We develop the view that inflection is driven partly by non-phonological analogy and that non-phonological information is of particular importance to the inflection of non-canonical roots, which in the view of [Marcus, G. F., Brinkmann, U., Clahsen, H., Wiese, R., & Pinker, S. (1995). German inflection: the exception that proves the rule. Cognitive Psychology, 29, 189-256.] are inflected by a symbolic rule process. We used the Dutch plural to evaluate these claims. An analysis of corpus data shows that a model using non-phonological information (orthography) produces significantly fewer errors on plurals of non-canonical Dutch nouns, in particular borrowings, than a model that includes only phonological information. Moreover, we show that a double default system, as proposed by Pinker [Pinker, S. (1999). Words and rules. London: Phoenix.], does not offer an advantage over the latter model. A second study, examining the use of orthography in an online plural production task, shows that, in Dutch, the chosen pseudoword plural is significantly affected by non-phonological information. A final simulation study confirms that these results are in line with a model of inflectional morphology that explains the inflection of non-canonical roots by non-phonological analogy instead of by a default rule process.

中文翻译：

---

荷兰语复数屈折：证明类比的例外☆

我们认为屈折部分是由非语音类比驱动的，并且非语音信息对非规范词根的屈折特别重要，在 [Marcus, GF, Brinkmann, U., Clahsen, H.、Wiese, R. 和 Pinker, S. (1995)。德语屈折：证明规则的例外。Cognitive Psychology, 29, 189-256.] 受到象征性规则过程的影响。我们使用荷兰语复数来评估这些声明。对语料库数据的分析表明，与仅包含语音信息的模型相比，使用非语音信息（正字法）的模型在非规范荷兰名词的复数形式（尤其是借用词）上产生的错误要少得多。此外，我们展示了平克 [Pinker, S. (1999) 提出的双重违约制度。词和规则。伦敦：凤凰城。]，不提供优于后一种模型的优势。第二项研究检查了在线复数生产任务中正字法的使用，表明在荷兰语中，选择的伪词复数受到非语音信息的显着影响。最终的模拟研究证实，这些结果与屈折形态模型一致，该模型通过非语音类比而不是默认规则过程来解释非规范词根的屈折。