符号认知的阐释

注：机翻，未校。

---

What is Symbolic Cognition?

Ronald J. Planer

© Springer Nature B.V. 2019

Abstract 摘要

Humans’ capacity for so-called symbolic cognition is often invoked by evolutionary theorists, and in particular archaeolo- gists, when attempting to explain human cognitive and behavioral uniqueness. But what is meant by “symbolic cognition” is often left underspecified. In this article, I identify and discuss three different ways in which the notion of symbolic cognition might be construed, each of them quite distinct. Getting clear on the nature of symbolic cognition is a necessary first step in determining what symbolic cognition might plausibly explain.
进化论家，特别是考古学家，在试图解释人类认知和行为的独特性时，经常援引人类的所谓符号认知能力。但 “符号认知” 的含义往往被低估了。在本文中，我确定并讨论了三种不同的方式来解释符号认知的概念，每一种都非常不同。弄清楚符号认知的本质是确定符号认知可能合理解释什么的必要第一步。

Keywords Symbolic cognition · The evolution of language · Inner speech · Language of thought · Meaning conventions

1 Introduction 1 引言

The question, “What makes humans unique?” admits of many answers. Among terrestrial vertebrates, we humans are unique in the range of environments we can inhabit, and the rate at which we come to successfully colonize them. We are also unique in the degree to which we engage in social learning. While many other animals socially learn, none show the high-fidelity, high-bandwidth forms of cultural learning that we do. Yet another answer would be that we are unique in the extent to which we cooperate and collaborate, particularly with non-relatives. Each of these statements is true and theoretically important, and many others could be added. But another way of interpreting the above question is as a request for that trait—or that set of traits—which explains the distinctive features of our species, or some sub- set of those features. So, for example, it is quite plausible to see human ecological success as explained by our unique forms of cultural learning (Boyd 2017). In turn, one might see human culture as explained by, or at least as a special case of, humans’ unique tendency to cooperate (including, but not limited to, our tendency towards altruism) (Bowles and Gintis 2013).
“是什么让人类独一无二”这个问题承认许多答案。在陆生脊椎动物中，我们人类在我们可以居住的环境范围以及我们成功定居它们的速度方面是独一无二的。我们在参与社交学习的程度上也是独一无二的。虽然许多其他动物进行社交学习，但没有一种动物表现出我们所做的高保真、高带宽形式的文化学习。然而，另一个答案是，我们在合作和协作的程度上是独一无二的，特别是与非亲属合作。这些陈述中的每一个都是正确的，并且在理论上很重要，还可以添加许多其他陈述。但对上述问题的另一种解释是，对那个特征——或那组特征——的要求，它解释了我们物种的独特特征，或这些特征的某些子集。因此，例如，用我们独特的文化学习形式来解释人类生态的成功是相当合理的（Boyd 2017）。反过来，人们可能会将人类文化视为人类独特的合作倾向（包括但不限于我们的利他主义倾向）来解释，或者至少是人类文化的一个特例（Bowles 和 Gintis 2013）。

So, the question at the start of this article can be divided into two more specific ones:
所以，本文开头的问题可以分为两个更具体的问题：

(1) What are the unique things that humans do / can do?
(1) 人类做 / 能做的独特事情是什么?

(2) What is the explanation for (1)?
(2) (1) 的解释是什么?

Researchers answer these questions from within their own disciplinary frameworks. In archaeology, for example, the focus is naturally on those capacities that leave mate- rial traces in the archaeological record—distinctive forms of technology, trade, mobility, ritual, geographic range and so on. The term “behavioral modernity” is often used to label this set of distinguishing features, some of which exist at the individual level, some at the level of groups.[1] Exactly which traits figure in this list remains a matter of dispute, with some archaeologists arguing that traits widely regarded as distinctively human were in fact exemplified by other hominins, such as the Neanderthals. (Two examples of this include Mithen 2005 on Neanderthal musicality and Stiner 2017 on Neanderthal mourning and commemoration prac- tices.) But there has been and continues to be something of a consensus that Late Pleistocene humans exhibited at least some behavioral complexities that were, for one reason or another, off limits to other hominins, including our nearest relative the Neanderthals with whom ancient humans co-existed for a significant period of time.
研究人员在他们自己的学科框架内回答这些问题。例如，在考古学中，重点自然是那些在考古记录中留下材料痕迹的能力——技术、贸易、流动性、仪式、地理范围等的独特形式。“行为现代性”一词经常被用来标记这组显着特征，其中一些存在于个人层面，一些存在于群体层面。[1]这份名单中究竟有哪些特征仍然是一个争议问题，一些考古学家认为，被广泛认为具有独特人类特征的特征实际上由其他古人类（如尼安德特人）所体现。（这方面的两个例子包括 Mithen 2005 关于尼安德特人的音乐性和 Stiner 2017 关于尼安德特人哀悼和纪念实践。但是，已经并将继续存在某种共识，即晚更新世人类至少表现出一些行为复杂性，由于某种原因，这些复杂性是其他古人类所禁止的，包括我们最近的亲戚尼安德特人，古人类与尼安德特人共存了很长一段时间。

This article is interested in a particular kind of account of what explains human uniqueness. In human evolution studies, and especially in archaeology circles, a common [1] See Chase (2003) for a discussion of this notion, and other, related ways of understanding behavioral modernity.
本文对解释人类独特性的一种特殊类型的解释感兴趣。在人类进化研究中，特别是在考古学界，一个常见的 [1] 参见 Chase （2003） 关于这个概念的讨论，以及理解行为现代性的其他相关方法。

answer has been that humans uniquely possess the capac- ity for symbolic cognition or symbolic thought (see, e.g., Noble and Davidson 1991; Stringer and Gamble 1993; Wad- ley 2001; Henshilwood et al. 2003; Tattersall 2008; Soffer 2009; Nowell 2010, and many others). Despite the high- profile status of this answer, scant attention has been paid to what symbolic cognition actually is. The purpose of this article is to address, or at least begin to address, that lacuna. I begin by discussing Peirce’s historically influential account of what a symbol is. Then, after sharpening that account, the sections that follow distinguish three different ways in which the notion of symbolic cognition might be taken, each hav- ing a legitimate claim to the title. (As will become clear, one of these notions—the last to be discussed—is much more psychologically basic than the others.) These more specific versions of the notion differ in the kinds of archaeologi- cally-visible behaviors they might explain. However, while interesting, and while clearly very important to archaeology, questions about the explanatory connection between these types of symbolic cognition and behavioral modernity are sidelined in what follows. The focus here is on what sym- bolic cognition is, an issue that must be tackled before these explanatory issues can be properly addressed, and one that leaves us with more than enough work to do in the present article. Over and above analyzing a key theoretical notion in archaeology, then, the main payoff of this article is to pave the way for a proper examination of whether symbolic cog- nition might explain all that it has been invoked to explain. But the latter is a project that must be left for another day.[2]
答案是，人类独特地拥有象征认知或象征性思维的能力（参见 Noble 和 Davidson 1991 年;Stringer 和 Gamble 1993;Wad- ley 2001 年;Henshilwood 等人，2003 年;Tattersall 2008 年;Soffer 2009 年;Nowell 2010 和许多其他人）。尽管这个答案的地位很高，但很少有人关注符号认知到底是什么。本文的目的是解决或至少开始解决这一空白。我首先讨论了皮尔斯对符号是什么的历史影响。然后，在进一步完善这一解释之后，接下来的章节区分了象征认知概念的三种不同方式，每种方式都有对标题的合法要求。（将会很清楚，其中一个概念——最后一个要讨论的概念——在心理学上比其他概念要基本得多。这些更具体的概念版本在它们可能解释的考古学可见行为类型上有所不同。然而，尽管很有趣，而且显然对考古学非常重要，但关于这些类型的符号认知和行为现代性之间的解释性联系的问题在接下来的内容中被搁置一旁。这里的重点是什么是符号认知，在正确解决这些解释性问题之前，必须解决这一问题，而这个问题在这篇文章中给我们留下了足够多的工作要做。因此，除了分析考古学中的一个关键理论概念之外，本文的主要成果是为正确检查符号认知是否可以解释它所引用的所有内容铺平了道路。但后者是一个必须留到另一天的项目。[2]

I shall pursue this task more from a cognitive science background than has been typical in archaeology, though the cognitive science I draw upon, it should be said, is by and large of the traditional variety. That variety explains mental phenomenon primarily in terms of computation over internal representations. There have been movements away from this computational framework, with greater emphasis being placed on dynamic connections between brain, body, and world. Indeed, some theorists go so far as to want to eschew talk of “representation” entirely, or almost entirely. The issues at play are bound to look quite different from the perspective of such alternative frameworks, especially the more extreme versions of them. Unpacking the lessons for archaeology and related fields of these alternative frame- works is a theoretically interesting and important project, and hence it is encouraging to see work along those very lines (Renfrew 2004; Malafouris 2013). But it is not my project here.
我将更多地从认知科学背景来追求这项任务，而不是典型的考古学，尽管我应该说，我借鉴的认知科学基本上是传统种类的。这种多样性主要从对内部表征的计算来解释心理现象。已经有远离这个计算框架的运动，更加强调大脑、身体和世界之间的动态连接。事实上，一些理论家甚至想完全或几乎完全避免谈论“表征”。从这些替代框架的角度来看，这些问题必然会大不相同，尤其是它们更极端的版本。解读这些替代框架作品对考古学和相关领域的经验教训在理论上是一个有趣且重要的项目，因此看到沿着这些路线进行的工作是令人鼓舞的（Renfrew 2004;Malafouris 2013 年）。但这不是我的项目。

2 Symbols à la Peirce 2 皮尔斯符号

In getting clear on what symbolic cognition is or might be, a reasonable place to start is with the question, “What is a symbol?” For whatever symbolic cognition is, it must be linked to symbols in some way. One—perhaps the most— influential answer to this question comes from Charles S. Peirce’s General Theory of Signs (Peirce 1883).[3] Peirce’s theoretical treatment of signs changed over his lifetime, and his mature thoughts were, to put it mildly, complex.[4] For pre- sent purposes, we shall limit ourselves only to the rudiments of his theory, which still leaves plenty to discuss.
在弄清楚什么是符号认知或可能是什么时，一个合理的起点是这个问题，“什么是符号？因为无论符号认知是什么，它都必须以某种方式与符号相关联。对这个问题的一个——也许是最具影响力的——答案来自查尔斯·皮尔斯 （Charles S. Peirce） 的《符号通论》（Peirce 1883）。[3] 皮尔斯对符号的理论处理在他的一生中发生了变化，委婉地说，他的成熟思想很复杂。[4] 为了预先发送的目的，我们只限于他的理论的基本内容，这仍然有很多值得讨论的地方。

