Consciousness and Cognition 30 (2014) 220–233

Contents lists available at ScienceDirect

Consciousness and Cognition journal homepage: www.elsevier.com/locate/concog

Control of spatial orienting: Context-specific proportion cued effects in an exogenous spatial cueing task Alex Gough a,⇑, Jesse Garcia a, Maryem Torres-Quesada b, Bruce Milliken a a b

McMaster University, Department of Psychology, Neuroscience and Behaviour, Hamilton, Ontario L8S4K1, Canada Universidad de Granada, Faculty of Psychology, Avda, del Hospicio, s/n, C.P. 18071 Granada, Spain

a r t i c l e

i n f o

Article history: Received 19 August 2013

Keywords: Cognitive control Spatial orienting Context-specific learning

a b s t r a c t Cognitive control refers to the ability to adjust strategy use based on the demands of a current context or task. Recent research using attentional filtering tasks has shown that cognitive control can adapt rapidly and automatically in accord with learning that is specific to particular tasks, items, and contexts (Crump, Gong, & Milliken, 2006; Fernandez-Duque & Knight, 2008; Jacoby, Lindsay, & Hessels, 2003). However, the role of context-specific control has not been investigated in detail in spatial orienting tasks. In a series of three experiments, the proportion of validly cued trials in an exogenous spatial cueing task was manipulated for one context but not for another context, with the two contexts intermixed randomly across trials. The results revealed that spatial/temporal contextual cues in conjunction, but not individually, produced context-specific control over spatial orienting. Ó 2014 Published by Elsevier Inc.

1. Introduction Cognitive control allows us to adjust processing to suit the demands of a current context, which is particularly important when cues to multiple tasks are present. For example, entering data in a spreadsheet requires attention to column headings, cursor position, data on the screen, progress in the task so far, and many other factors. On the other hand, organizing data in a spreadsheet to create a graph requires attention to many of the same spreadsheet features, but requires an entirely different set of processes. To the extent that an individual can adapt effectively to the two different task contexts, they are making effective use of cognitive control processes. The conflict monitoring hypothesis proposed by Botvinick, Braver, Barch, Carter, and Cohen (2001) attempts to model how and when cognitive control processes are implemented. The model suggests that the anterior cingulate cortex (ACC) monitors levels of conflict in information processing, and that this information is used to regulate the influence of centres responsible for goal-driven control. The conflict monitoring hypothesis fits well with a wide range of data, such as increased ACC activation in response to incongruent trials in the flanker task (Botvinick, Nystrom, Fissell, Carter, & Cohen, 1999), the Stroop task (Barch et al., 2001; Peterson et al., 2002), and the Simon task (Peterson et al., 2002), to name just a few. Greater ACC activation is also observed in response to conflict when participants must choose from a number of permissible responses (Barch, Braver, Sabb, & Noll, 2000), as well as following the commission of an error (Carter et al., 1998), when greater control over responses must be exerted to improve future performance. Clearly, a strength of the conflict monitoring model is that it describes a plausible mechanism for the monitoring and implementation of control in a wide array of task contexts. On the other hand, a possible weakness of the model is that it ⇑ Corresponding author. E-mail addresses: [email protected] (A. Gough), [email protected] (J. Garcia), [email protected] (M. Torres-Quesada), [email protected] (B. Milliken). http://dx.doi.org/10.1016/j.concog.2014.09.014 1053-8100/Ó 2014 Published by Elsevier Inc.

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

221

does not suggest a process or mechanism that would allow cognitive control settings to be associated with particular tasks, items, or contexts (see Blais, Robidoux, Risko, & Besner, 2007, for an adaptation of the conflict monitoring model that does implement conflict monitoring at the item-specific level; see also Verguts & Notebaert, 2008 for a discussion of this issue). Indeed, a range of recent studies have reported results that appear related to cognitive control, but that require some form of task-specific, item-specific, or context-specific learning. One of the tools used recently to study specificity in cognitive control is the proportion congruent effect (for a recent review, see Bugg & Crump, 2012). Proportion congruent effects are often measured in tasks that involve distractor conflict, such as the Stroop (Stroop, 1935) and flanker (Eriksen & Eriksen, 1974) tasks. Using the Stroop task as an example, many studies have manipulated proportion congruent between different blocks of trials, or between participants. In these studies, in a high proportion congruent condition, congruent trials (e.g., the word BLUE in blue) are more common than incongruent trials (e.g., the word BLUE in green). In contrast, in a low proportion congruent condition the opposite is true. The result typically observed is that the Stroop effect is larger for the high proportion congruent condition than for the low proportion congruent condition (Logan & Zbrodoff, 1979; Lowe & Mitterer, 1982). This result fits with the view that participants can engage cognitive control processes to adapt the degree to which word reading drives the selection of a response. When congruent trials are likely, word reading is likely to benefit performance, and therefore cognitive control processes allow word reading to affect response selection. In contrast, when congruent trials are unlikely, word reading is likely to interfere with performance, and therefore cognitive control processes prevent word reading from affecting response selection. This view of proportion congruency effects – that the likelihood of conflict leads to adaptive adjustments in the implementation of cognitive control – is consistent with the conflict monitoring hypothesis (Botvinick et al., 2001). However, several recent studies have pointed to the possibility that proportion congruency effects are not necessarily driven by adjustments in cognitive control made in response to global conflict-monitoring. Rather, it seems that proportion congruent effects can be driven by learning processes that are specific to items, contexts, or tasks that are associated with the experienced conflict. In the following section, four studies are summarized, all of which have used the proportion congruent method to tie adjustments in control to item-specific, context-specific, or task-specific learning. 1.1. Specificity in the proportion congruent effect Jacoby, Lindsay, and Hessels (2003) reported what they referred to as an item-specific proportion congruency (ISPC) effect. Participants performed a conventional Stroop task with integrated colour-word stimuli that were either congruent or incongruent, and participants were required to identify the colour in which the words were presented. Across the experimental session there were equal proportions of congruent and incongruent items. However, a proportion congruency manipulation was implemented across different sets of items within the experiment. For example, stimuli involving the colours blue and red (e.g., the word BLUE in blue, the word BLUE in red, the word RED in blue, the word RED in red) might be designated the high proportion congruent item set, while stimuli involving the colours GREEN and YELLOW might be designated the low proportion congruent item set. The key result in this study was that the Stroop effect was significantly larger for the high proportion congruent item set than for the low proportion congruent item set. Although there is some debate as to the nature of the learning processes responsible for the ISPC effect (see Bugg, Jacoby, & Chanani, 2011; Schmidt & Besner, 2008), there is broad agreement that proportion congruent effects are not a pure measure of voluntary shifts in cognitive control that align participants’ strategies with the proportion of congruent items. Rather, learning processes that are specific to particular items, and that differ for high and low proportion congruent contexts, must also contribute to proportion congruency effects. Fernandez-Duque and Knight (2008) reported a result that suggests a degree of task specificity to cognitive control processes. They asked participants to perform two different tasks, one that assessed the conventional colour-word Stroop effect, and the other that assessed a number Stroop effect. In the number Stroop task, participants identified how many stimuli were present in a display, with the stimuli themselves being digits. All of the digits in any given display depicted the same number, and that number was either congruent or incongruent with the number of digits in the display. Fernandez-Duque and Knight intermixed stimuli from the two tasks, and varied the proportion congruent only for the colour-word Stroop task. The key issue was whether the proportion congruent manipulation for the colour-word Stroop task would generalize to the number Stroop task. In fact, it did not; there was a proportion congruent effect for the colour-word Stroop task but none for the number Stroop task. The fact that changes in proportion congruent for a specific Stroop context did not generalize to another Stroop context again provides evidence that cognitive control processes can be tied to relatively specific aspects of processing, in this case to specific variants of the Stroop task. Crump, Gong, and Milliken (2006) reported a finding indicating that even task-irrelevant contextual properties of stimuli contribute to cognitive control. They used a consecutive trial variant of the colour-word Stroop task in which a colour word presented in white is followed by a coloured rectangle. The participants’ task was simply to name the colour of the rectangle. Critically, the target rectangle could appear in either of two locations, above or below fixation. One of these location contexts was associated with a high proportion congruency between the prime word and target colour, and the other location context was associated with a low proportion congruency. The key result was that congruency effects (i.e., faster responses for congruent than incongruent targets) were larger in magnitude for the high proportion congruent location context than for the low proportion congruent context. This result demonstrates that contextual control over congruency effects can involve associations between a task-irrelevant stimulus property (i.e., target location) and proportion congruency.

