Sentence-Level Errors Dominate Written English Struggles of Somali University Students

First-year students at SIMAD University in Mogadishu make far more sentence-level errors than word-level ones when they write academic English, according to a new descriptive study published in the Asian-Pacific Journal of Second and Foreign Language Education. Researchers Mustaf Ahmed Wehlie and Ahmed Abdullahi Mohamud analyzed 223 in-class essays of 100 to 150 words each, written on the topic of smartphone spending, and coded every observable deviation from standard written English. In total, the essays contained 2,270 coded error instances. Syntactic errors, which involve the arrangement, connection, and completeness of clauses and sentences, accounted for 49.87 percent of the total, followed by mechanical errors such as punctuation, capitalization, and spelling at 31.06 percent, and morphological errors involving verb tense at 19.07 percent. The study was grounded in Error Analysis theory and used a descriptive quantitative design, so the authors present the results as a taxonomy of written errors rather than a causal account of why the errors occur.

The participants were drawn from a target population of 528 first-year students enrolled in the English Skills Program at the university’s Institute of Modern Languages. Using Krejcie and Morgan’s sample size table, the researchers selected 223 participants through stratified random sampling, with faculty of study as the stratification variable. Proportional allocation applied an overall sampling fraction of 42.2 percent across seven faculties: Accountancy, Computing, Economics, Engineering, Management Sciences, Medicine and Health Sciences, and Social Sciences. Within each stratum, eligible students received anonymous serial numbers and were chosen by a computer-generated random list. The final sample was demographically balanced, with 50.2 percent male and 49.8 percent female students, and it was young: 74.9 percent were aged 18 to 20, 23.3 percent were 21 to 24, and only 1.8 percent were 25 to 29. The Faculty of Computing contributed the largest share of respondents at 22.0 percent, followed by Management Sciences at 19.7 percent and Accountancy at 18.8 percent.

Each participant wrote a short essay under controlled classroom conditions within 45 minutes, responding to the prompt: “People are spending too much on their smartphones. Why is this? Is it a positive or negative trend?” The researchers chose this topic to encourage authentic language use while eliciting varied morphological and syntactic structures. Essays were labeled with anonymous codes from P001 to P223, and all examples reported in the paper were extracted from the anonymized learner corpus rather than constructed for illustration. The authors emphasize that the examples retain the learners’ original wording as far as possible, which gives readers a direct view of the written forms first-year Somali university students actually produce in timed academic tasks.

Coding relied on an Error Identification Checklist developed from established Error Analysis and writing-assessment frameworks. It distinguished 11 observable error types: punctuation, tense, fragment, grammar, spelling, capitalization, word order, phrasing, preposition, run-on, and comma splice. These were grouped into three operational categories. Mechanical errors covered surface-level written conventions that affect visual form and readability, namely punctuation, capitalization, and spelling. Morphological errors covered word-level tense and verb-form marking, coded under a single tense category. Syntactic errors covered phrase-, clause-, and sentence-level problems, including word order, fragments, run-on sentences, comma splices, phrasing, preposition use, and general grammar errors. Where an error could plausibly overlap categories, coders classified it by the primary linguistic level affected: a missing full stop counted as mechanical, a fragment as syntactic, and a misspelled word as mechanical unless it changed an inflectional form.

The unit of analysis was the error instance, defined as one observable deviation from the target written form. Word-level errors such as spelling, capitalization, preposition, and tense errors were counted once per affected token, while clause- and sentence-level errors were counted once per affected clause or sentence rather than for every word. Punctuation errors were counted as separate mechanical instances, but a single sentence-boundary problem was not double-counted as both a punctuation error and a run-on or comma splice unless an additional distinct error was present. After coding, the frequency tables were audited to confirm that the 11 types were mutually exclusive and that category totals matched the overall count. The audited totals were 705 mechanical errors, 433 morphological errors, and 1,132 syntactic errors, summing to 2,270. Reliability was supported by two trained raters who independently coded 45 essays, roughly 20 percent of the corpus, reaching Cohen’s kappa of 0.82 and percentage agreement of 88.9 percent before resolving all disagreements by consensus.