Perhaps the most well-known feature of Peirce’s theory is his three-way distinction between icons, indexes, and sym- bols. Icons and indexes are often characterized as follows: an icon resembles its referent, and an index coincides—in space and time—with its referent. And symbols tend to be defined in one of two ways, either negatively or positively: the negative definition says that a symbol is a sign that is neither iconic nor indexical, while the positive one says that a symbol is a sign whose reference is determined by conven- tion. The actual situation is a bit more complicated than this, however.
也许皮尔斯理论最著名的特点是他对图标、索引和符号的三向区分。图标和索引通常具有以下特征：图标类似于其所指对象，索引在空间和时间上与其所指对象重合。符号往往以两种方式之一进行定义，要么是消极的，要么是积极的：否定的定义说符号是既不是标志性的也不是索引的符号，而积极的定义说符号是其指称是由协商决定的符号。然而，实际情况比这要复杂一些。

On Peirce’s theory, one entity, s, is a sign for some other entity, o, in virtue of causing some receiver to think of o. Peirce’s categories of signs are then distinguished based on how they cause the receiver to think of the relevant thing. s is an icon for o if and only if s causes the receiver to think of o in virtue of some perceived resemblance between s and
根据皮尔斯的理论，一个实体 s 是另一个实体 o 的符号，因为它会使一些接收者想到 o。然后，根据皮尔斯的符号类别如何使接收者思考相关事物来区分它们。s 是 o 的图标，当且仅当 s 由于 s 和 之间的某种感知相似性而使接收者想到 o

o. This is obviously different from resemblance full stop. A sign might resemble some object however closely we want and yet fail to be an icon for that object; most obviously, if the receiver is not capable of detecting the resemblance in question. Hence, Peirce’s account of an icon makes essential reference to an interpretative process on the part of some receiver.[5] The same is true of indexes. s is an index for o if and only if s causes the receiver to think of o in virtue of some “natural connection” between s and o. The stock exam- ple here is smoke and fire: seeing smoke often makes us think of fire, and this association between the two is under- pinned by a natural regularity.
o.这显然与 similarsance 句号不同。一个标志可能与某个对象非常相似，但无法成为该对象的图标;最明显的是，如果接收者无法检测到所讨论的相似性。因此，皮尔斯对圣像的描述必须提到某些接收者的解释过程。[5] 索引也是如此。S 是 o 的索引，当且仅当 S 导致接收者由于 S 和 O 之间的某种“自然联系”而想到 O。这里的股票例子是烟和火：看到烟经常让我们想到火，而两者之间的这种联系是由自然规律为基础的。

What about symbols? Well, the negative definition tells us that s is a symbol if and only if s is a sign for some entity o, and yet s does not cause the receiver to think of o, either in virtue of a perceived resemblance between s and o, nor in virtue of some natural connection between s and o. How, then, does s cause the receiver to think of o? An intuitive idea is that the positive definition is an answer to this question—it’s in virtue of a convention, and indeed, solely in virtue of a convention. Hence, in order to really understand symbolhood, it would seem that we need a better grip on conventions, and specifically, meaning conventions.
符号呢？嗯，否定定义告诉我们，当且仅当 s 是某个实体 o 的符号时，s 是一个符号，但 s 不会使接收者想到 o，无论是由于 s 和 o 之间感知到的相似性，还是由于 s 和 o 之间的某种自然联系。那么，s 是如何使接收者想到 o 的呢？一个直观的想法是，肯定定义是这个问题的答案——它基于一种约定，事实上，完全基于一种约定。因此，为了真正理解象征性，我们似乎需要更好地把握约定俗成，具体来说，就是意义约定俗成。

Table 1 Possible sender and receiver strategies in a signaling game
表 1 信令博弈中可能的发送方和接收方策略

State	Sender	Receiver
$w_1$	$s_1$	$a_1$
$w_2$	$s_2$	$a_2$

State	Sender	Receiver
$w_1$	$s_2$	$a_1$
$w_2$	$s_1$	$a_2$

David Lewis’s Convention (1969) remains the most com- prehensive and systematic treatment of the nature of conventions including meaning conventions—by far. Lewis developed the notion of a signaling game in order to elucidate the idea of meaning by convention. The simplest version of such a game involves a single sender-receiver pair, two signs, two acts, and two states of the world. The sender observes the state and produces one of her two signs, s1 or s2. The receiver observes the sign and produces one of his two acts in response, a1 or a2. The result is that the state is paired with some act. Each state-act pair produces a certain payoff to the sender and receiver, and (in the simplest case) they agree on which act ought to be performed in each state (i.e., their preferences align). There are thus two combinations of sender and receiver strategies that are equally good (taking numerical agreement among states and acts to indicate an optimal match) (see Table [1] ).
大卫·刘易斯 （David Lewis） 的《公约》（1969 年）仍然是迄今为止对公约性质（包括意义公约）最全面和最系统的论述。刘易斯发展了信号游戏的概念，以阐明约定俗成的意义概念。这种游戏的最简单版本涉及单个发送者-接收器对、两个符号、两个行为和两种世界状态。发送方观察状态并生成她的两个符号之一，即 s1 或 s2。接收者观察这个符号并产生他的两个行为之一作为回应，a1 或 a2。结果是 state 与某个 act 配对。每个状态-行为对都会给发送者和接收者带来一定的回报，并且（在最简单的情况下）他们同意在每个状态下应该执行哪个行为（即，他们的偏好一致）。因此，发送方和接收方策略有两种同样好的组合（在状态之间取得数字一致性并采取行动以指示最佳匹配）（见表 [1] ）。

All of the “action” takes place in the second column of Table [1]. Under the first strategy profile, s1 is used to commu- nicate the presence of state w1, and s2 the presence of w2, while under the second, things are the other way around. It does not matter which of these two strategy profiles they col- lectively adopt, so long as their strategies mesh. So, if sender and receiver happen to use s1 to mean w1 and s2 to mean w2, then there is a clear sense in which this choice is arbitrary and hence a matter of convention. And to reach the further claim that the meaning of these signs is purely conventional, we need only assume that neither sign resembles either state, and that there was no prior association between either state and either sign (nothing like a smoke-fire relationship).
所有 “action” 都发生在表 [1] 的第二列中。在第一种策略配置文件下，s1 用于表示状态 w1 的存在，s2 用于表示 w2 的存在，而在第二种策略下，情况正好相反。他们共同采用这两种策略中的哪一种并不重要，只要他们的策略相吻合即可。因此，如果发送方和接收方碰巧使用 s1 表示 w1，使用 s2 表示 w2，那么很明显，这种选择是任意的，因此是一个约定俗成的问题。为了进一步声称这些符号的含义是纯粹约定俗成的，我们只需要假设这两个符号都不像任何一种状态，并且任何一种状态和任何一种符号之间都没有事先的联系（不像烟火关系）。

When Lewis formulated these games, he had in mind cog- nitively complex agents—rational agents with beliefs about the world, about each other’s beliefs, about each other’s beliefs about each other’s beliefs, etc., as well as psycho-logical preferences over the state-act pairs. Thus, for a long time, Lewis’s work was not seen as having much application,if any, outside the human domain. That changed as a result of modelling work carried out by Brian Skyrms over the last several decades.6 These models have shown how Lew- is’s assumptions can be relaxed, and hence how meaning conventions can originate and stabilize in the absence of such cognitive complexity, e.g., through basic reinforcement learning or even “brute” natural selection. Accordingly, the work of Skyrms and its off-shoots shows that human com- munication cannot be regarded as the sole province of mean- ing conventions or even pure meaning conventions. A vari- ety of animal alarm call systems, for example, are plausibly viewed as being governed by such conventions.
当刘易斯制定这些游戏时，他想到的是认知复杂的主体——理性主体，他们对世界、彼此的信念、彼此的信念、彼此的信念等，以及对状态-行为对的心理偏好。因此，在很长一段时间内，刘易斯的工作在人类领域之外没有太多应用，如果有的话。由于 Brian Skyrms 在过去几十年中进行的建模工作，这种情况发生了变化。6 这些模型展示了如何放宽 Lew- is 的假设，因此在没有这种认知复杂性的情况下，意义约定俗成如何产生和稳定，例如，通过基本的强化学习甚至“蛮力”自然选择。因此，Skyrms 及其分支的工作表明，人类的交流不能被视为 意义约定俗成甚至纯粹意义约定的唯一领域。例如，各种动物警报呼叫系统似乎被视为受此类惯例的约束。

It is not obvious what to say of such signals from a (neo-) Peircean standpoint. On the one hand, they are paradigm cases of signs that are used in spatio-temporal association with their referents; a Vervet alarm call for a leopard (a “bark”) says something like leopard here now. On the other hand, they are also conventional. The connection between the alarm call and the leopard is not the result of a natural regularity or law, but has rather been established via a con- tingent, flexible process of sender-receiver co-adaptation. Some other Vervet call—a “cough” or a “chutter”—might have been used in its place, and may one day take over for the bark. There is more than one way to go at this point, but at least for the purposes of this article, I propose the following. A distinction can be drawn among two types of indexes: non-conventional and conventional. Smoke is a non-conventional index for fire; the Vervet leopard alarm call is a conventional index for leopards. Some non-conven- tional indexes are inorganic products used by receivers in some way, while others are features (broadly construed) of other organisms. Generally speaking, the latter set of cases are what ethologists would refer to as “cues.” Here, receivers are adapted to make use of some feature of another organ- ism, but this feature has not been adapted for the receiver’s use—mosquitos use CO2 omissions to locate animals to feed on, but animals do not produce CO2 for that reason. Conven- tional signs, by contrast, will always be cases where there is both a sender and receiver present, and where production and interpretation of the sign has been co-adapted.
从（新）皮尔斯的角度来看，如何看待这些信号并不明显。一方面，它们是用于与其所指对象的时空关联的符号的范例;豹子的 Vervet 警报呼叫（“吠叫”）现在说的是 leopard 现在。另一方面，它们也是传统的。警报声和豹子之间的联系不是自然规律或规律的结果，而是通过一个偶然的、灵活的发送者-接收者共同适应过程建立起来的。其他一些 Vervet 叫声——“咳嗽”或“咯咯”——可能已经被用来代替它，有一天可能会取代树皮。在这一点上，有不止一条路可以走，但至少为了本文的目的，我建议如下。可以区分两种类型的指数：非常规指数和常规指数。烟雾是火的非常规指标;长尾蛇警报呼叫是豹的常规索引。一些非 conmediation 指标是接收者以某种方式使用的无机产物，而另一些则是其他生物体的特征（广义解释）。一般来说，后一组情况是行为学家所说的 “线索”。在这里，接收器适应了利用另一个器官的某些特征，但这一特征尚未适应接收器的使用——蚊子利用 CO2 的遗漏来定位要捕食的动物，但动物因此不会产生 CO2。相比之下，召集符号总是同时存在发送者和接收者在场的情况，并且符号的产生和解释是共同适应的。

Table 2 An expanded Peircean taxonomy
表 2 扩展的 Peircean 分类法

	Spatio-temporal dependence	Spatio-temporal independence
Non-conventional	Non-conventional index (e.g., smoke)	Icons (e.g., a portrait)
Conventional	Conventional index (e.g., Vervet leopard call)	Symbols (e.g., “dog”)

