December 2016
Volume 16, Issue 15
Open Access
Article  |   December 2016
Expectations and visual aftereffects
Author Affiliations
Journal of Vision December 2016, Vol.16, 19. doi:10.1167/16.15.19
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Noga Pinchuk-Yacobi, Ron Dekel, Dov Sagi; Expectations and visual aftereffects. Journal of Vision 2016;16(15):19. doi: 10.1167/16.15.19.

      Download citation file:


      © 2017 Association for Research in Vision and Ophthalmology.

      ×
  • Supplements
Abstract

The tilt aftereffect (TAE) is traditionally regarded as a consequence of orientation-selective sensory adaptation, a low-level stimulus-driven process. Adaptation has been recently suggested to be the outcome of predictive coding. Here, we tested whether the TAE is modulated by predictability, and specifically, whether TAE depends on the congruency of adapted and expected orientations. Observers were presented with successive pairs of oriented Gabor patches. Pairs were arranged in blocks, forming two conditions with the orientation of the second pair member either predictable or not. For all pairs, the orientation of the first Gabor was tilted clockwise (CW) or counterclockwise (CCW) (±20° relative to vertical, randomized). In the “Expected” conditions, the orientation of the second Gabor was fixed relative to the first Gabor (the same or a mirror orientation, blocked). In the “no-expectation” condition, the orientation of the second Gabor was independent of the first Gabor (randomized ±20°). Intermixed test pairs were used to measure observers' perceived vertical, with the second pair member serving as a target, oriented around the vertical, permitting an estimate of the TAE produced by the presentation of the first Gabor. Results show an increase in TAE with the expected orientation matching the inducing orientation, but a decrease with the expected mirror orientation, consistent with additivity of the adaptation and the expectation effects. A second experiment, with the first oriented Gabor replaced by a colored circular blob, showed that expectation alone does not modulate the perceived orientation. These findings indicate a role for expectation in generating the perceptual TAE and are in line with predictive coding models of perception. We suggest that orientation dependent adaptation is affected by both the mean orientation (first order statistics) and by temporal contingencies (second order statistics).

Introduction
Visual adaptation refers to the visual system adjusting to changes in the statistics of the visual environment. Adaptation usually results in temporary biases in the perception of subsequently viewed stimuli, termed aftereffects (Thompson & Burr, 2009). For example, adaptation to an oriented stimulus induces a tilt aftereffect (TAE) in which the perceived orientation of a test stimulus appears slightly repelled away from the orientation of the adapted stimulus (Gibson & Radner, 1937). This repulsion effect is strongest for angles that are 10°–20° away from the adapting orientation (Campbell & Maffei, 1971). The TAE is primarily driven by mere exposure even when the stimulus is unattended or invisible (He & MacLeod, 2001). It was traditionally assumed to result from a bottom-up (stimulus-driven) process (De Baene & Vogels, 2010; Sekuler & Pantle, 1967), with longer adaptation causing stronger and longer lasting TAE (Magnussen & Johnsen, 1986). However, a growing body of evidence suggests that adaptation adjustments, including aftereffects, are not just automatic, but vary depending on the adapting procedure. For example, spatial attention was found to increase the magnitude of aftereffects following adaptation to a variety of features like motion (Chaudhuri, 1990) and orientation (Festman & Ahissar, 2004). Using optical distortions, it was shown that observers can “learn to adapt.” For example, with repeated adaptation on subsequent days, observers adjusted faster to a prism distortion (Yehezkel, Sagi, Sterkin, Belkin, & Polat, 2010). In accordance, four days of adaptation to an altered virtual reality, which lacked vertical information, revealed effects of days on adaptation, with different adaptation dynamics across days (Haak, Fast, Bao, Lee, & Engel, 2014). In an additional study, repeated adaptation to motion or contrast, across multiple daily sessions, induced aftereffects that were affected by previous adaptation sessions and task context (Dong, Gao, Lv, & Bao, 2016). It was recently shown that aftereffects interact with learning. Perceptual learning of texture discrimination was found to induce a TAE that was task dependent and persisted longer following training (Pinchuk-Yacobi, Harris, & Sagi, 2016). 
Prior expectations regarding the stimulus can also bias perception. Perception was shown to be biased in both directions: toward (attractive) prior expectations (Adams, Graf, & Ernst, 2004; Haijiang, Saunders, Stone, & Backus, 2006) and away from (repulsive) prior expectations (Brayanov & Smith, 2010). Attractive bias in perception can be explained by Bayesian inference, which is based on combining the incoming sensory statistics (likelihood) with the learned expectations (priors) in a probabilistically optimal manner (Box & Tiao, 1992). According to Bayes' law, under conditions of uncertainty, perception is biased towards prior expectations; thus, the value of unexpected information is discounted. However, in the TAE perception is biased away from prior expectation, thus, consistent with an exaggeration of unexpected sensory information and with predictive coding, according to which the visual cortex continuously generates predictions of incoming stimuli and transmits only the unpredicted part of them (Rao & Ballard, 1999). In recognition tasks, the currently recognized object is thought to form predictions relative to which errors are transmitted. In agreement with this proposal, TAE was found to decrease with the adapting stimulus duration when natural images were used as adapters in a recognition task (Dekel & Sagi, 2015). 
Adaptation is observed over multiple time scales, ranging from milliseconds to minutes, thus involving both, possibly interacting, fast and slow dynamics; it is suggested to implement an efficient coding strategy in sensory processing (Geisler, 2008). Neurophysiological and psychophysical studies have shown that many similar effects can be induced by both brief and prolonged adaptation (Bonds, 1991; Felsen et al., 2002; Foley & Boynton, 1993; Harris & Calvert, 1989). For example, V1 neurons' preferred orientation shifts away from the adapting orientation with both prolonged, several minutes (Dragoi, Sharma, & Sur, 2000), and brief (∼20 ms) adapting stimuli (Felsen et al., 2002), though the shift magnitude is smaller (about half) for brief presentations. Perceptual aftereffects were also found for both brief and prolonged exposures. Harris and Calvert (1989) showed TAE after adaptation as brief as 300 ms, with longer adaptation periods usually resulting in larger TAE values. These fast adaptive adjustments can allow the online discounting of predicted stimuli, as suggested by predictive coding models (Hosoya, Baccus, & Meister, 2005). 
In this work, we aimed to determine how stimulus-driven expectations regarding orientation affect perceived orientation. More specifically, how expecting a repetition of or a change in the orientation of a briefly presented adaptor stimulus modulates the TAE caused by the adaptor stimulation. In addition, we tested whether orientation expectation could modulate observers' perceived orientation even without bottom-up adaptive orientation stimulation. This was tested in a condition whereby orientation-expectation was created by color-orientation temporal contingency. 
Methods
Apparatus
The stimuli were presented on a 23.6-in VIEWPixx/3D monitor (1920 × 1080, 10 bit, 120 Hz, with “scanning backlight mode”) viewed at a distance of 100 cm. The mean luminance of the display was 47.26 cd/m2 in an otherwise dark environment. 
Observers
Six, five, and six observers, with normal or corrected-to-normal vision, participated in three separate experiments. All observers were naïve to the task and gave their written informed consent. The work was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). 
Stimuli and tasks
Observers were exposed to successive pairs of visual patterns. The first member in the pairs differed between experiments (see below), and the second member was always an oriented Gabor patch (σ = 0.6°, λ = 0.3°, random phase, carrier amplitude of 48.7%). Each member in the pair was presented for 50 ms at the center of the screen, and a blank screen (600 ms) was presented between pair members. Between pairs there was a blank screen presented for a time interval of 1–1.5 s, randomized across trials. Observers were instructed to monitor the stimulus stream for occasional test pairs, which were marked by four peripheral crosses in the second member of the pair (Figure 1C). In test pairs, the second member in the pair was a Gabor patch randomly tilted in one of seven orientations: 0°, ±1°, ±3°, and ±5° (relative to the vertical), and the observers were asked to report whether the Gabor was tilted clockwise or counterclockwise. Test pairs were used to measure the shift in the observers' perceived vertical, with the second pair member serving as a target, thus permitting a measurement of the change in perceived orientation produced by the presentation of the first pair member. Psychometric curves (percentage of clockwise reports as a function of target orientation) were interpolated to find the perceived vertical orientation, which is the orientation with an equal probability for clockwise and counterclockwise reports. Interpolation was performed by fitting a cumulative normal function (lapse rate = 0) using Psignifit 3.0 software for MATLAB (Fründ, Haenel, & Wichmann, 2011). 
Figure 1
 