222

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

Finally, a recent study reported by Funes, Lupiáñez, and Humphreys (2010) examined the specificity of proportion congruency effects to particular types of conflict within the same task. In their study, the target stimulus on each trial was an arrow, pointing up or down, and presented in one of four locations; top, bottom, left or right of fixation. The task in all cases was to indicate the direction the arrow was pointing by pressing a corresponding button on either the left or right side of a keyboard. A critical property of this procedure is that it allows one to measure either Simon interference (Simon & Rudell, 1967; see Lu & Proctor, 1995, for a review) or spatial Stroop interference (Logan & Zbrodoff, 1979), depending on the axis (horizontal or vertical) in which the target appears. In particular, when the target appeared in either the top or bottom location, its location was either congruent or incongruent with the direction in which the arrow pointed, which allowed a spatial Stroop effect to be measured for these targets. In contrast, when the target appeared in either the left or right location, its location was either congruent or incongruent with the hand required to respond to the target, which allowed a Simon effect to be measured on these trials. The key manipulation in this study was to vary the proportion congruent on one axis only (the left–right axis), and to examine whether proportion congruent effects generalized to the other axis. Interestingly, it seems that proportion congruent effects can generalize across the two axes (and thus across the two types of conflict) in some experimental contexts (Funes et al., 2010), but that in other experimental contexts the proportion congruent effect remains specific to the axis on which the proportion congruent manipulation was implemented (Torres-Quesada, Funes, & Lupiáñez, 2013). Thus, proportion congruency effects can be specific to a particular type of conflict even when the stimuli (upward and downward pointing arrows) and tasks (in what direction is the target arrow pointing?) are identical across the two conflict types. 1.2. The present study Clearly, there is broad evidence for specificity of cognitive control in tasks used to tap control over the influence of distractors (e.g., Stroop and flanker tasks), whether that specificity is defined in terms of items (Jacoby et al., 2003), tasks (Fernandez-Duque & Knight, 2008), task-irrelevant stimulus properties (Crump et al., 2006), or the type of conflict that is presumably subject to cognitive control (Funes et al., 2010; Torres-Quesada et al., 2013; see also Blais & Besner, 2006; Verguts & Notebaert, 2008 for broader discussions of this issue). Furthermore, research on the contextual cueing phenomenon has shown that implicit learning processes can produce remarkably specific control over orienting in visual search contexts (e.g., Chun & Jiang, 1998). However, to our knowledge, no prior study has addressed the issue of specificity of control in another task domain used to study spatial orienting, the spatial cueing paradigm. The primary aim of the current study was therefore to examine the specificity of control over orienting processes in a spatial cueing task. The spatial cueing task that we use is a variant of one introduced by Posner and Cohen (1984). In their task, two rectangular placeholders were presented to the left and right of fixation, and a third rectangular placeholder was placed at fixation. Following an initial fixation period one of the peripheral placeholders was briefly brightened. This peripheral cue was then followed by a brightening of the central placeholder to reorient attention to central fixation. Then, following a variable interval, an abrupt onset target appeared with equal likelihood at either of the two peripheral locations. Participants were required to make a response as quickly as possible upon detection of this target. When the temporal gap between the first peripheral event (the cue) and the target was less than approximately 300 ms, participants were faster to detect the target when it appeared at the same location as the cue (a validly cued trial) than when it appeared at the location opposite the cue (an invalidly cued trial). However, when the gap between cue and target exceeded 300 ms, the opposite result was observed – participants were faster to detect the target on invalidly cued trials. This latter effect was termed inhibition of return (IOR; Posner, Rafal, Choate, & Vaughan, 1985). The time course of exogenous spatial cueing effects described above has now been observed in many studies (for a review, see Lupianez, Klein, & Bartolomeo, 2006). Two recent studies have pointed out that control of the processes responsible for exogenous spatial cueing effects is surprisingly unrelated to overt strategic processes that participants can report voluntarily (Bartolomeo, Decaix, & Sieroff, 2007; Risko & Stolz, 2010). In both of these studies, the researchers manipulated the relative proportions of valid and invalid trials, and found that cueing effects varied as one might expect. That is, facilitation effects tended to be larger (or IOR effects smaller) when the proportion of validly cued trials was high. However, this result appeared to be entirely independent of participants’ knowledge of the proportion manipulation. These findings suggest that the processes controlling spatial cueing effects may well be stimulus-driven and subject to associative learning processes that are not open to introspection. As such, there seemed good reason to ask whether proportion valid effects in spatial cueing tasks might be driven by processes that are context-specific, and cued involuntarily by properties of stimuli and task. If such a result were to occur, it would imply that spatial orienting processes may be subject to some of the same stimulus-driven learning and control mechanisms as reported for various other tasks (see Bugg, Jacoby, & Toth, 2008; Cañadas, Rodríguez-Bailón, Milliken, & Lupiáñez, 2012; Heinemann, Kunde, & Kiesel, 2009; Leboe, Wong, Crump, & Stobbe, 2008; Lehle & Hübner, 2008; Wendt & Kiesel, 2011). 2. Experiment 1 In this experiment we asked whether varying the proportion of validly cued trials in an exogenous spatial orienting task would produce a context-specific influence on the cueing effect. To address this issue, two contexts were defined, one above the centre of the screen and one below the centre of the screen. In each of the two contexts, there were two marked locations (i.e., boxes) at which cues and targets could appear, one to the left and the other to the right of centre. The research strategy