This enables us to retain the idea that alarm calls and other plausible cases of conventional animal signals are in fact indexes. However, it means that the category of sym- bols can no longer be defined simply as signs that mean by convention, for that is also true of conventional indexes. It is tempting at this point to define symbols as conventional signs that do not coincide in space and time with their refer- ent, but that is not quite right. We frequently use symbols in the presence of the things they refer to; we use “dog” to refer to a dog that is here now all the time. What we should instead say is that symbols are conventional signs whose meaning is not dependent on spatio-temporal coincidence with their referents.[7] It is this fact which enables symbols to routinely refer to the elsewhere and elsewhen, to exhibit so-called “displaced reference.” Exactly how this spatio- temporal independence is cognitively achieved is difficult to say, and it is unlikely that there is a single answer that holds across the board. But it is not necessary to answer this question for symbols to be marked out as a useable theoreti- cal category.
这使我们能够保留这样的想法，即警报呼叫和其他传统动物信号的合理情况实际上是索引。然而，这意味着符号的类别不能再简单地定义为按约定俗成的符号，因为传统指数也是如此。在这一点上，人们很容易将符号定义为在空间和时间上与其所指对象不一致的约定俗成的符号，但这并不完全正确。我们经常在它们所指的事物存在时使用符号;我们使用 “dog” 来指代现在一直在这里的狗。相反，我们应该说的是，符号是约定俗成的符号，其意义不取决于与所指对象的时空巧合。[7] 正是这一事实使符号能够例行地指代别处，而 elsewhen 则表现出所谓的“移位指涉”。这种时空独立性究竟是如何在认知上实现的，这很难说，而且不太可能有一个单一的答案可以全面成立。但是，要把符号标记为一个可用的理论类别，就没有必要回答这个问题。

These features of signs are summarized in Table [2]. Note that spatio-temporal independence is a property that symbols share with icons (icons can also refer to the elsewhere and elsewhen). What distinguishes symbols from icons is the conventionality of the former.
这些迹象特征总结在表 [2] 中。请注意，时空独立性是符号与图标共享的属性（图标也可以引用 elsewhere 和 elsewhen）。符号与图标的区别在于前者的约定俗成。

With this account of symbolhood in place, let us now turn to the question of how the notion of symbolic cognition might be understood. In the sections that follow, I present three construals of the notion. The corresponding types of cognition are all linked to symbols in some close-knit way, but differently in each case.
有了对象征性的解释，现在让我们转向如何理解象征认知的概念的问题。在接下来的部分中，我将介绍该概念的三种解释。相应的认知类型都以某种紧密的方式与符号相关联，但在每种情况下都不同。

3 Talking to Others 3 与他人交谈

One simple—perhaps the most simple—thing we might have in mind by “symbolic cognition” is just the ability to learn to use symbols. The term would then pick out the cognition underpinning this learning and the competence to produce and understand the symbols to which such learning gives rise. Now some doubtlessly think that there is a fundamental architectural difference separating minds that can learn to use symbols and those that can’t, but that is far from obvi- ous. It seems just as likely that this ability comes in many degrees, with humans simply being much better at using symbols than any other species. That seems especially true when we shift our focus from the level of individual sym- bols to that of powerful, rule-governed systems of symbols. Indeed, mastery of a sign system that was not fully sym- bolic—say, because it made heavy use of iconicity—but which featured other complex attributes such as productivity and systematicity, would in many ways be more impressive than mastery of an expressively limited and haphazardly organized set of symbols.[8 ]And it also seems that the devel- opment of a communication system of the former type would transform cognition and behavior to a much greater extent than the development of a system of the latter type. What is clear, however, is that a system bringing together both sets of features would be more impressive still, and could be expected to have a significantly more dramatic impact on users’ lives.
我们可能想到的“符号认知”中的一个简单——也许是最简单的——就是学习使用符号的能力。然后，该术语将挑选出支撑这种学习的认知以及产生和理解这种学习产生的符号的能力。现在，毫无疑问，有些人认为，存在着根本的架构差异，将能够学习使用符号的思想与不能学习使用符号的思想区分开来，但这远非显而易见。这种能力似乎同样可能来自许多程度，人类只是比任何其他物种都更擅长使用符号。当我们把注意力从单个符号的层面转移到强大的、受规则支配的符号系统层面时，这似乎尤其正确。事实上，掌握一个不完全是对称性的——比如说，因为它大量使用象征性——但具有其他复杂属性，如生产力和系统性——的符号系统，在许多方面比掌握一组表现力有限且组织随意的符号更令人印象深刻。[8 ]而且，与后一种类型的系统的发展相比，前一种通信系统的发展似乎会更大程度地改变认知和行为。然而，很明显，将这两组功能结合在一起的系统将更加令人印象深刻，并且可以预期对用户的生活产生更显着的影响。

To imagine such a system is, of course, to imagine human language, or something very much like it. And hence the first notion of symbolic cognition that I think is worth sepa- rating out and giving a place of prominence to is this: the cognition involved in learning, producing, and understand- ing a language, or a language-like communication system. As is immediately clear, this notion of symbolic cognition remains neutral on the question of what form such cogni- tion takes. It is a functional conception, defining symbolic cognition in terms of its effects rather than in terms of any more specific capacities, powers, or mechanisms. Having such an abstract characterization is likely to prove useful. Most obviously, it enables us to unite into a single category all those theorists who think symbolic cognition was the key difference maker behind human behavioral modernity, but who otherwise disagree about its origins (Is it innate? Acquired? Scaffolded?) or its realization in the brain (Are we computers? Ultra-powerful associationist nets?). There is a unity in this disparity, and all of these theorists might be right (or wrong) in their primary claim. This should help to bring order to the literature on symbolic cognition. But it is also the least interesting way of interpreting symbolic cognition, amounting to little or no more than the capacity to acquire and use language or language-like communication systems. One often gets the sense from the literature that it is something over and above this “basic” cognitive capac- ity that is at stake in important debates in archaeology and cognitive science.
当然，想象这样一个系统就是想象人类语言，或者非常类似的语言。因此，我认为值得分离并给予突出位置的符号认知的第一个概念是：涉及学习、生产和理解一种语言或类似语言的交流系统的认知。显而易见，这种符号认知的概念在这种认知采取何种形式的问题上仍然是中立的。它是一个功能概念，根据其效果而不是任何更具体的能力、力量或机制来定义符号认知。拥有这样一个抽象的描述可能会被证明是有用的。最明显的是，它使我们能够将所有那些认为符号认知是人类行为现代性背后的关键差异制造者，但对其起源持不同意见的理论家统一为一类（它是天生的吗？后天的？脚手架？或其在大脑中的实现（我们是计算机吗？超强大的关联主义网络？这种差异是统一的，所有这些理论家的主要主张可能是对的（或错的）。这应该有助于为符号认知的文献带来秩序。但它也是解释符号认知的最无趣的方式，它只相当于获取和使用语言或类似语言的交流系统的能力。人们经常从文献中感觉到，在考古学和认知科学的重要辩论中，正是超越了这种 “基本 ”认知能力的东西。

4 Talking to Ourselves 4 自言自语

A more interesting way of understanding “symbolic cogni- tion” is this: it is our conscious experience of internally-pro- duced, natural language sentences. Such cognition is sym- bolic simply because natural language is symbolic. I suspect that many non-cognitive scientists have something like this notion in mind when they talk of “symbolic cognition.”
理解“符号认知”的一种更有趣的方式是：它是我们对内部产生的自然语言句子的有意识体验。这种认知是对称的，仅仅因为自然语言是象征性的。我怀疑许多非认知科学家在谈论“符号认知”时，脑海中都有类似这样的想法。

This phenomenon has come to be known as “inner speech.” This term is less than ideal, as it suggests that the relevant phenomenon is tied essentially to the auditory modality. It’s not: congenitally deaf individuals experience inner signing. In addition, it would appear that inner sign has a highly similar cognitive profile to inner speech in hearing individuals (e.g., in terms of its frequency, its grammati- cal properties, etc.).[9] Having drawn attention to this, I will follow standard usage and refer to both modes of silently communicating with oneself as inner speech.
这种现象被称为 “内心言语”。这个术语不太理想，因为它表明相关现象本质上与听觉模式有关。事实并非如此：先天性聋人体验到内在手语。此外，在健听个体中，内在符号似乎与内在言语具有高度相似的认知特征（例如，在频率、语法特性等方面）。[9] 在提请注意这一点之后，我将遵循标准用法，并将两种与自己默默交流的方式称为内心对话。

As virtually everyone who works on inner speech points out, it has received strikingly little theoretical attention, given inner speech’s frequency in our day-to-day lives and the great importance we tend to attach to our inner voices. Fortunately, considerably more work is now underway, with theorists focusing on the nuances of inner speech’s phenom- enology, its (possible) cognitive functions, and its potential evolutionary origins.[10] In this section, the aim is primarily to review some of these ideas, rather than to advance novel claims or arguments.
正如几乎所有研究内在言语的人所指出的那样，考虑到内在言语在我们日常生活中的频率以及我们往往非常重视我们的内在声音，它在理论上受到的关注非常少。幸运的是，现在有更多的工作正在进行中，理论家们专注于内心言语现象学的细微差别、其（可能的）认知功能及其潜在的进化起源。[10] 在本节中，目的主要是回顾其中一些想法，而不是提出新的主张或论点。

An early and important exception to this pattern of neglect can be found in the work of Lev Vygotsky and his disciples.[11] One of Vygotsky’s main interests was in the role of language in organizing and controlling action. He began by drawing attention to the pedestrian fact that we instruct children on how to complete various action sequences, for example, how to tie one’s shoes. He then pointed out that children often rehearse the same or similar instructions aloud to themselves as they carry out the same behaviors. In the Vygotskyean tradition, such verbal behavior is known as private speech. Vygotsky rightly claimed that these instruc- tions might play essentially the same role in organizing the flow of the child’s behavior. The final step, in his view, was for private speech to “go internal,” to be rehearsed silently to oneself, thereby giving rise to inner speech.
这种忽视模式的一个早期和重要的例外可以在列夫·维果茨基和他的弟子的工作中找到。[11]维果茨基的主要兴趣之一是语言在组织和控制行动中的作用。他首先提请注意一个行人的事实，即我们指导孩子们如何完成各种动作序列，例如，如何系鞋带。然后他指出，孩子们在执行相同的行为时，经常会大声地对自己排练相同或类似的指令。在维果茨基传统中，这种言语行为被称为私人言语。维果茨基正确地声称，这些指令在组织儿童行为的流程方面可能起着基本相同的作用。在他看来，最后一步是让私人言语 “进入内心”，对自己默默排练，从而产生内心言语。

