Retrieving words from the mental lexicon is generally conceptualized as a selection process, in which a central “lexical selection” mechanism adjudicates, through some rule or function or stochastic process, which of the many activated words will be prioritized for further processing. By “central” is meant that the selection mechanism functionally operates as a bottleneck, in that representations must be processed (in some way) by that mechanism in order to be prioritized for further processing. Historically, the different mechanisms of selection that have been proposed have emphasized different types of empirical evidence. The hypothesis of “selection by competition” emphasized observations of slower naming latencies and higher error rates when pictures are named in the context of within semantic-category coordinates, compared to semantically unrelated contexts (for discussion, see Damian & Bowers, 2003; La Heij, 1988; Schnur, Schwartz, Brecher, & Hodgson, 2006). The argument from semantic interference to lexical competition is that contexts that increase the activation of non-target words lead to a longer and noisier comparison process at selection, which translates to longer production times and/or lower accuracy. Noncompetitive models of lexical selection were initially developed to explain the types and distributions of errors observed in spontaneous speech and in patients with acquired brain injuries (e.g., Dell, 1986). More recently, there has been general recognition that speech production is not monolithically slowed by a semantic context, as there are a number of situations in which a related semantic context facilitates picturenaming latencies (Mahon, Costa, Peterson, Vargas, & Caramazza, 2007; Navarrete, Del Prato, Peressotti, & Mahon, 2014). Such semantic facilitation effects are in line with noncompetitive models and are contrary to the core expectation of selection by competition models. Productive debate and discussion ensued. In short, the onus is on competitive models to explain semantic facilitation effects, while the onus is on noncompetitive models to explain semantic interference effects. Proponents of selection by competition proposed additional mechanisms to account for observations of semantic facilitation. For instance, Abdel Rahman and Melinger (2009) argued that whether semantic interference or facilitation is observed depends on the balance of semantic-level-priming and lexicallevel-competition (for a critique, see Mahon & Caramazza, 2009). Roelofs and Piai (2013) argued that there is a threshold mechanism just prior to selection by competition that can override interference at the lexical level, thus leading to semantic facilitation (for a critique, see Mahon & Navarrete, 2014). Belke (2017) ascribed the semantic facilitation that is observed for the first presentation of pictures in the blocked-cyclic naming task to top-down influences and/or response strategies (for an alternative view, see Navarrete, Del Prato, & Mahon, 2012; Navarrete

中文翻译：

---

裁决言语产生中的冲突——我们需要一个中央选择机制吗？


从心理词典中检索单词通常被概念化为一个选择过程，其中中央“词汇选择”机制通过某种规则或功能或随机过程来裁决许多激活单词中的哪些将被优先考虑以进行进一步处理。 “中央”意味着选择机制在功能上作为瓶颈运行，因为表示必须由该机制（以某种方式）处理，以便确定优先级以进行进一步处理。从历史上看，所提出的不同选择机制都强调不同类型的经验证据。 “竞争选择”的假设强调，与语义不相关的上下文相比，当在语义类别坐标内的上下文中命名图片时，会观察到更慢的命名延迟和更高的错误率（有关讨论，请参见 Damian & Bowers，2003 年；La Heij ，1988 年；施努尔、施瓦茨、布雷彻和霍奇森，2006 年）。从语义干扰到词汇竞争的论点是，增加非目标词激活的上下文会导致选择时更长且更嘈杂的比较过程，这意味着更长的生产时间和/或更低的准确性。词汇选择的非竞争性模型最初是为了解释在自发言语和后天性脑损伤患者中观察到的错误的类型和分布而开发的（例如，Dell，1986）。最近，人们普遍认识到语音生成并不会因语义上下文而整体减慢，因为在许多情况下相关的语义上下文会促进图片命名延迟（Mahon、Costa、Peterson、Vargas 和 Caramazza，2007 年；纳瓦雷特、德尔普拉托、佩雷索蒂和马洪，2014）。 这种语义促进效应符合非竞争模型，与竞争模型选择的核心期望背道而驰。随后进行了富有成效的辩论和讨论。简而言之，竞争模型有责任解释语义促进效应，而非竞争模型有责任解释语义干扰效应。竞争选择的支持者提出了额外的机制来解释语义促进的观察结果。例如，Abdel Rahman 和 Melinger（2009）认为，是否观察到语义干扰或促进取决于语义层面启动和词汇层面竞争的平衡（有关批评，请参见 Mahon 和 Caramazza，2009）。 Roelofs 和 Piai (2013) 认为，在竞争选择之前存在一个阈值机制，可以克服词汇层面的干扰，从而导致语义便利化（批评参见 Mahon & Navarrete，2014）。Belke (2017) 将在块循环命名任务中首次呈现图片时观察到的语义促进归因于自上而下的影响和/或响应策略（另一种观点，参见 Navarrete, Del Prato, & Mahon, 2012纳瓦雷特