more general view that one is conscious of the products of thinking, but not of the thinking itself (Lashley, 1956; Miller, 1962). 1.1. Ironic processing One recent and dramatic example of (purportedly) unconscious processes generating conscious contents is that of ironic processing. In this kind of processing, one is more likely to think about something (e.g., white bears) when instructed to not think about that thing (Wegner, Schneider, Carter, & White, 1987). According to Wegner's (1994) influential model of ironic processing, the ironic effect arises as the result of the interplay between two distinct mechanisms. The first mechanism is an effortful and intentional operating process that searches for conscious contents that will maintain a desired mental state (e.g., to be happy). This operating process tends to be capacity- limited, effortful, and consciously mediated (Wegner, 1994). The second mechanism is an autonomous 'ironic' monitoring process that Scans mental contents to detect contents that signal a failure to maintain the desired mental state. The detection of contents that are incongruent with desired goals is essential for successful cognitive control (Cohen, Dunbar, & McClelland, 1990). If the monitor detects a content that con- flicts with the desired state, it then increases the probability that that particular content will enter consciousness. This then allows the operat- ing process to change its own operations in an appropriate manner when faced with the goal-incongruent content. Important for present purposes, the ironic monitor is not reflected in conscious thought; it is proposed to operate automatically. Although these two mechanisms usually work together in a coordinated fashion, when the current goal is to not activate a particular mental content (e.g., content X), then the interaction between the two processes becomes discordant and leads to the activation of content X in consciousness. This arises because, first, the ironic monitor must reflexively bring into consciousness those mental contents that are incongruent with the current goal (content X), and, second, because the operating process cannot actively exclude contents from entering consciousness. 1.2. Reflexive Imagery Task Recent experimental paradigms (e.g., Allen et al., 2013) have built on this classic research on ironic processing. In most of this current re- search, and consistent with Wegner (1994), entry into consciousness of the undesired content is considered to be automatic and reflex-like. One paradigm is the Reflexive Imagery Task (RIT; Allen et al., 2013), which builds, not only on the work of Wegner (1989), but also on the experiments by Ach (1905) and Gollwitzer (1999). In this task, partici- pants are instructed to not subvocalize (i.e., say in one's head) the names of objects that are presented on a computer screen. Interestingly, participants fail to suppress such subvocalizations on the majority of the trials. We will now present the reader with an example of the task and its effect. Your task is to not subvocalize the name of an object that will be presented somewhere in the next sentence. Here is the stimulus (+). Research involving the RIT reveals that the instruction, in combination with the action set it activates and the presentation of the stimulus ob- ject, renders it difficult for the subject to suppress activations of the pho- nological form of the object name, "bell" (Allen et al., 2013). In more complex variants of the RIT (e.g., Merrick, Farnia, Jantz, Gazzaley, & Morsella, 2015), participants are instructed to (a) not subvocalize the name of visual objects and (b) not count the number of letters compris- ing object names. On a substantial proportion of trials, participants ex- perience both kinds of involuntary thought. Importantly, each thought arises from distinct, high-level processes (i.e., that of object naming ver- sus object counting). Regarding ironic processing, the entry of unintentional contents into consciousness is not fully understood (Wegner & Schneider, 2003). Much seems to be at play in a phenomenon such as the RIT effect. The event is based on several component processes (Allen et al 2013) For example, in order for the effect to arise, there is first the induction o set (e.g., to not subvocalize the name of visual objects). For the set to be induced, an association between the category of stimuli (visua objects) and the response (subvocalizing) must be stored in memory The maintaining in mind of the association seems to require little to no working memory, which has led theorists to describe the delay perioc before the presentation of the stimulus as a form of imageless though (cf., Chen, Jantz, & Morsella, 2014; Woodworth, 1915). During thi period, action sets can influence ongoing behavioral dispositions ever though these sets do not produce any detectable conscious contents Interestingly, once the set is induced, the presentation of the stimulus i enough to trigger the phonological representation (i.e., subvocalization of the object name). Important for present purposes, in the standard RIT, the subvocaliza tions appear to 'pop into mind' unintentionally on a majority of trials Such anecdotal evidence and theorizing suggest that the effect 'just hap pens' to experimental participants and that the effect is not the result o experimental demand or of other, high-level strategic processes on the part of the participant. For example, it appears that the effect does no result from participants consciously thinking in the following manner "I was instructed to not think of the name of the object. The object i X. Therefore, I should not think of X." Instead, consistent with Wegner' model, entry into consciousness of the involuntary subvocalization ap pears to be experienced as immediate and as resulting spontaneously and automatically. Some suggestive evidence for this stems from the observation that the RIT effect survives at comparable rates ever when participants intentionally hold in mind verbal imagery (Cho Godwin, Geisler, & Morsella, 2014). In one condition of Cho et a (2014), participants were instructed to reiteratively subvocalize speech sound ("da, da, da") throughout the trial. This internally generat ed content was self-generated and intentional. Involuntary subvocaliza tions of object names still arose in over 80% of the trials. One could hypothesize that subvocalizations occurred because of the pauses be tween the intended speech sounds, but this is inconsistent with the ob servation that comparable results arose even when participant subvocalized a continuous, unbroken hum ("daaa ...") throughout the trial. In such a dual-task condition, it would have been difficult for par ticipants to carry out strategic processing that would lead to an artifac tual effect. In addition, in various variants of the RIT (e.g., Allen et al. 2013; Cho et al., 2014; Merrick et al., 2015), the latencies of the uninten tional subvocalizations appear, on some trials, to be too short to reflec strategic processing. 1.3. Limitations of extant paradigms Yet, at this stage of scientific inquiry, evidence for the importan claim that the kinds of mechanisms involved in ironic processing an in the RIT are automatic, reflex-like, and not due to high-level, strategi processes is, at best, only anecdotal and suggestive. Strategic processe and experimental demand remain as potential confounds in all research involving ironic processing. In addition, there is no documentation i the literature that participants experience these unintentional con scious contents as being "immediate." Data regarding such perceive immediacy (a dichotomous variable that is inherently phenomenologi cal) would be obtainable only through self-report measures, and, pref erably, on a trial-by-trial basis in which participants introspect righ after each trial whether the conscious content arose immediately when memory about trial events is most accurate (see discussion in Block, 2007). Stronger evidence for the proposal that ironic processes are reflex like would be that the RIT effect is influenced by a stimulus paramete such as word frequency. Such an effect of word frequency is unlikel to result from strategic processing or experimental demand. Regarding the latter, for example, such an artifact of demand would require fo participants to have a theory regarding how word frequency should in fluence responses in an experiment. Stronger evidence would als S. Bhangal et al. / Acta Psychologica 159 (2015) 33-40 consist of showing that, on trials in which involuntary subvocaliza- tions arise, participants often perceive these mental events as being immediate. 1.4. The current approach With these considerations in mind, we developed a version of the RIT in which the variable of word frequency (of the names associated with the visual objects) was manipulated experimentally (leading to double the number of stimuli found in the basic RIT; Allen et al., 2013). In addition, participants reported on each trial whether the effect was experienced as being immediate. Half of the trials consisted of visu- al objects associated with words of low frequency (e.g., KITE, CLOWN, and SCISSORS), and the other half of the trials consisted of visual objects associated with words of high frequency (e.g., DOG, BED, and ROSE). Our primary prediction was that, because high frequency words require a lower threshold of activation (Levelt, 1989), the RIT effect will occur more often in the high frequency condition than in the low frequency condition. This potential effect of word frequency on rates of involun- tary subvocalization would be difficult to explain in terms of just strate- gic processing or experimental demand. Our secondary prediction was that, on a majority of the trials in which the effect arises, the effect will be perceived as being immediate. Last, we predicted that we would replicate the basic RIT effect, with the effect arising in a majority of the trials. For several reasons, the