At the mercy of prior entry: Prior entry induced by invisible primes is not susceptible to current intentions

Abstract

If one of two events is attended to, it will be perceived earlier than a simultaneously occurring unattended event. Since 150 years, this effect has been ascribed to the facilitating influence of attention, also known as prior entry. Yet, the attentional origin of prior-entry effects¹ has been repeatedly doubted. One criticism is that prior-entry effects might be due to biased decision processes that would mimic a temporal advantage for attended stimuli. Although most obvious biases have already been excluded experimentally (e.g. judgment criteria, response compatibility) and prior-entry effects have shown to persist (Shore, Spence, & Klein, 2001), many other biases are conceivable, which makes it difficult to put the debate to an end. Thus, we approach this problem the other way around by asking whether prior-entry effects can be biased voluntarily. Observers were informed about prior entry and instructed to reduce it as far as possible. For this aim they received continuous feedback about the correctness of their temporal judgments. If elicited by invisible primes the effect could not be reduced at all and only by 12 ms if elicited by visible cues. This challenges decision biases as primary source of prior-entry effects — at least if attention is oriented exogenously.

Introduction

Imagine watching out at the night sky for shooting stars. Probably you will see a shooting star in an attended sky area earlier than a simultaneously appearing one in an unattended area. In other words, attended stimuli are perceived earlier than unattended stimuli. This prior-entry effect has been investigated in experimental psychology over the last 150 years (e.g. Boring, 1929, Scharlau, 2007b, Spence et al., 2001, Stone, 1926, Titchener, 1908; for a recent review see Spence & Parise, 2010). During this time period, prior-entry effects have been demonstrated within and between several sensory modalities (e.g. visual modality: Scharlau, 2007b, Shore et al., 2001, Stelmach and Herdman, 1991, Weiß and Scharlau, 2011; auditory modality: Kanai, Ikeda, & Tayama, 2007; somatosensory modality: Yates and Nicholls, 2009, Yates and Nicholls, 2011; between modalities: Spence et al., 2001, Zampini et al., 2007, Zampini et al., 2005).

As early as 1908, Edward B. Titchener assumed that “Unless, then I am unduly optimistic the negative displacement [that is, the prior-entry effect] need give psychologists no further trouble” (p. 259). Looking back, this statement was indeed unduly optimistic because the prior-entry effect has given psychologists some trouble. Especially one criticism was repeatedly raised: It was argued that the effect might be due to or might be enhanced by biases in decision or response criteria that would mimic an advantage in temporal perception for attended stimuli. But before we will explain this criticism in more detail, it is necessary to outline the basic experimental paradigm for assessing prior-entry effects.

Usually prior-entry effects are assessed either by a temporal order judgment (TOJ) or a simultaneity judgment (SJ) task. Observers judge either which of two rapidly presented stimuli was presented first (TOJ) or whether both stimuli were presented simultaneously or not (SJ). In both tasks two factors are manipulated. First, the temporal delay between the two targets is varied. Second, attention is directed toward one of the targets. The expected perceptual advantage for attended stimuli, that is, the prior-entry effect, occurs as a shift in the so-called point of subjective simultaneity (PSS). In the TOJ task, this is the temporal delay at which the two possible order judgments are given equally often. In the SJ task, this is the temporal delay at which observers judge “simultaneous” most frequently. Usually the PSS is shifted toward a temporal delay at which objectively the unattended stimulus leads the attended one. This shift's size quantifies the prior-entry effect.

For orienting attention in prior-entry studies several methods have been used; attention was oriented by instruction (e.g. Spence et al., 2001, Stelmach and Herdman, 1991, Yates and Nicholls, 2009) by peripheral location cues (e.g. Schneider & Bavelier, 2003, Experiment 1; Shore et al., 2001) or by central symbolic cues (e.g. Schneider & Bavelier, 2003, Experiment 2, Shore et al., 2001). Since the last decade Scharlau, 2002, Scharlau, 2004a, Scharlau, 2004b, Scharlau, 2004c, Scharlau, 2007a, Scharlau, 2007b, Scharlau, Ansorge, and Horstmann (2006), Scharlau and Neumann, 2003a, Scharlau and Neumann, 2003b and Weiß and Scharlau (2011) used a special prior-entry paradigm termed perceptual latency priming (PLP) in which peripheral cues are masked. Since we will also use PLP in the present study we will describe it in more detail.