Orientation cue experiment: stimuli and block design. Oriented Gabor patches were presented in successive pairs. In each pair, the first Gabor member was randomized (±20° relative to vertical). Observers monitored the stimulus stream for occasional test pairs, which were marked by four peripheral crosses in the second member of the pair. (A) Adaptor pairs in which the two Gabor members had the same orientation. (B) Adaptor pairs in which the two Gabor members had a mirror orientation. (C) Test pairs in which the second Gabor member was oriented around the vertical. Observers were requested to report whether the Gabor is tilted clockwise or counterclockwise. (D) Experimental block design. “Expected same orientation” blocks contained the same orientation Gabor pairs. “Expected mirror orientation” blocks contained mirror orientation Gabor pairs. “No expectation” blocks contained an equal number of the same orientation and mirror orientation Gabor pairs. In all blocks, a third of the pairs were test pairs. All pairs in each block were randomly intermixed.
Figure 1
 
Orientation cue experiment: stimuli and block design. Oriented Gabor patches were presented in successive pairs. In each pair, the first Gabor member was randomized (±20° relative to vertical). Observers monitored the stimulus stream for occasional test pairs, which were marked by four peripheral crosses in the second member of the pair. (A) Adaptor pairs in which the two Gabor members had the same orientation. (B) Adaptor pairs in which the two Gabor members had a mirror orientation. (C) Test pairs in which the second Gabor member was oriented around the vertical. Observers were requested to report whether the Gabor is tilted clockwise or counterclockwise. (D) Experimental block design. “Expected same orientation” blocks contained the same orientation Gabor pairs. “Expected mirror orientation” blocks contained mirror orientation Gabor pairs. “No expectation” blocks contained an equal number of the same orientation and mirror orientation Gabor pairs. In all blocks, a third of the pairs were test pairs. All pairs in each block were randomly intermixed.
Orientation cue experiment
The first member in each pair was an oriented Gabor patch randomly tilted, either 20° clockwise or counterclockwise relative to the vertical (Figure 1A through C). For adaptor pairs (i.e., not test pairs) the orientation of the second Gabor was either the same orientation as the first Gabor (Figure 1A) or the mirror orientation of the first Gabor (Figure 1B). 
Color cue experiment
The first member in each pair was a colored Gaussian blob (σ = 0.48°), either violet “red” (R = 227, G = 27, B = 127) or spring “green” (R = 27, G = 227, B = 127). For adaptor pairs, the second member was the same as in the orientation cue experiment (Gabor, ±20°; Figure 2). 
Figure 2
 
Color cue experiment stimuli. The layout of the pairs is the same as in Figure 1. (A) Adaptor pairs with a particular color-orientation contingency. (B) Adaptor pairs with the mirror contingency. (C) Test pairs. The task is the same as in Figure 1C. Experimental blocks contained either contingency pairs (“Expected” blocks) or contingency and mirror pairs (“No expectation” blocks), in addition to test pairs, similar to Figure 1D.
Figure 2
 
