The received wisdom is that truth-conditional semantics presupposes and entails certain externalist ontological commitments. The idea is simple enough: if a semantic theory specifies (compositional) truth conditions, then, for any given sentence that is true, the theory will specify the very conditions that hold such that the sentence is true. Presumably, such conditions are exactly ways the world might be that make our truths true and our falsehoods false. Semantics is world-involving merely by dint of being truth-involving. A general worry for this picture, no matter how otherwise tempting it might be, is the occurrence of referentially defective words and expressions, ones that appear properly to contribute to the truth conditions of sentences that may be variously true or false but which do not appear to have referents we should otherwise sanction. Chomsky and others have claimed that, far from being outliers, a class of peculiar cases, referentially defective expressions are the absolute norm, utterly ubiquitous. The focus of the following will be the apparently referentially defective expression the average American. I shall first spell out the problem it poses and upon that basis suggest a general two-tier approach to natural language semantics, under which the interpretation of syntax (the first tier) is not world-involving, but another level of interpretation is; this second tier of interpretation, however, is not specifically linguistic. Secondly, the most thorough analysis of ‘the average+N’ kind of construction will be presented due to Kennedy and Stanley (2009), whose aim is to defend the orthodoxy; that is, such an expression may be given a clear semantic analysis that renders it non-defective. Thirdly, a number of severe problems for the analysis will be presented. Fourthly, an alternative account of the peculiar properties of the construction will be offered in line with the two-tier proposal previously suggested. The account preserves the defectiveness of the average American, but renders it unproblematic from a theoretical perspective.

中文翻译：

---

普通美国人的语义学和本体论

公认的真理是条件真语义以某些外在的本体论承诺为前提并具有一定的外在主义本体论承诺。这个想法很简单：如果一个语义理论指定了（组成的）真实条件，那么对于任何给定的真实句子，该理论都会规定保持该句子真实的条件。大概，这些条件正是世界可能成为使我们的真理真实，我们的虚假错误的方式。语义学仅是因为涉入真理而涉入世界。不管这幅画有多诱人，这张图普遍担心的是出现了具有参考意义的词语和表达方式，似乎正确地助长了句子真实性条件的句子，这些句子可能是正确的，也可能是错误的，但似乎没有指称对象，否则我们应予以制裁。乔姆斯基和其他人声称，指称有缺陷的表达方式不是绝对值，而是绝对无处不在的绝对规范，而不是离群值。接下来的重点将是普通美国人明显具有指称性的表达方式。首先，我将阐明它所带来的问题，并在此基础上提出一种对自然语言语义学的通用两层方法，在这种方法下，语法（第一层）的解释不涉及世界，而另一层次的解释是：但是，第二层解释不是专门的语言学。其次，肯尼迪和斯坦利（Kennedy and Stanley，2009）提出了对“平均值+ N”类型的结构的最彻底的分析，其目的是捍卫正统观念。也就是说，可以对此类表达式进行清晰的语义分析，使其无缺陷。第三，将提出许多严重的分析问题。第四，将根据先前建议的两层建议提供对建筑特殊属性的替代说明。该帐目保留了普通美国人的缺陷，但从理论上讲使其毫无问题。将提出许多严重的分析问题。第四，将根据先前建议的两层建议提供对建筑特殊属性的替代说明。该帐目保留了普通美国人的缺陷，但从理论上讲使其毫无问题。将提出许多严重的分析问题。第四，将根据先前建议的两层建议提供对建筑特殊属性的替代说明。该帐目保留了普通美国人的缺陷，但从理论上讲使其毫无问题。