Earlier studies suggest that word length is influenced by the linguistic context to be precise and concise at the same time. The present study investigates whether the referential-situational context can also have an effect on the expected length of words. To test this assumption a salient property of the situational context, that is, the frequency of the unfamiliar referents was varied. The participants watched pictures of novel objects in the observational phase, presented either frequently or rarely. In the test phase they saw the same pictures of objects one by one and were asked to select one of two unfamiliar labels, which – according to them – could be the name of the object displayed. The two labels provided for each object at test had either short or long orthographic length. It was hypothesized that participants will select the long label more frequently when they had to guess the name of rare objects in contrast to frequent ones. The findings supported this hypothesis. Rare objects were paired with long labels significantly more often than frequent objects, resulting in a significant difference also when contrasted to chance-level. The results were similar if abbreviated or completely different label pairs were presented to the participants in the test phase suggesting that the situational context is taken into account when language users infer word form.