Working in an attentive and concentrated manner is an important prerequisite for successful behavior in many areas of life, including school and the workplace. To assess a person’s ability to work in an attentive and concentrated manner, psychologists have developed pen-and-paper tests examining selective attention and concentration, such as the d2 test (e.g., Brickenkamp,
1962,
2002) and the ‘Frankfurter Aufmerksamkeitsinventar’ (FAIR; Moosbrugger & Oehlschlägel,
1996). Because the results in such tests may have serious consequences for the tested person, these tests should be both highly reliable and highly valid. However, both the (retest) reliability and the validity of the d2, for example, are curtailed by effects of practice, that is, the fact that repeating the test considerably improves the test results (e.g., Hagemeister & Westhoff,
2011; Schmidt-Atzert, Büttner, & Bühner,
2004). The effects of practice decrease the (retest) reliability of a test because the results of a person in two subsequent tests may differ considerably. Moreover, the effects of practice also threaten the validity of a test because it is unclear whether practice or high ability has caused a good test result in situations where the practice history of the tested person is unknown (Hagemeister & Westhoff,
2011; Schmidt-Atzert et al.,
2004).
The FAIR test (Moosbrugger & Oehlschlägel,
1996) was constructed in response to some shortcomings of previous tests, such as the d2 (cf. Oehlschlägel & Moosbrugger,
1991). While both the existence and size of practice effects are empirically documented for the d2 (e.g., Steinborn, Langner, Flehmig, & Huestegge,
2018; review in Hagemeister & Westhoff,
2011), there are no published data on the size and characteristics of practice effects in the FAIR or FAIR-2 (Moosbrugger & Oehlschlägel,
1996,
2011). Hence, the main goal of this study is to close this gap and to start exploring the sources of practice effects in the FAIR-2.
Practice effects in conjunction-search tasks: methods, findings, and accounts
From the viewpoint of cognitive psychology, both the d2 and the FAIR-2 tests require a visual search for conjunction targets that are hidden between heterogenous distractor stimuli (for reviews of the rich literature on visual search, see Chan & Hayward,
2013, and Wolfe,
1998). In conjunction-search tasks, the targets are defined by a particular combination of features, whereas other combinations occur as distractor stimuli. The visual search for conjunction targets among distractors, which share features with the targets, requires the formation of internal target representations—the so-called “search templates” (e.g., Bravo & Farid,
2012; Wolfe & Horowitz,
2004). Simple visual search models assume that the stimulus display is then attentively searched and every stimulus (representation) is compared to the search template(s). Depending on whether this comparison produces a match or a mismatch, a corresponding response is made (e.g., Treisman & Gelade,
1980; Wolfe,
1998).
Research on the impact of practice on visual search has a long tradition, too. For example, in a classic study by Neisser (
1963), participants practiced searching for a variable number of targets for up to 31 days. At the beginning of the practice, search times increased with the number of targets being searched for, but these differences decreased with practice and had almost vanished after 31 days of practice. This finding suggests that, after sufficient practice, the processes of visual search may become automatic and, therefore, depend less on limited resources such as attention. According to this notion, whereas participants have to direct their attention to each item in unpracticed searches, the target items may (automatically) attract attention towards their position after extended practice (also see Shiffrin & Schneider,
1977).
Subsequent studies demonstrated that both the type of stimulus material and the stimulus–response mapping affect practice in visual conjunction-search. Concerning the stimulus material, substantial practice effects have been demonstrated when participants searched for arbitrary combinations of line segments (e.g., Czerwinski, Lightfoot, & Shiffrin,
1992; Lubow & Kaplan,
1997), letters (e.g., Neisser,
1963; Prinz,
1979) or words (e.g., Fisk, Lee, & Rogers,
1991; Rogers & Fisk,
1991). Moreover, whereas practice can improve searches for conjunctions of color and location (e.g., Frank et al.,
2014), practice does not seem to affect search for conjunctions of color and shape (e.g., Leonards et al.,
2002; Sireteanu & Rettenbach,
2000). Additionally, the consistency of the stimulus–response mapping can also affect practice in visual search. In consistent-mapping (CM) conditions, stimulus X is always a target and stimulus Y is always a distractor, whereas in variable-mapping (VM) conditions each stimulus is sometimes a target and sometimes a distractor. Several studies have identified that practicing a search task under CM conditions can produce much bigger improvements in performance than practice under VM conditions (e.g., Fisk et al.,
1991; Rogers & Fisk,
1991; Shiffrin & Schneider,
1977). Interestingly, though, it turned out later that the observation of larger practice effects with CM conditions, as compared to VM conditions, seems to be constrained to alphanumerical (i.e., familiar) stimuli, whereas practice can similarly improve performance in both CM and VM conditions with novel stimuli (e.g., Czerwinski et al.,
1992; Lightfoot, Czerwinski, & Shiffrin,
1993; Shiffrin & Lightfoot,
1997).
