Is social learning always locally adaptive?
Is social learning always locally adaptive?
复制标题
社会学习总是本地适应性的吗？
作者：
K. Laland
Galef（1995）令人信服地认为，社会习得的行为模式通常是适应性的，并证明了经验证据的权重支持这一立场。不幸的是，这篇值得称赞的文章包含了几个不准确的陈述，可能会导致读者对社会学习理论得出错误的结论。Galef的文章是对动物行为的社会传播数学模型的回应，他批评该模型得出了“反直觉（和反事实）的结论，即传统的行为模式往往不是局部适应性的”（第1325页）。Galef认为，这些模型将社会获得的特征视为“不受取决于其表达的奖励或惩罚的影响”（第1325页），从而达到了这种错误的立场。相反，我认为：（1）许多文化进化模型明确地纳入了个体经验对社会习得行为的修正效应;（2）应用于动物的文化进化模型并没有得出社会学习通常是适应不良的结论;（3）这些模型的发现并不意味着个体和社会获得的信息的相对寿命。尽管如此，我仍然认为动物的社会学习是否总是适应性的，这是一个悬而未决的问题，最好通过实证研究来解决。Galef的批评主要集中在两个文化进化模型上（Boyd & Richerson 1988; Rogers 1988），他认为这两个模型假设社会学习行为模式的表达频率不受个人经验的影响。这是对罗杰斯模型的有效批评，但不是对博伊德和理查德森的工作的批评，也不是对加莱夫忽略的其他相关理论研究的批评（Cavalli-Sforza & Feldman 1981; Vickery et al. 1991; Giraldeau et al. 1994）。加莱夫不恰当地把对罗杰斯论文的批评变成了该领域的不足。在早期的分析中，Boyd & Richerson（1985，pp. 81-131）提出了一个“引导变异”的模型，其中社会习得的行为是明确修改的个人经验的基础上。这种学习和社会传播的结合增加了那些受学习所青睐的特征的频率，博伊德和理查德森强调，这种力量是适应行为的来源。在一个动物种群中，这种特征可能是特定饮食的消耗量，或者在特定位置觅食所花费的时间比例。Galef（1326页）区分了Boyd & Richerson（1985）的“引导变异”模型，他赞成，从他们的1988年模型，他批评。事实上，这两个模型在概念上是相似的，都是引导变异模型，它们的差异在这里无关紧要。Boyd & Richerson（1988）假设“群体中的个体通过社会学习和个体学习的某种混合来获得他们的行为”（第32页）。具体来说，他们假设动物只在个体经验无法提供指导的情况下才依赖于社会线索，以确定哪种行为变体最有可能成功（即适应性）。Galef引用了他们的说法，“一旦一个人获得了[一种行为]，它就不会改变”（第33页），暗示社会习得的变体对个人经验是不可改变的。这种解释是错误的。为了解决进化问题，数学模型通常将一生的经历浓缩为一系列离散的阶段。Boyd & Richerson（1988）假设了一个简化的生命周期：（1）个人经验，如果不确定的话。N. Laland，剑桥大学动物行为学分系，Madingley，剑桥CB 3 8AA，英国。（电子邮件：knl1001@hermes.cam.ac.uk）。
Galef (1995) convincingly argued that socially learned behaviour patterns are typically adaptive, and demonstrated that the weight of empirical evidence supports this position. Unfortunately this otherwise commendable article contains several inaccurate statements that might lead readers to mistaken conclusions about social learning theory. Galef’s article is a response to mathematical models of the social transmission of behaviour in animals, which he criticized for reaching ‘the counterintuitive (and counterfactual) conclusion that traditional behaviour patterns are often not locally adaptive’ (page 1325). Galef argued that such models reach this erroneous position by treating socially acquired traits as ‘not affected by rewards or punishments contingent upon their expression’ (page 1325). In contrast, I argue that (1) many cultural evolution models explicitly incorporate the modifying effects of individual experience on socially learned behaviour, (2) cultural evolution models applied to animals do not conclude that social learning is often maladaptive, and (3) the findings of these models imply nothing about the relative longevity of individually and socially acquired information. None the less, I maintain that whether animal social learning is always adaptive is an open question, best resolved through empirical study. Galef’s critique focused primarily on two cultural evolution models (Boyd & Richerson 1988; Rogers 1988), which he argues assume that the frequency of expression of socially learned behaviour patterns is not affected by individual experience. This is a valid criticism of Rogers’ model, but not of Boyd & Richerson’s work, nor of other relevant theoretical studies that Galef ignores (Cavalli-Sforza & Feldman 1981; Vickery et al. 1991; Giraldeau et al. 1994). Galef inappropriately turned criticism of Rogers’ paper into a deficiency of the field. In an earlier analysis, Boyd & Richerson (1985, pp. 81–131) presented a model of ‘guided variation’ in which socially learned behaviour is explicitly modified on the basis of personal experience. This combination of learning and social transmission increases the frequency of those traits favoured by learning, and Boyd & Richerson emphasized that this force is a source of adaptive behaviour. In an animal population, such a trait might be the amount of a particular diet consumed, or the proportion of time spent foraging at a particular location. Galef (page 1326) distinguished Boyd & Richerson’s (1985) ‘guided variation’ model, which he favoured, from their 1988 model which he criticized. In fact, the models are conceptually similar, both being guided variation models, and their differences are of no relevance here. Boyd & Richerson (1988) assumed that ‘individuals in the population acquire their behavior by some mixture of social learning and individual learning’ (page 32). Specifically, they assumed that animals rely on social cues only when individual experience fails to provide guidance as to which behavioural variant is most likely to be successful (i.e. adaptive). Galef quoted their statement, ‘once an individual has acquired [a behaviour] it does not change’ (page 33), to imply that socially learned variants are immutable to individual experience. This interpretation is mistaken. To address evolutionary questions, it is common practice for mathematical models to condense a lifetime’s experience into a series of discrete stages. Boyd & Richerson (1988) assumed a simplified life cycle of (1) individual experience, which if inconclusive Correspondence: K. N. Laland, Sub-department of Animal Behaviour, University of Cambridge, Madingley, Cambridge CB3 8AA, U.K. (email: knl1001@hermes.cam.ac.uk).