[MUSIC] Our perspective uses the framework that multiple sources of information are influential in language processing and learning. These sources might also be considered constraints that help decrease the ambiguity of a word's meaning. A possible constraint in word learning, for example, might be the location where the word is likely to occur. Words referring to farm animals would tend to occur on farms, whereas words about sports would occur most likely in sports venues. The use of and the weight given to these sources of information or cues change across development. There is evidence that infants initially rely primarily on perceptual cues, and gradually begin to use a speaker's intent and linguistic cues to determine word reference. These cues and constraints are graded, not categorical in nature, suggesting further that they must be combine to give it more reliable understanding of the input. Evidence to date indicates that this combination process is highly efficient, or optimal, as described by a Bayesian like process. This theoretical framework does not predict or specify the magnitude of the various sources of influence. The relative contribution of potential influences is an empirical question. To study the potential influence of any potential source of information, it is also important to control for the contributions of other sources. Our research strategy is to study what influences the child to make this association between the spoken symbol and its meaning, one to be learned and remembered. To pursue this question, we will now review some possible influences in a child's vocabulary acquisition. Databases of speech occurrences are instrumental in this type of study. One assembled database consists of 5,000 transcriptions of children's speech, with a total of 3,500,000 word tokens. The ages of the children in this database range from birth to seven years. This database can be used to provide the dependent measure or an estimate of the words in a child's vocabulary at different ages. To the extent a word is found in the database, we can conclude that the word is more likely in a child's production vocabulary. You might think that just a linear account of word occurrences would be appropriate. However, most studies use logarithmic accounts, because it has been repeatedly demonstrated that the influence of a frequency variable on behavior tends to be logarithmic rather than linear. For example, if you were practicing some new activity, your first 10 practice trials gives you the same improvement as your next 100 trials. To achieve again the same improvement, you would need 1,000 more trials. When practice is plotted on a log scale, improvement can be seen as a linear function of practice. Some recent analysis carried out by colleagues and myself, will be used to illustrate this type of study. We chose a set of words known to be some of the first words that children understand and produce. We determined the number of times each of the 644 words in this set occurred across eight six-month age ranges between birth and four years of age in the database. Given that the size of the sample, of the total number of words at each age, differed, the dependent measure was the frequency count, normalized for 1 million occurrences. More frequent occurrences of a word would be analogous to a highly likelihood of the word being in the child's vocabulary. There has been a resurgence of research and applied interventions on the importance of the quality and quantity of language directed at children from birth onward. It is well established that many of the words children know are those that they have heard, especially in child-directed speech. Researchers first discovered that children before the age of three heard many more words in high than in low socioeconomic status families. Additional research revealed that the lexical richness and sentence complexity of mother's speech to their children also differed between these two types of families. Thus it appears that the quality as well as the quantity of language is important in children's language development. Thus it seems valuable to ask how the word frequency of child directed speech by parents and others influences vocabulary acquisition. As a measure of this parental input, a parental vocabulary corpus of spoken utterances from parents, caregivers, and experimenters in the presence of children between seven months and seven years of age, was used. The corpus provides a huge sample of the speech that children are exposed to, for example, dinner talk, talk during free play, and storytelling, even though not all of the utterances are strictly child directed. Given the potential influences of many sources, an evaluation of parental input frequency requires that these other potential influences be eliminated. This is achieved by a partial correlation in which the influence of other sources is eliminated. The partial correlation can range between 0 and 1 indexing the magnitude of influence of a given independent variable. [MUSIC]
