Higher order sequential effects in psychophysical judgments
Higher order sequential effects in psychophysical judgments
复制标题
心理物理学判断中的高阶序列效应
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
Justus
中科院分区：
文献类型：
--
作者：
Justus
当一系列刺激被呈现并且一个特定的属性被判断时，反应不仅取决于当前的刺激，而且还取决于在早期试验中出现的刺激-反应对。通常，当前响应与紧接在前的刺激负相关，并且与紧接在前的响应正相关（例如，Jesteadt，吕斯，&绿色，1977;Petzold，1981;Schifferstein & Frijters，1992）.一些早期的研究结果表明，连续效应延伸超过一个试验。例如，Ward和Lockhead（1971）对绝对判断中的反应错误进行的分析表明，序列效应的深度至少延伸到五次试验之前。然而，远程事件的影响可能是由于试验之间的响应传播。例如，试验t的反应影响试验t+1的反应，而试验t +1的反应又影响试验t+ 2的反应，等等。Jesteadt et al.（1977）试图通过使用线性回归模型来探索序贯效应的深度。通过分析多重相关性，他们发现，在回归方程中加入前一个刺激-反应对，多重相关性会增加，但加入多个事件并不会显著增加相关系数。从这一发现中，他们得出结论，序列效应仅限于前一次试验的刺激反应事件。然而，Staddon、King和Lockhead（1980）认为，至少前两次试验都涉及到了。他们的论点是基于绝对判断中的反应误差与滞后1的刺激正相关，而与滞后2的刺激负相关的发现。这样的交叉不能用只包含一次试验的事件的模型来解释。类似地，在混合模态心理量表中，Ward（1985，1986）发现了四次试验滞后的序贯效应。可能，顺序效应的深度取决于判断任务的性质。Staddon等人提到了关于绝对判断的研究，而Jesteadt等人则专注于三个关于震级估计的实验。如果一次以上的事件影响了对当前刺激的判断，那么这种高阶序列效应的性质是什么？更具体地说，是不同的顺序介导的不同过程的效果？Ward和Lockhead（1970）认为，前一个事件作为一个比较标准，而事件的影响超过一个审判回来是由其他过程调解。或者，人们可以认为，更早的事件也可以像前一个事件一样作为标准（Haubensak，1992 a）。本研究旨在探讨高阶序贯效应，特别是关于两个问题。首先，序列效应的深度是否取决于任务？数量估计和类别判断的判断过程可能不同，导致不同的深度。第二，如果更早的事件被纳入判断过程，它们是否满足与前一事件相同的功能？本文首先讨论的模型，使人们能够获得更具体的假设高阶序列的影响，特别是相互作用的先前事件。
When a series of stimuli is presented and a particular attribute is to be judged, responses depend not only on the current stimulus but also on which stimulus–response pairs occurred on earlier trials. In general, the current response is negativelycorrelated with the immediately preceding stimuli and positively correlated with the immediatelypreceding responses (e.g., Jesteadt, Luce, & Green, 1977;Petzold, 1981;Schifferstein & Frijters, 1992). Some earlier findings suggest that sequential effects extendover more than one trial. For instance, the analysis of response errors in absolute judgmentsby Ward and Lockhead (1971) suggests that the depth of sequential effects extends to at least five trials back. However, the effect of remote events could be due to response propagation from trial to trial. For instance, the response on trial t affects the response on trial t+1, which, in turn, affects the response on trial t+ 2, and so on. Jesteadt et al. (1977) attempted to explore the depth of sequential effects by using a linear regression model. Analyzing multiple correlations, they found that adding the immediately preceding stimulus– responsepair to the regression equationproducedan increment in the multiple correlation, but adding events more than one trial back did not significantly increase the correlation coefficient. From this finding they concludedthat sequential effects are limited to the stimulus–response event of the immediately preceding trial. Staddon, King, and Lockhead (1980), however, argued that at least the preceding two trials back are involved. Their arguments are based on the finding that the response error in absolute judgments is positivelycorrelated with the stimuliof lag 1, but negatively correlated with the stimuli of lag 2. Such a crossover cannot be accounted for by a model that incorporates only events one trial back. Similarly, in mixedmodalitypsychophysicalscaling,Ward (1985, 1986) found sequential effects for lags up to four trials. Possibly, the depth of sequential effects depends on the nature of the judgment task. Staddon et al. referred to studies on absolute judgments,whereas Jesteadt et al. focused on three experiments on magnitude estimation. If events more than one trial back affect the judgment of the current stimulus, what is the nature of such higher order sequential effects? More specifically, are effects of different order mediated by different processes? Ward and Lockhead (1970) argued that the immediately preceding event serves as a comparative standard, whereas the effect of events more than one trial back is mediated by other processes. Alternatively, one could argue that events further back can also function as standards in the same way as the immediately preceding event (Haubensak, 1992a). The present study was designed to investigate higher order sequential effects, especially concerning two questions. First, is the depth of sequential effects dependent on the task? Judgment processes in magnitude estimation and in category judgment might be different and cause different depths. Second, if events further back are incorporated into the judgment process, do they meet the same function as the immediately preceding event? The article begins with a discussion of models that allow one to derive more specific hypotheses concerning higher order sequential effects, especially interactions of prior events.