Subjective Familiarity of Words: Analysis of the Hoosier Mental Lexicon

We conducted a statistical analysis of several subsets of words from the Hoosier Mental Lexicon in order to examine some factors underlying the subjective familiarity ratings collected by Nusbaum, Pisoni, and Davis (1984). In this analysis, we grouped words into High-FAM (average familiarity rating greater than 6 on a 7-point scale), Mid-FAM (between 4.5 and 3.5), or Low-FAM (less than 2) sets, with primary interest being placed on the small set of Low-familiarity words (502 total), which were unknown to the majority of native listeners. This Low-FAM set consisted mainly of nonEnglish borrowings, technical terms, and archaic words. These items received much higher ratings from Speech Research Laboratory staff, who we assume have greater language experience, showing that the familiarity scale employed is not continuous for very low ratings, especially in terms of the properties of Low-FAM words. As expected, High-FAM words were the majority and had greater frequency of occurrence, greater neighborhood density and frequency, and were shorter on average than Mid or Low words. However, when these sets of words were controlled for length, we found that Highand Mid-FAM words did not differ significantly in neighborhood density and frequency, though High-FAM words were still significantly more frequent on average than Mid-FAM words, which were only slightly more frequent than the Low-FAM items. With word length equated, Low-FAM words still came from much sparser and lowerfrequency neighborhoods. This finding indicates that the difference in neighborhood density between Highand Mid-FAM items is due primarily to the generally shorter length of common, highly familiar words, whereas the lower density for Low-FAM words is genuinely due to differences in constituent structure. Words with Mid-range familiarity appear to comprise low frequency, midto high-density lexical items, which are most difficult to recognize according to the neighborhood activation model of spoken word recognition.