Color cue experiment stimuli. The layout of the pairs is the same as in Figure 1. (A) Adaptor pairs with a particular color-orientation contingency. (B) Adaptor pairs with the mirror contingency. (C) Test pairs. The task is the same as in Figure 1C. Experimental blocks contained either contingency pairs (“Expected” blocks) or contingency and mirror pairs (“No expectation” blocks), in addition to test pairs, similar to Figure 1D.
Procedures
Orientation cue experiment
The pairs were arranged in three block types (Figure 1D): (a) “Expected same orientation” blocks contained the “Same orientation” Gabor pairs; (b) “Expected mirror orientation” blocks contained the “Mirror orientation” Gabor pairs; and (c) “No-expectation” blocks contained an equal number of the “Same orientation” Gabor pairs and the “Mirror orientation” Gabor pairs; therefore, the orientation of the second Gabor was independent of the first Gabor (randomized ±20°). Observers were not informed about the block type. All blocks contained 252 pairs, a third of which were test pairs. In each block, adaptor and test pairs were randomly shuffled. The “Expected same orientation” and the “Expected mirror orientation” blocks were presented to two different groups of observers (N = 6 and N = 5, respectively). Each observer performed one to four daily control sessions that contained four “No expectation” blocks, and two to four test daily sessions that contained an equal number of “No expectation” and “Expected” (either the same orientation or mirror orientation) blocks. Control and test daily sessions were intermixed on different days. In test sessions, the block order was N-E-E-N, where N stands for “No expectation” blocks and E stands for “Expected” blocks. The duration of each block was about 10 min and there was a 2-min break between blocks. The first session was a control session used for familiarization with the task, and therefore its data was excluded from the analysis. 
Color cue experiment
The pairs were arranged according to two block types: (a) “Expected orientation” blocks contained the “Contingency” pairs (red followed by −20°, green followed by +20°); and (b) “No expectation” blocks contained an equal number of the “Contingency” pairs and the “Mirror contingency” pairs; therefore, the color of the first member (red or green) was independent of the orientation of the second member (±20). Observers (N = 6) performed one control session and five test sessions. Session structures and block durations were similar to the orientation cue experiment. 
Results
In the “Orientation cue experiment,” observers were presented with pairs of oriented Gabor patches. The orientation of the first Gabor served as an adaptor. The orientation of the second Gabor was either expected (“Expected same orientation” blocks or “Expected mirror orientation” blocks) or not expected (“No expectation” blocks) from the orientation of the first Gabor. We measured the shift in the observers' perceived vertical following adaptation to the first Gabor, termed the tilt aftereffect (TAE). The observer's perceived vertical was calculated separately for test pairs in which the first Gabor was tilted clockwise and for test pairs in which the first Gabor was tilted counterclockwise. The TAE was quantified as the average absolute value of the TAE generated by these two conditions (i.e., half the difference in the perceived vertical orientation between these two conditions), averaged across all block repetitions. 
Temporal integration of TAEs
The different block types contained different sequences of pairs of Gabor patches. Since adaptation is expected to be modulated by the number of repeated adaptor exposures (i.e., longer adaptation causing stronger and longer lasting effects; Greenlee, Georgeson, Magnussen, & Harris, 1991; Magnussen & Johnsen, 1986), and by intervening stimuli (i.e., de-adaptation by different orientations; Greenlee & Magnussen, 1988), we first analyzed how the basic TAE, in the absence of built-up expectations, was affected by the different recent stimulus histories. We expect remote history effects to average out due to the built-in symmetry in the experiments between clockwise and counterclockwise adaptors having opposite influences. For that, we measured the TAE separately for each stimulus history defined by the most recent three adaptors (the adaptor paired with the test and the previous adaptor pair), averaging across more remote histories. This formed eight history types (Table 1). The analysis included the “No expectation” blocks from both groups of observers (test sessions as well as control sessions). Stimulus history that contained only a test pair (i.e., test pair followed by another test pair) was not considered in “recent” history, since the time between the pairs was undefined and depended on the observer's response; therefore, trials with such history were excluded from the analysis (see also further discussion below). 
Table 1
 
TAE measurements based on data from the “No expectation” blocks.
Table 1
 
TAE measurements based on data from the “No expectation” blocks.
Table 1 presents TAE measurements that are based on data from the “No expectation” blocks, calculated separately for the different types of stimulus history. History is presented as {O3-O2-O1} triplets, where Oi stands for the orientation of the i-back adaptor, either CCW for counterclockwise orientation or CW for clockwise orientation. TAE estimates are based on measurements from 11 observers (13,356 trials overall). 
The analysis showed that the TAE was affected by stimulus history. The TAE magnitude was quite similar (but with opposite direction) for matching history types but with mirror orientations (CW↔CWW; e.g., in Table 1, the value of the TAE in row 1 is similar to that in row 2). 
When there were no expectations, the averaged TAE was about the same for trials with stimulus history corresponding to the “Expected same orientation” condition (TAE = 1.35°, average of values 1 to 4 in Table 1), and for trials with stimulus history corresponding to the “Expected mirror orientation” condition (TAE = 1.27°, average across stimulus history types 5 to 8 in Table 1). In addition, individual observer's results of the TAE calculated separately for history types corresponding to the two different block types (“Expected same orientation” and “Expected mirror orientation”), showed no systematic differences (see corresponding Figure A1 in the Appendix). 
Next, we tested whether the TAE values we measured were compatible with a model that assumes linearity in the influence of previous stimulus orientations on the final perceived TAE. This can be expressed by the following set of linear equations, one for each stimulus history j (j = 1:8).    
Where w0 stands for the perceived vertical (PV) when there is no adaptation, wi stands for the weight of the influence of the i-back adaptor, and oij stands for the orientation of the i-back adaptor in history sequence j (we assign “−1” for CW adaptor and “+1” for CCW adaptor). Best fitting solution of this overdetermined equation set to the TAE values presented in Table 1 gives w0 = −0.05, w1 = 1.31, w2 = 0.28, w3 = 0.1 (see Appendix for more details). The linear model was found to explain 99% of the variance in the data. This shows that to a first approximation a linear model explains the history effects in our data. Therefore, when the history of stimulus orientations is locally balanced (same number of CW and CCW adaptors), the average TAE magnitude is defined by the most recent adaptor, and is about the same for all conditions. 
Expectations: Effects of oriented cues
Next, we aimed to test whether TAE was modulated by the orientation expectations created in the different “Expected” conditions. For each group of observers we compared the TAE in the “Expected” blocks against the corresponding TAE in the “No expectation” blocks. The TAE was calculated separately for each stimulus history, similar to the analysis performed previously for all the “No expectation” blocks (see Table 1). In order to improve the accuracy of the fit (since there were much fewer “Expected” blocks than “No expectation” blocks), we combined and fitted the data for matching history types regardless of the direction CW or CCW (e.g., in Table 1, data for stimulus history 1 combined with data for stimulus history 2, 3, with 4, etc). In this way, the average PV is canceled out and so w0 can be omitted. 
Tables 2 and 3 show TAE estimates for the “Expected same orientation” group (based on measurements from six observers, 2856 trials overall for “Expected” blocks and 5376 trials overall for “No expectation” blocks) and the “Expected mirror orientation” group (based on measurements from five observers, 2520 trials overall for “Expected” blocks and 7896 trials overall for “No expectation” blocks), respectively, calculated separately for the different types of stimulus history. History is presented as {O3-O2-O1} or {T-O1}, where T stands for a test pair (immediately preceding the test for which the TAE is shown) and Oi stands for the orientation of the i-back adaptor, either CCW for counterclockwise orientation or CW for clockwise orientation. Interestingly, the TAEs measured for test trials that were preceded by a test pair (i.e., a test pair followed by another test pair) were larger, in most cases, than the TAEs measured for test trials that were preceded by an adaptor pair (line 5 in Tables 2 and 3). Performing a one-tailed paired t test (11 observers) between the TAEs for tests preceded by a test pair and the TAEs for tests preceded by an adaptor pair was found to be significant both when there was no expectation (mean difference = 0.29°, p = 0.03, “No expectation” blocks), and when there was an expectation (mean difference = 0.34°, p = 0.04, “Expected blocks). We speculate that the larger TAE results from increased adaptation to the adaptor stimulus that is preceded by a test. Adaptation might be increased because observers are probably more alert or attentive (as subjective report indicate) to the Gabor adaptor when it appears after a test (Spivey & Spirn, 2000). Enhanced attention following the presentation of a test is also supported by the finding of shorter reaction time (RT) for tests preceded by test pairs (RT = 737 ms) compared with tests preceded by adaptor pairs (RT = 763 ms, p = 0.02, pairwise two-tailed t test). 
Table 2
 