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

223

was to manipulate the relative proportions of validly cued and invalidly cued trials in just one of the two contexts (either the top or bottom half of the screen). The key research question was whether the effect of proportion valid on cueing effects would be specific to the context in which proportion valid was manipulated, or alternatively would generalize to the other context in which there was no manipulation of proportion valid. The key statistical result that would support the notion of context-specific control over spatial orienting would be a larger influence of proportion valid on cueing effects in the specific context in which proportion valid was manipulated, than in the context in which proportion valid was not manipulated (i.e., a three-way interaction between proportion valid, cueing, and context). 2.1. Method 2.1.1. Participants Thirty-six undergraduate students at McMaster University received course credit for their participation. All participants had normal or corrected-to-normal vision. 2.1.2. Apparatus and stimuli All stimuli were black presented on the white background of a 17-in. CRT monitor at a screen refresh rate of 60 Hz. The central fixation point was a cross measuring 2° of visual angle. The placeholders were square outlines measuring 3° on each side. The placeholders were located 11.5° to the left and right of the vertical midline of the screen, and were both presented either 7.7° above or 7.7° below the horizontal midline, varying from trial to trial. The cue was a brief thickening of one of the placeholders. The target was the letter ‘A’, 1.5° in height and width. 2.1.3. Procedure Participants were seated in front of the computer monitor at a viewing distance of approximately 60 cm. Instructions for the experiment were presented on-screen and read aloud to the participant. Participants were instructed to focus on the fixation cross throughout each trial. At the beginning of each trial, a fixation cross was displayed for 500 ms either above or below the horizontal midline, centred between the two locations at which the location placeholders would subsequently appear. The placeholders then appeared on either side of the fixation point for 100 ms, followed by a thickening of the border of one of the placeholders, the attentional cue, for 50 ms. After a 150 ms or 800 ms delay (depending on the participant’s assigned SOA condition) the target appeared in one of the placeholders, and remained until response. Participants were allowed a maximum of 4000 ms to respond to the target’s location. If the target appeared on the left side of the screen they were to press the ‘z’ key on the keyboard with their left hand, while if the target appeared on the right, they were to press the ‘/’ key with their right hand. A new trial began after a 1000 ms inter-trial interval. The experiment consisted of two blocks of trials, in each of which participants completed ten practice trials and then eight sub-blocks of 40 trials each. The proportion of validly cued trials was manipulated between blocks. As such, after completing the first block of 320 experimental trials with one proportion of validly cued trials (e.g., high), the participant was given a short break before beginning the second block of 320 experimental trials with the opposite proportion of validly cued trials (e.g., low). Within each sub-block the proportions of validly cued and invalidly cued trials appearing in each of the two contexts (above or below the horizontal midline) was the same as across the entire block. Following the practice trials and after each sub-block of 40 trials the participant was given a short, self-timed break. 2.1.4. Design Participants were assigned to one of two SOA conditions, with the cue-target SOA being 200 ms for one group of participants and 850 ms for the other group of participants. The proportion of validly cued trials was manipulated blocked within-subject, with half of the participants completing the high proportion valid block first and the low proportion valid block second, and the other half of participants completing the blocks in the opposite order. In the high proportion valid condition, the proportion of validly cued trials was .90 while the proportion of invalidly cued trials was .10. In the low proportion valid condition the opposite proportions were used. Proportion valid was manipulated for just one of the two location contexts (above or below the horizontal midline), which we called the biased context. The location context at which proportion cued was not manipulated was called the unbiased context. Which of the two contexts was the biased context and which was unbiased was counterbalanced across participants. To summarize, the experiment had five independent variables, each with two levels. Cueing (valid/invalid) varied randomly from trial to trial. On validly cued trials, the cue and target appeared in the same location, whereas on invalidly cued trials the cue and target appeared in different locations. Location context (biased/unbiased) also varied randomly across trials. In the biased condition, the trial was presented in the context (above or below the horizontal midline) in which the proportion valid was manipulated, whereas in the unbiased condition the trial was presented in the context in which the proportion valid was equal and constant across the two blocks. Proportion valid (high/low) was manipulated within-subjects between blocks of the experiment. In the high proportion valid condition, the proportion of validly cued trials was .90 in the biased context and .50 in the unbiased context. In the low proportion valid condition, the proportion of validly cued trials was .10 in the biased context and .50 in the unbiased context. As noted above, SOA (200 ms/850 ms) was manipulated between groups, and the location context (above or below the horizontal midline) that served as the biased context was

224

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

Table 1 Mean response times in milliseconds for Experiment 1, collapsed across participants. 200 ms SOA

850 ms SOA

Biased P(Valid)

Valid (V) Invalid (I) Cueing effect (I–V) P(Valid) effect

Unbiased P(Valid)

Biased P(Valid)

Unbiased P(Valid)

High

Low

High

Low

High

Low

High

Low

263 327 64

298 324 26

268 339 72

294 323 28

326 327 0

326 314 12

332 330 2

333 316 18

38

43

12

15

Note. The results of note were as follows: (1) As commonly seen in spatial cueing studies, cueing effects were larger (more positive) for the 200 ms SOA than for the 850 ms SOA; (2) cueing effects were larger for the high proportion valid than for the low proportion valid condition; and (3) this difference in cueing effects across the proportion valid conditions was as large for the unbiased context (i.e., the location at which proportion valid was not manipulated) as for the biased context (i.e., the location at which proportion valid was manipulated).

counterbalanced across participants. This counterbalancing variable was analyzed initially but produced no significant main effect and no significant interactions, and so it was excluded from the analyses reported here.1 2.1.5. Post-experiment awareness questionnaire After completing both blocks of the experiment, participants completed a questionnaire that queried their awareness of the proportion of validly cued and invalidly cued trials in each location context in each half of the experiment. Each participant was presented with a form on which a diagram of the experimental display appeared. Question 1 showed the two placeholders and fixation point on the top of the display, while Question 2 showed them on the bottom of the display. Beginning with the block the participant had just completed, participants were asked to estimate the probability associated with a target appearing in the same location as the cue for each of the two location contexts. After recording these estimates, participants were presented an identical form and asked to perform the same task based on their experience in the first block of the experiment. 2.2. Results Response times were initially submitted to an outlier exclusion procedure that ensured that different numbers of observations were not systematically excluded from cells of different sizes (the non-recursive with moving criterion procedure of Van Selst & Jolicoeur, 1994). This analysis removed 2.8% of the response times from subsequent analyses. Mean RTs were computed from the remaining observations, and these mean RTs and corresponding error rates were submitted to mixedfactor ANOVAs that treated cueing (valid/invalid), proportion valid (.90/.10), and location context (biased/unbiased) as within-subject variables, and SOA (200 ms/850 ms) as a between-subjects variable. The alpha level was set at .05 for all statistical effects. Mean RTs, collapsed across participants, are displayed in Table 1. The analysis of RTs revealed a significant two-way interaction between cueing and SOA, F(1, 34) = 43.80, MSe = 1256.28, p < .001, g2p = .39, as well as a significant three-way interaction between proportion valid, cueing and SOA, F(1, 34) = 8.77, p = .006, g2p = .11. To examine these interactions further, separate repeated measures ANOVAs were conducted for the two SOA conditions, each of which treated cueing, proportion cued, and location context as within-subject factors. For the 200 ms SOA condition, there was a significant main effect of cueing, F(1, 17) = 57.18, p < .001, g2p = .63. Responses were 47 ms faster for valid trials than for invalid trials. This benefit for valid trials is the result commonly observed for short SOAs in studies of exogenous spatial cueing. There was also a significant interaction between cueing and proportion valid, F(1, 17) = 29.39, p < .001, g2p = .46. Responses were 68 ms faster for valid than for invalid trials in the .90 proportion valid condition, t(17) = 10.37, p < .001, while responses were 27 ms faster for valid than for invalid trials in the .10 proportion valid condition, t(17) = 3.37, p = .004. For the 850 ms SOA condition, responses were 8 ms slower for valid trials than for invalid trials, although this difference was not statistically significant, p = .17, g2p = .06. This trend toward a cost for valid trials (i.e., IOR) is in line with the result commonly observed for long cue-target SOAs in studies of exogenous spatial cueing (Posner et al., 1985). The interaction between cueing and proportion valid was also significant, F(1, 17) = 7.48, p = .01, g2p = .18. Responses were 1 ms slower for valid than for invalid trials in the .90 proportion valid condition, t(17) = 0.16, p = .87, while responses were 15 ms slower for valid than for invalid trials in the .10 proportion valid condition, t(17) = 2.47, p = .02. Together, the results reveal opposite cueing effects for the 200 and 850 ms SOAs, thus accounting for the cueing by SOA interaction. The results also reveal similar proportion valid by cueing interactions for the two SOA conditions, and therefore the significant 3-way interaction owes to this interaction simply being larger in magnitude for the 200 ms SOA condition than for the 850 ms SOA condition.

1 This experiment was actually run as two separate experiments with one experiment using a 200 ms SOA and the other using a 850 ms SOA. The two experiments are treated as one here to produce a more concise description for the reader.

