Journal of Experimental Psychology:

Learning, Memory, and Cognition

© 2019 American Psychological Association 2020, Vol. 46, No. 3, 455– 475
0278-7393/20/$12.00 http://dx.doi.org/10.1037/xlm0000735

Perceptual Learning Produces Perceptual Objects

Michael J. Wenger and Stephanie E. Rhoten
The University of Oklahoma

In their seminal study of chess expertise, Simon and Chase (Chase & Simon, 1973; Simon & Chase,
1973) proposed that perceptual learning was a necessary component of skill acquisition. In their view,
acquisition of skill results from the strategic use of learning at multiple levels to adaptively overcome
inherent limitations. The knowledge acquired by way of perceptual learning that supported increasingly
sophisticated perceptual discrimination processes, according to Simon and Chase, was referred to as a
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

chunk. The chunk was conceptualized as a meaningful complex set of features that abstracted the notion
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of a perceptual object. Simon and Chase further suggested that meaningful combinations of chunks could
be combined to form configurations (Simon & Chase, 1973, p. 399). The present study addresses this idea
by framing the notion of a chunk in terms of two formal metatheories, one that addresses representation
(Ashby & Townsend, 1986) and one that addresses processing (Townsend & Nozawa, 1995), and tests
the prediction that perceptual learning produces organized perceptual objects (chunks). Two experiments
combine behavioral and electroencephelographic (EEG) measures to show that perceptual learning
produces (a) a shift from perceptual independence and separability to violations of separability, and (b)
shifts from limited-capacity serial processing to supercapacity parallel processing. The evidence from
both experiments is strong and consistent: perceptual learning does indeed induce chunking—the
production of perceptual objects, and the foundation of perceptual expertise.

Keywords: chunking, expertise, perceptual learning, skill acquisition

In 1963, Eleanor Gibson (Gibson, 1963) noted that the study of learning was a necessary component of skill acquisition. In their
perceptual learning had become a “healthily growing” (p. 50) field, view, acquisition of skill results from the strategic use of learning
having only three years earlier “acquired the status of an area at multiple levels to adaptively overcome inherent limitations.
worthy of separate review” (p. 29, in reference to Drever, 1960). They proposed that “specific perceptual knowledge acquired
Further, she noted that although perceptual learning must logically through long experience, stored in long-term memory, and ac-
be a component of the acquisition of higher-order skills, “[v]ery cessed by perceptual discrimination processes” (Simon & Chase,
little applied [sic] research on perceptual skills is to be found in 1973, p. 394) was necessary to alleviate the constraints of a
psychological journals” (Gibson, 1963, p. 49). More than 50 years limited-capacity short-duration memory. The knowledge acquired
later, comprehensive reviews of the literature (Kellman & Massey, by way of perceptual learning that supported increasingly sophis-
2013; Dosher & Lu, 2017; Seitz, 2017; Watanabe & Sasaki, 2015) ticated perceptual discrimination processes, according to Simon
suggest that Gibson’s observations still hold true. The literature on and Chase, was referred to as a chunk.
perceptual learning is healthy. However, it is also true that there is The chunk was conceptualized as a meaningful complex set of
relatively limited work on the explicit role of perceptual learning features that abstracted the notion of a perceptual object. Simon
in the acquisition of higher-level expertise (Kellman & Garrigan, and Chase further suggested that meaningful combinations of
2009; Kellman & Massey, 2013; Lu, Lin, & Dosher, 2016; Polat, chunks could be combined to form configurations (Simon &
2016). Chase, 1973, p. 399). Simon and Chase formalized this recursive
structure in a computational model that possessed the critical
Perceptual Learning and Perceptual Chunks characteristic of coding chunks by changing the weighted connec-
tions from input to output elements. The literature provides some
In their seminal study of chess, Simon and Chase (Chase & support for the idea that perceptual learning of any stimulus
Simon, 1973; Simon & Chase, 1973) proposed that perceptual involving more than one dimension produces effects that are
consistent with the regularities of learning and attending to objects
(as dependent conjunctions of features, e.g., Czerwinski, Light-
This article was published Online First June 20, 2019. foot, & Shiffrin, 1992; Li, Piech, & Gilbert, 2004; Li, Levi, &
Michael J. Wenger and Stephanie E. Rhoten, Department of Psychology, Klein, 2004; Schyns, Goldstone, & Thibaut, 1998). Indeed, mod-
Cellular and Behavioral Neurobiology, The University of Oklahoma.
ern conceptualizations, such as Goldstone’s notion of perceptual
This work was funded in part by National Institutes of Health Grant 1
R01 MH072795.
unitization (Goldstone, 1998, 2000), along with computational
Correspondence concerning this article should be addressed to Michael (e.g., Schyns et al., 1998) and conceptual (e.g., Kellman & Gar-
J. Wenger, Department of Psychology, Cellular and Behavioral Neurobi- rigan, 2009; Kellman & Massey, 2013) accounts of the construct
ology, The University of Oklahoma, 201 Stephenson Parkway, Suite 4100, of chunking share many structural aspects with successful models
Norman, OK 73019. E-mail: [email protected] of perceptual learning (e.g., Dosher, Jeter, Liu, & Lu, 2013;

455
 456 WENGER AND RHOTEN

Huang, Lu, & Dosher, 2012; Petrov, Dosher, & Lu, 2005; Zhaop- (O’Toole, Wenger, & Townsend, 2001; Townsend & Wenger,
ing, Herzog, & Dayan, 2003). 2015). With respect to encoded representations, the work relies on
The hierarchical/recursive structure proposed by Simon and the theoretical definitions of dependency and lack-of-separability
Chase (1973) implies a set of characteristics present in both the in general recognition theory (GRT, Ashby & Townsend, 1986;
empirical and theoretical literatures on perceptual learning. First, Townsend, Houpt, & Silbert, 2012). In that view, a perceptual
their proposal implies that learning can occur at multiple levels of object is one in which the encoded stimulus features evidence of a
representation. This is most critically true if the first-order orga- violation of perceptual independence (PI), a violation of perceptual
nization of features into perceptual objects is generalized to con- separability (PS), and/or a violation of decisional separability
figurations (and potentially) beyond. This is a characteristic of the (DS). A violation of PI would imply that encoded information
major theoretical conceptions of perceptual learning (e.g., Ahissar about the level of one dimension (e.g., the contrast of a gabor
& Hochstein, 2004; Dosher et al., 2013; Watanabe & Sasaki, 2015) patch) would be correlated with (not independent of) the encoded
and is suggested by some of the data on generalization and transfer information about another dimension (e.g., the gabor patch’s ori-
(e.g., Wang et al., 2016; Wang, Zhang, Klein, Levi, & Yu, 2012, entation) within an individual stimulus. A violation of PS would
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2014; Zhang, Cong, Klein, Levi, & Yu, 2014). Second, the notion imply that a change in the level of one dimension (e.g., an increase
This document is copyrighted by the American Psychological Association or one of its allied publishers.

that learning can occur at multiple levels of representation suggests in contrast) would change the level of encoded information about
the computational need for changing both feed-forward and feed- another dimension (e.g., an increase in perceived tilt), with this
back connections among levels of processing. This computational effect being obtained across stimuli. A violation of DS would
requirement is one that has been acknowledged in multiple theo- imply that a change in the level of one dimension (e.g., an increase
retical perspectives (see review in Schyns et al., 1998), is one that in contrast) would change the criterion for judging the state of
has been shown to be critical in maintaining stability in networks another dimension (e.g., becoming more predisposed to judge the
of neurons (e.g., Moldakarimov, Bazhenov, & Sejnowski, 2014), patch as tilted up, independent of the actual level of orientation),
and that is supported by various sources of empirical evidence with this effect also being obtained across stimuli. With respect to
(e.g., Rauss, Schwartz, & Pourtois, 2011; Ruff & Driver, 2006). the processing of encoded information, we rely on the theoretical
Third, implied in the original proposal by Simon and Chase, and characterization of processing architecture, stopping-rules, channel
elaborated more explicitly in later work (e.g., Chase & Ericsson, independence, and capacity developed in systems factorial theory
1981; Feltovich, Prietula, & Ericsson, 2006), is the notion that (SFT, Townsend & Nozawa, 1995; Little, Altieri, Yang, & Fific,
ongoing perceptual organization, or the formation of new chunks, 2017). In that view, a perceptual object is one that is processed in
is dependent on and assisted by existing representations. Support a parallel, dependent, exhaustive manner, and which evidences
for this idea can be found empirically in the Eureka phenomenon unlimited- to supercapacity processing. The advantages of using
(Ahissar & Hochstein, 1997), as well as in effects attributable to these two conceptions include rigorous mathematical definitions of
frames of reference or context (e.g., Rauss et al., 2011; Pourtois, their central constructs, well-developed and well-understood ex-
Grandjean, Sander, & Vuilleumier, 2004), and theoretically in perimental tasks, and highly refined statistical methodologies for
concepts such as Gibson’s predifferentiation (Gibson & Walk, linking theory and data.
1956) and the mnemonic encoding principle of skilled memory
theory (Chase & Ericsson, 1981; Wenger & Payne, 1995). Fourth,
Learning Perceptual Objects: GRT
according to Simon and Chase, the learning of perpetual objects
with extensive practice allowed for the inherent limitations of With respect to encoded representations, the notion of creating
declarative memory to be alleviated by a transition to procedural a perceptual object from two or more features by way of perceptual
memory. They conceived of this as stimulus-response learning, learning implies a practice-dependent shift from independence or
idealized computationally as a production, a pairing of a condition separability to violations of either or both. To test for this change, it
and an action. On this issue, Simon and Chase were consistent with is necessary to have a task that allows the perceptual state of the
the earlier conception of the stages of skill acquisition described by observer to be determined with respect to all of the features simulta-
Fitts and Posner (1967), in which the second stage was a transition neously (Ashby & Townsend, 1986; Kadlec & Townsend, 1992). The
to declarative retrieval. Repeated declarative retrieval provided the task that allows this is known as the complete identification (CID)
basis for the transition to the autonomous stage, sometimes re- paradigm. In this task, all possible combinations of each of the
ferred to as automaticity (e.g., Shiffrin & Schneider, 1977; Sch- features is presented with each requiring a unique response. For
neider & Shiffrin, 1977; Logan, 1988; Wenger, 1999), a stage example, in the experiments below, two contrast-defined features
characterized by extremely efficient responding, absent the need each are either absent or present and at either a low or high level of
for controlled attention. This conception is consistent with the contrast. In Experiment 1, this results in four possible stimulus states
empirical regularities associated with the learning and transfer of with each assigned a unique response. The data from this task are
pop-out (e.g., Ahissar, Laiwand, Kozminsky, & Hochstein, 1998; summarized as an identification/confusion matrix, with each cell
Schoups & Orban, 1996) and with contemporary computational having a corresponding set of RTs. The first of the critical measures
(e.g., Tenison & Anderson, 2016) and neurobiological (e.g., used with the response frequencies is a test of marginal response
Ashby, Ennis, & Spiering, 2007) theories. invariance (MRI), which is defined in terms of marginal responses on
each level of each dimension. MRI holds for a given level on a given
dimension if the marginal probability of identifying that level on
Defining and Detecting Perceptual Objects
that dimension is the same across levels of the other dimension
The approach taken here conceptualizes perceptual objects in (Ashby & Townsend, 1986; Silbert & Hawkins, 2016). If MRI
terms of foundational aspects of stimulus encoding and processing does not hold, then one infers a failure of PS, DS, or both. The
 PERCEPTUAL OBJECTS 457

second critical measure is a test of report independence (RI, defined as a ratio of the inverse cumulative hazard functions
originally referred to as sampling independence; see Ashby & (Chechile, 2011) for single- and double-target trials (Townsend &
Townsend, 1986). RI holds for a given stimulus if the probability Wenger, 2004b; Townsend & Eidels, 2011; Townsend, Wenger, &
of correctly identifying that stimulus is equal to the product of the Houpt, 2018). Values of this coefficient that are equal to 1 indicate
marginal probabilities of accurately identifying the level of each unlimited-capacity processing, values ⬍ 1 indicate limited capacity
component (Ashby & Townsend, 1986; Silbert & Hawkins, 2016). processing, and values ⬎ 1 indicate supercapacity processing, and
If RI does not hold, it suggests a violation of PI for that stimulus. statistical tools have been developed to assess the reliability of any
More recently, Townsend and colleagues (2012) have developed deviations from 1 (Houpt & Townsend, 2010). Inferences regarding
reaction time (RT)-based tests of MRI and RI (referred to respectively independence are guided by the inferences regarding capacity, as it
as timed MRI [tMRI] and timed RI [tRI]). On the basis of the has been shown that limited or supercapacity processing typically
inferences suggested by these tests, multidimensional Gaussian mod- results from channel dependences (Townsend & Wenger, 2004b;
els with these characteristics are fit to the data and compared (per Wenger & Townsend, 2006).
Thomas, 2001) to both more- and less-restrictive models to determine
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

whether the inferences drawn from the tests of MRI and RI are
This document is copyrighted by the American Psychological Association or one of its allied publishers.