TAE measurements for the “Expected same orientation” group.
Table 2
 
TAE measurements for the “Expected same orientation” group.
Table 3
 
TAE measurements for the “Expected mirror orientation” group.
Table 3
 
TAE measurements for the “Expected mirror orientation” group.
Results show that observers in the “Expected same orientation” group had on average a significantly larger TAE value in the “Expected same orientation” blocks (TAE = 1.70°, average across stimulus history types 1, 2, and 5 in Table 2) compared with the corresponding average TAE in the “No expectation” blocks (TAE = 1.25°, p = 0.004, pairwise two-tailed t test; Figure 3, black rhombuses). Considering the TAE values for the individual history types, the results reached statistical significance in only two history types (history type 1: TAE difference = 0.40°, p < 0.05, history type 2: TAE difference = 0.39, p = 0.06, history type 5: TAE difference = 0.56°, p = 0.02, pairwise one-tailed t test; Table 2). In accordance, observers in the “Expected mirror orientation” group had on average a significantly lower TAE value in the “Expected mirror orientation” blocks (TAE = 1.27°, average across stimulus history types 3, 4, and 5 in Table 3) compared with the “No expectation” blocks (TAE = 1.58°, p = 0.001; Figure 3, white rhombuses). Considering the TAE values for the individual history types, the results did not reach statistical significance (history type 3: TAE difference = −0.35°, p = 0.06, history type 4: TAE difference = −0.21, p = 0.2, history type 5: TAE difference = −0.37°, p = 0.07, pairwise one-tailed t test; Table 3). 
Figure 3
 
Orientation expectation modulates the perceived vertical orientation only in the presence of adaptation. Individual data of the perceived vertical (PV) difference (averaged across all block repetitions) measured in the “Expected” condition as a function of the “No expectation” condition. Observers' perceived vertical (PV) was calculated separately for two conditions, according to the different first pair members (orientation: −20 or +20; color: red or green). The difference in PV was quantified as the difference in PV between these two conditions divided by two. In the orientation cue experiment, when observers expected the orientation of the second Gabor in the pair to be the same (black rhombuses) or the mirror (white rhombuses) of the first Gabor, their TAE (PV difference due to orientation adaptation) was increased or decreased (respectively) relative to their TAE when they had no expectation. In the color cue experiment, the orientation-expectation resulting from the color-orientation contingency did not affect the observers' perceived vertical.
Figure 3
 