In the PLP paradigm, attention is directed by peripheral primes. These primes are peripheral cues which are metacontrast-masked by the target following at the same location. As a consequence, they are invisible or barely visible to the observer (e.g. Breitmeyer & Ögmen, 2006). Despite this strongly reduced visibility, primes in the PLP paradigm are effective in directing attention toward the target (e.g. Scharlau, 2004a), like cues in other prior-entry paradigms. This attention-directing property is suggested by several empirical findings. First and most importantly, the time course of PLP mirrors that typically reported for exogenous attention. With priming intervals below the duration of an exogenous attention shift − 100–200 ms (Nakayama and Mackeben, 1989, Suzuki and Cavanagh, 1997) PLP increases constantly; its peak is located between 100 and 300 ms, with longer priming intervals PLP decreases (e.g. Scharlau et al., 2006, Scharlau and Neumann, 2003b, for an overview about PLP see Scharlau, 2007a). Second, PLP does not seem to be due to sensory priming. Since prime and primed target are presented at the same location, PLP can partially or totally reflect an acceleration of the targets' sensory processing due to pre-activation of sensory receptors by the prime. In this case PLP's size should depend on prime–target similarity. But as Scharlau and Neumann (2003a, Experiment 4) demonstrated using a binary TOJ-task PLP has the same size for prime–target pairs with different degrees of similarity (shape and color). Third, PLP is not due to an averaging between prime and target onset since target leading primes but not target trailing primes cause PLP. Fourth, that PLP is due to confusion between prime and target onset is unlikely since the size of PLP remains the same for congruent and incongruent primes. Confusion should lead to a reduction in PLP for incongruent primes since in this case the prime specifies a response favoring the unattended target. Taken together, these properties of PLP argue for its attentional origin, thus making it a useful paradigm for studying prior-entry effects. So far we spared the question how a response bias could account for PLP since we will discuss this question in the next paragraph along with other prior-entry paradigms.

There are several ways imaginable in which a response bias² could account for a given prior-entry effect. First of all it is important to note that judging the temporal order of two stimuli occurring in very rapid succession is a difficult task. Therefore observers might be inclined to use other than temporal information for solving this task. For instance, if uncertain about the temporal properties of the stimulus sequence, observers could report the stimulus appearing at the attended location or sharing the attended property as the first. Here, the PSS-shift would be due to a judgment bias and not to attentional prioritization. In a more sophisticated form of a response bias, observers might actually ascribe the judgment criterion to any salient stimulus property, which would include the feature of being the attended one (cf. Frey, 1990, Pashler, 1998, Schneider and Bavelier, 2003). These so-called “second-order biases” or criterion biases are very difficult to avoid and cannot easily be distinguished from attention-based prior-entry effects.

Many studies (e.g. Kanai et al., 2007, Lester et al., 2009, Roberts and Humphreys, 2010, Scharlau, 2004a, Shore et al., 2001, Yates and Nicholls, 2009) took precautions against such second-order biases by varying the judgment criterion of the TOJ task. Observers judged either which stimulus was presented first or which stimulus was presented second. If they assigned any judgment criterion to the attended stimulus, the judgment “which first?” should lead to the usual prior-entry effect, but the judgment “which second?” should reveal an opposite advantage for the unattended stimulus. Half of the difference between the apparent temporal advantages derived from the two judgment criteria is an estimate of decision bias (Shore et al.). Using this method in a PLP-paradigm, Scharlau (2004a) found virtually no bias whereas Shore et al. found a bias of 13 ms with a prior-entry effect of 61 ms with visible peripheral cues. This finding is in accordance with the higher difficulty to distinguish between “attended” and “non-attended target” in a PLP-paradigm where attention is oriented with invisible primes. This makes a decision bias favoring the attended target less likely in PLP. Although these results can be counted as evidence against a prominent role of decision bias in prior-entry effects, not all researchers are convinced by this. For instance Schneider and Komlos (2008) argued that observers could use the same judgment criterion for both judgments and then invert their response in one of the conditions. Thus, the possible contribution of a decision bias to prior-entry effects seems a question very difficult to settle.