consistent with the best-fitting model. Converging Evidence From the Timing of
Neural Events
Learning Perceptual Objects: SFT The logic and statistical practice relating data to theory in both
With respect to foundational characteristics of information process- GRT and SFT relies on converging sources of evidence (as in
ing, the notion of creating a perceptual object from two or more Bridgman, 1945). In an effort to strengthen our inferences, we
features by way of perceptual learning implies a practice-dependent sought to use a neurophysiological variable along with response
shift from serial or parallel independent, self-terminating, limited- frequencies and latencies. Following Schweickert and Mounts
capacity processing to parallel, dependent, exhaustive, unlimited- to (1998), we sought to use a feature of the EEG signals that could be
supercapacity processing (O’Toole et al., 2001; Wenger & Townsend, readily interpreted in terms consistent with the assumptions of the
2001; Townsend & Wenger, 2015). To perform strong-inference tests time-based measures in GRT and the distributional measures in
for this change, it is necessary to use a task known as the double- SFT, in particular the assumption of selective influence. The most
factorial paradigm (DFP; Townsend & Nozawa, 1995; Townsend & promising candidate in this regard was the lateralized readiness
Wenger, 2004a,2004b). The prototypical version of this task involves potential (LRP). The LRP is a negative-going waveform, measured
two features, each of which can be present or absent (Ingvalson & in central electrodes contralateral to the motor response that it
Wenger, 2005; Townsend & Nozawa, 1995). In addition, when both precedes, and is interpreted as being an indicator that sufficient
features are present, their relative speeds of processing are manipu- processing has been completed to program the motor response
lated so that each can be processed slowly or quickly. The response (Coles, 1989; Hackley & Miller, 1995; Miller & Hackley, 1992;
instruction used here is to give one response when both features are Mordkoff & Grosjean, 2001; Ray, Slobounov, Mordkoff, John-
present; otherwise a second response is given. This is referred to as an ston, & Simon, 2000). As well, the LRP should behave in accord
AND task. with the assumptions regarding processing times and observable
Inferences about processing architecture and stopping rule are RTs in GRT and SFT. The LRP is estimated by subtracting the
drawn using RT interaction contrasts calculated for the four cells of ipsilateral from the contralateral potential for each hand. Our focus
the design in which both features are present. The interaction contrasts will be on the onset time for the LRP, operationalized as the
are calculated at the level of the mean, and then at the level of the earliest time at which the LRP becomes reliably less than 0 (as in
survivor function of the RT distribution (the complement of the Kuefner, Jacques, Prieto, & Rossion, 2010; Von Der Heide,
cumulative distribution function). The signs of these two interaction Wenger, Bittner, & Fitousi, 2018). We test the hypothesis that
contrasts allow for unique inferences regarding architecture and stop- when the timing of the onset of the LRP is analyzed using the RT
ping rule, and statistical tests are available to check the signs of both tests specified for GRT and SFT, the conclusions will be consistent
(Houpt & Townsend, 2010). Inferences regarding capacity are drawn with those drawn from the analyses of the RT data, providing an
on the basis of a capacity coefficient which, for the AND task, is additional source of converging evidence for our inferences.

Figure 1. Examples (here at 40% contrast) of the four stimulus types used in Experiments 1 and 2. Dimensions
are in degrees of visual angle. (a) Target-absent, (b) bottom only, (c) top only, (d) top and bottom.
 458 WENGER AND RHOTEN
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 2. Experiment 1: Relative thresholds (THR) and false alarm rates (FAR) for the three stimulus types
for each of the four observers (Obs). Relative thresholds were calculated as threshold on block n divided by
threshold on block 1.

Experiment 1 Board of the University of Oklahoma (IRB approval number

3908).
Experiment 1 uses the theory and methods of GRT to test the
Design. Two tasks were used in Experiment 1: a complete
hypothesis that perceptual learning results in a change in encoded
identification task before and after perceptual practice, and a
representations such that they become perceptual objects. Specif-
three-down one-up staircase task augmented with catch trials for
ically, we predict that prior to perceptual learning, observers will
encode the features of a two-element stimulus as independent perceptual practice. The complete identification task was a feature-
and/or separable, both perceptually and decisionally. After percep- complete factorial design, involving the presence and absence of
tual learning, observers will encode those features as dependent two contrast-defined features, presented just above or below fix-
and/or nonseparable, perceptually or decisionally. This will be ation in the center of the stimulus. This resulted in four possible
consistent with the theoretical conception of an object from the stimuli, each of which were assigned a unique response. These
GRT perspective. Critically, these inferences should be supported stimuli were presented at two levels of (Michelson) contrast (10%
by the analyses of the response frequencies, RTs, and the onset and 50%) before practice, and three levels of contrast (final thresh-
times of the LRP. old for each participant, 10%, and 50%) after practice.1 The
three-down one-up staircase task involved the three stimuli in
Method
Participants. A total of four (all female, ages 21–24) partic- 1
It has been noted that this basic version of the complete identification
ipants were recruited from the University of Oklahoma commu- task is subject to identifiability problems (e.g., Silbert & Thomas, 2013),
nity. All were right-handed, reported normal or corrected-to- and that an expanded design along with novel statistical approaches (Soto
normal vision, normal hearing, and unencumbered use of both et al., 2015) has the ability to solve those identifiability problems. We used
the limited design because data collection had begun prior to the publica-
hands. Participants were reimbursed by gift card at the completion tion of the last of these papers. Consequently, we here rely on the logic of
of their participation. The protocol and procedures for this exper- converging evidence to support our inferences, noting where necessary
iment were reviewed and approved by the Institutional Review where there may limitations to the strength of our inferences.
 PERCEPTUAL OBJECTS 459

which at least one of the contrast-defined targets was present. Table 1

Separate staircases were run for each stimulus type. Results of the Regression Analyses on Relative Thresholds and
Materials. The stimuli were created from a digital version of False Alarm Rates From the Practice Sessions for all Four
the line drawing of the Cheshire Cat from the original edition of Observers and all Three Target-Present Stimuli in Experiment 1
Alice in Wonderland (Carroll, 1865). Figure 1 presents examples
of each type of stimulus. The units of the dimensions noted in Relative threshold False alarm rate
Figure 1 are degrees of visual angle at a constrained viewing Observer Stimulus Intercept ␤ˆ R2
Intercept ␤ˆ R2
distance of 78 cm. The target-absent stimulus was created from a
1 Top .13 ⫺1.25 .93 ⫺2.18 .30 .58
randomly selected and arranged set of 15 ⫻ 15 pixels from the Bottom ⫺.07 ⫺1.30 .91 ⫺1.80 .14 .57
forehead of the cat. Each of the two targets was created from the Both ⫺.25 ⫺1.17 .90 ⫺2.32 .35 .48
left eye of the cat, rotated by 90°. Pattern masks were created by 2 Top ⫺.71 ⫺.45 .36 ⫺2.34 .12⫹ .34
Bottom ⫺.15 ⫺1.21 .80 ⫺3.05 .51 .66
scrambling (in 15 ⫻ 15 pixel tiles) the stimulus presented on each Both ⫺.81 ⫺.59 .46 ⫺3.07 .51 .62
trial. Gaussian noise (␮ ⫽ 0, ␴ ⫽ 0.10) was added to all stimuli 3 Top 1.07 ⫺1.82 .91 ⫺2.44 .33 .42
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and masks. Contrast for all stimuli and masks ranged from 1% to Bottom .38 ⫺1.53 .91 ⫺2.03 .18 .17
⫺1.44 ⫺2.13
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Both .34 .92 .20 .15

60%, in steps of 0.1%. All stimuli were presented on a 55-cm 4 Top .91 ⫺1.59 .87 ⫺2.00 .25 .58
(diagonal) gray-scale CRT, and all responses were made using a Bottom .79 ⫺.93 .62 ⫺1.70 .10⫹ .36
four-button response box with button presses timed to ⫾1 ms. Both 1.22 ⫺1.55 .77 ⫺2.07 .26 .48
៮
X .78 .45
Both the complete identification trials and staircase practice trials
were presented and timed using E-prime (Psychology Software Note. All ␤ˆ values were reliably different from 0 (p ⬍ .05) except for
Tools, Sharpsburg PA). those marked with ⫹.
Procedure. The experiment was performed across 14 days,
and consisted of one day of a baseline (BL) complete identification
task, followed by 10 days of practice using the three-down one-up
was divided into six blocks of 250 trials per block, with each block
modified staircase procedure, followed by an endline (EL) set of
involving only one of the three target-present stimuli. A short (eyes
three days of sessions with the complete identification task. Con-
current EEG was collected during the BL and EL complete iden- closed) break occurred between blocks. All participants began the
tification sessions. All testing took place in light- and sound- first block of practice with each stimulus type at 60% contrast.
shielded chambers. Testing with concurrent EEG occurred in an Each subsequent block of each stimulus type began at 110% of the
identical chamber that was also electromagnetically shielded (i.e., threshold from the previous block with that stimulus type. Contrast
a Faraday cage). All sessions began with a five minute period of decrements and increments during each staircase were at 10% of
dark adaptation and all testing occurred between 8:00 a.m. and the current contrast level. Catch trials involved presentation of a
5:00 p.m. local time. Participants performed all trials with their target-absent stimulus at the current level of contrast. The se-
heads in a chin rest positioned 78 cm from the monitor. quence and timing of events were the same as those in the com-
Each session of the complete identification task required ap- plete ID. Participants indicated the perceived presence of a stim-
proximately 90 min, including time for EEG setup and clean-up. ulus by pressing the button under the index finger of their
The BL session involved 300 presentations of each test stimulus, dominant hand and indicated the perceived absence of a stimulus
each at two levels of contrast: 10% and 50%. The EL sessions each by pressing the button under the index finger of their nondominant
involved 300 presentations of each test stimulus, each at three hand. No feedback was provided.
levels of contrast: 10%, 50%, and each participant’s final threshold EEG data were collected using a 128-channel EGI system (EGI
level of contrast from the perceptual practice sessions. Each ses- Philips, Eugene OR). EEG was recorded continuously with hard-
sion was divided into blocks of 400 trials, with a short (5 min) ware filters set from 0.1 to 100 Hz, a sampling rate of 1,000 Hz,
break between blocks, during which impedances were checked and and an online vertex reference. Impedances were kept ⱕ50 K⍀ for
adjusted as needed. Order of trials was randomized by session for the entire session. Continuous EEG was epoched around the cues
each participant. (⫺800 to 1500 ms). Data were preprocessed and analyzed using
Each trial of the complete identification task included the fol-
EEGLab (Delorme & Mackeig, 2004). Data were first inspected
lowing events. A small dot signaled the beginning of each trial,
visually, and bad channels were deleted. Artifacts were then re-
which was self-initiated by the observer by pressing the key under
jected and any additional bad channels were deleted on the basis of
the index finger of their dominant hand. A fixation cross was
probability using EEGLab. The data were then band-pass filtered
presented centrally for a random period on each trial, determined
by an exponential distribution with a mean of 700 ms, censored at (1 to 90 Hz), using a notch filter centered at 60 Hz (59 – 61 Hz).
400 and 1000 ms. This was followed by three randomly selected The data were then epoched, and independent components analysis
pattern masks, at the level of contrast for the test stimulus, each (ICA) was performed and components corresponding to artifacts
presented for 25 ms. The test stimulus was then presented for 25 were removed. The data were then low-pass filtered at 8 Hz to
ms, followed by a set of three randomly selected pattern masks. determine the LRP. A set of six electrodes, three to the left of
The participant then was allowed up to 2 s to respond. Participants midline and three to the right, were used to determine the LRP,
were instructed to respond as quickly and as accurately as possible, following previously used procedures (Kuefner et al., 2010; Von
and no feedback was provided. Der Heide et al., 2018). The start-time for the LRP was defined as
Each session of perceptual practice required between 45 and the time point at which the negative-going difference wave was
60 min; no concurrent EEG was collected. Each session of practice reliably different from 0.
 460 WENGER AND RHOTEN