Vygotsky seemed to regard inner speech as the basis of thought and thinking, as do many other theorists interested in language and cognition. Much, if not all, of what we pre- theoretically regard as reasoning, decision-making, plan- ning, and so on consists in talking to ourselves. I’m unsure whether I will have time to go to the party on Friday night; I engage in some inner dialogue: “Well, if I leave the party by 11 pm, I can be in bed by 12 am. Then I can get up early and a squeeze in a few extra hours of work before midday.” So I go to the party. Can it really be doubted that this inner dialogue was the basis for my behavior?
维果茨基似乎将内心语言视为思想和思考的基础，许多其他对语言和认知感兴趣的理论家也是如此。我们先验地认为的推理、决策、计划等，如果不是全部的话，很大一部分是与自己交谈。我不确定我是否有时间去参加周五晚上的派对;我进行一些内心对话：“好吧，如果我在晚上 11 点之前离开派对，我可以在凌晨 12 点之前上床睡觉。然后我就可以早点起床，在中午之前多工作几个小时。所以我去参加派对。真的可以怀疑这种内心对话是我行为的基础吗？

It can, though one would certainly be forgiven for failing to see how. The orthodox view in cognitive science is that most of our thinking takes place “behind closed doors”: the mind is composed of a large and varied group of cogni- tive mechanisms,[12] where the internal operations of most of these mechanisms are inaccessible to introspection. We are consciously aware of only (some of) the final products. These thoughts and thought processes are not couched in natural language terms at all, but rather some innately pos- sessed “language of thought” (Fodor 1975), or multiple such languages, produced and understood by different parts of the brain. On this view, to token a thought is to token a mental expression—a representation couched in “mentalese”— where the content of the whole is determined by the con- tent of the expression’s parts, together with their mode of composition. And to engage in a process of thinking is to token a sequence of thoughts, where there is some inter- esting semantic relation that these thoughts stand into one another (e.g., the content of thoughts t1 and t2 might logi- cally entail the content of t3, or increase the probability of t3). This picture might need abandoning, but it continues to enjoy a place of prominence in both philosophy and cogni- tive science—and for good reasons, this author would argue. One reason is that it is extremely difficult to make sense of the cognition of non-linguistic animals in any other way.
它可以，尽管人们没有看到如何做肯定是可以原谅的。认知科学的正统观点是，我们的大部分思考都是在“闭门造车”进行的：心智由一大群不同的认知机制组成，[12]其中大多数这些机制的内部运作是内省无法进入的。我们有意识地只了解（部分）最终产品。这些思想和思维过程根本不是用自然语言来表达的，而是一些天生就拥有的“思想语言”（Fodor 1975），或多种这样的语言，由大脑的不同部分产生和理解。根据这种观点，象征一个思想就是象征一种心理表达——一种用 “mentalese” 表达的表征——其中整体的内容是由表达的各个部分的内容以及它们的组成方式决定的。而参与思考过程就是标记一系列思想，其中这些思想之间存在一些相互交织的语义关系（例如，思想 t1 和 t2 的内容可能在逻辑上包含 t3 的内容，或增加 t3 的可能性）。这幅图景可能需要放弃，但它在哲学和认知科学中仍然享有突出的地位——笔者认为，这是有充分理由的。一个原因是，以任何其他方式来理解非语言动物的认知是极其困难的。

One obvious thing inner speech does is enable us to men- tally rehearse sentences in advance of their delivery, such as we might do when practicing an important speech. But what might its function be beyond this? How one answers this question will depend on one’s views about the mind’s cognitive architecture. Suppose one thinks that the innate structure of the mind largely consists in powerful associative learning mechanisms. Then one might think inner speech is how the mind comes to be a genuinely representational system in the first place.[13] Inner speech would make possible a range of fundamentally new forms of cognition, including rational or reason-respecting cognition. But suppose, on the other hand, one subscribes to some version of the traditional account of our cognitive architecture outlined in the last paragraph. Then the question of what the function of inner speech is or might be becomes more complex. As noted in the introduction, the approach taken in this article is to bring a mainstream cognitive science perspective to bear on the topic of symbolic cognition, hence it is the second of these scenarios on which I will focus here.[14 ] In my view, there are two suggestions for how inner speech might function within such an architecture that stand out in terms of their plau- sibility and consequentialness. First, it has been suggested that inner speech might play a role in making information available to certain cognitive mechanisms that would other- wise lack access to that information (Fields 2002; Carruthers 2006). Second, it has been suggested that inner speech can improve our concentration, thereby enhancing our ability to reason, plan, and make decisions, by keeping certain pri- orities high on the mind’s “to-do” list (Fields 2002). Let us explore each idea in some more detail, respectively.
内心言语所做的一件明显的事情是使我们能够在句子发表之前定期排练句子，就像我们在练习重要演讲时所做的那样。但除此之外，它的功能可能是什么呢？一个人如何回答这个问题将取决于一个人对心智认知结构的看法。假设一个人认为心智的先天结构主要由强大的联想学习机制组成。那么，人们可能会认为内在言语是心灵首先成为一个真正的表征系统的方式。[13] 内在言语将使一系列全新的认知形式成为可能，包括理性或尊重理性的认知。但是，另一方面，假设一个人赞同上一段中概述的关于我们认知架构的传统解释的某个版本。那么，内心言语的功能是什么或可能是什么的问题就变得更加复杂了。正如引言中所指出的，本文采用的方法是将主流认知科学的观点引入符号认知的话题，因此这是我在这里要关注的第二种情况。[14 ] 在我看来，关于内心语言在这样的架构中如何运作，有两个建议在合理性和后果性方面很突出。首先，有人提出，内心言语可能在为某些认知机制提供信息方面发挥作用，而这些认知机制则无法获得这些信息（Fields 2002;Carruthers 2006 年）。其次，有人认为，内心语言可以提高我们的注意力，从而增强我们推理、计划和做出决定的能力，方法是将某些原则保持在大脑的“待办事项”清单上（Fields 2002）。让我们分别更详细地探讨每个想法。

On Bernard Baars’ (1997, 2005) famous “global work- space” model, there is a central location in the brain where most if not all of the brain’s cognitive mechanisms write to and read from—an “informational commons” if you will. But the real situation might be much more complicated than that. There are many possibilities. Perhaps each mechanism is capable of accessing information stored in such a work- space at least some of the time, but access is like a revolving door, with only some subset of the mechanisms enjoying access at a given time depending on the cognitive task(s) being pursued. Alternatively, our architecture might be quite disjointed, with subsets of mechanisms communicating with other subsets, one-way communication channels, etc. In any case, the idea behind the first of the above suggestions is that inner speech might function to get information to mecha- nisms that would not (currently) enjoy access to it otherwise by re-presenting that information to the mind via sensory (specifically auditory) channels. That information would eventually filter back through the whole system (undergoing various transformations and translations along the way)— just as it would if it were being picked up for the first time via inter-personal communication, say. (See Fig. [1].)
在 Bernard Baars （1997， 2005） 著名的“全球工作空间”模型中，大脑中有一个中心位置，如果不是全部的话，大脑的大部分认知机制都在这里写入和读取——如果你愿意的话，这是一个“信息共享”。但实际情况可能比这复杂得多。有很多可能性。也许每种机制都能够至少在某些时候访问存储在这种工作空间中的信息，但访问就像一扇旋转门，根据所追求的认知任务，只有部分机制子集在给定时间享有访问权。或者，我们的架构可能非常脱节，机制子集与其他子集通信、单向通信通道等。无论如何，上述第一个建议背后的想法是，内在言语可能起到将信息传递给（当前）无法访问信息的机制的作用，否则它们将通过感觉（特别是听觉）渠道将信息重新呈现给大脑。这些信息最终会在整个系统中过滤回来（在此过程中经历各种转换和翻译）——就像它第一次通过人际交流被拾取一样。（见图[1].)

Fig. 1 A simple feed-forward information-processing network sup- plemented by an inner speech loop. Absent this loop, information out- put by each column of mechanisms would no longer be accessible to those mechanisms
图 1 由内部语音循环支持的简单前馈信息处理网络。如果没有这个循环，这些机制将无法再访问每列机制输出的信息

In his Prehistory of the Mind (1996), Steven Mithen defended a proposal like this, combined with the idea that prior to language (and hence prior to inner speech), the mind was massively fragmented; it consisted in a large-array of mental “modules” with limited means of exchanging infor- mation with one another. A highly similar view was later defended in more detail by Peter Carruthers (2006). For both of these authors, modern thinking and behavior are seen as flowing from the integration of the mind’s different mecha- nisms, something made possible by the availability of a new representational medium—inner speech—in the brain.
在他的《心灵史前史》（1996 年）中，史蒂文·米森 （Steven Mithen） 为这样的提议辩护，并结合了这样一种观点，即在语言之前（因此在内心语言之前），心灵是大规模分裂的;它由大量的心理“模块”组成，彼此之间交换信息的方式有限。Peter Carruthers （2006） 后来更详细地捍卫了高度相似的观点。对于这两位作者来说，现代思维和行为都被视为源于心灵不同机械的整合，这是由于大脑中一种新的表征媒介——内心语言——的可用性而成为可能的。

There are other ways of unpacking the main idea here which do not take such a strong stand on the mind’s archi- tecture, past or present. To give just one example: Shanon (1998) points out that inner speech enables us to simulate conversations of the sort we might have while collectively trying to problem-solve or make a decision with peers. Put differently, it allows us to adopt multiple perspectives on some content; we first serve as sender and then as receiver. If the set of cognitive mechanisms we use to generate com- municative content is somewhat different than the set we use to evaluate incoming content (and it would appear that is the case: see Mercier and Sperber 2017), it easy to imagine how switching between sender and receiver roles might enhance reasoning. One reason it would make sense for there to be such a difference in cognitive mechanisms is that often we are on the receiving end of speech from agents whose hon- esty cannot be taken for granted; hence, the need to more thoroughly vet it.
还有其他方法可以解读这里的主要思想，这些方法并没有对过去或现在的心灵结构采取如此强烈的立场。仅举一个例子：Shanon （1998） 指出，内心语言使我们能够模拟我们在集体尝试解决问题或与同伴一起做出决定时可能进行的那种对话。换句话说，它允许我们对某些内容采用多种观点;我们首先充当发送者，然后充当接收者。如果我们用来生成通信内容的认知机制集与我们用来评估传入内容的认知机制集有些不同（看起来确实如此：参见 Mercier 和 Sperber 2017），那么很容易想象在发送者和接收者角色之间切换如何增强推理。认知机制存在如此差异的一个原因是，我们经常处于代理人的言语接收端，他们的 honest 不能被认为是理所当然的;因此，需要更彻底地审查它。