225

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

Finally, returning to the overall analysis, the most important result was the clearly non-significant three-way interaction between cueing, proportion valid, and context, F < 1, g2p = .003. This result implies that although the cueing effect was modulated by the proportion valid manipulation, this modulation was not specific to the location context at which proportion valid was manipulated. Rather, this effect generalized to both of the location contexts. Mean error rates, collapsed across participants, are presented in Table 2. There were no significant effects in the analysis of error rates. 2.2.1. Awareness data Participants’ estimates of proportion valid, collected in the post-experiment awareness questionnaires described above, were submitted to a 2  2  2 repeated measures ANOVA that treated proportion valid (.90/.10), spatial context (biased/ unbiased) and SOA (200 ms/850 ms) as factors. Mean estimates of proportion valid, collapsed across participants, are displayed in Table 3. There were no significant effects in the analysis, indicating that participants’ estimates of proportion valid did not vary systematically across conditions. To examine whether participants’ estimates in each condition differed from chance, estimates of proportion valid in each condition were subtracted from chance level (.5), and these differences were submitted to one-sample t-tests. None of these comparisons were significant, even with an uncorrected-for-multiple-comparisons alpha level of .05. Taken together, the analyses of the awareness data offer no support for the idea that participants were consciously aware of the proportion valid manipulation. Participants seemed unaware that proportion valid varied across blocks for one of the two contexts, and indeed seemed unaware that proportion valid varied across blocks at all. 2.3. Discussion The results of Experiment 1 revealed three useful findings. First, performance in the localization task used here produced a pattern of data that was similar to that observed in many other exogenous spatial cueing studies (Posner & Cohen, 1984; see Klein, 2000 for a review). In particular, responses were faster for validly cued than for invalidly cued trials with a short cue-target SOA, while the opposite result was observed for a longer cue-target SOA. Second, the cueing effects for both short and long SOAs were sensitive to the proportion of validly cued trials, with larger facilitation effects (or smaller IOR effects) in the high proportion valid condition. Third, and most important, this modulation of cueing effects by proportion valid was entirely insensitive to the location context manipulation. That is, the proportion valid by cueing interaction was no different in size for the biased location context (at which the proportion valid manipulation was implemented) than for the unbiased Table 2 Mean error rates for Experiments 1–3, collapsed across participants. 200 ms SOA

850 ms SOA

Biased P(Valid)

Unbiased P(Valid)

Biased P(Valid)

Unbiased P(Valid)

High

Low

High

Low

High

Low

High

Low

Experiment 1

Valid Invalid

0.00 0.06

0.00 0.03

0.00 0.07

0.01 0.03

0.00 0.01

0.01 0.00

0.00 0.01

0.00 0.00

Experiment 2

Valid Invalid

0.00 0.02

0.00 0.01

0.01 0.00

0.00 0.00

0.00 0.02

0.00 0.01

0.00 0.02

0.01 0.02

Experiment 3

Valid Invalid

0.00 0.02

0.01 0.01

0.01 0.00

0.00 0.00

0.00 0.01

0.01 0.00

0.00 0.02

0.00 0.01

Note. Figures given are as a proportion of all trials. Table 3 Participants’ mean estimates of proportion valid (%), collapsed across participants, for Experiments 1–3. Location context

P(Valid) High

Low

Experiment 1 (Collapsed across SOAs)

Biased Unbiased

51.5 53.6

47.4 45.3

Experiment 1 (Short SOA)

Biased Unbiased

53.9 53.5

48.6 43.8

Experiment 1 (Long SOA)

Biased Unbiased

49.1 53.8

46.3 47.8

Experiment 2

Biased Unbiased

50.8 50.8

50.4 54.6

Experiment 3

Biased Unbiased

55.3 53.5

51.5 51.8

Note. There were no significant effects in the analyses of these data. Taken together, these data suggest that participants were not consciously aware of the proportion valid manipulation at a given context within a block, or indeed the overall change in proportion valid across blocks.

226

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

location context (at which there was no proportion valid manipulation). This result suggests that the learning and control processes responsible for the modulation of cueing effects by proportion valid were not specific to spatial context. In the next experiment, we asked whether a manipulation of temporal context might reveal context-specific control over the processes responsible for spatial cueing effects. 3. Experiment 2 In Experiment 1, a proportion valid manipulation was applied to one of two spatial contexts. The critical research question was whether the effect of proportion valid on cueing effects would be specific to the context in which proportion valid was manipulated, or alternatively would generalize to the other context in which there was no manipulation of proportion valid. Experiment 2 followed the same logic, but in this case context was defined temporally rather than spatially. In particular, all stimuli were presented on the horizontal midline of the display, and proportion valid was manipulated selectively for one of two cue-target SOAs that were intermixed randomly across trials. 3.1. Method 3.1.1. Participants Eighteen undergraduate students at McMaster University received course credit for their participation. All participants had normal or corrected-to-normal vision. 3.1.2. Apparatus, stimuli and procedure Stimuli were identical to Experiment 1 with one exception. The location context variable from Experiment 1 was eliminated from the current procedure and all stimuli were presented on the horizontal midline of the screen. 3.1.3. Design The design of Experiment 2 was similar to that of Experiment 1, with the exception that proportion valid was manipulated separately for two temporal contexts (200 ms and 850 ms SOAs) rather than two spatial contexts. As in Experiment 1, the two contexts were intermixed randomly from trial to trial. Whereas in Experiment 1 the location context (above or below the horizontal midline) at which proportion valid was manipulated (i.e., the biased location context) was counterbalanced across participants, here the temporal context (short or long SOA) at which proportion valid was manipulated was counterbalanced across participants. So for one group of participants, the proportion of validly cued trials was .90 in one block and .10 in the other block for the 200 ms SOA condition only. For this group, in both blocks, the proportion cued was .50 for the 850 ms SOA condition. For the other group of participants, the proportion of validly cued trials was .50 in both blocks for the 200 ms SOA condition, but varied as a function of block (.90/.10) in the 850 ms SOA condition. 3.1.4. Post-experiment awareness questionnaire Immediately after the experiment, participants completed questionnaires identical to those described in Experiment 1, except that instead of the placeholders being presented on the top or bottom of the experimental display diagram, they were presented in the middle of the diagram. After an initial question asking whether or not participants had noticed a difference in SOA between trials (all participants indicated that they were aware that different SOAs were presented), Question 1 asked participants to estimate the probability that the target appeared in the same location as the preceding cue when the time between presentation of the cue and target was short, and Question 2 asked participants to estimate the probability that the target appeared in the same location as the preceding cue when the time between presentation of the cue and target was long. 3.2. Results Response times were submitted to the same outlier exclusion procedure as in Experiment 1 (Van Selst & Jolicoeur, 1994), which resulted in the removal of 2.9% of the RTs from subsequent analyses. Mean RTs for each condition were then computed from the remaining observations. These mean RTs, collapsed across participants, are displayed in Table 4. Whether a proportion valid manipulation implemented on short SOA trials impacted cueing effects on long SOA trials could not be established by comparing cueing effects for short and long SOAs within group, as cueing effects for short and long SOAs are expected to differ a priori, without a proportion valid manipulation. As such, our first pair of analyses looked at performance at a particular SOA, either short or long, across the two groups. In each of these analyses, proportion valid was manipulated (the biased context) for one group and not manipulated (the unbiased context) for the other group. As such, each of these analyses treated cueing (valid/invalid) and proportion valid (high/low) as within-subject variables, and temporal context (biased/unbiased) as a between-subject variable. After establishing that the usual pattern of spatial cueing effects in these preliminary analyses was observed, the data were then collapsed together to provide the most sensitive analysis possible of whether proportion valid effects were specific to the temporal context for which proportion valid was manipulated. For the 200 ms SOA condition, there was a significant main effect of cueing, F(1, 16) = 8.85, MSe = 637.26, p = .008, g2p = .36. Valid trials were responded to 18 ms faster than invalid trials. The interaction between cueing and proportion valid was not

227

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233 Table 4 Mean response times in milliseconds for Experiment 2, collapsed across participants. 200 ms SOA

850 ms SOA

Biased P(Valid)

Valid (V) Invalid (I) Cueing effect (I–V) P(Valid) effect

Unbiased P(Valid)

Biased P(Valid)

Unbiased P(Valid)

High

Low

High

Low

High

Low

High

Low

342 368

353 363

350 373

354 366

343 347

350 334

328 324

335 324

11

22

12

4

16

3

26 16

11

20

12 8

Note. As in Experiment 1, the results of note were: (1) Cueing effects were larger for the high proportion valid than for the low proportion valid condition; and (2) this difference in cueing effects across the proportion valid conditions was as large for the unbiased temporal context (i.e., the SOA at which proportion valid was not manipulated) as for the biased temporal context (i.e., the SOA at which proportion valid was manipulated). Note that for a given participant one of the two SOAs was biased and the other was unbiased, and the data from two groups of participants are displayed here. One of the groups contributed data to the 200 ms SOA Biased and 850 ms SOA Unbiased conditions, while the other group contributed data to the 200 ms SOA Unbiased and 850 ms SOA Biased conditions.