Results however, is that the literature on the analysis of complete identifica-

tion data is a vibrant and evolving one, with regular additions and
Practice. Thresholds in each block for each stimulus type
evolutions (e.g., Silbert & Hawkins, 2016; Soto, Vucovich, Mus-
were determined by calculating the geometric mean of the contrast
grave, & Ashby, 2015), including the extension of GRT to RTs
level associated with last 15 reversals in each block. False alarm
(Townsend et al., 2012). The approach taken here is to take advantage
rates were calculated for each block. To account for individual
of the strengths of a variety of approaches by combining them, with
differences, thresholds were converted to relative thresholds for
the goal of using the inferences drawn from each approach as sources
each observer by dividing the threshold in block n by that observ-
of potentially converging evidence (see, e.g., Cornes, Donnelly, God-
er’s threshold in block 1. Reliability of change for both threshold
win, & Wenger, 2011; Von Der Heide et al., 2018, for a similar
and false alarm rates was assessed using linear regression on
log-transformed values as a function of log-transformed block approach).
number. Plots of threshold and false alarm rates as a function of Figure 3 illustrates the manner in which our analyses proceeded.
block are presented for each of the four observers in Figure 2 and One path in this process was used for preliminary inferences
the results of the regression analyses are presented in Table 1. regarding PS and DS, and a second was used for preliminary
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Practice produced reliable decreases in threshold for all four ob- inferences regarding PI. The critical quantities for the first path
This document is copyrighted by the American Psychological Association or one of its allied publishers.

servers for each stimulus type. Practice also produced reliable were the following. The first is a test of marginal response invari-
increases in false alarm rates for all four observers and for all ance (MRI) for the response frequencies, and its corresponding test
stimulus types, with the exception of the single-target top stimulus in RTs (timed MRI or tMRI; Townsend et al., 2012). MRI is
for observer 2 and the single-target bottom stimulus for observer 4. defined in terms of marginal responses on each level of each
Baseline and endline identification. dimension. MRI holds for a given level on a given dimension if the
Data analysis. Since the inception of GRT (Ashby & Townsend, marginal probability of identifying that level on that dimension is
1986), there have been a range of approaches to analyzing data with the same across levels of the other dimension (Ashby &
respect to inferences regarding PI, PS, and DS, with a range of Townsend, 1986; Silbert & Hawkins, 2016). If MRI and tMRI do
strengths and weaknesses noted for each (e.g., Silbert & Thomas, not hold, then this suggests a failure of PS, DS, or both. The
2013; Thomas, 2001, 2003; Thomas & Silbert, 2014). A critical point, second set of quantities are the marginal measures of sensitivity

Figure 3. Process for deriving inferences from the complete identification data. Boxes shaded in gray represent
preliminary inferences that are tested/confirmed using hierarchical model fitting.
 PERCEPTUAL OBJECTS 461

(d=) and criterion (c) as defined in signal detection theory (Green All of the preliminary inferences drawn to this point were then
& Swets, 1966; Macmillan & Creelman, 2005). Equality of these used to guide hierarchical model fitting (per Thomas, 2001).
marginal measures across the levels of the other dimension (e.g., Specifically, the set of preliminary inferences regarding PI, PS,
marginal d= for the top feature across the two levels of the bottom) and DS suggested an hypothesized model. This hypothesized
is tested using 95% confidence intervals (Gourevitch & Galanter, model was used to specify a hierarchy of possible models, starting
1967). Inequality of the marginal d=s suggests a possible failure of with the simplest possible model (PI, PS, and DS all holding). The
PS and inequality of the marginal cs suggests a possible failure of most complex model in any hierarchy was more complex than the
DS. The critical quantities on the second path is a test of report hypothesized model, and was constructed by relaxing one assump-
independence (RI, originally referred to as sampling indepen- tion of the hypothesized model. In addition, in cases in which the
dence; see Ashby & Townsend, 1986) and its corresponding test hypothesized model contained either a violation of PI or a viola-
for RTs (timed RI or tRI, Townsend et al., 2012). RI holds for a tion of DS, an alternative model was constructed using a violation
given stimulus if the probability of correctly identifying that stimulus of the alternative (Silbert & Thomas, 2013). In cases in which the
is equal to the product of the marginal probabilities of accurately hypothesized model contained both violations of PI and DS, alter-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

identifying the level of each component (Ashby & Townsend, 1986; native models in which only PI or only DS were violated were also
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Silbert & Hawkins, 2016). If RI and tRI do not hold, it suggests a fit. Finally, a completely unconstrained (all parameters free) was
violation of PI for that stimulus. Here it should be noted that we also fit to the data. The best model (based on a ␹2 statistic
applied the tests of tMRI and tRI to both the RT data and the start calculated on the negative log likelihoods of each model, per
times of the LRPs. Thomas, 2001) selected in this process was used for the final set of
Prior to analysis, the RT data were censored at 200 and 2000 ms. inferences.
Any LRP start times that were less than 100 ms before the Tests of MRI and tMRI. Table 2 presents a summary of the
observed RT were deleted prior to analysis. This eliminated less inferences for all of the analyses of the data from Experiment 1. To
than 1% of the observations for each observer. Overall correlations conserve space, we present the results only for the cases in which
between RTs and LRP start times for the four observers ranged either MRI or tMRI failed. For the tests of MRI, the critical test
from r ⫽ .78 –.89. statistic is a ␹2 (Silbert & Hawkins, 2016) and for the tests of tMRI

Table 2
Experiment 1: Tests of MRI (Response Frequencies) and TMRI (RTs and Start Times of the LRPs) for all Four Observers

Tests of MRI Tests of tMRI

D
Level of other
Obs Test Contrast Target dimension dimension ␹2
Comparison RT LRP

1 EL 10 Top Absent 5.617 Top, absent .136 .183

Present 25.734 Top, present .136 .163
Bottom Absent 1.539 Bottom, absent .053 .245
Present .375 Bottom, present .028 .193
50 Top Absent .000 Top, absent .070 .206
Present 33.689 Top, present .205 .128
2 EL Threshold Top Absent 2.347 Top, absent .079 .072
Present 21.425 Top, present .138 .024
10 Top Absent .137 Top, absent .065 .229
Present 45.405 Top, present .242 .390
Bottom Absent .852 Bottom, absent .041 .187
Present .833 Bottom, present .326 .235
50 Top Absent 6.005 Top, absent .219 .193
Present 38.487 Top, present .362 .259
Bottom Absent .791 Bottom, absent .091 .284
Present .059 Bottom, present .163 .127
3 BL 50 Top Absent .084 Top, absent .120 .128
EL Threshold Top Absent 3.607 Top, absent .162 .165
Present .012 Top, present .135 .201
10 Top Absent .118 Top, absent .260 .226
Present 3.350 Top, present .352 .135
Bottom Absent 1.769 Bottom, absent .325 .150
Present .450 Bottom, present .380 .143
50 Top Present 21.007 Top, present .366 .270
4 EL Bottom Absent 2.286 Bottom, absent .111 .204
Present .142 Bottom, present .084 .164
10 Top Absent 1.141 Top, absent .050 .029
Present 3.334 Top, present .111 .400
50 Top Absent 3.687 Top, absent .041 .177
Present 23.525 Top, present .198 .209
Note. Cells shaded in gray indicate cases in which MRI or TMRI failed. BL ⫽ baseline; EL ⫽ endline.
 462 WENGER AND RHOTEN

(using the RT and the LRP data), the critical test statistic is D, feature. This suggests that one of the effects of perceptual practice is
scaled by the number of observations in the cumulative distribu- to alter the representation of the top feature as a function of whether
tion functions that are being compared (Townsend et al., 2012, p. the bottom feature was present or absent. This can be interpreted as a
485). In Table 2, cells highlighted in gray indicate where the tests violation of PS in nine of the 12 possible cases. Third, in six of 12
for MRI or tMRI fail. In all but one case, MRI and tMRI (for both cases, the response criterion for the top feature was altered as a
RTs and LRPs) held for all observers at both contrast levels at BL function of the state of the bottom feature. This can be interpreted as
(before perceptual practice). However, there were numerous violations of DS in half of the 12 possible cases.
failures of both MRI and tMRI at endline (after perceptual prac-
tice), occurring most frequently at the two superthreshold levels of
Discussion
contrast and far more frequently for the top rather than the bottom
feature. This suggests that PS and/or DS may both have held prior Experiment 1 was intended to test the hypothesis that per-
to practice, but that either or both may have been violated after ceptual learning leads to the creation of perceptual objects,
practice. In addition, the tests of tMRI on the RTs and LRPs defined in terms of the GRT constructs of PI, PS, and DS.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

showed a high level of consistency. Of the 40 possible inferences, Specifically, we tested the prediction that perceptual learning
This document is copyrighted by the American Psychological Association or one of its allied publishers.