Turn now to the other suggestion, namely, inner speech’s ability to enhance our concentration. The thought here is this: even in the extreme case where inner speech serves as no more than a reflection of thought and thinking, it might nevertheless have a significant impact on the latter by sustaining the activity of certain cognitive mechanisms. This would in effect enhance our ability to stay focused on a particular task. This idea fits most naturally with com- petitive or “demonesque” models of cognitive function. The notion of a “demon” was introduced by Selfridge and Nessier (1960) in the context of their “pandemonium” model of visual letter recognition. These authors envisioned let- ter recognition as emerging from competitive interactions among hierarchically-organized layers of cognitive units (i.e., demons) where each unit specialized in the recogni- tion of a particular letter feature, or complex of features. Each demon would emit a “shout” where the volume of its shout was a function of how closely the stimulus matched the demon’s input condition. The shouting of lower-level demons determined the shouting of higher-level demons, eventually resulting a final “chorus” of shouts at the high- est level of the hierarchy, corresponding to a perceptual judgment.
现在转向另一个建议，即内心言语能够增强我们的注意力。这里的想法是这样的：即使在极端情况下，内心语言只不过是思想和思维的反映，但它仍然可能通过维持某些认知机制的活动而对后者产生重大影响。这实际上会增强我们专注于特定任务的能力。这个想法最自然地适合于认知功能的竞争性或 “恶魔式 ”模型。“恶魔”的概念是由 Selfridge 和 Nessier （1960） 在他们的视觉字母识别的“混乱”模型的背景下引入的。这些作者设想了 让识别来自分层组织的认知单元（即恶魔）层之间的竞争性互动，其中每个单元都专门识别特定的字母特征或特征复合体。每个恶魔都会发出“喊叫”，其中它的喊叫音量是刺激与恶魔输入条件的匹配程度的函数。低级恶魔的喊叫决定了高级恶魔的喊叫，最终导致等级制度最高层的最后“合唱”，对应于感性判断。

It is not too hard to imagine how inner speech might influ- ence the operation of a network of cognitive mechanisms governed by a scaled-up version of this logic. By inter- nally rehearsing sentences that report on thought contents recently processed by reasoning mechanisms (or planning mechanisms, etc.), that information might be kept salient in the system, biasing the chances that it will be processed further (which will inevitably mean that other information in the system will take a back seat). Phenomenologically, we have the experience of one “thought” prompting another “thought,” etc., but these inner speech acts might in fact be a reflection of thinking taking place in consciously inac- cessible parts of the brain, rather than constitutive of that reasoning itself. (The same might be said of planning and decision-making.)
不难想象，内心言语如何影响由这种逻辑的放大版本支配的认知机制网络的运作。通过内部排练报告最近由推理机制（或计划机制等）处理的思想内容的句子，这些信息可能会在系统中保持突出，从而影响它被进一步处理的机会（这将不可避免地意味着系统中的其他信息将退居二线）。从现象学上讲，我们有一个 “思想 ”促使另一个 “思想 ”等的经验，但这些内在的言语行为实际上可能是 思想发生在大脑中有意识的不可触及的部分的反映，而不是这种 推理本身的构成。（规划和决策也是如此。

These two suggestions for the cognitive functioning of inner speech are not in competition; both might be true. Moreover, the functions are complementary in the sense that both would bolster our ability to reason, to plan, and to make decisions, but in different ways: the first would result in a broader swath of information being accessible to the relevant cognitive mechanisms at a given time, while the second would enhance the processing of that information by enabling us to focus on particular problems for prolonged periods. If either of these suggestions is right, the effects of inner speech on the lives of our hominin ancestors would have been profound. And all the more so if both are true.
这两个关于内心言语认知功能的建议并不是竞争的;两者都可能是真的。此外，这些功能是互补的，因为两者都会增强我们推理、计划和决策的能力，但方式不同：前者将导致在给定时间相关认知机制可以访问更广泛的信息，而后者将使我们能够长时间专注于特定问题，从而增强对这些信息的处理。如果这些建议中的任何一个是正确的，那么内心语言对我们人类祖先的生活的影响将是深远的。如果两者都是真的，那就更是如此。

There are other possible ideas about the role of inner speech in our cognitive lives which might be relevant to explaining the origins of behavioral modernity, depending on how exactly the latter is understood. In particular, one might think that inner speech—through the near constant,narrative commentary it provides—plays a critical role in the construction of the autobiographical self, a sense of self as an agent who persists through time, has a unique life history, and a unique collection of personality traits (Fields 2002). To the extent that one thinks having such a rich sense of self is “part of what it is to be Human,”[15] one might regard inner speech as essential.
关于内心语言在我们的认知生活中的作用，还有其他可能的想法可能与解释行为现代性的起源有关，这取决于人们如何理解后者。特别是，人们可能会认为内心言语——通过它提供的近乎恒定的叙事性评论——在自传式自我的构建中起着关键作用，自传式自我是一种持续存在的自我意识，具有独特的生活史和独特的人格特征集合（Fields 2002）。在某种程度上，一个人认为拥有如此丰富的自我意识是“作为人类的一部分”，[15]一个人可能会认为内心的语言是必不可少的。

I want to finish out this section with a brief discussion of a thought-provoking question raised by Chris Fields. He asks why our “inner voice” should be inner at all, given that self- directed external speech—or what we above called “private speech”—would presumably be as effective at performing the functions claimed for inner speech. (Or in Vygotskyean terms: Why do we need inner speech in addition to private speech?) Field’s contention is that inner speech probably evolved under selection to keep the contents of our minds hidden in social cheating scenarios. That is plausible. How- ever, the ability to keep our thoughts to ourselves would have doubtlessly oiled the wheels of cooperation too (as evi- denced by the fact that those who completely lack a “social filter” seldom have many friends). It may also be that there are cognitive savings. For example, by constructing inner speech in higher auditory regions straightaway, we bypass the operation of many lower-level perceptual mechanisms involved in speech recognition (mechanisms that parse, detect prosody). Of course, these savings might be negligi- ble in any given case, but they might add up to something substantial over time, given the near omnipresence of inner speech in our waking lives.
我想通过对 Chris Fields 提出的一个发人深省的问题进行简要讨论来结束本节。他问道，既然自我导向的外部言语——或者我们上面所说的“私人言语”——在执行内心言语所声称的功能方面，大概同样有效，为什么我们的“内心声音”应该是内心的。（或者用维果茨基的术语来说：为什么除了私人言语之外，我们还需要内心的言语？菲尔德的论点是，内心语言可能是在选择下进化而来的，以便在社交作弊场景中隐藏我们的思想内容。这是有道理的。然而，将我们的思想藏在心里的能力无疑也会给合作的车轮上油（事实上，那些完全缺乏 “社会过滤器 ”的人很少有很多朋友）。也可能是存在认知储蓄。例如，通过直接在高级听觉区域构建内部语音，我们绕过了语音识别中涉及的许多较低层次的感知机制（解析、检测韵律的机制）的操作。当然，这些节省在任何特定情况下都可能微不足道，但随着时间的推移，考虑到我们清醒生活中几乎无处不在的内心语言，它们可能会累积起来。

Contemplation of this question raises another: was inner speech something that evolved subsequently to the evolu- tion of natural language? One possibility is that, as natural language evolved (in the service of interpersonal commu- nication), inner speech came online immediately as a by- product. Among other things, such a scenario would require a standing capacity for rich, combinatorial auditory imagery. Another possibility is that inner speech—at least as we now know it—involved significant changes in neural circuitry. (Most obviously, one might think it required novel forms of connectivity enabling the flow of information from language production parts of the brain to auditory areas.) I will not weigh in on which scenario is more plausible here, mainly because it would require a long detour into neuroscience. Suffice it to say that it is unlikely that the material record would ever be able to tell between these two scenarios. Even on the (I think, absurd) assumption that language and inner speech both evolved rapidly, independently, and at a signifi- cant temporal remove from one another, the availability of private speech during the interim undercuts the idea that these two events would leave separate pulses in the record. All of this raises still further fascinating questions and possibilities. In particular, it may be that there were interesting co-evolutionary connections between public lan- guage and inner speech, with the cognitive role of the latter imposing selection pressure on natural language forms. On this view, even if language first evolved for interpersonal communication, it may have been subsequently shaped in significant ways by its role in cognitive functioning.
对这个问题的思考提出了另一个问题：内心语言是不是随着自然语言的进化而进化而来的？一种可能性是，随着自然语言的进化（为人际交往服务），内心语言立即作为副产品在线出现。除其他外，这样的场景需要丰富的组合听觉图像的站立能力。另一种可能性是内心言语——至少我们现在知道的——涉及神经回路的重大变化。（最明显的是，人们可能会认为它需要新颖形式的连接，使信息能够从大脑的语言生产部分流向听觉区域。我不会在这里权衡哪种情况更合理，主要是因为它需要绕道进入神经科学。我只想说，材料记录不太可能区分这两种情况。即使假设语言和内心语言都是快速的、独立的，并且在时间上彼此相距甚远，但在此期间私人言语的可用性削弱了这两个事件会在记录中留下不同脉搏的想法。所有这些都提出了更多有趣的问题和可能性。特别是，公共语言和内部语言之间可能存在有趣的协同进化联系，后者的认知作用对自然语言形式施加了选择压力。根据这种观点，即使语言最初是为了人际交流而进化的，它也可能随后受到它在认知功能中的作用的重要影响。

Time to sum up. In this section, I have sketched a sec- ond way in which the notion of symbolic cognition might be taken: symbolic cognition might be understood as inner speech. Inner speech involves signs that are paradigmati- cally symbolic in Peirce’s sense, namely, natural language words and sentences. Phenomenologically, thought and thinking would seem to just be episodes of inner speech; however, that impression may be misleading. But even if it is, it is likely that the evolution of inner speech significantly enhanced human cognition, for example, by upgrading infor- mation flow and mental focus.
是时候总结一下了。在本节中，我勾勒了一种可以采用符号认知概念的第二种方式：符号认知可以被理解为内心语言。内心语言涉及皮尔斯意义上具有典型象征意义的符号，即自然语言的单词和句子。从现象学上讲，思想和思考似乎只是内心言语的插曲;然而，这种印象可能具有误导性。但即使是这样，内心语言的进化也可能显著增强了人类的认知能力，例如，通过升级信息流和精神集中度。

5 Symbol Crunching 5 符号运算

In this final section, I want to articulate yet a third way the notion of symbolic cognition might be taken. Some addi- tional conceptual groundwork must be carried out before this notion can be clearly expressed, however.
在最后一节中，我想阐明象征认知概念的第三种方式。然而，在清楚地表达这一概念之前，必须进行一些额外的概念基础工作。

The term “symbol” is used differently in cognitive sci- ence than it is in the context of Peirce’s theory of signs. Sometimes it is used to mean no more than a mental repre- sentation of something—a symbol for X is simply a brain state (or process) that is about or stands proxy for X. Often, however, “symbol” is reserved for particular forms of rep- resentation, and there are some interesting forms of overlap between these uses of the term and the Peircean notion of a symbol.
“符号”一词在认知科学中的使用方式与在皮尔斯符号理论中的使用方式不同。有时，它被用来表示对某物的心理代表——X 的符号只是关于 X 或代表 X 的大脑状态（或过程）。然而，“符号”通常被保留用于特定形式的代表，并且该术语的这些用法与皮尔西式的符号概念之间存在一些有趣的重叠形式。