Several accounts for the effects of practice in visual search tasks have been proposed. A first account assumes that practice in visual search tasks mainly changes the attentional “weights” of both the targets and the distractors. In particular, in their “
attention-attraction” model, Shiffrin and Schneider (
1977) assume that CM practice increases the attentional weight (or attention-attraction strength) of targets, and decreases the attentional weights of distractors (also see Czerwinski et al.,
1992; Rogers,
1992; Shiffrin,
1988). Among other observations (which will be discussed in due course), this account can explain why CM practice leads to larger improvements in performance than VM practice. In fact, according to Shiffrin and colleagues, larger training effects with CM conditions than with VM conditions are a hallmark of the automatization of visual search (e.g., Czerwinski et al.,
1992; Lightfoot et al.,
1993).
A second group of accounts assumes that practice in visual search tasks improves the perceptual processing of targets, or the perceptual discrimination between targets and distractors (cf. Goldstone,
1998, for a review). According to one account, practice improves the perceptual processing of targets by stimulating the formation of new processing units—a process called ‘unitization—for previously unpracticed stimulus combinations (Czerwinski et al.,
1992; Frank et al.,
2014; Lightfoot et al.,
1993; Shiffrin & Lightfoot,
1997).
According to the third account of perceptual learning, practice improves the perceptual discrimination of targets and distractors, that is, participants learn to detect and process those features that distinguish between targets and distractors in a given task (e.g., Cousineau & Larochelle,
2004; Duncan & Humphreys,
1989; Fisher,
1982; Rabbitt,
1964). If a participant has to search for the letters E and H among distractors K and Z, a short horizontal line in the middle of the stimulus would be a critical feature that could be used to discriminate between targets and distractors.
A series of studies with words provided evidence for the attention-attraction account of practice in visual search (Fisk et al.,
1991; Rogers & Fisk,
1991; Rogers,
1992).
2 Participants first practiced visual search tasks for 10 days under both CM and VM conditions. The task consisted of searching for a word from a pre-cued semantic category among distractor words. After the practice phase, participants were tested in several tests or transfer conditions. A first notable result was that performance improved more strongly during practice with CM conditions than with VM conditions, and the search was faster in CM conditions than in VM conditions at the end of the training. The results from the test conditions typically showed that the repetition of the target or the distractor from practice to test produced a better performance as compared to a control condition, in which neither the target nor the distractor were repeated. In contrast, when either the target, the distractor, or both (i.e., role reversal) switched their role from practice to test, performance deteriorated compared to the control condition. The pattern of findings is consistent with an attention-attraction account (e.g., Shiffrin & Schneider
1977; Shiffrin,
1988), and inconsistent with a perceptual-discrimination account of practice in visual search (e.g., Cousineau & Larochelle,
2004; Duncan & Humphreys,
1989; Fisher,
1982; Rabbitt,
1964). For example, the strong disruption of performance in the role-reversal condition is compatible with the attention-attraction account because, according to this account, participants would have to search for weak targets (i.e., previous distractors) among strong distractors (i.e., previous targets) in this condition. In contrast, according to the perceptual-discrimination account, reversing the roles of targets and distractors should not significantly impair performance because the learned features for discriminating targets and distractors remain the same.
In another series of studies, Shiffrin and colleagues demonstrated the limits of automatization in visual search and provided evidence for the unitization of stimuli as a result of extended practice (e.g., Czerwinski et al.,
1992; Lightfoot et al.,
1993; Shiffrin & Lightfoot,
1997). These authors reasoned that the stimulus material used in a search task determines how practice would affect performance. In particular, they assumed that the automatization of visual search would mainly occur when targets and distractors were dissimilar, and when the unitization of stimuli was unlikely to occur. Both of these conditions are met with alphanumeric stimuli. In contrast, with unfamiliar stimuli and high similarity between targets and distractors, the practice could improve performance by developing higher-order units for stimulus processing. Thus, “if subjects initially process stimuli at the feature level and later learn to process them holistically, the number of comparisons required to discriminate targets from distractors would go down considerably as training proceeds” (Czerwinski et al.,
1992, p 296). The novel stimuli consisted of a rectangular frame that contained three line segments in different configurations. Participants extensively practiced the visual search for a pre-cued target in displays containing a target (or no target) and several distractors. A (unique) combination of two features distinguished a given target from each distractor in each stimulus set, hence, a conjunction search was required. There were several notable results. First, at the beginning of practice, performance was much worse than usually reported in studies with alphanumeric stimuli, suggesting an effect of different familiarity with stimulus sets. Second, practice improved performance to a similar degree in both CM and VM conditions, suggesting that practice did not automatize visual search under these conditions. Third, although practice strongly improved performance, a closer examination of performance suggested that participants were still serially searching for targets after practice, albeit at a much higher rate than at the beginning of practice. Fourth, when participants were transferred to a condition with new sets (i.e., combinations) of practiced stimuli, which still required a conjunction search, performance did not appear to suffer (e.g., Experiment 3 in Shiffrin & Lightfoot,
1997). Fifth, when participants were transferred to a condition with new sets of practiced stimuli that no longer required a conjunction search, because a single feature distinguished each pair of stimuli, performance did not improve as would be expected if participants were still comparing individual features. In summary, the results of these studies provide evidence for the hypothesis that practicing the search for unfamiliar geometrical stimuli can lead to a unitization of the stimulus representations guiding such searches.