significant, F(1, 16) = 2.67, p = .12, but the cueing effects were in the predicted direction; valid trials were responded to 24 ms faster than invalid trials in the high proportion valid condition, whereas the advantage for valid trials was only 11 ms in the low proportion valid condition. Importantly, the three-way interaction between temporal context, proportion valid, and cueing was not significant (F < 1). In other words, there was no evidence that the effect of proportion valid differed for the biased and unbiased contexts. For the 850 ms SOA condition, the main effect of cueing was not significant (F < 1), but the trend was again in the expected direction. Invalid trials were responded to 7 ms faster than valid trials. The interaction between proportion valid and cueing approached significance F(1, 16) = 3.17, p = .09, and again the cueing effects were in the predicted direction; whereas RTs for valid and invalid trials did not differ in the high proportion valid condition, RTs were 14 ms slower for valid than for invalid trials in the low proportion valid condition. The three-way interaction between temporal context, proportion valid and cueing was not significant (F < 1), again offering no evidence that the influence of proportion valid on cueing effects differed for the biased and unbiased contexts. Together, these analyses demonstrate the standard results found in exogenous spatial cueing studies, with faster responses for valid than for invalid trials in the 200 ms SOA condition, and slower responses for valid than for invalid trials in the 850 ms SOA condition. The next analysis combined the data from both SOA conditions. This analysis treated cueing (valid/invalid), proportion valid (high/low), and temporal context (biased/unbiased) as within-subject variables. Note that SOA was not a variable in this analysis, as the biased SOA was 200 ms for one group of participants and 850 ms for the other group of participants. As a result, the cueing effects in this analysis are a blend of those observed for short and long SOAs. Nonetheless, this analysis addresses the two most important conceptual issues in this experiment: (1) When all the data are considered, do cueing effects vary as a function of proportion valid; and (2) When all the data are considered, does this effect of proportion valid on cueing effects differ for the biased and unbiased contexts. The analysis revealed a significant interaction between proportion valid and cueing F(1, 16) = 9.56, MSe = 176.51, p = .007, g2p = .23. To examine this interaction further, the effect of cueing was evaluated separately for the high and low proportion valid conditions. In the high proportion valid condition, the effect of cueing was significant, F(1, 16) = 4.54, MSe = 597.15, p = .049, with valid trials being responded to 12 ms faster than invalid trials. In contrast, in the low proportion valid condition, the 1 ms difference in RT for valid and invalid trials was not significant (F < 1). Overall, this interaction indicates that the proportion valid manipulation was effective in modulating the size of cueing effects. Most important, the three-way interaction between temporal context, proportion valid, and cueing was not significant (F < 1). This result indicates that the effect of proportion valid on cueing effects did not differ for the biased and unbiased contexts. 3.2.1. Awareness data The analyses of the awareness data in Experiment 2 were conducted in an analogous manner to those in Experiment 1. Mean estimates of proportion valid, collapsed across participants, are displayed in Table 3. As in Experiment 1, there were no significant effects in the analysis, indicating that participants’ estimates of proportion valid did not differ across conditions. As in Experiment 1, t-tests were performed to determine whether or not participants’ estimates of proportion valid varied systematically from chance. None of these comparisons were significant. These analyses offer no evidence that participants were aware of the proportion valid manipulation in Experiment 2. 3.3. Discussion The results of Experiment 2 were similar to those of Experiment 1. Again, responses were faster for valid than for invalid trials with a short cue-target SOA, while the opposite result was observed for a longer cue-target SOA. Also, just as in Experiment 1, the cueing effects for both short and long SOAs were sensitive to the proportion of valid trials, with larger

228

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

facilitation effects (or smaller IOR effects) in the high proportion valid condition. Finally, this modulation of cueing effects by proportion valid was again insensitive to the context manipulation, with context in this case defined by the two cue-target SOAs. This result suggests that the learning and control processes responsible for the modulation of cueing effects by proportion valid were not specific to temporal context. In the next experiment, we asked whether increasing the distinctiveness of the two contexts, by manipulating spatial and temporal attributes of the contexts simultaneously, would lead to contextspecific control over spatial orienting effects. 4. Experiment 3 In this experiment, context was defined by simultaneous manipulation of a spatial parameter (above or below the horizontal midline) and a temporal parameter (short vs long cue-target SOA). Specifically, for half of the participants the cuetarget SOA was 200 ms for trials appearing on the top of the screen, and 850 ms for trials appearing on the bottom of the screen, whereas the opposite assignment of SOAs to spatial positions held for the other half of the participants. Proportion valid was manipulated for just one of these two contexts. Again, the critical research question was whether the effect of proportion valid on cueing effects would be specific to the context in which proportion valid was manipulated. 4.1. Method 4.1.1. Participants Eighteen undergraduate students at McMaster University received course credit for their participation. All participants had normal or corrected-to-normal vision. 4.1.2. Apparatus, stimuli and procedure The stimuli and procedure for Experiment 3 were the same as Experiment 1, except that the context manipulation included both spatial (as in Experiment 1) and temporal (as in Experiment 2) dimensions. In other words, trials that appeared above fixation (the spatial context) were always associated with one cue-target SOA (the temporal context), while trials that appeared below fixation were always associated with the other cue-target SOA. The assignment of a particular combination of the spatial and temporal context dimensions was counterbalanced across participants and remained the same for both halves of the experiment for each participant. 4.1.3. Design The design of Experiment 3 was similar to that of Experiment 1, with the exception that the two contexts were defined not merely by their spatial position, but also by the cue-target SOA for trials that appeared in each spatial position. Proportion valid was then manipulated for just one of the two contexts; for half of the participants proportion valid was manipulated for the spatial position with 200 ms SOA trials, while for the other half of participants proportion valid was manipulated for the spatial position with 850 ms SOA trials. As in Experiments 1 and 2, proportion valid was manipulated in a counterbalanced manner between blocks, with .90 proportion valid in one block and .10 proportion valid in the other block, for the biased context only. Proportion valid was .50 in both blocks for the unbiased context. As in Experiment 1, the two contexts were intermixed randomly from trial to trial. 4.1.4. Post-experiment awareness questionnaire Immediately after the experiment, participants completed questionnaires identical to those described in Experiment 1. 4.2. Results RTs were submitted to the same outlier exclusion procedure as in Experiment 1 (Van Selst & Jolicoeur, 1994), which resulted in the removal of 2.6% of the response times from subsequent analyses. Mean RTs for each condition were then computed from the remaining observations. These mean RTs, collapsed across participants, are displayed in Table 5. For the same reasons outlined in the Results section of Experiment 2, our analyses focused on performance at a particular SOA as a function of whether proportion valid was manipulated at that SOA (the biased context) or not (the unbiased context). Each of these analyses treated cueing (valid/invalid) and proportion valid (high/low) as within-subject variables, and spatial/temporal context (biased/unbiased) as a between-subjects variable.2

2 Although we did carry out an omnibus ANOVA that included both SOAs, the results of this ANOVA were omitted from the body of the article for the following reason. As was the case in Experiment 2, this analysis did not include SOA as a factor; each of the two SOAs was biased for one group and unbiased for the other group, and thus context (biased/unbiased) rather than SOA was included as a factor in this analysis. The key finding in this analysis was a significant three-way interaction between cueing, proportion valid, and context, F(1, 16) = 4.63, p = .047, g2p = .22. However, as the findings reported in the main text make clear, this three-way interaction is carried by the results on short SOA trials. As such, given the unconventional nature of our design, it seemed misleading to report the results of the omnibus ANOVA; the three-way interaction in this analysis misleadingly implies that proportion valid influences on cueing were specific to the biased context overall, rather than for the short SOA trials only. For this reason, we opted to report only the results of the ANOVAs for the separate SOAs.

