Listening Assessment: Literal Success, Inference Failure
Mr. Alcázar's eighth-grade Spanish 1B class has 22 students, including 7 with IEPs and 3 newcomers who have been in the country less than a year. He uses a 2:40 podcast interview recorded by a speaker from coastal Colombia at about 145 words per minute. The speaker uses sinalefa and drops final /s/, and says 'se me hace que' twice. Students listen twice and answer 10 multiple-choice items. Results: 18 of 22 answer literal-detail items correctly; 12 of 22 answer the main-idea item; only 5 of 22 answer the two inference items; 4 of 22 identify the meaning of 'se me hace que.' Mr. Alcázar's first plan is to replay the whole recording a third time and then move on. His department chair asks him to look at the item-level data before deciding. The next class is 50 minutes, and he has the transcript available.
1. Analyze the item-level pattern. What does it suggest, and why is replaying the whole recording unlikely to fix the problem?
Read the worked answer
The pattern is uneven, not global: literal details are strong (18/22), main idea is moderate (12/22), and inference is weak (5/22). That suggests students can catch words and phrases but are not building a situation model or linking evidence to an implied meaning. Replaying the whole recording mostly gives another pass at the same stream; it does not teach them what an inference is or how to justify one. The accent is probably not the main barrier, because literal items would also collapse if decoding were the primary issue. Mr. Alcázar should chunk the audio into 20–30 second segments, pre-teach 'se me hace que' as 'it seems to me,' model a think-aloud with one inference, and use a claim-evidence organizer: 'The speaker implies ___ because she says ___.' He can replay only the relevant segment, not the whole interview, and then have pairs retell the segment before answering a new inference item.
2. How can Mr. Alcázar protect access for students with IEPs and newcomers without turning the inference task into a literal lookup task?
Read the worked answer
He needs to separate access supports from construct changes. Access supports include glossed key vocabulary, a visual timeline of the interview, one replay of a shorter segment, partner retell before individual response, sentence frames for evidence, and extended time or preferential seating as specified in IEPs. Those supports let students hear and organize the content. The construct-making an inference and supporting it with evidence-should remain. He should not provide the inference answer in the gloss or accept a copied literal sentence as the response. For newcomers, a bilingual glossary for a few concrete nouns is reasonable, but the inference question should still require linking two ideas. He can assess the same construct with a scaffolded oral response or a selected-evidence item. This distinction keeps the task fair without lowering the cognitive demand.
3. One inference item asks why the speaker laughs after mentioning her grandmother. Several students choose 'because she is angry.' What should Mr. Alcázar conclude?
Read the worked answer
He should first check whether the item depends on something students were never taught: prosody, cultural expectations about laughter, or a rapid shift in the interview. If the laugh follows a warm anecdote and the distractors include a literal word from the text, students may be matching words rather than using tone and context. That is an inference-skill issue, not just a vocabulary gap. He should model how tone, discourse markers, and the next sentence constrain meaning: 'She laughs, and then she says her grandmother always fed everyone-so the laugh is affectionate, not angry.' Then give a parallel item with evidence. He should also review the item itself: if the correct answer is not defensible from the segment, revise it. The goal is to teach students to use contextual and prosodic cues, and to avoid keeping a flawed item just because it discriminates.