the analyses of the RTs and LRPs agreed on 35 (88%). for two arbitrary, contrast-defined features would produce a
Tests of marginal signal detection measures. Figure 4 plots shift from separability and independence to violations of per-
the differences in the marginal hit and false alarm rates, before and ceptual and/or decisional separability, and violations of percep-
after practice, at each level of contrast. The values plotted are the tual independence. Performance (as measured by detection
difference in each measure for one of the features (top, bottom) across thresholds) reliably decreased for all observers, with this being
the two levels of the other feature (present, absent). The points labeled accompanied by small but reliable increases in false alarm rates.
as Top are the differences between the measure when the bottom Prior to practice, there was consistent evidence for PS, DS, and
feature was present minus the value of the measure when the bottom PI. In contrast, after practice, there was reasonably consistent
feature was absent. Figure 5 plots the differences in the marginal evidence for violations of PS, accompanied in a number of
measures of sensitivity (d=) and criterion for the four observers at each cases by violations of DS, suggesting perceptual learning in-
level of contrast, before and after practice. Prior to practice, equality volves changes in both encoding and decision-making. There
of both marginal measures held for all four observers. However, after was, however, no strong evidence that perceptual learning pro-
practice, equality of the marginal d=s failed 10 times, with the majority duced violations of PI. Thus, the results can be interpreted in
of those failures occurring at the two suprathreshold levels of contrast terms of creating perceptual representations in which the fea-
for the top feature. In addition, after practice, equality of the marginal tures are nonseparable, rather than dependent.
cs failed eight times, with at least one failure for every observer except Critically, these conclusions were reached on the basis of three
observer 1. The majority of the failures of equality occurred for the sources of data: choice frequencies, choice RTs, and the onset time
top feature at the two suprathreshold levels of contrast. In all cases the of the LRP. Although frequencies and latencies have been used in
failures of equality involved a more liberal criterion when the other conjunction in previous work with GRT (see Townsend et al.,
feature was present relative to when it was absent. All of this suggests 2012), this is the first case to our knowledge that adds a neural
that perceptual practice resulted in possible failures of both PS and variable in a theory-based way as a source of converging evidence.
DS. Although there were points of disagreement across these three
Tests of RI and tRI. Table 3 summarizes the tests of RI and variables, the consistency was generally high. This combination of
tRI, listing only the cases in which RI or tRI failed. As can be seen, evidence allowed us to efficiently identify a set of candidate
there were no failures of any of the tests of RI prior to practice and models, and then use hierarchical model-fitting to adjudicate the
only a small number (five) of possible failures of RI or tRI after final inferences.
practice. For those five failures, four were observed at the two Finally, we should note that this approach allowed us to identify
suprathreshold levels of contrast. In addition, in four of the five individual differences as a function of learning (see also Fific,
cases, the failures were observed for the stimulus in which both Nosofsky, & Townsend, 2008). This has long been a critical
features were present. Finally, there was one observer (4) for strength of the GRT approach, allowing for examination of indi-
whom there was no evidence of any failures of RI or tRI, at either vidual differences in strategy and learning. In this case, the ap-
stage of practice or at any level of contrast. The consistency among proach revealed that observers varied with respect to the preva-
the three tests here was more limited than for the tests of MRI and lence of violations of both PS and DS, suggestive of variations in
tMRI. In four of the five cases, the tests on the response frequen- strategies across observers.
cies and RTs were in agreement with each other but in disagree-
ment with the test on the LRP start times. The source of this
Experiment 2
inconsistency is not readily apparent. In sum, there is at best
limited evidence for practice-induced violations of PI. Experiment 2 uses the theory and methods of SFT (Townsend &
Hierarchical model fits. Table 4 summarizes the results of the Nozawa, 1995; Townsend & Wenger, 2004a,2004b) to test the
hierarchical model fitting. Here it can be seen that, prior to practice, hypothesis that perceptual learning results in a change in the
there was a uniform preservation of PI, PS, and DS, for all observers, processing of encoded representations such that they become per-
at both levels of contrast. Practice produced three sets of changes. ceptual objects. Specifically, we predict that prior to practice,
First, none of the possible violations of PI suggested by the tests of RI observers will process the elements of a stimulus in ways that
and tRI were supported by the model fitting. Second, in nine of 12 would not be associated with the processing of a set of features
cases, PS was violated, and in all cases the violations were for the top bound together as an object: in serial, exhaustively, independently,
 PERCEPTUAL OBJECTS 463
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 4. Experiment 1: Differences (Diff) in marginal hit rates (HR, panels a– d) and false alarm rates (FR,
panels e– h) at baseline (BL) and endline (EL) at each level of contrast, for each observer.
 This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

464

at baseline (BL) and endline (EL) at each level of contrast.

WENGER AND RHOTEN

Figure 5. Experiment 1: Differences (Diff) in marginal sensitivity (d=, panels a– d) and criterion (c, panels e– h)
 PERCEPTUAL OBJECTS 465

Table 3
Experiment 1: Failures of Tests of RI (Response Frequencies) or
TRI (RTs and LRP Start Times)

tRI: D
Obs Test Contrast Top Bottom RI ␹ 2
RT LRP

1 Endline 10 Present Present 1.000 .051 .145

2 Baseline 50 Present Present 2.500 .016 .289
3 Baseline 50 Present Present 3.000 .065 .226
Endline Threshold Present Absent .300 .016 .240
10 Present Present 9.400 .111 .273

Counts 0 1 5
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Note. Cells shaded in gray indicate the test that suggested the failure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

and with limited capacity. After practice, observers should process

the elements quite differently, in a way that would be associated
with a set of features bound together as an object: in parallel,
exhaustively, nonindependently, and with unlimited to supercapac- Figure 6. Experiment 2: Schematic of the design for the AND version of
ity. We test this hypothesis using the same set of stimuli that were the double-factorial paradigm. Regions marked with yes and no indicate the
used in Experiment 1 with a new set of observers. correct response to the designated stimulus. F ⫽ fast, S ⫽ slow.

Method
Institutional Review Board of the University of Oklahoma (IRB
Participants. A total of five observers (three females, 20 –24 y) approval number 3908).
were recruited from the University of Oklahoma community and were Design. Two tasks were used in Experiment 2. The first was
financially compensated for their participation. All reported normal or the modified staircase procedure used in Experiment 1; this task
corrected-to-normal vision and had unencumbered use of both hands. was used for practice. The second task was a double-factorial
One of the participants discontinued participation midway through the paradigm redundant targets task, using an AND response rule
experiment. Consequently, we report only the data from the four (Townsend & Nozawa, 1995, see Figure 6). The double-factorial
participants who contributed complete data sets. The protocol and paradigm (DFP) consists of a set of two nested 2 ⫻ 2 factorials.
procedures for this experiment were reviewed and approved by the The outer factorial in this case was the presence or absence of

Table 4
Experiment 1: Final Inferences From the Hierarchical Model Fitting to the Data From Each Observer, at Each Time Point and Level
of Contrast

Preservation of
Test Obs Contrast PI PS DS Parms ln(L) AIC

Baseline 1 10 T T T 4 ⫺1453.76 2915.51

50 T T T 4 ⫺1210.74 2429.48
2 10 T T T 4 ⫺1508.57 2915.51
50 T T T 4 ⫺1336.18 2680.35
3 10 T T T 4 ⫺1472.05 2952.11
50 T T T 4 ⫺1226.02 2460.05
4 10 T T T 4 ⫺1541.55 3091.11
50 T T T 4 ⫺1299.68 2607.36
Violations 0 0 0
Endline 1 Threshold T T T 4 ⫺1502.66 3013.31
10 T F: Top T 5 ⫺1069.95 2149.90
50 T F: Top T 5 ⫺965.45 1940.90
2 Threshold T F: Top T 5 ⫺1469.53 2949.05
10 T F: Top F: Top 6 ⫺1289.52 2591.03
50 T F: Top F: Top 6 ⫺1047.88 2107.76
3 Threshold T T T 4 ⫺1245.90 2499.80
10 T F: Top F: Top 6 ⫺969.70 1957.41
50 T F: Top F: Top 6 ⫺653.46 1320.92
4 Threshold T T T 4 ⫺1541.55 3091.11
10 T F: Top F: Top 6 ⫺1205.35 2422.70
50 T F: Top F: Top 6 ⫺1205.33 2424.65
Violations 0 9 6
 466 WENGER AND RHOTEN
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 7. Relative thresholds and false alarm rates for each of the four observers (Obs) in Experiment 2.

either the top or bottom feature. The task was structured so that a levels contrast (threshold plus the two levels used at BL). Each of
positive (yes) response was given only when both features were the DFP sessions lasted approximately 90 min, including time for
perceived as present; the three other possibilities required a no EEG setup and cleanup. Contrast levels in the DFP were blocked,
response. Nested within this factorial was a second 2 ⫻ 2 factorial, and presentation of all the possible stimuli within were randomized
relevant to the stimuli in which both features were present. Here, by observer and session. The events on each trial were identical to
a manipulation designed to selectively influence the speed of those for the complete identification task in Experiment 1, with the
processing was used (described below), such that all four possible exception that observers made either a positive or negative re-
combinations of fast (F) and slow (S) processing times were sponse, using the index finger of their dominant hand for the
represented. Trial frequencies were set (Mordkoff, Yantis, & positive response and the index finger of the nondominant hand for
Egeth, 1990; Townsend & Nozawa, 1995) to equate the number of the negative response. Concurrent EEG was collected during the
total positive and negative trials, as well as the number of single- BL and EL DFP sessions, using the equipment and parameters
and double-target trials. used in Experiment 1. Finally, all details of the practice sessions in
Materials. The test stimuli and masks used in the practice Experiment 2 were identical to those in Experiment 1.
sessions of Experiment 2 were identical to those used in the
practice sessions of Experiment 1. The test stimuli and masks used Results
in the DFP testing in Experiment 2 were presented at 30% and
60% contrast at BL and EL, with an additional set of stimuli at Practice. Relative thresholds and FA rates as a function of
each observer’s final threshold at EL. The higher levels of contrast, practice are plotted in Figure 7, and the results of the analyses of
relative to Experiment 1, were selected during pilot work to reduce the changes in relative threshold and FA rates are presented in
the number of errors, as the SFT analyses are done on correct RTs Table 5. Reliable reductions in threshold were obtained for all
and LRP start times. stimuli for all four observers, though the reductions for observer 3
Procedure. As in Experiment 1, Experiment 2 took place over were much more modest than those of the other three observers.
14 days. Day 1 was a baseline assessment of performance in the Reliable increases in FA rates were obtained for all stimuli for two
DFP at two suprathreshold levels of contrast, Days 2–11 were
practice using the modified staircase procedure,2 and Days 12–14 2
Because of schedule constraints, observer 2 completed only 15 blocks
were an endline assessment of performance in the DFP at three of practice.
 PERCEPTUAL OBJECTS 467

Table 5
Results of the Regression Analyses on Relative Thresholds and False Alarm Rates From the
Practice Sessions for all Four Observers and all Three Target-Present Stimuli in Experiment 2

Threshold False alarms

Observer Stimulus Intercept ␤ˆ R 2
Intercept ␤ˆ R2

1 Top .476 ⫺1.764 .92 ⫺2.471 .378 .58

Bottom .128 ⫺1.376 .91 ⫺2.187 .277 .50
Both ⫺.467 ⫺1.184 .86 ⫺2.187 .280 .52
2 Top .875 ⫺1.929 .92 ⫺1.834 .225 .58
Bottom .876 ⫺1.836 .92 ⫺1.890 .231 .51
Both .143 ⫺1.405 .97 ⫺3.088 .825 .61
3 Top ⫺.129 ⫺.224 .24 ⫺1.655 ⫺.179⫹ .12
Bottom .003 ⫺.231 .28 ⫺1.827 ⫺.081⫹ .02
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Both .834 ⫺.269 .18 ⫺1.764 ⫺.221⫹ .20

This document is copyrighted by the American Psychological Association or one of its allied publishers.