229

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233 Table 5 Mean response times in milliseconds for Experiment 3, collapsed across participants. 200 ms SOA

850 ms SOA

Biased P(Valid)

Valid (V) Invalid (I) Cueing effect (I–V) P(Valid) effect Difference in P(Valid) effects (Biased–Unbiased)

Unbiased P(Valid)

Biased P(Valid)

Unbiased P(Valid)

High

Low

High

Low

High

Low

High

Low

350 380 30

380 372 8

315 336 21

328 343 15

308 302 6

325 307 18

350 331 19

361 333 28

38

6 32

12

9 3

Note. As in Experiments 1 and 2, cueing effects were larger for the high proportion valid than for the low proportion valid condition. However, for the short SOA trials only, this difference in cueing effects across the proportion valid conditions was larger for the biased spatial/temporal context (i.e., the context for which proportion valid was manipulated) than for the unbiased spatial/temporal context (i.e., the SOA for which proportion valid was not manipulated). Note that for a given participant one location paired with a particular SOA was biased while the other location paired with the other SOA was unbiased, and the data from two groups of participants are displayed here. Given the labelling scheme used in the table, one of the groups contributed data to the 200 ms SOA Biased and 850 ms SOA Unbiased conditions, while the other group contributed data to the 200 ms SOA Unbiased and 850 ms SOA Biased conditions.

For the 200 ms SOA condition, there was a significant main effect of cueing, F(1, 16) = 7.72, MSe = 485.71, p = .01, g2p = .33. Valid trials were responded to 15 ms faster than invalid trials. The interaction between cueing and proportion valid was significant, F(1, 16) = 16.51, p < .0.001, g2p = .51. Valid trials were responded to 25 ms faster than invalid trials in the high proportion valid condition, whereas the advantage for valid trials was only 4 ms in the low proportion valid condition. However, and of most importance, here the three-way interaction between spatial/temporal context, proportion valid, and cueing was significant, F(1, 16) = 8.90, p = .009, g2p = .36. This interaction was examined further by conducting separate two-way ANOVAs for the biased and unbiased contexts. For the biased context, there was a significant interaction between cueing and proportion valid, F(1, 8) = 16.98, p = .0.003, g2p = .68. Responses were 30 ms faster for valid trials than for invalid trials in the high proportion valid condition t(8) = 3.78, p = .005, but 8 ms slower for valid trials than for invalid trials in the low proportion valid condition, p > .10. For the unbiased context, the interaction between cueing and proportion valid was not significant, F = 1.08, p > .30, with faster responses for valid trials in both the high proportion (21 ms) and low proportion (15 ms) valid conditions. This advantage for valid trials overall was captured by a significant main effect of cueing, F(1, 8) = 21.54, p = .002, g2p = .73. For the 850 ms SOA condition, the main effect of cueing was significant, F(1, 16) = 6.31, p = .02, g2p = .28. Invalid trials were responded to 18 ms faster than valid trials. The interaction between proportion valid and cueing approached significance, F(1, 16) = 3.36, p = .09, with the cueing effects in the predicted directions; RTs were 13 ms slower for valid than for invalid trials in the high proportion valid condition and 23 ms slower for valid than for invalid trials in the low proportion valid condition. Unlike in the 200 ms SOA condition, the three-way interaction between temporal context, proportion cued, and cueing was not significant (F < 1). 4.2.1. Awareness data The analyses of the awareness data collected in Experiment 3 were conducted in an analogous manner to Experiments 1 and 2. Mean estimates of proportion valid, collapsed across participants, are displayed in Table 3. As in Experiments 1 and 2, there were no significant effects in the analysis, indicating that participants’ estimates of proportion valid did not differ across conditions. Following the same logic as in Experiment 1, t-tests were performed to compare participants’ estimates of proportion valid to chance level (.50). These analyses revealed no significant effects, offering no evidence that participants’ were aware of the proportion valid manipulations in Experiment 3. 4.3. Discussion Many of the same effects observed in Experiments 1 and 2 were replicated in Experiment 3. Responses were again faster for valid than for invalid trials with a short cue-target SOA, while the opposite result was observed for a longer cue-target SOA. Again, cueing effects for both short and long SOAs were sensitive to the proportion of valid trials, with larger facilitation effects (or smaller IOR effects) in the high proportion valid condition. However, unlike in the previous experiments, the modulation of cueing effects by proportion valid was specific to the context in which proportion cued was manipulated, at least for the short SOA trials. This result constitutes the first evidence of which we are aware that learning and control processes that modulate cueing effects can be context-specific in nature. 5. General discussion The results from numerous recent studies suggest that adjustments in cognitive control are influenced by learning that is specific to the items, task, or context associated with experienced conflict (Crump et al., 2006; Fernandez-Duque & Knight,

230

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

2008; Funes et al., 2010; Jacoby et al., 2003; Verguts & Notebaert, 2008). The goal of the present study was to evaluate whether performance in a spatial orienting task is subject to context-specific control. To our knowledge, the experiments presented in this paper are the first to explore this issue in a spatial orienting task. In Experiment 1, the proportion of validly cued trials was manipulated across blocks for just one of two spatial contexts. The key issue was whether the influence of proportion valid on spatial cueing effects would be specific to the context in which proportion valid was manipulated, or would generalize to the context in which proportion valid was not manipulated. Indeed, proportion valid affected spatial cueing effects equivalently for the biased and unbiased contexts. In Experiment 2, a similar pattern of results was observed when context was defined by cue-target SOA rather than spatial location. However, in Experiment 3, when the biased and unbiased contexts were associated with both a unique spatial context and different SOAs, proportion valid effects for short SOA trials differed for the biased and unbiased contexts. In particular, when proportion valid was manipulated for the short SOA trials presented in a specific spatial context, cueing effects for those short SOA trials were larger in the high proportion valid than in the low proportion valid condition. In contrast, when proportion valid was manipulated for the long SOA trials presented in a specific spatial context, cueing effects for the accompanying short SOA trials were no different for the high proportion valid than for the low proportion valid condition. Taken together, these experiments suggest that context-specific control of spatial orienting can occur when proportion valid is associated with two distinctly different contexts. Moreover, the results of the awareness questionnaires indicate that control over spatial orienting in this study was unaccompanied by awareness of the proportion valid manipulation (see also Bartolomeo et al., 2007; Risko & Stolz, 2010). 5.1. Proportion valid effects: Generalization across contexts In Experiments 1 and 2, proportion valid influenced cueing effects to the same degree across contexts in which proportion valid was and was not manipulated. Although we have described this effect as one of generalization across the two contexts, there are at least two ways in which this type of generalization could occur. One possibility is that participants learn nothing about the different proportion valid manipulations in the two contexts, and instead learn that proportion valid is high or low independent of the two contexts. For example, in Experiment 1, when 90% of trials were valid in the biased context and 50% of trials were valid in the unbiased context, participants might have learned that, generally speaking, 70% of trials were valid. An attentional strategy could in turn be implemented based on this broad knowledge of the proportion valid across both contexts. A second possibility is that participants did learn about the proportion valid manipulation specifically in the biased context, and this knowledge transferred to performance in the unbiased context. By this view, across many trials in the high proportion valid block participants may have learned to expect the appearance of a target in the valid location. Similarly, across many trials in the low proportion valid block participants may have learned to expect the target to appear opposite the cue. Both of these learning processes may have stemmed from processing that was specific to the biased context. In contrast, the unbiased context would not have supported this kind of cue-target learning, as the cue was not a predictive signal for the location of the target. Given the absence of a predictive strategy stemming from processing of unbiased context trials, the predictive strategy learned on biased context trials may generalize to unbiased context trials, in effect filling the strategic void for those trials. The design and results of the present study do not allow us to distinguish between the above two possibilities, but the results from the awareness questionnaires do constrain both accounts. In particular, results of the awareness questionnaires offered no evidence that participants were aware that proportion valid was manipulated separately for the two contexts, and indeed offered no evidence that participants were aware of the proportion valid manipulations at all. As such, whether learning was general across both contexts, or specific to one context and transferred to the other context, the results of the present study converge with those of other recent studies in suggesting that control over spatial orienting need not hinge on aware attentional strategies (Bartolomeo et al., 2007; Risko & Stolz, 2010). 5.2. Proportion valid effects: Specific to a particular context In Experiment 3, on short SOA trials, proportion valid influenced cueing effects to a larger extent for the context in which proportion valid was manipulated than for the context in which proportion valid was not manipulated. This result constitutes the first example of which we are aware that control over orienting in a spatial cueing paradigm can be guided by context-specific learning processes. Such a result also implies that spatial orienting processes may be subject to some of the same stimulus-driven learning and control mechanisms as reported for various other tasks (see Bugg et al., 2008; Cañadas et al., 2012; Heinemann et al., 2009; Leboe et al., 2008; Lehle & Hübner, 2008; Wendt & Kiesel, 2011). Although the precise mechanisms that drive this effect require further study, we speculate below on how such a context-specific control result might occur in the present study, and in particular on why it might occur only for short SOA trials. The context-specific control result of interest here can be described precisely as follows: The spatial cueing effect for short SOA trials was sensitive to proportion valid when manipulated on those same short SOA trials, but was not sensitive to proportion valid when manipulated for a companion set of long SOA trials. At the core of this finding could be a learning mechanism that is more sensitive to cue-target contingencies when cue and target are separated by a short than by a long temporal interval. However, there must be two constraints on such a learning mechanism. First, the data collected in

