Limits...
A computational model associating learning process, word attributes, and age of acquisition.

Hidaka S - PLoS ONE (2013)

Bottom Line: Simply put, this model formulates how different learning processes, characterized by change in learning rate over time and/or by the number of exposures required to acquire a word, likely result in different AoA distributions depending on word type.The first analysis showed that the proposed model accounts for empirical AoA distributions better than a standard alternative.We further discuss the theoretical implications of our model-based approach.

View Article: PubMed Central - PubMed

Affiliation: Japan Advanced Institute of Science and Technology (JAIST), Ishikawa, Japan.

ABSTRACT
We propose a new model-based approach linking word learning to the age of acquisition (AoA) of words; a new computational tool for understanding the relationships among word learning processes, psychological attributes, and word AoAs as measures of vocabulary growth. The computational model developed describes the distinct statistical relationships between three theoretical factors underpinning word learning and AoA distributions. Simply put, this model formulates how different learning processes, characterized by change in learning rate over time and/or by the number of exposures required to acquire a word, likely result in different AoA distributions depending on word type. We tested the model in three respects. The first analysis showed that the proposed model accounts for empirical AoA distributions better than a standard alternative. The second analysis demonstrated that the estimated learning parameters well predicted the psychological attributes, such as frequency and imageability, of words. The third analysis illustrated that the developmental trend predicted by our estimated learning parameters was consistent with relevant findings in the developmental literature on word learning in children. We further discuss the theoretical implications of our model-based approach.

Show MeSH

Related in: MedlinePlus

Proportion of words that best fit each model for different age groups.Ages are grouped based on median AoA from 16 to >30 (shown as 31).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3815221&req=5

pone-0076242-g004: Proportion of words that best fit each model for different age groups.Ages are grouped based on median AoA from 16 to >30 (shown as 31).

Mentions: Do AoA distributions change during the course of development? As AoA distributions reflect the underlying learning process, if the answer is yes, it would imply a developmental change in the learning process. Here we provide basic descriptive statistics from the model-fitting for each age interval. Average estimated AoA based on the AoA distributions correlated to conventional median AoA (r = 0.952, 0.954, and 0.923 (p<0.001 for all) for cumulative, rate-change, and cumulative-and-rate-change learning models, respectively). Because of these high correlations, we adopted the conventional median AoA for the age bins in the following analysis so that our results would be comparable with those of other studies that utilized conventional AoA. Figure 4 shows the proportions of words that best fit with each model for each median AoA interval, which was defined as the first month where an acquisition rate of ≥50% was observed for a given word. The proportion of the rate-change learning model having the best fit clearly increases with age, while conversely that of the cumulative-and-rate-change learning model declines. In particular, from 20 to 25 months of age, a sharp peak shift from the cumulative-and-rate-change to the rate-change learning model can be observed. This indicates that for words learned later as median AoA (not as late learners in the tail of AoA distribution), the late learners learn them “faster” than early learners did, because learning rate increases with time in the rate-change learning model (see Figure 5; to be explained later). By “faster”, we mean a higher rate of new learners per unit time out of children who do not acquire the words. Moreover, around 20 months of age the cumulative-and-rate-change learning model peaks, an age that corresponds approximately to the vocabulary spurt period – the putative onset of fast vocabulary growth. In addition, it is worthwhile to consider that 18 months of age or later is known as the period when children start to show systematic generalizations for novel words [14]. We will discuss the implications of this systematic change in AoA distributions as they relate to underlying developmental learning processes and the vocabulary spurt period in the later section. Note that the patterns of AoA distributions for each age group are only descriptive, and meaningful discussion must be preceded by careful and in-depth analyses since the distributions are also highly dependent on word class (Figure 3), and correlate with psychological factors such as word frequency (See also Study 2).


A computational model associating learning process, word attributes, and age of acquisition.

Hidaka S - PLoS ONE (2013)

Proportion of words that best fit each model for different age groups.Ages are grouped based on median AoA from 16 to >30 (shown as 31).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3815221&req=5

pone-0076242-g004: Proportion of words that best fit each model for different age groups.Ages are grouped based on median AoA from 16 to >30 (shown as 31).
Mentions: Do AoA distributions change during the course of development? As AoA distributions reflect the underlying learning process, if the answer is yes, it would imply a developmental change in the learning process. Here we provide basic descriptive statistics from the model-fitting for each age interval. Average estimated AoA based on the AoA distributions correlated to conventional median AoA (r = 0.952, 0.954, and 0.923 (p<0.001 for all) for cumulative, rate-change, and cumulative-and-rate-change learning models, respectively). Because of these high correlations, we adopted the conventional median AoA for the age bins in the following analysis so that our results would be comparable with those of other studies that utilized conventional AoA. Figure 4 shows the proportions of words that best fit with each model for each median AoA interval, which was defined as the first month where an acquisition rate of ≥50% was observed for a given word. The proportion of the rate-change learning model having the best fit clearly increases with age, while conversely that of the cumulative-and-rate-change learning model declines. In particular, from 20 to 25 months of age, a sharp peak shift from the cumulative-and-rate-change to the rate-change learning model can be observed. This indicates that for words learned later as median AoA (not as late learners in the tail of AoA distribution), the late learners learn them “faster” than early learners did, because learning rate increases with time in the rate-change learning model (see Figure 5; to be explained later). By “faster”, we mean a higher rate of new learners per unit time out of children who do not acquire the words. Moreover, around 20 months of age the cumulative-and-rate-change learning model peaks, an age that corresponds approximately to the vocabulary spurt period – the putative onset of fast vocabulary growth. In addition, it is worthwhile to consider that 18 months of age or later is known as the period when children start to show systematic generalizations for novel words [14]. We will discuss the implications of this systematic change in AoA distributions as they relate to underlying developmental learning processes and the vocabulary spurt period in the later section. Note that the patterns of AoA distributions for each age group are only descriptive, and meaningful discussion must be preceded by careful and in-depth analyses since the distributions are also highly dependent on word class (Figure 3), and correlate with psychological factors such as word frequency (See also Study 2).

Bottom Line: Simply put, this model formulates how different learning processes, characterized by change in learning rate over time and/or by the number of exposures required to acquire a word, likely result in different AoA distributions depending on word type.The first analysis showed that the proposed model accounts for empirical AoA distributions better than a standard alternative.We further discuss the theoretical implications of our model-based approach.

View Article: PubMed Central - PubMed

Affiliation: Japan Advanced Institute of Science and Technology (JAIST), Ishikawa, Japan.

ABSTRACT
We propose a new model-based approach linking word learning to the age of acquisition (AoA) of words; a new computational tool for understanding the relationships among word learning processes, psychological attributes, and word AoAs as measures of vocabulary growth. The computational model developed describes the distinct statistical relationships between three theoretical factors underpinning word learning and AoA distributions. Simply put, this model formulates how different learning processes, characterized by change in learning rate over time and/or by the number of exposures required to acquire a word, likely result in different AoA distributions depending on word type. We tested the model in three respects. The first analysis showed that the proposed model accounts for empirical AoA distributions better than a standard alternative. The second analysis demonstrated that the estimated learning parameters well predicted the psychological attributes, such as frequency and imageability, of words. The third analysis illustrated that the developmental trend predicted by our estimated learning parameters was consistent with relevant findings in the developmental literature on word learning in children. We further discuss the theoretical implications of our model-based approach.

Show MeSH
Related in: MedlinePlus