4 Top .247 ⫺1.327 .58 ⫺1.893 .042⫹ .02

Bottom .286 ⫺1.619 .95 ⫺1.835 .034⫹ .02
Both ⫺.342 ⫺.507 .66 ⫺2.631 .266 .34
៮
X .70 .34
Note. All ␤ˆ values were reliably different from 0 (p ⬍ .05) except for those marked with⫹.

of the observers (1 and 2). Observer 4 showed no change in FA pacity coefficient for an AND version of the DFP (Townsend &
rates, and observer 5 showed no change with the top-only stimulus. Wenger, 2004b):
Baseline and endline DFP performance.
Dependent measures. The RTs and LRP start times for correct Kt(t) ⫹ Kb(t)
Ca(t) ⫽ (2)
trials were analyzed with respect to inferences regarding process- Ktb(t)
ing architecture, stopping rule, capacity, and independence in the
following way. With respect to architecture and stopping rule, the where K ⫽ ⫺ln[F(t)] (the reverse cumulative hazard function,
data from the four shaded cells in Figure 6 were analyzed using Chechile, 2011), and the subscripts indicate the presence of the top
two interaction contrasts of the form (t), bottom (b), or both (tb) features. A value of 1 for this coeffi-
cient indicates unlimited capacity processing, values ⬍ 1 indicate
IC ⫽ (Tss ⫺ Tsf ) ⫺ (T fs ⫺ T f f ). (1) limited capacity processing, and values ⬎ 1 indicate supercapacity
T represents either the mean RT (or LRP start time) or survivor processing. Reliability of any departures from 1 was assessed
function of the RT (or LRP start time) distribution, with the using the methods described in Houpt and Townsend (2012).
survivor function being the complement of the cumulative distri- Censoring and selection of the RT and LRP start time data were
bution function, S(t) ⫽ 1 – F(t). Townsend and Nozawa (1995) done as in Experiment 1, and the correlations between the LRP
showed that when the interaction contrast at these two levels are start times and the RTs ranged from r ⫽ .56 –.69.
used together, it is possible to uniquely identify the processing Mean RTs and LRP start times. Results of the analyses of the
architecture and stopping rule (see Table 6). Reliability of the mean RTs and mean LRP start times are presented in Tables 7 and
interaction contrasts at the level of the mean was assessed using a 8, respectively. At both BL and EL, the main effects for each
2 ⫻ 2 analysis of variance, separately for each observer at each feature were reliable in all but one case, which was marginal. The
contrast level at BL and EL. Reliability of the interaction contrasts orderings associated with each of these main effects were consis-
at the level of the survivor functions was analyzed using the tent with the speed of processing manipulation being effective,
methods described in Houpt and Townsend (2010). with this being true for both RTs and LRP start times. With respect
With respect to capacity, we analyzed the RT and LRP start time to the mean RTs before practice, the interaction was reliable in
data from both the single- and double-target trials using the ca- only two cases, with both of these being at the highest level of
contrast. After practice, however, the interaction was reliable for
all observers at both suprathreshold levels of contrast, and was not
Table 6 reliably different from 0 at threshold. With respect to the mean
Allowable Inferences When the Signs of the Mean Interaction LRP start times before practice, only one interaction was reliably
Contrast (MIC) and Survivor Function Interaction Contrast different from 0 (observer 1 at 60% contrast); however, after
(SFIC) are Used Together practice, the interaction was reliable for all observers at the two
suprathreshold contrast levels and was not reliably different from
MIC SFIC Architecture Stopping-rule 0 at threshold. In all cases, for both the RTs and the LRP start
0 0 Serial Self-terminating
times, when the interaction was reliably different from 0, the sign
0 ⬍ 0¡⬎0 Serial Exhaustive of the MIC was negative. This suggests that (see Table 6), in most
⬎0 ⬎ 0 Parallel Self-terminating cases, practice produced a shift from some form of serial process-
⬍0 ⬍ 0 Parallel Exhaustive ing to some form of parallel processing, with these inferences
⬎0 ⬍ 0¡⬎0 Coactive
being supported by both the behavioral and the EEG data.
 468 WENGER AND RHOTEN

Table 7
Experiment 2: Analyses of the RTs at the Level of the Mean

Threshold 30% contrast 60% contrast

Obs Test Effect F MSE MIC F MSE MIC F MSE MIC
ⴱ ⴱⴱⴱ
1 BL Top (t) 4.53 29751 24 16.61 20825 ⫺53
Bottom (b) 4.25ⴱ 7.71ⴱⴱ
t⫻b .31 3.99ⴱ
EL t 29.55ⴱⴱⴱ 34284 7 80.74ⴱⴱⴱ 4437 ⫺65 38.43ⴱⴱⴱ 5716 ⫺53
b 31.93ⴱⴱⴱ 66.30ⴱⴱⴱ 34.82ⴱⴱⴱ
t⫻b .03 21.55ⴱⴱⴱ 10.71ⴱⴱ
2 BL Top (t) 2.24ⴱ 248981 10 10.94ⴱⴱ 73504 ⫺71
Bottom (b) 9.01 3.62ⴱ
t⫻b 1.36 6.00ⴱⴱ
EL t 13.81ⴱⴱⴱ 69154 29 19.71ⴱⴱⴱ 83080 ⫺107 22.95ⴱⴱⴱ 37725 ⫺51
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

b 10.40ⴱⴱ 3.23⫹ 16.85ⴱⴱⴱ

t⫻b 5.07ⴱ 5.51ⴱ
This document is copyrighted by the American Psychological Association or one of its allied publishers.

.09
3 BL Top (t) 3.64ⴱ 535271 ⫺13 3.74ⴱ 167296 23
Bottom (b) 4.06ⴱ 2.74⫹
t⫻b .78 1.87
EL t 34.07ⴱⴱⴱ 77527 ⫺19 52.46ⴱⴱⴱ 15032 ⫺95 60.62ⴱⴱⴱ 6599 ⫺84
b 12.97ⴱⴱⴱ 22.85ⴱⴱⴱ 98.33ⴱⴱⴱ
t⫻b .10 13.80ⴱⴱⴱ 23.81ⴱⴱⴱ
4 BL Top (t) 4.01ⴱ 23212 7 8.31ⴱⴱ 11313 4
Bottom (b) 6.48ⴱ 11.64ⴱⴱⴱ
t⫻b .04 .03
EL t 11.87ⴱⴱⴱ 24014 24 63.42ⴱⴱⴱ 9328 ⫺74 14.54ⴱⴱⴱ 82574 ⫺125
b 20.78ⴱⴱⴱ 71.41ⴱⴱⴱ 6.19ⴱ
t⫻b 1.12 13.14ⴱⴱⴱ 4.21ⴱ

Note. All effects were df ⫽ 1; MIC ⫽ mean interaction contrast (in ms).
⫹
.05 ⱕ p ⬍ .10. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.

Distributions of RTs and LRP start times. We consider first presents the results of the pairwise Kolmogorov–Smirnov tests on
the data from the double-target trials, as these will guide our the orderings. Here it can be seen that, for all pairwise comparisons
inferences regarding architecture and stopping rule. A first ques- except those involving the fast-slow and slow-fast trials, the ma-
tion has to do with the orderings on the distributions, and Table 9 jority of orderings were reliable, with those that did not reach

Table 8
Experiment 2: Analyses of the LRP Start Times at the Level of the Mean

Threshold 30% contrast 60% contrast

Obs Test Effect F MSE MIC F MSE MIC F MSE MIC
ⴱⴱⴱ ⴱⴱⴱ
1 BL Top (t) 80.45 6105 14 91.06 5118 ⫺58
Bottom (b) 114.85ⴱⴱⴱ 99.29ⴱⴱⴱ
t⫻b 1.37 14.64ⴱⴱⴱ
EL t 36.55ⴱⴱⴱ 5196 2 49.59ⴱⴱⴱ 3346 ⫺35 88.46ⴱⴱⴱ 3451 ⫺41
b 27.28ⴱⴱⴱ 25.20ⴱⴱⴱ 88.57ⴱⴱⴱ
t⫻b .03 4.77ⴱ 11.13ⴱⴱⴱ
2 BL Top (t) 193.88ⴱⴱⴱ 2787 16 27.63ⴱⴱⴱ 4521 ⫺37
Bottom (b) 103.28ⴱⴱⴱ 82.80ⴱⴱⴱ
t⫻b 1.84 2.87⫹
EL t 3.75ⴱ 13835 27 15.18ⴱⴱⴱ 17961 ⫺49 5.73ⴱ 18331 ⫺64
b 2.92⫹ 11.50ⴱⴱⴱ 11.76ⴱⴱ
t⫻b .80 5.20ⴱ 4.34ⴱ
3 BL Top (t) 40.40ⴱⴱⴱ 13355 29 27.82ⴱⴱⴱ 10444 26
Bottom (b) 35.22ⴱⴱⴱ 53.64ⴱⴱⴱ
t⫻b 1.63 1.56
EL t 4.22ⴱ 7158 ⫺2 172.05ⴱⴱⴱ 3342 ⫺31 123.63ⴱⴱⴱ 4626 ⫺43
b 5.19ⴱ 68.63ⴱⴱⴱ 142.45ⴱⴱⴱ
t⫻b .05 5.23ⴱ 14.44ⴱⴱⴱ
4 BL Top (t) 11.37ⴱⴱⴱ 9579 11 24.91ⴱⴱⴱ 6406 6
Bottom (b) 16.97ⴱⴱⴱ 32.10ⴱⴱⴱ
t⫻b .30 .28
EL t 33.84ⴱⴱⴱ 4508 15 98.23ⴱⴱⴱ 5557 ⫺40 105.56ⴱⴱⴱ 4042 ⫺41
b 54.09ⴱⴱⴱ 40.76ⴱⴱⴱ 101.82ⴱⴱⴱ
t⫻b 1.54 8.98ⴱⴱ 13.54ⴱⴱⴱ

Note. All effects were df ⫽ 1; MIC ⫽ mean interaction contrast (in ms).
⫹
.05 ⱕ p ⬍ .10. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.
 PERCEPTUAL OBJECTS 469

Table 9
Experiment 2: KS Test Statistics Testing the Orderings of the Distributions of RTs and LRP Start Times in the Double-Target Trials,
as a Function of Practice and Contrast

Before practice After practice

30 60 Threshold 30 60
Obs Comp RT LRP RT LRP RT LRP RT LRP RT LRP
ⴱⴱⴱ ⴱⴱⴱ ⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱ ⴱⴱⴱ ⴱ
1 FF vs. FS 2.79 4.94 1.69 3.96 3.32 3.02 1.71 2.56 1.56 3.08ⴱⴱⴱ
FF vs. SF 2.65ⴱⴱⴱ 4.42ⴱⴱⴱ 1.86ⴱⴱ 3.17ⴱⴱⴱ 3.70ⴱⴱⴱ 3.55ⴱⴱⴱ 1.49ⴱ 3.41ⴱⴱⴱ 1.49ⴱ 3.16ⴱⴱⴱ
FF vs. SS 3.60ⴱⴱⴱ 5.67ⴱⴱⴱ 2.56ⴱⴱⴱ 6.10ⴱⴱⴱ 4.92ⴱⴱⴱ 4.96ⴱⴱⴱ 4.84ⴱⴱⴱ 4.87ⴱⴱⴱ 4.55ⴱⴱⴱ 5.57ⴱⴱⴱ
FS vs. SF .69 1.17 .53 1.58ⴱⴱ .89 1.19 .52 1.54ⴱ .45 1.61ⴱ
FS vs. SS 2.06ⴱⴱⴱ 3.85ⴱⴱⴱ 1.71ⴱⴱ 4.57ⴱⴱⴱ 2.91ⴱⴱⴱ 4.42ⴱⴱⴱ 3.95ⴱⴱⴱ 4.55ⴱ 3.88ⴱⴱⴱ 5.24ⴱⴱⴱ
SF vs. SS 2.32ⴱⴱⴱ 4.22ⴱⴱⴱ 1.89ⴱⴱ 5.13ⴱⴱⴱ 2.74ⴱⴱⴱ 4.09ⴱⴱⴱ 4.02ⴱⴱⴱ 4.39ⴱⴱⴱ 4.17ⴱⴱⴱ 4.81ⴱⴱⴱ
2.83ⴱⴱⴱ 4.76ⴱⴱⴱ 1.73ⴱⴱ 3.59ⴱⴱⴱ 1.92ⴱⴱ 2.19ⴱⴱⴱ 1.58ⴱ 1.31⫹ 2.13ⴱⴱⴱ 1.94ⴱⴱ
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2 FF vs. FS
FF vs. SF 2.38ⴱⴱⴱ 5.62ⴱⴱⴱ 1.73ⴱⴱ 2.43ⴱⴱⴱ 2.05ⴱⴱⴱ 2.45ⴱⴱⴱ 1.49ⴱ 1.54ⴱ 2.71ⴱⴱⴱ 1.57ⴱ
This document is copyrighted by the American Psychological Association or one of its allied publishers.