Orientation expectation modulates the perceived vertical orientation only in the presence of adaptation. Individual data of the perceived vertical (PV) difference (averaged across all block repetitions) measured in the “Expected” condition as a function of the “No expectation” condition. Observers' perceived vertical (PV) was calculated separately for two conditions, according to the different first pair members (orientation: −20 or +20; color: red or green). The difference in PV was quantified as the difference in PV between these two conditions divided by two. In the orientation cue experiment, when observers expected the orientation of the second Gabor in the pair to be the same (black rhombuses) or the mirror (white rhombuses) of the first Gabor, their TAE (PV difference due to orientation adaptation) was increased or decreased (respectively) relative to their TAE when they had no expectation. In the color cue experiment, the orientation-expectation resulting from the color-orientation contingency did not affect the observers' perceived vertical.
Thus, in both groups (“Expect same orientation” and “Expect mirror orientation” groups) the TAE was modulated by orientation expectations but in opposite direction. Accordingly, TAE modulations (“Expected”–“No expectation”) were significantly different between the two “Expected” groups (p < 0.0001, two-tailed unequal-variance t test). The results for the two groups suggest that the orientation expectation interacts additively with the TAE and therefore can be added as an additional variable to the linear equation of the TAE, giving the following elaborated equation:  where we stands for the weight of the influence of the orientation expectation, and oe stands for the sign of the expected orientation.  
Fitting this set of overdetermined equations to the TAE values of the “Expected same orientation” group, using least squares method, gives we = 0.4, w1 = 1.11, w2 = 0.24, w3 = 0.09. The linear model was found to explain 99% of the variance in the data (see Appendix for detailed equations based on Table 2). For the “Expected mirror orientation” group (TAE values of Table 3) we find we = 0.28, w1 = 1.53, w2 = 0.32, w3 = 0.21. The linear model was found to explain 99% of the variance in the data (see Appendix for detailed equations based on Table 3). 
Expectations: Effects of color cues
Next, we inquired whether expecting a stimulus orientation can modulate the perceived vertical orientation even when there is no orientation adaptation. To test this, we conducted a second experiment in which the first pair member was a colored Gaussian blob instead of a Gabor patch (“Color cue experiment”). The second pair member, a Gabor patch, had an orientation that was either expected (“Expected” blocks) or not expected (“No expectation” blocks) from the color of the Gaussian blob. In this experiment there is no TAE because there is no adaptation to the orientation of the first pair member. 
Results show that the perceived vertical orientation of the test Gabor, averaged across all block repetitions, was unaffected by the color of the previous Gaussian blob, in both the “Expected orientation” (p = 0.5, two-tailed t test) and “No expectation” blocks (p = 0.36). In addition, there was no difference between blocks (p = 0.72, pairwise two-tailed t test, Figure 3, gray asterisks). This shows that the orientation expectation resulting from the color-orientation contingency did not affect the perceived orientation. 
Discussion
Here we showed that an adaptation-induced shift in the perceived orientation (TAE) is modulated by rapidly learned stimulus statistics. When observers expected the test orientation to repeat the adapted orientation, their TAE was increased; however, when they expected the test to assume the mirror orientation, their TAE was decreased compared with their TAE when they had no expectations. Interestingly, in the absence of orientation adaptation, expecting a stimulus orientation (which resulted from the color orientation temporal contingency) did not affect the observers' perceived orientation. 
Results show that observers manage to learn over time the temporal contingencies between the orientation of the first cue Gabor and the orientation of the subsequent Gabor. These learned contingencies enable a later presented cue Gabor to induce expectation regarding the orientation of the subsequent Gabor, which biases the observer's perceived vertical in a direction opposite to that originally associated with the cue. This is similar to the orientation-contingent color McCollough effect (McCollough, 1965) in which sustained association of a specific orientation and a specific color causes a subsequent presentation of the same orientation to appear in a color complementary to the color originally associated with the orientation. 
Our results can be interpreted in the framework of hierarchical predictive coding, according to which, the visual cortex constantly generates predictions of incoming stimuli at multiple levels of processing (Rao & Ballard, 1999). Each level maintains an internal model of the environment based on prior statistical regularities and communicates with other levels by signaling deviations between these assumed models and the actual input (prediction error). In this framework, repulsive biases in the perception of an oriented stimulus can be viewed as forms of predictive errors relative to the predicted orientation. That is, when a test stimulus has an orientation that violates the predicted orientation, the predicted part of its activity is subtracted or “explained away,” thus causing its perceived orientation to be repelled away from its actual orientation (Lochmann, Ernst, & Deneve, 2012). The finding of TAE modulation by rapidly learned orientation expectations can be interpreted as an addition of two expectations or predictions: first-order and second-order expectations. Accordingly, orientation adaptation introduces a first-order expectation for continuity of the adapted orientation. This expectation is based on first-order statistical regularities of natural stimuli, to be continuous in space and time. This component is usually measured in traditional adaptation experiments. The rapidly learned orientation expectation in our experiment introduces a second-order expectation based on statistical regularities that are more complex, such as the relationship or the correlation between pairs of orientations. This component modulates the size of the first-order effect. In the “Expected same orientation” condition, the two expectations were congruent, leading to an enhanced orientation expectation and thus, an enhanced TAE when this expectation was violated. On the other hand, in the “Expected mirror orientation” condition the two expectations were incongruent, leading to a reduced TAE magnitude. Similar additive interactions were previously shown between the TAE and the short-term memory of an oriented grating (Saad & Silvanto, 2013). These interactions suggest that our second-order orientation expectations induced repulsive tilt biases—away from the expected orientation. These biases, which exaggerate the unexpected sensory information, are inconsistent with Bayes' law. To account for repulsive effects in a Bayesian framework, it was argued that adaptation might change the likelihood rather than the prior, by reallocation of orientation-selective representation, which increases the measurement's precision around the adapted orientation (Stocker & Simoncelli, 2006). However, since our expectation effect exists in a balanced orientation presentation (equal presentation of clockwise and counterclockwise orientations), it cannot be explained by asymmetric resources allocation, unless the allocation can be made dynamic, based on orientation expectation, induced by a cue stimulus. 
