Neuropsychologia 47 (2009) 3125–3133 Contents lists available at ScienceDirect Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia What’s behind an inference? An EEG study with conditional arguments Mathilde Bonnefond, Jean-Baptiste Van der Henst ∗ Laboratoire Langage, Cerveau et Cognition (L2C2), Institut des Sciences Cognitives - CNRS, Université de Lyon, 67, Boulevard Pinel, 69675 Bron Cedex, France article info Article history: Received 19 August 2008 Received in revised form 17 July 2009 Accepted 21 July 2009 Available online 28 July 2009 Keywords: Inference Conditional reasoning N2 P3b PSW CNV abstract Conditional reasoning studies typically involve presenting a major conditional premise (If P then Q), a minor premise (P) and a conclusion (Q). We describe how most fMRI studies investigate reasoning and point out that these studies neglect to take into consideration the temporal sequence of cognitive steps generated by the interaction of the premises. The present study uses EEG to address this issue and com- pares the processing of the minor premise P when it is presented before vs. after the conditional statement (P ; If P then Q vs. If P then Q; P ). When the minor premise comes after the conditional statement and matches the antecedent its processing results in a P3b component, known to reflect the satisfaction of expecta- tions, and in two later components, a PSW component and a CNV component. These two components are discussed in light of a conclusion generation phase and a maintenance phase. We also investigated the effect of violating expectations through the presentation of a minor premise that mismatches the antecedent of the conditional statement (If P then Q; R ). The data indicate that the processing of such a premise yields an N2 component which is known to reflect perceptual conflict. © 2009 Elsevier Ltd. All rights reserved. The study of reasoning is often reduced to inference making and, most often, of the logical kind. Consider a prototypical example, Modus Ponens, in (1) below: (1) If Menachem goes to Camp David then Anwar goes to Jerusalem. (If P then Q) Menachem goes to Camp David. (P) Therefore Anwar goes to Jerusalem. 1 (Q) The conclusion Anwar goes to Jerusalem derives from the inter- action of the two premises. In other words, it cannot be drawn from an individual premise. The present study uses EEG technique as a means to investigate premise interaction while participants carry out such basic logical inferences. In what follows, we describe why an investigation of interac- tions among premises is called for. Our Introduction is divided into four parts. First, we present a brief historical background to the study of reasoning in order to show how the issue of interaction was addressed in the past. Second, we show how the advent of neuroimaging has allowed the field to distinguish between dif- ferent cognitive processes and we present some limitations of functional magnetic resonance imaging (fMRI) studies. In partic- ular we describe how, once individual premises are taken into account, dependent variables garnered from EEG are advantageous for studying human reasoning because its temporal resolution ∗ Corresponding author. E-mail addresses: mbonnefond@isc.cnrs.fr (M. Bonnefond), vanderhenst@isc.cnrs.fr (J.-B. Van der Henst). 1 For the purposes of exposition, we will refer to the conditional premise in (1) as the Major Premise, the second statement of fact, as exemplified in the second line of (1), as the Minor Premise (and, as is clear, the Conclusion speaks for itself). allows one to make hypotheses about the cognitive procedures that are occurring on line. Finally, we present in detail the paper’s experi- mental paradigm and the EEG components that are expected to play a role in the analyses. 1. How does psychology investigate inferences? Most psychological accounts that investigate inferential pro- cesses focus on difficulty and thus compare the performance of different kinds of reasoning problems. For example, one can com- pare the rates of correct answers to Modus Ponens in (1) to equally valid conditional arguments, such as Modus Tollens (If P then Q; non-Q; non-P). Whereas Modus Ponens behavioral data show that 90–100% participants endorse the conclusion in (1), about 60% of participants correctly evaluate Modus Tollens. Both of the main historical approaches to deductive reasoning – the mental logic theory (Braine & O’Brien, 1998) and the mental model theory (Johnson-Laird & Byrne, 1991, Johnson-Laird, 2001)– aimed to account for this sort of finding: according to Mental Logic, difficulty is linked to the number of inference rules needed to reach the conclusion and, according to Mental Model Theory, difficulty is linked to the number of models (i.e. possibilities) a participant is required to consider in order to properly represent the premises and to evaluate the conclusion. Unsurprisingly, when psychologists address the issue of premise interaction, they do it by considering how difficult it is to com- bine different premises. For instance, proponents of mental model theory argue that people are less likely to carry out the Modus Tol- lens inference than they are to make the Modus Ponens inference 0028-3932/$ – see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2009.07.014