Fig 1.
Test-grid from Xu & Tanenbaum.
Example word learning trial with test-grid shown to participants. Figure created by the author as an illustrative adaptation of the paradigm in [14–16]; not reproduced from the original.
Table 1.
All trial data from [16].
Table 2.
Second-block trial data from [16].
Table 3.
First-block trial data from [16].
Fig 2.
Computation of mental representation from single training example and subsequent comparison to test objects.
Exact values are schematic and for illustration only.
Fig 3.
Algorithmic flow chart highlighting a possible path of NGM behavior under parallel-presentation.
Typical path of meaning extracted from parallel-presentation trial. All exemplars contribute to initial hypothesized meaning.
Fig 4.
Algorithmic flow chart highlighting a possible path of NGM behavior under sequential-presentation.
Typical path of meaning extracted from sequential-presentation trial. Subsequent “training” stimuli do not affect initial hypothesis so long as consistent. If inconsistent, then a new hypothesis is generated (not depicted here).
Fig 5.
Implementation of distance computation between an object and mental representation under the NGM.
Table 4.
Example learning and evaluation trial under the NGM.
Table 5.
Major patterns to be captured by models of word learning and generalization.
Fig 6.
Chart of all seven training configurations.
Conditions used for parameter tuning shown in light red. Time during training is indicated within each block vertically; the objects in the parallel condition are co-present at the same time, while the “sequential” trials training objects are never co-present.
Fig 7.
NGM predictions and empirical human data after training on a single item.
Gold bars represent human experimental results from Spencer et al. [15]; grey bars show output from the NGM. Error bars indicate standard deviations.
Fig 8.
NGM predictions and empirical human data across learning configurations with items presented sequentially.
(left) Subordinate training items (e.g., all dalmatians); (center) Basic training items (e.g., all dogs); (right) Superordinate training items (e.g., all animals). Gold bars represent human experimental results from Spencer et al. [15]; grey bars show output from the NGM. Error bars indicate standard deviations.
Fig 9.
NGM predictions and empirical human data across learning configurations with items presented simultaneously.
(left) Subordinate training items (e.g., all dalmatians); (center) Basic training items (e.g., all dogs); (right) Superordinate training items (e.g., all animals). Gold bars represent human experimental results from Spencer et al. [15]; grey bars show output from the NGM. Error bars indicate standard deviations.