One reasonably common distinction is between signals and symbols (see, e.g., Gallistel and King 2011). Signals are signs—information carrying vehicles—whose function it is to transmit information over space. Symbols, in contrast, are signs whose function it is to carry information forward in time. This is a distinction worth marking out, whether in this terminology or some other, for the two roles make quite different demands. Typically, signals should be fast, and they should decay rapidly so as not to “clog up” the communication channel (but not too rapidly). Symbols, on the other hand, are typically stored in some location and read from when the message they carry is needed, render- ing speed moot. And they tend to be adapted to endure for indefinite periods, for there is often no knowing in advance just when one might need some information. It is because of these differing job demands that the cooption of a pre- existing signal for use as a symbol—as in “reverberating loop” models of memory where a signal is sent around and around a circuit until the information it carries is needed— are generally profligate.
一个相当常见的区别是信号和符号之间（参见，例如 Gallistel 和 King 2011）。信号是标志 — 信息携带工具，其功能是在太空中传输信息。相比之下，符号是其功能是将信息向前推进的标志。这是一个值得区分的区别，无论是在这个术语还是其他术语中，因为这两个角色提出了完全不同的要求。通常，信号应该很快，并且它们应该迅速衰减，以免“堵塞”通信通道（但不要太快）。另一方面，符号通常存储在某个位置，并在需要它们携带的消息时读取，这使得速度没有意义。而且他们往往适应了无限期的忍耐，因为通常无法提前知道何时可能需要一些信息。正是由于这些不同的工作需求，将预先存在的信号用作符号——就像在“混响循环”的内存模型中，信号被发送出去，并且绕电路一圈，直到需要它所携带的信息——通常是挥霍无度的。

But the important point for present purposes is this. As a general rule, symbols are decoupled in space and time from the states of affairs they carry information about. (The cases where this is not true are degenerate ones.) They store information about events elsewhere and elsewhen. While signals can have this property too—after all, interpersonal linguistic communication exemplifies this property—it is seen as being uncommon within the brain of the organism (which is generally the relevant context in the case of cogni- tive science, at least in its traditional guise). Thus, there is a clear affinity between this cognitive scientific usage of the term “symbol” and the Peircean notion of a symbol: both feature spatio-temporal independence or displacement. We might denote symbols in the former sense as “symbolsM,” where “M” signifies that this notion of a symbol has its home in discussions of memory.
但就目前的目的而言，重要的一点是这一点。作为一般规则，符号在空间和时间上与它们所承载信息的事态是解耦的。（这不是真的情况是退化的。它们将有关事件的信息存储在 elsewhere 和 elsewhen 中。虽然信号也可以具有这种特性——毕竟，人际语言交流就是这种特性的例证——但它在生物体的大脑中被视为不常见（在认知科学的情况下，这通常是相关的背景，至少在其传统形式下）。因此，“符号”一词的这种认知科学用法与皮尔斯的符号概念之间存在明显的相似性：两者都具有时空独立性或位移。我们可以将前一种意义上的符号表示为“symbolsM”，其中“M”表示符号的概念在关于记忆的讨论中有其归宿。

The signal-symbol distinction is similar to another dis- tinction originally drawn by the cognitive scientist Peter Gärdenfors (1995). This distinction is between cued and detached representations. For Gärdenfors, a “cued repre- sentation stands for something that is present in the current external situation of the representing organism,” whereas a “detatched representation may stand for objects or events that are neither present in the current situation nor triggered by some recent situation” (1995, p. 266; emphases in origi- nal). The idea behind calling the first type of representation “cued” is that representations of this type are essentially tied to the presence of certain perceptual cues for their tokening. Detatched representations, by contrast, are not cued in this sense. When a dog hears the jingling of its owner’s keys, and on that basis forms a representation that its owner is home, that is a cued representation. When it dawns on the owner in the middle of a dinner party that he has forgotten to leave the back door open for his furry friend, that’s a detached representation.
信号-符号的区别类似于认知科学家 Peter Gärdenfors （1995） 最初绘制的另一个区别。这种区别在于 cueed 和 detached 表示。对 Gärdenfors 来说，“暗示的再现代表存在于所代表的有机体的当前外部状况中的事物”，而“被描述的再现可能代表既不存在于当前情况下，也不由某些近期情况触发的物体或事件”（1995 年，第 266 页;将第一种类型的表示称为 “cueed” 的想法是，这种类型的表示本质上与其标记的某些感知线索的存在有关。相比之下，从这个意义上说，Detatched representations 并不是 cues。当一只狗听到主人钥匙的叮当声，并在此基础上形成它的主人在家的表示时，这就是一个提示表示。当业主在一次晚宴中突然意识到他忘记为他毛茸茸的朋友打开后门时，这是一种超然的表现。

So, cued representations are tied to the here and now in a way that detached representations are not. The former are index-like in Peirce’s sense; the latter, symbol-like. Let us denote detached representations as “symbols D” (where “D” stands for “detached”). The two notions, symbol M and symbol D, are similar but not identical. A memory that can be evoked only when the animal is presented with the same context in which the memory was laid down would argu- ably be an instance of a symbolM that is not a symbol D. And going the other way: in representing an imaginary state of affairs, for example, we are led to token a representation that is a symbol D but not a symbol M.
因此，提示性表征与当前情境的联系紧密，而这点是 detached 表征所不具备的。前者在皮尔斯的意义上类似于指示符；后者则更像是符号。我们将 detached 表征称为“symbols D”（其中“D”代表“detached”）。这两种概念——symbol M 和 symbol D——是相似但并不完全相同的。一个只能在动物处于与记忆形成时相同的情境中才能被唤起的记忆，可以说是一个 symbol M 的实例，但不是一个 symbol D。而反过来，在表征一个虚构状态时，我们会产生一个表示，这个表示是 symbol D，但不是 symbol M。

There is a different distinction which is often drawn among representations, this one focusing on the format of the representations. Specifically, some representations are said to be modal or perceptual while others are said to be amodal. This distinction has been most discussed in connec- tion with so-called “grounded” theories of cognition (see, e.g., Barsalou 1999; Goldstone and Barsalou 1998). The rough idea is that modal representations refer in virtue of being realized by the same or highly similar neural activity which is occasioned by perception of the relevant object or property. In other words, they refer in virtue of resembling a perceptual experience of the thing. Amodal representations lack this feature, and so are described as being “arbitrary.” While both sorts of representations are frequently described as symbols—there are modal and amodal symbols—it is the latter that are seen as paradigmatically symbolic due to their close association with Classicalism[16 ] about our cogni- tive architecture, and with what John Haughland once called “Good-Old-Fashioned-Artificial-Intelligence” (“GOFAI”). Here, then, is another similarity between a traditional cog- nitive scientific usage of the term “symbol,” and Peirce’s notion of a symbol, though this time the similarity is differ- ent. Now it concerns a sign’s lack of iconicity, or its arbitrar- iness. Let us call this third type of cognitive science symbol “symbolA,” where “A” stands for “amodal.”
在表示之间通常有不同的区别，这个区别侧重于表示的格式。具体来说，一些表征被称为模态或感知，而另一些则被称为非模态。这种区别与所谓的“扎根”认知理论（参见 Barsalou 1999;Goldstone 和 Barsalou 1998）。粗略的想法是，模态表征是指由对相关对象或属性的感知引起的相同或高度相似的神经活动实现。换句话说，他们指的是类似于事物的感知体验。非模态表示缺乏此功能，因此被描述为 “任意”。虽然这两种表征经常被描述为符号——有模态符号和非模态符号——但后者被视为范式符号，因为它们与关于我们认知架构的古典主义[16]以及约翰·豪兰德（John Haughland）曾经称之为“好的老式人工智能”（“GOFAI”）密切相关。因此，这是“符号”一词的传统认知科学用法与皮尔斯的符号概念之间的另一个相似之处，尽管这次的相似性不同。现在它涉及标志缺乏象征性或任意性。让我们把这第三种类型的认知科学符号称为 “symbolA”，其中 “A” 代表 “amodal”。

Table 3 A taxonomy of cognitive science symbols
表 3 认知科学符号的分类法

Symbol Type	Description
$Symbo{l}_M$	A representation whose function it is to carry information forward in time (memory). Contrasts with signals
$Symbo{l}_D$	A representation that can be tokened in the absence of its referent. Contrasts with cued representations
$Symbo{l}_A$	An arbitrary representation. Contrasts with modal or perceptual representations
$Symbo{l}_S$	A representation exhibiting both arbitrariness and spatio-temporal independence

Thus, there are a variety of usages of the term “symbol” in cognitive science which overlap with the Peircean notion in one way or another. Most often, when cognitive scientists are discussing symbolsM or symbolsD, they are assuming that these symbols are also symbolsA, and vice versa. Or in other words: “symbol talk” in cognitive science has typi- cally expressed (and continues to express) the idea that we are dealing with internal signs that exhibit spatio-temporal independence from their referents and which do not stand in any theoretically interesting resemblance relation to their referents. We might denote symbols in this sense as “sym- bolsS,” where “S” stands for “standard cognitive science.” (Table [3 ] summarizes each of these notions.)
因此，在认知科学中，“符号”一词有多种用法，它们以某种方式与皮尔斯的概念重叠。大多数情况下，当认知科学家讨论 symbolsM 或 symbolsD 时，他们假设这些符号也是 symbolsA，反之亦然。或者换句话说：“认知科学中的”符号谈话“已经典型地表达了（并将继续表达）这样一种想法，即我们正在处理的内部符号，这些符号表现出与所指对象的时空独立性，并且与它们的所指对象没有任何理论上有趣的相似关系。我们可以将这种意义上的符号表示为“sym- bolsS”，其中“S”代表“标准认知科学”。（表 [3 ] 总结了这些概念中的每一个。

One question this terminology enables us to ask is this: are symbolsS properly regarded as Peircean symbols? The issue would seem to boil down to whether the meanings of symbolsS are governed by convention. At one point in time, the idea that such symbols might mean by convention would have seemed absurd. To think this would have been to com- mit the homunculus fallacy on an epic scale. But the recent work of Skyrms and colleagues that was mentioned above suggests it is not absurd. For that work shows how agents lacking in intelligence—and in the limit, totally individu- ally mindless agents—might develop signaling conventions. That includes parts of an individual organism. If two parts of the brain—two cognitive mechanisms, say—can stand to one another as sender and receiver in a signaling game, then the signs they exchange can mean by convention. It is not difficult to see how cognitive mechanisms might play these roles. Simply put, all that is required is for the output of one cognitive mechanism C to serve as the input to another C*; for C to have access to information that C* would otherwise lack were it not for the inputs it receives from C; and for C and C* to have “interests” in the sense that some ways of coordinating their operation better promote the stabilization and/or spread of the mechanisms than others.[17]
这个术语使我们能够提出的一个问题是：符号是否正确地被视为皮尔斯符号？问题似乎归结为符号的含义是否受惯例支配。在某个时间点，认为这些符号可能意味着约定俗成的想法似乎很荒谬。如果这样想，那就等于以史诗般的规模来谬误人类谬误。但上面提到的 Skyrms 及其同事最近的工作表明这并不荒谬。因为这项工作展示了缺乏智能的代理——并且在极限下，完全没有头脑的代理——如何发展出信号约定。这包括单个生物体的各个部分。如果大脑的两个部分（例如两种认知机制）可以在信号游戏中作为发送者和接收者相互代表，那么它们交换的符号可以按照惯例表示。不难看出认知机制如何发挥这些作用。简单地说，所需要的只是让一个认知机制 C 的输出作为另一个认知机制 C* 的输入;让 C 能够访问如果不是 C* 从 C 那里接收到的输入，它就会缺乏的信息;C 和 C* 具有“利益”，即某些协调其运作的方式比其他方式更能促进机制的稳定和/或传播。[17]