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

231

post-experiment questionnaires offer little evidence that participants were aware of the learning mechanisms that produced the proportion valid effects. Across all conditions, and in all three experiments, participants’ estimates of proportion valid did not differ across conditions, or from chance (.5 across both halves of an experiment). An additional analysis of the awareness data from Experiment 1, where SOA was manipulated between subjects, was conducted to investigate whether participants’ awareness and strategic use of the cue differed across SOAs. Mean estimates of proportion valid for the short SOA condition are displayed in rows three and four of Table 3, and for the long SOA condition in rows five and six of Table 3. As in all other analyses of the awareness data, there were no significant effects, offering no support for the view that awareness and strategic use of the cue differed for short and long SOAs. Similarly, t-tests that compared participants’ estimates of proportion valid to a chance level (.50) revealed no significant effects, again offering no evidence that participants’ awareness and strategic use of cues differed across SOAs. As such, we conclude that some unaware learning mechanism that produces stronger learning effects for short cue-target SOAs could be responsible for the result of interest here. At the same time, the context-specificity of the proportion valid effect in Experiment 3 points to a second constraint on this learning process. Specifically, the robust learning associated with proportion valid for short cue-target SOAs in Experiment 3 did not generalize to the companion long SOA trials that were mixed together in the same block; the proportion valid effect for long SOA trials was as large when proportion valid was manipulated for short SOA trials as when it was manipulated for long SOA trials. One candidate mechanism that could conceivably produce this type of unaware, context-specific learning for only shortSOA trials relates to the object file updating framework of Kahneman, Treisman, and Gibbs (1992). Kahneman et al. proposed that onset of a target event can cue the retrieval of spatio-temporally similar episodic representations, called object files. The current target might then be linked with the retrieved object file, and interpreted as an updated version of a previously presented object. In effect, object files are thought of as episodic representations that can support the integration of successive states of an object across time. In the context of the present experiments, it is possible that cue and target displays were subject to these object integration processes, and further that these integration processes were subject to learning processes that were sensitive to the relative proportions of valid and invalid trials. In other words, participants may have learned when proportion valid was manipulated for short SOA trials that there was either a high or low likelihood that consecutive events (i.e., cue and target) would belong to the same object. Processing of trials in the short SOA context only, where object integration processes play a large role, might then be subject to this learning effect. Note also that this context-specific learning process that affects object integration might well be qualitatively different from other learning processes capable of producing proportion valid effects under other conditions (e.g., at long SOAs). Although the above account is speculative in nature, there does appear to be a need for a framework that explains a range of results, both here and elsewhere, in which spatial cueing effects have been shown to be sensitive to proportion valid in the absence of awareness of variation in proportion valid (Bartolomeo et al., 2007; Lambert, Naikar, McLachlan, & Aitken, 1999; Risko & Stolz, 2010).3 In addition, studies outside the domain of spatial cueing have also pointed to the important role of unaware context-specific learning and control mechanisms. For example, Crump et al. (2006), Crump, Vaquero, & Milliken, 2008) examined whether context-specific learning processes might underlie, at least in part, proportion congruent effects in the Stroop task. They presented colour patch probes in one of two location contexts, either above or below fixation, following presentation of a congruent or incongruent colour word. Their results demonstrated a context-specific proportion congruent effect and, as in the present study, participants showed no awareness of the proportion congruent manipulation. The conflict monitoring framework could perhaps also be modified to account for the types of context-specific control over spatial orienting observed here and elsewhere (Bartolomeo et al., 2007; Risko & Stolz, 2010). According to the conflict-monitoring framework, activation in the ACC is sensitive to conflict encountered in the processing of information (Botvinick et al., 2001). If we assume that mismatches in the location of cue and target produce conflict, then activity in the ACC might reliably track the relative proportions of valid and invalid trials, which could in turn impact cueing effects. To explain context-specific proportion cued effects, ACC activation would have to be context-sensitive, resulting in different levels of cognitive control over responding as a function of the relevant context. Indeed, Blais et al. (2007) proposed a variant of Botvinick et al.’s (2001) conflict-monitoring model that incorporated control at the item-specific level, and that could therefore simulate an item-specific proportion congruent Stroop effect (Jacoby et al., 2003). A similar such adaptation of the conflict monitoring model might well account for contextual specificity of proportion valid effects observed in the present experiments.4

3 Risko and Stolz (2010) used proportion manipulations that were smaller in magnitude than those used in the present experiments, and proposed that the use of more extreme proportions may engage voluntary and strategic control processes. The present data fail to support this hypothesis. 4 A related issue that has received attention in studies of context-specific control is whether adjustments in control impact performance on-line in response to the contextual cues associated with the current trial, or instead carry over from the preceding trial and therefore only impact performance when contextual cues repeat from one trial to the next (King, Korb, & Egner, 2012). We examined this issue in the present study by re-analyzing the short SOA data from Experiment 3, focusing exclusively on the biased context condition (i.e., the condition that produced a proportion valid effect), and including previous trial context (biased/unbiased) as an additional factor. The question of interest was whether the proportion valid by cueing interaction would vary as a function of the context (biased/unbiased) of the preceding trial. Although the 3-way interaction did not approach significance, there was a numerical trend in the direction of a larger proportion valid by cueing interaction when the preceding trial context was biased (i.e., a short SOA trial in the same location context as the current short SOA trial) than when it was unbiased (i.e., a long SOA trial in the opposite location context to the current short SOA trial). Whether this trend reflects a true effect must await further research with a more sensitive design.

