Recall and recognition describe two distinct memory tasks: recall requires producing a target phrase from internal memory alone, while recognition requires identifying that target among visible alternatives. In collocation practice, the distinction shapes everything from exercise format to apparent difficulty. A recall-style drill shows a complement such as "heed" or "breath" and asks the learner to write or speak the full phrase from memory with no on-screen verbs to choose from. A recognition-style drill shows the same complement but also lists "take," "catch," "make," and "give" as numbered alternatives, and the learner picks the verb that completes the source-backed phrase. The recognition version reduces the mental search because every plausible answer is on the page; the recall version forces the learner to retrieve the verb without any prompt beyond the noun phrase itself. That difference explains why a four-choice collocation game feels substantially easier than a fill-in-the-blank test, and why neither one alone can fully describe what you know about English phrases.
Why the Two Tasks Feel Different in Phrase Learning
Cognitive psychology has long described recall as the harder of the two tasks. Recall demands that the learner reconstruct the target from internal cues alone — no labels, no options, no hints beyond the trigger word. Recognition, by contrast, requires only a judgment: does this displayed alternative match what I already know? The mental work shifts from generation to discrimination, which is a much lighter demand on working memory.
That shift matters most for collocations, because collocations are arbitrary pairings. There is no grammatical rule that forces "pay" to attach to "heed" rather than "take heed"; both forms appear historically, and modern usage strongly prefers "pay heed." A learner encountering "heed" with no options has to search through every plausible partner, weigh frequency, and commit. A learner seeing the same complement with four numbered verbs only has to ask which one is correct for the displayed target. The first task tests productive knowledge; the second tests discriminative knowledge. Both are useful, and they answer different questions about what you know.
| Dimension | Recall | Recognition |
|---|---|---|
| Task shape | Produce the full phrase | Select the correct phrase from visible options |
| Information on screen | Only the complement (e.g., "heed") | Complement plus four numbered verb choices |
| Mental operation | Generate and verify from memory | Match a known target to an on-screen option |
| Typical failure | Cannot retrieve the verb at all | Retrieves a verb but picks the wrong one |
| Common format | Fill-in-the-blank, free writing, oral production | Multiple choice, matching, four-option selection |
| Cue strength | Low — only the trigger word | High — verb candidates are visible |
| What it measures | Productive availability | Discriminative knowledge of the listed set |
| Sensitivity to interference | Higher — multiple partners compete | Lower — competition is bounded by the options |
The bottom row of the table points to a real limit: the recognition format constrains interference. A recall task exposes you to the full space of possible partners; a recognition task only exposes you to the four the designer chose. That makes recognition easier, more reproducible, and far easier to score automatically — which is why almost every browser-based phrase game uses it.
How Collocation Match Uses a Recognition Format
Collocation Match is built entirely around the recognition task. Each round displays a complement — "heed," "a glimpse," "a breath," and so on — and pairs it with four numbered verb choices. All four choices are verbs, so the visible shape of the answer does not give away which one belongs to the displayed phrase; you cannot rely on grammar or word-class filtering to narrow the field. The correct answer is the source-backed target for the complement shown, never a claim that the other three verbs are universally ungrammatical.
The eight fixed targets are sourced from Princeton WordNet 3.1 and independently confirmed on dedicated Cambridge phrase pages. Cambridge's own grammar reference defines collocation as words that are commonly used together, which is the broad idea this game isolates into eight concrete phrase pairs. Because the route is fixed, every match you complete can be checked against the same external data, not against an unseen random deck. That fixed-route property is what makes the recognition task auditable: another player sitting at the same round token will see the same complement and the same four choices you did.
Recognition format also lets the game keep its scoring transparent. A correct selection against the current round token advances exactly one round; a wrong selection increments the error counter and leaves the same prompt visible so you can try again. There is no hidden penalty, no silent question swap, and no language grade attached to the result. The errors and points you see at the end are product-authored numbers that describe only how you fared on this eight-round route.