RIT is a fruitful paradigm with which to de- termine if the effects of ironic processing are due to strategic effects/ experimental demand or are better construed as being reflex-like. First, the RIT effect occurs in the majority of subjects and across the ma- jority of experimental trials, which makes the effect reliable and easy to replicate. For example, in Allen et al. (2013), the effect arose in 86% of the trials. Second, the conscious content (i.e., the phonological form) as- sociated with the initial versions of the RIT is well-studied and has well examined properties (Miller, 1996). Third, the paradigm also affords one the opportunity to measure, on a trial-by-trial basis, several aspects of the RIT response (e.g., immediacy and latency). Fourth, the RIT pro- vides a way of examining the mechanisms underlying the phenomenon of entry into consciousness, something that is challenging to tackle ex- perimentally and remains one of the greatest enigmas in science (Crick & Koch, 2003). Last, it is worth mentioning that this is the kind of incremental research that, building on prior research (e.g., Allen et al., 2013; Wegner, 1989) and involving a phenomenon that is robust, multifaceted, and reliable, is important for progress in the field of psy- chological science (Nosek, Spies, & Motyl, 2012). 2. Method 2.1. Participants San Francisco State University students (n = 33, 25 Females, Mage 25.4, SD Age= 9.6) participated for course credit. The involvement of human participants in our project was approved by the Institutional Review Board at San Francisco State University. 2.2. Stimuli and apparatus Stimuli were presented on an Apple iMac computer monitor (50.8 cm) with a viewing distance of approximately 48 cm. Stimulus pre- sentation was controlled by PsyScope software (Cohen, MacWhinney, Flatt, & Provost, 1993). All questions and instructions were presented in black 36-point Chicago font on a white background. Participants were shown a series of 80 well-known objects (Fig. 1, Appendix A) displayed within a visual angle of 5.96 x 6.56 (5 x 5.5 cm), occupying the center of the screen. We selected visual objects that (a) are the kinds of objects that children learn to identify and name aloud in elementary school, (b) had been used successfully Low Frequency Condition 35 High Frequency Condition Fig. 1. Sample stimuli from each of the two conditions. (Not drawn to scale.) in previous research (Morsella & Miozzo, 2002; Snodgrass & Vanderwart, 1980), (c) resemble the visual objects used in previous RIT-based studies, and, critically for the low frequency stimuli, (d) yield high 'name agreement.' According to Snodgrass and Vanderwart (1980), the percentage of their sample of participants that could correctly identify our low frequency items is high (89.03%, SE = 2.53). Omitted from this analysis are four of our low frequency visual objects (BANDAGE, CRIB, SKULL, and WITCH). Importantly, the findings presented below are the same with or without these four items, tBy-Item(74) = 2.92, p = .0046; [By-Subject(32) = 3.05, p = .0046. With our present sample of participants, we identified gross misidentifications of the visual objects (e.g., responding "pasta" for the object HAIR and "filer" for the object BANDAGE) in only 76 (2.9%) of the 2640 trials (80 trials for each of the 33 participants). When removing from this analysis the data from a participant who misidentified 18 of the objects (a rate of misidentification that was over 4.6 SDs above the general misidentification rate of our sample [M 2.30, SE= 0.59]), the number of misidentifications was lower, 58 (2.3%) out of 2560 trials. Two of our visual objects (TOP and WELL), though depicted very well and having suitable levels of name agreement, are associated with phonological forms that have homo- phones possessing different word frequencies. As mentioned in the Re- sults section, the same pattern of results was obtained with or without the inclusion of the data associated with these two stimuli. If we had asked participants, after each trial, whether they could identify and name the object that had appeared during the trial (regardless of whether they had experienced the relevant phonological imagery), then this would have increased the experimental demand to think of the object names on the subsequent trials, thereby invalidating our pri- mary finding. 