FF vs. SS 3.80ⴱⴱⴱ 5.67ⴱⴱⴱ 3.10ⴱⴱⴱ 5.18ⴱⴱⴱ 2.57ⴱⴱⴱ 2.49ⴱⴱⴱ 4.33ⴱⴱⴱ 2.43ⴱⴱⴱ 5.45ⴱⴱⴱ 2.28ⴱⴱⴱ
FS vs. SF .61 1.25⫹ .55 1.73ⴱⴱ .77 .52 .82 .20 .98 .59
FS vs. SS 1.87ⴱⴱ 3.64ⴱⴱⴱ 2.35ⴱⴱⴱ 3.78ⴱⴱⴱ 1.44ⴱ 2.18ⴱⴱⴱ 3.40ⴱⴱⴱ 1.94ⴱⴱ 4.78ⴱⴱⴱ 1.90ⴱⴱ
SF vs. SS 1.99ⴱⴱⴱ 3.42ⴱⴱⴱ 2.47ⴱⴱⴱ 4.51ⴱⴱⴱ 1.79ⴱⴱ 2.19ⴱⴱⴱ 3.36ⴱⴱⴱ 1.94ⴱⴱ 4.73ⴱⴱⴱ 1.77ⴱⴱ
3 FF vs. FS 2.01ⴱⴱⴱ 2.908ⴱⴱⴱ 2.65ⴱⴱⴱ 3.320ⴱⴱⴱ 3.38ⴱⴱⴱ 2.02ⴱⴱ 1.89ⴱⴱ 3.25ⴱⴱⴱ 1.57ⴱ 4.12ⴱⴱⴱ
FF vs. SF 2.92ⴱⴱⴱ 3.441ⴱⴱⴱ 2.74ⴱⴱⴱ 3.104ⴱⴱⴱ 3.19ⴱⴱⴱ 2.43ⴱⴱⴱ 1.94ⴱⴱ 4.66ⴱⴱⴱ 1.42ⴱ 4.30ⴱⴱⴱ
FF vs. SS 3.72ⴱⴱⴱ 5.415ⴱⴱⴱ 3.80ⴱⴱⴱ 4.852ⴱⴱⴱ 4.98ⴱⴱⴱ 3.69ⴱⴱⴱ 5.07ⴱⴱⴱ 7.24ⴱⴱⴱ 5.14ⴱⴱⴱ 7.24ⴱⴱⴱ
FS vs. SF 1.10 .858 .82 .608 1.03 1.09 .55 .09 .82 .553
FS vs. SS 2.25ⴱⴱⴱ 3.040ⴱⴱⴱ 1.86ⴱⴱ 2.387ⴱⴱⴱ 2.81ⴱⴱⴱ 2.41ⴱⴱⴱ 3.85ⴱⴱⴱ 5.22ⴱⴱⴱ 4.10ⴱⴱⴱ 5.512ⴱⴱⴱ
SF vs. SS 2.41ⴱⴱⴱ 2.829ⴱⴱⴱ 1.67ⴱⴱ 2.433ⴱⴱⴱ 2.09ⴱⴱⴱ 1.73ⴱ 3.73ⴱⴱⴱ 5.11ⴱⴱⴱ 4.70ⴱⴱⴱ 5.678ⴱⴱⴱ
4 FF vs. FS 1.552ⴱ 1.948ⴱⴱ 1.688ⴱ 2.34ⴱⴱⴱ 2.012ⴱⴱ 3.958ⴱⴱⴱ 2.162ⴱⴱⴱ 1.46⫹ 2.053ⴱⴱ 2.45ⴱⴱⴱ
FF vs. SF 1.643ⴱⴱ 2.476ⴱⴱⴱ 1.713ⴱ 2.58ⴱⴱⴱ 1.714⫹ 3.423ⴱⴱⴱ 1.938ⴱⴱ 2.35ⴱⴱⴱ 1.789ⴱ 2.42ⴱⴱⴱ
FF vs. SS 2.921ⴱⴱⴱ 3.116ⴱⴱⴱ 2.282ⴱⴱⴱ 4.57ⴱⴱⴱ 3.429ⴱⴱⴱ 4.980ⴱⴱⴱ 5.116ⴱⴱⴱ 5.15ⴱⴱⴱ 5.278ⴱⴱⴱ 6.28ⴱⴱⴱ
FS vs. SF .548 1.033⫹ .728 .568 .820 .907 .522 1.35⫹ .838 .84
FS vs. SS 1.643ⴱ 1.832ⴱ 1.523⫹ 2.80ⴱⴱⴱ 1.566ⴱ 3.051ⴱⴱⴱ 3.702ⴱⴱⴱ 4.51ⴱⴱⴱ 3.605ⴱⴱⴱ 5.11ⴱⴱⴱ
SF vs. SS 2.008ⴱⴱ 1.815ⴱ 1.596⫹ 2.82ⴱⴱⴱ 2.087ⴱⴱ 2.821ⴱⴱⴱ 3.768ⴱⴱⴱ 4.68ⴱⴱⴱ 4.326ⴱⴱⴱ 5.64ⴱⴱⴱ
Note. FF ⫽ fast–fast; FS ⫽ fast–slow; SF ⫽ slow–fast; SS ⫽ slow–slow (see Figure 6).
⫹
.05 ⱕ p ⬍ .10. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.

criterion for reliability being marginally reliable (.05 ⱕ p ⬍ .10). Discussion

These outcomes reinforce the conclusion that the speed manipu-
lation had its intended effect on both RTs and LRP start times, and Experiment 2 was designed to test the hypothesis that perceptual
allows the survivor function interaction contrasts (SFIC) results for learning leads to the creation of perceptual objects, in terms of the
both dependent variables to be interpretable. This is the first time, basic characteristics of information processing addressed by systems
to our knowledge, that this regularity has been assessed for a factorial theory. Specifically, this experiment tested the prediction that
neurophysiological rather than a behavioral variable, and furthers practice should produce a transition from serial, exhaustive, limited-
the extent to which predictions from SFT can be applied to the capacity processing to parallel, exhaustive, supercapacity processing,
timing of a neural event. indicative of positively dependent channels. As in Experiment 1,
Figure 8 plots the SFICs for all four observers, at BL and EL, at perceptual practice produced reliable decreases in thresholds, with
each level of contrast, for both the RTs and LRP start times. Table mixed results with respect to increases in false alarm rates. At BL, the
10 presents the test statistics for the positive (D⫹) and negative data suggested in most cases serial, exhaustive, limited capacity
(D⫺) deflections of the SFICs for both data sources. The overall processing, with the exceptions observed at the highest level of
pattern across observers was a transition from serial exhaustive to contrast. In comparison, at EL, at both suprathreshold levels of con-
parallel exhaustive processing, particularly at the two suprathresh- trast, processing was parallel, exhaustive, and supercapacity. All of
old levels of contrast, and this pattern was present for both the RTs this is consistent with the proposal that perceptual practice leads to the
and the LRP start times. creation of perceptual objects, as defined in SFT.
Figure 9 presents the capacity coefficients3 for each of the As in Experiment 1, the inferences drawn from the RTs and the
observers at each contrast level, at baseline and at endline, and
start times of the LRPs were consistent with one another. This, to
Table 11 presents the test statistics (from Houpt & Townsend, 2012)
our knowledge, is the first demonstration of the use of behavioral
assessing the extent to which those coefficients deviate from 1. The
and neural data as converging sources of evidence to test predic-
coefficients and tests are presented for both the RTs and the start times
tions drawn from SFT. Finally, the variations across observers in
of the LRPs. With a few exceptions, with these being at the highest
contrast, BL capacity was limited. At endline, capacity was generally terms of inferences demonstrate the ability of this approach to
limited at threshold and supercapacity at both suprathreshold levels of
contrast. Finally, in all but one case (observer 1, BL 60% contrast), the 3
Note that we did not use the more recent version of the capacity
inferences drawn from the RTs and the inferences drawn from the coefficient (Townsend & Altieri, 2012), one that can be conditionalized on
start times of the LRPs were in agreement. accuracy, based on the fact that performance was near ceiling in all cases.
 This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

470

Figure 8.
WENGER AND RHOTEN

Experiment 2: SFICs for the RTs (column a) and LRP start times (column b).
 PERCEPTUAL OBJECTS 471

Table 10
Experiment 2: Tests of the Positive and Negative Deflections of the SFICs, as Calculated From the RTs and the Start Times of
the LRPs

Baseline Endline
RT LRP RT LRP
Obs Contrast D⫹ D⫺ D⫹ D⫺ D⫹ D⫺ D⫹ D⫺
ⴱ ⴱ ⴱ
1 Threshold .36 ⫺.37 .53 ⫺.58ⴱ
30 .33ⴱ ⫺.32ⴱ .52ⴱ ⫺.51ⴱ .01 ⫺.47ⴱ .25 ⫺.58ⴱ
60 .00 .31ⴱ .16 ⫺.66ⴱ .03 ⫺.49ⴱ .03 ⫺.49ⴱ
2 Threshold .38ⴱ ⫺.39ⴱ .65ⴱ ⫺.50ⴱ
30 .29⫹ ⫺.31ⴱ .47ⴱ ⫺.50ⴱ .04 ⫺.29ⴱ .37 ⫺.67ⴱ
60 .05 ⫺.39ⴱ .27 ⫺.62ⴱ .15 ⫺.49ⴱ .32 ⫺.56ⴱ
.32⫹ ⫺.35ⴱ ⫺.13
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

3 Threshold .05
30 .33⫹ ⫺.38ⴱ .33ⴱ ⫺.35⫹ .01 ⫺.43ⴱ .12 ⫺.53ⴱ
This document is copyrighted by the American Psychological Association or one of its allied publishers.