Theoretically, one can argue that the differences we found in TAE strength can result from the difference in adaptor sequences in the different conditions, regardless of built-up expectations. For example, the same-expectation condition contains on average longer sequences of adaptors with matching orientation. Accordingly, the larger TAE in the same-expectation condition (and smaller TAE in mirror-expectation condition, relative to the no-expectation condition) could be explained by the presence of longer sequences of equal orientation adaptors preceding a test. Such explanation suggests that adaptors effects do not sum-up linearly (regardless of adaptor sign), but that the strengthening effect of adaptors with an orientation that matches the orientation of the last adaptor is higher than the mitigating effect of adaptors with opposite orientation. This explanation seems unlikely given previous results (Magnussen & Johnsen, 1986) and the results presented here. Magnussen & Johnsen (1986) showed, using successive adaptation (adapt-recover-adapt, though on a longer time scale, of minutes), increased TAE by previous adaptation to the same tilt direction and decreased TAE with previous adaptation having an opposite tilt. While it is difficult to quantitatively compare these opposing effects, the additivity assumption seems to hold. Our results, presented in Tables 115343, strongly support additivity (Equations 1, 2). Most importantly, an analysis performed separately for each group, showed differences in TAE strength according to the expectation (larger TAE in “Expected same” group and lower TAE in “Expected mirror” group), regardless of the stimulus history. 
Of relevance to our experiment are two recent studies that have examined the influence of prior presentations of Gabor stimuli (stimulus history) on the perception of subsequently viewed test Gabor (Chopin & Mamassian, 2012; Fischer & Whitney, 2014). These studies found in addition to the typical repulsive perceptual bias away from stimuli presented in the recent past (negative aftereffects), also an attractive bias towards stimuli presented more remotely in the past. Fischer and Whitney (2014) used an orientation matching task to show that perception is biased towards the orientation of a preceding Gabor that appeared 5–15 s before the test Gabor. They interpret their results as a “serial dependency” effect used to promote visual stability over time. Chopin and Mamassian (2012) used both a binocular rivalry test and measurements of TAE and found attractive biases in perception towards a Gabor stimulus that appeared a few minutes in the past (but also see Fischer & Whitney, 2014). In accordance with our interpretation, these studies suggest that owing to the natural world being mostly stable and continuous over time, past stimulus statistics are used to derive expectations regarding the next percept. However, in contrast to those studies in which long-term stimulus statistics created expectations that directly changed subsequent perception, in our experiment, stimulus statistics were used to learn over time the Gabor temporal contingencies (second order statistics). These learned contingencies were later used to form expectation regarding the next stimulus. This dynamic expectation was short termed, based on the recently occurred cue that was presented only 600 ms prior to the test Gabor. Therefore, our repulsive expectation effect is consistent with adaptation short-term effects being also repulsive. According to the predictive coding framework, repulsive effects can be explained by subtraction or “explaining away” of the activity of the predicted stimulus from the activity of the present stimulus. Perhaps when prediction regarding subsequent stimulus is made too far in the past, (i.e., based on stimuli presented more than 5 s in the past, Chopin & Mamassian, 2012; Fischer & Whitney, 2014) the low-level memory of the predicted stimulus decays, and no longer interacts with the low-level representation of the currently present stimuli (no overlap in the activity). In such cases, there is no subtraction of the predicted stimulus activity and accordingly no repulsive biases. 
In contrast to the effects found in the orientation cue experiment, in the color cue experiment, the second-order orientation expectation resulting from the color orientation contingency did not affect the observers' perceived orientation. We believe that this lack of effect is due to the absence of first-order orientation expectation (from orientation adaptation) in this experiment. This suggests that second-order expectation can influence orientation perception, only when first-order orientation expectation exists. It is possible that color cannot produce orientation expectation. However, this seems unlikely since grating orientation was made contingent on color when the two were presented simultaneously (Held & Shattuck, 1971). 
An interesting questing is whether the TAE modulations we found are due to a genuine change in observer's perceived orientation, like in typical vision adaptation, or due to a change in the observer's decisional process. Our results show a translation of the psychometric functions without a change in slope (data not shown), thus ruling out effects due to changes of uncertainty level. It is still possible that observers change their decision criterion (their subjective vertical) according to condition, but distinguishing between perceptual biases and decision biases is not easy, if at all possible (i.e., distinguishing between a shift in the decision value and a shift of the decision axis). However, our experimental structure makes a decision bias seem less likely. The stimulus expectations were formed dynamically and were changed randomly from trial to trial according to the preceding Gabor cue and the specific contingencies of the block type (either “Expected same orientation” or “Expected mirror orientation”). Randomly interleaving different stimulus expectations, in contrary to using a specific single expectation, cannot be implemented by a simple shift in the decision criterion Morgan, 2013; Morgan, 2014), even when the different stimuli are clearly marked (Zak, Katkov, Gorea, & Sagi, 2012). Thus, making it more complicated for the observer to use a decisional bias. 
The design of our study enabled us to examine adaptation dynamics in conditions that were rarely studied before. Most past work on orientation adaptation focuses on a single prolonged exposure to a stimulus with a specific orientation. Only a few recent studies have tested how perception is affected by the statistics of recent stimulus history (Dahmen, Keating, Nodal, Schulz, & King, 2010; Price & Prescott, 2012). Here we examined how adaptation to multiple exposures of clockwise and counterclockwise stimuli accumulates over time to affect perception of subsequent test stimulus. We found that TAE measurements were compatible with a model that assumes linearity in the temporal integration of adaptor effects on the final perceived TAE. Since the influence of the third-back adaptor (∼2.5 s before the test) was less than 8% of the average final effect, the influences of more remote adaptors are estimated to be negligible (∼4 s before the test), and to be averaged out due to the sign randomization (CW/CCW). The linear model was further expanded to include the additional variable of orientation expectation that was found to interact additively with the TAE. 
Acknowledgments
This work was supported by the Basic Research Foundation administered by the Israel Academy of Sciences and Humanities. We thank the reviewers and the editor, Professor Mark Georgeson, for critical comments that significantly improved the paper, and Dr. Mikhail Katkov for comments on an earlier version of the paper. 
Commercial relationships: none. 
Corresponding author: Dov Sagi. 
Email: dov.sagi@weizmann.ac.il. 
Address: Department of Neurobiology, Weizmann Institute of Science, Rehovot, Israel. 
References
Adams, W. J, Graf, E. W, & Ernst, M. O. (2004). Experience can change the'light-from-above'prior. Nature Neuroscience, 7 (10), 1057–1058.
Bonds, A. B. (1991). Temporal dynamics of contrast gain in single cells of the cat striate cortex. Visual Neuroscience, 6 (03), 239–255.
Box, G. E. P, & Tiao, G. C. (1992). Bayesian inference in statistical analysis (Vol. 40). Hoboken, NJ: John Wiley & Sons, doi.org/10.1002/9781118033197.