36 S. Bhangal et al. / Acta Psychologica 159 (2015) 33-40 We strove to have visual objects from various semantic categories (e.g., animals, artifacts, and food) and to not have too many stimuli from one particular semantic category. The 80 visual objects were split into two word frequency groups. According to the Kuera and Francis (1967) word frequency corpus, the mean frequency for the low fre- quency words was 2.15 per million (SD = 1.97, range = 0-7) and that of the high frequency words was 179.18 per million (SD = 160.37, range = 11-897). Similar conclusions were found with the SUBTLEX us Word Frequency Database (Brysbaert & New, 2009). The 40 low frequency objects had a word frequency range of 0.14 to 27.65 per million, and an average of 6.29 per million (SE = 0.98). The 40 high frequency objects had a range of 10.92 to 2990.65, and an average of 234.46 (SE = 74.21). In a by-item analysis, in which visual object served as the unit of analysis, the mean gross misidentication rates by our participants were comparable across the frequency conditions: MLow Frequency = .04 (SE = .01); MHigh Frequency = .02 (SE = .01), t(78) 1.42, p = .16. After removing the misidentification data from the participant with the excessive misidentification rates (see above), the misidentification rates were even more comparable: MLow Frequency=.03 (SE = .01); MHigh Frequency = .02 (SE = .004), t(78)=0.83, p= 41. Incorrect naming of the stimuli could arise in par- ticipants who are non-native speakers of English. We investigated this possibility with our funneled debriefing questions (see below). 2.3. Procedures Instructions were presented on the computer screen, which in- formed participants that they would be shown a series of objects and that their task was to not think of the name of the objects. Participants were instructed to, upon being presented with an object, press the spacebar if they happened to think of the name of the object and as soon as they thought of the name. Participants were instructed to keep their eyes focused on the center of the screen at all times and to keep their left hand rested on the space bar at all times. If participants did not happen to think of the name of the object that was presented, they did not respond in any way. It was emphasized to participants to press the space bar as quickly as possible as soon as they happened to think of the name of the object. Participants were informed that objects would remain on the screen for a fixed period of time, whether or not they pressed the space bar. Specifically, each trial went as follows. First, the phrase "Do Not Think of the Name of the Object" was displayed in the center of the screen, serving as a ready prompt; participants indicated their readiness by pressing the return key. Once participants indicated their readiness, a fixation-cross (+) appeared in the center of the screen (700 ms), pre- paring participants for the presentation of the stimulus. Following fixa- tion, an object appeared for 4 s, during which time participants could indicate, by pressing the space bar, if they happened to think of the name of the object. Participants were instructed to press the spacebar once, and only if they happened to think of the name of the object. After each trial, participants were presented with two questions: "If you thought of the name of the object, please type the name that came to mind" and "If you thought of the name of the object, did the name come to mind immediately?" In response to the first question, participants had the opportunity to input by keyboard the name that came to mind. In response to the second question, participants indicated their response by typing "y" for yes and "n" for no. Again, if we had asked participants, after each trial, whether they could identify and name the object that appeared during the trial, then this would have increased the experimental demand to think of the object names on the subse- quent trials, thereby invalidating our primary finding. Participants first completed a practice trial in which they were instructed to not think of the name of the object. The stimulus used in the practice trial was the same for all participants and was not a stimu- lus used during the critical, experimental trials. During the practice trial, the experimenter observed the participant and verified that he or she was following instructions. Once participants completed the experiment, they responded to a series of funneled debriefing questions (following the procedures of Bargh & Chartrand, 2000), which included general questions to assess whether participants (a) were aware of the purpose of the study, (b) had any strategies for completing the task, and (c) had anything in- terfere with their performance on the task. Importantly, we also asked participants about whether or not they used a strategy when trying to perform well on the task, when trying to not think of the name of the object. Because the study included participants who were non-native speakers of English, we also included a series of questions to assess whether participants (a) thought of the name of the object in a language other than English, (b) pressed the space bar in such a situation, and (c) had a strategy for completing the task if they happened to think of the name of the object in more than one language. The data from two participants were excluded from analysis because the participants did not follow instructions. Both participants did not press the spacebar on any of the trials, despite indicating that they did experience the relevant mental imagery. 