How to Complete the Eight-Round Recognition Route
- Start the route. Open Collocation Match and select Start. The status text updates to announce that round one is ready, and the first complement appears with four numbered verb choices.
- Read the displayed complement and the four verb options. Each round shows a short noun phrase (such as "heed" or "a glimpse") and four verb buttons labelled one through four. All four buttons are verbs, so the answer is not revealed by word class alone.
- Choose the source-backed verb for the displayed complement. Click the matching button or press the corresponding number key 1 to 4. For a closer look at how recall and recognition relate to fixed-route drills, the immediate vs delayed sentence recall walkthrough uses a similar transparent route structure.
- Watch the round advance on a correct choice. A correct verb for the current round token moves you to the next round. The matched pair is then disclosed so you can confirm the source-backed target against an external reference.
- Recover from a wrong choice by trying again on the same prompt. An incorrect button adds one error and keeps the complement and four choices visible. There is no penalty beyond the error counter, and you can re-enter your selection until it matches the route target.
- Finish all eight rounds. The full route covers pay heed, catch a glimpse, take a breath, make sense, give birth, draw a blank, keep pace, and have a look. Each pair appears once, in a fixed order, with eight unique route keys and eight unique complements.
- Read the result panel for matches, errors, and recreational points. Once the eighth target is accepted, the result panel shows the totals. Points are a transparent product rule rather than a score published by WordNet or Cambridge.
- Restart if you want another pass. Restart repeats the same verified route. Familiarity with the order is part of why repetition changes performance, and that change should not be read as standardized improvement.
What Recognition Cannot Verify About Your Collocations
Recognition makes a deliberately narrow claim. The four-choice format can confirm only the exact eight collocations on the route, and only for the senses WordNet 3.1 and the corresponding Cambridge phrase pages record. A different verb may form another valid expression in a different sentence, dialect, tense, or context — for example, "take heed" appears in older usage even though the route target is "pay heed." The contract is explicit: the page does not claim every distractor is universally ungrammatical, and it does not parse arbitrary writing.
Recognition also cannot tell you whether you can produce the collocation in a real sentence. Selecting "take" beside "a breath" tests discriminative knowledge of the phrase "take a breath"; it does not test whether you can compose a paragraph that uses the same phrase naturally. If you want to test productive availability, you need a recall-style task — a fill-in-the-blank prompt, a free writing exercise, or a delayed sentence recall drill — and the score on a recognition task is not a substitute for that evidence.
Finally, recognition cannot reach beyond the route. Inflected alternatives such as "takes a breath" or "took a breath," regional variants like "have a look" versus "take a look" in some dialects, and any collocation not in the eight-item list fall outside the table the game verifies. The Cambridge entry for catch a glimpse, for instance, carries usage notes the four-choice format cannot surface. Recognition is precise about what it covers and silent about what it does not.
Why This Is a Game, Not a Proficiency Test
Eight disclosed phrases cannot support a professional, school, employment, treatment, eligibility, or health decision. The route is fixed, the distractors are bounded, and there is no norm group, CEFR scale, placement protocol, or vocabulary-size estimate behind the result. The page does not assign a language grade, block the session, or silently substitute a different question, and it does not diagnose a learning condition, measure vocabulary size, rate intelligence, or show whether your brain has improved.
What the game does offer is an auditable phrase-matching puzzle with full keyboard access, transparent scoring, and a clearly disclosed route. Every accepted answer is tied to the current round token; repeated key presses are no-ops, and stale clicks from a previous round cannot score a hidden phrase. The compact choice grid wraps on small screens and works with pointer, touch, or the keys 1 through 4 without depending on colour. Those are reliability properties of a recreation tool, not claims about your English.
Used as a recognition warm-up before a recall drill, or as a quick check on the eight collocations it covers, Collocation Match sits comfortably inside the larger practice of collocation study. Used as the only evidence of your phrase knowledge, it under-represents what you know, because recall and recognition answer different questions and a single four-choice game can only answer the recognition one.