In a more recent study, Frank et al. (
2014) investigated the neuroanatomical correlates of practice effects in a visual search task. In their study, participants practiced searching for a conjunction of color and location on eight successive days.
3 Participants were then probed in several test (or transfer) conditions, including role reversal and a control condition involving a conjunction search for a new set of stimuli.
4 Functional magnetic resonance imaging was used to measure the brain activity of three participants during both the training and test sessions. On the behavioral level, performance improved during practice and seemed to reach an asymptote after 4–5 days. Moreover, performance almost dropped to pre-training levels in the role-reversal condition, as well as in the control condition. On the neuronal level, the authors observed that practice-induced changes in performance were correlated with increasing activity in visual areas (i.e., V1–V4), but were not correlated with activity in areas related to the control of eye movement (i.e., frontal eye fields, supplementary eye fields, superior colliculi). The latter areas were supposed to measure the effects of practice on the control of attention. To account for their results, Frank et al. (
2014) suggested that practice effects in their search task had a perceptual locus (e.g., unitization) rather than an attentional one (e.g., automatization of target detection), supporting the earlier results of Shiffrin and colleagues (e.g., Shiffrin & Lightfoot,
1997).
Contextual cueing
In addition to changes in the attentional weights of stimuli or changes in the cognitive representations of stimuli, participants might also learn to use contextual cues to improve performance in visual search tasks with practice. For example, in an influential study, Chun and Jiang (
1998; see also Chun,
2000) demonstrated that participants can implicitly learn the correlation between the configuration of distractor items and the position of the target. In their Experiment 1, participants searched for a rotated ‘T’ under heterogeneously rotated ‘L’ distractors. Each display contained a target, and participants reported the orientation of the target by pressing a key. Importantly, each block of trials contained a set of “old” displays and a set of “new” displays in random order. The old set of displays consisted of 12 stimulus configurations that were repeated throughout the whole experiment, once per block. The new set of displays consisted of 12 different configurations that were newly generated for each block; these served as a control condition. Thirty blocks of trials were grouped into 6 epochs of 5 blocks each. The results showed that starting with epoch 2, RTs were faster for old displays compared to new displays, demonstrating contextual cueing. Hence, participants were able to quickly learn to locate the target in repeated stimulus configurations. A follow-up experiment revealed that learning the relationship between the spatial configuration of distractors and the target location was important for contextual cueing, whereas distractor identities did not play a role (but see Makovski,
2016 for different findings). Subsequent studies on the locus of the contextual-cueing effect suggested that contextual cueing facilitates the attentional guidance of visual searches, rather than facilitating a response to the target (e.g., Harris & Remington,
2017; see Sisk, Remington, & Jiang,
2019 for review).
While contextual cueing may play a role in the d2 test of attention, it cannot do so in the FAIR-2 test. In the d2 test of attention, each of three stimulus lines is repeated several times to obtain a total of 14 lines. In contrast, in the FAIR-2, stimuli are randomly ordered in each line, and no stimulus line occurs twice in the test. As such, the learning of stimulus configurations (or target locations), in the sense of contextual cueing, is not possible in the FAIR-2 test.
Implications of findings from research on practice effects in the visual search for pen-and-paper tests of visual attention
The empirical evidence suggests that practice may affect visual search at different stages, including changes in the perceptual processing of target stimuli (i.e., unitization) and changes in the attentional guidance of visual search (cf. Czerwinski et al.,
1992; Goldstone,
1998; Lightfoot et al.,
1993). Despite these findings, different mechanisms seem to be triggered in different situations. Whereas unitization seems to occur in the search for novel stimuli when the similarity of targets and distractors is high, changes in the attentional attractiveness of stimuli seem to occur for familiar stimuli when the similarity of targets and distractors is low (e.g., Lightfoot et al.,
1993). In contrast, transfer studies of practice in visual search tasks produced little evidence for the learning of critical features distinguishing targets from distractors.
At present, one can only speculate about the possible sources of practice effects in pen-and-paper tests of visual attention, such as the FAIR-2. In fact, there are many methodological differences between these tests and the typical search tasks used in the laboratory, and empirical studies on the sources of practice effects in pen-and-paper tests of attention do not exist. The FAIR-2 test, however, involves unfamiliar stimuli and high levels of similarity between targets and distractors, creating conditions under which, according to Shiffrin and Lightfoot (
1997), practice may lead to unitization rather than changes of attentional guidance.
Finally, besides altering the processing of stimuli, practicing a visual search task may also cause stimulus-independent learning. Stimulus-independent effects of practice, sometimes called “task learning” (e.g., Frank et al.,
2014), might benefit performance in different ways, such as improving familiarity with the test situation (i.e., reducing test anxiety), developing more efficient search strategies and improving the execution of motor responses (e.g., Frank et al.,
2014; Rogers,
1992; Sireteanu & Rettenbach,
2000; Wühr,
2019).