There is much more that might be said at this point, but in the remainder of this section I want to limit my focus to just the following question, namely: what connection might there be between having a mind that computes symbolsS and the capacity to use external Peircean symbols? What might the connection be between internal and external Peircean symbols?
在这一点上可以说更多，但在本节的其余部分，我想将我的注意力限制在以下问题上，即：拥有计算符号的头脑和使用外部皮尔斯符号的能力之间可能存在什么联系？内部和外部皮尔斯符号之间可能有什么联系？

Here is one possible answer. According to Gärdenfors, detached representations play a crucial role in explaining the evolution of language. As I understand him, the main idea is that we can explain our linguistic capacity to communicate about the elsewhere and elsewhen in terms of our anteced- ent capacity to think about the elsewhere and elsewhen. In other words: detached minds make displaced communication possible. However, as Gärdenfors realizes, simply having detached representations isn’t enough for displaced com- munication, as many animals possess the former but not the latter. Hence, he suggests that the difference maker lies in humans’ readiness, or facility for, associating public signals to “corresponding references in [our] inner environments” (1995, p. 271).[18]
这是一个可能的答案。根据 Gärdenfors 的说法，分离的表征在解释语言的演变中起着至关重要的作用。根据我的理解，主要思想是，我们可以根据我们思考 elsewhere 和 elsewhen 的先验能力来解释我们交流 elsewhere 和 elsewhen 的语言能力。换句话说：超然的思想使流离失所的交流成为可能。然而，正如 Gärdenfors 所意识到的那样，仅仅拥有分离的表征对于流离失所的交流来说是不够的，因为许多动物拥有前者，但没有后者。因此，他认为差异制造者在于人类准备好或能够将公共信号与“[我们]内部环境中的相应参考”联系起来（1995 年，第 271 页）。[18]

Perhaps Gärdenfors is right. But even if he is, this gets us only part of the way. It provides an explanation for the human ability to learn the symbolic (in the Peircean sense) primitives of human languages. That is done (so the story goes) by mapping these items onto what I have called symbolsD. But of course we learn a lot more than that; we learn how to produce and understand an infinity of public language expressions, as Chomsky, Fodor, and others have repeatedly pointed out over the years. And to explain that, it seems we must appeal to an internal representational system with some very special properties indeed—displaced refer- ence, yes, but also productivity and systematicity.
也许 Gärdenfors 是对的。但即使他是，这也只能让我们走得更远。它解释了人类学习人类语言的象征性（在皮尔斯意义上）原始语言的能力。这是通过将这些项目映射到我称之为 symbolsD 的东西来完成的（故事是这样的）。但当然，我们学到的远不止这些;我们学习如何产生和理解无穷无尽的公共语言表达，正如乔姆斯基、福多尔和其他人多年来反复指出的那样。为了解释这一点，我们似乎必须诉诸于一个确实具有一些非常特殊属性的内部表征系统——是的，取代了指涉，但也包括生产力和系统性。

It is at this point that we again meet the idea of a language of thought: an internal representational system governed by a combinatorial syntax and semantics. This system (or each if there are many) is conceived of as being governed by principles that specify both how complex expressions are to be built up out of less complex expressions, and of how the meaning of a whole expression is determined by the meaning of the parts, together with how those parts are assembled. This internal language is also conceived of as featuring recursion, thereby explaining how a finite system of discrete symbols and rules can yield an infinity of mean- ingful expressions (i.e., thoughts).[19] So, on this view, one or another story is told about how the primitives in a language of thought get their meaning (e.g., computational role, a cer- tain history of selection, asymmetric causal relations, etc.), but the meaning of complex expressions is determined by the compositional semantics of the system. Displaced refer- ence on the part of the primitives receives one treatment; displaced reference on the part of the wholes another.
正是在这一点上，我们再次遇到了一种思想语言的概念：一个由组合句法和语义支配的内部表征系统。这个系统（如果有很多，则每个系统）被认为是受原则支配的，这些原则既指定了复杂的表达式如何从不太复杂的表达式中构建出来，又指定了整个表达式的含义如何由各部分的含义决定，以及这些部分是如何组装的。这种内部语言也被设想为具有递归的特点，从而解释了一个由离散符号和规则组成的有限系统如何产生无穷无尽的卑鄙表达（即思想）。[19]因此，根据这种观点，关于思想语言中的原语如何获得其意义（例如，计算作用、选择的记录历史、不对称的因果关系等），讲述了一个或另一个故事，但复杂表达的意义是由系统的组合语义决定的。原语部分的移位指涉受到一种处理;整体部分的移位参考是另一个。

But what about the arbitrariness of the representations in a language of thought? Does this feature do any work in explaining our facility with language? At first blush, it’s not obvious that arbitrariness has any role to play here. Suppose one thinks, for example, that most or even all of our con- ceptual representations are actually modal representations. Couldn’t one still accept Gärdenfors’ account of Peircean symbol acquisition? And more generally, couldn’t one still hold that our minds contain a powerful language of thought, in virtue of which we are capable of learning to use powerful public symbol systems?
但是，在思想语言中，表征的任意性呢？此功能在用语言解释我们的设施方面有什么作用吗？乍一看，武断在这里扮演任何角色并不明显。例如，假设一个人认为我们的大部分甚至全部的感性表征实际上是模态表征。难道人们仍然不能接受 Gärdenfors 关于皮尔斯符号获取的描述吗？更一般地说，难道人们仍然认为我们的思想包含一种强大的思想语言，凭借这种语言，我们能够学会使用强大的公共符号系统吗？

The answer to the first question is an unequivocal “yes.” A modal representation of a cat might resemble the percep- tual experience of a cat, but this representation need not be tied to—in the sense of only being capable of being evoked by—some perceptual cue of a cat. (Were all modal represen- tations cued representations, they would at most play a very limited role in our psychologies.) More generally, there is no tension between being a modal representation and exhibiting spatio-temporal independence.
第一个问题的答案是明确的“是”。猫的模态表征可能类似于猫的感知体验，但这种表征不需要与猫的某种感知线索相关联——在某种意义上只能被唤起。（如果所有的情态再现都是暗示性的，它们在我们的心理学中最多只能发挥非常有限的作用。更一般地说，作为模态表示和表现出时空独立性之间没有紧张关系。

The second question is more complicated. Here I can do little more than present a statement of the worries one might have. Fodor (1998) argued—and many philosophers and cognitive scientists have agreed with him—that whatever else primitive representations are, they had better compose to form complex representations, or else we are at a loss to explain the productivity of thought. Fodor then used this claim to argue against a variety of theories of concepts, most famously the prototype theory, arguing that prototypes just don’t compose in the right kind of way. His main example was the (complex) concept PET FISH. He said that our pro- totype of a pet—say, a dog—and our prototype of a fish— say, a trout—do not compose to form our prototype of a pet fish—something like a goldfish. The example is light- hearted, but the point it illustrates is serious. It would seem that an analogous worry arises for a modal conception of primitive representations.[20] Moreover, modal representa- tions impose further vehicular constraints that a prototype theory does not, or at least need not, impose. This is because modal representations are perceptual representations, ones realized in the same nervous tissue that the correspond- ing perceptual events are realized in. (A prototype could be an abstract data structure by contrast.) Hence, compos- ing modal representations requires “blending” two or more quasi-perceptual vehicles. That may well be possible, but it is unlikely to be anywhere near as cheap, nor as fast, as the construction of a complex representation utilizing arbitrary representations. So, one might think that arbitrary mental representations serve to make cognition more computation- ally efficient, including cognition underlying the production and interpretation of natural language expressions. And that may well be important for explaining our great proficiency for learning and using complex public symbol systems like language, even if it is not necessary for our basic ability to learn public symbols.
第二个问题要复杂得多。在这里，我只能陈述一个人可能有的担忧。Fodor （1998） 认为——许多哲学家和认知科学家都同意他的观点——无论其他原始表征是什么，它们最好组合起来形成复杂的表征，否则我们将无法解释思维的生产力。然后，Fodor 用这个主张来反驳各种概念理论，最著名的是原型理论，认为原型就是没有以正确的方式组合。他的主要例子是（复杂的）概念 PET FISH。他说，我们的宠物原型（比如狗）和鱼的原型（比如鳟鱼）并不能构成我们的宠物鱼原型——比如金鱼。这个例子很轻松，但它说明的要点是严肃的。似乎对原始表征的模态概念也出现了类似的担忧。[20]此外，模态表示施加了原型理论没有或至少不需要施加的进一步车辆约束。这是因为模态表征是知觉表征，在相应的知觉事件实现的同一神经组织中实现的表征。（相比之下，原型可以是抽象数据结构。因此，合成模态表示需要“混合”两个或多个准知觉载体。这很可能是可能的，但它不太可能像使用任意表示构造复杂表示那样便宜，也不像那样快速。因此，人们可能会认为，任意的心理表征有助于提高认知的计算效率，包括自然语言表达的产生和解释的基础认知。这对于解释我们学习和使用复杂的公共符号系统（如语言）的高度熟练程度可能非常重要，即使对于我们学习公共符号的基本能力来说不是必需的。

While this barely scratches the surface, it is enough to illustrate how both the spatio-temporal independence and the arbitrariness of symbolsS might bear upon an explanation of humans’ capacity to learn and use language. One simple view linking all the issues we’ve been discussing together might run as follows. The evolution of natural language, a system thoroughly symbolic in Peirce’s sense, both at the level of primitives and at the level of complex expressions, required the prior existence of a rich, combinatorial system of internal symbolsS. Possession of such a symbol system constitutes yet another way in which “symbolic cognition” might be understood: one “has” symbolic cognition when one possesses a rich, combinatorial symbolS system for thought and thinking. Inner speech then followed in the wake of language evolution, either directly (if it was a by-product of the evolution of language), or indirectly (if some new neural hardware was required). On this view, there would thus be a chain of dependencies linking the different forms of symbolic cognition we have distinguished.
虽然这仅仅触及了表面，但它足以说明时空独立性和符号的任意性如何影响对人类学习和使用语言能力的解释。将我们一直在讨论的所有问题链接在一起的简单视图可能如下所示。自然语言的演变，一个在皮尔斯意义上完全具有符号性的系统，无论是在原始表达的层面还是在复杂表达的层面，都需要一个丰富的、内部符号的组合系统S的先验存在。拥有这样的符号系统构成了另一种可以理解“符号认知”的方式：当一个人拥有一个丰富的、用于思考和思考的组合符号系统时，他就“拥有”符号认知。然后，内心语言紧随语言进化，要么是直接的（如果它是语言进化的副产品），要么是间接的（如果需要一些新的神经硬件）。因此，根据这种观点，将存在一个依赖关系链，将我们已经区分的不同形式的符号认知联系起来。