60 .38ⴱ ⫺.32⫹ .33ⴱ ⫺.38ⴱ .06 ⫺.60ⴱ .13 ⫺.57ⴱ

4 Threshold .11 ⫺.14 .20 ⫺.23⫹
30 .28⫹ ⫺.33ⴱ .20 ⫺.27⫹ .05 ⫺.43ⴱ .12 ⫺.51ⴱ
60 .13 ⫺.36ⴱ .12 ⫺.28⫹ .06 ⫺.41ⴱ .04 ⫺.61ⴱ
⫹ ⴱ
.05 ⱕ p ⬍ .10. p ⬍ .05.

quantify individual differences at both behavioral and neural lev- Experiment 2, we showed that these same two features were
els. generally processed in a serial and exhaustive manner, with limited
capacity, prior to perceptual learning. Perceptual practice produced
General Discussion a shift to parallel, dependent, supercapacity processing, also con-
sistent with the learning of a perceptual object. Our use of GRT
In work that both set an agenda and coalesced a set of theoretical and SFT allowed us to illustrate that these changes were not
perspectives on skill (e.g., Fitts & Posner, 1967), Simon and Chase (1973) completely uniform, with there being important differences across
advanced the notion that the learning of perceptual chunks was a
individuals (Smith & Little, 2018). This ability to quantify indi-
necessary first step in the acquisition of expertise. They held that this
vidual differences has been a consistent strength of these two
learning of chunks allowed the capacity limitations of a short-duration
metatheories (e.g., Fific et al., 2008; Fifić, Townsend, & Eidels,
memory to be avoided and allowed for the transition from the slow
2008; Fific & Townsend, 2010).
use of declarative memory to the fast use of procedural memory.
In addition to testing our hypotheses with behavioral data—re-
Indeed, the notion of chunk has remained conceptually influential,
sponse frequencies and latencies—we were able to make use of
with a number of more recent theories having a role for perceptual and
neurophysiological data as a source of converging evidence. Specif-
conceptual chunks (e.g., Czerwinski et al., 1992; Goldstone, 1998,
ically, we analyzed the onset times of the lateralized readiness poten-
2000; Kellman & Garrigan, 2009; Kellman & Massey, 2013; Li et al.,
tials in terms of the temporal measures associated with both GRT and
2004; Li et al., 2004; Schyns et al., 1998).
A critical problem, however, is that a chunk has at best been SFT and found that the inferences drawn on the basis of the patterns
given operational definition. The two experiments reported here in this feature of the EEG were generally consistent with those drawn
took a very different approach. Specifically, we used two formal from on the basis of the patterns in the RT. The extent to which
metatheories— general recognition theory and systems factorial theories such as GRT and SFT can make principled predictions for
theory—to apply rigorous definitions of the characteristics of a measures of neural activity has been of interest at least since the work
chunk—a perceptual object—in terms of aspects of representation of Schweickert (Schweickert, 1989; Schweickert & Mounts, 1998).
and processing. In terms of representation, we defined a perceptual Schweickert’s results were mixed, whereas ours are reasonably con-
object as one that produced violations of independence and sepa- sistent. We would suggest that the difference lies in the level of
rability. In terms of processing, we defined a perceptual object as analysis: whereas Schweickert focused on the amplitude of a single
one processed exhaustively, in a parallel dependent manner, that ERP feature derived at individual electrodes, we relied on a specific
exhibits supercapacity. Note that these are not novel uses of the property of EEG activity that signals the programming of a motor
constructs of GRT and SFT, as they have been used in previous output on the basis of completion of cognitive processing. GRT and
theorizing to characterize facial holism (O’Toole et al., 2001) and SFT are intentionally general and describe regularities that theoreti-
the general notion of a gestalt (Townsend & Wenger, 2015). One cally apply to any information processing system, not just one that can
contribution of the present effort is to bring them together to bear be analyzed in terms of response frequencies and latencies. We
on a foundational question in perceptual learning and expertise. believe our results suggest that questions such as those raised by
In Experiment 1, we documented that two arbitrary contrast- Schweickert (1989; Schweickert & Mounts, 1998), along with the
defined features were represented as independent and separable regularities that can be derived as predictions from GRT and SFT, can
before perceptual learning. Perceptual practice produced a shift to be addressed with neural data at the appropriate level of analysis (see,
violations of independence and separability, both perceptually and e.g., Q. Zhang, Walsh, & Anderson, 2018, for another compelling
decisionally, consistent with the learning of a perceptual object. In example). Further, we believe that our results suggest that, at an
 This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

472

Figure 9.
WENGER AND RHOTEN

Experiment 2: C(t)s for the RTs (column a) and LRP start times (column b).
 PERCEPTUAL OBJECTS 473

Table 11
Experiment 2: Tests of the Deviations of the Capacity Coefficients From 1, as Calculated From
the RTs and the Start Times of the LRPs

Baseline contrast Endline contrast

Obs Data 30 60 Threshold 30 60
ⴱⴱⴱ ⴱ ⴱⴱⴱ ⴱⴱⴱ
1 RT ⫺5.95 ⫺1.29 ⫺2.96 5.39 2.53ⴱⴱⴱ
LRP start times ⫺1.77ⴱ .62 ⫺2.41ⴱⴱ 5.05ⴱⴱⴱ 5.56ⴱⴱⴱ
2 RT ⫺22.32ⴱⴱⴱ ⫺22.92ⴱⴱⴱ ⫺2.71ⴱ 8.09ⴱⴱⴱ 8.57ⴱⴱⴱ
LRP start times ⫺1.98ⴱ 3.41ⴱⴱ ⫺1.66ⴱ 6.81ⴱⴱⴱ 4.26ⴱⴱⴱ
3 RT ⫺1.65ⴱ ⫺1.64ⴱ ⫺1.86ⴱ 3.72ⴱⴱⴱ 3.96ⴱⴱⴱ
LRP start times ⫺12.24ⴱⴱⴱ ⫺10.51ⴱⴱⴱ ⫺13.94ⴱⴱⴱ 2.62ⴱⴱ 2.04ⴱ
4 RT ⫺1.80ⴱ ⫺1.80ⴱ ⫺2.87ⴱⴱ 11.89ⴱⴱⴱ 13.88ⴱⴱⴱ
LRP start times ⫺2.65ⴱⴱ ⫺1.98ⴱ ⫺1.73ⴱ 3.69ⴱⴱⴱ 3.96ⴱⴱⴱ
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ⴱ ⴱⴱ ⴱⴱⴱ
p ⬍ .05. p ⬍ .01. p ⬍ .001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

appropriate level of analysis, neural data can provide a theoretically Ashby, F. G., & Townsend, J. T. (1986). Varieties of perceptual indepen-
motivated source of converging evidence, strengthening the connec- dence. Psychological Review, 93, 154 –179.
tion between theory and data. Bridgman, P. W. (1945). Some general principles of operational analysis.
There are, to be sure, limitations to the present effort. Primary Psychological Review, 52, 246.
among these is the use of the 2 ⫻ 2 design used in Experiment 1, Carroll, L. (1865). Alice in wonderland. London, UK: Macmillan.
which has been shown to be subject to identifiability problems with Chase, W. G., & Ericsson, K. A. (1981). Skilled memory. In J. R. Anderson
respect to violations of perceptual independence and decisional sep- (Ed.), Cognitive skills and their acquisition (pp. 141–190). Hillsdale, NJ:
Erlbaum.
arability (Silbert & Thomas, 2013; Thomas & Silbert, 2014).
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive
Future work should use an expanded design and take advantage of
Psychology, 4, 55– 81.
newer statistical developments (Soto et al., 2015). A second lim-
Chechile, R. A. (2011). Properties of reverse hazard functions. Journal of
itation is that the tests of representation and processing, although
Mathematical Psychology, 55, 203–222.
using identical stimuli, were performed across rather than within Coles, M. G. (1989). Modern mind-brain reading: Psychophysiology, phys-
subjects. Future work should test the extent to which the regular- iology, and cognition. Psychophysiology, 26, 251–269.
ities observed across the two experiments reported here hold Cornes, K., Donnelly, N., Godwin, H., & Wenger, M. J. (2011). Perceptual
within individual observers. and decisional factors affecting the detection of the Thatcher illusion.
This latter point also raises an important theoretical opportunity. Journal of Experimental Psychology: Human Perception and Perfor-
Specifically, the regularities predicted using GRT and SFT should mance, 37, 645– 668.
be relatable within a single, coherent metatheory. Work with linear Czerwinski, M., Lightfoot, N., & Shiffrin, R. M. (1992). Automatization
systems theory (e.g., Wenger & Townsend, 2006; Townsend et al., and training in visual search. American Journal of Psychology, 105,
2012) suggests one avenue for exploration. Conceptually, there is 271–315.
a necessity for relating the notion of learning a perceptual object Delorme, A., & Mackeig, S. (2004). eeglab: An open source toolbox for
within a single theoretical perspective. analysis of single-trial eeg dynamics including independent component
In summary, in two experiments we demonstrated that percep- analysis. Journal of Neuroscience Methods, 134, 9 –21.
tual learning for two arbitrary contrast-defined features produces a Dosher, B. A., Jeter, P., Liu, J., & Lu, Z.-L. (2013). An integrated
shift from independence, separability, seriality, and limitations in reweighting theory of perceptual learning. Proceedings of the National
capacity to violations of dependence and separability, exhaustive Academy of Sciences of the United States of America, 110, 13678 –
parallelism, and supercapacity, and that evidence for these 13683.
Dosher, B. A., & Lu, Z.-L. (2017). Visual perceptual learning and models.
learning-dependent shifts can be observed in both behavioral and
Annual Review of Vision Science, 3, 343–363.
neurophysiological data. The evidence is strong and consistent:
Drever, J. (1960). Perceptual learning. Annual Review of Psychology, 11,
perceptual learning does indeed induce chunking—the production
131–160.
of perceptual objects.
Feltovich, P. J., Prietula, M. J., & Ericsson, K. A. (2006). Studies of
expertise from psychological perspectives. In K. A. Ericsson, N. Char-
References ness, P. J. Feltovich, & R. R. Hoffman (Eds.), The Cambridge handbook
of expertise and expert performance (pp. 41– 67). New York, NY:
Ahissar, M., & Hochstein, S. (1997). Task difficulty and the specificity of
perceptual learning. Nature, 387, 401– 406. Cambridge University Press.
Ahissar, M., & Hochstein, S. (2004). The reverse hierarchy theory of visual Fific, M., Nosofsky, R. M., & Townsend, J. T. (2008). Information-
perceptual learning. Trends in Cognitive Sciences, 8, 457– 464. processing architectures in multidimensional classification: A validation
Ahissar, M., Laiwand, L., Kozminsky, G., & Hochstein, S. (1998). Learn- test of the systems factorial technology. Journal of Experimental Psy-
ing pop-out detection: Building representions for conflicting target- chology: Human Perception and Performance, 34, 356 –375.
distractor relationships. Vision Research, 38, 3095–3107. Fific, M., & Townsend, J. T. (2010). Information-processing alternatives to
Ashby, F. G., Ennis, J. M., & Spiering, B. J. (2007). A neurobiological holistic perception: Identifying the mechanisms of secondary-level ho-
theory of automaticity in perceptual categorization. Psychological Re- lism within a categorization paradigm. Journal of Experimental Psychol-
view, 114, 632– 656. ogy: Learning, Memory, and Cognition, 36, 1290 –1313.
 474 WENGER AND RHOTEN

Fifić, M., Townsend, J. T., & Eidels, A. (2008). Studying visual search Miller, J., & Hackley, S. A. (1992). Electrophysiological evidence for
using systems factorial methodology with target-distractor similarity as temporal overlap among contingent mental processes. Journal of Exper-
the factor. Perception & Psychophysics, 70, 583– 603. imental Psychology: General, 121, 195.
Fitts, P. M., & Posner, M. I. (1967). Human performance. Oxford, Eng- Moldakarimov, S., Bazhenov, M., & Sejnowski, T. J. (2014). Top-down
land: Brooks/Cole. inputs enhance orientation selectivity in neurons of the primary visual
Gibson, E. J. (1963). Perceptual learning. Annual Review of Psychology, cortex during perceptual learning. PLoS Computational Biology, 10,
14, 29 –56. e1003770.
Gibson, E. J., & Walk, R. D. (1956). The effect of prolonged exposure to Mordkoff, J. T., & Grosjean, M. (2001). The lateralized readiness potential
visual patterns on learning to discriminate similar and different patterns. and response kinetics in response-time tasks. Psychophysiology, 38,
Journal of Comparitive and Psysiological Psychology, 51, 584 –587. 777–786.
Goldstone, R. L. (1998). Perceptual learning. Annual Review of Psychol- Mordkoff, J. T., Yantis, S., & Egeth, H. E. (1990). Detecting conjunctions
ogy, 49, 585– 612. of color and form in parallel. Perception & Psychophysics, 48, 157–168.
Goldstone, R. L. (2000). Unitization during category learning. Journal of O’Toole, A. J., Wenger, M. J., & Townsend, J. T. (2001). Quantitative
Experimental Psychology: Human Perception and Performance, 26, models of perceiving and remembering faces: Precedents and possibil-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

