In a feature net model, knowledge of spelling patterns:Question 16Answera.can influence the perception of whole words but not the perception of bigrams.b.is distributed across the model, and therefore the knowledge is detectable only in the overall functioning of the network.c.is locally represented, allowing the network to draw inferences about partially viewed stimuli.d.is overshadowed by the parallel processing employed by the net.

Electrophysiological studies have investigated neural indices of such effects. Holcomb et al. (2002) showed that the N400, a marker of lexical and semantic processing usually observed between 350 and 500 ms (e.g., Kutas et al., 2006), was larger when targets had a high compared to low ND. This effect was found in both a lexical decision (for both words and non-words) and semantic categorization task, which suggests that similar mechanisms are at work in the two tasks, and was recently replicated by Müller et al. (2010) and Laszlo and Federmeier (2011). The larger N400 to targets with high, compared to low, ND has been interpreted in terms of increased lexico-semantic activation of, and competition among, neighbors, according to Holcomb and colleagues, and increased semantic activation of neighbors according to Laszlo and Federmeier (2011). Because ND effects in the N400 time window have been found for both words and pseudowords, Laszlo and Federmeier have concluded that access to meaning is attempted regardless of the orthographic status of the target. According to the authors, these data therefore argue against staged models of word recognition (e.g., Forster, 1999) and support cascade models (e.g., Harm and Seidenberg, 2004).Both behavioral and electrophysiological studies have shown that ND effects can also be observed cross-linguistically. For example, in van Heuven et al.’s (1998) first experiment, proficient Dutch-English bilinguals performed a progressive demasking task on both Dutch (L1) and English (L2) words. Identification speed in both languages was negatively influenced by the number of orthographic neighbors in the other language (i.e., the higher the ND, the longer the RT). In Experiment 4, a different group of proficient Dutch-English bilinguals performed a lexical decision task on English (L2) words. Again, RTs were longer for English words that had a high number of neighbors in Dutch (L1). These and other data (e.g., Alternberg and Cairns, 1983; Frenck-Mestre, 1993; Bijeljac-Babic et al., 1997) suggest that orthographic representations for the first and the second languages might be organized together in highly proficient bilinguals and trigger a complex series of activation and inhibition processes among words belonging to different languages (Dijkstra and Van Heuven, 2002).The N400 modulation by ND has also been observed cross-linguistically (Midgley et al., 2008). In a categorization experiment, late French-English bilinguals, all proficient in L2 (English), were asked to perform a go/no-go task and press a button when an animal name was presented on the screen. Participants were presented with two separate lists (French and English words) whose order was counterbalanced across subjects. Cross-language (CL) ND was manipulated in the following way: 50% of the French words had a high number of neighbors in English and 50% had a low CLND. Similarly, 50% of the English words had a high number of neighbors in French and 50% had a low CLND. In general, event-related potentials (ERPs) were more negative for targets with high, compared to low, CLND. However, the pattern of effects depended on the target language. The N400 (300–500 ms) effect peaked later and was less widely distributed for L1 than L2 targets. Furthermore, early effects (P2/N2, 175–275 ms) were present only for L2 targets. These effects were absent in a group of monolingual English speakers.Midgley et al. (2008) interpreted the difference in CLND effects between the two languages in terms of frequency of exposure: the participants were more proficient in French, French being their first language; therefore, the connection strength between lexical representations was stronger for L1 than L2. As a consequence, French neighbors were more easily activated by English targets than English neighbors by French targets. A similar interpretation was proposed to explain the presence of early effects (P2/N2) for L2 targets (which were present in Holcomb et al., 2002, but only in the categorization task). According to the authors, differences in frequency between the targets and their neighbors in the two studies would explain the discrepancy in results. In Holcomb et al. (2002), both target and neighboring words had a high subjective written frequency, whereas, in Midgley et al. (2008), L2 targets had a lower subjective frequency than their L1 neighbors. Therefore, in the second study, the activation and competition from high-frequency neighbors would have started earlier.

The finding in the last paragraph regarding the retrieval of related words supports:A.spreading activation.B.depth of processing.C.the serial position effect.D.the existence of visuospatia

Which of the following is NOT a component of AI?Answer areaMachine LearningNeural NetworksBlockchainNatural Language Processing

The _____ is the universe of network-accessible information, an embodiment of human knowledge.

Characteristics of knowledge