232

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

5.3. Generalization of cognitive control in natural environments Spalek and Hammad (2005) produced evidence that in an exogenous cueing task, like that used in the present experiments, English-reading participants are faster to detect a target that appears on the right after a cue on the left, and Arabic-reading participants are faster to detect targets appearing on the left after a cue on the right. In other words, performance was biased by the direction that participants customarily read. Why should experience in reading be reflected in a completely different, non-naturalistic task that is designed, controlled and completed in a laboratory environment? This result is especially surprising given the results of the current Experiment 3, in which context-specific learning was observed across two contexts that differ merely in terms of subtle spatial and temporal parameters within a single task. Given these results, the generalization of orienting behaviour from years of experience in reading to a spatial orienting task reported by Spalek and Hammad (2005) is all the more remarkable. To account for these findings, we note that reading is a behaviour that people perform regularly from early childhood. The potential training effect of all that experience may bias behaviour without recourse to the kind of cognitive control mechanisms explored in the present experiments. Perhaps the lack of top-down engagement in these behaviours reduces the need for the recruitment of cognitive control processes, and instead the transfer of training from an abundance of experience with reading to a spatial orienting task is driven by processes that are distinct from those that drove performance in our study. 5.4. Summary The results of the present study demonstrate for the first time that a context-specific manipulation of proportion valid in an exogenous spatial orienting task can modulate spatial cueing effects. This context-specific control over spatial orienting appears not to be due to strategic use of the cue, as participants showed no awareness of the proportion cued manipulation. These results converge with those from studies of selective attention (e.g., Stroop, flankers) in demonstrating that control over attention processes may be cued rapidly and without awareness in response to processing of contextual features that are not directly task-relevant. Moreover, whereas the time between cue and target is conventionally thought of as defining a temporal window in which cognitive control might operate, the present results suggest that such temporal features can serve as contextual cues that directly impact cognitive control (see also Wendt & Kiesel, 2011). Acknowledgments This research was supported by an NSERC Discovery Grant awarded to B.M. References Barch, D. M., Braver, T. S., Akbudak, E., Conturo, T., Ollinger, J., & Snyder, A. (2001). Anterior cingulate cortex and response conflict: Effects of response modality and processing domain. Cerebral Cortex, 11, 837–848. Barch, D. M., Braver, T. S., Sabb, F. W., & Noll, D. C. (2000). Anterior cingulate and the monitoring of response conflict: Evidence from an fMRI study of overt verb generation. Journal of Cognitive Neuroscience, 12, 298–309. Bartolomeo, P., Decaix, C., & Sieroff, E. (2007). The phenomenology of endogenous orienting. Consciousness and Cognition, 16, 144–161. Blais, C., & Besner, D. (2006). Reverse Stroop effects with untranslated responses. Journal of Experimental Psychology: Human Perception and Performance, 32(6), 1345–1353. Blais, C., Robidoux, S., Risko, E. F., & Besner, D. (2007). Item-specific adaptation and the conflict-monitoring hypothesis: A computational model. Psychological Review, 114, 1076–1086. Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624–652. Botvinick, M., Nystrom, L. E., Fissell, K., Carter, C. S., & Cohen, J. D. (1999). Conflict monitoring versus selection-for-action in anterior cingulate cortex. Nature, 402, 179–181. Bugg, J. M., & Crump, M. J. C. (2012). In support of a distinction between voluntary and stimulus-driven control: A review of the literature on proportion congruent effects. Frontiers in Psychology, 3, 367. Bugg, J. M., Jacoby, L. L., & Chanani, S. (2011). Why it is too early to lose control in accounts of item-specific proportion congruency effects. Journal of Experimental Psychology: Human Perception and Performance, 37(3), 844–859. Bugg, J. M., Jacoby, L. L., & Toth, J. P. (2008). Multiple levels of control in the Stroop task. Memory & Cognition, 36, 1484–1494. Cañadas, E., Rodríguez-Bailón, R., Milliken, B., & Lupiáñez, J. (2012). Social categories as a context for the allocation of attentional control. Journal of Experimental Psychology: General. http://dx.doi.org/10.1037/a0029794. Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D., & Cohen, J. D. (1998). Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 280, 747–749. Chun, M. M., & Jiang, Y. (1998). Contextual cueing: Implicit learning and memory of visual context guides spatial attention. Cognitive Psychology, 36, 28–71. Crump, M. J. C., Gong, Z., & Milliken, B. (2006). The context-specific proportion congruent Stroop effect: Location as a contextual cue. Psychonomic Bulletin & Review, 13(2), 316–321. Crump, M. J. C., Vaquero, J., & Milliken, B. (2008). Context specific learning and control: The role of awareness, task relevance, and relative salience. Consciousness & Cognition, 17, 22–36. Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16, 143–149. Fernandez-Duque, D., & Knight, M. (2008). Cognitive control: Dynamic, sustained, and voluntary influences. Journal of Experimental Psychology: Human Perception and Performance, 34(2), 340–355. Funes, M. J., Lupiáñez, J., & Humphreys, G. (2010). Analyzing the generality of conflict adaptation effects. Journal of Experimental Psychology: Human Perception and Performance, 36(1), 147–161. Heinemann, A., Kunde, W., & Kiesel, A. (2009). Context-specific prime-congruency effects: On the role of conscious stimulus representations for cognitive control. Consciousness and Cognition, 18, 966–976.

A. Gough et al. / Consciousness and Cognition 30 (2014) 220–233

233

Jacoby, L. L., Lindsay, D. S., & Hessels, S. (2003). Item-specific control of automatic processes: Stroop process dissociations. Psychonomic Bulletin & Review, 10(3), 638–644. Kahneman, D., Treisman, A., & Gibbs, B. (1992). The reviewing of object files: Object-specific integration of information. Cognitive Psychology, 24, 175–219. King, J. A., Korb, F., & Egner, T. (2012). Priming of control: Implicit contextual cuing of top-down attentional set. Journal of Neuroscience, 32, 8192–8200. Klein, R. M. (2000). Inhibition of return. Trends in Cognitive Sciences, 4(4), 138–147. Lambert, A., Naikar, N., McLachlan, K., & Aitken, V. (1999). A new component of visual orienting: Implicit effects of peripheral information and subthreshold cues on covert attention. Journal of Experimental Psychology: Human Perception & Performance, 25, 321–340. Leboe, J. P., Wong, J., Crump, M., & Stobbe, K. (2008). Probe-specific proportion task repetition effects on switching costs. Perception & Psychophysics, 70, 935–945. Lehle, C., & Hübner, R. (2008). On-the-fly adaptation of selectivity in the flanker task. Psychonomic Bulletin & Review, 15, 814–818. Logan, G. D., & Zbrodoff, N. J. (1979). When it helps to be misled: Facilitative effects of increasing the frequency of conflicting stimuli in Stroop-like tasks. Memory and Cognition, 7, 166–174. Lowe, D., & Mitterer, J. O. (1982). Selective and divided attention in a Stroop task. Canadian Journal of Psychology, 36, 684–700. Lu, C.-H., & Proctor, R. W. (1995). The influence of irrelevant location information on performance: A review of the Simon effect and Congruency effects. Psychonomic Bulletin & Review, 2, 174–207. Lupianez, J., Klein, R. M., & Bartolomeo, P. (2006). Inhibition of return: Twenty years after. Cognitive Neuropsychology, 23, 1003–1014. Peterson, B. S., Kane, M. J., Alexander, G. M., Lacadie, C., Skudlarski, P., Leung, H., et al (2002). An event-related functional MRI study comparing interference effects in the Simon and Stroop tasks. Cognitive Brain Research, 13, 427–440. Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H. Bouma & D. G. Bouwhuis (Eds.), Attention and performance X: Control of language processes (pp. 531–556). Hillsdale, NJ: Erlbaum. Posner, M. I., Rafal, R. D., Choate, L., & Vaughan, J. (1985). Inhibition of return: Neural basis and function. Cognitive Neuropsychology, 2, 211–228. Risko, E. F., & Stolz, J. A. (2010). The proportion valid effect in covert orienting: Strategic control or implicit learning? Consciousness and Cognition, 19, 432–442. Schmidt, J. R., & Besner, D. (2008). The Stroop effect: why proportion congruent has nothing to do with congruency and everything to do with contingency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(3), 514. Simon, J. R., & Rudell, A. P. (1967). Auditory S-R compatibility: The effect of an irrelevant cue on information processing. Journal of Applied Psychology, 51, 300–304. Spalek, T. M., & Hammad, S. (2005). The left-to-right bias in inhibition of return is due to the direction of reading. Psychological Science, 16(1), 15–18. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643–662. Torres-Quesada, M., Funes, M. J., & Lupiáñez, J. (2013). Dissociating proportion congruent and conflict adaptation effects in a Simon-Stroop procedure. Acta Psychologica, 142, 203–210. Van Selst, M., & Jolicoeur, P. (1994). A solution to the effect of sample size on outlier elimination. Quarterly Journal of Experimental Psychology, 47A, 631–650. Verguts, T., & Notebaert, W. (2008). Hebbian learning of cognitive control: Dealing with specific and nonspecific adaptation. Psychological Review, 115(2), 518–525. Wendt, M., & Kiesel, A. (2011). Conflict adaptation in time: Foreperiods as contextual cues for attentional adjustment. Psychonomic Bulletin & Review, 18, 910–916.

Control of spatial orienting: context-specific proportion cued effects in an exogenous spatial cueing task.

Cognitive control refers to the ability to adjust strategy use based on the demands of a current context or task. Recent research using attentional fi...
325KB Sizes 0 Downloads 8 Views