Apart from post-perceptual decision processes a response bias might also arise on the level of sensory or motor processing. With respect to motor processing, if the required response is compatible with allocation of attention – e.g. attention is directed to the left or right and the observers have to judge whether the stimulus on the right or left appeared first – motor priming might facilitate judgments in favor of the attended target. Other response-relevant features of the attention-grabbing prime can also elicit responses which might mimic prior entry, for example if the prime has the same shape as the primed target (Scharlau & Neumann, 2003a, Experiment 1). This type of motor priming can be excluded either by primes with irrelevant features (Scharlau & Neumann, 2003a, Experiments 3 and 4) or by using primes which share their response-relevant features with the unprimed target (e.g. Weiß & Scharlau, 2011).

Another possible bias in prior-entry paradigms using peripheral primes or cues is a sensory bias. Primes that are presented close to the primed target might pre-activate sensory receptors concerned with target processing. Acceleration of the target is then caused by faster sensory processing. As mentioned above a large contribution of sensory priming to PLP is unlikely since PLP's size is not dependent on prime–target similarity.

So far, we have only spoken about how response biases could account for prior-entry effects in TOJs. Usually smaller prior-entry effects are found with the SJ task (e.g. Schneider and Bavelier, 2003, Van der Burg et al., 2008, Yates and Nicholls, 2011). This led some authors to assume that SJs provide a more exact, bias free measure of prior-entry effects. In SJs a tendency favoring one of the judgments would affect the frequency of “simultaneous” judgments and thereby the width and the height of their bell-shaped distribution but importantly not their peak, that is the PSS. Additionally, the SJ task is less prone to sensorimotor biases since a judgment about simultaneity or succession is neither compatible with attentional allocation nor can it be specified by a prime. Although it seems very convincing at first glance that the SJ task is less prone to response biases, this can be criticized for two reasons. First, the argument presupposes that judgments of temporal order and judgments of simultaneity can be used interchangeably, implying that they measure the same underlying processes. This implies that the percepts of temporal order and simultaneity rely on the same internal event. This assumption is at odds with many studies which revealed different consequences of experimental manipulations on TOJs and SJs (e.g. Guerrini et al., 2003, Mitrani et al., 1986, Petrini et al., 2010, Shore et al., 2005, Stelmach and Herdman, 1991, Van Eijk et al., 2008, Vatakis et al., 2008, Weiß and Scharlau, 2011) indicating that judgments of temporal order and judgments of simultaneity have – at least partially – different underlying mechanisms. For instance Weiß and Scharlau, revealed that prior-entry assessed from order judgments is accompanied by a substantial reduction in simultaneity perception. This finding is predicted by the temporal profile model of prior entry by Stelmach and Herdmann which assumes that perception of temporal order and perception of simultaneity rely on different mechanisms (for a more thorough discussion of this topic see the General discussion). Second, even if TOJ and SJ would measure completely the same underlying processes, impaired sensitivity of the SJ task in comparison to the TOJ task could account for the smaller prior-entry effects found in SJs as well (e.g. Anton-Erxleben, Abrams, & Carrasco, 2010). Hence, assessing prior-entry effects with an SJ task instead of a TOJ task is not a completely satisfying solution for the bias argument which even creates new problems.