The individual error rankings sharpen the picture of where difficulty concentrates. Tense errors were the single most frequent type at 433 instances, or 19.07 percent of all coded errors, indicating substantial difficulty with English verb inflection. Punctuation followed at 355 instances, or 15.64 percent, and sentence fragments came third at 265 instances, or 11.67 percent. General grammar errors, a syntactic subtype that included problems such as subject-verb agreement failures, accounted for 234 instances, or 10.31 percent. Spelling errors occurred 185 times, or 8.15 percent, and capitalization errors 165 times, or 7.27 percent. Word-order errors appeared 141 times, or 6.21 percent, phrasing errors 138 times, or 6.08 percent, preposition errors 120 times, or 5.29 percent, run-on sentences 119 times, or 5.24 percent, and comma splices 115 times, or 5.07 percent.

The paper illustrates each category with corrected examples from the corpus. Tense errors included forms such as “Many people will spending too much time on the internet,” where the gerund follows the auxiliary “will” instead of the base form, and “She is spending too much time on scrolling last night,” where the present progressive clashes with a past-time expression. Fragment errors included dependent clauses punctuated as full sentences, such as “Because he was tired.” Comma splices joined two independent clauses with a comma alone, as in “People are spending too much on their smartphones, this is a growing concern in society.” Word-order errors displaced objects and time expressions, as in “People spend online too much time on social media.” Preposition errors included “addicted with” instead of “addicted to,” and spelling errors included the transposed form “peopel.” The authors stress that these explanations describe the target-language rule violated, not the psycholinguistic origin of each mistake.

A methodological boundary runs deliberately through the study. Although the literature review discusses structural differences between Somali and English, such as Somali’s commonly verb-final clause organization and more flexible subject expression, the researchers did not code any individual error as interlingual or intralingual. They note that such source attribution would require interviews, think-aloud protocols, learner histories, or systematic contrastive evidence, none of which the essay-only dataset provides. The study therefore implemented only the descriptive stages of Error Analysis: data collection, error identification, description and classification, and frequency analysis. The authors also caution that the findings measure written grammatical accuracy in short essays, not broad English proficiency, oral performance, listening, reading, vocabulary knowledge, or pragmatic competence, and that lexical, semantic, rhetorical, and discourse-level errors fell outside the coding scheme.

For instruction, the authors argue that the distribution supports teaching sentence-level structure, tense accuracy, and written mechanical conventions as separate but connected skills. They suggest that instructors use students’ actual error patterns to build sentence-revision exercises: rewriting non-standard word order into standard subject-verb-object order, combining fragments into complete clauses, editing tense and verb forms, and revising paragraphs for punctuation, capitalization, spelling, and coherence. Curriculum planners, they write, should integrate explicit syntax-focused instruction and feedback-based writing practice alongside focused work on verb morphology and mechanical conventions, and policymakers might consider institutional writing support for first-year learners. The authors recommend longitudinal, cross-institutional, and mixed-method studies, including comparative work across Somalia and the wider Horn of Africa, to trace how written accuracy develops and to determine whether particular errors are interlingual, intralingual, instruction-related, or tied to other learning conditions.

Subject of Research: Written English error patterns among first-year EFL students at a Somali university

Article Title: Assessing mechanical, morphological, and syntactic error patterns in written English grammatical accuracy among English language learners at SIMAD University, Somalia

Article References: Wehlie, M. A., & Mohamud, A. A. (2026). Assessing mechanical, morphological, and syntactic error patterns in written English grammatical accuracy among English language learners at SIMAD University, Somalia. Asian-Pacific Journal of Second and Foreign Language Education, 11(1), Article 81. https://doi.org/10.1186/s40862-026-00457-x

Image Credits: AI Generated

DOI: 10.1186/s40862-026-00457-x

Keywords: error analysis, written English, EFL writing, grammatical accuracy, syntactic errors, verb tense, punctuation, sentence fragments, Somalia, SIMAD University, higher education, descriptive quantitative study