86 –112. ities. In M. J. Wenger & J. T. Townsend (Eds.), Computational, geo-

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Gourevitch, V., & Galanter, E. (1967). A significance test for one param- metric, and process perspectives on facial cognition: Contexts and
eter isosensitivity functions. Psychometrika, 32, 25–33. challenges (pp. 1–38). Mahwah NJ: Erlbaum.
Green, D. M., & Swets, J. A. (1966). Signal detection theory and psycho- Petrov, A., Dosher, B., & Lu, Z.-L. (2005). The dynamics of perceptual
physics. New York, NY: Wiley. learning: An incremental reweighting model. Psychological Review,
Hackley, S. A., & Miller, J. (1995). Response complexity and precue 112, 715–743.
interval effects on the lateralized readiness potential. Psychophysiology, Polat, U. (2016). Advanced perceptual learning techniques induce neuro-
32, 230 –241. plasticity to enable improved visual functions. Current Ophthalmology
Houpt, J. W., & Townsend, J. T. (2010). The statistical properties of the Reports, 4, 1–7.
survivor interaction contrast. Journal of Mathematical Psychology, 54, Pourtois, G., Grandjean, D., & Sander, P., & Vuilleumier, D. (2004).
446 – 453. Electrophysiological correlates of rapid spatial orienting towards fearful
Houpt, J. W., & Townsend, J. T. (2012). Statistical measures for workload faces. Cerebral Cortex, 14, 619 – 633.
capacity analysis. Journal of Mathematical Psychology, 56, 341–355. Rauss, K., Schwartz, S., & Pourtois, G. (2011). Top-down effects on early
Huang, C.-B., Lu, Z.-L., & Dosher, B. A. (2012). Co-learning analysis of visual processing in humans: A Predictive coding framework. Neuro-
two perceptual learning tasks with identical input stimuli supports the science & Biobehavioral Reviews, 35, 1237–1253.
reweighting hypothesis. Vision Research, 61, 25–32. Ray, W., Slobounov, S., Mordkoff, J., Johnston, J., & Simon, R. (2000).
Ingvalson, E. M., & Wenger, M. J. (2005). A strong test of the dual mode Rate of force development and the lateralized readiness potential. Psy-
hypothesis. Perception & Psychophysics, 67, 14 –35. chophysiology, 37, 757–765.
Kadlec, H., & Townsend, J. T. (1992). Signal detection analysis of dimen- Ruff, C. C., & Driver, J. (2006). Attentional preparation for a lateralized
sional interactions. In F. G. Ashby (Ed.), Multidimensional models of visual distractor: Behavioral and fMRI evidence. Journal of Cognitive
perception and cognition (pp. 181–228). Hillsdale, NJ: Erlbaum. Neuroscience, 18, 522–538.
Kellman, P. J., & Garrigan, P. (2009). Perceptual learning and human Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human
expertise. Physics of Life Reviews, 6, 53– 84. information processing: Detection, search, and attention. Psychological
Kellman, P. J., & Massey, C. M. (2013). Perceptual learning, cognition, Review, 84, 1– 66.
and expertise. Psychology of learning and motivation (Vol. 58, pp. Schoups, A. A., & Orban, G. A. (1996). Interocular transfer in perceptual
117–165). San Diego, CA: Elsevier. learning of a pop-out discrimination task. Proceedings of the National
Kuefner, D., Jacques, C., Prieto, E., & Rossion, B. (2010). Electrophysi- Academy of Sciences of the united States of America, 93, 7358 –7362.
ological correlates of the composite face illusion: Disentangling percep- Schweickert, R. (1989). Separable effects of factors on activation functions
tual and decisional components of holistic face processing in the human in discrete and continuous models: D’ and evoked potentials. Psycho-
brain. Brain and Cognition, 74, 225–238. logical Bulletin, 106, 318 –328.
Li, R. W., Levi, D. M., & Klein, S. A. (2004). Perceptual learning improves Schweickert, R., & Mounts, J. (1998). Additive effects of factors on
efficiency by re-tuning the decision ’template’ for position discrimina- reaction time and evoked potentials in continuous-flow models. In C. E.
tion. Nature Neuroscience, 7, 178 –183. Dowling & F. S. Roberts (Eds.), Recent progress in mathematical
Li, W., Piech, V., & Gilbert, C. D. (2004). Perceptual learning and psychology: Psychophysics, knowledge, representation, cognition, and
top-down influences in primary visual cortex. Nature Neuroscience, 7, measurement (pp. 311–327). Mahwah, NJ: Erlbaum.
651– 657. Schyns, P. G., Goldstone, R. L., & Thibaut, J.-P. (1998). The development
Little, D., Altieri, N., Yang, C., & Fific, M. (2017). Systems factorial of features in object concepts. Behavioral and Brain Sciences, 21, 1–17.
technology: A theory driven methodology for the identification of per- Seitz, A. R. (2017). Perceptual learning. Current Biology, 27, R631–R636.
ceptual and cognitive mechanisms. San Diego, CA: Elsevier Science & Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human
Technology Books. Retrieved from https://books.google.com/ information processing: Ii. perceptual learning, automatic attending and
books?id⫽OQICMQAACAAJ a general theory. Psychological review, 84, 127.
Logan, G. D. (1988). Toward an instance theory of automatization. Psy- Silbert, N. H., & Hawkins, R. X. D. (2016). A tutorial on general recog-
chological Review, 95, 492–527. nition theory. Journal of Mathematical Psychology, 73, 94 –109.
Lu, Z.-L., Lin, Z., & Dosher, B. A. (2016). Translating perceptual learning Silbert, N. H., & Thomas, R. D. (2013). Decisional separability, model
from the laboratory to applications. Trends in Cognitive Sciences, 20, identification, and statistical inference in the general recognition theory
561–563. framework. Psychonomic Bulletin & Review, 20, 1–20.
Macmillan, N. A., & Creelman, C. D. (2005). Detection theory: A user’s Simon, H. A., & Chase, W. G. (1973). Skill in chess. American Scientist,
guide (2nd ed.). Mahwah, NJ: Erlbaum. 61, 394 – 403.
 PERCEPTUAL OBJECTS 475

Smith, P. L., & Little, D. R. (2018). Small is beautiful: In defense of the cognition and action. In E. J. Wagenmaker (Ed.), The stevens handbook
small-N design. Psychonomic Bulletin & Review. Advance online pub- of experimental psychology and cognitive neuroscience. London, UK:
lication. Oxford University Press.
Soto, F. A., Vucovich, L., Musgrave, R., & Ashby, F. G. (2015). General Von Der Heide, R. J., Wenger, M. J., Bittner, J. L., & Fitousi, D. (2018).
recognition theory with individual differences: A new method for ex- Converging operations and the role of perceptual and decisional influ-
amining perceptual and decisional interactions with an application to ences on the perception of faces: Neural and behavioral evidence. Brain
face perception. Psychonomic Bulletin & Review, 22, 88 –111. and Cognition, 122, 59 –75.
Tenison, C., & Anderson, J. R. (2016). Modeling the distinct phases of skill Wang, R., Wang, J., Zhang, J.-Y., Xie, X.-Y., Yang, Y.-X., Luo, S.-H., . . .
acquisition. Journal of Experimental Psychology: Learning, Memory, Li, W. (2016). Perceptual learning at a conceptual level. Journal of
and Cognition, 42, 749. Neuroscience, 36, 2238 –2246.
Thomas, R. D. (2001). Characterizing perceptual interactions in face iden- Wang, R., Zhang, J.-Y., Klein, S. A., Levi, D. M., & Yu, C. (2012). Task
tification using multidimensional signal detection theory. In M. J. relevancy and demand modulate double-training enabled transfer of
Wenger & J. T. Townsend (Eds.), Computational, geometric, and pro- perceptual learning. Vision Research, 61, 33–38.
cess perspectives on facial cognition: Contexts and challenges (pp. Wang, R., Zhang, J.-Y., Klein, S. A., Levi, D. M., & Yu, C. (2014). Vernier
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

193–228). Mahwah, NJ: Erlbaum. perceptual learning transfers to completely untrained retinal locations
after double training: A “piggybacking” effect. Journal of Vision,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Thomas, R. D. (2003). Further considerations of a general d= in multidi-

mensional space. Journal of Mathematical Psychology, 47, 220 –224. 14(13), 12.
Watanabe, T., & Sasaki, Y. (2015). Perceptual learning: Toward a com-
Thomas, R. D., & Silbert, N. H. (2014). Technical clarification to Silbert
prehensive theory. Annual Review of Psychology, 66, 197–221.
and Thomas (2013): “Decisional separability, model identification, and
Wenger, M. J. (1999). On the whats and hows of retrieval in the acquisition
statistical inference in the general recognition theory framework.” Psy-
of a simple skill. Journal of Experimental Psychology: Learning, Mem-
chonomic Bulletin & Review, 21, 574 –575.
ory, and Cognition, 25, 1137–1160.
Townsend, J. T., & Altieri, N. (2012). An accuracy-response time capacity
Wenger, M. J., & Payne, D. G. (1995). On the acquisition of mnemonic
assessment function that measures performance against standard parallel
skill: Application of skilled memory theory. Journal of Experimental
predictions. Psychological Review, 119, 500 –516.
Psychology: Applied, 3, 194 –215.
Townsend, J. T., & Eidels, A. (2011). Workload capacity spaces: A unified
Wenger, M. J., & Townsend, J. T. (2001). Faces as Gestalt stimuli: Process
methodology for response time measures of efficiency as workload is
characteristics. In M. J. Wenger & J. T. Townsend (Eds.), Computa-
varied. Psychonomic Bulletin & Review, 18, 659 – 681. tional, geometric, and process perspectives on facial cognition (pp.
Townsend, J. T., Houpt, J. W., & Silbert, N. H. (2012). General recognition 229 –284). Mahwah, NJ: Erlbaum.
theory extended to include response times: Predictions for a class of Wenger, M. J., & Townsend, J. T. (2006). On the costs and benefits of
parallel systems. Journal of Mathematical Psychology, 56, 476 – 494. faces and words. Journal of Experimental Psychology: Human Percep-
Townsend, J. T., & Nozawa, G. (1995). On the spatio-temporal properties tion and Performance, 32, 755–779.
of elementary perception: An investigation of parallel, serial, and coact- Zhang, J.-Y., Cong, L.-J., Klein, S. A., Levi, D. M., & Yu, C. (2014).
ive theories. Journal of Mathematical Psychology, 39, 321–359. Perceptual learning improves adult amblyopic vision through rule-based
Townsend, J. T., & Wenger, M. J. (2004a). The serial-parallel dilemma: A cognitive compensation. Investigative Ophthalmology & Visual Science,
case study in a linkage of theory and method. Psychonomic Bulletin & 55, 2020 –2030.
Review, 11, 391– 418. Zhang, Q., Walsh, M. M., & Anderson, J. R. (2018). The Impact of
Townsend, J. T., & Wenger, M. J. (2004b). A theory of interactive parallel inserting an additional mental process. Computational Brain & Behav-
processing: New capacity measures and predictions for a response time ior, 1–14.
inequality series. Psychological Review, 111, 1003–1035. Zhaoping, L., Herzog, M. H., & Dayan, P. (2003). Nonlinear ideal obser-
Townsend, J. T., & Wenger, M. J. (2015). On the dynamic perceptual vation and recurrent preprocessing in perceptual learning. Network:
characteristics of gestalten: Theory-based methods. In J. Wagemans Computation in Neural Systems, 14, 233–247.
(Ed.), The Oxford handbook of perceptual organization (pp. 948 –968).
Oxford, UK: Oxford University Press. Received October 9, 2018
Townsend, J. T., Wenger, M. J., & Houpt, J. W. (2018). Uncovering mental Revision received May 6, 2019
architecture and related mechanisms in elementary human perception, Accepted May 6, 2019 䡲