Thus, it is difficult if not even impossible to show that no kind of criterion or decision bias contributes to a given prior-entry effect. Therefore, we decided to approach this problem the other way around. Instead of asking which amount of a given prior-entry effect might be ascribed to biases in decision criteria, we ask whether observers are able to bias a given prior-entry effect in accordance with their current intentions. If observers would fail to bias a prior-entry effect substantially, this would be a strong argument against the practical relevance of decision biases in prior-entry effects. If the prior-entry effect arises on the level of post-perceptual decision processes, it seems plausible that these processes should be susceptible to current intentions.

If observers can exert voluntary control on a given prior-entry effect, this influence should be reflected in two parameters. First, we expect that the bias should change the PSS, that means, the size of the prior-entry effect. This is the heart of the bias argument against the attentional explanation of prior-entry effects. The second parameter which can reflect a strategic influence is the difference limen (DL) as measure of discrimination accuracy. Observers should be at least able to impair their discrimination accuracy by paying less attention to the temporal task. Whether observers might be able to improve their temporal discrimination accuracy by investing more effort into the TOJ is a question more difficult to answer. Although the bias argument focuses on the size of the prior-entry effect changes in discrimination accuracy might go along with changes in PSS. For instance favoring the attended target via a bias should reduce discrimination accuracy.

Theoretically, the size of a given prior-entry effect can be biased in two directions, enlargement and reduction. At first glance, enlarging the prior-entry effect seems to be perfect for testing the bias hypothesis. At a closer look it has some disadvantages. First, the range of target delays constricts the range of measurement and thus the possible bias effect. Second, how large must the bias effect be – that is how much must the prior-entry effect grow – for concluding that it is exclusively or mostly due to a bias effect? Fortunately, the predictions are clear-cut for a reduction of the prior-entry effect. If we bias observers against prior entry, the apparent facilitation effect should be completely eliminated if the effect is due to voluntary control alone³; otherwise the prior-entry effect would be reduced by the amount which is susceptible to voluntary control.

Since we were interested in a most exact estimate of the maximal amount of the prior-entry effect which could be controlled voluntarily, it is very important to create circumstances most beneficial for a reduction of the prior-entry effect. For this means, observers were informed about the mechanism of prior entry. They further received feedback about the correctness of their TOJs; this feedback was given trial-by-trial and in regular intervals for performance in all trials completed up to the respective interval.

The prior-entry effect is inevitably accompanied by a high number of wrong judgments, and the enlargement of the number of wrong judgments is proportional to the size of the prior-entry effect (see Fig. 1 for an illustration). Conversely, making fewer errors reduces the prior-entry effect.

If the bias hypothesis is correct, this feedback should make it easy to overcome any judgment tendency favoring the attended stimulus and to learn giving the correct response even under conditions of temporal uncertainty. Elimination or at least a large reduction of the prior-entry effect should be the consequence. Also, the feedback will improve temporal discrimination accuracy. If the attentional hypothesis is correct, the task would be much more difficult. Observers would have to overcome their own temporal perception, and it should be far more difficult to profit from the provided feedback. Note that response remapping — that is choosing the judgment alternative “unattended stimulus first” under the impression that actually the “attended” stimulus had occurred first cannot reduce the prior-entry effect. It would, by contrast, lead to a negative prior-entry effect favoring the unattended target. For reducing or even eliminating the prior-entry effect observers have to identify the trials in which attentional prioritization had led to illusory reversals of the targets' temporal order and remap their response only in these trials. Additionally, as in most recent prior-entry studies, response dimension and dimension of attentional allocation are orthogonally varied (cf. Shore et al., 2001) in our study, thus observers must also indentify the “attended” and the “unattended” target before remapping their responses.

Thus, to overcome one's actual temporal perception would be a difficult task and we expect a minor or no reduction of the prior-entry effect at all. As the observers will probably try several strategies to distrust their perception and due to increased task difficulty, we might find a further decrease in temporal discrimination accuracy in comparison to primed trials where no feedback is provided.