A plausible misreading: scoring inference against an ironic article
Your students read an authentic 450-word article from a Montreal weekly about a borough's summer pedestrian-street pilot. It includes a bar graph of merchant revenue before and after the pilot (+8% on the main street, −3% on the parallel street) and closes with an ironic line about 'le triomphe tranquille des tables à pique-nique.' Question 4 asks students to infer the author's attitude toward the project; Question 5 asks what 'achalandage' means in context. Sami's literal comprehension questions are all correct, and his answers are written in complete, well-formed sentences. For Q4 he writes: 'L'auteur est contre le projet parce qu'il mentionne la baisse de 3 % sur la rue parallèle.' For Q5 he writes: 'les achats.' The marking key gives 2 points for Q4 (attitude plus justification) and 1 point for Q5.
1. Score Q4 and Q5. Where you award partial credit, justify it, and say what the key itself should have allowed.
Read the worked answer
Q4 - 1/2. Sami's justification is textually grounded and factually accurate: the −3% figure does appear. His inference is wrong, though. The article's overall movement is a modest improvement, and the closing image of picnic tables 'triumphing quietly' is affectionate irony, not opposition. The key's two components let me separate the operations: 0 for attitude, 1 for citing relevant textual evidence. That split is the honest score, because it records both what he can do and what he cannot.
A single score of 2 would reward a misreading because it was well written; a 0 would hide that he located and used evidence.
Q5 - 0/1. 'Les achats' confuses the thing counted with the counting. 'Achalandage' refers to the flow of customers or foot traffic through a commercial area; the surrounding sentence typically contrasts it with sales figures. The item tests deriving meaning from co-text, and he did not do it.
The key should have specified partial credit for Q4 and given a model paraphrase rather than a single English gloss for Q5.
2. What does this pattern tell you about Sami as a reader, and what does it not tell you?
Read the worked answer
What it tells you: he decodes fluently and locates explicitly stated information - every literal item is correct. He does not yet integrate a graph with prose, does not use co-text to resolve a low-frequency word, and does not register evaluative tone. That combination is a text-level comprehension profile, not a word-level one.
What it does not tell you: that his vocabulary is generally weak. 'Achalandage' is low-frequency and specific to Québec commercial French; missing it may reflect exposure, not lexical deficit. Nor can you conclude he cannot reason inferentially, because Q4 shows reasoning applied to accurate but partial evidence - a different failure from guessing.
One confound needs flagging: his answers are polished sentences. Strong written production can make a weak comprehension answer look more competent than it is, and the reverse happens to students who understand but write poorly. Before finalising, ask him orally, in French, to explain what the author thinks of the project and which words signalled it. That second sample separates comprehension from expression.
3. Propose one change to Q4 and one to Q5 that would separate inference from vocabulary and from graph reading.
Read the worked answer
Q4 change: split it into two scored parts - 'L'auteur est-il favorable, défavorable ou ironiquement favorable au projet? Cite une expression du texte (et non du graphique) qui appuie ta réponse.' The first part isolates stance; the second forces the student to cite evaluative language in the prose rather than a data point. The graph then becomes its own item, where reading a −3% bar and deciding what it does and does not support is scored as data interpretation.
Q5 change: replace the single gloss with a two-stage item - 'Dans la phrase, remplace « achalandage » par une expression de ton choix qui garde le même sens' plus 'Quel mot de la même phrase t'a aidé?' This makes the strategy visible and gives partial credit to a student who reasons correctly but chooses an approximate paraphrase.
Both changes shorten the marking key's work by making each item test one operation. That also produces usable diagnostic information: a class-wide pattern of correct stance and incorrect evidence citation is a different teaching problem from the reverse.