The standard vocabulary-learning loop has a design flaw. You see a word in your target language, look up its translation, read the definition in your native language, and then try to commit the link to memory. The problem is that your brain never builds a direct connection between the sound of the word and the thing it refers to. It builds a connection between two pieces of text — which means every time you hear the word spoken, you have to route it through a translation proxy before you understand it. That half-second of translation is exactly what makes fluent listening feel out of reach.

The listening games here bypass that proxy entirely. You hear a spoken word and you select a picture — no text, no translation, no native-language intermediary. The brain is forced to build a direct neural link between the acoustic signal and the concept. This is the mechanism that Paivio's Dual-Coding Theory (1971) describes: verbal and visual information are processed in separate channels, and learning is most durable when both channels are activated at once by the same event. Hearing "basketball" while looking at 🏀 is a richer encoding than reading "basketball = baloncesto" will ever produce.

Mayer's Cognitive Theory of Multimedia Learning (2002) adds a practical constraint: working memory is limited, and anything on screen that isn't the audio signal and the pictures is competing for the mental resources you need for the actual task. This is why the games have no translation boxes, no vocabulary lists, no decorative panels. Every removed element is cognitive load returned to the learner. Sweller (1988) called this principle extraneous load reduction — the design job is to eliminate everything that doesn't directly serve the learning objective.

Krashen's Input Hypothesis (1982) provides the third pillar. Acquisition happens when you receive comprehensible input slightly above your current level — what Krashen called i+1. The adaptive level system in both games operationalises this directly: three correct answers raises the level (harder words, faster speech, more choices), one miss lowers it. The game continuously hunts for the edge of your comprehension and sits there.

There is also a practical argument for games over static pages in the current web environment. A table of vocabulary translations can be summarised by an AI assistant in one sentence. An interactive game that speaks words at you, responds to your answers, tracks your missed words, varies the voice and the speed, and remembers your progress across sessions cannot be replaced by a text summary — because the value is in the doing, not the knowing. The 700 spoken words across these two games exist to be heard and identified repeatedly, not read once. That distinction is the whole point.