But there are other options as well. In particular, it may well be that these different forms of symbolic cognition co- evolved over long periods of time. For example, it is not hard to imagine how a primitive language of thought—perhaps one that was shared with other primates—might have been gradually scaled up over time as communicative complexity increased in our ancestral line; how an upgraded language of thought might have, in turn, further increased communi- cative complexity; and so on. Inner speech might have also exerted evolutionary influence on these other traits, with the latter feeding back to influence the evolution of inner speech itself. An account along these lines, if less neat than the foregoing, seems more plausible on evolutionary grounds.
但也有其他选择。特别是，这些不同形式的符号认知很可能是在很长一段时间内共同进化的。例如，不难想象，随着我们祖先血统中交流复杂性的增加，一种原始的思想语言——也许是与其他灵长类动物共享的语言——是如何随着时间的推移而逐渐扩大的;升级后的思想语言如何反过来进一步增加交际的复杂性;等等。内心言语也可能对这些其他特征产生进化影响，后者反噬影响内心言语本身的进化。沿着这些思路的解释，即使不如前面那么简洁，但在进化论的基础上似乎更合理。

We have seen, then, that there is a variety of ways in which the notion of symbolic cognition might be taken. Here I have elucidated three natural construals, though there are probably others. Progress will be made on “the” problem of the origins of human symbolic cognition and its role in explaining the distinctive features of human lifeways by being clearer about which version of this notion we have in mind in a given context.
因此，我们已经看到，符号认知的概念可以有多种方式被采用。在这里，我阐明了三种自然解释，尽管可能还有其他解释。通过更清楚地了解我们在特定背景下所想到的这个概念的哪个版本，我们将在人类符号认知的起源及其在解释人类生活方式的独特特征方面的作用这一“问题”上取得进展。

Acknowledgements

I would like to thank Kim Sterelny and Lauren Reed for comments on an early draft of this article. I would also like to thank the anonymous referees for their feedback.

Compliance with Ethical Standards

Conflict of interest The author reports no conflict of interest.

Ethical Approval This article does not contain any studies with human participants or animals performed by any of the authors.

References

---

[1] School of Philosophy and Centre of Excellence

for the Dynamics of Language, The Australian National University, Canberra, ACT, Australia

[2] See Currie and Killin (2019) for an in-depth discussion of the rela- tionship between material remains and hominin cognitive capacities that is likely to prove useful in carrying out this project.

[3] For some prominent examples of evolutionary theorists who have drawn on Peirce’s theory to answer this question, see Deacon (1998), Rossano (2010), and Everett (2017).

[4] See Atkin 2013 for a good summary.

[5] See Planer and Kalkman (2019) for a discussion and an improved definition of iconic signs.

[6] See Skyrms (2010) for an excellent overview of this work.

[7] This idea is very similar to a proposal in Deacon (1998).

[8] A wonderful real-life example of the former kind of system: in many Australian Aboriginal cultures, storytelling makes heavy use of maps drawn in the sand (Wilkins 1997). Events within a story are localized to areas within the simultaneously produced sand map.

[9] See Vicente and Martinez Manrique (2011) for a discussion and references.

[10] Again, see Vicente and Martinez Manrique (2011) for a nice over- view.

[11] See, in particular, Vygotsky (2012 [1932]).

[12] Mechanisms, not “modules.” I take it that everyone agrees there are many cognitive mechanisms. Whether many, or most, of these mechanisms are “modular” in some interesting sense remains a highly contentious issue.

[13] Dan Dennett (1991) is perhaps an instance of this view.

[14] There are of course intermediate possibilities. Were one of these true, then inner speech would play a kind of hybrid role.

[15] I am here thinking of those authors who, in McBrearty and Brooks’ (2000) words, see themselves as “search[ing] for the soul, for the inventive spark that distinguishes humans from the rest of the ani- mal kingdom” (p. 533). Antonio Damasio (2012) would be such an example.

[16] For canonical statements of Classicalism, see, e.g., Fodor (1975), Newell (1980), and Fodor and Pylyshyn (1988).

[17] For a more detailed treatment of these ideas, see Planer (forthcom- ing).

[18] See also James Hurford (2014) who approvingly cites Gärdenfors, and tells a very similar story.

[19] As recent work by Chomsky (e.g., Berwick and Chomsky 2016) has made clear, language for him just is humans’ language of thought. We alone possess such a language (so the view goes). Fodor, in con- trast, would claim that languages of thought are much more wide- spread in the animal kingdom. Though he would presumably claim, along with Chomsky, that humans’ language of thought has some special formal properties (e.g., hierarchical structure, recursion).

[20] See Prinz and Barsolou (2002) for a response to this kind of worry.

---

Atkin A (2013) Peirce’s theory of signs. In: Zalta EN (ed) The stanford encyclopedia of philosophy, summer edition. https://plato.stanford.edu/archives/sum2013/entries/peirce-semiotics/

Baars BJ (1997) In the theatre of consciousness. Global workspace theory, a rigorous scientific theory of consciousness. J Conscious Stud 4(4):292–309

Baars BJ (2005) Global workspace theory of consciousness: toward a cognitive neuroscience of human experience. Prog Brain Res 150:45–53

Barsalou LW (1999) Perceptions of perceptual symbols. Behav Brain Sci 22(4):637–660

Berwick RC, Chomsky N (2016) Why only us: language and evolution?

MIT Press, Cambridge

Bowles S, Gintis H (2013) A cooperative species: human reciprocity and its evolution. Princeton University Press, Princeton

Boyd R (2017) A different kind of animal: how culture transformed our species. Princeton University Press, Princeton

Carruthers P (2006) The architecture of the mind. Oxford University Press, Oxford

Chase P (2003) Comment on “The origin of modern behavior: a review and critique of models and test implications” by Henshilwood et al. Curr Anthropol 44(5):637

Currie A, Killin A (2019) From things to thinking: cognitive archaeol- ogy. Mind Lang 34(2):263–279

Damasio A (2012) Self comes to mind: constructing the conscious brain. Vintage, New York

Deacon TW (1998) The symbolic species: the co-evolution of language and the brain. WW Norton & Company, New York

Dennett DC (1991) Consciousness explained. Little, Brown and Com- pany, Boston

Everett D (2017) How language began: the story of humanity’s greatest invention. Profile Books, London

Fields C (2002) Why do we talk to ourselves? J Exp Theor Artif Intell 14(4):255–272

Fodor JA (1975) The language of thought. Harvard University Press, Cambridge

Fodor JA (1998) Concepts: where cognitive science went wrong.

Oxford University Press, Oxford

Fodor JA, Pylyshyn ZW (1988) Connectionism and cognitive architec- ture: a critical analysis. Cognition 28(1–2):3–71

Gallistel CR, King AP (2011) Memory and the computational brain: why cognitive science will transform neuroscience, vol 6. Wiley, New York, p 2011

Gärdenfors P (1995) Cued and detached representations in animal cog- nition. Behav Proc 35(1–3):263–273

Goldstone RL, Barsalou LW (1998) Reuniting perception and concep- tion. Cognition 65(2-3):231–262

Henshilwood CS et al (2003) The origin of modern human behavior: critique of the models and their test implications. Curr Anthropol 44(5):627–651

Hurford JR (2014) Origins of language: a slim guide. Oxford Univer- sity Press, Oxford

Lewis D (1969) Convention Cambridge. Harvard University Press, Cambridge

Malafouris L (2013) How things shape the mind. MIT Press, Cambridge

McBrearty S, Brooks AS (2000) The revolution that wasn’t: a new interpretation of the origin of modern human behavior. J Hum Evol 39(5):453–563

Mercier H, Sperber D (2017) The enigma of reason. Harvard University Press, Cambridge

Mithen S (1996) The prehistory of the mind: a search for the origins of art, religion and science. Thames & Hudson, Great Britain

Mithen S (2005) The singing Neanderthals: the origins of music, lan- guage, mind and body. Weidenfeld & Nicolson, Great Britain

Newell A (1980) Physical symbol systems. Cognit Sci 4(2):135–183 Noble W, Davidson I (1991) The evolutionary emergence of mod-ern human behaviour: language and its archaeology. Man 1991:223–253

Nowell A (2010) Defining behavioral modernity in the context of Neandertal and anatomically modern human populations. Annu Rev Anthropol 39:437–452

Peirce CSS (ed) (1883) Studies in logic. By members of the Johns Hopkins University. Little, Brown, and Company, New York

Planer RJ (forthcoming) The evolution of languages of thought. Biol Philos

Planer RJ, Kalkman D (2019) Arbitrary signals and cognitive complex- ity. Br J Philos Sci. https://doi.org/10.1093/bjps/axz018

Prinz JJ, Barsolou LW (2002) Acquisition and productivity in percep- tual symbol systems: an account of mundane creativity. Creativity, cognition, and knowledge: an interaction. Praeger, Santa Barbara, p 105

Renfrew C (2004) Towards a theory of material engagement. Rethink- ing materiality: the engagement of mind with the material world. McDonald Institute, Cambridge, pp 23–31

Rossano MJ (2010) Making friends, making tools, and making sym- bols. Curr Anthropol 51(S1):S89–S98

Selfridge OG, Neisser U (1960) Pattern recognition by machine. Sci Am 203(2):60–69

Shanon B (1998) What is the function of consciousness? J Conscious Stud 5(3):295–308

Skyrms B (2010) Signals: evolution, learning, and information. Oxford University Press, Oxford

Soffer O (2009) Defining modernity, establishing rubicons, imagining the other—and the Neanderthal enigma. Sourcebook of Paleo- lithic transitions. Springer, New York, pp 43–64

Stiner MC (2017) Love and death in the Stone Age: what constitutes first evidence of mortuary treatment of the human body? Biol Theory 12(4):248–261

Stringer C, Gamble C (1993) In search of the Neanderthals: solving the puzzle of human origins. Thames and Hudson, New York

Tattersall I (2008) An evolutionary framework for the acquisition of symbolic cognition by Homo sapiens. Compar Cognit Behav Rev 3:99–114

Vicente A, Martinez Manrique F (2011) Inner speech: nature and func- tions. Philos Compass 6(3):209–219

Vygotsky LS (2012 [1932]) Thought and language. MIT Press, Cambridge

Wadley L (2001) What is cultural modernity? A general view and a South African perspective from Rose Cottage Cave. Camb Archaeol J 11(2):201–221

Wilkins DP (1997) Alternative representations of space: arrernte nar- ratives in sand. The visual narrative reader. Bloomsbury, London, pp 252–281

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Ronald J. Planer Ronald.Planer@anu.edu.au

---

via：

• What is Symbolic Cognition? | Topoi © Springer Nature B.V. 2019

  https://link.springer.com/article/10.1007/s11245-019-09670-5