3. Results 3.1. Primary dependent measures Out of the 80 trials per participant, the mean proportion of trials in which involuntary subvocalizations arose was .70 (SE = .05, range = .04-1), a proportion that was significantly different from zero, t(32) = 14.48, p < .001. The same result was found with arcsine trans- formations of the proportion data, t(32) = 16.97, p < .001. (Arcsine transformations are often used to statistically normalize data that are in the form of proportions.) With the present sample of participants and stimuli, we replicated the RIT effect successfully. For the 40 Low Frequency trials, the mean proportion of trials in which involuntary subvocalizations arose was .67 (SE = .05); for the High Frequency trials, it was .74 (SE = .05), a difference that was statis- tically significant, t(32) = 2.81, p = .008 (n = 20). (Each of these two proportions was also significantly different from zero, ts > 12.94, ps S. Bhangal et al. / Acta Psychologica 159 (2015) 33-40 obtained in a by-item analysis, in which visual object served as the unit of analysis, t(79) =37.62, p 38 S. Bhangal et al. / Acta Psychologica 159 (2015) 33-40 are highly likely to enter consciousness do so more quickly than object names that are less likely to enter consciousness. Together, these find- ings are difficult to explain as reflecting artifacts from experimental demand or from other kinds of high-level, strategic processes. In line with our secondary hypothesis, and consistent with the view that these ironic processes are automatic and reflex-like, the unintentional subvocalizations experienced by participants were perceived as being immediate on a majority (75%) of the trials. Last, with the current sample and stimuli (which in number was twice as many as that of the original RIT study; Allen et al., 2013), we replicated the RIT effect, with involuntary subvocalizations occurring on a majority (70%) of the trials. 4.1. Limitations One limitation of the present study is that the effects are based on the experimental techniques of introspection and self-report. These techniques have well known limitations. For example, trial-by-trial self-reports might be inaccurate because participants may forget what actually occurred on a given trial (Block, 2007) or because participants base their responses on heuristics or on beliefs about how participants should conduct themselves in an experiment concerning psychophysics (see discussion of this limitation in Morsella et al., 2009). Some of these limitations associated with self-report and introspection could be remedied to some extent in future studies that couple our dependent measures with those of neuroimaging, which could provide objective indices regarding whether, and when, unintentional subvocalizations transpire within a trial. It is worth mentioning that, regarding the valid- ity of the RIT effect, evidence from various sources, including neuroim- aging studies, corroborate that, in paradigms in which participants must report about the experience of a particular conscious content, it is un- likely that participants confabulate about the occurrence of these men- tal events (Mason et al., 2007; McVay & Kane, 2010; Mitchell et al., 2007; Wyland, Kelley, Macrae, Gordon, & Heatherton, 2003). Future studies could also investigate whether, given that ironic processes may be reflex-like, the RIT effect could be thwarted through the tech- nique of habituation. 4.2. Theoretical considerations Although the focus of the present experimental project is on the manner in which high-level conscious contents can be controlled exter- nally by the combination of instructions (e.g., to not perform some mental act) and the presentation of stimuli (e.g., visual objects), it is im- portant to appreciate that the RIT effect is dependent in large part on the activation of high-level processes that are internal in nature. These internal processes include, at the least, action sets (e.g., for naming/ subvocalizing) and, perhaps, the activities of the operating and monitoring processes described in Wegner's (1994) model. Thus, the RIT effect, though initiated by external control, arises from the internal dynamics of an interesting and peculiar system, the conscious mind. Regarding the conscious mind, our findings are consistent with Baars's (1988) Global Workspace Theory, in which disparate conscious contents are integrated in the conscious field. In this model, the field is conceptualized as a kind of workspace. In the conscious field, the con- tents are available 'globally' to a plethora of brain systems/processes. In this way, consciousness contributes to cognition and behavior by inte- grating information-processing structures that would otherwise be in- dependent. Pertinent to the RIT effect, in Baars's model (1988; see also Baars & McGovern, 1996), one of the functions of the conscious field is to represent errors. According to Baars and McGovern (1996), error de- tection is largely an unconscious process. When there is an error in ac- tion control, in the obtaining of