Adults regularly use word-gesture combinations in communication, and meaningful gestures facilitate word learning. However, it is not clear if this benefit of gestures is due to the speaker’s movement increasing the listener’s attention or if it needs to be a meaningful gesture, if the difficulty of the task results in disparate reliance on gestures, and if word classes are differentially affected by gestures. In the present research, participants were measured on their novel word learning across four gesture conditions: meaningful gesture, beat gesture, nonsense gesture, and no gesture with extended training (Study 1, n = 139) and brief training (Study 2, n = 128). Overall, meaningful gestures and high frequency words led to the highest word learning accuracy. This effect of word frequency did not hold true for beat gestures after brief training suggesting that adding rhythmic information — if not adding semantic information — may detract from word learning. This research highlights the importance of considering task difficulty when analyzing the effects of gestures.