Universal Features in Phonological Neighbor Networks

Abstract

Human speech perception involves transforming a countinuous acoustic signal into discrete linguistically meaningful units (phonemes) while simultaneously causing a listener to activate words that are similar to the spoken utterance and to each other. The Neighborhood Activation Model posits that phonological neighbors (two forms [words] that differ by one phoneme) compete significantly for recognition as a spoken word is heard. This definition of phonological similarity can be extended to an entire corpus of forms to produce a phonological neighbor network (PNN). We study PNNs for five languages: English, Spanish, French, Dutch, and German. Consistent with previous work, we find that the PNNs share a consistent set of topological features. Using an approach that generates random lexicons with increasing levels of phonological realism, we show that even random forms with minimal relationship to any real language, combined with only the empirical distribution of language-specific phonological form lengths, are sufficient to produce the topological properties observed in the real language PNNs. The resulting pseudo-PNNs are insensitive to the level of lingustic realism in the random lexicons but quite sensitive to the shape of the form length distribution. We therefore conclude that "universal" features seen across multiple languages are really string universals, not language universals, and arise primarily due to limitations in the kinds of networks generated by the one-step neighbor definition. Taken together, our results indicate that caution is warranted when linking the dynamics of human spoken word recognition to the topological properties of PNNs, and that the investigation of alternative similarity metrics for phonological forms should be a priority.

Lexical Networks in !Xung

“…This disparity between languages is caused by the large phonemic inventory, which creates fewer minimal pair matches when randomly sorted, as in the uniform, Zipfian, scrambled, and unigram pseudolexicons. !Xung thus serves as a counterexample to recent claims that simplistic random lexicons created with unigram sampling can mimic the properties of real LNs (Brown et al, 2018). The disparity begins to shrink as the pseudolexicons become more natural, suggesting that disparities due to the large phonemic inventory are reduced by phonological structure and that phonotactic constraints on word forms in !Xung lead the lexicon to include more minimal pairs.…”

“…Vitevitch (2008) argues that the high connectivity and tendency toward clustering found in the English language lexicon are important aids to word learning and retrieval; later work finds similar properties in other lexicons (Arbesman et al, 2010;Shoemark et al, 2016). Some claims about the linguistic relevance of LNPs have been qualified by experiments showing that certain property values are inherent to the construction process of the network and can be replicated even when words are sampled from simple generative processes (Stella and Brede, 2015;Gruenenfelder and Pisoni, 2009;Turnbull and Peperkamp, 2016;Brown et al, 2018), though all these studies except Brown et al point out that the LNs of natural languages maintain some distinctive properties.…”

Database of word-level statistics for Mandarin Chinese (DoWLS-MAN)

“…As such, the word-level value of PND is the same as the commonly used network measure known as degree (i.e., the number of edges per node). Topological features extracted from phonological networks have been used to analyze both participant-level verbal productions (Neergaard, Luo, et al 2019), and whole-vocabularies (Arbesman et al 2010b, 2010aBrown et al 2018;Dautriche et al 2017;Shoemark et al 2016;Siew 2013;Siew and Vitevitch 2019;Stella et al 2018;Stella and Brede 2015;Turnbull and Peperkamp 2016;Vitevitch 2008). Meanwhile, word-level network values extracted from whole-vocabularies, have given insight into phonological processes through several network measures.…”

“…Analyses of phonological networks that go beyond the word-level, referred to as the network's topology, have been illustrative of the role that segmentation plays in speech processing. Topological features extracted from phonological networks have been used to analyze both participant-level verbal productions (Neergaard, Luo, et al 2019), and whole-vocabularies (Arbesman, Strogatz, and Vitevitch 2010b, 2010aBrown et al 2018;Dautriche et al 2017;Shoemark et al 2016;Siew 2013;Siew and Vitevitch 2019;Stella et al 2018;Stella and Brede 2015;Turnbull and Peperkamp 2016;Vitevitch 2008). constructed the sixteen Mandarin phonological networks seen in Table 1.…”

What Do Cognitive Networks Do? Simulations of Spoken Word Recognition Using the Cognitive Network Science Approach

“…Despite the success of the cognitive network approach in accounting for certain aspects of spoken word recognition (and other language-related and memory processes) this approach has been criticized because “…these networks do not ‘do’ anything; they have no function” [ 26 ] (pg. 16).…”