some goal, or in some other context, the error is detected by unconscious rule systems, which then force the error into consciousness. Hence, the error can be dealt with by the many systems/processes that respond to the conscious contents in the workspace (see similar proposal in Norman & Shallice, 1980). This view complements nicely Wegner's (1994) account of the interplay be- tween the (unconscious) monitoring process and the (conscious) oper- ating process. From the standpoint of Baars (1988), the RIT effect arises in part because the unintentional subvocalization is itself an error. For the error to be dealt with properly by the cognitive apparatus, the error must be represented in the conscious field. It is clear that error detection is associated with conscious process- ing, because errors, once detected by certain mechanisms, are repre- sented as conscious contents. With this in mind, it is important to appreciate two other ideas pertaining to the liaison between conscious- ness and errors/discrepancies. First, it has been proposed that conscious contents, because of their very nature, can never be ambiguous at one moment in time (Merker, 2012). (Ambiguous stimuli and binocular rivalry are ambiguous only across time; Merker, 2012.) Second, it has been hypothesized that the mind as a whole appears to be a dy- namic, affinity-based system (Bindra, 1959; Chomsky, 1988; Varela, Thompson, & Rosch, 1991) that "prefers" discrepancy-free (or conflict- free) over conflict-ridden situations. This may be because the latter is energy-inefficient (Bargh & Morsella, 2008). For example, action con- flicts require energy and resources to do both x and to counteract doing x (Lorenz, 1963). Because the subjective cost (Morsella, 2005) from conflict is inherently unpleasant (Dreisbach & Fischer, 2012), even approach-approach conflicts are to be avoided (Lewin, 1935). By "blindly" avoiding the subjective cost of conflict/discrepancy, a form of harmony is incidentally achieved among component systems. For ex- ample, when carrying hot objects, the organism wears oven gloves. In this way, the organism becomes more adaptable, suffering fewer depri- vations and less tissue damage (Dempsey, 1951). The foregoing reveals that the mind can be construed as a dynamic system that, because of its affinities toward certain states over others (e.g., discrepancy-free states over conflict-ridden states), has the ability to self-organize in adaptive ways in ontogeny (Dempsey, 1951; Varela et al., 1991). According to Wegner (1994) and Baars (1988), the error detection at play in the RIT effect is usually, in the grand scheme of ontogeny, an adaptive process. 4.3. Encapsulation and action options The observation that stimulus-triggered subvocalization arose despite participants' intentions is consistent with theorizing about the 'encapsulated' nature of the generation of conscious contents (Fodor, 1983; Krisst, Montemayor, & Morsella, in press). (Perceptual processes giving rise to illusions are often said to be encapsulated, because one's knowledge about the true nature of the perceptual stimuli cannot 'turn off the illusion.) For example, from the standpoint of Krisst et al. (in press), it would be maladaptive for the generation of conscious con- tents to be controlled completely by one's beliefs or desires (see also Firestone & Scholl, 2014; Pylyshyn, 1984). Because of this, encapsulation is part of the architecture of the cognitive apparatus. In most cases, this architecture, which includes encapsulation, is adaptive. According to Morsella (2005), action-related urges are often encapsulated. For exam- ple, when holding one's breath while underwater, one cannot avoid the conscious inclination to inhale (Morsella, 2005). Thus, action-related conscious contents triggered by stimulus environments cannot be weakened or eliminated by one's desires, even when doing so would be adaptive (Morsella, 2005; hman & Mineka, 2001). Thus, although inclinations triggered by external stimuli are behaviorally suppressible, they are not mentally suppressible (Bargh & Morsella, 2008; Morsella, 2005). One can think of many instances in which externally triggered conscious contents are more difficult to control than is overt behavior. In the RIT, for example, it is more difficult for participants to suppress subvocalizing than to suppress uttering the object name aloud (discussed in Allen et al., 2013). The present finding could also be con- strued as consistent with theoretical approaches (Krisst et al., in press) that regard conscious contents as 'action options' that, though activated iProspecesible- ELSEVIER Acta Psychologica 159 (2015) 33-40 Contents lists available at ScienceDirect Acta Psychologica journal homepage: www.elsevier.com/locate/actpsy Ironic effects as reflexive responses: Evidence from word frequency effects on involuntary subvocalizations Sabrina Bhangala, Christina Merrick a, Ezequiel Morsella a,b,* b * Department of Psychology, San Francisco State University, United