Given the lack of empirical corpus-based frequency counts in many languages, it would be useful and of theoretical interest if judgements of relative frequency of words in a language by proficient speakers of that language could substitute objective frequency counts for the purposes of devising language teaching materials, tests, and research instruments. The present paper reports on three investigations of frequency judgements. Using different methodologies and with varying sizes of word samples, it is shown that judgements by professional linguists do not correlate highly with corpus-based frequency counts. There are considerable individual variations in judgements of frequency, and the most reliable results are likely to be achieved by aggregating and averaging judgements. However, even averaged judgements do not predict frequency counts particularly well. This research does not inspire confidence in word frequency judgements as surrogates for objective frequency measures. It is suggested either that judgements of word frequency, even by highly educated native speakers of a language, may not be very accurate estimates of the frequency of words in their language, or that even large modern corpora are inadequate indicators of word frequency as experienced by individuals and groups of individuals. Further research is needed into the nature of intuitions about word frequency.