NB Francophone G2 Reading common mistakes
NB Francophone G2 Reading · Understand the traps and how to avoid them.
When reading aloud a pseudoword or unfamiliar word such as « brou » or « plir », the candidate guesses a whole familiar word from the first letter or the overall shape (says « boule » for « brou ») instead of sounding out each grapheme in order.
The candidate has been trained to rely on sight-word recognition and global shape cues, so when no lexical entry matches, they fall back on guessing rather than applying graphème-phonème correspondances; French complex graphemes (ou, on, an, ch, oi, eau) are not yet automatised.
What to do instead
Practise daily decoding of lists of non-words built only from taught correspondances (pa, ri, chon, vau, blin), requiring the candidate to point to each grapheme, say its sound, then blend left to right. Model the difference between a guess and a decode, and correct any global guessing immediately by covering the word and revealing it sound by sound.
The candidate reads « les amis » as « les zamis » or « un ours » as « un nours », and confuses silent final consonants, reading « petit » as « petite » or pronouncing the -ent of « ils parlent ».
The candidate applies a one-letter-one-sound rule and has not internalised French elision, liaison, and silent-terminal-letter conventions; these are orthographic regularities rather than pure phonics, and they are often taught incidentally rather than practised.
What to do instead
Run short focused drills contrasting minimal pairs (le héros / les héros, un ami / des amis, il parle / ils parlent) and have the candidate mark silent letters with a light cross before reading the word aloud. Include liaison practice in short memorised phrases read repeatedly until fluent.
Reading is accurate but halting and word-by-word, with long pauses at line breaks and no attention to punctuation; the candidate finishes a 60-word passage far under expected pace and cannot answer a question about it without rereading every word.
The candidate decodes each word as a separate problem, so working memory is consumed by decoding and none is left for meaning; they have practised accuracy but not repeated reading for automaticity.
What to do instead
Use repeated reading of the same short passage three to four times, timing the first and last read and charting the gain. Pair this with phrase-level chunking: mark slash boundaries at natural phrase units and have the candidate read the phrase, not the word, before moving on.
The candidate reads high-frequency irregular words such as « monsieur », « femme », « fils », « est », « au », « du » letter by letter and produces a distorted pronunciation, then loses the thread of the sentence.
These words violate regular correspondances and must be learned as whole units; the candidate has not built an automatic sight vocabulary for them and keeps trying to decode them phonically.
What to do instead
Build a small set of 20 to 30 high-frequency irregular words and practise them as flash recognition until response is under two seconds, then embed them in sentences. Explicitly tell the candidate which words are 'words we learn by heart' so they stop applying decoding rules to them.
The candidate meets an unknown word such as « énorme » or « grogner » in context and stops, or asks an adult, instead of using the surrounding sentence to infer a plausible meaning and continue reading.
The candidate treats unknown vocabulary as a decoding failure rather than a meaning problem, and has no repertoire of context strategies; they may also fear making an error.
What to do instead
Teach and rehearse a fixed three-step routine: read to the end of the sentence, ask 'what would make sense here?', then check the word's first letters against the guess. Practise on passages where the target word is strongly supported by context, and praise sensible guesses even when imperfect.
The candidate can read a sentence aloud correctly but cannot answer a simple 'qui fait quoi, où, quand' question about it, or answers with a word that appeared in the text but does not answer the question.
The candidate is decoding without constructing a mental model; when asked a question they search the surface string for a matching word rather than retrieving the meaning of the sentence they just read.
What to do instead
Insert a stop-and-tell pause after every one or two sentences: the candidate must say in their own words who did what. Use very short texts initially and ask only literal wh-questions whose answer is stated in one sentence, requiring the candidate to point to the exact sentence that supports the answer.
Asked an inferential question such as 'Pourquoi le garçon est-il triste?', the candidate says 'ce n'est pas écrit' and gives up, or invents an answer unrelated to the text.
The candidate does not distinguish 'stated in the text' from 'I can work it out from clues', and has not been taught to combine a text clue with background knowledge; inferencing is an invisible process that is rarely modelled.
What to do instead
Model inference aloud: state the clue from the text, state what you already know, then state the conclusion, using a two-column 'indice dans le texte / ce que je sais' chart. Begin with highly constrained inferences (a character puts on a coat and shivers: what is the weather?) before moving to emotion and motive questions.
The candidate confuses the purpose of a text's parts: treats a caption, a title, a bold word, or a diagram label as part of the running narrative, or ignores headings and the table of contents entirely when locating information.
The candidate reads all print as a single stream and has not learned that titles, headings, captions, and labelled diagrams carry different kinds of information; these features are often present but never explicitly taught.
What to do instead
Do a features walk-through before reading: name the title, headings, captions, and any labelled diagram, and state what each is for. Then set locating tasks ('find the page that tells you about…', 'which caption explains the picture?') that can only be solved by using those features.
The candidate reads a simple narrative and cannot retell it in order, mixing up the beginning and end, or omitting the problem and the resolution; when asked what happens first, they answer with an event from the middle.
The candidate has no internalised story schema (personnage, lieu, problème, solution) and reads for individual word meaning rather than for the arc of the text; sequencing words such as « d'abord », « ensuite », « enfin » are not noticed.
What to do instead
Use a graphic organiser with four boxes (qui/où, problème, que se passe-t-il, comment cela finit) completed after each reading. Highlight séquence markers in the text and have the candidate number events on strips of paper and physically reorder them before retelling.
When a sentence does not make sense, the candidate keeps reading to the end of the passage without self-correcting, and when asked 'did that make sense?' says yes; they also never reread after a decoding error that changed the meaning.
The candidate monitors only whether words were pronounced, not whether the text is meaningful; self-correction requires the reader to hold meaning in mind and compare it against what was decoded, a skill that must be explicitly taught and modelled.
What to do instead
Teach a 'ça a du sens?' check after each sentence, with the candidate giving a thumbs up or down and rereading any sentence marked down. Deliberately insert a nonsense word into a familiar sentence and have the candidate catch it, then transfer the same monitoring to their own reading with a short self-correction log.