Brayanov, J. B, & Smith, M. A. (2010). Bayesian and “anti-Bayesian” biases in sensory integration for action and perception in the size–weight illusion. Journal of Neurophysiology, 103 (3), 1518–1531.
Campbell, F. W, & Maffei, L. (1971). The tilt after-effect: a fresh look. The Physiological Laboratory, 11(November 1970), 833–840, doi.org/10.1016/0042-6989(71)90005-8.
Chaudhuri, A. (1990). Modulation of the motion aftereffect by selective attention. Nature, 344 (6261), 60–62.
Chopin, A, & Mamassian, P. (2012). Predictive properties of visual adaptation. Current Biology, 22 (7), 622–626.
Dahmen, J. C, Keating, P, Nodal, F. R, Schulz, A. L, & King, A. J. (2010). Adaptation to stimulus statistics in the perception and neural representation of auditory space. Neuron, 66 (6), 937–948.
De Baene, W, & Vogels, R. (2010). Effects of adaptation on the stimulus selectivity of macaque inferior temporal spiking activity and local field potentials. Cerebral Cortex, 20 (9), 2145–2165.
Dekel, R, & Sagi, D. (2015). Tilt aftereffect due to adaptation to natural stimuli. Vision Research, 117, 91–99.
Dong, X, Gao, Y, Lv, L, & Bao, M. (2016). Habituation of visual adaptation. Scientific Reports, 6, 19152, http://doi.org/10.1038/srep19152.
Dragoi, V, Sharma, J, & Sur, M. (2000). Adaptation-induced plasticity of orientation tuning in adult visual cortex. Neuron, 28 (1), 287–298, doi.org/10.1016/S0896-6273(00)00103-3.
Felsen, G, Shen, Y. S, Yao, H, Spor, G, Li, C, & Dan, Y. (2002). Dynamic modification of cortical orientation tuning mediated by recurrent connections. Neuron, 36 (5), 945–954, doi.org/10.1016/S0896-6273(02)01011-5.
Festman, Y, & Ahissar, M. (2004). Attentional states and the degree of visual adaptation to gratings. Neural Networks: The Official Journal of the International Neural Network Society, 17 (5-6), 849–60, doi.org/10.1016/j.neunet.2004.02.006.
Fischer, J, & Whitney, D. (2014). Serial dependence in visual perception. Nature Neuroscience, 17 (5), 738–743.
Foley, J. M, & Boynton, G. M. (1993). Forward pattern masking and adaptation: Effects of duration, interstimulus interval, contrast, and spatial and temporal frequency. Vision Research, 33 (7), 959–980, doi.org/10.1016/0042-6989(93)90079-C.
Fründ, I, Haenel, N. V, & Wichmann, F. A. (2011). Inference for psychometric functions in the presence of nonstationary behavior. Journal of Vision, 11 (6): 16, 1–19, doi: 10.1167/11.6.16. [PubMed] [Article]
Geisler, W. S. (2008). Visual perception and the statistical properties of natural scenes. Annual Review of Psychology, 59, 167–192.
Gibson, J. J, & Radner, M. (1937). Adaptation, after-effect and contrast in the perception of tilted lines. I. Quantitative studies. Journal of Experimental Psychology, 20 (5), 453.
Greenlee, M. W, Georgeson, M. A, Magnussen, S, & Harris, J. P. (1991). The time course of adaptation to spatial contrast. Vision Research, 31 (2), 223–236.
Greenlee, M. W, & Magnussen, S. (1988). Interactions among spatial frequency and orientation channels adapted concurrently. Vision Research, 28 (12), 1303–1310.
Haak, K. V, Fast, E, Bao, M, Lee, M, & Engel, S. A. (2014). Four days of visual contrast deprivation reveals limits of neuronal adaptation. Current Biology, 24 (21), 2575–2579.
Haijiang, Q, Saunders, J. A, Stone, R. W, & Backus, B. T. (2006). Demonstration of cue recruitment: Change in visual appearance by means of Pavlovian conditioning. Proceedings of the National Academy of Sciences, USA, 103 (2), 483–488.
Harris, J. P, & Calvert, J. E. (1989). Contrast, spatial frequency and test duration effects on the tilt aftereffect: Implications for underlying mechanisms. Vision Research, 29 (1), 129–135, doi.org/10.1016/0042-6989(89)90179-X.
He, S, & MacLeod, D. I. A. (2001). Orientation-selective adaptation and tilt after-effect from invisible patterns. Nature, 411 (6836), 473–476.
Held, R, & Shattuck, S. R. (1971). Color-and edge-sensitive channels in the human visual system: Tuning for orientation. Science, 174 (4006), 314–316.
Hosoya, T, Baccus, S. A, & Meister, M. (2005). Dynamic predictive coding by the retina. Nature, 436 (7047), 71–77.
Lochmann, T, Ernst, U. A, & Deneve, S. (2012). Perceptual inference predicts contextual modulations of sensory responses. Journal of Neuroscience, 32 (12), 4179–4195, http://doi.org/10.1523/JNEUROSCI.0817-11.2012.
Magnussen, S, & Johnsen, T. (1986). Temporal aspects of spatial adaptation. A study of the tilt aftereffect. Vision Research, 26 (4), 661–72, doi.org/10.1016/0042-6989(86)90014-3.
McCollough, C. (1965). Color adaptation of edge-detectors in the human visual system. Science, 149 (3688), 1115–1116.
Morgan, M. (2013). Sustained attention is not necessary for velocity adaptation. Journal of Vision, 13 (8): 26, 1–11, doi: 10.1167/13.8.26. [PubMed] [Article]
Morgan, M. J. (2014). A bias-free measure of retinotopic tilt adaptation. Journal of Vision, 14 (1): 7, 1–9, doi: 10.1167/14.1.7. [PubMed] [Article]
Pinchuk-Yacobi, N, Harris, H, & Sagi, D. (2016). Target-selective tilt aftereffect during texture learning. Vision Research, 124, 44–51.
Price, N. S. C, & Prescott, D. L. (2012). Adaptation to direction statistics modulates perceptual discrimination. Journal of Vision, 12 (6): 32, 1–17, doi: 10.1167/12.6.32. [PubMed] [Article]
Rao, R. P. N, & Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2 (1), 79–87, doi.org/10.1038/4580.
Saad, E, & Silvanto, J. (2013). How visual short-term memory maintenance modulates subsequent visual aftereffects. Psychological Science, 24 (5), 803–808, doi.org/10.1177/0956797612462140.
Sekuler, R, & Pantle, a. (1967). A model for after-effects of seen movement. Vision Research, 7 (5), 427–439. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/5613304
Spivey, M. J, & Spirn, M. J. (2000). Selective visual attention modulates the direct tilt aftereffect. Perception & Psychophysics, 62 (8), 1525–1533.
Stocker, A, & Simoncelli, E. P. (2006). Sensory adaptation within a Bayesian framework for perception. Advances in Neural Information Processing Systems, 18, 1291–1298.
Thompson, P, & Burr, D. (2009). Visual aftereffects. Current Biology, 19, 11–14, doi.org/10.1016/j.cub.2008.10.014.
Yehezkel, O, Sagi, D, Sterkin, A, Belkin, M, & Polat, U. (2010). Learning to adapt: Dynamics of readaptation to geometrical distortions. Vision Research, 50 (16), 1550–1558, doi.org/10.1016/j.visres.2010.05.014.
Zak, I, Katkov, M, Gorea, A, & Sagi, D. (2012). Decision criteria in dual discrimination tasks estimated using external-noise methods. Attention, Perception, & Psychophysics, 74 (5), 1042–1055.
Appendix
Equation A1
Image not available
Linear equation constructed from TAE values of Table 1. The sign “+” is assigned for a CCW adaptor, and the sign “−” is assigned for a CW adaptor. The term w0 stands for the mean perceived vertical when there is no adaptation, and wi stands for the weight of the influence of the i-back adaptor. The equation has eight measurements and four unknowns, thus overdetermined. The estimated solution, found using least squares method, gives w0 = −0.05, w1 = 1.31, w2 = 0.28, w3 = 0.16. This linear model explains 99% of the variance in the data (obtained by Display FormulaImage not available ).  
Equation A2
Image not available
Linear equation constructed from TAE values of Table 2. The sign “+” is assigned when the orientation of the adaptor or the expected orientation is CCW, and the sign “−” is assigned when the orientation of the adaptor or the expected orientation is CW. wi stands for the weight of the influence of the i-back adaptor, and we stands for the weight of the influence of the orientation expectation. The equation has six measurements and four unknowns. Solving it using least squares gives w1 = 1.11, w2 = 0.24, w3 = 0.09, we = 0.4. This linear model explains 99% of the variance in the data. 
Equation A3
Image not available
Linear equation for the “Expected mirror orientation” group, constructed from TAE values of Table 3 (format the same as in Equation A2). Solving the equation using least squares gives w1 = 1.53, w2 = 0.32, w3 = 0.21, we = 0.28. This linear model explains 99% of the variance in the data. 
Figure A1
 
