Age of Acquisition Ratings validated by actual vocabulary scores

Abstract

In this talk, we present new empirical support for the claim that Age of Acquisition (AOA) word ratings are a valid measurement of the age at which a word has actually been acquired. We show that these AOA values are the best predictors of lexical knowledge as tested with 139 Dutch-speaking children. The Age of Acquisition (AOA) effect (Morrison and Ellis 1997, Brysbeart et-al 2000, Juhasz 2005) has received much attention in recent literature. The basic finding is that AOA ratings of words strongly correlate with accessibility values – the earlier a word has been acquired, the faster it is accessed. This phenomenon has been demonstrated for word recognition, word production, recall, lexical access in aphasic patients and other populations with brain damage. It has been also demonstrated with second language speakers and with non-words learning. Objections to the AOA phenomenon divide into two types. The first claims that the AOA effect is a secondary effect that can be reduced to frequency. The second type of objections cast doubt on the methodology through which the AOA ratings are collected. These ratings are obtained through subjective evaluations of a (usually small) group of respondents who are asked to estimate for a large list of words, at which age they think they have acquired which word. Apart from the fact that this method seems highly subjective, claims have been made that participants’ estimations are affected exactly by the ease in which they can access a word and are therefore not a valid measurement of the words’ real AOA. However, recent findings (Brysbeart and Biemiller, 2017) have shown that subjective English AOA ratings correlate with actual test-based values of English words from 37 years ago, as measured in a vocabulary test conducted in 1981. This supports the claim that AOA ratings are real and cannot be explained by the method of collection. We report additional evidence for the relation between AOA ratings and actual test-based values: Using a new digital coloring method, which allows for a fast and reliable measurement of children’s vocabulary, we tested 139 Dutch-speaking children, all of the same age group (5-6 years old) on their knowledge of 74 Dutch words. This “knowledge-value” of the words was correlated with six different measurements, which are currently used for word frequency and vocabulary assessment in the Netherlands. Five of them are frequency-based and the sixth is the subjective AOA ratings (Brysbeart et al 2014). The frequency-based lists were: (1 and 2) Schrooten and Vermeer 1994 estimation of childrens input based on 15000 words taken from children books (we used both the general frequency and the specific 4-6 year old frequency). (3) The BAK list (Kuiken and Droge 2010) which is the official list used as a guideline by Dutch schools. (4) SubtlexNL (Keuleers et-al 2010) – a general frequency measurement based on TV and movies subtitles. (5) BASILEX (Tellings et-al 2014) – a list based on 11,4 million words collected from children books. Surprisingly, children’s actual knowledge values correlated the strongest with the AOA ratings (r=0,63; p<0,01). Partial correlation analysis shows that this relation between AOA ratings and children’s current vocabulary knowledge remains significant even when the frequency-based values are controlled for. We claim that these results strengthen the view of AOA as a real phenomenon, and of AOA rating as a valid measurement of the age at which a word has actually been acquired. Further, we claim that guidelines and predictions for vocabulary development in children should be not based on frequency of words in the input, but rather on empirical measurements of actual word knowledge.