemini 2.5 Pro, and DeepSeek R1, in extracting and synthesizing regulatory and clinical information using antibiotics approved for complicated urinary tract infections (cUTIs) between 2010 and 2025 as a case study. Methods: LLM models were evaluated using general (short, direct) and detailed (structured, guidance-referencing) prompts across five domains including accuracy (precision and recall), explanation quality, error type (hallucination rate, misclassification, and omission), efficiency (response time, correct answers per second, and seconds per correct answer), and consistency with responses generated in duplicate runs. Two investigators independently reviewed outputs against FDA guidance, resolving discrepancies by consensus. Statistical analyses included χ², Wilcoxon, and Kruskal–Wallis tests with false discovery rate correction. Results: Among 324 responses, accuracy differed significantly across models (χ², p <0.001) with Gemini 2.5 Pro achieving the highest accuracy (66.7%), followed by ChatGPT-4o (51.9%) and DeepSeek R1 (37.0%). General prompts outperformed detailed prompts (59.3% vs 44.4%; p =0.011). Gemini 2.5 Pro showed highest explanation quality and most consistent outputs, while ChatGPT-4o had shortest response times and highest efficiency. Hallucination was the most frequent error type across models. Conclusion: LLMs showed variable capability in extracting regulatory and clinical trial information. Gemini 2.5 Pro showed the strongest overall performance, while ChatGPT-4o was faster but less accurate, and DeepSeek R1 underperformed across most domains. These findings highlight both the promise and limitations of LLMs in regulatory science and support complementary use alongside human review to streamline evidence synthesis. Author Summary: Our research addresses a critical question in artificial intelligence for healthcare: how well do generative Generative Artificial Intelligence (GenAI) tools extract and synthesize regulatory and clinical information to inform decision-making? We assessed ChatGPT-4o, Gemini 2.5 Pro, and DeepSeek R1 performance in extracting and synthesizing information from regulatory documents and clinical trial data using all FDA approved antibiotics for the treatment of complicated urinary tract infections. We compared LLMs outputs directly with the original data sources. We assessed the models’ performance using both broad and detailed prompts across several areas, including accuracy of the information (precision and recall), quality of explanation, type of errors (hallucination, misclassification, and omission), efficiency and speed (response time, correct answers per second, and seconds per correct answer), and consistency of responses across repeated runs. The results suggest that while the models were generally fast and efficient extracting large volumes of information, they also produced errors and omissions that could limit their reliability. These findings highlight both the promise and the current limitations of GenAI, underscoring its potential value as a human supervised tool for safely supporting regulatory science and clinical decision-making.

A notably high percentage (33%) of students with mathematics learning difficulties (MLD) has been reported in Tibet, which can partly be explained by the current state of mathematics teaching. Less known is what kind of errors these students are making in different types of mathematical tasks. This study investigated the types and frequency of errors made in mathematics tasks by Tibetan seventh-grade students ( n = 30) with MLD, as well as whether gender and school type had any effects on these errors. The novel Mathematics Error Pattern Identification Test (MEPIT) was used to identify the following eight different error types: visual-spatial, comprehension, transformation, relevance, fact, procedural, measurement, and presentation. Students identified as having MLD completed the MEPIT in two sessions due to its length (64 items). Regression analysis and t-tests were used. The most frequent error types were fact and comprehension errors. Compared to the boys in the study, the girls seemed to be more vulnerable to fact and relevance errors. The students in the rural school made more comprehension errors compared to the students in the urban school. Our exploratory study calls for further research on assessing errors that students with MLD make in mathematical tasks. MEPIT was shown to be a promising assessment tool for Tibetan teachers to identify errors that students with MLD may make in mathematics tasks.

Papers to Appear in Forthcoming Issues

This study concerns one of the processes involved in written verbal production: text revising. Our aim was to experimentally test the specificity of reading for revision as compared to reading for comprehension, and to determine the cognitive cost of initiating and performing critical-reading processes. A two-session experiment was conducted on students who had to perform a comprehension task and a revising task on a text presented in its basic version, or in a version containing either syntax errors or spelling errors. An analysis of the cognitive effort associated with critical-reading and comprehension-reading processes, and of the participants’ comprehension and revising performance, showed that critical reading was more effortful than comprehension reading. It showed also that critical reading was more effortful for the text version with syntax errors than with spelling mistakes. In addition, the working memory span of the readers/revisers had a different impact on effort and performance, depending on the type of errors in the text (spelling or syntax).

This corpus-based study aimed to investigate the presence of context-dependent linguistic errors in a corpus of clinical reports. The data were taken from a corpus comprising more than 2 million words and made up of clinical reports from emergency medicine, intensive care unit, general surgery, and psychiatry. Quantitative and qualitative analyses were carried out. A language model based on n-grams was developed for the detection of errors, parameters for the selection of cases were defined, and a classification tool was implemented. The findings indicated that emergency medicine was the medical specialty with the highest number of context-dependent errors and that the most frequent type of error was omission of written accent. Furthermore, the analysis revealed the presence of errors of competence due to the incorrect application of the linguistic norm of Spanish, phenomena of phonetic similarity, and composition of words; it is also worth noting that performance errors occurred due to rapid typing on the keyboard. This study constituted the first analysis and creation of a typology of context-dependent errors for the medical domain in Spanish. It contributed to the design of a module based on linguistic knowledge that can be used for the development and improvement of automatic correction systems that, in turn, are used for data processing in medicine.

Antiretroviral medications have reduced the morbidity and mortality associated with HIV. In contrast to these benefits, medication errors involving antiretrovirals represent opportunities for causing harm. We analyzed 400 medication errors that involved at least one single or combined HIV antiretroviral product that had been reported to a national medication error reporting program (MEDMARX). Our analysis revealed that 3% of the errors were harmful. Most of the errors (45%) occurred in the dispensing phase of the medication use process, a finding that differs significantly from many published studies. The most frequent types of errors were wrong dose (37.5%) and wrong medication (32%). Lamivudine (Epivir, Glaxo-SmithKline, Research Triangle Park, NC) was the most commonly identified product to be involved in the errors. Community hospitals were more likely to have prescribing errors than teaching hospitals. Similar brand and generic names were associated with many of the errors. With frequent dosing of many HIV medications, health care organizations must have a process to clarify orders rapidly and maintain current references (including photos) of antiretrovirals. Prescribers should clearly spell out the intended product and avoid abbreviations. Participating in voluntary medication error reporting programs increases the awareness of the threats of medication errors involving antiretrovirals.