Single observer measurements of TAE based on “No expectation” data. No systematic differences were found between the TAE calculated for trials with history types corresponding to the “Expected same orientation” blocks and the TAE calculated for trials with history types corresponding to the “Expected mirror orientation” blocks.
Figure A1
 
Single observer measurements of TAE based on “No expectation” data. No systematic differences were found between the TAE calculated for trials with history types corresponding to the “Expected same orientation” blocks and the TAE calculated for trials with history types corresponding to the “Expected mirror orientation” blocks.
Figure 1
 
Orientation cue experiment: stimuli and block design. Oriented Gabor patches were presented in successive pairs. In each pair, the first Gabor member was randomized (±20° relative to vertical). Observers monitored the stimulus stream for occasional test pairs, which were marked by four peripheral crosses in the second member of the pair. (A) Adaptor pairs in which the two Gabor members had the same orientation. (B) Adaptor pairs in which the two Gabor members had a mirror orientation. (C) Test pairs in which the second Gabor member was oriented around the vertical. Observers were requested to report whether the Gabor is tilted clockwise or counterclockwise. (D) Experimental block design. “Expected same orientation” blocks contained the same orientation Gabor pairs. “Expected mirror orientation” blocks contained mirror orientation Gabor pairs. “No expectation” blocks contained an equal number of the same orientation and mirror orientation Gabor pairs. In all blocks, a third of the pairs were test pairs. All pairs in each block were randomly intermixed.
Figure 1
 
Orientation cue experiment: stimuli and block design. Oriented Gabor patches were presented in successive pairs. In each pair, the first Gabor member was randomized (±20° relative to vertical). Observers monitored the stimulus stream for occasional test pairs, which were marked by four peripheral crosses in the second member of the pair. (A) Adaptor pairs in which the two Gabor members had the same orientation. (B) Adaptor pairs in which the two Gabor members had a mirror orientation. (C) Test pairs in which the second Gabor member was oriented around the vertical. Observers were requested to report whether the Gabor is tilted clockwise or counterclockwise. (D) Experimental block design. “Expected same orientation” blocks contained the same orientation Gabor pairs. “Expected mirror orientation” blocks contained mirror orientation Gabor pairs. “No expectation” blocks contained an equal number of the same orientation and mirror orientation Gabor pairs. In all blocks, a third of the pairs were test pairs. All pairs in each block were randomly intermixed.
Figure 2
 
Color cue experiment stimuli. The layout of the pairs is the same as in Figure 1. (A) Adaptor pairs with a particular color-orientation contingency. (B) Adaptor pairs with the mirror contingency. (C) Test pairs. The task is the same as in Figure 1C. Experimental blocks contained either contingency pairs (“Expected” blocks) or contingency and mirror pairs (“No expectation” blocks), in addition to test pairs, similar to Figure 1D.
Figure 2
 
Color cue experiment stimuli. The layout of the pairs is the same as in Figure 1. (A) Adaptor pairs with a particular color-orientation contingency. (B) Adaptor pairs with the mirror contingency. (C) Test pairs. The task is the same as in Figure 1C. Experimental blocks contained either contingency pairs (“Expected” blocks) or contingency and mirror pairs (“No expectation” blocks), in addition to test pairs, similar to Figure 1D.
Figure 3
 
Orientation expectation modulates the perceived vertical orientation only in the presence of adaptation. Individual data of the perceived vertical (PV) difference (averaged across all block repetitions) measured in the “Expected” condition as a function of the “No expectation” condition. Observers' perceived vertical (PV) was calculated separately for two conditions, according to the different first pair members (orientation: −20 or +20; color: red or green). The difference in PV was quantified as the difference in PV between these two conditions divided by two. In the orientation cue experiment, when observers expected the orientation of the second Gabor in the pair to be the same (black rhombuses) or the mirror (white rhombuses) of the first Gabor, their TAE (PV difference due to orientation adaptation) was increased or decreased (respectively) relative to their TAE when they had no expectation. In the color cue experiment, the orientation-expectation resulting from the color-orientation contingency did not affect the observers' perceived vertical.
Figure 3
 
Orientation expectation modulates the perceived vertical orientation only in the presence of adaptation. Individual data of the perceived vertical (PV) difference (averaged across all block repetitions) measured in the “Expected” condition as a function of the “No expectation” condition. Observers' perceived vertical (PV) was calculated separately for two conditions, according to the different first pair members (orientation: −20 or +20; color: red or green). The difference in PV was quantified as the difference in PV between these two conditions divided by two. In the orientation cue experiment, when observers expected the orientation of the second Gabor in the pair to be the same (black rhombuses) or the mirror (white rhombuses) of the first Gabor, their TAE (PV difference due to orientation adaptation) was increased or decreased (respectively) relative to their TAE when they had no expectation. In the color cue experiment, the orientation-expectation resulting from the color-orientation contingency did not affect the observers' perceived vertical.
Figure A1
 
Single observer measurements of TAE based on “No expectation” data. No systematic differences were found between the TAE calculated for trials with history types corresponding to the “Expected same orientation” blocks and the TAE calculated for trials with history types corresponding to the “Expected mirror orientation” blocks.
Figure A1
 
Single observer measurements of TAE based on “No expectation” data. No systematic differences were found between the TAE calculated for trials with history types corresponding to the “Expected same orientation” blocks and the TAE calculated for trials with history types corresponding to the “Expected mirror orientation” blocks.
Table 1
 
TAE measurements based on data from the “No expectation” blocks.
Table 1
 
TAE measurements based on data from the “No expectation” blocks.
Table 2
 
TAE measurements for the “Expected same orientation” group.
Table 2
 
TAE measurements for the “Expected same orientation” group.
Table 3
 
TAE measurements for the “Expected mirror orientation” group.
Table 3
 
TAE measurements for the “Expected mirror orientation” group.
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×