States Department of Neurology, University of California, San Francisco, United States acta psychologica CrossMark ARTICLE Article history: INFO Received 29 January 2015 Received in revised form 28 April 2015 Accepted 30 April 2015 Available online 22 May 2015 PsycInfo classification: Cognitive processes (2340) Attention (2346) Consciousness states (2380) Keywords: Ironic processes Consciousness Mind wandering Cognitive control Involuntary processing ABSTRACT In ironic processing, one is more likely to think about something (e.g., white bears) when instructed to not think about that thing. To further investigate this phenomenon involving cognitive control, in the Reflexive Imagery Task (RIT), participants are instructed to not subvocalize the names of visual objects. On the majority of the trials, par- ticipants fail to suppress such subvocalizations. This finding supports theonzing that conscious thoughts can be triggered by external stimuli in a manner that is nontrivial, involuntary, and, importantly, reflex-like. These conclusions challenge intuitions that consciousness is unpredictable, whimsical, and somewhat insulated from external control. Perhaps these thoughts arise, not in a reflex-like manner, but from experimental demand or other high-level, strategic processes. This prevalent criticism would be inconsistent with the observation that the RIT effect is influenced by a stimulus parameter such as word frequency. Regarding demand characteristics, such an artifact would require participants to have a theory regarding how word frequency should influence responses. We introduce evidence that stimuli associated with high frequency names are more likely to yield involuntary subvocalizations than stimuli associated with low frequency names. These theoretically-relevant data suggest that ironic effects in paradigms such as the RIT resemble reflex-like processes 2015 Elsevier B.V. All rights reserved. 1. Introduction Unlike reflexes, which are susceptible to external control, conscious thoughts appear to be somewhat insulated from the influence of the ex- ternal world. While attending a lecture, for example, one's mind might be occupied by thoughts, not about the content of the lecture (despite the intentions and best efforts of the lecturer), but, say, about a future vacation or a childhood memory. These thoughts are private, un- inferable to those around one. Oftentimes, they appear to arise out of the blue unpredictable even to their creator. Thus, the conscious mind has been construed as having the capacity to be 'offline,' unpre- dictable, and shielded from external influence (Barron, Riby, Greer, & Smallwood, 2011; Fodor, 1975; Fodor, 1983; Smallwood & Schooler, 2006; Wegner & Bargh, 1998). Indeed, the notion that one's conscious thoughts could be controlled externally in any meaningful way has been the topic only of science fiction. In short, the conscious mind has been likened to an irrepressible, mercurial stream, one that, unlike re- flexes, can be free from the reins of external influence (James, 1890). This view about consciousness stems from casual observations, Corresponding author at: Department of Psychology, San Francisco State University, 1600 Holloway Avenue, EP 301, San Francisco, CA 94132-4168, United States. E-mail address: morsella@sfsu.edu (E. Morsella). theorizing, and intuitions about the mind (see review in Allen, Wilkins, Gazzaley, & Morsella, 2013). Despite these views, some theorists (e.g., Freud, 1938; Helmholtz, 1956; James, 1890; Miller, 1959; Vygotsky, 1962; Wegner, 1989) have proposed that, under the appropriate, controlled circumstances con- scious thoughts are more predictable and reflex-like than what might appear to be the case at first glance and when considering what occurs in everyday, uncontrolled environments (Allen et al., 2013). It is worth mentioning that the basic reflex, too, requires certain conditions for its generation. For instance, the patellar reflex can fail to arise if, for exam- ple, it is actively counteracted by the subject (e.g., by contracting the leg muscles). Regarding the reflex-like generation of conscious thoughts (henceforth, 'conscious contents'), Helmholtz (1856) was intrigued by the fact that the written word could automatically trigger in an observer's consciousness the phonological representation of the word, regardless of the observer's intentions. The orthograph is a visual stimulus, but the phonological form is based instead on auditory pro- cessing (Levelt, 1989). That the former triggers the latter requires vast, sophisticated processing, most of which is consciously-impenetrable. Helmholtz (1956) referred to this process by which sophisticated un- conscious mechanisms generate conscious contents as an unconscious inference. Such inferences are evident in perceptual processes and sensory illusions. Helmholtz's (1956) reasoning is consistent with the http://dx.doi.org/10.1016/j.actpsy.2015.04.006 0001-6918/ 2015 Elsevier B.V. All rights reserved.