Broadly, the set of games that have been developed for LANGA can be divided in two categories: teaching and practice games. New words are taught in groups of four;  for each new group of words, teaching games are always played first. In the teaching games, the four pictures corresponding to the words to be learned are shown. The learner clicks on any picture to hear its spoken form (produced by a male or female native speaker of the language to be learned). After hearing the word, the learner must repeat the word. LANGA’s speech recognition engine (SRE) then matches this input against its “grammar” (set of possible recognition targets), which consists of phonetic encodings of each of the four words being taught. The SRE returns the word that it matches the input to in the grammar (i.e., its “guess” as to which word was spoken out of the possible options), as well as a confidence score for that guess. The SRE’s guess matches the target word, and the confidence score is over a threshold level (determined empirically through pilot testing), the learner receives feedback in the form of a green “thumbs up” icon and a high-pitched tone; otherwise, a red, “thumbs down” icon is shown along with a low-pitched tone. If the learner’s speech does not yield a “correct” response from the SRE for three trials in a row, the program displays the written form of the word as a further aid to pronunciation. Learners must produce each word correctly at least once to advance to the practice games. Practice games challenge the user to speak the target learning words in different contexts; all use the same SRE scoring procedure as the learning game. All involve presenting the pictures corresponding to the learning words, but in different gamified contexts that have been borrowed from popular casual gaming apps (based on the hypothesis that the popularity of these game mechanics reflects the fact that people generally find them engaging, which will in turn facilitate learners investing the necessary amount of practice required for successful learning). Currently, the version of LANGA that has been tested includes three practice games: Missing, Doubles and Match-3. The Missing game consists of four trials.  In each trial, pictures of three of the four words from the learning set are presented (see Illustration of the four games used to teach secon...), prompting the user to name the missing one (which changes from trial to trial). In the Doubles game four pictures are presented, but two of them are the same item; the user has to name the word corresponding to the picture that appears twice. In the Match-3 game, the pictures were arranged as tiles in grid. The goal is to swap pairs of adjacent tiles in order to align three copies of a picture, at which point those tiles disappear and new  tiles enter the grid. To swap tiles, the user must tap/click on a tile, correctly speak its name, then tap on an adjacent tile and correctly speak its name. The game ends when each picture has been named successfully at least four times.